Modelos estatísticos para previsão de metástases axilares ... · mama em pacientes submetidos à biópsia de linfonodo sentinela [tese]. São Paulo: Faculdade de Medicina, Universidade

JOSÉ LUIZ BARBOSA BEVILACQUA

Modelos estatísticos para previsão de metástases

axilares em câncer de mama em pacientes

submetidos à biópsia de linfonodo sentinela

Tese apresentada à Faculdade de Medicina da

Universidade de São Paulo para obtenção do

título de Doutor em Ciências

Área de concentração: Clínica Cirúrgica

Orientador: Prof. Dr. Ruy Geraldo Bevilacqua

Co-orientador: Prof. Dr. Aldo Junqueira

Rodrigues Junior

São Paulo

2005

FICHA CATALOGRÁFICA

Preparada pela Biblioteca da Faculdade de Medicina da Universidade de São Paulo

©reprodução autorizada pelo autor

DEDICATÓRIA

Ao Prof. Dr. Ruy Geraldo Bevilacqua,

Minha bússola, meu orientador, meu professor e, principalmente, pai;

À Roberta,

Minha esposa, minha maior descoberta e companheira de todos os momentos,

com quem venho aprendendo as outras dimensões do significado de família;

À Elisa e Camila,

Minhas filhas, com quem aprendi o verdadeiro significado da vida;

À Maria José,

Minha mãe e tudo o que isso representa.

AGRADECIMENTOS

À Dra. Kimberly J. Van Zee, pela sua inestimável colaboração, por todos os seus

ensinamentos e amizade.

À Jane V. Fey, pela essencial ajuda, e pelo maravilhoso trabalho de coleta e

manutenção do banco de dados.

Ao Dr. Patrick I. Borgen e ao Dr. Murray F. Brennan, aos quais devo minha

eterna gratidão, por tornarem possível meu fellowship no Memorial Sloan-

Kettering Cancer Center (MSKCC).

Aos outros médicos do MSKCC, Hiram S. Cody III, Alexandra S. Heerdt,

Jeanne A. Petrek (in memoriam), Leslie L. Montgomery, Elisa Port e Virgilio

Sacchini, pelo carinho e atenção.

Ao Dr. Michael W. Kattan, estatístico do MSKCC, pela ajuda no

desenvolvimento dos nomogramas.

Ao Prof. Dr. Aldo Junqueira Rodrigues Jr. e à Profa. Dra. Consuelo Junqueira

Rodrigues, pelo apoio e estímulo científico.

À Alex MacDonald e à Laís de Almeida Cardoso, pelos excelentes trabalhos de

formatação e revisão de textos.

À equipe da VinhaNet pela ajuda no desenvolvimento do software, e ao Marco

Skita, pela programação visual do CD.

Esta tese está de acordo com:

Referências: adaptado de International Committee of Medical Journals Editors

(Vancouver)

Universidade de São Paulo. Faculdade de Medicina. Serviço de Biblioteca e

Documentação. Guia de apresentação de dissertações, teses e monografias.

Elaborado por Anneliese Carneiro da Cunha, Maria Julia de A. L. Freddi, Maria F.

Crestana, Marinalva de Souza Aragão, Suely Campos Cardoso, Valéria Vilhena. São

Paulo: Serviço de Biblioteca e Documentação; 2004

Abreviaturas dos títulos dos periódicos de acordo com List of Journals Indexed in Index Medicus

SUMÁRIO

InícioLista de abreviaturas Lista de símbolos Lista de siglas Lista de tabelas Lista de figuras Resumo Summary 1.INTRODUÇÃO.................................................................................................... 01 2. OBJETIVOS........................................................................................................ 04 3. REVISÃO DA LITERATURA........................................................................... 05 4. CASUÍSTICA E MÉTODOS.............................................................................. 16 4.1. Técnica de Biópsia e de Avaliação de LS Sentinela do MSKCC.................... 17 4.1.1. Mapeamento Linfático e Biópsia de LS........................................................ 17 4.1.2. Avaliação Anatomopatológica....................................................................... 18 4.2. Descrição das variáveis estudadas.................................................................... 19 4.3. Modelos Estatísticos......................................................................................... 21 4.3.1. Modelo de Metástase em Linfonodo Sentinela............................................. 22 4.3.1.1. Casuística do modelo de metástase em LS................................................. 22 4.3.1.2. Estatística do modelo de metástase em LS................................................. 23 4.3.2. Modelo de Metástase em Linfonodo Não-Sentinela..................................... 25 4.3.2.1. Casuística do modelo de metástase em Não-LS......................................... 25 4.3.2.2. Estatística do modelo de metástase em Não-LS......................................... 27 5. RESULTADOS................................................................................................... 29 5.1. Modelo de Metástase em Linfonodo Sentinela................................................ 29 5.2. Modelo de Metástase em Linfonodo Não-Sentinela........................................ 61 5.2.1. Utilizando-se os nomogramas.......................... ............................................ 69 6. DISCUSSÃO....................................................................................................... 75 6.1. Metástase em linfonodo sentinela.................................................................... 75 6.2. Metástase em linfonodo não-sentinela............................................................. 86 7. CONCLUSÕES................................................................................................... 96 8. REFERÊNCIAS.................................................................................................. 97 Apêndices

LISTA DE ABREVIATURAS 99mTc isótopo do tecnécio de número atômico 99 metaestável

et al. e outros

H&E hematoxilina e eosina

IC intervalo de confiança

IHQ imunoistoquímica(o)

IVL invasão vásculo-linfática

LS linfonodo(s) sentinela(s)

Max valor máximo

Min valor mínimo

MRM mastectomia radical modificada

MT mastectomia total

Não-LS linfonodo(s) axilar(es) não-sentinela(s) ou linfonodo(s) axilar(es)

adicional(is)

ND não disponível

NR não relatada (estatística)

NS não significante

QIL quadrante ínfero-lateral

QIM quadrante ínfero-medial

QSL quadrante súpero-lateral

QSM quadrante súpero-medial

RE receptor(es) de estrógeno

RP receptor(es) de progesterona

vs versus

LISTA DE SÍMBOLOS

< menor que

= igual a

> maior que

≥ maior ou igual a

≤ menor ou igual a

+ positivo

µCi microcurie(s)

µm micrômetro(s)

χ2 chi-quadrado

cm centímetro(s)

ml mililitro(s)

mm milímetro(s)

LISTA DE SIGLAS

ACOSOG American College of Surgeons Oncology Group

CD compact disc

IBCSG International Breast Cancer Study Group

IRB Institutional Review Board

MSKCC Memorial Sloan-Kettering Cancer Center

NSABP National Surgical Adjuvant Breast and Bowel Project

ROC receiver operating characteristic curve ou curva de características

operacionais

SEER Surveillance, Epidemiology, and End Results

LISTA DE TABELAS

Tabela 1 – Estudos com análise multivariada de fatores preditivos para metástase axilar linfonodal..................................................................................... 06

Tabela 2 – Estudos sobre fatores preditivos de metástase em Não-LS em axila com LS positivo............................................................................................. 09

Tabela 3 – Estudos que relatam a incidência de metástase em Não-LS em axila com LS positivo, por tamanho do tumor primário........................................ 11

Tabela 4 – Estudos que relatam a incidência de metástase em Não-LS em axila com LS positivo, por presença de IVL no tumor primário............................ 12

Tabela 5 – Estudos que relatam a incidência de metástase em Não-LS em axila com LS positivo, por tamanho da metástase no LS....................................... 13

Tabela 6 – Estudos que relatam a incidência de metástase em Não-LS em pacientes com metástase em LS detectada exclusivamente por IHQ.................... 14

Tabela 7 – Características dos pacientes e tumores – Amostras de Modelagem e de Validação – Modelo de Metástase em LS............................................. 30

Tabela 8A – Incidência de metástases em LS – Amostras de Modelagem e de Validação – Modelo de Metástase em LS.......................................... 32

Tabela 8B – Incidência de metástases em LS – Amostras de Modelagem e de Validação – Modelo de Metástase em LS........................................... 33

Tabela 9A – Análise univariada – Variáveis contínuas – Amostra de Modelagem – Modelo de Metástase em LS............................................................... 38

Tabela 9B – Análise univariada – Variáveis categóricas – Amostra de Modelagem – Modelo de Metástase em LS............................................................... 39

Tabela 10 – Análise univariada – Comparação dos tipos histológicos – Amostra de Modelagem – Modelo de Metástase em LS......................................... 40

Tabela 11 – Características dos pacientes e tumores – Ductal versus Lobular – Amostra de Modelagem – Modelo de Metástase em LS...................... 41

Tabela 12 – Localização do Tumor – Análise univarida: Ductal versus Lobular – Amostra de Modelagem – Modelo de Metástase em LS..................... 42

Tabela 13 – Análise multivariada: Regressão logística – Comparação Ductal versus Lobular (N=4066) – Amostra de Modelagem – Modelo de Metástase em LS........................................................................................................ 44

Tabela 14 – Comparações múltiplas entre quadrantes – Amostra de Modelagem – Modelo de Metástase em LS – Análise univariada.............................. 48

Tabela 15 – Localização do Tumor – Análise univarida: Medial versus Tumores de outros quadrantes – Amostra de Modelagem – Modelo de Metástase em LS......................................................................................................... 49

Tabela 16 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo de Metástase em LS incluindo Grau Nuclear + Lobular (N=3723).............................................................................................. 51

Tabela 17 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo Categórico (N=4066) – Modelo de Metástase em LS......................................................................................................... 53

Tabela 18 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo Contínuo (N=4066) – Modelo de Metástase em LS............... 54

Tabela 19 – Cálculo de probabilidade de metástase em LS – Valores a serem aplicados na equação do Modelo Contínuo......................................... 59

Tabela 20 – Características dos pacientes e tumores – Amostras de Modelagem e de Validação – Modelo de Metástase em Não-LS.................................... 62

Tabela 21 – Incidência de metástases em Não-LS – Amostras de Modelagem e de Validação – Modelo de Metástase em Não-LS.................................... 63

Tabela 22 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo Com Congelação (N=669*) – Modelo de Metástase em Não-LS......................................................................................................... 65

Tabela 23 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo Sem Congelação (N=702*) – Modelo de Metástase em Não-LS......................................................................................................... 66

Tabela 24 – Recidiva axilar após LS Positivo em biópsia de LS sem esvaziamento complemetar – Revisão da literatura................................................... 93

Tabela 25 – Recidiva axilar após linfadenectomia sistemática – Revisão da literatura............................................................................................... 94

LISTA DE FIGURAS

Figura 1 – Freqüência de metástase em LS por estádio T e método de detecção – Amostra de Modelagem – Modelo de Metástase em LS....................... 31

Figura 2 – Freqüência de metástase em LS em carcinoma ductal (incluindo subtipos especiais) por estádio T e método de detecção – Amostra de Modelagem – Modelo de Metástase em LS.................................................................. 35

Figura 3 – Freqüência de metástase em LS em carcinoma lobular por estádio T e método de detecção – Amostra de Modelagem – Modelo de Metástase em LS........................................................................................................... 36

Figura 4 – Freqüência de metástase em LS por localização do tumor (quadrante) – Amostra de Modelagem – Modelo de Metástase em LS........................ 46

Figura 5 – Freqüência de metástase em LS por localização do tumor (quadrante) – Amostra de Modelagem – Modelo de Metástase em LS........................ 47

Figura 6 – Curva ROC – Modelo Categórico – Amostra de Modelagem – Modelo de Metástase em LS.................................................................................... 55

Figura 7 – Curva ROC – Modelo Contínuo – Modelo de Metástase em LS........... 56

Figura 8 – Probabilidade observada versus probabilidade calculada – Modelo de Metástase em LS.................................................................................... 60

Figura 9 – Nomograma com congelação, N=669 – Modelo de Metástase em Não-LS........................................................................................................... 67

Figura 10 – Nomograma sem congelação, N=702 – Modelo de Metástase em Não-LS......................................................................................................... 68

Figura 11 – Curva ROC – Nomograma com congelação, N=669 – Modelo de Metástase em Não-LS.......................................................................... 71

Figura 12 – Curva ROC – Nomograma (sem congelação), N=702 – Modelo de Metástase em Não-LS – Nomograma sem congelação........................ 72

Figura 13 – Probabilidade observada versus probabilidade calculada – Modelo de Metástase em Não-LS – Nomograma com congelação....................... 73

Figura 14 – Probabilidade observada versus probabilidade calculada – Modelo de Metástase em Não-LS – Nomograma sem congelação....................... 74

RESUMO Bevilacqua JLB. Modelos estatísticos para previsão de metástases axilares em câncer de mama em pacientes submetidos à biópsia de linfonodo sentinela [tese]. São Paulo: Faculdade de Medicina, Universidade de São Paulo; 2005. 112p. INTRODUÇÃO: O status axilar é o fator prognóstico mais importante em câncer de mama. A biópsia de linfonodo sentinela (BLS) tornou-se o procedimento padrão no estadiamento axilar em pacientes com axila clinicamente negativa. O procedimento recomendado para pacientes com metástase em linfonodo sentinela (LS) inclui a linfadenectomia axilar completa (LAC). Porém questiona-se a necessidade da LAC em todos os pacientes com metástase em LS (MLS), particularmente naqueles com baixo risco de metástase em linfonodos axilares adicionais (Não-LS). Estimar de forma precisa a probabilidade de MLS e metástase em Não-LS (MNão-LS) pode auxiliar em muito o processo de decisão terapêutica. Os dados publicados referentes aos fatores preditivos de MLS e MNão-LS são de certa forma escassos. A forma de como esses dados são apresentados – geralmente expressos como razão de chance (odds-ratio) – dificulta o cálculo de probabilidade de MLS ou MNão-LS para um paciente em específico. Nesta tese, dois modelos de previsão computadorizados de fácil utilização foram desenvolvidos, utilizando-se um grande número de casos, para facilitar o cálculo de probabilidade de MLS e MNão-LS. MÉTODOS: Todos os dados clínico-patológicos foram coletados do banco de dados prospectivo de LS do Memorial Sloan-Kettering Cancer Center (MSKCC), Nova York, EUA. Os projetos de desenvolvimento de ambos os modelos foram aprovados pelo Institutional Review Board do MSKCC. Dois modelos foram desenvolvidos: Modelo de MLS (Modelo LS) e Modelo de metástase em Não-LS (Modelo Não-LS). No Modelo LS, os achados clínico-patológicos de 4.115 procedimentos subseqüentes de BLS (amostra de modelagem) foram submetidos à análise de regressão logística multivariada para se criar um modelo de previsão de MLS. Um software baseado nesse modelo foi desenvolvido utilizando-se as variáveis: idade, tamanho do tumor, tipo histológico, invasão vásculo-linfática (IVL), localização do tumor e multifocalidade. Esse modelo foi validado em uma amostra distinta (amostra de validação) com 1.792 BLSs subseqüentes. No Modelo Não-LS, os achados patológicos do tumor primário e das MLS, obtidas de 702 procedimentos de BLS (amostra de modelagem) em pacientes submetidos à LAC, foram submetidos à análise de regressão logística multivariada para se criar um modelo de previsão de MNão-LS. Um nomograma e um software baseados nesse modelo foram desenvolvidos utilizando-se as variáveis: tamanho do tumor, tipo histológico, grau nuclear, IVL, multifocalidade, receptor de estrógeno, método de detecção da MLS, número de LS positivos, número de LS negativos. Esse modelo foi validado em uma amostra distinta (amostra de validação) com 373 casos subseqüentes. RESULTADOS: O software do Modelo LS na amostra de modelagem mostrou-se adequado, com a área sob a curva de características operacionais (ROC) de 0,76. Quando aplicado na amostra de validação, o Modelo LS também previu de forma acurada as probabilidades de MLS (ROC = 0,76). O software e o nomograma do Modelo Não-LS na amostra de modelagem apresentaram uma área sob a curva ROC de 0,76 e, na amostra de validação, 0,77. CONCLUSÕES: Dois softwares de fácil utilização foram desenvolvidos, utilizando-se informações comumente disponíveis pelo cirurgião para calcular para o paciente a probabilidade de MLS e MNão-LS de forma precisa, fácil e individualizada. O software do Modelo LS não deve ser utilizado, porém, para se evitar a BLS. Descritores: 1.NEOPLASIAS MAMÁRIAS/epidemiologia 2.MODELOS ESTATÍSTICOS 3.PROGNÓSTICO 4.BIÓPSIA DE LINFONODO SENTINELA 5.LINFONODOS 6.METÁSTASE NEOPLÁSICA.

SUMMARY Bevilacqua JLB. Statistical models for predicting axillary metastases in breast cancer patients submitted to sentinel lymph node biopsy [thesis]. São Paulo: Faculdade de Medicina, Universidade de São Paulo; 2005. 112p. INTRODUCTION: Axillary lymph node status is the most significant prognostic factor in breast cancer. Sentinel lymph node biopsy (SLNB) has become the standard of care as the axillary staging procedure in clinically node negative patients. The standard procedure for patients with sentinel lymph node (SLN) metastasis includes complete axillary lymph node dissection (ALND). However, many experts question the need for complete ALND in every patient with detectable SLN metastases, particularly those perceived to have a low risk of additional lymph node (Non-SLN) metastasis. Accurate estimates of the likelihood of SLN metastases and additional disease in the axilla could greatly assist in decision-making treatment. The published data on predictive factors for SLN and Non-SLN metastases is somewhat scarce. It is also difficult to apply these data – usually expressed as odds ratio – to calculate the probability of SLN or Non-SLN metastases for a specific patient. In this thesis, two user-friendly computerized prediction models based on large datasets were developed, to assist the prediction of the presence of SLN and Non-SLN metastases. METHODS: All clinical and pathological data were collected from the prospective SLN database of Memorial Sloan-Kettering Cancer Center (MSKCC), New York, USA. The development projects of both models were approved by the Institutional Review Board of MSKCC. Two models were developed: Model for predicting SLN metastases (SLN Model) and Model for predicting Non-SLN metastases (Non-SLN Model). In the SLN Model, clinical and pathological features of 4,115 sequential SLNB procedures (modeling sample) were assessed with multivariable logistic regression to predict the presence SLN metastases. A software based on the logistic regression model was created using age, tumor size, tumor type, lymphovascular invasion, tumor location and multifocality. This model was subsequently applied to another set of 1,792 sequential SLNBs (validation sample). In the Non-SLN Model, pathological features of the primary tumor and SLN metastases, identified in 702 SLNBs (modeling sample) on patients who underwent complete ALND, were assessed with multivariable logistic regression to predict the presence of additional disease in the Non-SLNs of these patients. A nomogram and a software were created using tumor size, tumor type and nuclear grade, lymphovascular invasion, multifocality, and estrogen-receptor status of the primary tumor; method of detection of SLN metastases; number of positive SLNs; and number of negative SLNs. This model was subsequently applied to another set of sequential 373 procedures (validation sample). RESULTS: The software of the SLN Model for the modeling sample was accurate and discriminating, with an area under the receiver operating characteristic (ROC) curve of 0.76. When applied to the validation sample, the SLN-Model accurately predicted likelihood of SLN metastases (ROC = 0.76). The software and nomogram of the Non-SLN Model was also accurately predicted likelihood of Non-SLN metastases, with an area under ROC curve of 0.76 for the modeling sample, and 0.77 for the validation sample. CONCLUSION: Two user-friendly softwares were developed, using information commonly available to the surgeon to easily and accurately calculate the likelihood of having SLN metastases or additional, Non-SLN metastases for individual patients. However, the software of the SLN Model should not be used to avoid SLNB. Keywords: 1.BREAST NEOPLASMS /epidemiology 2.MODELS, STATISTICAL 3.PROGNOSIS 4.SENTINEL LYMPH NODE BIOPSY 5.LIMPH NODES 6.NEOPLASM METASTASIS.

1

1. INTRODUÇÃO

A metástase linfonodal no câncer de mama é um evento relativamente

comum, constituindo o principal fator prognóstico nesse tipo de neoplasia. A

informação da presença ou ausência de mestástase linfonodal é de fundamental

importância clínica para determinar o tratamento.1-4

Diversos fatores clínico-patológicos associados à presença de metástase

axilar em câncer de mama são descritos (Tabela 1, página 6), mostrando a natureza

multifatorial desse evento. Entretanto, as formas de como esses estudos expressam os

resultados dificulta muito a aplicação prática e clínica dos resultados. Até o presente,

poucos estudos especificamente estudaram a presença de metástase em linfonodo

sentinela (LS).

Mesmo havendo vários trabalhos na literatura médica sobre fatores preditivos

de metástase axilar, é muito difícil responder de forma precisa e numérica a

perguntas do tipo “Doutor, qual é minha chance (ou probabilidade) de metástase

axilar?” ou “Minha paciente tem um linfonodo sentinela positivo identificado

somente na imunoistoquímica, será que é necessário realizar o esvaziamento axilar?

Qual é o risco de essa paciente ter outros linfonodos comprometidos?” A

dificuldade em responder a tais perguntas é devida à natureza multifatorial da

metástase linfonodal. Embora se conheçam alguns dos fatores associados à presença

de metástase, a questão é: como aplicá-los de forma prática para responder a

perguntas desse tipo?

2

As maneiras como geralmente são expressos os resultados dos modelos

estatísticos – a razão de chance, ou o risco relativo em relação a uma subvariável –

são de difícil aplicação clínica. Os resultados, porém, deveriam ser facilmente

aplicáveis, apresentados de forma prática e mensurável – preferencialmente em

forma de números percentuais – para atingir os objetivos primordiais da pesquisa

translacional, o enfoque da presente tese.

Os processos de crescimento tumoral e de metástase ainda não são

completamente conhecidos. Apesar do enorme progresso tecnológico e do

conhecimento genético e molecular biológico, ainda não foram descobertos todos os

fatores-chave que determinam a gênese e a progressão tumoral. Os eventos naturais

são multifatoriais. O uso da matemática para se entenderem os fenômenos naturais é

uma das mais antigas aplicações desta ciência e o câncer de mama não é exceção.

Isso facilita a aplicação dos resultados na prática clínica.

Um dos primeiros exemplos do uso da matemática em eventos biológicos se

refere à curva de crescimento tumoral.5-9 Outros estudos tentam avaliar a resposta

desse crescimento a tratamentos específicos.10-12

Outros modelos matemáticos, utilizando-se de critérios prognósticos clínico-

patológicos simples, são descritos para prever o comportamento e evolução de

pacientes com câncer.13-15

Em epidemiologia, existem modelos estatísticos de grande aplicação clínica

para o cálculo de risco para câncer de mama e sua prevenção.16-21 O mais conhecido

é o modelo de Gail et al.,19 sendo um dos únicos com validação em outros estudos.22-

26

3

A validação de um modelo estatístico, em uma amostra distinta da que foi

desenvolvido, é de extrema importância, mostrando o poder do modelo. Infelizmente

a maioria dos modelos publicados não tem validação prospectiva em amostras

distintas.

A íntima relação entre a matemática (estatística) e os novos avanços na

biologia molecular tumoral, com banco de tecidos e dados clínicos interligados, torna

o potencial da pesquisa translacional imensurável.

4

2. OBJETIVOS

1) O objetivo principal é desenvolver modelos estatísticos que sejam de fácil

aplicação no cotidiano médico, para determinar a probabilidade de

metástases linfonodais em câncer de mama.

2) O primeiro modelo tem o objetivo de calcular a probabilidade de

metástase em LS.

3) O segundo modelo tem o objetivo de calcular a probabilidade de

metástase em linfonodos não-sentinela (Não-LS) quando o LS for

positivo para metástase, ou seja, calcular a probabilidade de se

identificarem metástases em outros linfonodos axilares que não o

sentinela.

4) Como parte do objetivo principal está a validação dos modelos em

amostras distintas das quais foram geradas.

5) Os objetivos secundários da presente tese são: identificar a freqüência e

os fatores associados à metástase axilar e explicar possíveis diferenças

para a maior incidência de metástases axilares na era do LS.

5

3. REVISÃO DA LITERATURA

A informação sobre a presença de metástase axilar é o principal fator

prognóstico em pacientes com carcinoma invasor de mama27-32 e a linfadenectomia

axilar tem sido utilizada na abordagem dessa doença há vários séculos.33-38

Até a década de 1980, a importância da linfadenectomia axilar sistemática,

como parte integrante do tratamento cirúrgico, era inquestionável. O estudo NSABP-

B04 é o primeiro a questionar o valor terapêutico da linfadenectomia axilar, sendo

colocado exclusivamente como fator prognóstico da progressão pós-cirúrgica da

doença.39,40

Com a disseminação do uso da mamografia como método de diagnóstico

precoce do câncer de mama nos países desenvolvidos, na década de 1980, houve um

aumento significativo do número de casos diagnosticados em fase inicial, ou seja, de

tumores pequenos e não palpáveis. Como conseqüência, os diagnósticos passaram a

ser feitos numa situação em que a freqüência de metástases axilares era muito menor.

Nessa época, diversos autores passaram a questionar a necessidade da realização da

linfadenectomia axilar nesses tumores pequenos, uma vez que esse procedimento não

é isento de morbidades.41 Na tentativa de selecionar os pacientes para

linfadenectomia, vários estudos, todos retrospectivos, tentaram identificar os fatores

preditivos de metástase axilar. (Tabela 1).

6

Tabela 1 – Estudos com análise multivariada de fatores preditivos para

metástase axilar linfonodal

Variável Número de estudos

Referências

Tamanho 20 42-61 Idade 10 45,47,49,52-55,60,62,63 Invasão vasculo-linfática 7 43,44,46-48,53,56 Tipo histológico 7 46,49,52,53,58-60 Localização 4 48,49,52,60 Palpabilidade 3 43,48,53 Grau Nuclear 3 28,43,56 Grau Histológico 3 52,55,64 Grau Tumoral 2 49,61 Multifocalidade 2 46,47 Receptor de estrógeno 2 49,54 Status linfonodal clínico 2 47,53 Status menopausal 1 47 Receptor de progesterona 1 49 HER-2/neu 1 51 Status das margens 1 53 Bordas espiculadas 1 46 S-phase 1 54 Ploidia 1 49 Raça 1 49 Microcalcificação 1 61 Número de linfonodos examinados 1 60 Invasão de mamilo 1 46 Paridade 1 47

7

Com o advento do conceito de LS,65,66 assim como sua aplicação no

estadiamento da axila em câncer de mama,67-69 foi possível selecionar os pacientes

para a realização do esvaziamento axilar. Propôs-se sua indicação apenas quando

havia metástase em LS. A biópsia do LS ganhou importância na medida em que

mostrou-se ser procedimento tecnicamente simples, muito eficaz no estadiamento da

axila e associado à mínima morbidade.70

A biópsia do LS tem sido adotada como um procedimento padrão no

estadiamento axilar. Substitui a dissecção linfonodal axilar sistemática quando os LS

assim removidos não apresentam metástases.51,68,70-74 A biópsia do LS tem alta

acurácia no estadiamento axilar,51,68,70-74 sendo esta no mínimo igual à dissecção

linfonodal axilar, principalmente quando realizada a análise anatomopatológica

refinada dos LS. Isso inclui múltiplos cortes seriados e estudo imunoistoquímico

(IHQ) com anticorpos anticitoqueratina.75 Esse procedimento tornou-se viável na era

do LS devido a um pequeno número de linfonodos necessários para a análise.

A freqüência de metástases axilares linfonodais por estádio T (TNM3) varia

de acordo com os estudos da era pré-LS. Nos estudos maiores, a freqüência é de 10%

em T1mic,55 9% a 13% em T1a,47,52,53,63,76 13% a 19% em T1b,47,52,53,60,63,76 26% a

29% em T1c,47,53,60,76 39% a 50% em T2 menor ou igual a 3 cm,47,53,60,76 48% a 59%

em T2 maior que 3 cm,53,60 e 71% a 80% em T3.53,76

Vários estudos sobre a identificação dos fatores preditivos de metástases

axilares são identificados na literatura.42-64,76-84

A maioria desses estudos coleta os dados retrospectivamente, não menciona

como foram coletados ou não tem revisão anatomopatológica centralizada. Nesses

estudos42-64,76-84 as variáveis mais freqüentes e independentemente associadas a

8

metástase axilar são tamanho do tumor,42-61,77-84 idade,45,47,49,52-55,60,62,63 invasão

vascular linfática (IVL)43,44,46-48,53,56 e subtipo histológico do tumor.46,49,52,53,58-60

Somente Gann et al.49 validam retrospectivamente seu modelo estatístico. A Tabela 1

resume todos os fatores independentes associados à metástase axilar em câncer de

mama descritos na literatura.

O número de linfonodos axilares comprometidos é um importante fator

prognóstico, sendo inclusive adotado na atual versão do TNM.3 Portanto, somente a

informação do LS como positivo ou negativo é algo limitado. A informação do

número de linfonodos comprometidos pode, em certas circunstâncias, ajudar na

decisão terapêutica do tratamento de rádio ou quimioterapia adjuvante.

Embora o tratamento padrão atual seja completar o esvaziamento axilar

(linfadenectomia axilar complementar) frente a um LS positivo, discute-se essa

obrigatoriedade em todos os pacientes, principalmente naqueles em que o risco de

metástase adicional em Não-LS é baixo.85,86

Os estudos relatam que apenas aproximadamente 50% dos pacientes com LS

positivo apresentam metástase em Não-LS.51,67-69,73,87-92

Vários fatores clínico-patológicos estão associados à presença de doença

adicional em Não-LS (Tabela 2). Podemos observar que a casuística dos estudos

anteriores é limitada, que nem todos realizam análise multivariada das observações e

que em nenhum há validação dos resultados em amostras distintas.

O tamanho do tumor e o tamanho da metástase em LS são os fatores mais

comumente analisados. Entretanto, nenhum desses estudos incluídos na Tabela 2 é

capaz de identificar com segurança um grupo isento de risco para metástase em Não-

LS.

9

Tabela 2 – Estudos sobre fatores preditivos de metástase em Não-LS em axila com LS positivo

(Os números da tabela são valores de P relatados nos trabalhos originais)

LS, linfonodo sentinela; Não-LS, linfonodo axilar não-sentinela; IVL, invasão vásculo-linfática; RE, receptor de estrógeno; RP, receptor de progesterona N, número de pacientes com LS positivo submetidos à dissecção axilar —, não analisado; NS, não significante; NR, estatística não relatada

Estudo Pacientes Características do tumor primário Características do LS

Autor Ano Análise

Multiva-riada

N Idade Tama-

nho (cm)

Grau Nuclear

Grau Histo-lógico

Tipo Histo-lógico

IVL Multifo-calidade RE RP

Método de

Deteção

Tamanho da Metásta-

se no LS

Extensão Extranodal

Nº de LS

positivos

Nº total de LS

negativos Chu85 1999 Sim 157 NS 0,01 NS — NS NS ― NS NS NS <0,0001 ― NS ― Reynolds90 1999 Sim 60 NS 0,0004 NS — NS NS ― 0,03 NS ― 0,002 ― ― ― Teng91 2000 Sim 26 — 0,001 ― — — ― ― ― — 0,02 ― ― ― ― Turner93 2000 Sim 194 NS 0,03 NS NS — 0,03 ― NS NS ― 0,01 0,0001 NS ―

Abdessalam87 2001 Não 100 NS NS NS — NS 0,004 ― NS NS ― 0,01 <0,001 NS ―

Canavese94 2001 Não 72 — NR — — — — — — — — NR ― ― ―

Cserni95 2001 Sim 69 — <0,01 — NS — — — — — — <0,005 <0,05 ― ―

Kamath86 2001 Não 101 — 0,005 ― — — ― ― ― — 0,02 <0,001 ― ― ― Liang96 2001 Não 78 — — — — — — — — — — NR ― ― ― Rahusen89 2001 Não 93 NS NS — NS — NS ― NS — ― 0,05 ― 0,002 ― Viale97 2001 Sim 109 NS NS — NS NS NS ― NS NS ― 0,02 ― ― ― Weiser98 2001 Sim 206 NS 0,007 NS NS NS NS ― ― — ― 0,0002 ― ― ― Wong99 2001 Sim 389 NS <0,001 ― — NS ― ― ― — ― ― ― <0,001 ― den Bakker100 2002 Não 32 — 0,03 NR — NR — — — — — — ― ― ― Jakub101 2002 Não 62 — NS — — — — — — — — — ― 0,06 ― Mignotte102 2002 Não 120 — NR — — — — — — — NR NR ― ― ― Sachdev103 2002 Sim 55 NS 0,0001 ― — NS 0,001 ― ― — ― 0,02 ― ― ― Hwang104 2003 Sim 131 NS 0,009 NS — NS 0,03 — NS NS — 0,024 NS NS 0,04 Fleming105 2004 Sim 54 NS NS — NS NS — — — — — 0,02 0,02 ― ―

Presente Estudo106,107 Sim 702+373 — 0,001 NS — NS 0,003 0,02 0,08 — <0,001 — ― <0,001 <0,001

10

A Tabela 3 resume a literatura que mostra a influência do tamanho do tumor

na incidência de metástase em Não-LS. Conforme esperado, quanto maior o tumor

maior a probabilidade de metástase em Não-LS.

A Tabela 4 refere-se às informações existentes sobre o impacto da invasão

vásculo-linfática na positividade de Não-LS, que é bem mais elevado quando essa

invasão é identificada.

A influência do tamanho da metástase em LS como fator preditivo de

metástase em Não-LS está evidente nos dados da literatura resumidos na Tabela 5.

Quanto maior a metástase no LS, ou quanto maior o número de focos metastáticos

identificados no LS, maior o risco de comprometimento de Não-LS.

Embora a casuística seja pequena, mesmo nos casos em que a metástase no

LS é identificada exclusivamente por IHQ, a incidência de metástase em Não-LS é

relevante, variando de 8% a 16% (Tabela 6).

A linfadenectomia axilar tem sido questionada quanto ao valor terapêutico108

e esse procedimento não é inócuo, estando associado a 78% de alteração de

sensibilidade na mama, 25% de dor local, 15% a 25% de linfedema e 17% de

restrição de movimento.41,109

O estudo mais clássico a questionar o valor terapêutico da linfadenectomia

axilar é o NSABP-B04 (National Surgical Adjuvant Breast and Bowel Project –

B04). Esse estudo comparou mastectomia radical, mastectomia total e mastectomia

total seguida de radioterapia. Os dados de 25 anos de acompanhamento revelam que

não há nenhum benefício em sobrevida associado ao tratamento cirúrgico axilar.110 A

crítica ao estudo do NSABP-B04 é que este não teria poder estatístico em mostrar

diferenças, devido ao pequeno número de pacientes em cada grupo.

11

Tabela 3 – Estudos que relatam a incidência de metástase em Não-LS em axila com LS positivo, por tamanho do tumor primário

Estudo

T1a (≤5cm)

T1b (0.6-1.0 cm)

T1c (1.1-2.0 cm)

T2 (2.1-5.0 cm)

T3 (>5.0 cm)

Chu85 0% 13% 29% 38% 71% Reynolds90 [______________ 25% para T1 ________________] [______ 79% para T2/3 _______] Turner93 17% 20% 46% 48% 73% Kamath86 25% 30% 40% 46% 80% Rahusen89 50% 50% 49% 50% ― Weiser98 8% 21% 37% 48% ― Wong99 14% 22% 30% 45% 57% Viale97 100% 14% 25% 24% ― Sachdev103 [______________ 13% para T1 ________________] [______ 33% para ≥T2 _______] Mignotte102 [____________ 14% ____________] 54% 52% ―

Não-LS, linfonodo axilar não-sentinela LS, linfonodo sentinela

12

Tabela 4 – Estudos que relatam a incidência de metástase em Não-LS em axila

com LS positivo, por presença de IVL no tumor primário

Estudo Sem IVL IVL Reynolds90 43% 62% Turner93 37% 65% Abdessalem87 31% 62% Rahusen89 42% 30% Weiser98 26% 41% Viale97 21% 26% Sachdev103 12% 32%

Não-LS, linfonodo axilar não-sentinela LS, linfonodo sentinela IVL, invasão vásculo-linfática

13

Tabela 5 – Estudos que relatam a incidência de metástase em Não-LS em axila

com LS positivo, por tamanho da metástase no LS

Estudo

<1 mm

≥1 mm

<2 mm

>2 mm

>2 cm ou “Doença extensa”

Chu85 ― ― 7% 55% ― Reynolds90 ― ― 22% 67% ― Turner93 ― ― 26% 63% ― Abdessalam87 ― ― 20% 47% 75% Kamath86 ― ― 15% 58% 65% Rahusen89 27% 50% ― ― ― Viale97 16% ― 22% 45% ― Weiser98 ― ― 18% 45% ― Mignotte102 ― ― 22% 79% ― Sachdev103 17% 49% ― ― ―

Não-LS, linfonodo axilar não-sentinela LS, linfonodo sentinela

14

Tabela 6 – Estudos que relatam a incidência de metástase em Não-LS em

pacientes com metástase em LS detectada exclusivamente por IHQ

Estudo Proporção % Teng91 3/26 12% Kamath86 2/26 8% Wong99 3/28 11% Mignotte102 7/44 16% Jakub101 9/62 15%

Não-LS, linfonodo axilar não-sentinela LS, linfonodo sentinela IHQ, imunoistoquímica

15

Orr et al.,111 em meta-análise de seis estudos randomizados, comparando

tratamentos com e sem linfadenectomia axilar, encontram um benefício de 5,4% na

sobrevida associada aos pacientes com axilas clinicamente negativas submetidos à

linfadenectomia. Esses achados, porém, também estão sujeitos a críticas. Nenhum

dos pacientes desses estudos recebeu tratamento adjuvante, embora hoje em dia a

grande maioria desses pacientes seria submetida a tratamento sistêmico adjuvante.

Portanto esse benefício teria um impacto menor para os padrões atuais.

A questão de a linfadenectomia ter papel terapêutico ou somente prognóstico

ainda hoje é assunto controverso e está sendo objeto do estudo prospectivo Z0011 do

Grupo de Oncologia do Colégio Americano de Cirurgiões (American College of

Surgeons Oncology Group – ACOSOG Z0011).112

16

4. CASUÍSTICA E MÉTODOS

O protocolo de estudo prospectivo de LS foi aprovado pelo Institutional

Review Board (IRB) do Memorial Sloan-Kettering Cancer Center – MSKCC, Nova

York, NY, EUA (Protocolo # 96-049).73 O IRB equivale ao Comitê de Ética. É o

órgão interno daquela instituição responsável pela aprovação de todos e quaisquer

projetos que envolvam pacientes e/ou seus respectivos dados. Todos os pacientes

foram tratados no MSKCC e também assinaram um Termo de Consentimento Livre

e Esclarecido para serem submetidos a cirurgias e para que seus dados e materiais

pudessem ser objetos de estudos pelos Serviços de Mastologia e Patologia do

MSKCC. Esse protocolo exclui pacientes com axila clinicamente positiva para

metástase axilar para a realização de biópsia de LS. Todos os dados utilizados nesta

tese são provenientes do banco de dados de LS do MSKCC, tendo sido coletados de

forma pré-definida e prospectiva.

Os projetos de estudos envolvidos nesta tese também foram aprovados pelo

IRB. Todos os dados, antes de serem extraídos do banco de dados original, foram

anonimizados de forma a impedir qualquer identificação dos pacientes, sendo,

portanto, dispensado o uso do termo de Consentimento Livre e Esclarecido

especificamente para esses projetos.

17

4.1. Técnica de Biópsia e de Avaliação de LS Sentinela do MSKCC113

4.1.1. Mapeamento Linfático e Biópsia de LS

Na véspera ou na manhã da cirurgia, 300 µCi de 99mTc (isótopo do tecnécio

de número atômico 99 metaestável) ligado a colóide sulfúrico não filtrado foi

administrado no paciente. Nos primeiros 224 casos essa injeção foi feita no

parênquima mamário (intraparenquimatosa). A partir de então a injeção passou a ser

feita de forma intradérmica na projeção cutânea do tumor.114-116 Cinqüenta a sessenta

minutos após a injeção foi realizado o estudo de linfocintilografia.

No centro cirúrgico, 4 ml de corante isosulfano azul (Lymphazurin; Zenith

Parentarals, Rosemont, IL, EUA) foram injetados de forma intraparenquimatosa ao

redor do tumor ou no subcutâneo na projeção deste, de 5 a 10 minutos antes da

incisão cutânea na axila. A biópsia do LS foi guiada pela sonda de radiação gama (C-

Trak, Care-Wise Medical, Morgan, CA, EUA) e pelo aparecimento do corante azul

no linfonodo ou linfonodos. A contagem de radiação gama dos linfonodos foi feita in

situ ou ex vivo, e posteriormente foi realizada a contagem de radiação na axila

(contagem de fundo), na procura de outros linfonodos, até que uma redução de 4

vezes fosse observada. Posteriormente, essa conduta foi mudada e todos os

linfonodos captantes foram ressecados, independentemente da contagem de fundo.

Como parte da fase de validação do procedimento pela equipe cirúrgica, em 326

casos foram realizados esvaziamentos axilares independentemente do resultado do

laudo anatomopatológico. A palpação intra-operatória do campo cirúrgico da axila

foi realizada de forma rotineira após a retirada dos LS identificados. Foram

18

chamados de LS qualquer linfonodo contendo material radioativo e/ou azul e/ou

suspeito à palpação.

4.1.2. Avaliação Anatomopatológica

Nos primeiros 78 casos, os espécimes patológicos foram examinados de

forma rotineira e corados por hematoxilina e eosina (H&E), não tendo sido realizada

a avaliação histopatológica intra-operatória de congelação.

Após esses casos iniciais a avaliação intra-operatória com congelação foi

realizada de forma rotineira. Quando o tamanho do LS não permitia a avalição

histopatológica de congelação, todo o material foi submetido a estudo de rotina em

parafina. Não foi realizada a citologia de contato (do inglês – imprint) em nenhum

dos casos.

De modo geral, assim que o patologista recebia o LS fresco, este foi cortado

em fatias de aproximadamente 2 a 3 mm de espessura. O corte central foi congelado

em nitrogênio líquido e realizado corte histológico de 5µm corado por H&E – estudo

histopatológico de congelação (chamado H&E congelação). Todo o material,

congelado ou não, foi posteriormente fixado em formalina a 10% e embebido em

parafina. Posteriormente, outro corte de 5 µm foi realizado no nível correspondente

ao controle da congelação (chamado estudo histopatológico de rotina ou H&E

rotina). Se nesse estudo histopatológico de rotina fosse identificada metástase,

nenhum outro estudo mais detalhado seria realizado no material dos LS. Caso o

estudo histopatológico de rotina fosse negativo, fazia-se um estudo mais

aprofundado. Esse estudo mais aprofundado consistia de: cortes seriados a cada 50

µm, sendo feitos dois pares de lâminas (5 µm cada) em cada nível de 50 µm, uma

19

lâmina corada por H&E e outra submetida a IHQ com anticorpos anticitoqueratina

CAM 5.2 ou AE1:AE2 (Becton Dickinson Immunocytometry Systems, San Jose,

CA, EUA) de forma intercalada. Os primeiros 78 casos foram retrospectivamente

submetidos ao estudo por IHQ e cortes seriados. Uma média de 10 a 20 níveis por

linfonodo foi realizada. O protocolo de biópsia de LS previa que os pacientes com

metástase em LS fossem submetidos à linfadenectomia axilar (níveis I e II ou níveis

I, II e III). Os linfonodos adicionais (Não-LS) dissecados foram analisados através do

estudo histopatológico de rotina descrito acima.74,106,107

Todos os tumores foram submetidos ao estudo anatomopatológico de rotina.

Todos os casos em que foram feitas biópsias em outras instituições foram revistos

por patologistas no MSKCC. Sempre que possível foi realizado IHQ nos tumores

para receptores de estrógeno e progesterona e HER2/neu (c-erb-B2) no MSKCC.

4.2. Descrição das variáveis estudadas

O tamanho patológico do tumor primário corresponde ao tamanho do

componente invasivo em centímetros e, quando não expresso de forma contínua, teve

seu valor expresso na forma categórica de acordo com o sistema de estadiamento

TNM.3

A multifocalidade foi definida como foco tumoral separado do foco do tumor

primário, não sendo feita distinção entre multifocalidade e multicentricidade. Nos

casos de tumores multifocais ou naqueles uniforais que comprometiam mais de um

quadrante, foram realizadas reclassificações das localizações de forma a priorizar a

seguinte ordem decrescente: quadrante súpero-lateral (QSL), “quadrante” central,

quadrante ínfero-lateral (QIL), quadrante ínfero-medial (QIM), quadrante súpero-

20

medial (QSM). Por exemplo, para um tumor em QSM e central, a localização seria

reclassificada como “central”; na situação de tumor em região central e QSL, este

seria reclassificado como “QSL”. Se todos os quadrantes fossem envolvidos, a

localização seria classificada como “QSL”.

Com o objetivo de análise dos dados, e não como classificação

anatomopatológica clássica, os tipos histológicos foram reclassificados (ou

agrupados) como Ductal, Lobular ou Especial. Os tipos histológicos classificados por

carcinoma medular atípico e carcinoma com características ducto-lobulares foram

agregados ao grupo denominado Ductal. Na amostra analisada havia um caso de

carcinoma secretor que foi agrupado no subgrupo Especial, que inclui os tipos

histológicos tubular, colóide e medular típico.

Como padrão no Departamento de Patologia do MSKCC, o carcinoma lobular

não recebe graduação histológica ou nuclear.

O grau nuclear I foi definido como nenhuma ou pequena variação no tamanho

ou formato do núcleo da célula neoplásica; o grau nuclear II, variação moderada no

tamanho ou formato do núcleo; o grau nuclear III, variação acentuada do tamanho ou

formato do núcleo.117

O grau histológico foi definido através do sistema de Scarff-Bloom-

Richardson (SBR) modificado. Este foi determinado pela combinação do grau de

diferenciação, da extensão do pleomorfismo e do índice mitótico. O grau histológico,

dessa forma, foi classificado em I, II ou III.117

A invasão vásculo-linfática foi definida como a presença (positiva) ou

ausência (negativa) de uma ou mais células tumorais dentro de vaso linfático ou

21

sangüíneo. Os patologistas do MSKCC não diferenciam nos laudos o envolvimento

vascular linfático de sangüíneo.

Os receptores hormonais, receptor de estrógeno e progesterona, foram

considerados negativos quando sua expressão na IHQ foi menor que 10% das células

coradas e, positivos, quando essa expressão foi maior que 10%.

Os métodos de detecção de metástases nos linfonodos foram classificados

em: Somente IHQ, Cortes Seriados, H&E Congelação, H&E Rotina. Seus

respectivos significados são: “Somente IHQ”, quando as metástases em linfonodo

foram detectadas somente por IHQ; “Cortes Seriados”, quando as metástases foram

detectadas em cortes seriados corados por H&E ou nos cortes em H&E após

identificação na IHQ; “H&E Congelação”, quando detectadas no exame

anatomopatológico intra-operatório de congelação; “H&E Rotina”, quando

detectadas somente no controle pós-operatório da congelação ou no estudo rotineiro

em H&E com um único corte na metade do linfonodo.

4.3. Modelos Estatísticos

Dois modelos estatísticos foram desenvolvidos. O primeiro modelo tem o

objetivo de calcular a probabilidade de metástase em LS e o segundo, a

probabilidade de metástase em Não-LS quando o LS for positivo para metástase.

Ambos os modelos foram validados em amostras distintas. Portanto, para

cada modelo existem amostras denominadas de “modelagem” e de “validação”.

Para facilitar o uso dos modelos, foram desenvolvidos softwares, que podem

ser encontrados no CD anexo a esta tese.

22

4.3.1. Modelo de Metástase em Linfonodo Sentinela

4.3.1.1. Casuística do modelo de metástase em LS

Neste estudo foram avaliados 6.645 procedimentos.

A amostra de modelagem consiste nos primeiros 4.608 procedimentos de

biópsia de LS realizados entre 12 de setembro de 1996 e 31 de julho de 2002.

Os critérios de exclusão foram: sarcoma, neoplasia lobular, carcinoma

intracístico ou adenóide invasivo, quimioterapia neoadjuvante, ausência de tamanho

tumoral (Tx), LS não identificado ou falso-negativo e doença benigna.

Foram excluídos 163 casos em que não foi identificada neoplasia maligna, 16

casos de sarcomas, 3 casos de carcinoma intracístico invasivo, 3 casos de carcinoma

adenóide cístico, 53 casos de quimioterapia neoadjuvante, 86 casos em que o

tamanho exato não pôde ser determinado (Tx), 169 casos em que nenhum LS foi

ressecado ou o LS foi negativo e os Não-LS foram positivos (falso-negativos). Após

as exclusões, os 4.115 procedimentos remanescentes constituíram a amostra de

modelagem sobre a qual foi desenvolvido o modelo estatístico.

A amostra de validação consiste em 2.037 procedimentos de biópsia de LS

realizados entre 1 de agosto de 2002 e 21 de maio de 2004.

Foram excluídos 114 casos em que não foi identificada neoplasia maligna, 9

casos de sarcomas, 4 casos de neoplasia lobular, 47 casos de quimioterapia

neoadjuvante, 31 casos em que o tamanho exato não pode ser determinado (Tx), 36

casos “falso-negativos”, em que o LS foi negativo e os Não-LS foram positivos e 33

23

casos em que nenhum LS foi ressecado. Após as exclusões, os 1.763 procedimentos

remanescentes constituíram a amostra de validação do modelo.

As variáveis estudadas foram: idade, sexo, tamanho do tumor (T),

lateralidade, bilateralidade, palpabilidade, localização do tumor na mama

(quadrante), tipo de cirurgia, tipo histológico do tumor, grau histológico, grau

nuclear, receptor de estrógeno, receptor de progesterona, HER-2/neu,

multifocalidade, invasão vásculo-linfática, metástase em LS e método de detecção

das metástases.

Inicialmente foram analisados os 4.115 procedimentos realizados entre 12 de

setembro de 1996 e 31 de julho de 2002, para se determinarem os fatores que

apresentam correlação estatística com metástase em LS e se desenvolver um modelo

estatístico para prever a probabilidade de metástase em LS.

Posteriormente, esse modelo de regressão logística de previsão de metástase

em LS foi validado nos demais 1.763 procedimentos consecutivos após 1 de agosto

de 2002.

4.3.1.2. Estatística do modelo de metástase em LS

As análises estatísticas foram feitas por meio dos programas SPSS, versão

10.0.5 (SPSS Inc. Chicago, IL, EUA), e StatXact, vesão 4.0.1 (Cytel Software

Corporation, Cambridge, MA, EUA). A terminologia estatística obedece às

recomendações de Bevilacqua.118

Análise Univariada

24

O teste de χ2 Pearson bicaudal foi utilizado para determinar as correlações

entre as variáveis categóricas. O teste soma de postos de Wilcoxon foi utilizado para

determinar as correlações entre as variáveis contínuas, como idade e tamanho do

tumor, uma vez que ambas não apresentavam distribuição normal.

Análise Multivariada

As variáveis que demonstraram correlação estatística com metástase em LS

na análise univariada (definida como P < 0,05) foram analisadas pela regressão

logística binária utilizando-se os métodos de entrada e por etapas. Modelos de

regressão logística para prever metástase em LS foram desenvolvidos com base na

amostra de modelagem. Um modelo contínuo foi criado usando as variáveis, idade e

tamanho tumoral, de forma contínua. Esse modelo foi posteriormente validado com a

amostra de validação.

A capacidade dos modelos em discriminar corretamente as probabilidades

observadas e calculadas, ou seja, discriminar corretamente a probabilidade de

metástase em LS foi medida pelo cálculo da área sob a curva de características

operacionais, mais comumente conhecida como curva ROC (do inglês ROC curve –

receiver operating characteristic curve) e pelo gráfico de plotagem das

probabilidades observadas e calculadas.

Para facilitar a aplicação do modelo na prática, foi desenvolvido um software

com base no modelo contínuo. Esse software encontra-se no CD anexo a esta tese.

25

4.3.2. Modelo de Metástase em Linfonodo Não-Sentinela

4.3.2.1. Casuística do modelo de metástase em Não-LS

Neste estudo foram avaliados 4.790 procedimentos.

A amostra de modelagem consiste nos primeiros 3.207 procedimentos de

biópsia de LS realizados entre 12 de setembro de 1996 e 24 de abril de 2001.

Os critérios de exclusão foram: sarcoma, neoplasia lobular, carcinoma

intracístico ou adenóide invasivo, carcinoma in situ, ausência de neoplasia maligna,

axila clinicamente positiva, tratamento neoadjuvante, ausência de tamanho tumoral

(Tx), LS negativo, LS não identificado ou falso-negativo, não ter sido realizada

linfadenectomia complementar.

Foram excluídos 78 casos em que não foram realizados IHQ no LS de forma

previamente programada, 33 casos de quimioterapia neoadjuvante, 82 casos em que

nenhum linfonodo axilar foi ressecado, 2.263 casos de LS negativos ou, mais

raramente, nos casos em que não houve intenção de se realizar dissecção axilar,

como por exemplo, em situações em que o LS era positivo e por opção do paciente e

cirurgião optou-se por não completar o esvaziamento. Excluíram-se também 34 casos

em que menos de 10 linfonodos foram dissecados, 15 casos por motivos variados

(por exemplo: carcinoma ductal in situ, ausência de neoplasia maligna, outros tipos

histológicos raros ou axila clinicamente suspeita). Após as exclusões, os 702

procedimentos remanescentes constituíram a amostra de modelagem sobre a qual foi

desenvolvido o modelo estatístico.

26

A amostra de validação consiste em 1.583 procedimentos de biópsia de LS

realizados entre 25 de abril de 2001 e 20 de setembro de 2002.

Foram excluídos 26 casos de quimioterapia neoadjuvante, 33 casos em que

nenhum linfonodo axialr foi ressecado, 1.117 casos de LS negativos ou em que não

houve intenção de se realizar dissecção axilar por qualquer motivo (como descrito

acima), 21 casos em que menos de 10 linfonodos foram dissecados, 9 casos por

motivos variados (por exemplo: carcinoma ductal in situ, ausência de neoplasia

maligna, outros tipos histológicos raros ou axila clinicamente suspeita), 4 casos com

antecedente de carcinoma na mesma mama. Após as exclusões, os 373

procedimentos remanescentes constituíram a amostra de validação.

As variáveis estudadas foram: idade, tamanho do tumor (T), tipo histológico,

grau nuclear, invasão vásculo-linfática, multifocalidade, receptor de estrógeno,

método de detecção de metástase em LS, número de LS positivos e negativos.

As variáveis como receptor de progesterona e o grau histológico estão muito

correlacionadas a receptor de estrógeno e grau nuclear, respectivamente, por isso não

foram consideradas neste modelo.

Nesse modelo, em particular, o tipo histológico e o grau nuclear foram

agrupados em uma única variável, sendo que os tipos histológicos Ductal e Especial

foram agrupados e classificados pela graduação nuclear. Dessa forma, as categorias

nesta “nova” variável são: Ductal I – correspondendo à histologia ductal ou especial

com grau nuclear I; Ductal II – correspondendo à histologia ductal ou especial com

grau nuclear II; Ductal III – histologia ductal ou especial com grau nuclear III;

Lobular – correspondendo à histologia lobular sem classificação nuclear.

27

Primeiramente foram analisados os 702 procedimentos realizados entre 12 de

setembro de 1996 e 24 de abril de 2001, para determinar os fatores que apresentam

correlação estatística com metástase em Não-LS e se desenvolver um modelo

estatístico para prever o risco de metástase em Não-LS.

Posteriormente, esse modelo de previsão de metástase em Não-LS foi

validado nos demais 373 procedimentos consecutivos realizados após 25 de abril de

2001.

4.3.2.2. Estatística do modelo de metástase em Não-LS

As análises estatísticas e os nomogramas foram desenvolvidos por meio do

programa S-Plus, Profissional Edition with the Design Library versão 2000

(Mathsoft, Data Analysis Products Division, Seattle, WA, EUA).119,120 A

terminologia estatística obedece à recomendações de Bevilacqua.118

Nomograma, segundo o Novo Dicionário Aurélio da Língua Portuguesa,121 é

um “gráfico, com curvas apropriadas, mediante o qual se podem obter as soluções de

uma equação determinada pelo simples traçado de uma reta”.

Análise Multivariada

A associação de cada variável com a presença de metástase em Não-LS foi

analisada por regressão logística multivariada na amostra de modelagem. Um

nomograma foi desenvolvido com base nessa análise. Esse nomograma foi validado

posteriormente pela amostra de validação.

28

A capacidade dos modelos em discriminar corretamente entre as

probabilidades observadas e calculadas, ou seja, discriminar corretamente a

probabilidade de metástase em Não-LS foi medida pelo cálculo da área sob a curva

ROC e pelo gráfico de plotagem das probabilidades observadas e calculadas.

Dois modelos de análises multivariadas por regressão logística foram

desenvolvidos: Modelo com congelação e Modelo sem congelação. O que diferencia

os modelos é o método de detecção de metástase em LS. No modelo com

congelação, o estudo de congelação está separado do método de rotina por H&E; já

no modelo sem congelação, os casos de identificação de metástase por “congelação”

foram reclassificados como “rotina”. Optou-se por criar esses dois tipos de modelos,

uma vez que o estudo de congelação não está disponível ou não é realizado em todos

os centros.

Dois nomogramas foram desenvolvidos baseados nesses modelos de análises

multivariadas. Esses nomogramas foram denominados “Nomograma com

Congelação” e “Nomograma sem Congelação”.

Os nomogramas foram desenvolvidos por Michael W. Kattan, PhD,

estatístico do MSKCC.

Para facilitar o uso dos nomogramas, o setor de estatística do MSKCC

desenvolveu um software que se encontra no CD anexo a esta tese.

29

5. RESULTADOS

5.1. Modelo de Metástase em Linfonodo Sentinela

As variáveis analisadas para o desenvolvimento da modelagem e da validação

estão expressas na Tabela 7. De maneira geral, pode-se observar que não existiram

grandes diferenças entre as duas amostras.

Ao se analisarem ambas as amostras (Tabela 7), observa-se que: o tamanho

mediano dos tumores foi de 1,1 a 1,2 cm, sendo a maioria menor que 3 cm; a idade

mediana foi de aproximadamente 55 anos, principalmente entre a faixa etária de 41 e

69 anos; a maioria dos tumores era unilateral e no quadrante súpero-lateral; o

tratamento conservador foi o principal tratamento cirúrgico realizado; a histologia

ductal foi a mais incidente; o grau nuclear tipo III e o grau histológico tipo II foram

os mais freqüentes; houve predomínio da positividade dos receptores hormonais;

houve menor freqüência de tumores unifocais e sem invasão vásculo-linfática.

Aproximadamente um terço dos casos apresentava metástase em LS, sendo a maioria

dessas diagnosticada pelo estudo de congelação.(Figura 1).

As Tabelas 8A e 8B mostram a incidência de metástase em LS de acordo com

as variáveis estudadas nas amostras de modelagem, validação e em ambas.

30

Tabela 7 – Características dos pacientes e tumores – Amostras de Modelagem e de Validação – Modelo de Metástase em LS

Modelagem N (%)

Validação N(%)

Modelagem N (%)

Validação N(%)

Número de casos 4115 (100,0) 1763 (100.0) Tipo de Cirurgia Conservadora 2799 (68,0) 1092 (61,9) Idade em anos Mastectomia (MRM ou MT) 1316 (32,0) 668 (37,9) Mediana (Min-Max) 56 (20-91) 55 (23-90) Desconhecido ― 3 (0,2) Tamanho do tumor (cm) Tipo de Tumor Mediana (Min-Max) 1,2 (0-11,0) 1,1 (0-9,5) Ductal 3574 (86,9) 1556 (88,3) Lobular 372 (9,0) 166 (9,4) Sexo Colóide/medular/tubular 169 (4,1) 41 (2,3) Feminino 4079 (99,1) 1747 (99,1) Masculino 36 (0,9) 16 (0,9) Grau Histológico* I 331 (8,0) 97 (5,5) Idade (Anos) II 1034 (25,1) 378 (21,4) ≤ 40 419 (10,2) 197 (11,2) III 1785 (43,4) 822 (46,6) 41-69 2978 (72,4) 1239 (70,3) Desconhecido/Não aplicável 965 (23,5) 466 (26,4) ≥ 70 718 (17,4) 75 (18,5) Grau Nuclear* Tamanho do Tumor (T) I 323 (7,8) 112 (6,4) Tis 329 (8,0) 218 (12,4) II 1869 (45,4) 750 (42,5) T1mic 131 (3,2) 51 (2,9) III 1206 (29,3) 635 (36,0) T1a 422 (10,3) 199 (11,3) Desconhecido/Não aplicável 717 (17,4) 266 (15,1) T1b 963 (23,4) 362 (20,5) T1c 1582 (38,4) 624 (35,4) Receptor de Estrógeno T2 ≤ 3,0 cm 495 (12,0) 215 (12,2) Negativo 649 (15,8) 277 (15,7) T2 > 3,0 cm 160 (3,9) 80 (4,5) Positivo 2485 (60,4) 1191 (67,6) T3 33 (0,8) 14 (0,8) Desconhecido 981 (23,8) 295 (16,7) Lateralidade Receptor de Progesterona Direito 2063 (50,1) 861 (48,8) Negativo 1239 (30,1) 605 (34,3) Esquerdo 2052 (49,9) 902 (51,2) Positivo 1878 (45,6) 862 (48,9) Desconhecido 998 (24,3) 296 (16,8) Bilateralidade Não 3809 (92,6) 1607 (91,2) HER-2/neu Sim 306 (7,4) 156 (8,8) Negativo 1888 (45,9) ND Positivo 615 (14,9) ND Palpabilidade Desconhecido 1612 (39,2) ND Não 1514 (36,8) ND Sim 1637 (39,8) ND Multifocalidade Desconhecido 964 (23,4) ND Não 3112 (75,6) 1284 (72,8) Sim 1003 (24,4) 479 (27,2) Localização do Tumor QSL 2311 (56,2) 1101 (56,8) Invasão Vásculo-Linfática QIL 521 (12,7) 236 (13,4) Não 3316 (80,6) 1422 (80,7) QSM 612 (14,9) 289 (16,4) Sim 799 (19,4) 341 (19,3) QIM 299 (7,3) 163 (9,2) Central 323 (7,8) 74 (4,2) Mestástase em LS Desconhecido 49 (1,2) ― Não 2829 (68,7) 1160 (65,8) Sim 1286 (31,3) 603 (34,2)

LS, linfonodo sentinela Min, valor mínimo; Max, valor máximo cm, centímetro ND, não disponível QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial MRM, mastectomia radical modificada MT, mastectomia total *, classificado somente na histologia Ductal

31

Estádio T N Somente

IHQ N (%)

Cortes Seriados*

N (%)

H&E Congelação

N(%)

H&E Rotina N (%)

Total N (%)

Tis 329 14 (4,3) 15 (4,6) 1 (0,3) 5 (1,5) 35 (10,6) T1mic 131 7 (5,3) 4 (3,1) 2 (1,5) 1 (0,8) 14 (10,7) T1a 422 14 (3,3) 22 (5,2) 11 (2,6) 10 (2,4) 57 (13,5) T1b 963 32 (3,3) 53 (5,5) 98 (10,2) 27 (2,8) 210 (21,8) T1c 1582 64 (4,0) 102 (6,4) 340 (21,5) 58 (3,7) 564 (37,5) T2 (≤3cm) 495 22 (4,4) 39 (7,9) 178 (36,0) 32 (6,5) 271 (54,7) T2 (>3cm) 160 9 (5,6) 10 (6,3) 77 (48,1) 10 (6,3) 106 (63,3) T3 33 3 (9) 1 (3) 20 (61) 5 (15) 29 (88) Total 4115 165 (4,0) 246 (6,0) 727 (17,7) 148 (3,6) 1286 (31,3)

* Cortes Seriados: Inclui metástases vistas em cortes seriados corados por hematoxilina e eosina (H&E) ou vistos nos cortes em H&E após identificação na imunoistoquímica (IHQ) LS, linfonodo sentinela

Figura 1 – Freqüência de metástase em LS por estádio T e método de detecção –

Amostra de Modelagem – Modelo de Metástase em LS (A tabela acima faz parte da Figura 1 e expressa os valores numéricos desta)

0

25

50

75

100

Met

ásta

se e

m L

S (%

)

T2(>3cm) T3 T1b T1a T1mic T2(≤ 3cm) T1c N=422 N=963 N=33 N=131 N=1582 N=495 N=160

13,5%

21,8%

37,5%

54,7%

63,3%

88%

Tis N=329

10,6 10,7

Total N=4115

31,3%

32

Tabela 8A – Incidência de metástases em LS – Amostras de Modelagem e de Validação – Modelo de Metástase em LS

LS, linfonodo sentinela LS+, linfonodo sentinela positivo ND, não disponível QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial MRM, mastectomia radical modificada

Modelagem Validação Total Número

(N) LS + (N)

LS + (%)

Número

(N) LS + (N)

LS + (%)

Número

(N) LS + (N)

LS + (%)

Número de casos 4115 1286 31,3 1763 603 34,2 5878 1889 32,1 Sexo Feminino 4079 1271 31,2 1747 596 33,9 5826 1867 32,0 Masculino 36 15 42 16 10 10/16 52 25 48 Idade (Anos) ≤ 40 419 175 41,8 197 88 44,7 616 263 42,7 41-69 2978 935 31,4 1239 426 34,4 4217 1361 32,3 ≥ 70 718 176 24,5 75 89 27 793 265 33,4 Tamanho do Tumor (T) Tis 329 35 10,6 218 24 11,0 547 59 10,8 T1mic 131 14 10,7 51 11 21 182 25 13,7 T1a 422 57 13,5 199 31 15,6 621 88 14,2 T1b 963 210 21,8 362 84 23,2 1325 294 22,2 T1c 1582 564 35,7 624 263 42,1 2206 827 37,5 T2 ≤ 3,0 cm 495 271 54,7 215 121 56,3 710 392 55,2 T2 > 3,0 cm 160 106 66,3 80 56 70 240 162 67,5 T3 33 29 88 14 13 13/14 47 42 89 Lateralidade Direito 2063 619 30,0 861 290 33,7 2924 909 31,1 Esquerdo 2052 667 32,5 902 313 34,7 2954 980 33,2 Bilateralidade Não 3809 1206 31,7 1607 554 34,5 5416 1760 32,5 Sim 306 80 26,1 156 49 31,4 462 129 27,9 Palpabilidade Não 1514 320 21,1 ND ND ND 1514 320 21,1 Sim 1637 642 39,2 ND ND ND 1637 642 39,2 Desconhecido 964 324 33,6 ND ND ND 964 324 33,6 Localização do Tumor QSL 2311 742 32,1 1101 347 34,7 3412 1089 31,9 QIL 521 188 36,1 236 93 39,4 757 281 37,1 QSM 612 147 24,0 289 74 25,6 901 221 24,5 QIM 299 84 28,1 163 51 31,3 462 135 29,2 Central 323 103 31,9 74 38 51 397 141 35,5 Desconhecido 49 22 45 ― ― ― 49 22 45

33

Tabela 8B – Incidência de metástases em LS – Amostras de Modelagem e de

Validação – Modelo de Metástase em LS

LS, Linfonodo sentinela LS +, Linfonodo sentinela positivo ND, Não disponível MRM, Mastectomia radical modificada MT, Mastectomia total *, Classificado somente na histologia Ductal


(N) LS + (N)

LS + (%)

Número

(N) LS + (N)

LS + (%)

Número

(N) LS + (N)

LS + (%)

Número de casos 4115 1286 31,3 1763 603 34,2 5878 1889 32,1 Tipo de Cirurgia Conservadora 2799 743 26,5 1092 314 28,8 3891 1057 27,2 Mastectomia (MRM ou MT) 1316 543 41,3 668 289 43,3 1984 832 41,9 Desconhecido ― ― ― 3 0 0/3 3 0 0/3 Tipo de Tumor Ductal 3574 1127 31,5 1556 538 34,6 5130 1665 32,5 Lobular 372 150 40,3 166 61 36,7 538 211 39,2 Colóide/medular/tubular 169 9 5,3 41 4 10 210 13 6,2 Grau Histológico* I 331 41 12,4 97 11 11,3 428 52 12,1 II 1034 342 33,1 378 141 37,3 1412 483 34,2 III 1785 690 38,7 822 351 42,7 2607 1041 39,9 Desconhecido/Não aplicável 965 213 22,1 466 100 21,5 1431 313 21,9 Grau Nuclear* I 323 40 12,4 112 23 20,5 435 63 14,5 II 1869 569 30,4 750 249 33,2 2619 818 31,2 III 1206 464 38,5 635 246 38,7 1841 710 38,6 Desconhecido/Não aplicável 717 213 29,7 266 85 31,9 983 298 30,3 Receptor de Estrógeno Negativo 649 221 34,1 277 101 36,5 926 322 34,8 Positivo 2485 891 35,9 1191 466 39,1 3676 1357 36,9 Desconhecido 981 174 17,7 295 36 12,2 1276 210 16,5 Receptor de Progesterona Negativo 1239 407 32,8 605 221 36,5 1844 628 34,1 Positivo 1878 699 37,2 862 346 40,1 2740 1045 38,1 Desconhecido 998 180 18,0 296 36 12,1 1294 216 16,7 HER-2/neu Negativo 1888 679 36,0 ND ND ND 1888 679 36,0 Positivo 615 242 39,3 ND ND ND 615 242 39,3 Desconhecido 1612 365 22,6 ND ND ND 1612 365 22,6 Multifocalidade Não 3112 893 28,7 1284 412 32,1 4396 1305 29,7 Sim 1003 393 39,2 479 191 39,9 1482 584 39,4 Invasão Vásculo-Linfática Não 3316 800 24,1 1422 368 25,9 4738 1168 24,7 Sim 799 486 60,8 341 235 68,9 1140 721 63,2

34

Ao se analisarem as Tabelas 8A e 8B, observa-se que não existiu nenhuma

grande diferença entre as amostras em relação às variáveis estudadas. Destaca-se, ao

se observarem os percentuais, que existiu uma tendência a aumento na incidência de

metástase em LS nos pacientes do sexo masculino, idade jovem, tumores grandes

(por exemplo, 10,6% em Tis versus 88% em T3 – Figura 1), palpáveis, de

localização central e lateral na mama, submetidos a mastectomia, histologias ductal e

lobular, grau nuclear e histológico III e II, positividade dos receptores hormonais,

presença de multifocalidade e invasão vásculo-linfática.

O estudo detalhado dos LS (cortes seriados mais IHQ) detectou 10% a mais

de metástases que não seriam identificadas pelo estudo do LS sem a utilização desses

recursos (Figura 1). Na Figura 1, observa-se que 4,0% das metástases foram

identificadas exclusivamente por IHQ e 6,0% nos cortes seriados dos LS. A

identificação de metástase em LS por IHQ e cortes seriados ocorreu em 3,6% e 5,7 %

nos carcinomas de histologia ductal (Figura 2), e 8,6% e 9,1% nos carcinomas de

histologia lobular (Figura 3). Isso mostra que a identificação de metástases em LS

com a utilização de estudo anatomopatológico mais refinado foi ainda mais relevante

para os carcinomas lobulares.

35


IHQ N (%)

Cortes Seriados*

N (%)

H&E Congelação

N(%)

H&E Rotina N (%)

Total N (%)

Tis 329 14 (4,3) 15 (4,6) 1 (0,3) 5 (1,5) 35 (10,6)

T1mic 129 7 (5,4) 4 (3,1) 2 (1,6) 1 (0,8) 14 (10,9)

T1a 382 13 (3,4) 19 (5,0) 9 (2,4) 9 (2,4) 50 (13,1)

T1b 884 31 (3,5) 46 (5,2) 92 (10,4) 26 (2,9) 195 (22,1)

T1c 1426 51 (3,6) 87 (6,1) 311 (21,8) 55 (3,9) 504 (35,3)

T2 (≤3cm) 433 11 (2,5) 32 (7,4) 157 (36,3) 29 (6,7) 229 (52,9)

T2 (>3cm) 138 5 (3,6) 8 (5,8) 68 (49,3) 9 (6,5) 89 (65,2)

T3 22 1 (5) 1 (5) 13 (59) 4 (18) 19 (86)

Total 3743 133 (3,6) 212 (5,7) 653 (17,4) 138 (3,7) 1136 (30,6)


Figura 2 – Freqüência de metástase em LS em carcinoma ductal (incluindo

subtipos especiais) por estádio T e método de detecção – Amostra de Modelagem – Modelo de Metástase em LS (A tabela acima faz parte da Figura 2 e expressa os valores numéricos desta)

0

25

50

75

100

Met

ásta

se e

m L

S (%

)


13,1%

22,1%

35,3%

52,9%

65,2%

86%

Tis N=329

10,6 10,9

30,6%

Total N=3743

36


IHQ N (%)

Cortes Seriados*

N (%)

H&E Congelação

N(%)

H&E Rotina N (%)

Total N (%)

T1mic 2 0 0 0 0 0

T1a 40 1 (3) 3 (8) 2 (5) 1 (3) 7 (18)

T1b 79 1 (1) 7 (9) 6 (8) 1 (1) 15 (19)

T1c 156 13 (8,3) 15 (9,6) 29 (18,6) 3 (1,9) 60 (38,5)

T2 (≤3cm) 62 11 (18) 7 (11) 21 (34) 3 (5) 42 (68)

T2 (>3cm) 22 4 (18) 2 (9) 9 (41) 1 (5) 16 (73)

T3 11 2 0 7 1 10

Total 372 32 (8,6) 34 (9,1) 74 (19,9) 10 (2,7) 150 (40,3)


Figura 3 – Freqüência de metástase em LS em carcinoma lobular por estádio T

e método de detecção – Amostra de Modelagem – Modelo de Metástase em LS (A tabela acima faz parte da Figura 3 e expressa os valores numéricos desta)

0

25

50

75

100

Met

ásta

se e

m L

S (%

)


0/2

18% 19%

38,5%

68% 73%

10/11

Total N=372

40,3%

37

Foram realizadas análises univariadas para se testar estatisticamente as

correlações das variáveis com a presença de metástase em LS (Tabelas 9A, 9B e 10).

Essas análises foram realizadas somente na amostra de modelagem. As variáveis

lateralidade e tipo de cirurgia não foram submetidas a análises univariadas, uma vez

que não existe substrato lógico para esta inclusão. A bilateralidade também foi

excluída uma vez que não foi possível diferenciar sincronicidade e metacronicidade

nesses casos.

As variáveis que mostraram correlação estatística com metástase em LS, pela

análise univariada, foram: idade (forma contínua ou discriminada), tamanho do

tumor (forma contínua ou discriminada – Estádio T), palpabilidade, localização do

tumor, invasão vásculo-linfática, multifocalidade, receptor de progesterona, grau

histológico e grau nuclear (Tabelas 9A, 9B e 10).

A variável tipo de tumor também mostrou diferenças em relação à presença

de metástase em LS, principalmente os subtipos especiais que sabidamente

apresentam menor incidência de metástase linfonodal.122 Ao contrário do que se

poderia esperar, foi observado que a histologia lobular apresentava maior incidência

de metástase em LS que a histologia ductal (40,3% versus 31,5%, respectivamente,

P = 0,001). Para se caracterizar essa inesperada ou eventual diferença, foram

realizadas análises mais detalhadas em relação às histologias tumorais (Tabelas 10,

11 e 12 e Figuras 2 e 3).

38

Tabela 9A – Análise univariada – Variáveis contínuas – Amostra de Modelagem – Modelo de Metástase em LS

LS Negativo LS Positivo P† Mediana (Min-Max) Idade (anos) 57 (20-91) 53 (21-89) <0,001 Tamanho do tumor (cm) 1,0 (0-6,1) 1,5 (0-11) <0,001

† Teste soma de postos de Wilcoxon Min, valor mínimo Max, valor máximo cm, centímetro LS, linfonodo sentinela

39

Tabela 9B – Análise univariada – Variáveis categóricas – Amostra de Modelagem – Modelo de Metástase em LS

Número

(N) LS +

N (%) P†

Número (N)

LS + N (%)

P†

Idade–anos RE ≤ 40 419 175 (41,8) Negativo 649 221 (34,1) 0,393 41-69 2978 935 (31,4) <0,001 Positivo 2485 891 (35,9) ≥ 70 718 176 (24,5) EP Sexo Negativo 1239 407 (32,8) 0,013 Feminino 4079 1271 (31,2) 0,176 Positivo 1878 699 (37,2) Masculino 36 15 (42) HER-2/neu Palpabilidade Negativo 1888 679 (36,0) 0,131 Não 1514 320 (21,1) <0,001 Positivo 615 242 (39,3) Sim 1637 329 (38,1) Estádio T Localização do Tumor Tis 329 35 (10,6) QSL 2311 742 (32,1) T1mic 131 14 (10,7) QIL 521 188 (36,1) T1a 422 57 (13,5) QSM 612 147 (24,0) <0,001 T1b 963 210 (21,8) <0,001 QIM 299 84 (28,1) T1c 1582 564 (35,7) Central 323 103 (31,9) T2 (≤ 3cm) 495 271 (54,7) T2 (> 3cm) 160 106 (63,3) Localização do Tumor T3 33 29 (88) Medial 911 231 (25,4) <0,001 Outro 3155 1033 (32,7) Grau Histológico I 331 41 (12,4) Invasão Vásculo-Linfática II 1034 342 (33,1) <0,001 Não 3316 800 (24,1) <0,001 III 1785 690 (38,7) Sim 799 486 (60,8) Grau Nuclear Multifocalidade I 323 40 (12,4) Não 3112 893 (28,7) <0,001 II 1869 569 (30,4) <0,001 Sim 1003 393 (39,2) III 1206 464 (38,5)

† Teste de χ2 Pearson bicaudal – desconhecidos foram excluídos da análise estatística LS, linfonodo sentinela LS +, linfonodo sentinela positivo RE, receptor de erstrógeno RP, receptor de progesterona QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial

40

Tabela 10 – Análise univariada – Comparação dos tipos histológicos – Amostra

de Modelagem – Modelo de Metástase em LS

Número (N)

LS Positivo N (%)

P†

Tipo do Tumor Lobular 372 150 (40,3) Ductal 3574 1127 (31,5) <0,001 Colóide/medular/tubular 169 9 (5,3) Tipo do Tumor Ductal 3574 1127 (31,5) <0,001 Colóide/medular/tubular 169 9 (5,3) Tipo do Tumor Lobular 372 150 (40,3) 0,001 Ductal 3574 1127 (31,5)

† Teste de χ2 Pearson bicaudal LS, linfonodo sentinela

41

Tabela 11 – Características dos pacientes e tumores – Ductal versus Lobular – Amostra de Modelagem – Modelo de Metástase em LS

LS, linfonodo sentinela Min, valor mínimo Max, valor máximo cm, centímetro QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial MRM, mastectomia radical modificada; MT, mastectomia total

Ductal N (%)

Lobular N(%)

Ductal N (%)

Lobular N(%)

Número de casos 3743 (100,0) 372 (100.0) Tipo de Cirurgia Conservadora 2560 (68,4) 239 (64,2) Idade em anos Mastectomia (MRM ou MT) 1183 (31,6) 133 (35,8) Mediana (Min-Max) 55 (20-89) 59 (34-91) Tipo de Tumor Tamanho do tumor (cm) Ductal 3574 (95,5) - Mediana (Min-Max) 1,1 (0-7,5) 1,5 (0,1-11) Colóide/medular/tubular 169 (4,5) - Sexo Grau Histológico Feminino 3709 (99,1) 370 (99,5) I 331 (8,8) - Masculino 34 (0,9) 2 (0,5) II 1034 (27,6) - III 1785 (47,7) - Idade (Anos) Desconhecido 593 (15,8) - ≤ 40 410 (11,0) 9 (2.4) 41-69 2690 (71,9) 288 (77,4) Grau Nuclear ≥ 70 643 (17,2) 75 (20,2) I 323 (8,6) - II 1869 (49,9) - Tamanho do Tumor (T) III 1206 (32,2) - Tis 329 (8,8) - Desconhecido 345 (9,2) - T1mic 129 (3,4) 2 (0,5) T1a 282 (10,2) 40 (10,8) Receptor de Estrógeno T1b 884 (23,6) 79 (21,2) Negativo 627 (16,8) 22 (5,9) T1c 1426 (38,1) 156 (41,9) Positivo 2182 (58,3) 303 (81,5) T2 ≤ 3,0 cm 571 (15,3) 62 (16,7) Desconhecido 934 (25,0) 47 (12,6) T2 > 3,0 cm 22 (0,6) 22 (5,9) T3 7 (0,4) 11 (3,0) Receptor de Progesterona Negativo 1126 (30,1) 113 (30,4) Lateralidade Positivo 1667 (44,5) 211 (56,7) Direito 1867 (49,9) 196 (52,7) Desconhecido 950 (25,4) 48 (12,9) Esquerdo 1976 (50,1) 176 (47,3) HER-2/neu Bilateralidade Negativo 1680 (44,9) 208 (55,9) Não 3466 (92,6) 343 (92,2) Positivo 570 (15,2) 45 (12,1) Sim 277 (7,4) 29 (7,8) Desconhecido 1493 (39,9) 119 (32,0) Palpabilidade Multifocalidade Não 1394 (37,2) 120 (32,3) Não 2894 (77,3) 218 (58,6) Sim 1478 (39,5) 159 (42,7) Sim 849 (22,7) 154 (41,4) Desconhecido 871 (23,3) 93 (25,0) Invasão Vásculo-Linfática Localização do Tumor Não 2994 (80,0) 322 (86,6) QSL 2064 (55,1) 247 (66,4) Sim 749 (20,0) 50 (13,4) QIL 486 (13,0) 35 (9,4) QSM 571 (15,3) 41 (11,0) Mestástase em LS QIM 279 (7,5) 20 (5,4) Não 2607 (69,7) 222 (59,7) Central 301 (8,0) 22 (5,9) Sim 1136 (30,3) 150 (40,3) Desconhecido 42 (1,1) 7 (1,9)

42

Tabela 12 – Localização do Tumor— Análise univarida: Ductal versus Lobular

– Amostra de Modelagem – Modelo de Metástase em LS

Variável Ductal (%) N=3143

Lobular (%) N=372

P†

Idade (anos) ≤ 40 11,0 2,4 41-69 71,9 77,4 <0,001 ≥ 70 17,2 20,2 Sexo Fenimino 99,1 99,6 0,557 Masculino 0,9 0,5 Palpabilidade (+) * 51,4 57,0 0,072 Invasão Váculo-Linfática (+) 20,0 13,4 0,002 Multifocalidade (+) 22,7 41,4 <0,001 RE (+) * 77,7 86,1 <0,001 RP (+) * 59,7 65,1 0,063 HER-2/neu (+) * 25,3 17,8 0,009 Estádio T Tis 8,8 ― T1mic 3,4 0,5 T1a 10,2 18,8 T1b 23,6 21,2 <0,001 T1c 38,1 41,9 T2 (≤3cm) 15,3 16,7 T2 (>3cm) 0,6 5,9 T3 0,4 3,0 Localização do Tumor * QSL 55,7 67,7 QIL 13,1 9,6 QSM 15,4 11,2 <0,001 QIM 7,5 5,5 Central 8,1 1,9 Grau Nuclear * I 10,5 ND II 32,8 ND NA III 56,7 ND Grau Histológico * I 9,5 ND II 55,0 ND NA III 35,5 ND

† Teste de χ2 Pearson bicaudal +, positivo LS, linfonodo sentinela QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial RE, receptor de estrógeno; RP, receptor de progesterona

43

Algumas diferenças nas características dos pacientes com carcinoma ductal

(incluindo os subtipos especiais) e carcinoma lobular foram observadas (Tabelas 11 e

12). Foram identificadas diferenças nas seguintes variáveis: idade, representada pela

distribuição por faixa etária ou pela mediana (55 anos em ductal e 59 anos em

lobular); tamanho do tumor, representado pelo estádio T e a mediana (1,1 cm em

ductal e 1,5 cm em lobular); positividade dos receptores hormonais de estrógeno

(77,7% em ductal e 86,1% em lobular) e HER2/neu (25,3% em ductal e 17,8% em

lobular); distribuição da localização do tumor (22,9% mediais em ductal e 16,7%

mediais em lobular); presença de multifocalidade (22,7% em ductal e 41,4% em

lobular) e presença de invasão vásculo-linfática (20% em ductal e 13,4% em

lobular).

Pela comparação gráfica das Figuras 1, 2 e 3 pode-se observar que as

diferenças nas identificações de metástases em LS por IHQ foi de 8,6% versus 3,6%

(P < 0,001) e de 9,1% versus 5,7% (P = 0,008) por cortes seriados entre lobular e

ductal, respectivamente.

Na análise multivariada por regressão logística pelo método de etapas, porém,

a diferença entre carcinoma ductal e lobular desapareceu (Tabela 13), ou seja não

existiu diferença estatística em relação à metástase em LS entre carcinoma ductal e

lobular (P = 0,142). Ou seja, a diferença observada na análise univariada não foi

independente e pode ser justificada pelas diferenças em multifocalidade, localização

ou idade dos pacientes entre esses dois tipos de tumor. Por esse motivo esses tipos

histológicos (ductal e lobular) foram agrupados nas análises subseqüentes.

44

Tabela 13 – Análise multivariada: Regressão logística – Comparação Ductal

versus Lobular (N=4066) – Amostra de Modelagem – Modelo de Metástase em LS

Variável β Razão de

Chance (e β)

95,0 % IC (Inferior–Superior)

P

Tipo histológico do tumor Ductal 0 1 Lobular 0,187 1,206 0,939 – 1,548 0,142 Colóide/medular/tubular -1,707 0,181 0,181 – 0,091 <0,001 Invasão vásculo-linfática Não 0 1 Sim 1,149 3,156 2,641 – 3,777 <0,001 Estádio T Tis+T1mic+T1a 0 1 Tis+T1mic+T1a vs. T1b 0,758 2,134 1,633 – 2,789 <0,001 Tis+T1mic+T1a vs. T1c 1,269 3,558 2,785 – 4,544 <0,001 Tis+T1mic+T1a vs. T2 (≤3cm) 1,879 6,547 4,884 – 8,775 <0,001 Tis+T1mic+T1a vs. T2 (>3cm) 2,321 10,118 6,734 – 15,414 <0,001 Tis+T1mic+T1a vs. T3 3,870 47,927 14,020 – 163,835 <0,001 Localização do tumor Medial 0 1 Outro 0,307 1,360 1,132 – 1,633 0,001 Idade (anos) ≥ 70 0 1 ≥ 70 vs. ≤ 40 0,601 1,824 1,365 – 2,437 <0,001 ≥ 70 vs. 41-69 0,388 1,474 1,199 – 1,810 <0,001 Multifocalidade Não 0 1 Sim 0,608 1,836 1,541 – 2,188 <0,001

Constante -2,870 0,057 <0,001

LS, linfonodo sentinela Medial, quadrantes mediais Outro, outros quadrantes incluindo central IC, intervalo de confiança vs, versus

45

Outras análises mais detalhadas foram realizadas para avaliar as diferenças

entre a localização do tumor na mama (quadrante) em relação à presença de

metástase em LS (Figura 4 e 5 e Tabelas 14, 15).

A Figura 4 mostra de forma gráfica as incidências de metástase em LS por

quadrante, que foram: 24,0% no QSM; 28,1% no QIM; 32,1% no QSL; 36,1% no

QIL e 31,9% no “quadrante” central. A análise univariada mostrou que a incidência

de metástase em LS é diferente entre esses quadrantes (P < 0,001). Para se

identificarem quais quadrantes são responsáveis pela diferença observada realizaram-

se comparações múltiplas entre os quadrantes (Tabela 14).

Quando são feitas comparações múltiplas, deve-se ajustar o nível de

significância, ou seja, corrigir o valor de P. O método utilizado para tal foi o de

Bonferroni, no qual se divide o valor de significância (definido como P < 0,05) pelo

número de subgrupos a serem comparados na análise estatística.

Observando-se os dados da Tabela 14, fica claro que os quadrantes diferentes

eram os súpero-medial (QSM) e o ínfero-medial (QIM), ou seja os quadrantes

mediais. Portanto, as incidências de metástases em LS distribuídas entre quadrantes

mediais e outros foram 25,4% e 32,7%, respectivamente (P < 0,001) (Figura 5).

Na análise univariada, quando se comparam tumores mediais com os de

outros quadrantes, observou-se que as únicas diferenças significativas entre esses

grupos ocorreram nas variáveis sexo e tipo histológico (Tabela 15).

46

Excluídos 49 casos com localização desconhecida P, Teste de χ2 Pearson bicaudal LS, linfonodo sentinela Figura 4 – Freqüência de metástase em LS por localização do tumor

(quadrante) – Amostra de Modelagem – Modelo de Metástase em LS

Medial Lateral

Inferior

Superior

24,0% (147/612)

32,1% (742/2311)

36,1% (188/521)

28,1% (84/299)

31,9% (103/323)

Comparação entre quadrantes P < 0,001

QSM QSL

QIM QIL

Central

47

Excluídos 49 casos com localização desconhecida P, Teste de χ2 Pearson bicaudal LS, linfonodo sentinela Figura 5 – Freqüência de metástase em LS por localização do tumor

(quadrante) – Amostra de Modelagem – Modelo de Metástase em LS

Medial

Inferior

Superior

25,4% (231/911)

Medial versus Outros Quadrantes

P < 0,001

Outros 32,7%

(1033/3155)

Medial Lateral

48

Tabela 14 – Comparações múltiplas entre quadrantes – Amostra de Modelagem

– Modelo de Metástase em LS – Análise univariada

Comparação P (χ2) *

Medial vs. Outro <0,001* Medial vs. QSL <0,001 ** Medial vs. QIL <0,001 ** Medial vs. Central 0,023 QSL vs. QIL 0,088 QSL vs. Central 0,937 QIL vs. Central 0,213 QSM vs. QSL <0,001 *** QSM vs. QIL <0,001 *** QSM vs. Central 0,010 QSM vs. QIM 0,184 QIM vs. Central 0,302 QIM vs. QIL 0,019 QIM vs. QSL 0,160 QSL vs. QIL 0,088 QSL vs. Central 0,937 QIL vs. Central 0,213

* Nivel de significância – estatisticamente significante quando P < 0.05 ** Nível de significância ajustada – estatisticamente significante quando P < 0.008 *** Nível de significância ajustada – estatisticamente significante quando P < 0.005 Medial, QSM+QIM Outro, QSL+QIL+Central QSM, quadrante súpero-medial; QSL, quadrante súpero-lateral; QIL, quadrante ínfero-lateral; QIM, quadrante ínfero-medial vs, versus

49

Tabela 15 – Localização do Tumor – Análise univarida: Medial versus Tumores

de outros quadrantes – Amostra de Modelagem – Modelo de Metástase em LS

Variável Medial (N=911)

Outro (N=3155)

P†

Idade (anos) ≤ 40 9,5 10,4 41-69 73,2 72,0 0,702 ≥ 70 17,2 17,6 Sexo Fenimino 99,9 99,0 0,004 Masculino 0,1 1 Palpabilidade (+) 51,4 52,3 0,684 Invasão Váculo-Linfática (+) 18,4 19,7 0,392 Multifocalidade (+) 22,2 24,5 0,153 RE (+) 79,2 79,4 0,919 RP (+) 60,3 60,4 0,950 HER-2/neu (+) 24,3 24,8 0,775 Estádio T Tis 8,3 7,5 T1mic 3,1 3,2 T1a 10,9 10,2 T1b 25,2 23,1 T1c 38,6 38,6 0,292 T2 (≤3cm) 10,2 12,5 T2 (>3cm) 3,2 4,0 T3 0,4 0,9 Tipo Histológico Lobular 6,7 9,6 Ductal 88,4 86,4 0,012 Colóide/medular/tubular 4,9 3,9 Grau Nuclear I 9,0 9,6 II 54,8 55,1 0,835 III 36,2 35,3 Grau Histológico I 10,9 10,4 II 32,0 33,0 0,848 III 57,1 56,6

† Teste de χ2 Pearson bicaudal +, positivo QSM, quadrante súpero-medial RE, receptor de estrógeno; RP, receptor de progesterona LS, linfonodo sentinela

50

Na análise multivariada por regressão logística pelo método de passos,

confirmou-se a diferença entre tumores mediais e de outros quadrantes (Tabela 16),

ou seja, a localização do tumor na mama foi uma variável independente de metástase

em LS.

As variáveis que mostraram correlação estatística com metástase em LS, pela

análise univariada (Tabelas 9A, 9B e 10) – idade, tamanho do tumor, palpabilidade,

localização do tumor, invasão vásculo-linfática, multifocalidade e grau nuclear –

foram submetidas à análise multivariada por regressão logística pelo método de

passos (Tabela 16). A variável grau nuclear foi reclassificada e as variáveis receptor

de progesterona e grau histológico foram excluídas da análise multivariável.

Conforme exposto em “Casuística e Métodos”, como padrão no

Departamento de Patologia do MSKCC, carcinoma lobular não recebe graduação

histológica ou nuclear. No cálculo de regressão logística, todos os casos em que o

valor da variável está ausente são excluídos da análise. Para se evitar a perda de

informações dos casos de histologia lobular, foi acrescentada na variável grau

nuclear uma nova categoria chamada “lobular”.

A variável receptor de progesterona não foi incluída nessa análise, pois em

24,3% dos casos (998 casos) os dados em relação aos receptores de progesterona

eram desconhecidos.

Optou-se, também, por não incluir na análise multivariada a variável grau

histológico, pois em tumores in situ essa classificação não é realizada e em tumores

microinvasores (T1mic), pode ser prejudicada.

51

Tabela 16 – Análise multivariada: Regressão logística – Amostra de Modelagem

– Modelo de Metástase em LS incluindo Grau Nuclear + Lobular (N=3723)

Variável β Razão de Chance

(e β)


P

Tipo histológico do tumor Colóide/medular/tubular 0 1 Ductal ou Lobular 1,378 3,967 1,771 – 8,887 0,001 Invasão vásculo-linfática Não 0 1 Sim 1,146 3,145 2,619 – 3,778 <0,001 Estádio T Tis+T1mic+T1a 0 1 Tis+T1mic+T1a vs. T1b 0,716 2,046 1,534 – 2,730 <0,001 Tis+T1mic+T1a vs. T1c 1,250 3,489 2,677 – 4,548 <0,001 Tis+T1mic+T1a vs. T2 (≤3cm) 1,832 6,243 4,569 – 8,532 <0,001 Tis+T1mic+T1a vs. T2 (>3cm) 2,298 9,955 6,404 – 15,474 <0,001 Tis+T1mic+T1a vs. T3 3,824 45,779 13,328 – 157,247 <0,001 Localização do tumor Medial 0 1 Outro 0,330 1,390 1,149 – 1,682 0,001 Idade (anos) ≥ 70 0 1 ≥ 70 vs. ≤ 40 0,563 1,756 1,299 – 2,374 <0,001 ≥ 70 vs. 41-69 0,386 1,471 1,190 – 1,819 <0,001 Grau Nuclear I 0 1 II 0,482 1,619 1,100 – 2,384 0,015 III 0,484 1,623 1,088 – 2,420 0,018 Lobular 0,623 1,865 1,197 – 2,906 0,006 Multifocalidade Não 0 1 Sim 0,632 1,882 1,568 – 2,259 <0,001

Constante -4,682 0,009 <0,001

LS, linfonodo sentinela Medial, quadrantes mediais; Outro, outros quadrantes incluindo central IC, intervalo de confiança vs, versus

52

As variáveis que se correlacionaram com a presença de metástase em LS de

forma independente pela regressão logística acima foram: tipo histológico, invasão

vascular linfática, tamanho do tumor, localização do tumor, idade, grau nuclear

(+lobular) e multifocalidade (Tabela 16).

O modelo da Tabela 16 excluiu ainda, por falta de informações, 392 casos

(9,5% do total de casos), sendo 343 casos por falta de grau nuclear e 49 casos por

falta de dados na localização do tumor.

Como o objetivo da presente tese foi desenvolver um modelo de fácil

aplicação e não somente identificar fatores preditivos de metástase em LS, optou-se

por excluir também a variável grau nuclear do modelo final (Tabelas 17 e 18).

O modelo final de regressão logística baseado em 98,8% da amostra de

modelagem é apresentado nas Tabelas 17 e 18. A Tabela 17 apresenta o modelo

utilizando as variáveis tipo histológico, invasão vascular linfática, tamanho do tumor,

localização do tumor, idade e multifocalidade de forma categórica (Modelo

Categórico). A Tabela 18 apresenta o mesmo modelo utilizando todas as variáveis

acima, entretanto tamanho do tumor e idade são apresentadas de forma contínua

(Modelo Contínuo).

A capacidade dos modelos em discriminar corretamente a probabilidade de

metástase em LS foi medida pelo cálculo da área sob a curva de características

operacionais, mais comumente conhecida como curva ROC (do inglês ROC curve –

receiver operating characteristic curve) e pelo gráfico de plotagem das

probabilidades observadas e calculadas (Figuras 6 e 7).

53


– Modelo Categórico (N=4066) – Modelo de Metástase em LS


(e β)


P

Tipo histológico do tumor Colóide/medular/tubular 0 1 Ductal ou Lobular 1,726 5,621 2,829 – 11,169 <0,001 Invasão vásculo-linfática Não 0 1 Sim 1,136 3,115 2,609 – 3,719 <0,001 Estádio T Tis+T1mic+T1a 0 1 Tis+T1mic+T1a vs. T1b 0,772 2,164 1,657 – 2,827 <0,001 Tis+T1mic+T1a vs. T1c 1,287 3,623 2,840 – 4,623 <0,001 Tis+T1mic+T1a vs. T2 (≤3cm) 1,903 6,703 5,009 – 8,971 <0,001 Tis+T1mic+T1a vs. T2 (>3cm) 2,344 10,426 6,899 – 15,756 <0,001 Tis+T1mic+T1a vs. T3 3,934 51,105 14,989 – 174,247 <0,001 Localização do tumor Medial 0 1 Outro 0,312 1,366 1,138 – 1,641 0,001 Idade (anos) ≥ 70 0 1 ≥ 70 vs. ≤ 40 0,582 1,790 1,341 – 2,388 <0,001 ≥ 70 vs. 41-69 0,464 1,468 1,195 – 1,804 <0,001 Multifocalidade Não 0 1 Sim 0,628 1,874 1,576 – 2,228 <0,001

Constante -4,594 0,010 <0,001

LS, linfonodo sentinela Medial, quadrantes mediais Outro, outros quadrantes incluindo central IC, intervalo de confiança vs, versus

54


– Modelo Contínuo (N=4066) – Modelo de Metástase em LS

Variável β Razão de

Chance (e β )


P

Idade (anos) -0, 016 0,985 0,979 – 0,990 <0,001 Tamanho do tumor (cm) 0,669 1,953 1,795 – 2,125 <0,001 Tipo histológico do tumor Ductal ou Lobular 0 1 Colóide/medular/tubular -1,727 0,178 0,089 – 0.354 <0,001 Invasão vásculo-linfática Não 0 1 Sim 1,142 3,134 2,625 – 3,742 <0,001 Multifocalidade Não 0 1 Sim 0,571 1,771 1,494 – 2,098 <0,001 Localização do tumor Outro 0 1 Medial -0,314 0,731 0,608 – 0.878 0,001

Constante -1,144 0,319 <0,001

Medial, quadrantes mediais Outro, outros quadrantes incluindo central IC, intervalo de confiança

55

1 - Especificidade (Taxa de Falso Positivo)

1,00,75,50,250,00

Sens

ibilid

ade

(Tax

a de

Ver

dade

iro P

ositi

vo)

1,00

,75

,50

,25

0,00

Figura 6 – Curva ROC – Modelo Categórico – Amostra de Modelagem –

Modelo de Metástase em LS

Área sob a curva ROC Desvio Padrão P 95,0 % IC

(Inferior-Superior) 0,756 0,008 <0,001 0,740 - 0,772

56


1,00,75,50,250,00

Sens

ibilid

ade

(Tax

a de

Ver

dade

iro P

ositi

vo)

1,00

,75

,50

,25

0,00

Figura 7 – Curva ROC – Modelo Contínuo – Modelo de Metástase em LS

Amostra Área sob a curva ROC

Desvio Padrão P 95,0 % IC

(Inferior-Superior) Modelagem 0,758 0,008 <0,001 Modelagem Validação 0,763 0,012 <0,001 Validação

57

A Figura 6 mostra a curva ROC do modelo categórico aplicado à amostra de

modelagem. A área abaixo da curva foi 0,756, com um intervalo de confiança entre

0,740 e 0,772. A Figura 7 mostra a curva ROC do modelo contínuo aplicado à

amostra de modelagem. A área abaixo da curva foi 0,758, com intervalo de confiança

entre 0,742 e 0,774.

Ambos os modelos categórico e contínuo, por corresponderem à área acima

de 0,75, podem ser considerados adequados.123 O modelo contínuo, pela sua

praticidade, foi aplicado na amostra de validação.

O resultado numérico da equação da regressão logística dada pelo modelo é a

probabilidade de metástase em LS. Para se calcular essa probabilidade é necessário

fornecer ao modelo (equação) os valores das variáveis idade, localização do tumor,

tamanho do tumor, multifocalidade, tipo histológico e invasão vascular linfática. Os

valores a serem aplicados na equação para o cálculo de probabilidade em LS

encontram-se na Tabela 19.

Os valores de cada caso da amostra de validação foram aplicados

individualmente na equação de regressão logística simplificada; uma probabilidade

individual de cada caso foi calculada e tabulada com base nesse modelo contínuo.

Aplicando-se as probabilidades calculadas e observadas na amostra de validação

calculou-se a área abaixo da curva ROC desse modelo aplicado a essa amostra. A

área abaixo da curva ROC foi 0,763, com intervalo de confiança entre 0,739 e 0,786

(Figura 7).

A Figura 8 mostra uma forma mais simples de expressar os resultados e a

acurácia do modelo plotando a probabilidade observada (real) e a probabilidade

58

calculada, tendo como base a curva ideal. Observa-se que o modelo está bem

próximo ao modelo ideal.

Para se facilitar o cálculo de probabilidade de metástase em LS pode-se

utilizar o software que se encontra anexo. Esse software calcula a probabilidade

individual de um paciente apresentar metástase em LS baseado no modelo contínuo.

59

Tabela 19 – Cálculo de probabilidade de metástase em LS – Valores a serem

aplicados na equação do Modelo Contínuo

P, Probabilidade de metástase em linfonodo sentinela (LS)

Variável Valores a serem aplicados na equação

A. Idade Anos T. Tamanho do Tumor cm H. Tipo Histológico do Tumor Colóide/medular/tubular 0 Ductal ou Lobular 1 I. Invasão Vásculo-Linfática Não 0 Sim 1 L. Localização do Tumor Medial 0 Outro 1 M. Multifocalidade Não 0 Sim 1

Medial, quadrantes mediais Outro, outros quadrantes incluindo central

P = 1+ e (1.144 + 0.016xA – 0.669xT + 1,727xH – 1.142xI + 0.314xL – 0.571xM)

1

60

Figura 8 – Probalilidade observada versus probabilidade calculada – Modelo de Metástase em LS

Probabilidade Calculada

1,0,8,6,4,20,0

Prob

abilid

ade

Obs

erva

da

1,0

,8

,6

,4

,2

0,0

Curva IdealGrupo de observações

Probabilidade Calculada

1,0,8,6,4,20,0

Prob

abilid

ade

Obs

erva

da

1,0

,8

,6

,4

,2

0,0

Curva IdealGrupo de observações

61

5.2. Modelo de Metástase em Linfonodo Não-Sentinela

As variáveis analisadas no desenvolvimento deste modelo estão expressas na

Tabela 20. De maneira geral, pode-se observar que não existem grandes diferenças

entre as amostras de modelagem de validação.

Ao se analisarem ambas as amostras (Tabela 20), observa-se que a maioria

dos tumores apresentava entre 1 e 3 cm; a maioria dos pacientes tinha mais de 50

anos; eram carcinomas ductais grau nuclear tipo II ou III; houve um predomínio da

positividade dos receptores hormonais; houve uma menor freqüência de tumores

unifocais e sem invasão vásculo-linfática. Aproximadamente 70% dos casos

apresentavam metástase em somente um LS.

A Tabela 21 mostra a incidência de metástase em Não-LS de acordo com as

variáveis estudadas nas amostras de modelagem, de validação e em ambas.

Identificou-se metástase em Não-LS em 38,9% dos casos em que os LS estavam

comprometidos, ou seja, em 61,1% dos casos os LS eram os únicos linfonodo

comprometidos na axila.

Ao se analisar a Tabela 21, observa-se também que não existiu nenhuma

grande diferença entre essas amostras em relação às variáveis individualmente.

Destaca-se, ao se observarem os percentuais, que existiu uma tendência a um

aumento na incidência de metástase em Não-LS nos pacientes em idade jovem (<50

anos), tumores grandes (por exemplo, 24% em tumores menores em T1a versus 57%

em T2 maiores que 3 cm), histologias ductal grau III e lobular, presença de

multifocalidade, invasão vásculo-linfática e achado de metástase no estudo de

congelação do LS no intra-operatório.

62

Tabela 20 – Características dos pacientes e tumores – Amostras de Modelagem e de Validação – Modelo de Metástase em Não-LS

Número de casos 702 (100,0) 373 (100,0) Método de detecção Somente IHQ 63 (9,0) 18 (4,8) Idade (Anos) Cortes seriados (H&E) 78 (11,1) 40 (10,7) ≤ 50 290 (41,3) 157 (42,1) Rotina 65 (9,3) 23 (6,2) > 50 412 (58,7) 216 (57,9) Congelação 463 (66,0) 273 (73,2) Congelação não realizada 33 (4,7) 19 (5,1) Tamanho do Tumor (T) ≤ T1a 33 (4,7) 13 (3,5) Número de LS Positivos T1b 122 (17,4) 49 (13,1) 1 488 (69,5) 265 (71,0) T1c 312 (44,4) 166 (44,5) 2 161 (22,9) 75 (20,1) T2 ≤ 3,0 cm 154 (21,9) 93 (24,9) 3 35 (5,0) 21 (5,6) T2 > 3,0 cm 65 (9,3) 41 (11,0) 4 12 (1,7) 8 (2,1) T3 16 (2,3) 11 (2,9) 5 3 (0,4) 3 (0,8) 6 1 (0,1) 0 Tipo de Tumor/Grau Nuclear 7 2 (0,3) 0 Ductal, I 22 (3,1) 11 (2,9) >8 0 1 (0,3) Ductal, II 321 (45,7) 175 (46,9) Ductal, III 275 (39,2) 129 (34,6) Número de LS Negativos Lobular 84 (12,0) 58 (15,5) 0 271 (38,6) 132 (35,4) 1 183 (26,1) 79 (21,2) Invasão Vásculo-Linfática 2 102 (14,5) 72 (19,3) Não 418 (59,5) 219 (58,7) 3 68 (9,7) 41 (11,0) Sim 284 (40,5) 154 (41,3) 4 34 (4,8) 22 (5,9) 5 16 (2,3) 7 (1,9) Multifocalidade 6 6 (0,9) 10 (2,7) Não 505 (71,9) 214 (64,6) 7 8 (1,1) 2 (0,5) Sim 197 (28,1) 132 (35,4) >8 14 (2,0) 8 (2,1) Receptor de Estrógeno Negativo 135 (19,2) 83 (22,3) Positivo 567 (80,8) 290 (77,7)

LS, Linfonodo sentinela IHQ, imunoistoquímica H&E, hemetoxilina e eosina

Modelagem N (%)

Validação N(%)

Modelagem N (%)

Validação N(%)

63

Tabela 21 – Incidência de metástases em Não-LS – Amostras de Modelagem e de Validação – Modelo de Metástase em Não-LS


(N) Não- LS + (N)

Não-LS + (%)

Número

(N) Não-LS + (N)

Não-LS + (%)

Número

(N) Não-LS + (N)

Não-LS + (%)

Número de procedimentos 702 264 37,6 373 154 41,3 1075 418 38,9 Idade (Anos) ≤ 50 290 114 39,3 157 64 40,8 447 178 39,8 > 50 412 150 36,4 216 90 41,7 628 240 38,2 Tamanho do Tumor (T) ≤ T1a 33 8 24 13 1 1/13 46 9 20 T1b 122 32 26,2 49 13 27 171 45 26,3 T1c 312 111 35,6 166 60 36,1 478 171 35,8 T2 ≤ 3,0 cm 154 62 40,3 93 44 47,3 247 106 42,9 T2 > 3,0 cm 65 37 57 41 26 63 106 63 59,4 T3 16 14 14/16 11 10 10/11 27 24 89 Tipo de Tumor e Grau Nuclear Ductal, I 22 6 27 11 3 3/11 33 9 27 Ductal, II 321 100 31,2 175 57 32,6 496 157 31,7 Ductal, III 275 121 44,0 129 67 51,9 404 188 46,5 Lobular 84 37 44 58 27 47 142 64 45,1 Invasão Vásculo-Linfática Não 418 125 29,9 219 71 32,4 637 196 30,8 Sim 284 139 48,9 154 83 53,9 438 222 50,7 Multifocalidade Não 505 176 34,9 214 89 41,6 746 265 35,5 Sim 197 88 44,7 132 65 49,2 329 153 46,5 Receptor de Estrógeno Negativo 135 50 37,0 83 37 45 218 87 39,9 Positivo 567 214 37,7 290 117 40,3 857 331 38,6 Método de detecção Somente IHQ 63 6 10 18 4 4/18 81 10 12 Cortes seriados (H&E) 78 12 15 40 3 8 118 15 12,7 Rotina 65 13 20 23 6 26 88 19 21 Congelação 463 221 47,7 273 128 46,9 736 349 47,4 Congelação não realizada 33 12 36 19 13 13/19 52 25 48 Número de LS Positivos 1 488 155 31,8 265 97 36,6 753 252 33,5 2 161 72 44,7 75 34 45 236 106 44,9 3 35 23 66 21 15 71 56 38 67,9 4 12 9 9/12 8 5 5/8 20 14 70 5 3 2 2/3 3 2 2/3 6 4 4/6 6 1 1 1/1 0 0 - 1 1 1/1 7 2 2 1/2 0 0 - 2 2 2/2 >8 0 0 - 1 1 1/1 1 1 1/1 Número de LS Negativos 0 271 143 52,8 132 83 62,9 403 217 53,8 1 183 67 36,6 79 25 32 262 92 35,1 2 102 27 26,5 72 18 25 174 45 25,9 3 68 18 27 41 14 34 109 32 29,4 4 34 11 32 22 10 46 56 21 38 5 16 3 3/16 7 0 0/7 23 3 13 6 6 1 1/6 10 3 3/10 16 4 4/16 7 8 2 2/8 2 1 1/2 10 3 3/10 >8 14 1 1/14 8 0 0/8 22 1 5

Não-LS +, Linfonodo não-sentinela positivo; IHQ, imunoistoquímica; H&E, hemetoxilina e eosina LS, linfonodo sentinela

64

Dois modelos de análises multivariadas por regressão logística foram

desenvolvidos (Tabelas 22 e 23). O que diferencia os modelos é o método de

detecção de metástase em LS. No modelo da Tabela 22, o estudo de congelação foi

separado do método de rotina por H&E (Modelo com congelação); já no modelo da

Tabela 23, os casos de identificação de metástase por “congelação” foram

reclassificados como “rotina” (Modelo sem congelação). Optou-se por criar esses

dois tipos de modelos, uma vez que o estudo de congelação não está disponível ou

não é realizado em todos os centros.

Nas análises multivariadas por regressão logística (Tabelas 22 e 23), o

tamanho do tumor, a invasão vásculo-linfática, o método de detecção da metástase, o

número de LS positivos e o número de LS negativos mostraram-se estar

independentemente associados à metástase em Não-LS (todos com P < 0,05). A

multifocalidade mostrou correlação com a positividade em Não-LS somente no

modelo sem congelação da Tabela 23 (P = 0,02), embora essa correlação tenha sido

marginal no modelo com congelação da Tabela 22 (P = 0,069). O receptor de

estrógeno e o tipo do tumor/grau nuclear não mostraram correlação com metástase

em Não-LS (Tabelas 22 e 23).

A variável idade mostrou correlação estatística com positividade em Não-LS;

entretanto, intencionalmente não foi incluída nos modelos finais, pois seu efeito,

além de insignificante no resultado final, dificultava a expressão gráfica dos

nomogramas (Figuras 9 e 10).

65

Tabela 22 – Análise multivariada: Regressão logística – Amostra de Modelagem – Modelo Com Congelação (N=669*) – Modelo de Metástase em Não-LS


(e β)

95,0 % IC (Inferior–Superior) P

Tamanho do Tumor 0,293 1,341 1,130 – 1,591 0,001 Tipo de Tumor e Grau Nuclear Ductal, grau nuclear I vs. II 0,004 1,004 0,333 – 3,022 0.994 Ductal, grau nuclear I vs. III 0,233 1,263 0,411 – 3,879 0,683 Ductal, grau nuclear I vs. Lobular 0,096 1,101 0,330 – 3,672 0,875 Invasão Vásculo-Linfática Não 0 1 Sim 0,537 1,711 0,189 – 2,461 0,003 Multifocalidade Não 0 1 Sim 0,378 1,459 0,970 – 2,914 0,069 Receptor de Estrógeno Negativo 0 1 Positivo 0,427 1,533 0,952 – 2,468 0,078 Método de detecção Congelação vs. IHQ -2,040 0,130 0,053 – 0,319 <0,001 Congelação vs. Cortes seriados (H&E) -1,491 0,225 0,114 – 0,443 <0,001 Congelação vs. Rotina -1,128 0,324 0,166 – 0,630 <0,001 Número de LS Positivos 0,407 1,502 1,192 – 1,893 <0,001 Número de LS Negativos -0,233 0,792 0,708 – 0,887 <0,001

Constante -1,767 0,171 0,006

* Excluídos os 33 casos em que não foi feito estudo de congelação. LS, Linfonodo sentinela Não-LS, linfonodo não-sentinela IHQ, imunoistoquímica H&E, hemetoxilina e eosina IC, intervalo de confiança vs, versus

66


– Modelo Sem Congelação (N=702*) – Modelo de Metástase em Não-LS

* Os estudos de congelação foram classificados como “Rotina” LS, Linfonodo sentinela Não-LS, linfonodo não-sentinela IHQ, imunoistoquímica H&E, hemetoxilina e eosina IC, intervalo de confiança vs, versus


(e β)

95,0 % IC (Inferior–Superior) P

Tamanho do Tumor 0,290 1,337 1,139 – 1,568 <0,001 Tipo de Tumor e Grau Nuclear Ductal, grau nuclear I vs. II 0,187 1,205 0,421 – 3,454 0,8 Ductal, grau nuclear I vs. III 0,504 1,655 0,567 – 4,829 0,4 Ductal, grau nuclear I vs. Lobular 0,345 1,411 0,446 – 4,467 0,6 Invasão Vásculo-Linfática Não 0 1 Sim 0,522 1,685 1,187 – 2,391 0,003 Multifocalidade Não 0 1 Sim 0,453 1,573 1,063 – 2,327 0,02 Receptor de Estrógeno Negativo 0 1 Positivo 0,331 1,393 0,883 – 2,198 0,16 Método de detecção* Rotina vs. IHQ -1,647 0,193 0,087 – 0,427 <0,001 Rotina vs. Cortes seriados (H&E) -1,431 0,239 0,122 – 0,468 <0,001 Número de LS Positivos 0,455 1,577 1,256 – 1,979 <0,001 Número de LS Negativos -0,226 0,798 0,715 – 0,890 <0,001

Constante -2,117 0,120 <0001

67

Figura 9 – Nomograma com congelação, N=669 – Modelo de Metástase em Não-

LS

68

Figura 10 – Nomograma sem congelação, N=702 – Modelo de Metástase em

Não-LS

69

Dois nomogramas foram desenvolvidos (Figuras 9 e 10) baseados nesses

modelos de análises multivariadas. Ambos incorporaram as oito variáveis de cada

modelo. O nomograma da Figura 9 diferenciou em método de detecção a

“congelação” de “rotina”; já o nomograma da Figura 10 reclassificou “congelação”

como “rotina”, pelos motivos anteriormente descritos.

A Figura 11 mostra a curva ROC do modelo com congelação aplicado à

amostra de modelagem. Nessa amostra a área abaixo da curva foi 0,758, com um

intervalo de confiança entre 0,732 e 0,784. Na amostra de validação, a área abaixo da

curva foi 0,771, com um intervalo de confiança entre 0,735 e 0,808.

A Figura 12 mostra a curva ROC do modelo sem congelação aplicado à

amostra de modelagem. Nessa amostra a área abaixo da curva foi 0,751, com

intervalo de confiança entre 0,714 e 0,788. Na amostra de validação, a área abaixo da

curva foi 0,778, com um intervalo de confiança entre 0,742 e 0,814.

As Figuras 13 e 14 também expressam os resultados de acurácia dos modelos

acima plotando a probabilidade observada (real) e a probabilidade calculada, tendo

como base a curva ideal. Observa-se que os dois modelos estão também bem

próximos aos modelos ideais.

5.2.1 Utilizando-se os nomogramas

Cada versão do nomograma consiste em 11 linhas. A primeira linha (Pontos)

é a pontuação dada para cada variável. As linhas 2 a 9 representam as variáveis

incluídas nos modelos. Para cada paciente, tendo por base suas características

clínico-histopatológicas individuais, é assinalado um valor para cada variável na

70

escala de pontos da primeira linha. A pontuação de cada variável na escala de pontos

é determinada por uma linha vertical feita entre o valor apropriado da variável e a

linha de pontos. Por exemplo, tamanho tumoral de 1 centímetro corresponde a 10

pontos (Figura 9).

Em seguida, realiza-se a somatória dos pontos das oito variáveis. O valor da

somatória é identificado na linha “Total de pontos”. Uma outra linha vertical é

traçada entre o total de pontos e a linha “probabilidade de Não-LS Positivo”. O

número correspondente nessa última linha é a probabilidade de metástase em Não-LS

positivo.

Em adição ao nomograma gráfico, para se facilitar o cálculo de probabilidade

de metástase em Não-LS, pode-se utilizar o software que se encontra anexo.

71


1,00,75,50,250,00

Sen

sibi

lidad

e (T

axa

de V

erda

deiro

Pos

itivo

)

1,00

,75

,50

,25

0,00

Figura 11 – Curva ROC – Nomograma com congelação, N=669 – Modelo de

Metástase em Não-LS




72


1,00,75,50,250,00

Sen

sibi

lidad

e (T

axa

de V

erda

deiro

Pos

itivo

)

1,00

,75

,50

,25

0,00

Figura 12 – Curva ROC – Nomograma (sem congelação), N=702 – Modelo de

Metástase em Não-LS – Nomograma sem congelação




73

Figura 13 – Probabilidade observada versus probabilidade calculada – Modelo

de Metástase em Não-LS – Nomograma com congelação

74

Figura 14 – Probabilidade observada versus probabilidade calculada – Modelo de Metástase em Não-LS – Nomograma sem congelação

75

6. DISCUSSÃO

6.1 Metástase em linfonodo sentinela

A freqüência de metástase em LS axilar é maior do que a freqüência de

metástases axilares observadas na era pré-LS.47,52,53,55,60,63,76 Esse fato pode ser

explicado pela maneira como o LS é analisado histopatologicamente, com cortes

seriados e uso de IHQ para células epiteliais.

Outra explicação para os achados de aumento da incidência de positividade

no LS, conforme sugerido por Hansen et al.124 e Moore et al.125, pode decorrer da

manipulação prévia do tumor por meio de punções ou biópsias, atualmente mais

difundidas. Esses estudos sugerem que essas manipulações prévias podem aumentar

a positividade do LS, embora o significado clínico desses achados não seja

claro.124,125

A presença de células epiteliais em LS na ausência de neoplasia maligna é

muito rara, conforme demostrado por King et al.126 Nesse estudo, analisando biópsias

de LS em pacientes submetidas à mastectomia profilática, os autores demonstraram

que a manipulação cirúrgica de mama normal (sem neoplasia maligna) através de

biópsia prévia não aumenta os achados de células epiteliais nos LS identificadas por

IHQ.126

A análise das distribuições do método de identificação das metástases em LS

observada nas Figuras 1, 2 e 3 comprova que o estudo anatomopatológico mais

detalhado, com IHQ e cortes seriados, aumenta a taxa de identificação de metástases

em LS. Isso foi ainda mais importante para os carcinomas lobulares.

76

Antes do advento do LS, Cote et al.127 mostraram diferenças entre as taxas de

identificação de metástases ocultas em linfonodos axilares em pacientes

supostamente com axilas negativas quando comparados com carcinomas ductais e

lobulares. Nos carcinomas ductais invasivos, 8% das metástases foram detectadas

por cortes seriados e 19%, por IHQ. Nos carcinomas lobulares invasivos, 4% foram

identificadas por cortes seriados e 39% por IHQ. De Mascarel et al.128 relataram 10%

de metástases (13/129) identificadas por IHQ em carcinoma ductal invasivo e 41%

(37/89) em carcinoma lobular invasivo. McGuckin et al.129 relataram 25% de

metástases (41/163) identificadas por IHQ em carcinoma ductal invasivo e 38%

(11/29) em carcinoma lobular invasivo.

A identificação de células metastáticas nos linfonodos em algumas

circunstâncias pode ser difícil, especialmente em casos de carcinoma lobular

invasivo, por causa das características das células lobulares e de suas semelhanças

com os histiócitos e vice-versa. Rosen122 observou que, quando o envolvimento do

linfonodo é esparso, ou quando há extenso envolvimento do linfonodo limitado aos

sinusóides, pode ser difícil a distinção entre carcinoma lobular e histiócitos. Rosen122

também descreveu que o carcinoma lobular metastático se assemelha a histiócitos

reacionais dos sinusóides do linfonodo. Rosai130 comentou ainda sobre essa

dificuldade, observando que as células de carcinoma lobular freqüentemente

apresentam núcleos vesiculares com citoplasma amplo que se assemelham às células

histiocíticas reacionais e mimetizam seios histiocitários.

Vários estudos131-136 demostram um aumento de 10% a 52% na detecção de

metástases em LS com o uso de cortes seriados e IHQ. Embora todos esses estudos

utilizem anticorpos anticitoqueratina, os anticorpos utilizados variam entre eles, o

77

que dificulta a comparação dos resultados entre si. Ao que parece, os anticorpos

anticitoqueratina CAM 5.2 e AE1:AE2 são os mais comumente utilizados hoje, e

foram esses os utilizados no presente estudo.

Com a biópsia de LS, o patologista necessita analisar somente um ou alguns

poucos linfonodos, o que viabiliza a utilização de um estudo mais detalhado em cada

linfonodo. Com esse tipo de análise mais detalhada do linfonodo é possível

identificar metástases ocultas que não são identificadas na análise usual de todos os

linfonodos axilares, na qual se realiza somente um corte histológico corado por H&E

por linfonodo.

Numa extensa revisão da literatura, Liberman137 demonstrou que a análise

mais detalhada do LS, principalmente com IHQ, diminui significativamente as taxas

de falso-negativos da biópsia de LS como procedimento de estadimento axilar (8%

sem IHQ versus 3% com IHQ). 137

O achado de micrometástases e células neoplásicas isoladas pode ocorrer com

a análise anatomopatológica detalhada do LS, porém esse tipo de análise tem como

principal valor aumentar a eficácia da biópsia de LS; daí sua importância.

O impacto clínico das micrometástases ou metástases ocultas tem sido alvo

de vários estudos, revisões e meta-análises.128,138-143

Dowlatshahi et al.138 e Liberman140 revisaram a literatura sobre a detecção de

metástases ocultas em linfonodos e sua importância clínica na era pré-LS. Nessas

duas revisões fica explícito que a análise histopatológica mais detalhada, com cortes

seriados e IHQ, é capaz de identificar mais metástases em linfonodos. Embora a

relevância clínica desse tipo de metástase ainda seja controversa, esses autores

concluem que, em séries grandes com mais de 100 pacientes analisados, a

78

identificação de metástases ocultas atinge até 20% e se correlaciona com pior

prognóstico, pior sobrevida global e livre de doença.138,140

Em um outro estudo realizado no MSKCC,143 analisamos 373 casos de

pacientes com câncer de mama e linfonodos axilares negativos submetidos à

mastectomia e esvaziamento axilar entre 1976 e 1978. Nenhum desses pacientes

recebeu tratamento adjuvante. O tempo mediano de acompanhamento foi de 17,6

anos. Uma análise detalhada de todos os linfonodos com cortes seriados e IHQ foi

feita num total de 2.470 blocos de parafina (aproximadamente 8.000 linfonodos e

20.000 lâminas). Foram analisados dois níveis em cada linfonodo, sendo os cortes

seriados realizados a cada 50 µm. Por último, esses cortes foram corados por H&E e

IHQ para receptores de células epiteliais (AE1:AE3, Ventana Medical Systems, Inc.,

Tucson, AZ). Identificamos 23% (84/373) de linfonodos positivos por H&E ou IHQ;

33% (28/84) eram positivos somente por IHQ, enquanto 67% (56/84) eram positivos

nos cortes seriados corados por H&E. Em 71% dos casos somente um linfonodo

axilar era positivo. As estimativas de sobrevida livre de doença (SLD) e de morte por

câncer de mama em 20 anos foram, respectivamente, 79% e 20% nos casos em que

IHQ e H&E eram negativas (N = 289), 66% e 37% em que IHQ era positiva e H&E

negativa (N = 28) e 49% e 50% em que a IHQ e o H&E eram positivas (N = 56). A

sobrevida livre de doença e a morte por câncer de mama foram associadas à

positividade de linfonodos positivos (P < 0,01). Entre os pacientes com IHQ e H&E

negativas, a SLD nos pacientes com tumores menores ou iguais a 2 cm foi

significativamente melhor do que nos pacientes com tumores maiores que 2 cm, e

equivalente à sobrevida de pacientes com tumores menores ou iguais a 1 cm das

séries históricas. Concluímos nesse estudo que a presença de metástases ocultas ou

79

micrometástases detectadas por IHQ está associada a uma pior evolução associada ao

câncer de mama. A análise por IHQ dos linfonodos em pacientes com axila negativa

pode identificar um subgrupo de risco aumentado para recidiva e morte por câncer de

mama. Recentemente, um nomograma foi desenvolvido com esses dados para

predizer a mortalidade em pacientes tratados sem quimioterapia adjuvante.139

Millis et al.142 analisaram o significado de metástases axilares detectadas por

IHQ em 477 pacientes com linfonodos previamente negativos por H&E. Sessenta

pacientes (12,6%) foram identificados com metástases diagnosticadas

exclusivamente por IHQ. Após um acompanhamento mediano de 18,9 anos não

houve diferenças estatísticas no grupo com ou sem metástase oculta diagnosticada

por IHQ.

De Mascarel et al.128 avaliaram 218 pacientes com linfonodos previamente

negativos por H&E. O estudo identificou 50 pacientes com metástases vistas

exclusivamente por IHQ. Apesar do longo acompanhamento (18 a 24 anos), não

houve diferenças de sobrevida entre os grupos com e sem metástases por IHQ.

Maibenco et al.141 utilizaram o banco de dados da SEER (Surveillance,

Epidemiology, and End Results) e avaliaram o valor prognóstico dos pacientes com

micrometástases (pN1mi). Eles identificaram um total de 42.197 pacientes

classificados como pN0, e 1.724 como pN1mi. Comparando-se o grupo de pacientes

pN0 com os pN1mi com uma micrometástase e pN1mi com mais de uma

micrometástase, independentemente do grau do tumor, a sobrevida em 12 anos

respectivamente foi de 92,8%, 87,8% e 89,2% (P < 0,001). O maior impacto na

sobrevida dos pacientes com micrometástases ocorreu em pacientes com tumores

grau 3 (pN0 89,8%, pN1mi com uma micrometástase 76,1% e pN1 com mais de uma

80

micrometástase 80,7%). A presença de micrometástase não teve efeito prognóstico

em pacientes com tumores menores de 1 cm (96,4 versus 96,2%).

Três grandes estudos prospectivos multicêntricos estão em andamento e

definirão em breve o significado das micrometástases em câncer de mama. São eles:

o National Surgical Adjuvant Breast and Bowel Project, estudo B32 (NSABP-B32),

o American College of Surgeons Oncology Group, estudo Z0010 (ACOSOG-Z0010)

e o International Breast Cancer Study Group, estudo 23-01 (IBCSG 23-1).

O presente estudo demonstrou que idade, palpabilidade, localização do

tumor na mama, invasão vásculo-linfática, multifocalidade, positividade do receptor

de progesterona, tamanho do tumor, subtipo histológico, grau histológico e grau

nuclear são fatores preditivos de metástase em LS na análise univariada.

Na análise multivariada identificamos que os fatores preditivos independentes

de metástase em LS em câncer de mama foram: tipo histológico, invasão vascular

linfática, tamanho do tumor, localização do tumor, idade, grau nuclear (+lobular) e

multifocalidade (Tabela 16). Forçou-se a inclusão da histologia lobular no grau

nuclear para que este não fosse automaticamente da regressão logística.

Outros estudos na era pré-LS também demonstraram que idade,44,45,47-49,52-

55,60,62,63,76,78 localização do tumor na mama,48,49,52,53,59,60,77,79 invasão vásculo-

linfática,43,44,46-48,53,56,57,64,76,80 multifocalidade,46,47 tamanho do tumor,42-50,52-

61,63,64,76,78,81-84 subtipo histológico46,49,52,58-60,63,78 e grau nuclear43,44,56,59,63,76,82 são

fatores preditivos de metástase axilar na análise univariada ou multivariada.

Uma lista de variáveis independentes (análise multivariada) associadas à

metástase axilar é apresentada na Tabela 1. Esses fatores associados à metástase

81

axilar não mudaram drasticamente com o advento do LS, o que reforça a confiança

no procedimento de biópsia do LS em câncer de mama.

O tamanho do tumor, a invasão vásculo-linfática, a idade e o subtipo

histológico são os fatores independentes mais comumente descritos associados à

metástase axilar entre esses estudos; portanto, a identificação dessas variáveis em

nossos resultados não surpreende.

Somente uma minoria dos estudos publicados identifica grau nuclear e grau

histológico (ou grau tumoral) como fatores preditivos de metástase em axilar (Tabela

1). A graduação nuclear é a avaliação citológica do núcleo do tumor em comparação

ao núcleo da célula epitelial mamária normal, enquanto grau histológico ou grau

tumoral descrevem os padrões de crescimento microscópico do carcinoma ductal

invasivo assim como os achados citológicos de diferenciação.122 Essas variáveis

podem estar ausentes nos resultados de outros estudos porque muitos patologistas e

instituições, incluindo o MSKCC, não classificam carcinoma lobular invasivo em

grau nuclear ou histológico.

Nossos dados mostraram que a multifocalidade foi um fator preditivo

independente de metástase em LS. Somente outros dois estudos mostram

multifocalidade como fator preditivo de metástase axilar.46,47 Uma outra explicação

para esse achado é que os carcinomas multifocais geralmente têm um volume

tumoral maior que os tumores similares estadiados como unifocais, e tamanho

tumoral é um fator preditivo de metástase axilar. Portanto, talvez a multifocalidade

ajude a melhor determinar o volume tumoral em tumores multifocais.

82

Em outro estudo desenvolvido no MSKCC, demonstramos que

multifocalidade é também um fator independente de pior sobrevida livre de

doença.144

Lohrish et al.,79 Asaga et al.145 e Zucali et al.146 demonstraram um pior

prognóstico, em termos de sobrevida livre de doença e sobrevida global, associado a

tumores de quadrantes mediais comparado a quadrantes laterais. Israel et al.147

relataram diferenças nos padrões de metástase à distância em tumores mediais.

Todos esses estudos79,145-147 explicam que essas diferenças podem ser devidas à

maior probabilidade de metástase para os linfonodos da cadeia linfonodal de artéria

mamária interna associada aos tumores mediais, o que é comprovado por Bevilacqua

et al.,148 em uma extensa revisão da literatura referente à abordagem da metástase em

cadeia mamária interna em câncer de mama.

Os tumores em quadrantes mediais têm pior sobrevida global e livre de

doença comparados aos do quadrantes laterais.79,145,146 Outros estudos relatam

diferenças nos padrões de metástases a distâncias associadas à localização medial dos

tumores.147 Os autores desses estudos atribuem seus resultados à maior probabilidade

de metástases ocultas nos linfonodos da cadeia mamária interna que ocorrem em

tumores mediais.

Conforme já mostrado em nossos resultados (Tabela 9B, Figuras 4 e 5), a

incidência de metástase axilar difere entre os tumores mediais e de outros quadrantes,

respectivamente 24,1% e 32,7% (P < 0,001). Para termos certeza de que esses

achados não eram acidentais, foi realizada uma análise mais aprofundada (Tabelas

14, 15 e 18). Não houve diferenças significantes entre as diversas variáveis

comparando-se os tumores mediais com aqueles de outros quadrantes (Tabela 15).

83

Houve uma insignificante diferença em relação ao número de casos do sexo

masculino, 0,1% em mediais e 1% em outros quadrantes. Houve também menos

casos de histologia lobular nos quadrantes mediais em comparação com os outros

quadrantes, 6,7% e 9,6%. Essas duas diferenças identificadas não justificam as

diferenças na incidência de metástase associada à localização. Foram realizadas

também comparações univariadas múltiplas entre os quadrantes (Tabela 14). Através

da análise Tabela 14, fica nítido que os quadrantes mediais são os responsáveis pela

diferença encontrada. A localização do tumor comprovou ser também um fator

independente de metástase em LS na análise multivariada (Tabela 18).

Outros estudos já publicados suportam nossos achados.48,49,52,53,59,60,77,147

Quatro desses estudos48,49,52,60 citam a localização do tumor como fator preditivo

independente para metástase axilar e relatam uma maior freqüência de metástase

axilar associada aos tumores laterais.

Gann et al.49 analisaram 18.000 pacientes do banco de dados do Colégio

Americano de Cirurgiões (Patient Care Evaluation Survey Database of the American

College of Surgeons) e mostraram que a taxa de metástase em linfonodos axilares de

tumores em quadrantes súpero-mediais é 0,54 vezes a taxa de metástase para os

tumores súpero-laterais. Em estudo semelhante, utilizando-se do banco de dados do

SEER (Surveillance, Epidemiology, and End Results Database), Maibenco et al.52

analisaram 13.000 pacientes com tumores menores ou iguais a 1 cm e demonstraram

que a metástase axilar em tumores mediais é 0,76 vezes menor que os tumores de

outros quadrantes. Esse número é muito próximo ao que encontramos em nossos

dados (Tabela 9B), que foi de 0,78 (25,4% / 32,7%).

84

A menor freqüência de metástase em linfonodos axilares, incluindo os

linfonodos sentinelas, observada em tumores mediais em nosso estudo e identificada

na literatura, reforça a hipótese de rotas alternativas do fluxo linfático nessa

localização.148

Outro fato curioso observado em nossos resultados foi a aparente diferença

nas taxas de metástase em linfonodos sentinelas entre os carcinomas ductais e

lobulares, respectivamente 31,5% e 40,3% (P < 0,0001), observada na Tabela 10.

Frente a esses achados uma análise mais detalhada foi realizada. Uma possível

explicação seria diferenças associadas a outras variáveis, como por exemplo,

multifocalidade, sabidamente muito mais freqüente em tumores de histologia

lobular.122 Conforme já descrito em nossos resultados, foram identificadas diferenças

nas seguintes variáveis: idade, representada pela mediana e pela distribuição por

faixa etária; tamanho do tumor, representado pelo estádio T e a mediana;

positividade dos receptores hormonais de estrógeno e HER2/neu; distribuição da

localização do tumor; multifocalidade e invasão vásculo-linfática. Entretanto, na

análise multivariada, a diferença entre carcinoma ductal e lobular desaparece (Tabela

13), ou seja, não há diferença estatística em relação à metástase em LS entre

carcinoma ductal e lobular (P = 0,142). A diferença observada na análise univariada

não é independente e pode ser justificada pelas diferenças em outras variáveis. Dessa

forma, nos modelos de regressão logística subseqüentes (Tabelas 16, 17 e 18) os

tumores de histologia ductal e lobular foram agrupados em um único grupo e

comparados aos tumores de histologia especial, como colóide, tubular e medular

típico.

85

Os modelos finais de regressão logística, categóricos e contínuos (Tabelas 17

e 18), mostraram-se adequados para predição de metástase em LS conforme as

análises da curva ROC e a área sob essas curvas (Figuras 6 e 7). Entretanto, pela

praticidade, o modelo contínuo da Tabela 18, expresso na forma de equação na

Tabela 19, é mais adequado para o uso, principalmente para o cálculo de

probabilidades.

Uma crítica a esses modelos é que a linfadenectomia axilar não foi realizada

em todos os casos e, portanto, existe alguma contaminação de casos de falso-

negativo em LS. Porém, de certa forma, esse mesmo aspecto pode ser considerado

algo a favor dos modelos, já que na prática clínica, geralmente, não se realiza

linfadenectomia nos casos de LS negativo e certamente existem falso-negativos

nessa amostra. Isso torna o modelo mais próximo da realidade.

Em 2002, descrevemos um modelo de metástase axilar validado

prospectivamente.149 Naquele estudo, incluímos os primeiros 2.000 procedimentos de

biópsia de LS do MSKCC, independentemente dos casos de falso-negativo. As

restrições e críticas àquele modelo são o fato de termos incluído os casos de falso-

negativo de LS e excluído casos de carcinoma lobular.149 Por esses motivos optamos

por desenvolver novos modelos, especificamente para prever metástase em LS e

Não-LS, que são os objetos da presente tese.

Outra crítica aos modelos é a possível limitação da aplicação destes em

relação aos achados pré-operatórios, clínicos ou patológicos. Por exemplo: para a

correta aplicação dos modelos é necessário que seja realizada biópsia do tumor

primário, e a invasão vásculo-linfática pode não ser identificada na biópsia

percutânea. Outra possível limitação se refere à diferença entre o tamanho

86

radiológico e patológico. Vários estudos, porém, demonstram que a avaliação

radiológica pré-operatória do tumor através de mamografia, ultra-sonografia e/ou

ressonância magnética tem excelente correlação com o tamanho patológico.150-152

Com base na fórmula matemática da Tabela 19 foi desenvolvido um software

que facilita, em muito, o cálculo de probabilidade de LS para cada paciente.

Esse modelo contínuo (Tabelas 18 e 19) foi aplicado numa amostra

subseqüente submetida à biópsia de LS e se mostrou também bastante adequado em

prever as metástases (Figuras 7 e 8).

O estudo de Gann et al.49 é o único a validar os seus resultados em amostras

distintas. Entretanto, esse estudo somente descreve os fatores associados de forma

independente à metástase axilar e não é um estudo com base em linfonodos sentinela.

A maneira como Gann et al.49 expressam os resultados impede que se possa aplicá-

los de forma prática a calcular individualmente a probabilidade de metástase axilar.

Dessa forma, o presente modelo estatístico e a maneira como este é

apresentado é peculiar e inovador, não existindo similar algum na literatura médica

até hoje.

6.2 Metástase em linfonodo não-sentinela

O tratamento padrão em pacientes com metástase em LS tem sido a

linfadenectomia axilar complementar. Em nossa casuística 38% dos casos em que o

LS era positivo apresentavam outros linfonodos axilares (Não-LS) também

comprometidos, comparados com taxas que chegam a 50% na literatura.51,67-69,73,87-92

Isso pode ser explicado pelo número de pacientes nesses estudos, muito inferior ao

nosso.

87

Identificamos tamanho tumoral, invasão vásculo-linfática, multifocalidade,

métotodo de detecção de metástase em LS, número de LS positivos e número de LS

negativos como fatores preditivos de metástase em Não-LS (Tabelas 22 e 23). Esses

achados não diferem substancialmente dos encontrados na literatura (Tabela 2), à

exceção de multifocalidade (P = 0,006) e do número de LS negativos (<0,001) que

não haviam sido descritos anteriormente.

Demonstramos que multifocalidade é um fator preditivo de metástase em LS.

O achado de que esse fator é também um fator independente de metástase em Não-

LS não nos parece surpreendente pelas mesmas justificativas descritas anteriormente.

A associação inversa e independente de metástase em Não-LS e o número de

LS negativos é algo inédito, embora, seja um tanto óbvia. O achado de metástase em

um único LS frente a outros LS ressecados nos induz a pensar que o “volume” ou a

“quantidade” de metástase não é suficiente para atingir a todos os LS e, portanto,

muito menos provável que outros linfonodos axilares possam estar comprometidos.

A associação direta e independente de metástase em Não-LS e o número de

LS positivos também é um tanto lógica e óbvia. Ao identificarmos uma situação em

que múltiplos LS estão comprometidos, nos parece muito provável que outros

linfonodos axilares também possam estar comprometidos. Esse achado também é

descrito por outros autores.89,99,101

O tamanho da metástase em LS está associado à presença metástase em Não-

LS.85-87,89,90,93,95,97,98,103-105 Para quantificar o volume de metástase no LS utilizamos

de forma indireta o método de detecção da metástase, uma vez que alguns

patologistas – incluindo os do MSKCC – referem dificuldade em medir o tamanho

das metástases quando estas se apresentam de forma difusa ou em grupamento de

88

células metastáticas nos LS. Isso poderia influenciar de forma subjetiva a

mensuração do tamanho da metástase. Dessa forma, tentamos eliminar a

subjetividade da medida e por isso o método de detecção nos parece menos

subjetivo, mais prático e mais reprodutível. Essa forma de quantificar a metástase

em LS pode ser alvo de críticas, uma vez que a atual versão do TNM preconiza o uso

do tamanho de metástase no linfonodo.3

Kamath et al.86 e Rausen et al.89 demonstraram que, quantitativamente, o

método de detecção de metástase está correlacionado com o tamanho da metástase

em LS. Por exemplo, metástases vistas somente na IHQ se correlacionam

diretamente com metástases muito pequenas, provavelmente menores que 1 mm ou

somente células isoladas. Já as metástases vistas no exame de congelação ou no

controle de H&E são metástases maiores, geralmente maiores que 1 a 2 mm.

Embora o receptor de estrógeno e o grau nuclear modificado (que inclui

histologia lobular) não tenham sido identificados como fatores preditivos

independentes de metástase em Não-LS na análise multivariada, optamos por incluí-

los no nomograma, pois sua inclusão melhorou a habilidade preditiva global do

modelo.

Gann et al.49demonstraram que a positividade dos receptores de estrógeno

está associada também à presença de metástase em linfonodos axilares. (Tabela 1)

De acordo com nossos achados, em 62% dos pacientes com LS positivo, o LS

foi o único linfonodo positivo na axila, o que torna a obrigatoriedade da

linfadenectomia axilar em todos os casos questionável.

A questão do valor terapêutico da linfadenectomia axilar é ainda um assunto

controverso. O estudo do American College of Surgeons Oncology Group, estudo

89

Z0011 (ACOSOG-Z0011), em andamento, poderá futuramente esclarecer essa

questão.112 Esse estudo, que é braço do estudo em LS (Z0010),153 sorteia de forma

randômica os pacientes com LS positivo em observação e complementação com

linfadenectomia.112

Os pesquisadores favoráveis à realização da linfadenectomia complementar

após um LS positivo argumentam que o esvaziamento axilar é crítico na abordagem

terapêutica, já que a informação do número de linfonodos axilares positivos

influencia no prognóstico,32,154,155 muda o estadiamento3 e orienta a decisão quanto

aos tratamentos adjuvantes.3,4,27,156,157 Somente a metade dos pacientes com

metástase em LS apresenta metástase em Não-LS. Os favoráveis à linfadenectomia

complementar argumentam que esse procedimento pode influenciar na sobrevida via

controle locorregional, eliminando um sítio potencial de doença residual e uma fonte

de metástase à distância.158-160 Uma meta-análise dos estudos randomizados da era

pré-LS revela que há um benefício da sobrevida de 5,4% associado ã linfadenectomia

axilar em pacientes com axila clinicamente negativa.

Os pesquisadores que são oponentes à realização da linfadenectomia

complementar após um LS positivo argumentam que o benefício terapêutico da

complementação do esvaziamento é mínimo.108 Os dados de 25 anos de

acompanhamento do estudo NSABP-B04 revelam que não há nenhum benefício em

sobrevida associado ao tratamento cirúrgico axilar.110 Além disso, somente 50% dos

pacientes com LS positivo – ou 38 % de acordo com nossos dados – apresentam

metástase em Não-LS.51,67-69,73,87-92 Conseqüentemente, a maioria dos pacientes (50%

a 62%) é submetida a esvaziamento axilar desnecessariamente, sem benefício

terapêutico ou informações de estadiamento adicionais. Os oponentes à

90

linfadenectomia complementar também argumentam que, por já apresentarem

metástase em LS, esses pacientes geralmente já têm o tratamento sistêmico definido,

independentemente da informação adicional de doença residual na axila, e se houver

doença residual, o tratamento sistêmico poderia erradicar esse eventual foco residual.

Nesse mesmo aspecto, a radioterapia após um tratamento conservador poderia,

teoricamente, também contribuir no controle da doença axilar residual, pois parte da

axila á também irradiada nos campos de radioterapia da mama, principalmente o

nível 1 da axila.

Naik et al.74 apresentaram os dados de 4.008 casos consecutivos de biópsia de

LS realizadas no MSKCC, com acompanhamento mediano de 31 meses. Nesse

estudo, em 210 casos o LS era positivo e não foi realizada linfadenectomia axilar

(tratamento não-convencional). Nesse grupo em específico houve somente 3 casos de

recidivas axilares (1,4%), sendo somente 1 caso exclusivo em axila, 1 caso

coincidente com recidiva na mama e o outro coincidente com recidiva sistêmica. Em

contrapartida, dos 3.798 casos restantes em que os pacientes foram submetidos ao

tratamento convencional houve 7 casos de recidivas axilares (0,18%). Entende-se por

tratamento convencional: pacientes com LS negativo não submetidos à

linfadenectomia, pacientes com metástase em LS submetidos à linfadenectomia e

pacientes com LS negativos submetidos à linfadenectomia como parte dos casos de

validação do procedimento. Comparando-se as taxas de recidivas axilares entre os

pacientes submetidos ao tratamento convencional e não-convencional,

respectivamente 0,18% e 1,4%, conclui-se que essa diferença é estatisticamente

significante (P = 0,013). Vale a pena ressaltar que o grupo de pacientes submetidos

ao tratamento não-convencional é, geralmente, muito selecionado, ou seja, de baixo

91

risco de metástase em Não-LS. Entretanto, numericamente, num tempo de

acompanhamento de quase 3 anos, o número de recidivas axilares não é alto como se

esperaria. Guenter et al.,161 Fant et al.162 e Winchester et al.163, em casuísticas

menores (respectivamente 46, 31 e 73 pacientes) e com tempo de acompanhamento

semelhante ao estudo de Naik et al.,74 não relataram nenhuma recidiva no grupo de

pacientes com metástase em LS que não foram submetidos à linfadenectomia axilar

complementar (Tabela 24).

Na Tabela 25 estão listados os dados de recidiva axilar após linfadenectomia

axilar clássica, da era pré-LS. Nota-se que, após um tempo médio de

acompanhamento longo, a ordem de grandeza das recidivas axilares é de

aproximadamente 1%. O que não difere muito dos achados de 1,4% relatados por

Naik et al.74 Porém comparações desse tipo são muito questionáveis, já que são

amostras distintas, em períodos e tratamentos adjuvantes diferentes e, neste último, o

acompanhamento mediano é inferior aos dos dados históricos.

Nesse contexto, para ajudar médicos e pacientes, desenvolvemos modelos de

predição de metástase em Não-LS (Figuras 9 e 10) que estão disponíveis também em

software. Esses modelos foram prospectivamente validados e se mostraram bastante

acurados (Figuras 11,12,13 e 14).

Não identificamos nenhuma ferramenta semelhante com tal acurácia na

literatura até o momento. Entretanto, existem limitações em nossos resultados, já que

os Não-LS obtidos após a linfadenectomia complementar foram analisados de forma

rotineira, ou seja, sem o estudo detalhado com cortes seriados e IHQ. Outros estudos

demonstram uma proporção maior de metástases identificadas nos Não-LS quando

estes são examinados com cortes seriados e IHQ.93,164 Portanto, uma avaliação mais

92

detalhada dos Não-LS certamente alterariam nossos resultados. Porém, conforme

discutido anteriormente o valor da doença linfonodal micrometastática ainda é objeto

de discussão e estudos prospectivos em andamento.

Nosso modelo também não avalia a extensão extranodal, que é descrita como

um fator preditivo de metástase em Não-LS.91,93,95,105

Outra potencial crítica aos nossos resultados é o fato de que alguns dos

pacientes, principalmente aqueles com baixo risco de doença em Não-LS, não foram

submetidos à linfadenectomia axilar e conseqüentemente não foram incluídos no

desenvolvimento deste modelo.74 Entretanto, os gráficos das Figuras 13 e 14

demonstram que, aplicando-se os modelos na amostra de validação, estes predizem a

probabilidade de forma bastante acurada, mesmo para aqueles grupos de predição de

risco baixo.

Nossos modelos de predição de metástase e LS e Não-LS não atingem a

perfeição. Em todos os nossos modelos a área abaixo da curva ROC está ao redor de

0,77 na amostra de validação. Para explicar o que isso significa, usaremos como

exemplo o modelo de metástase em Não-LS. Isso significa que, se selecionarmos

randomicamente dois pacientes, dos quais um tem pelo menos um Não-LS positivo e

o outro tem Não-LS negativo, há uma chance de 77% de que o nomograma predirá

uma probabilidade maior para o paciente com Não-LS positivo. Isso ocorre numa

escala que varia entre 0,5 – que seria atingida de fizéssemos um sorteio “tipo cara e

coroa” – a 1, que seria atingido num modelo perfeito em diferenciar um paciente com

Não-LS positivo de um negativo.

93

Tabela 24 – Recidiva axilar após LS Positivo em biópsia de LS sem

esvaziamento complemetar – Revisão da literatura

LS, linfonodo sentinela

Estudo N Acompanhamento Mediano (Meses)

Recidiva Axilar N (%)

Gunther et al.161 46 39 0 Fant et al.162 31 14 0 Winchester et al.163 73 24 0 Naik et al.74 210 25 3 (1,4) Total 360 14-39 3 (0,8)

94

Tabela 25 – Recidiva axilar após linfadenectomia sistemática – Revisão da

literatura

Estudo N Acompanhamento Mediano (Meses)

Recidiva Axilar N (%)

Haagensen165 935 ND 2 (0,2) Fisher et al.39 654 54 8 (1,2) Fowble et al.166 914 40 17 (1,9) Veronesi et al.167 1232 72 0 Siegel et al.168 259 27 2 (0,8) Recht et al.169 420 77 9 (2,1) Halverson et al. 170 309 55 1 (0,3) Chua et al.171 782 88 8 (1,0) Louis-Sylvestre et al.172 320 180 5 (1,6) Total 5825 27–180 52 (0,9)

95

Nenhum de nossos modelos de predição de metástase e LS ou Não-LS

determina condutas a serem tomadas. Esses modelos e os correspondentes softwares

são ferramentas de uso para discussão pré-operatória entre médico e paciente, quer

seja antes da biópsia do LS ou após a identificação de metástase em LS.

Fora de protocolos de estudo, como o ACOSOG Z0011,112 os cirurgiões

devem continuar a recomendar a linfadenectomia complementar na presença de LS

positivo. Entretanto, alguns médicos e pacientes ficam relutantes em complementar a

cirurgia devido ao baixo risco de comprometimento de linfonodos adicionais e pelo

receio das possíveis morbidades associadas a esse procedimento. Nesses casos, os

nomogramas (Figuras 9 e 10) fornecem uma estimativa acurada do risco de doença

linfonodal adicional, permitindo uma decisão terapêutica de forma mais consciente.

Por fim, nossos modelos representam um significante avanço para se estimar

a doença metastática axilar em câncer de mama, ao se comparar com nossa intuição

ou a teorização fundada no subjetivismo do “eu acho que”, também conhecido como

achismo, segundo o Dicionário Houaiss.

.

96

7. CONCLUSÕES

1) Os modelos desenvolvidos e seus respectivos softwares mostraram-se

eficazes para definir a probabilidade metástases em LS e em Não-LS.

2) Os fatores preditivos independentes associados à metástase em LS foram:

tipo histológico, invasão vásculo-linfática, tamanho do tumor, localização do

tumor na mama, idade, grau nuclear, multifocalidade.

3) Os fatores preditivos independentes associados à metástase em Não-LS

foram: tamanho do tumor, invasão vásculo-linfática, multifocalidade, método

de detecção da metástase em LS, número de LS positivos, número de LS

negativos.

4) O estudo anatomopatológico mais detalhado com cortes seriados e

imunoistoquímica para células epiteliais, aumentou significativamente a

identificação de metástase em LS.

97

8. REFERÊNCIAS 1. Recommendations for the reporting of breast carcinoma. Association of

Directors of Anatomic and Surgical Pathology. Hum Pathol 1996; 27(3):220-224.

2. Diseases of the breast. 2nd ed ed. Philadelphia: Lippincott Williams & Wilkins, 2002.

3. TNM (2002) - Breast. In: Sobin LH, Wittekind C, International Union Against Cancer (UICC), editors. New York: Wiley-Liss, 2002: 131-142.

4. Multi-agent chemotherapy for early breast cancer. Cochrane Database Syst Rev 2002;(1):CD000487.

5. Gompertz B. On the nature of the function expressive of the law of human mortality ans on a mew model of determining the value of life contigencies. Phil Trans Roy Soc London 1825; 115:513-585.

6. Laird AK. Dynamics of tumor growth. Br.J.Cancer 18, 490-502. 1964. 7. Heuser L, Spratt JS, Polk HC, Jr. Growth rates of primary breast cancers.

Cancer 1979; 43(5):1888-1894.

8. Schwartz MA. A biomathematical approach to clinical tumor growth. Cancer 1961; 14:1272-1294.

9. Norton L, Simon R, Brereton HD, Bogden AE. Predicting the course of Gompertzian growth. Nature 1976; 264(5586):542-545.

10. Priore RL. Using a mathematical model in the evaluation of human tumor response to chemotherapy. J Natl Cancer Inst 1966; 37(5):635-647.

11. Gregory WM, Reznek RH, Hallett M, Slevin ML. Using mathematical models to estimate drug resistance and treatment efficacy via CT scan measurements of tumour volume. Br J Cancer 1990; 62(4):671-675.

12. Cameron DA, Ritchie AA, Miller WR. The relative importance of proliferation and cell death in breast cancer growth and response to tamoxifen. Eur J Cancer 2001; 37(12):1545-1553.

13. D'Eredita' G, Giardina C, Martellotta M, Natale T, Ferrarese F. Prognostic factors in breast cancer: the predictive value of the Nottingham Prognostic Index in patients with a long-term follow-up that were treated in a single institution. Eur J Cancer 2001; 37(5):591-596.

14. Sundquist M, Thorstenson S, Brudin L, Nordenskjold B. Applying the Nottingham Prognostic Index to a Swedish breast cancer population. South

98

East Swedish Breast Cancer Study Group. Breast Cancer Res Treat 1999; 53(1):1-8.

15. Anderson TJ, Alexander FE, Lamb J, Smith A, Forrest AP. Pathology characteristics that optimize outcome prediction of a breast screening trial. Br J Cancer 2000; 83(4):487-492.

16. Berry DA, Parmigiani G, Sanchez J, Schildkraut J, Winer E. Probability of carrying a mutation of breast-ovarian cancer gene BRCA1 based on family history. J Natl Cancer Inst 1997; 89(3):227-238.

17. Bodian CA, Perzin KH, Lattes R. Lobular neoplasia. Long term risk of breast cancer and relation to other factors. Cancer 1996; 78(5):1024-1034.

18. Claus EB, Risch N, Thompson WD. Autosomal dominant inheritance of early-onset breast cancer. Implications for risk prediction. Cancer 1994; 73(3):643-651.

19. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, Mulvihill JJ. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989; 81(24):1879-1886.

20. Gilpin CA, Carson N, Hunter AG. A preliminary validation of a family history assessment form to select women at risk for breast or ovarian cancer for referral to a genetics center. Clin Genet 2000; 58(4):299-308.

21. Shattuck-Eidens D, Oliphant A, McClure M, McBride C, Gupte J, Rubano T, Pruss D, Tavtigian SV, Teng DH, Adey N, Staebell M, Gumpper K, Lundstrom R, Hulick M, Kelly M, Holmen J, Lingenfelter B, Manley S, Fujimura F, Luce M, Ward B, Cannon-Albright L, Steele L, Offit K, Thomas A, . BRCA1 sequence analysis in women at high risk for susceptibility mutations. Risk factor analysis and implications for genetic testing. JAMA 1997; 278(15):1242-1250.

22. Bondy ML, Lustbader ED, Halabi S, Ross E, Vogel VG. Validation of a breast cancer risk assessment model in women with a positive family history. J Natl Cancer Inst 1994; 86(8):620-625.

23. Costantino JP, Gail MH, Pee D, Anderson S, Redmond CK, Benichou J, Wieand HS. Validation studies for models projecting the risk of invasive and total breast cancer incidence. J Natl Cancer Inst 1999; 91(18):1541-1548.

24. McGuigan KA, Ganz PA, Breant C. Agreement between breast cancer risk estimation methods. J Natl Cancer Inst 1996; 88(18):1315-1317.

25. Rockhill B, Spiegelman D, Byrne C, Hunter DJ, Colditz GA. Validation of the Gail et al. model of breast cancer risk prediction and implications for chemoprevention. J Natl Cancer Inst 2001; 93(5):358-366.

99

26. Spiegelman D, Colditz GA, Hunter D, Hertzmark E. Validation of the Gail et al. model for predicting individual breast cancer risk. J Natl Cancer Inst 1994; 86(8):600-607.

27. NIH consensus conference. Treatment of early-stage breast cancer. JAMA 1991; 265(3):391-395.

28. Mustafa IA, Cole B, Wanebo HJ, Bland KI, Chang HR. Prognostic analysis of survival in small breast cancers. J Am Coll Surg 1998; 186(5):562-569.

29. Rosen PP, Groshen S, Saigo PE, Kinne DW, Hellman S. Pathological prognostic factors in stage I (T1N0M0) and stage II (T1N1M0) breast carcinoma: a study of 644 patients with median follow- up of 18 years. J Clin Oncol 1989; 7(9):1239-1251.

30. Rosen PP, Groshen S. Factors influencing survival and prognosis in early breast carcinoma (T1N0M0-T1N1M0). Assessment of 644 patients with median follow-up of 18 years. Surg Clin North Am 1990; 70(4):937-962.

31. Rosen PP, Groshen S, Kinne DW, Norton L. Factors influencing prognosis in node-negative breast carcinoma: analysis of 767 T1N0M0/T2N0M0 patients with long-term follow-up. J Clin Oncol 1993; 11(11):2090-2100.

32. Carter CL, Allen C, Henson DE. Relation of tumor size, lymph node status, and survival in 24,740 breast cancer cases. Cancer 1989; 63(1):181-187.

33. Halsted WS. The results of operation for the cure of cancer the breast performed at Johns Hopkins Hospital from June 1889 to January. Johns Hopkins Hosp Rep 1894; 4:297-350.

34. Meyer W. An improved method of the radical operation for carcinoma of the breast. Med Rec 1894; 46:746-749.

35. Halsted WS. A clinical and histological study of certain adenocarcinomata of the breast: And a brief consideration of the supraclavicular operation and the results of operations for cancer of the breast from 1889 to 1898 at John Hopkins Hospital. Ann Surg 1898; 28:557-576.

36. Scarff RW, Handley RS. Prognosis in carcinoma of the breast. Lancet 1938;(2):582-583.

37. Patey DH, Dyson WH. Prognosis of carcinoma of the breast in relation to the type of operation performed. Br J Cancer 1948; 2:7-13.

38. Urban JA, Baker HW. Radical mastectomy in continuity with en bloc resection of the internal mammary lymph node chain - A new procedure for primary operable cancer of the breast. Cancer 1952; 5(5):992-1008.

39. Fisher B, Redmond C, Fisher ER, Bauer M, Wolmark N, Wickerham DL, Deutsch M, Montague E, Margolese R, Foster R. Ten-year results of a

100

randomized clinical trial comparing radical mastectomy and total mastectomy with or without radiation. N Engl J Med 1985; 312(11):674-681.

40. Fisher B, Anderson S, Bryant J, Margolese RG, Deutsch M, Fisher ER, Jeong JH, Wolmark N. Twenty-year follow-up of a randomized trial comparing total mastectomy, lumpectomy, and lumpectomy plus irradiation for the treatment of invasive breast cancer. N Engl J Med 2002; 347(16):1233-1241.

41. Petrek JA, Senie RT, Peters M, Rosen PP. Lymphedema in a cohort of breast carcinoma survivors 20 years after diagnosis. Cancer 2001; 92(6):1368-1377.

42. Ahlgren J, Stal O, Westman G, Arnesson LG. Prediction of axillary lymph node metastases in a screened breast cancer population. South-East Sweden Breast Cancer Group. Acta Oncol 1994; 33(6):603-608.

43. Barth A, Craig PH, Silverstein MJ. Predictors of axillary lymph node metastases in patients with T1 breast carcinoma [see comments]. Cancer 1997; 79(10):1918-1922.

44. Chadha M, Chabon AB, Friedmann P, Vikram B. Predictors of axillary lymph node metastases in patients with T1 breast cancer. A multivariate analysis. Cancer 1994; 73(2):350-353.

45. Choong PL, deSilva CJ, Dawkins HJ, Sterrett GF, Robbins P, Harvey JM, Papadimitriou J, Attikiouzel Y. Predicting axillary lymph node metastases in breast carcinoma patients. Breast Cancer Res Treat 1996; 37(2):135-149.

46. De Laurentiis M, Gallo C, De Placido S, Perrone F, Pettinato G, Petrella G, Carlomagno C, Panico L, Delrio P, Bianco AR. A predictive index of axillary nodal involvement in operable breast cancer. Br J Cancer 1996; 73(10):1241-1247.

47. Fehr MK, Kochli OR, Helfenstein U, Schar G, Bussmann C, Fink D, Haller U. [Multivariate analysis of clinico-pathologic predictors of axillary lymph node metastasis in invasive breast carcinoma]. Geburtshilfe Frauenheilkd 1995; 55(4):182-188.

48. Fein DA, Fowble BL, Hanlon AL, Hooks MA, Hoffman JP, Sigurdson ER, Jardines LA, Eisenberg BL. Identification of women with T1-T2 breast cancer at low risk of positive axillary nodes. J Surg Oncol 1997; 65(1):34-39.

49. Gann PH, Colilla SA, Gapstur SM, Winchester DJ, Winchester DP. Factors associated with axillary lymph node metastasis from breast carcinoma: descriptive and predictive analyses. Cancer 1999; 86(8):1511-1519.

50. Giuliano AE, Barth AM, Spivack B, Beitsch PD, Evans SW. Incidence and predictors of axillary metastasis in T1 carcinoma of the breast [see comments]. J Am Coll Surg 1996; 183(3):185-189.

101

51. Giuliano AE, Jones RC, Brennan M, Statman R. Sentinel lymphadenectomy in breast cancer. J Clin Oncol 1997; 15(6):2345-2350.

52. Maibenco DC, Weiss LK, Pawlish KS, Severson RK. Axillary lymph node metastases associated with small invasive breast carcinomas. Cancer 1999; 85(7):1530-1536.

53. Olivotto IA, Jackson JS, Mates D, Andersen S, Davidson W, Bryce CJ, Ragaz J. Prediction of axillary lymph node involvement of women with invasive breast carcinoma: a multivariate analysis [see comments]. Cancer 1998; 83(5):948-955.

54. Ravdin PM, De Laurentiis M, Vendely T, Clark GM. Prediction of axillary lymph node status in breast cancer patients by use of prognostic indicators [see comments]. J Natl Cancer Inst 1994; 86(23):1771-1775.

55. Rivadeneira DE, Simmons RM, Christos PJ, Hanna K, Daly JM, Osborne MP. Predictive factors associated with axillary lymph node metastases in T1a and T1b breast carcinomas: analysis in more than 900 patients. J Am Coll Surg 2000; 191(1):1-6.

56. Shoup M, Malinzak L, Weisenberger J, Aranha GV. Predictors of axillary lymph node metastasis in T1 breast carcinoma. Am Surg 1999; 65(8):748-752.

57. Van Lancker M, Goor C, Sacre R, Lamote J, Van Belle S, De Coene N, Roelstraete A, Storme G. Patterns of axillary lymph node metastasis in breast cancer. Am J Clin Oncol 1995; 18(3):267-272.

58. Velanovich V. Predictors of lymph node metastasis for breast cancer (Meeting abstract). Proc Annu Meet Am Soc Clin Oncol 1997; 16:A687.

59. Velanovich V, Szymanski W. Lymph node metastasis in breast cancer: common prognostic markers lack predictive value. Ann Surg Oncol 1998; 5(7):613-619.

60. Voogd AC, Coebergh JW, Repelaer van Driel OJ, Roumen RM, van Beek MW, Vreugdenhil A, Crommelin MA. The risk of nodal metastases in breast cancer patients with clinically negative lymph nodes: a population-based analysis. Breast Cancer Res Treat 2000; 62(1):63-69.

61. Walls J, Boggis CR, Wilson M, Asbury DL, Roberts JV, Bundred NJ, Mansel RE. Treatment of the axilla in patients with screen-detected breast cancer. Br J Surg 1993; 80(4):436-438.

62. Gajdos C, Tartter PI, Bleiweiss IJ, Bodian C, Brower ST. Stage 0 to stage III breast cancer in young women. J Am Coll Surg 2000; 190(5):523-529.

102

63. Mustafa IA, Cole B, Wanebo HJ, Bland KI, Chang HR. The impact of histopathology on nodal metastases in minimal breast cancer. Arch Surg 1997; 132(4):384-390.

64. Yiangou C, Shousha S, Sinnett HD. Primary tumour characteristics and axillary lymph node status in breast cancer. Br J Cancer 1999; 80(12):1974-1978.

65. Cabanas RM. An approach for the treatment of penile carcinoma. Cancer 1977; 39(2):456-466.

66. Morton DL, Wen DR, Wong JH, Economou JS, Cagle LA, Storm FK, Foshag LJ, Cochran AJ. Technical details of intraoperative lymphatic mapping for early stage melanoma. Arch Surg 1992; 127(4):392-399.

67. Giuliano AE, Kirgan DM, Guenther JM, Morton DL. Lymphatic mapping and sentinel lymphadenectomy for breast cancer [see comments]. Ann Surg 1994; 220(3):391-398.

68. Krag D, Weaver D, Ashikaga T, Moffat F, Klimberg VS, Shriver C, Feldman S, Kusminsky R, Gadd M, Kuhn J, Harlow S, Beitsch P. The sentinel node in breast cancer--a multicenter validation study [see comments]. N Engl J Med 1998; 339(14):941-946.

69. Veronesi U, Paganelli G, Galimberti V, Viale G, Zurrida S, Bedoni M, Costa A, De Cicco C, Geraghty JG, Luini A, Sacchini V, Veronesi P. Sentinel-node biopsy to avoid axillary dissection in breast cancer with clinically negative lymph-nodes [see comments]. Lancet 1997; 349(9069):1864-1867.

70. Veronesi U, Paganelli G, Viale G, Luini A, Zurrida S, Galimberti V, Intra M, Veronesi P, Robertson C, Maisonneuve P, Renne G, De Cicco C, De Lucia F, Gennari R. A randomized comparison of sentinel-node biopsy with routine axillary dissection in breast cancer. N Engl J Med 2003; 349(6):546-553.

71. Cody HS, III. Sentinel lymph node mapping in breast cancer. Oncology (Huntingt) 1999; 13(1):25-34.

72. Cody HS, III. Management of the axilla in early stage breast cancer: will sentinel node biopsy end the debate? [editorial]. J Surg Oncol 1999; 71(3):137-139.

73. O'Hea BJ, Hill AD, El Shirbiny AM, Yeh SD, Rosen PP, Coit DG, Borgen PI, Cody HS, III. Sentinel lymph node biopsy in breast cancer: initial experience at Memorial Sloan-Kettering Cancer Center [see comments]. J Am Coll Surg 1998; 186(4):423-427.

74. Naik AM, Fey J, Gemignani M, Heerdt A, Montgomery L, Petrek J, Port E, Sacchini V, Sclafani L, VanZee K, Wagman R, Borgen PI, Cody HS, III. The risk of axillary relapse after sentinel lymph node biopsy for breast cancer is

103

comparable with that of axillary lymph node dissection: a follow-up study of 4008 procedures. Ann Surg 2004; 240(3):462-468.

75. Giuliano AE, Dale PS, Turner RR, Morton DL, Evans SW, Krasne DL. Improved axillary staging of breast cancer with sentinel lymphadenectomy [see comments]. Ann Surg 1995; 222(3):394-399.

76. Jackson JS, Olivotto IA, Wai MDE, Grau C, Mates D, Ragaz J. A decision analysis of the effect of avoiding axillary lymph node dissection in low risk women with invasive breast carcinoma. Cancer 2000; 88(8):1852-1862.

77. Halverson KJ, Taylor ME, Perez CA, Garcia DM, Myerson R, Philpott G, Levy J, Simpson JR, Tucker G, Rush C. Management of the axilla in patients with breast cancers one centimeter or smaller [see comments]. Am J Clin Oncol 1994; 17(6):461-466.

78. Iwasaki Y, Fukutomi T, Akashi-Tanaka S, Nanasawa T, Tsuda H. Axillary node metastasis from T1N0M0 breast cancer: possible avoidance of dissection in a subgroup. Jpn J Clin Oncol 1998; 28(10):601-603.

79. Lohrisch C, Jackson J, Jones A, Mates D, Olivotto IA. Relationship between tumor location and relapse in 6,781 women with early invasive breast cancer. J Clin Oncol 2000; 18(15):2828-2835.

80. Port ER, Tan LK, Borgen PI, Van Zee KJ. Incidence of axillary lymph node metastases in T1a and T1b breast carcinoma. Ann Surg Oncol 1998; 5(1):23-27.

81. Reger V, Beito G, Jolly PC. Factors affecting the incidence of lymph node metastases in small cancers of the breast. Am J Surg 1989; 157(5):501-502.

82. Saiz E, Toonkel R, Poppiti RJ, Jr., Robinson MJ. Infiltrating breast carcinoma smaller than 0.5 centimeters: is lymph node dissection necessary? [see comments]. Cancer 1999; 85(10):2206-2211.

83. Seidman JD, Schnaper LA, Aisner SC. Relationship of the size of the invasive component of the primary breast carcinoma to axillary lymph node metastasis. Cancer 1995; 75(1):65-71.

84. Silverstein MJ, Gierson ED, Waisman JR, Colburn WJ, Gamagami P. Predicting axillary node positivity in patients with invasive carcinoma of the breast by using a combination of T category and palpability. J Am Coll Surg 1995; 180(6):700-704.

85. Chu KU, Turner RR, Hansen NM, Brennan MB, Bilchik A, Giuliano AE. Do all patients with sentinel node metastasis from breast carcinoma need complete axillary node dissection? [see comments]. Ann Surg 1999; 229(4):536-541.

104

86. Kamath VJ, Giuliano R, Dauway EL, Cantor A, Berman C, Ku NN, Cox CE, Reintgen DS. Characteristics of the sentinel lymph node in breast cancer predict further involvement of higher-echelon nodes in the axilla: a study to evaluate the need for complete axillary lymph node dissection. Arch Surg 2001; 136(6):688-692.

87. Abdessalam SF, Zervos EE, Prasad M, Farrar WB, Yee LD, Walker MJ, Carson WB, Burak WE, Jr. Predictors of positive axillary lymph nodes after sentinel lymph node biopsy in breast cancer. Am J Surg 2001; 182(4):316-320.

88. Albertini JJ, Lyman GH, Cox C, Yeatman T, Balducci L, Ku N, Shivers S, Berman C, Wells K, Rapaport D, Shons A, Horton J, Greenberg H, Nicosia S, Clark R, Cantor A, Reintgen DS. Lymphatic mapping and sentinel node biopsy in the patient with breast cancer. JAMA 1996; 276(22):1818-1822.

89. Rahusen FD, Torrenga H, van Diest PJ, Pijpers R, van der WE, Licht J, Meijer S. Predictive factors for metastatic involvement of nonsentinel nodes in patients with breast cancer. Arch Surg 2001; 136(9):1059-1063.

90. Reynolds C, Mick R, Donohue JH, Grant CS, Farley DR, Callans LS, Orel SG, Keeney GL, Lawton TJ, Czerniecki BJ. Sentinel lymph node biopsy with metastasis: can axillary dissection be avoided in some patients with breast cancer? J Clin Oncol 1999; 17(6):1720-1726.

91. Teng S, Dupont E, McCann C, Wang J, Bolano M, Durand K, Peltz E, Bass SS, Cantor A, Ku NN, Cox CE. Do cytokeratin-positive-only sentinel lymph nodes warrant complete axillary lymph node dissection in patients with invasive breast cancer? Am Surg 2000; 66(6):574-578.

92. Veronesi U, Paganelli G, Viale G, Galimberti V, Luini A, Zurrida S, Robertson C, Sacchini V, Veronesi P, Orvieto E, De Cicco C, Intra M, Tosi G, Scarpa D. Sentinel lymph node biopsy and axillary dissection in breast cancer: results in a large series [see comments]. J Natl Cancer Inst 1999; 91(4):368-373.

93. Turner RR, Chu KU, Qi K, Botnick LE, Hansen NM, Glass EC, Giuliano AE. Pathologic features associated with nonsentinel lymph node metastases in patients with metastatic breast carcinoma in a sentinel lymph node. Cancer 2000; 89(3):574-581.

94. Canavese G, Gipponi M, Catturich A, Vecchio C, Tomei D, Nicolo G, Carli F, Spina B, Bonelli L, Villa G, Buffoni F, Bianchi P, Agnese A, Mariani G. Technical issues and pathologic implications of sentinel lymph node biopsy in early-stage breast cancer patients. J Surg Oncol 2001; 77(2):81-87.

95. Cserni G. Sentinel lymph-node biopsy-based prediction of further breast cancer metastases in the axilla. Eur J Surg Oncol 2001; 27(6):532-538.

105

96. Liang WC, Sickle-Santanello BJ, Nims TA. Is a completion axillary dissection indicated for micrometastases in the sentinel lymph node? Am J Surg 2001; 182(4):365-368.

97. Viale G, Maiorano E, Mazzarol G, Zurrida S, Galimberti V, Luini A, Renne G, Pruneri G, Maisonneuve P, Veronesi U. Histologic detection and clinical implications of micrometastases in axillary sentinel lymph nodes for patients with breast carcinoma. Cancer 2001; 92(6):1378-1384.

98. Weiser MR, Montgomery LL, Tan LK, Susnik B, Leung DY, Borgen PI, Cody HS, III. Lymphovascular invasion enhances the prediction of non-sentinel node metastases in breast cancer patients with positive sentinel nodes. Ann Surg Oncol 2001; 8(2):145-149.

99. Wong SL, Edwards MJ, Chao C, Tuttle TM, Noyes RD, Woo C, Cerrito PB, McMasters KM. Predicting the status of the nonsentinel axillary nodes: a multicenter study. Arch Surg 2001; 136(5):563-568.

100. den Bakker MA, van Weeszenberg A, de Kanter AY, Beverdam FH, Pritchard C, van der Kwast TH, Menke-Pluymers M. Non-sentinel lymph node involvement in patients with breast cancer and sentinel node micrometastasis; too early to abandon axillary clearance. J Clin Pathol 2002; 55(12):932-935.

101. Jakub JW, Diaz NM, Ebert MD, Cantor A, Reintgen DS, Dupont EL, Shons AR, Cox CE. Completion axillary lymph node dissection minimizes the likelihood of false negatives for patients with invasive breast carcinoma and cytokeratin positive only sentinel lymph nodes. Am J Surg 2002; 184(4):302-306.

102. Mignotte H, Treilleux I, Faure C, Nessah K, Bremond A. Axillary lymph-node dissection for positive sentinel nodes in breast cancer patients. Eur J Surg Oncol 2002; 28(6):623-626.

103. Sachdev U, Murphy K, Derzie A, Jaffer S, Bleiweiss IJ, Brower S. Predictors of nonsentinel lymph node metastasis in breast cancer patients. Am J Surg 2002; 183(3):213-217.

104. Hwang RF, Krishnamurthy S, Hunt KK, Mirza N, Ames FC, Feig B, Kuerer HM, Singletary SE, Babiera G, Meric F, Akins JS, Neely J, Ross MI. Clinicopathologic factors predicting involvement of nonsentinel axillary nodes in women with breast cancer. Ann Surg Oncol 2003; 10(3):248-254.

105. Fleming FJ, Kavanagh D, Crotty TB, Quinn CM, McDermott EW, O'Higgins N, Hill AD. Factors affecting metastases to non-sentinel lymph nodes in breast cancer. J Clin Pathol 2004; 57(1):73-76.

106. Bevilacqua JLB, Manasseh DM, Kattan MW, Tan LK, Cody HS, III, Borgen PI, Van Zee KJ. A mulitivariate nomogram to predict likelihood of non-

106

sentinel node metastases in breast cancer patients with positive sentinel nodes. (Meeting Abstract # 84). 56th Annual Cancer Symposium of Society of Surgical Oncology 2003.

107. Van Zee KJ, Manasseh DM, Bevilacqua JL, Boolbol SK, Fey JV, Tan LK, Borgen PI, Cody HS, III, Kattan MW. A nomogram for predicting the likelihood of additional nodal metastases in breast cancer patients with a positive sentinel node biopsy. Ann Surg Oncol 2003; 10(10):1140-1151.

108. Cady B. Case against axillary lymphadenectomy for most patients with infiltrating breast cancer. J Surg Oncol 1997; 66(1):7-10.

109. Petrek JA, Blackwood MM. Axillary dissection: current practice and technique. Curr Probl Surg 1995; 32(4):257-323.

110. Fisher B, Jeong JH, Anderson S, Bryant J, Fisher ER, Wolmark N. Twenty-five-year follow-up of a randomized trial comparing radical mastectomy, total mastectomy, and total mastectomy followed by irradiation. N Engl J Med 2002; 347(8):567-575.

111. Orr RK. The impact of prophylactic axillary node dissection on breast cancer survival--a Bayesian meta-analysis. Ann Surg Oncol 1999; 6(1):109-116.

112. American College of Surgeons Oncology Group (ACOSOG). A randomized trial of axillary node dissection in women with clinical T1 - 2 N0 M0 breast cancer who have a positive sentinel node (Protocol Z0011). http://www.acosog.org/studies/synopses/Z0011_Synopsis.pdf . 2005. 28-5-2005.

113. Cody HS, III, Borgen PI. State-of-the-art approaches to sentinel node biopsy for breast cancer: study design, patient selection, technique, and quality control at Memorial Sloan-Kettering Cancer Center [see comments]. Surg Oncol 1999; 8(2):85-91.

114. Boolbol SK, Fey JV, Borgen PI, Heerdt AS, Montgomery LL, Paglia M, Petrek JA, Cody HS, Van Zee KJ. Intradermal isotope injection: a highly accurate method of lymphatic mapping in breast carcinoma. Ann Surg Oncol 2001; 8(1):20-24.

115. Linehan DC, Hill AD, Akhurst T, Yeung H, Yeh SD, Tran KN, Borgen PI, Cody HS, III. Intradermal radiocolloid and intraparenchymal blue dye injection optimize sentinel node identification in breast cancer patients [see comments]. Ann Surg Oncol 1999; 6(5):450-454.

116. Martin RC, Derossis AM, Fey J, Yeung H, Yeh SD, Akhurst T, Heerdt AS, Petrek J, VanZee KJ, Montgomery LL, Borgen PI, Cody HS, III. Intradermal isotope injection is superior to intramammary in sentinel node biopsy for breast cancer. Surgery 2001; 130(3):432-438.

107

117. Fitzgibbons PL, Page DL, Weaver D, Thor AD, Allred DC, Clark GM, Ruby SG, O'Malley F, Simpson JF, Connolly JL, Hayes DF, Edge SB, Lichter A, Schnitt SJ. Prognostic factors in breast cancer. College of American Pathologists Consensus Statement 1999. Arch Pathol Lab Med 2000; 124(7):966-978.

118. Bevilacqua RG. Estatística na tese. In: Cunha AC, Freddi MJAL, Crestana MF, Aragão MS, Cardoso SC, Vilhena V, editors. Guia de apresentação de dissertações, teses e monografias. Serviço de Biblioteca e Documentação da Faculdade de Medicina da Universidade de São Paulo - SBD/FMUSP, 2004: 98-107.

119. Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15(4):361-387.

120. Harrell FE, Jr. Regression Modeling Strategies With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York: Springer-Verlag, 2001.

121. Ferreira ABH. Novo Dicionário Aurélio da Língua Portuguesa. 3a. ed. Regis Ltda, 2004.

122. Rosen PP. Rosen's Breast Pathology. 1st ed. Philadelphia,PA: Lippincott-Ravers Publishers, 1997.

123. Hosmer DWJ, Lemeshow S. Applied logistic regression. 2nd ed. Hoboken,NJ, USA: John Wiley & Sons, Inc., 2000.

124. Hansen NM, Ye X, Grube BJ, Giuliano AE. Manipulation of the primary breast tumor and the incidence of sentinel node metastases from invasive breast cancer. Arch Surg 2004; 139(6):634-639.

125. Moore KH, Thaler HT, Tan LK, Borgen PI, Cody HS, III. Immunohistochemically detected tumor cells in the sentinel lymph nodes of patients with breast carcinoma: biologic metastasis or procedural artifact? Cancer 2004; 100(5):929-934.

126. King TA, Ganaraj A, Fey JV, Tan LK, Hudis C, Norton L, Cody HS, III, Borgen PI. Cytokeratin-positive cells in sentinel lymph nodes in breast cancer are not random events: experience in patients undergoing prophylactic mastectomy. Cancer 2004; 101(5):926-933.

127. Cote RJ, Peterson HF, Chaiwun B, Gelber RD, Goldhirsch A, Castiglione-Gertsch M, Gusterson B, Neville AM. Role of immunohistochemical detection of lymph-node metastases in management of breast cancer. International Breast Cancer Study Group [see comments]. Lancet 1999; 354(9182):896-900.

108

128. de Mascarel I, MacGrogan G, Picot V, Mathoulin-Pelissier S. Prognostic significance of immunohistochemically detected breast cancer node metastases in 218 patients. Br J Cancer 2002; 87(1):70-74.

129. McGuckin MA, Cummings MC, Walsh MD, Hohn BG, Bennett IC, Wright RG. Occult axillary node metastases in breast cancer: their detection and prognostic significance. Br J Cancer 1996; 73(1):88-95.

130. Rosai J. Ackerman's Surgical Pathology. St. Louis, Toronto: The C.V. Mosby Company, 1981.

131. Cserni G. Metastases in axillary sentinel lymph nodes in breast cancer as detected by intensive histopathological work up [see comments]. J Clin Pathol 1999; 52(12):922-924.

132. Czerniecki BJ, Scheff AM, Callans LS, Spitz FR, Bedrosian I, Conant EF, Orel SG, Berlin J, Helsabeck C, Fraker DL, Reynolds C. Immunohistochemistry with pancytokeratins improves the sensitivity of sentinel lymph node biopsy in patients with breast carcinoma. Cancer 1999; 85(5):1098-1103.

133. Dowlatshahi K, Fan M, Bloom KJ, Spitz DJ, Patel S, Snider HC, Jr. Occult metastases in the sentinel lymph nodes of patients with early stage breast carcinoma: A preliminary study [see comments]. Cancer 1999; 86(6):990-996.

134. Pendas S, Dauway E, Cox CE, Giuliano R, Ku NN, Schreiber RH, Reintgen DS. Sentinel node biopsy and cytokeratin staining for the accurate staging of 478 breast cancer patients. Am Surg 1999; 65(6):500-505.

135. Turner RR, Ollila DW, Krasne DL, Giuliano AE. Histopathologic validation of the sentinel lymph node hypothesis for breast carcinoma. Ann Surg 1997; 226(3):271-276.

136. Weaver DL, Krag DN, Ashikaga T, Harlow SP, O'Connell M. Pathologic analysis of sentinel and nonsentinel lymph nodes in breast carcinoma: a multicenter study [see comments]. Cancer 2000; 88(5):1099-1107.

137. Liberman L, Schneider L. Review of published experience. In: Cody HS, editor. Sentinel Lymph Node Biopsy. London, UK: Martin Dunitz Ltd., 2002: 285-310.

138. Dowlatshahi K, Fan M, Snider HC, Habib FA. Lymph node micrometastases from breast carcinoma: reviewing the dilemma. Cancer 1997; 80(7):1188-1197.

139. Kattan MW, Giri D, Panageas KS, Hummer A, Cranor M, Van Zee KJ, Hudis CA, Norton L, Borgen PI, Tan LK. A tool for predicting breast carcinoma mortality in women who do not receive adjuvant therapy. Cancer 2004; 101(11):2509-2515.

109

140. Liberman L. Pathologic analysis of sentinel lymph nodes in breast carcinoma [editorial; comment]. Cancer 2000; 88(5):971-977.

141. Maibenco DC, Dombi GW, Kau TY, Severson RK. Impact of micrometastases on the long-term survival of women with T1 breast cancer (Meeting Abstract# 2006). Breast Cancer Res Treat 2004; 88 Suppl 1:S78.

142. Millis RR, Springall R, Lee AH, Ryder K, Rytina ER, Fentiman IS. Occult axillary lymph node metastases are of no prognostic significance in breast cancer. Br J Cancer 2002; 86(3):396-401.

143. Tan LK, Giri D, Pangeas K, Cranor M, Bevilacqua JL, Hummer A, Brogi E, Norton L, Hudis C, Borgen PI. Occult micrometastases in axillary lymph nodes of breast cancer patients are significant: a retrospective study with long term follow up. (Meeting Abstract #146). Proc Annu Meet Am Soc Clin Oncol 2002; 21:37A.

144. Giri D, Thaler H, Bevilacqua JLB, Cranor M, Tan LK. Multifocal versus unifocal node-negative breast carcinomas: Are they prognostically different? (Meeting Abstract #132). 90th Annual Meeting of United States and Canadian Academy of Pathology, Atlanta, Georgia, March 2001 2001.

145. Asaga T, Masuzawa T, Yoshida A, Kawahara S. Location of breast cancer as a prognostic factor (Meeting Abstract). Breast Cancer Res Treat 1994; 32:S48.

146. Zucali R, Mariani L, Marubini E, Kenda R, Lozza L, Rilke F, Veronesi U. Early breast cancer: evaluation of the prognostic role of the site of the primary tumor. J Clin Oncol 1998; 16(4):1363-1366.

147. Israel L, Gogeul E, Morere JF, Breau JL. Distinct metastatic pattern in breast tumor of inner and central areas (Meeting abstract). Proc Annu Meet Am Soc Clin Oncol 1995; 14:A110.

148. Bevilacqua JL, Gucciardo G, Cody HS, MacDonald KA, Sacchini V, Borgen PI, Van Zee KJ. A selection algorithm for internal mammary sentinel lymph node biopsy in breast cancer. Eur J Surg Oncol 2002; 28(6):603-614.

149. Bevilacqua J, Cody H, III, MacDonald KA, Tan LK, Borgen PI, Van Zee KJ. A prospective validated model for predicting axillary node metastases based on 2,000 sentinel node procedures: the role of tumour location [corrected]. Eur J Surg Oncol 2002; 28(5):490-500.

150. Fiorentino C, Berruti A, Bottini A, Bodini M, Brizzi MP, Brunelli A, Marini U, Allevi G, Aguggini S, Tira A, Alquati P, Olivetti L, Dogliotti L. Accuracy of mammography and echography versus clinical palpation in the assessment of response to primary chemotherapy in breast cancer patients with operable disease. Breast Cancer Res Treat 2001; 69(2):143-151.

110

151. Yang WT, Lam WW, Cheung H, Suen M, King WW, Metreweli C. Sonographic, magnetic resonance imaging, and mammographic assessments of preoperative size of breast cancer. J Ultrasound Med 1997; 16(12):791-797.

152. Davis PL, Staiger MJ, Harris KB, Ganott MA, Klementaviciene J, McCarty KS, Jr., Tobon H. Breast cancer measurements with magnetic resonance imaging, ultrasonography, and mammography. Breast Cancer Res Treat 1996; 37(1):1-9.

153. American College of Surgeons Oncology Group (ACOSOG). A prognostic study of sentinel node and bone marrow micrometastases in women with clinical T1 or T2 N0 M0 breast cancer (Protocol Z0010). https://www.acosog.org/studies/synopses/Z0010_Synopsis.pdf . 2005. 28-5-2005.

154. Cabanes PA, Salmon RJ, Vilcoq JR, Durand JC, Fourquet A, Gautier C,

Asselain B. Value of axillary dissection in addition to lumpectomy and radiotherapy in early breast cancer. The Breast Carcinoma Collaborative Group of the Institut Curie [see comments]. Lancet 1992; 339(8804):1245-1248.

155. Moore MP, Kinne DW. Axillary lymphadenectomy: a diagnostic and therapeutic procedure. J Surg Oncol 1997; 66(1):2-6.

156. Polychemotherapy for early breast cancer: an overview of the randomised trials. Early Breast Cancer Trialists' Collaborative Group [see comments]. Lancet 1998; 352(9132):930-942.

157. Tamoxifen for early breast cancer: an overview of the randomised trials. Early Breast Cancer Trialists' Collaborative Group [see comments]. Lancet 1998; 351(9114):1451-1467.

158. Harris JR, Osteen RT. Patients with early breast cancer benefit from effective axillary treatment. Breast Cancer Res Treat 1985; 5(1):17-21.

159. Hayward J, Caleffi M. The significance of local control in the primary treatment of breast cancer. Lucy Wortham James clinical research award. Arch Surg 1987; 122(11):1244-1247.

160. Sosa JA, Diener-West M, Gusev Y, Choti MA, Lange JR, Dooley WC, Zeiger MA. Association between extent of axillary lymph node dissection and survival in patients with stage I breast cancer. Ann Surg Oncol 1998; 5(2):140-149.

161. Guenther JM, Hansen NM, DiFronzo LA, Giuliano AE, Collins JC, Grube BL, O'Connell TX. Axillary dissection is not required for all patients with breast cancer and positive sentinel nodes. Arch Surg 2003; 138(1):52-56.

111

162. Fant JS, Grant MD, Knox SM, Livingston SA, Ridl K, Jones RC, Kuhn JA. Preliminary outcome analysis in patients with breast cancer and a positive sentinel lymph node who declined axillary dissection. Ann Surg Oncol 2003; 10(2):126-130.

163. Winchester DJ, Sener SF, Winchester DP, Perlman RM, Goldschmidt RA, Motykie G, Martz CH, Rabbitt SL, Brenin D, Stull MA, Moulthrop JM. Sentinel lymphadenectomy for breast cancer: experience with 180 consecutive patients: efficacy of filtered technetium 99m sulphur colloid with overnight migration time. J Am Coll Surg 1999; 188(6):597-603.

164. Chu KU, Turner RR, Hansen NM, Brennan MB, Giuliano AE. Sentinel node metastasis in patients with breast carcinoma accurately predicts immunohistochemically detectable nonsentinel node metastasis. Ann Surg Oncol 1999; 6(8):756-761.

165. Haagensen CD. Disease of the Breast. New York: Sounders, 1986: 911-913.

166. Fowble B, Solin LJ, Schultz DJ, Goodman RL. Frequency, sites of relapse, and outcome of regional node failures following conservative surgery and radiation for early breast cancer. Int J Radiat Oncol Biol Phys 1989; 17(4):703-710.

167. Veronesi U, Salvadori B, Luini A, Banfi A, Zucali R, Del Vecchio M, Saccozzi R, Beretta E, Boracchi P, Farante G, . Conservative treatment of early breast cancer. Long-term results of 1232 cases treated with quadrantectomy, axillary dissection, and radiotherapy. Ann Surg 1990; 211(3):250-259.

168. Siegel BM, Mayzel KA, Love SM. Level I and II axillary dissection in the treatment of early-stage breast cancer. An analysis of 259 consecutive patients. Arch Surg 1990; 125(9):1144-1147.

169. Recht A, Pierce SM, Abner A, Vicini F, Osteen RT, Love SM, Silver B, Harris JR. Regional nodal failure after conservative surgery and radiotherapy for early-stage breast carcinoma. J Clin Oncol 1991; 9(6):988-996.

170. Halverson KJ, Taylor ME, Perez CA, Garcia DM, Myerson R, Philpott G, Levy J, Simpson JR, Tucker G, Rush C. Regional nodal management and patterns of failure following conservative surgery and radiation therapy for stage I and II breast cancer. Int J Radiat Oncol Biol Phys 1993; 26(4):593-599.

171. Chua B, Ung O, Boyages J. Competing considerations in regional nodal treatment for early breast cancer. Breast J 2002; 8(1):15-22.

172. Louis-Sylvestre C, Clough K, Asselain B, Vilcoq JR, Salmon RJ, Campana F, Fourquet A. Axillary treatment in conservative management of operable

112

breast cancer: dissection or radiotherapy? Results of a randomized study with 15 years of follow-up. J Clin Oncol 2004; 22(1):97-101.

113

APÊNDICES

Câncer Hoje 2004; 2(4): 32-41 Ann Surg Oncol 2003; 10(10): 1140-51 Ann Surg Oncol 2003; 10(10): 1128-30 (Editorial) Proc Annu Meet Am Soc Clin Oncol 2002; 21: 37A Eur J Surg Oncol 2002; 28(5): 490-500 Eur J Surg Oncol 2002; 28(6): 603-14 Eur J Surg Oncol 2002; 28(8): 899 (Erratum) J Am Coll Surg 2001; 193(5): 547-55 Proc 13th International Congress on Senology 2004; Meeting Abstract OP-032 Ann Surg Oncol 2003; 10(1): S34 Breast Cancer Res Treat 2000; 64(1): 36-7

Câncer Hoje 2(4): 32-41, 2004 32

Linfonodo sentinela em câncer de mama

José Luiz B.Bevilacqua*, * André B. Dias** e Alfredo Carlos Barros*

* Núcleo de Mastologia, Hospital Sírio Libanês, São Paulo, SP, Brasil ** Oncologia Clínica, Hospital Sírio Libanês, São Paulo, SP, Brasil

Introdução

Nas últimas duas décadas vem diminuindo consideravelmente a agressividade no manejo cirúrgico em pacientes com carcinoma de mama. Existe, na literatura, uma tendência a cirurgias mais conservadoras sem, no entanto, comprometer o controle loco-regional e sobrevida dos pacientes.1,2

O comprometimento axilar ainda é um dos fatores prognósticos mais importantes no câncer de mama.

A dissecção da axila (ou linfadenectomia axilar ou esvaziamento axilar) foi durante mais de um século o tratamento padrão no câncer de mama devido ao seu papel bem estabelecido no estadiamento, prognóstico e controle loco-regional; embora, seu impacto no aumento de sobrevida seja discutível.3 Além disso, o conhecimento do grau de envolvimento axilar é tópico fundamental na orientação do tratamento quimioterápico adjuvante.4

Tumores pequenos têm uma chance reduzida de envolvimento axilar.5,6 O número de pacientes nesta situação vem crescendo gradativamente principalmente devido ao rastreamento mamográfico que aumenta a quantidade de diagnósticos em estádio precoce.7 Neste grupo com baixo risco de doença axilar, tenta-se diminuir a morbidade do tratamento, melhorando assim a qualidade de vida. Como a chance dos linfonodos estarem comprometidos é baixa, é difícil justificar a morbidade da dissecção linfonodal clássica.

“O comprometimento linfonodal prediz a sobrevida, orienta a adjuvância e contribui para o controle loco-regional. A taxa de comprometimento linfonodal em estádios iniciais é baixa”.4

O conceito de mapeamento e identificação do linfonodo sentinela (LS) é um grande avanço na oncologia na tentativa de se resolver esta questão. O objetivo deste artigo é revisar a importância clínica deste procedimento bem como as questões técnicas, anatomopatológicas e os principais estudos de validação.

Importância clínica

O comprometimento linfonodal prediz a sobrevida, orienta a adjuvância e contribui para o controle loco-regional.4 A taxa de comprometimento linfonodal em estádios iniciais é baixa. Classicamente, o carcinoma in situ tem 0-1% de chance de comprometimento axilar, assim como T1a e T1b têm respectivamente 3-5% e 6-17% de chance de ter a axila positiva. Já nos estádios mais avançados esses números sobem para 23-48% no T2 e 29-64% no T3.5,8

A linfadenectomia axilar não é inócua, estando associada a 78% de alteração de sensibilidade na mama, 25% de dor local, 15% a 25% de linfedema e 17% de restrição de movimento.9

A proposta da biópsia do linfonodo sentinela (BLS) adiciona uma nova estratégia no manejo das pacientes com câncer de mama em estádio precoce. Esta opção visa diminuir a morbidade do procedimento, sem comprometer o estadiamento e a sobrevida da população. Desta forma a BLS está indicada para pacientes com neoplasia de mama em estádios precoces (Tis, T1 e T2) sem comprometimento axilar clínico.10-12 (Figura 1)

Definição e histórico

O conceito de sentinela é a identificação, e subseqüente ressecção do linfonodo inicial de drenagem do tumor,

Câncer Hoje 2(4): 32-41, 2004 33

análise pode predizer o status da cadeia acometida.10 Para tanto, duas pressuposições devem ser feitas: (I) a mama deve ter um padrão definido de drenagem, para que sempre um linfonodo ou um pequeno grupo receba toda a drenagem do órgão. (II) O primeiro linfonodo da cadeia deve atuar como filtro, não deixando as células neoplásicas migrarem para um outro nível linfonodal.

Historicamente, este conceito foi desenvolvido por Gould13, que descreveu o conceito anatômico, e Cabanas,14,15 que descreveu o conceito fisiológico em estudos com tumor de pênis. Em 1991, Morton16 descreveu a pesquisa de sentinela em animais e, posteriormente, em humanos portadores de melanoma. Osbourne17 aplicou este conceito nos tumores de mama. Krag,18 em 1993, desenvolveu uma das principais técnicas de mapeamento, utilizando colóide sulfurado marcado com Tc99. Em 1994, Giuliano19 descreveu a técnica com azul patente.

Figura 1

Técnica

As principais técnicas para o mapeamento do LS são: o radioisótopo, o corante azul e a técnica combinada.20,21

A técnica do azul consiste em injetar, no ato operatório, antes da incisão cutânea, um corante azul (azul patente, isossulfan azul ou azul de metileno) seguido de localização e dissecção de um vaso linfático corado até a localização do primeiro linfonodo da drenagem.

Entretanto, alguns aspectos devem ser ressaltados: (I) o tempo é fator crucial na técnica, pois caso exista um longo tempo entre a injeção e a dissecção, o corante poderá marcar vários linfonodos, ou passar por todos eles, sendo impossível identificar qual deles recebe primariamente a drenagem da mama; (II) o método não avalia linfonodos das cadeias mamária interna, supraclavicular e subclavicular; (III) a área de dissecção é maior, pois se desseca o vaso linfático para procurar o linfonodo em meio ao tecido gorduroso axilar; (IV) há uma pequena probabilidade (<1%) de reação alérgica do tipo anafilática ao corante.20,21

A técnica do radioisótopo consiste em injetar, pelo menos 2 horas antes da cirurgia, uma partícula marcada com Tc99 (colóide sulfúrico, dextran, fitato ou albumina) e encontrar o LS através de linfocintilografia associada a uma sonda de localização (do inglês probe), utilizada durante a dissecção. Esta técnica tem a vantagem de estudar drenagens anômalas como a cadeia linfonodal dos vasos mamários internos (cadeia mamária interna), e outras drenagens não usuais, como fossa supraclavicular. Algumas particularidades ocorrem com esta técnica: (I) a migração do radioisótopo pode não ocorrer devido às características da partícula marcada (geralmente maior e mais pesada que o corante azul). (II) algumas vezes é difícil localizar o sentinela devido à radiação emitida pelos tecidos adjacentes, quer seja pela área ao redor do linfonodo, quando se usa probe, como pelo sítio primário de punção, que algumas vezes pode esconder o sentinela a linfocintilografia.20,21

O método combinado é a somatória das duas técnicas, o radioisótopo e corante azul.20,21

Os estudos comparativos entre as técnicas mostram resultados semelhantes, com acurácia ao redor de 97%22 (Tabela 1). A técnica combinada é mais eficaz, pois a complementação das informações obtidas pelo emprego de duas técnicas diferentes apresenta resultados um pouco melhores.

Resultados cumulativos com biópsia de linfonodo sentinela

Método Linfonodos achados Falso negativo Acurácia

Tc99 2112/2292 (92%) 54/779 (7%) 1942/1996 (97%)

Azul Patente 714/886 (81%) 23/245 (9%) 691/717 (96%)

Combinação 1071/1155 (93%) 21/471 (5%) 1042/1063 (98%)

Total 3897/4333 (90%) 98/1441 (7%) 3675/3776 (97%)

Modificado: Cody HS, III. Breast Cancer Res 2001; 3(2):104-108.

Tabela 1: Comparação entre as técnicas de mapeamento do linfonodo sentinela

Câncer Hoje 2(4): 32-41, 2004 34

Há alguns detalhes não completamente esclarecidos em literatura. Pelo menos sete locais diferentes de punção foram descritos e não há nenhum estudo randomizado comparando em conjunto todas as técnicas (Tabela 2).

Local de injeção Estudo

Intratumoral Serin, 1996(84)

Subperiostal Terui, 1989(85)

Cavidade de biópsia Harlow, 1999(86)

Subcutâneo Veronesi, 1997(71)

Peritumoral Krag, 1998(87)

Intradérmico Lineham, 1999(88)

Subareolar Klimberg, 1999(89)

Tabela 2: Locais de punção e injeção do radioisótopo e/ou corante azul para mapeamento do linfonodo sentinela

Os locais de punção podem ser didaticamente divididos em 2 grupos: (I) intraparenquimatosos (englobando sítios intratumorais, peritumorais e na cavidade de biópsia), e (II) “subdérmicos” (referentes aos sítios subareolares, intradérmico e subcutâneo).10,23-25 Conseqüentemente algumas pequenas diferenças irão ocorrer nos resultados. Grosseiramente, os locais de punção intraparenquimatosos parecem possuir uma taxa de identificação do LS um pouco menor, no entanto, evidenciam locais menos freqüentes de drenagem como mamária interna, supra ou subclavicular. Os sítios “subdérmicos”, por sua vez, têm aparentemente maior taxa de identificação do LS, mas partem do pressuposto que a glândula e a pele possuem a mesma drenagem linfática.10,23-25

Borgstein26 tenta definir melhor esta questão em seu estudo com 217 pacientes, injetando radioisótopo peritumoral e corante azul subareolar. Foram identificados 210 linfonodos sendo que a concordância entre o azul e o radioisótopo foi de 94%. O autor conclui que a mama funciona como uma unidade linfática na maioria das vezes explicando porquê uma grande variação de técnicas pode apresentar resultados semelhantes. Atualmente ainda não existe uma preferência entre os locais de punção, sendo que qualquer um dos métodos descritos é aceitável na identificação do LS.

As tentativas de aperfeiçoamento das técnicas visam diminuir ao máximo a taxa de falso negativo, que seria a identificação de um LS negativo, quando não são dissecados outros linfonodos, num paciente com axila positiva. Quando o LS é positivo, é obrigatória a exploração axilar, embora alguns estudos mais recentes tentem selecionar pacientes de bom prognóstico para observação exclusiva nesta situação27,28.

A taxa de falso negativo varia de acordo com as maiores séries entre 0% e 12% conforme observado na tabela 3. Isso depende, obviamente, da experiência da equipe, sendo que a curva de aprendizado é de suma importância para reprodução desses resultados. Cox em seu estudo mostra que são necessários 23 casos operados para se atingir 90% de sucesso com a técnica e 53 casos para 95% de sucesso.29 Classe por sua vez, acompanhando os primeiros 100 casos de dois cirurgiões, mostra que a taxa de sucesso aumenta para 85% após 10 casos e conclui que a curva de aprendizado tem papel crucial na validação e nos resultados da BLS.30 No encontro de consenso em LS ocorrido na Filadélfia nos EUA em abril de 2001, a maioria dos participantes concordou que seria razoável realizar de 20 a 30 de procedimentos de BLS concomitante ao esvaziamento axilar para se atingir uma taxa de falso negativo e falha de detecção do linfonodo menor que 5%.31

Câncer Hoje 2(4): 32-41, 2004 35

AUTOR Nº DE PACIENTES LN POSITIVOS (%) TAXA DE FALSO NEGATIVO (%)

Giuliano19 114 37/114 (32) 5/42 (12)

Giuliano90 100 42/100 (42) 0/42 (0)

Turner67 103 43/103 (42) 1/44 (10)

Guenter91 103 28/103 (27) 3/31 (10)

Veronesi71 160 81/160 (51) 4/85 (5)

Borgstein92 104 44/104 (42) 1/45 (2)

Krag93 405 101/405 (25) 13/114 (11)

Veronesi94 371 168/371 (45) 12/180 (7)

Bass95 173 53/173 (31) 1/54 (2)

Winchester96 162 44/162 (27) 4/48 (8)

Hill97 104 47/104 (45) 5/52 (10)

Wong25 2206 641/2206 (29) 56/697 (8)

LN = Linfonodos Taxa de falso negativo = FN/FN+VP FN = Número de falsos negativos VP = Número de verdadeiros positivos

Tabela 3: Biópsia de linfonodo sentinela comparada a dissecção axilar completa. Taxa de falso negativo.

Quando não é possível identificar nenhum LS, pela falha do método de mapeamento ou por qualquer outro motivo, é recomendado que seja realizada a dissecção linfonodal axilar clássica.31,32

A técnica depende também da seleção dos pacientes. Os tumores T1 e T2 constituem os principais candidatos à exploração do sentinela.31-33 Os tumores iniciais como carcinoma ductal in situ e tipos histológicos favoráveis como medular, mucinoso, papilar e adenóide cístico têm taxa reduzida de comprometimento axilar. Os autores defendem, no entanto, a realização da BLS mesmo nos casos de baixo risco uma vez que o BLS tem baixa morbidade e o eventual achado de metástase axilar nesses tipos de tumor muda drasticamente a abordagem no tratamento adjuvante. O carcinoma ductal in situ com alto risco para microinvasão, classicamente aqueles associados a alto grau nuclear, comedonecrose, extensos, palpáveis ou aqueles com indicação de mastectomia, têm um risco de metástase axilar que chega a 10%34-36, portanto também são candidatos à realização da BLS.

Os estudos mais recentes com grande número de pacientes envolvidos (2206 pacientes) mostram que este procedimento é tão eficiente no estadiamento axilar em tumores T3 quanto em tumores menores.37 Entretanto, o número de casos de tumores grandes sem comprometimento axilar é bastante escasso5,37 o que torna o valor da BLS nestes casos ainda discutível.

A realização da técnica em pacientes que já sofreram biópsia excisional era considerada contra-indicação relativa devido à alteração da via linfática pelo ato cirúrgico. Porém estudos mais recentes mostram que a técnica pode ser realizada seguramente.25,38

Outra contra-indicação relativa é o tumor multicêntrico ou multifocal. Alguns estudos iniciais mostraram uma taxa de falso negativo muito alta. Porém estudos mais recentes mostram resultados confiáveis mesmo nessa situação.39,40

O tratamento sistêmico neo-adjuvante ou de indução é outra contra-indicação relativa para a BLS. Os estudos iniciais mostravam resultados proibitivos quando se realizava BLS após o tratamento sistêmico neo-adjuvante.41,42 Entretanto, estudos mais recentes mostram que o tratamento sistêmico neo-adjuvante não interfere na acurácia ou taxas de falso negativo da BLS43-56. Para se evitar falso negativos, alguns autores propõem que, nos casos de axila clinicamente negativas que serão submetidos à tratamento sistêmico neo-adjuvante, a BLS seja realizada previamente a este.57,58 Essa é uma conduta muito interessante, entretanto a desvantagem é a paciente ser submetida a um procedimento cirúrgico adicional.

As pacientes com comprometimento axilar clínico perfazem o único grupo com contra-indicação mais formal para a BLS. O comprometimento dos linfáticos pelo tumor leva a uma obstrução mecânica dos sinusóides causando uma mudança no fluxo linfático levando o corante a outro linfonodo que não o sentinela. Isso ressalta a importância da palpação digital do conteúdo axilar pela incisão, após a retirada do LS. Ao ser identificado um linfonodo duro, este deve também ser dissecado e, se possível, realizar o estudo anatomopatológico de congelação. A palpação é parte

Câncer Hoje 2(4): 32-41, 2004 36

integrante do procedimento de BLS.

A figura 1 contém o fluxograma do NCCN (National Comprehensive Cancer Network) que resume muito bem os critérios de indicação para BLS e abordagem da axila nos dias atuais.

A metodologia da identificação radioguiada do LS permitiu a criação de uma nova técnica para localização de lesões não palpáveis, conhecida como ROLL (Radioguided Occult Lesion Localization).59-62 Este método fundamenta-se na injeção do radiofármaco através da orientação por estereotaxia ou por ultra-sonografia nas lesões não palpáveis suspeitas e posterior biópsia radioguiada através do probe. O método tem resultados similares em relação ao agulhamento, entretanto os estudos mostram que a quantidade de material ressecado é menor,61 as lesões ficam mais centralizadas no espécime61, menor incidência de comprometimento das margens cirúrgicas60, a marcação da lesão pode ser realizada na véspera da cirurgia, facilitando o agendamento da cirurgia.60

As duas técnicas, ROLL e BLS, podem ser realizadas simultaneamente no mesmo procedimento cirúrgico, sendo chamada de SNOLL (Sentinel Node and Occult Lesion Localization) e tem evidentes vantagens práticas.63-66

Avaliação Histopatológica

Sensibilidade VPN Acurácia Falso Negativo

Com IHQ 97%(245/253) 98%(388/396) 99%(633/641 ) 3%(8/253 )

Sem IHQ 92%(885/963) 96%(1724/1802) 97%(2609/2687) 8%(78/963)

TOTAL 93%

(1130/1216) 96%(2112/2198) 97%(3242/3328) 7%(86/1216)

VPN, valor preditivo negativo *Modificado: Liberman L, Schneider L. Review of published experience. In: Cody HS, editor. Sentinel Lymph Node Biopsy. London, UK: Martin Dunitz Ltd., 2002: 285-310.

Câncer Hoje 2(4): 32-41, 2004 37

Tabela 4: Comparação entre as técnicas de avaliação histopatológica do linfonodo sentinela com e sem imuno-histoquímica (Meta-análise de 40 estudos entre 1993-1999)*

“O comprometimento dos linfáticos pelo tumor leva a uma obstrução mecânica dos sinusóides causando uma mudança no fluxo linfático levando o corante a outro linfonodo que não o sentinela”.

Aspectos anatomopatológicos

Para que os resultados da técnica sejam consistentes, não só o cirurgião e o médico nuclear devem seguir corretamente o protocolo, como o patologista deve estar em íntimo contato com a equipe cirúrgica. O processamento adequado do material é importante e altera as taxas de sensibilidade e especificidade do procedimento.

Na dissecção axilar clássica, sem a análise do LS, o exame do conglomerado de linfonodos pode deixar de examinar algum espécime comprometido visto que o processamento de grande quantidade de material resulta em uma análise mais superficial de cada linfonodo, pois, geralmente, se analisa somente um corte histológico em cada linfonodo dissecado.

A diminuição da quantidade de material, que ocorre com o LS, permite um estudo mais completo por linfonodo, aumentando a detecção de metástases e reduzindo custos operacionais.20,67,68

A taxa de falso negativo global é da ordem de 2% a 10% conforme dados da tabela 3. Esses números podem se modificar de acordo com a técnica anatomopatológica utilizada. (Tabela 5)

A análise linfonodal pode ser feita em dois momentos distintos: no intra-operatório ou no pós-operatório.

No intra-operatório, podem ser realizadas a citologia de contato (do inglês - imprint) e a avaliação histopatológica de congelação. Ambos têm como vantagem serem custo efetivo. Caso o sentinela seja positivo, permite a dissecção da axila no mesmo ato cirúrgico, poupando a doente de outra intervenção.69 Conseqüentemente, o tempo gasto com o procedimento é menor. Em contrapartida, essas técnicas cursam com alguma perda de volume do tecido linfonodal. Esta perda pode chegar a 50% do material, em alguns estudos, o que compromete a identificação de doença de pequeno volume. Devido a isso, teoricamente, incidência de falso negativo parece aumentar um pouco, chegando até 36% dependendo da série. Os autores sugerem que linfonodos pequenos, onde a perda de material pode alterar o resultado final do procedimento, devem ser estudados no pós-operatório. Os linfonodos maiores podem ser estudados tanto no intra como no pós-operatório. O falso positivo devido a artefatos, também pode ocorrer nessas circunstâncias, aumentando o número de dissecações desnecessárias.70,71 Atualmente, tanto o estudo intra-operatório quanto o pós-operatório são aceitos como padrão.69

“A melhor maneira de se atingir o resultado final primário é fazer a BLS e dissecar a axila em um mesmo tempo cirúrgico”.

A análise pós-operatória é mais completa. A inclusão do linfonodo inteiro permite cortes em diferentes níveis que comprovadamente aumentam a detecção de metástases, como exemplifica a figura 2. A adição do estudo de imuno-histoquímica, com estudo das citoqueratinas CAM 5.2, AE-1:AE-3 e MAK-6, aumenta ainda mais a taxa de linfonodos positivos20,67,68,72. Isso diminui o falso negativo para doença macrometastática73,74 (Tabela 4) mas, em contrapartida, aumenta o diagnóstico da doença micrometastática. Isso é exemplificado em vários estudos da era pré-LS.73,75 Desses, o estudo com maior casuística, examinou todos os linfonodos axilares de 921 pacientes consideradas primariamente axila negativa, fazendo cortes em seis níveis diferentes do mesmo bloco de parafina. Foram encontradas metástases em 83 pacientes (9% da amostra).76 Num segundo estudo, os mesmos autores analisaram a mesma população de pacientes descrita anteriormente e realizaram imunoistoquímica nos linfonodos que permaneceram negativos apesar dos cortes em diferentes níveis. Foi observado um acréscimo de 20% na detecção de metástases.77

A aplicação clínica do diagnóstico da doença micrometastática ainda é controversa. Dois estudos recentes mostram resultados conflitantes. No primeiro estudo, publicado no ASCO 2002, foram estudados 373 pacientes durante um

Câncer Hoje 2(4): 32-41, 2004 38

período de 20 anos.78 Dessas 289 tinham H&E negativo e imuno-negativa com sobrevida livre de doença (SLD) de 79%. Já as pacientes que tinham H&E negativo e imuno-positiva tiveram SLD 66% e as que tinham ambos marcadores positivos a SLD foi de 49%. Já um segundo estudo com 218 pacientes, 50 delas com imuno-histoquímica positiva, e seguimento de 18 anos, não apresentou diferença de sobrevida entre os dois grupos.79

Estudos de validação

A validação da técnica da BLS, sua sensibilidade e especificidade são bem conhecidas.74 Poucos são os estudos de validação clínica da BLS. Quando se procura por estudos randomizados este número reduz-se ainda mais, havendo somente um estudo randomizado publicado até hoje.80

Neste estudo, Veronesi80 randomizou 516 pacientes em dois grupos: um que realizou no mesmo ato BLS e dissecção axilar (Grupo 1), e outro que realizou BLS e a dissecção axilar nos casos de LS positivo (Grupo 2). O resultado final primário do estudo foi avaliar a capacidade da BLS predizer o status da cadeia acometida. Os resultados finais secundários foram: qualidade de vida; número de metástases axilares no Grupo 2; período livre de doença e sobrevida global.

Algumas críticas podem ser feitas a este estudo. Uma delas é referente à metodologia do estudo, conforme Krag81 argumentou em seu editorial. A melhor maneira de se atingir o resultado final primário é fazer a BLS e dissecar a axila em um mesmo tempo cirúrgico. Esta estratégia permite o estudo de todos os linfonodos podendo avaliar a capacidade do sentinela em predizer o comprometimento da axila. Fica, portanto, a dúvida da necessidade da randomização visto que todas as pacientes foram estudadas pela mesma técnica e que não se espera resultados diferentes entre os dois braços.

Quanto aos outros resultados finais, o estudo tem pequeno poder estatístico para detectar qualquer diferença de sobrevida entre os dois braços. Atualmente dois grandes estudos multicêntricos estão em andamento para resolver essa questão, o National Surgical Adjuvant Breast and Bowel Project (NSABP) protocolo B-3282 e o American College of Surgeons Oncology Group (ACOSOG) protocolo Z001183.

No presente momento, não temos estudos com poder suficiente para responder a pergunta “podemos estar comprometendo a sobrevida de nossas pacientes se optarmos por uma avaliação menos mórbida da axila?” Contudo a BLS é uma técnica cada vez mais difundida e estudada, e em breve todas essas dúvidas serão resolvidas.

Conclusões

Importância clínica: diminuir a morbidade da dissecção axilar sem comprometer o estadiamento e sobrevida das doentes.

Técnicas: corante azul, radioisótopo e a técnica combinada (preferível).

Indicações: estádios iniciais T1 e T2 considerar em carcinoma in situ (alto grau, extenso ou palpáveis) e histologias especiais (taxa reduzida de comprometimento axilar).

Contra-indicações absolutas: comprometimento clínico da axila e falha na identificação do LS.

Contra-indicações relativas: tumores grandes (T3 e T4 raro), multicêntricos, biópsia excisional prévia com cavidade de biópsia grande e tratamento sistêmico neo-adjuvante.

Análise patológica: intra-operatória (congelação, citologia de contato), pós-operatória, com múltiplos níveis de corte e imuno-histoquímica (relevância clínica ainda não definida).

Referências bibliográficas

1. Fisher B, Ravdin RG, Ausman RK, Slack NH, Moore GE, Noer RJ. Surgical adjuvant chemotherapy in cancer of the breast: results of a decade of cooperative investigation. Ann.Surg. 1968;168:337-56. 2. Fisher B, Slack NH. Number of lymph nodes examined and the prognosis of breast carcinoma. Surg.Gynecol.Obstet. 1970;131:79-88. 3. Fisher B, Jeong JH, Anderson S, Bryant J, Fisher ER, Wolmark N. Twenty-five-year follow-up of a randomized trial

Câncer Hoje 2(4): 32-41, 2004 39

4. Margolese RG, Hortobagyi GN, Buchholz TA. Neoplasms of the breast. In: Kufe DW, Pollock RE, Weichselbaum RR, Bast RC, Jr., Gansler TF, Holland JF et al. Holland-Frei Cancer Medicine, 6th Edition ed. Hamilton, London,UK: BC Decker Inc, 2003:1879-972. 5. Bevilacqua J, Cody H, III, MacDonald KA, Tan LK, Borgen PI, Van Zee KJ. A prospective validated model for predicting axillary node metastases based on 2,000 sentinel node procedures: the role of tumour location [corrected]. Eur.J.Surg.Oncol. 2002;28:490-500. 6. Hellman S, Harris JR. Natural History of Breast Cancer. In: Harris JR, Lippman ME, Morrow M, Osborne CK. Diseases of the breast, 2nd ed ed. Philadelphia: Lippincott Williams & Wilkins, 2002:407-24. 7. Nystrom L, Andersson I, Bjurstam N, Frisell J, Nordenskjold B, Rutqvist LE. Long-term effects of mammography screening: updated overview of the Swedish randomised trials. Lancet 2002;359:909-19. 8. Greco M, Agresti R, Raselli R, Giovanazzi R, Veronesi U. Axillary dissection can be avoided in selected breast cancer patients: analysis of 401 cases. Anticancer Res. 1996;16:3913-7. 9. Petrek JA, Blackwood MM. Axillary dissection: current practice and technique. Curr.Probl.Surg. 1995;32:257-323. 10. Bonnema J, van de Velde CJ. Sentinel lymph node biopsy in breast cancer. Ann.Oncol. 2002;13:1531-7. 11. Cutuli B, Velten M, Martin C. Assessment of axillary lymph node involvement in small breast cancer: analysis of 893 cases. Clin.Breast Cancer 2001;2:59-65. 12. Koizumi M, Makita M, Yoshimoto M, Kasumi F, Sakamoto G, Ogata E. Indications for sentinel lymph node biopsy in patients with breast cancer: retrospective and simulation analyses. Jpn.J.Clin.Oncol. 2002;32:517-24. 13. Gould EA, Winship T, Philbin PH, Kerr HH. Observations pf a sentinel node in cancer of the parotid. Cancer 1960;13:77-8. 14. Cabanas RM. An approach for the treatment of penile carcinoma. Cancer 1977;39:456-66. 15. Cabanas RM. Anatomy and biopsy of sentinel lymph nodes. Urol.Clin.North Am. 1992;19:267-76. 16. Morton DL, Wen DR, Wong JH, Economou JS, Cagle LA, Storm FK et al. Technical details of intraoperative lymphatic mapping for early stage melanoma. Arch.Surg. 1992;127:392-9. 17. Osborne MP, Payne JH, Richardson VJ, McCready VR, Ryman BE. The preoperative detection of axillary lymph node metastases in breast cancer by isotope imaging. Br.J.Surg. 1983;70:141-4. 18. Krag DN, Weaver DL, Alex JC, Fairbank JT. Surgical resection and radiolocalization of the sentinel lymph node in breast cancer using a gamma probe. Surg.Oncol. 1993;2:335-9. 19. Giuliano AE, Kirgan DM, Guenther JM, Morton DL. Lymphatic mapping and sentinel lymphadenectomy for breast cancer [see comments]. Ann.Surg. 1994;220:391-8. 20. Cody HS, III. Sentinel lymph node biopsy. London, UK: Martin Dunitz Ldt., 2002. 21. Morita ET, Chang J, Leong SP. Principles and controversies in lymphoscintigraphy with emphasis on breast cancer. Surg.Clin.North Am. 2000;80:1721-39. 22. Cody HS, III. Clinical aspects of sentinel node biopsy. Breast Cancer Res. 2001;3:104-8. 23. Boolbol SK, Fey JV, Borgen PI, Heerdt AS, Montgomery LL, Paglia M et al. Intradermal isotope injection: a highly accurate method of lymphatic mapping in breast carcinoma. Ann.Surg.Oncol. 2001;8:20-4. 24. Martin RC, Derossis AM, Fey J, Yeung H, Yeh SD, Akhurst T et al. Intradermal isotope injection is superior to intramammary in sentinel node biopsy for breast cancer. Surgery 2001;130:432-8. 25. Wong SL, Edwards MJ, Chao C, Tuttle TM, Noyes RD, Carlson DJ et al. The effect of prior breast biopsy method and concurrent definitive breast procedure on success and accuracy of sentinel lymph node biopsy. Ann.Surg.Oncol. 2002;9:272-7. 26. Borgstein PJ, Meijer S, Pijpers RJ, van Diest PJ. Functional lymphatic anatomy for sentinel node biopsy in breast cancer: echoes from the past and the periareolar blue method. Ann.Surg. 2000;232:81-9. 27. Bevilacqua JLB, Manasseh DM, Kattan MW, Tan LK, Cody HS, III, Borgen PI et al. A mulitivariate nomogram to predict likelihood of non-sentinel node metastases in breast cancer patients with positive sentinel nodes. (Meeting Abstract # 84). 56th Annual Cancer Symposium of Society of Surgical Oncology 2003. 28. Van Zee KJ, Manasseh DM, Bevilacqua JL, Boolbol SK, Fey JV, Tan LK et al. A nomogram for predicting the likelihood of additional nodal metastases in breast cancer patients with a positive sentinel node biopsy. Ann.Surg.Oncol. 2003;10:1140-51. 29. Cox CE, Bass SS, Boulware D, Ku NK, Berman C, Reintgen DS. Implementation of new surgical technology: outcome measures for lymphatic mapping of breast carcinoma [see comments]. Ann.Surg.Oncol. 1999;6:553-61. 30. Classe JM, Curtet C, Campion L, Rousseau C, Fiche M, Sagan C et al. Learning curve for the detection of axillary sentinel lymph node in breast cancer. Eur.J.Surg.Oncol. 2003;29:426-33. 31. Schwartz GF, Giuliano AE, Veronesi U. Proceedings of the consensus conference on the role of sentinel lymph node biopsy in carcinoma of the breast, April 19-22, 2001, Philadelphia, Pennsylvania. Cancer 2002;94:2542-51. 32. NCCN National Comprehensive Cancer Network. Practice guidelines in oncology - version 3.2003. www.nccn.org . 2003. 33. Keshtgar MR, Ell PJ. Clinical role of sentinel-lymph-node biopsy in breast cancer. Lancet Oncol. 2002;3:105-10. 34. Intra M, Veronesi P, Mazzarol G, Galimberti V, Luini A, Sacchini V et al. Axillary sentinel lymph node biopsy in patients with pure ductal carcinoma in situ of the breast. Arch.Surg. 2003;138:309-13. 35. Klauber-DeMore N, Tan LK, Liberman L, Kaptain S, Fey J, Borgen P et al. Sentinel lymph node biopsy: is it indicated in patients with high-risk ductal carcinoma-in-situ and ductal carcinoma-in-situ with microinvasion? Ann.Surg.Oncol. 2000;7:636-42. 36. McMasters KM, Chao C, Wong SL, Martin RC, III, Edwards MJ. Sentinel lymph node biopsy in patients with ductal carcinoma in situ: a proposal. Cancer 2002;95:15-20. 37. Wong SL, Chao C, Edwards MJ, Tuttle TM, Noyes RD, Carlson DJ et al. Accuracy of sentinel lymph node biopsy for patients with T2 and T3 breast cancers. Am.Surg. 2001;67:522-6. 38. Miner TJ, Shriver CD, Jaques DP, Maniscalco-Theberge ME, Krag DN. Sentinel lymph node biopsy for breast cancer: the role of previous biopsy on patient eligibility. Am.Surg. 1999;65:493-8. 39. Schrenk P, Wayand W. Sentinel-node biopsy in axillary lymph-node staging for patients with multicentric breast

Câncer Hoje 2(4): 32-41, 2004 40

cancer. Lancet 2001;357:122. 40. Tousimis E, Van Zee KJ, Fey JV, Hoque LW, Tan LK, Cody HS, III et al. The accuracy of sentinel lymph node biopsy in multicentric and multifocal invasive breast cancers. J.Am.Coll.Surg. 2003;197:529-35. 41. Fernandez A, Cortes M, Benito E, Azpeitia D, Prieto L, Moreno A et al. Gamma probe sentinel node localization and biopsy in breast cancer patients treated with a neoadjuvant chemotherapy scheme. Nucl.Med.Commun. 2001;22:361-6. 42. Nason KS, Anderson BO, Byrd DR, Dunnwald LK, Eary JF, Mankoff DA et al. Increased false negative sentinel node biopsy rates after preoperative chemotherapy for invasive breast carcinoma. Cancer 2000;89:2187-94. 43. Aihara T, Munakata S, Morino H, Takatsuka Y. Feasibility of sentinel node biopsy for breast cancer after neoadjuvant endocrine therapy: A pilot study. J.Surg.Oncol. 2004;85:77-81. 44. Bedrosian I, Reynolds C, Mick R, Callans LS, Grant CS, Donohue JH et al. Accuracy of sentinel lymph node biopsy in patients with large primary breast tumors. Cancer 2000;88:2540-5. 45. Brady EW. Sentinel lymph node mapping following neoadjuvant chemotherapy for breast cancer. Breast J. 2002;8:97-100. 46. Breslin TM, Cohen L, Sahin A, Fleming JB, Kuerer HM, Newman LA et al. Sentinel lymph node biopsy is accurate after neoadjuvant chemotherapy for breast cancer. J.Clin.Oncol. 2000;18:3480-6. 47. Cohen LF, Breslin TM, Kuerer HM, Ross MI, Hunt KK, Sahin AA. Identification and evaluation of axillary sentinel lymph nodes in patients with breast carcinoma treated with neoadjuvant chemotherapy. Am.J.Surg.Pathol. 2000;24:1266-72. 48. Julian TB, Patel N, Dusi D, Olson P, Nathan G, Jasnosz K et al. Sentinel lymph node biopsy after neoadjuvant chemotherapy for breast cancer. Am.J.Surg. 2001;182:407-10. 49. Julian TB, Dusi D, Wolmark N. Sentinel node biopsy after neoadjuvant chemotherapy for breast cancer. Am.J.Surg. 2002;184:315-7. 50. Kuerer HM, Sahin AA, Hunt KK, Newman LA, Breslin TM, Ames FC et al. Incidence and impact of documented eradication of breast cancer axillary lymph node metastases before surgery in patients treated with neoadjuvant chemotherapy. Ann.Surg. 1999;230:72-8. 51. Mamounas EP. Sentinel lymph node biopsy after neoadjuvant systemic therapy. Surg.Clin.North Am. 2003;83:931-42. 52. Miller AR, Thomason VE, Yeh IT, Alrahwan A, Sharkey FE, Stauffer J et al. Analysis of sentinel lymph node mapping with immediate pathologic review in patients receiving preoperative chemotherapy for breast carcinoma. Ann.Surg.Oncol. 2002;9:243-7. 53. Piato JR, Barros AC, Pincerato KM, Sampaio AP, Pinotti JA. Sentinel lymph node biopsy in breast cancer after neoadjuvant chemotherapy. A pilot study. Eur.J.Surg.Oncol. 2003;29:118-20. 54. Reitsamer R, Peintinger F, Rettenbacher L, Prokop E. Sentinel lymph node biopsy in breast cancer patients after neoadjuvant chemotherapy. J.Surg.Oncol. 2003;84:63-7. 55. Schwartz GF, Meltzer AJ. Accuracy of axillary sentinel lymph node biopsy following neoadjuvant (induction) chemotherapy for carcinoma of the breast. Breast J. 2003;9:374-9. 56. Stearns V, Ewing CA, Slack R, Penannen MF, Hayes DF, Tsangaris TN. Sentinel lymphadenectomy after neoadjuvant chemotherapy for breast cancer may reliably represent the axilla except for inflammatory breast cancer. Ann.Surg.Oncol. 2002;9:235-42. 57. Sabel MS, Schott AF, Kleer CG, Merajver S, Cimmino VM, Diehl KM et al. Sentinel node biopsy prior to neoadjuvant chemotherapy. Am.J.Surg. 2003;186:102-5. 58. Schrenk P, Hochreiner G, Fridrik M, Wayand W. Sentinel node biopsy performed before preoperative chemotherapy for axillary lymph node staging in breast cancer. Breast J. 2003;9:282-7. 59. Gennari R, Galimberti V, De Cicco C, Zurrida S, Zerwes F, Pigatto F et al. Use of technetium-99m-labeled colloid albumin for preoperative and intraoperative localization of nonpalpable breast lesions. J.Am.Coll.Surg. 2000;190:692-8. 60. Gray RJ, Salud C, Nguyen K, Dauway E, Friedland J, Berman C et al. Randomized prospective evaluation of a novel technique for biopsy or lumpectomy of nonpalpable breast lesions: radioactive seed versus wire localization. Ann.Surg.Oncol. 2001;8:711-5. 61. Luini A, Zurrida S, Paganelli G, Galimberti V, Sacchini V, Monti S et al. Comparison of radioguided excision with wire localization of occult breast lesions. Br.J.Surg. 1999;86:522-5. 62. Zurrida S, Galimberti V, Monti S, Luini A. Radioguided localization of occult breast lesions. Breast J. 1998;7:11-3. 63. Barros A, Cardoso MA, Sheng PY, Costa PA, Pelizon C. Radioguided localisation of non-palpable breast lesions and simultaneous sentinel lymph node mapping. Eur.J.Nucl.Med.Mol.Imaging 2002;29:1561-5. 64. De Cicco C, Trifiro G, Intra M, Marotta G, Ciprian A, Frasson A et al. Optimised nuclear medicine method for tumour marking and sentinel node detection in occult primary breast lesions. Eur.J.Nucl.Med.Mol.Imaging 2003;31:349-54. 65. Feggi L, Basaglia E, Corcione S, Querzoli P, Soliani G, Ascanelli S et al. An original approach in the diagnosis of early breast cancer: use of the same radiopharmaceutical for both non-palpable lesions and sentinel node localisation. Eur.J.Nucl.Med. 2001;28:1589-96. 66. Gray RJ, Giuliano R, Dauway EL, Cox CE, Reintgen DS. Radioguidance for nonpalpable primary lesions and sentinel lymph node(s). Am.J.Surg. 2001;182:404-6. 67. Turner RR, Ollila DW, Krasne DL, Giuliano AE. Histopathologic validation of the sentinel lymph node hypothesis for breast carcinoma. Ann.Surg. 1997;226:271-6. 68. Turner RR, Ollila DW, Stern S, Giuliano AE. Optimal histopathologic examination of the sentinel lymph node for breast carcinoma staging. Am.J.Surg.Pathol. 1999;23:263-7. 69. Turner RR, Hansen NM, Stern SL, Giuliano AE. Intraoperative examination of the sentinel lymph node for breast carcinoma staging [see comments]. Am.J.Clin.Pathol. 1999;112:627-34. 70. van Diest PJ, Torrenga H, Borgstein PJ, Pijpers R, Bleichrodt RP, Rahusen FD et al. Reliability of intraoperative

Câncer Hoje 2(4): 32-41, 2004 41

frozen section and imprint cytological investigation of sentinel lymph nodes in breast cancer. Histopathology 1999;35:14-8. 71. Veronesi U, Paganelli G, Galimberti V, Viale G, Zurrida S, Bedoni M et al. Sentinel-node biopsy to avoid axillary dissection in breast cancer with clinically negative lymph-nodes [see comments]. Lancet 1997;349:1864-7. 72. Dowlatshahi K, Fan M, Bloom KJ, Spitz DJ, Patel S, Snider HC, Jr. Occult metastases in the sentinel lymph nodes of patients with early stage breast carcinoma: A preliminary study [see comments]. Cancer 1999;86:990-6. 73. Liberman L. Pathologic analysis of sentinel lymph nodes in breast carcinoma [editorial; comment]. Cancer 2000;88:971-7. 74. Liberman L, Schneider L. Review of published experience. In: Cody HS. Sentinel Lymph Node Biopsy. London, UK: Martin Dunitz Ltd., 2002:285-310. 75. Dowlatshahi K, Fan M, Snider HC, Habib FA. Lymph node micrometastases from breast carcinoma: reviewing the dilemma. Cancer 1997;80:1188-97. 76. Prognostic importance of occult axillary lymph node micrometastases from breast cancers. International (Ludwig) Breast Cancer Study Group [see comments]. Lancet 1990;335:1565-8. 77. Cote RJ, Peterson HF, Chaiwun B, Gelber RD, Goldhirsch A, Castiglione-Gertsch M et al. Role of immunohistochemical detection of lymph-node metastases in management of breast cancer. International Breast Cancer Study Group [see comments]. Lancet 1999;354:896-900. 78. Tan LK, Giri D, Pangeas K, Cranor M, Bevilacqua JL, Hummer A et al. Occult micrometastases in axillary lymph nodes of breast cancer patients are significant: a retrospective study with long term follow up. (Meeting Abstract #146). Proc.Annu.Meet.Am.Soc.Clin.Oncol. 2002;21:37A. 79. de Mascarel I, MacGrogan G, Picot V, Mathoulin-Pelissier S. Prognostic significance of immunohistochemically detected breast cancer node metastases in 218 patients. Br.J.Cancer 2002;87:70-4. 80. Veronesi U, Paganelli G, Viale G, Luini A, Zurrida S, Galimberti V et al. A randomized comparison of sentinel-node biopsy with routine axillary dissection in breast cancer. N.Engl.J.Med. 2003;349:546-53. 81. Krag D, Ashikaga T. The design of trials comparing sentinel-node surgery and axillary resection. N.Engl.J.Med. 2003;349:603-5. 82. Krag DN, Julian T. Expert perspectives: update on NSABP-32, a randomized phase III clinical trial to compare sentinel node resection to conventional axillary dissection for clinically node-negative breast cancer. Breast Dis.Year Book Q. 2002;13:113-4. 83. American College of Surgeons Oncology Group (ACOSOG). A randomized trial of axillary node dissection in women with clinical T1 - 2 N0 M0 breast cancer who have a positive sentinel node (Protocol Z0011). http://www.acosog.org/studies/synopses/Z0011_Synopsis.pdf . 2004. 84. Serin D, Vinot JM, Martin P, Reboul F, Robin M, Chevallier J. [The value of breast lymphoscintigraphy in the definition of axillary staging in cancer of the breast]. Bull.Cancer 1986;73:299-304. 85. Terui S, Yamamoto H. New simplified lymphoscintigraphic technique in patients with breast cancer. J.Nucl.Med. 1989;30:1198-204. 86. Harlow S, Krag D, Weaver D, Ashikaga T. Extra-Axillary Sentinel Lymph Nodes in Breast Cancer. Breast Cancer 1999;6:159-65. 87. Krag D, Weaver D, Ashikaga T, Moffat F, Klimberg VS, Shriver C et al. The sentinel node in breast cancer--a multicenter validation study [see comments]. N.Engl.J.Med. 1998;339:941-6. 88. Linehan DC, Hill AD, Akhurst T, Yeung H, Yeh SD, Tran KN et al. Intradermal radiocolloid and intraparenchymal blue dye injection optimize sentinel node identification in breast cancer patients [see comments]. Ann.Surg.Oncol. 1999;6:450-4. 89. Klimberg VS, Rubio IT, Henry R, Cowan C, Colvert M, Korourian S. Subareolar versus peritumoral injection for location of the sentinel lymph node [see comments]. Ann.Surg. 1999;229:860-4. 90. Giuliano AE, Jones RC, Brennan M, Statman R. Sentinel lymphadenectomy in breast cancer. J.Clin.Oncol. 1997;15:2345-50. 91. Guenther JM, Collins JC, Barnes G, Jr., O’Connell TX. Selective lymphoscintigraphy: a necessary adjunct to dye-directed sentinel node biopsy for breast cancer? Arch.Surg. 2000;135:1101-5. 92. Borgstein PJ, Pijpers R, Comans EF, van Diest PJ, Boom RP, Meijer S. Sentinel lymph node biopsy in breast cancer: guidelines and pitfalls of lymphoscintigraphy and gamma probe detection. J.Am.Coll.Surg. 1998;186:275-83. 93. Krag DN, Ashikaga T, Harlow SP, Weaver DL. Development of sentinel node targeting technique in breast cancer patients. Breast J. 1998;4:67-74. 94. Veronesi U, Paganelli G, Viale G, Galimberti V, Luini A, Zurrida S et al. Sentinel lymph node biopsy and axillary dissection in breast cancer: results in a large series [see comments]. J.Natl.Cancer Inst. 1999;91:368-73. 95. Bass SS, Cox CE, Ku NN, Berman C, Reintgen DS. The role of sentinel lymph node biopsy in breast cancer. J.Am.Coll.Surg. 1999;189:183-94. 96. Winchester DJ, Sener SF, Winchester DP, Perlman RM, Goldschmidt RA, Motykie G et al. Sentinel lymphadenectomy for breast cancer: experience with 180 consecutive patients: efficacy of filtered technetium 99m sulphur colloid with overnight migration time. J.Am.Coll.Surg. 1999;188:597-603. 97. Hill AD, Tran KN, Akhurst T, Yeung H, Yeh SD, Rosen PP et al. Lessons learned from 500 cases of lymphatic mapping for breast cancer. Ann.Surg. 1999;229:528-35.

A Nomogram for Predicting the Likelihood of Additional NodalMetastases in Breast Cancer Patients With a Positive Sentinel

Node Biopsy

Kimberly J. Van Zee, MS, MD, Donna-Marie E. Manasseh, MD, Jose L. B. Bevilacqua, MD,Susan K. Boolbol, MD, Jane V. Fey, MPH, Lee K. Tan, MD, Patrick I. Borgen, MD,

Hiram S. Cody III, MD, and Michael W. Kattan, PhD

Background: The standard of care for breast cancer patients with sentinel lymph node (SLN)metastases includes complete axillary lymph node dissection (ALND). However, many question theneed for complete ALND in every patient with detectable SLN metastases, particularly thoseperceived to have a low risk of non-SLN metastases. Accurate estimates of the likelihood ofadditional disease in the axilla could assist greatly in decision-making regarding further treatment.

Methods: Pathological features of the primary tumor and SLN metastases of 702 patients whounderwent complete ALND were assessed with multivariable logistic regression to predict thepresence of additional disease in the non-SLNs of these patients. A nomogram was created usingpathological size, tumor type and nuclear grade, lymphovascular invasion, multifocality, andestrogen-receptor status of the primary tumor; method of detection of SLN metastases; number ofpositive SLNs; and number of negative SLNs. The model was subsequently applied prospectivelyto 373 patients.

Results: The nomogram for the retrospective population was accurate and discriminating, with anarea under the receiver operating characteristic (ROC) curve of 0.76. When applied to the prospec-tive group, the model accurately predicted likelihood of non-SLN disease (ROC, 0.77).

Conclusions: We have developed a user-friendly nomogram that uses information commonlyavailable to the surgeon to easily and accurately calculate the likelihood of having additional,non-SLN metastases for an individual patient.

Key Words: Axillary metastases—Breast cancer—Nomogram—Prediction—Sentinel node.

The sentinel lymph node (SLN) biopsy procedure hasbeen validated by numerous studies1–6 and found to be

accurate for assessing regional lymph node involvement.For those whose SLN biopsy specimen is histopatholog-ically negative, the risk of “missed” axillary disease isextremely low.7,8 Therefore, SLN biopsy alone, withoutcomplete axillary lymph node dissection (ALND), hasbeen adopted at many institutions as an accurate methodof staging the axilla while avoiding much of the morbid-ity associated with a complete ALND. Although thestandard of care for breast cancer patients with SLNmetastases remains performance of complete ALND, manyquestion the need for complete ALND in every patient withdetectable SLN metastases, particularly those in whom theperceived risk of additional disease is low.9,10

Proponents of performance of completion ALND aftera positive SLN biopsy argue that further axillary clear-ance is critical to management. The total number of

Received March 6, 2003; accepted August 19, 2003.From the Departments of Surgery (KJVZ, DMEM, JLBB, SKB,

JVF, PIB, HSC), Pathology (LKT), and Urology, Epidemiology andBiostatistics (MWK), Memorial Sloan-Kettering Cancer Center, NewYork, New York.

Drs. Manasseh and Bevilacqua contributed equally to the work.Dr. Bevilacqua is currently affiliated with Hospital Sírio Libanes,

Instituto Brasileiro de Controle do Câncer, and Disciplina de CirurgiaGeral, Departamento de Cirurgia, Faculdade de Medicina da Univer-didade de Sao Paulo. São Paulo, Brazil; Dr. Boolbol is currentlyaffiliated with Beth Israel Medical Center, New York, New York.

Address correspondence and reprint requests to: Kimberly J. VanZee, MS, MD, FACS, 1275 York Ave., MRI-1026, New York, NY10021; Fax: 212-794–5812; E-mail: [email protected].

Published by Lippincott Williams & Wilkins © 2003 The Society of SurgicalOncology, Inc.

Annals of Surgical Oncology, 10(10):1140–1151

DOI: 10.1245/ASO.2003.03.015

1140

involved nodes is important prognostic information, asan increasing number of positive nodes portends worsesurvival.11–13 This is reflected in the new American JointCommittee on Cancer (AJCC) sixth edition staging sys-tem, wherein the number of positive nodes defines N1,N2, and N3 disease and ultimately the stage to which thepatient is assigned.14 In addition, proponents of the per-formance of complete ALND after positive SLN biopsyargue that the additional information can benefit patientsby guiding decisions about adjuvant chemotherapy. Forthe approximately one-half of patients in whom there isresidual nodal disease, it is also argued that completeALND can influence survival via local-regional controlof the axilla,15–17 thereby eliminating a potential site ofrecurrent disease and, ultimately, a source for distantdisease. A meta-analysis of randomized trials revealed a5.4% survival benefit associated with ALND for clini-cally node-negative patients.18

Opponents of complete ALND performed after posi-tive SLN biopsy argue that the therapeutic benefit ofcomplete ALND is minimal.19 Furthermore, approxi-mately 50% of patients with positive SLNs are found tohave no other nodal metastases.1–6,20–24 Therefore, manypatients are undergoing unnecessary ALND, with noadditional therapeutic benefit or further staging informa-tion provided. It is also argued that because patients withSLN metastases will generally receive systemic therapy,regardless of the presence of any additional nodal me-tastases, any residual disease does not influence choiceof therapy and may itself be eradicated by the systemictherapy. In addition, radiation therapy after breast-con-serving surgery may contribute to control of any addi-tional nodal disease. It is this debate that physicians and

their patients are faced with in the office setting when apositive SLN is discovered on final pathology.

Several groups have identified histopathological vari-ables of the primary tumor and its metastasis that caninfluence risk of additional disease in the non-SLNs9,10,21–29 (Table 1). Size of the primary tumor andsize of the SLN metastasis are the two variables mostcommonly analyzed. Previous investigation at our insti-tution found that size of the primary tumor and of theSLN metastasis are significantly predictive of likelihoodof additional, non-SLN metastases.26 Most other studiesthat examined at least one of these two variables showeda statistically significant correlation with risk of addi-tional, non-SLN disease,9,10,21–25,27–29 yet most have notbeen able to identify a subset that has no risk of addi-tional disease in the non-SLNs.10,22–24,27,28,30 Thosestudies that have identified favorable subsets with anapparently negligible risk of additional nodal diseasehave all involved very small subsets (i.e., 5–24patients).9,21,26,30 –32

Furthermore, it is difficult to estimate the risk ofadditional, non-SLN metastases for an individual patientby using the literature. First, the estimates of risk for anygiven characteristic vary considerably among studies.Tables 2 through 5 show reported incidences by primarytumor size, presence of lymphovascular invasion (LVI)in the primary tumor size of SLN metastasis, and pres-ence of immunohistochemically (IHC-) detected SLNmetastases. This variation in reported incidences may beattributable to relatively small sample sizes or the resultof differences among the study populations in terms ofother variables influencing risk. Second, it is difficult toapply risk estimates to several patient characteristics

TABLE 1. Studies reporting predictors of non-SLN metastases in patients with a positive SLN biopsy; values listed are Pvalues reported in study

Study

nSize(cm)

Primary tumor characteristics

Estrogenreceptorstatus

Method ofdetection

SLN charactertistics

Author YearNucleargrade

Lymphovascularinvasion Multifocality

Size ofSLN met

No. ofpositive

SLN

No. ofnegative SLN

examined

aChu9 1999 157 .014 NS NS — NS NS �.0001 NS —aReynolds21 1999 60 .0004 NS NS — .03 — .002 — —Teng22 2000 26 .001 — — — — .02 — — —aTurner25 2000 194 .03 NS .03 — NS — .01 NS —Abdessalam23 2001 100 NS NS .004 — NS — .01 NS —Kamath10 2001 101 .005 — — — — — .001 — —Rahusen24 2001 93 NS NS NS — NS — .05 .002 —aViale28 2001 109 NS NS NS — NS — .02 — —aWeiser26 2001 206 .007 NS NS — — — .0002 — —aWong27 2001 389 �.001 — — — — — — �.001 —aSachdev29 2002 55 .0001 — .001 — — — .02 — —

NS, not statistically significant; —, not included in analysis. n, number of patients with positive SLN biopsy and complete axillary dissection; SLN,sentinel lymph node.

a Denotes studies reporting P values for multivariate analysis.

1141PREDICTING NON-SLN METASTASES AFTER POSITIVE SLNB

Ann Surg Oncol, Vol. 10, No. 10, 2003

simultaneously because of the generally univariatemethod of reporting in the literature.

In an attempt to achieve a more precise prediction forthe individual patient than is readily available by usingthese published estimates of risk, we used a multivariablelogistic-regression analysis of a large data set to modelthe association between selected variables and the like-lihood of metastases in non-SLNs in patients with apositive SLN biopsy. We examined pathological size ofthe primary tumor, the method of detection of the SLNmetastasis, and several other variables that are readilyavailable and theoretically related to risk of additionalnodal disease. We used 702 cases from our large pro-spective sentinel lymph node database to develop themodel, and we developed a user-friendly nomogram topredict the likelihood of finding additional positivenodes at completion ALND. We then tested our modelby prospectively applying it in an additional study groupcomprising 373 patients.

The goal of our study was to develop a tool that wouldallow greater individualization of a patient’s risk esti-mate by simultaneously taking into account several per-tinent characteristics specific to the patient. With a moreprecise and individualized estimate, both physician andpatient would be better able to weigh the pros and consof further axillary dissection.

MATERIALS AND METHODS

Between September 12, 1996, and September 20,2002, 4790 consecutive cases of SLN biopsy at Memo-rial Sloan-Kettering Cancer Center (MSKCC) were en-tered prospectively into the MSKCC Breast Cancer Sen-tinel Lymph Node Database. Our study populationinvolved the subset of 1075 cases that fulfilled the fol-lowing criteria: primary invasive breast carcinoma withclinically negative axilla and no prior systemic treat-ment; successful SLN biopsy in which metastatic diseasewas identified; and completion ALND with at least 10nodes examined. A total of 140 cases were excludedbecause a completion ALND was not performed. Theoverall study population comprised the patients meetingthe selection criteria, who were then divided into twogroups: a retrospective group who had undergone SLNbiopsy between September 12, 1996, and April 24, 2001,and a prospective group undergoing SLN biopsy be-tween April 25, 2001, and September 20, 2002. Thisproject was reviewed and approved by the MSKCCInstitutional Review Board.

TABLE 3. Studies reporting incidence of non-SLNmetastases in axillae with positive SLN(s), by presence of

LVI in primary tumor

Author No LVI LVI

Reynolds21 43% 62%Turner25 37% 65%Abdessalem23 31% 62%Rahusen24 42% 30%Weiser26 26% 41%Viale28 21% 26%Sachdev29 12% 32%

LVI, lymphovascular invasion; SLN, sentinel lymph node.

TABLE 4. Studies reporting incidence of non-SLNmetastases in axillae with positive SLN(s), by size of

SLN metastasis

Author �1 mm �1 mm �2 mm �2 mm�2 cm or

“gross disease”

Chu9 — — 7% 55% —Reynolds21 — — 22% 67% —Turner25 — — 26% 63% —Abdessalem23 — — 20% 47% 75%Kamath10a — — 15% 58% 65%Rahusen24 27% 50% — — —Viale28 16% — 22% 45% —Weiser26 — — 18% 45% —Mignotte30a — — 22% 79% —Sachdev29 17% 49% — — —

SLN, sentinel lymph node.a Categories used: �2 mm and �2 mm.

TABLE 2. Studies reporting incidence of non-SLN metastases in axillae with positive SLN(s), by primary tumor size

Author T1a (�.5 cm) T1b (.6–1.0 cm) T1c (1.1–2.0 cm) T2 (2.1–5.0 cm) T3 (�5.0 cm)

Chu9 0% 13% 29% 38% 71%Reynolds21 [ 25% for T1 ] [ 79% for T2/3 ]Turner25 17% 20% 46% 48% 73%Kamath10 25% 30% 40% 46% 80%Rahusen24 50% 50% 49% 50% —Weiser26 8% 21% 37% 48% —Wong27 14% 22% 30% 45% 57%Viale28 100% 14% 25% 24% —Sachdev29 [ 13% for T1 ] [ 33% for �T2 ]Mignotte30 [ 14% ] 54% 52% —

SLN, sentinel lymph node.

1142 K. J. VAN ZEE ET AL.


Our technique for SLN biopsy includes the use of bothblue dye and radioisotope, as previously outlined inearlier studies at our institution.33

SLN Histopathological EvaluationWhenever possible, the SLN was bisected and sec-

tioned at 2–3-mm intervals. The nodal tissue was quick-frozen in liquid nitrogen, and a single 5-�m-thick sectionstained with hematoxylin and eosin (H&E) was exam-ined intraoperatively (frozen-section analysis). If the sec-tion was positive, a complete ALND was done immedi-ately. After the frozen-section analysis, the remainingfrozen tissue was fixed in formalin and embedded inparaffin. Another 5-�m-thick H&E-stained section wasevaluated as a frozen-section control (routine histopa-thology). If this section showed evidence of metastaticdisease, no further pathological workup of the SLN wasperformed. If the routine H&E section remained nega-tive, enhanced pathological analysis was performed inthe following fashion: two pairs of H&E- and cytokeratinIHC–stained sections with a distance of 50 �m betweenthe pairs were prepared from the paraffin block. At onelevel, the cytokeratin antibody CAM 5.2 (Becton Dick-inson Immunocytometry Systems, San Jose, CA) wasused, whereas the cytokeratin cocktail AE1:AE3 (Ven-tana Medical Systems, Tucson, AZ) was applied for theother level. Patients with SLN metastases not detected byfrozen-section analysis generally underwent completionALND at a later date. For all additional nodes identifiedby completion ALND, routine H&E analysis was doneon a single section of each node.

Data AnalysisClinical data collected for each case from the database

included age; pathological size of the invasive carci-noma, defined in centimeters; tumor type (ductal orlobular carcinoma); nuclear grade (I: slight or no varia-tion in size and shape of nucleus; II: moderate variationin size and shape; III: marked variation in size andshape); presence of lymphovascular invasion (presenceof one or more tumor cells in a lymphatic or vascularstructure); multifocality of primary tumor (foci of carci-

noma separate from primary tumor); estrogen-receptor(ER) status (negative, �10% of cells staining positive);method of detection of SLN metastases (frozen-sectionanalysis [frozen], routine histopathology [routine], H&Estains of serial sections [serial HE], IHC); number of

TABLE 5. Studies reporting incidence of non-SLNmetastases in patients with IHC-detected SLN metastases

Author Proportion %

Teng22 3/26 12Kamath10 2/26 8Wong27 3/28 11Mignotte30 7/44 16Jakub31 9/62 15

SLN, sentinel lymph node; IHC, immunohistochemistry.

TABLE 6. Descriptive characteristics of the retrospective(9/12/1996–4/24/2001) and prospective

(4/25/2001–9/20/2002) patient populations

Retrospective(n � 702)

Prospective(n � 373)

n % n %

Age�50 290 41.3 157 42.1�50 412 58.7 216 57.9

Pathologic size (cm)�.5 33 4.7 13 3.5.6–1.0 122 17.4 49 13.11.1–2.0 312 44.4 166 44.52.1–3.0 154 21.9 93 24.93.1–5.0 65 9.3 41 11.0�5.1 16 2.3 11 2.9

Tumor type and nuclear gradeDuctal, I 22 3.1 11 2.9Ductal, II 321 45.7 175 46.9Ductal, III 275 39.2 129 34.6Lobular 84 12.0 58 15.5

Lymphovascular invasionNo 418 59.5 219 58.7Yes 284 40.5 154 41.3

MultifocalNo 505 71.9 241 64.6Yes 197 28.1 132 35.4

Estrogen-receptor statusNegative 135 19.2 83 22.3Positive 567 80.8 290 77.7

Method of detectionIHC only 63 9.0 18 4.8Serial H&E 78 11.1 40 10.7Routine H&E 65 9.3 23 6.2Frozen 463 66.0 273 73.2Frozen not done 33 4.7 19 5.1

No. of positive SLN1 488 69.5 265 712 161 22.9 75 20.13 35 5.0 21 5.64 12 1.7 8 2.15 3 .4 3 .86 1 .1 0 07 2 .3 0 0�8 0 0 1 .3

No. of negative SLN0 271 38.6 132 35.41 183 26.1 79 21.22 102 14.5 72 19.33 68 9.7 41 11.04 34 4.8 22 5.95 16 2.3 7 1.96 6 .9 10 2.77 8 1.1 2 .5�8 14 2.0 8 2.1

SLN, sentinel lymph node; IHC, immunohistochemistry; H&E, he-matoxylin and eosin.



positive SLNs; and number of negative SLNs. Becauselobular carcinomas generally are not assigned a nucleargrade, the tumor type and nuclear grade were combinedinto the following four categories: ductal carcinoma,nuclear grade I; ductal carcinoma, nuclear grade II; duc-tal carcinoma, nuclear grade III; and lobular carcinoma.To allow use of our model by groups that do not rou-

tinely perform frozen-section analysis, a second modelwas developed with only three levels for the method ofdetection variable: routine histopathology (routine), se-rial sectioning (serial HE), and immunohistochemistry(IHC). In this model, a node in which metastatic diseasewas detected by either frozen-section analysis or routinehistopathology was categorized as routine. Data on ad-

TABLE 7. Incidence of additional, non-SLN metastases for retrospective and prospective patient populations, by primary andSLN pathologic characteristics

Retrospective (n � 702) Prospective (n � 373)Entire population

(n � 1075)

Proportion % Proportion % Proportion %

Age�50 114/290 39 64/157 41 178/447 40�50 150/412 36 90/216 42 240/628 38

Pathologic size (cm)�.5 8/33 24 1/13 8 9/46 20.6–1.0 32/122 26 13/49 26 45/171 261.1–2.0 111/312 36 60/166 36 171/478 362.1–3.0 62/154 40 44/93 47 106/247 433.1–5.0 37/65 57 26/41 63 63/106 59�5.1 14/16 88 10/11 91 24/27 89

Tumor type and nuclear gradeDuctal, I 6/22 27 3/11 27 9/33 27Ductal, II 100/321 31 57/175 33 157/496 32Ductal, III 121/275 44 67/129 52 188/404 47Lobular 37/84 44 27/58 47 64/142 45

Lymphovascular invasionNo 125/418 30 71/219 32 196/637 31Yes 139/284 49 83/154 54 222/438 51

MultifocalityNo 176/505 35 89/241 37 265/746 36Yes 88/197 45 65/132 49 153/329 47

Estrogen-receptor statusNegative 50/135 37 37/83 45 87/218 40Positive 214/567 38 117/290 40 331/857 39

Method of detectionIHC only 6/63 10 4/18 22 10/81 12Serial H&E 12/78 15 3/40 8 15/118 13Routine H&E 13/65 20 6/23 26 19/88 22Frozen 221/463 48 128/273 47 349/736 47Frozen not done 12/33 36 13/19 68 25/52 48

No. of positive SLN1 155/488 32 97/265 37 252/753 332 72/161 45 34/75 45 106/236 453 23/35 66 15/21 71 38/56 684 9/12 75 5/8 62 14/20 705 2/3 67 2/3 67 4/6 676 1/1 100 0/0 — 1/1 1007 2/2 100 0/0 — 2/2 100�8 0/0 — 1/1 100 1/1 100

No. of negative SLN0 134/271 49 83/132 63 217/403 541 67/183 37 25/79 32 92/262 352 27/102 26 18/72 25 45/174 263 18/68 26 14/41 34 32/109 294 11/34 32 10/22 45 21/56 385 3/16 19 0/7 0 3/23 136 1/6 17 3/10 30 4/16 257 2/8 25 1/2 50 3/10 30�8 1/14 7 0/8 0 1/22 5

SLN, sentinal lymph node; IHC, immunohistochemistry; H&E, hematoxylin and eosin.



ditional variables such as progesterone-receptor status,histologic grade, and AJCC T stage were also collected;however, because these variables are highly correlatedwith ER status, nuclear grade, and pathological size,respectively, they were not considered to be of substan-tial benefit to the model. HER-2/neu amplification datawere also collected but were not included because theywere incomplete and variable, owing to evolving meth-ods of assessment during the years of the study.

A nomogram was developed based on the patients inthe retrospective group and then was validated with thepatients in the prospective group. In the retrospectivepopulation (n � 702), multivariable logistic regressionwas used to analyze the association of each variable withthe likelihood of non-SLN metastases, and a nomogramwas created with all variables. This model was used inthe prospective group (n � 373) to predict each individ-ual patient’s probability of having positive non-SLNs.The discrimination of the model was measured by usingthe area under the receiver operating characteristic(ROC) curve. The calibration of the model was assessedgraphically. Women were grouped into deciles based ontheir nomogram predictions. For each decile, the meannomogram-predicted probability was compared with theproportion of women who actually had positive non-SLNs (actual probability). All analyses were performedwith S-Plus Software Version 2000 Professional Editionwith the Design Library (Mathsoft, Data Analysis Prod-ucts Division, Seattle, WA).34

RESULTS

Descriptive characteristics of the study population arelisted in Table 6. Table 7 shows the incidence of addi-tional, non-SLN metastases for retrospective, prospec-tive, and total patient populations by primary and SLNpathological characteristics. On multivariable logistic-regression analysis, pathological size, lymphovascularinvasion, method of detection, number of positive SLNs,and number of negative SLNs were each associated withthe likelihood of additional, non-SLN metastases (P �.05 for each). Multifocality was of borderline signifi-cance, and neither tumor type and nuclear grade nor ERstatus had a statistically significant association with thelikelihood of non-SLN metastases (Tables 8 and 9). Agewas not included in the final nomograms because itseffect was too small to be seen on the nomograms.

A nomogram based on this model and developedin the retrospective population (n � 702) appears inFigure 1. The overall predictive accuracy of a modelincorporating the eight variables, as measured by thebootstrap corrected ROC curve, was 0.76. To address the

calibration accuracy of the nomogram (i.e., the absoluteerror of its prediction), we conducted additional boot-strapping and plotted the probabilities predicted by thenomogram against the corresponding observed propor-tions in the prospective population (n � 373) (Fig. 2).The area under the ROC curve for the model applied tothe prospective population is 0.77. For those users whodo not routinely perform frozen-section analysis on theSLN, a separate analysis without frozen-section informa-tion was performed and illustrated in the nomogram inFigure 3. The ROC of this version of the nomogram is0.75 in the retrospective population and 0.78 in the

TABLE 9. Results of multivariable logistic-regressionanalysis testing the relationship between primary tumor and

SLN characteristics and the incidence of non-SLN metastasesa

Variables P value

Pathology size .0006Tumor type and nuclear grade .4

Ductal, nuclear grade I vs. II .8Ductal, nuclear grade I vs. III .4Ductal, nuclear grade I vs. lobular .6

Lymphovascular invasion .003Multifocality .02Estrogen-receptor status .16Method of detection (no frozen section

available)�.0001

Routine vs. IHC .0001Routine vs. serial H&E �.001

No. of positive SLN .0001No. of negative SLN �.0001

H&E, hematoxylin and eosin; IHC, immunohistochemistry; SLN,sentinel lymph node.

aThis model is for use when no frozen section data is available.Method of detection has three values: routine H&E, serial H&E and

IHC.

TABLE 8. Results of multivariable logistic-regressionanalysis testing the relationship between primary tumor and

SLN characteristics and the incidence of additional, non-SLN metastases

Variables P value

Pathology size .001Tumor type and nuclear grade .7

Ductal, nuclear grade I vs. II 1.0Ductal, nuclear grade I vs. III .7Ductal, nuclear grade I vs. lobular .8

Lymphovascular invasion .003Multifocality .06Estrogen-receptor status .08Method of detection �.001

Frozen vs. IHC �.001Frozen vs. serial HE �.001Frozen vs. routine �.001

No. of positive SLN �.001No. of negative SLN �.001

HE, hematoxylin and eosin; IHC, immunohistochemistry; SLN, sen-tinel lymph node.



prospective population; the corresponding calibrationcurve is depicted in Figure 4.

Using the NomogramEach version of the nomogram consists of 11 rows.

The first row (POINTS) is the point assignment for eachvariable. Rows 2 through 9 represent the variables in-cluded in the model. For an individual patient, eachvariable is assigned a point value (uppermost scale[POINTS]) based on the histopathological characteris-tics. To determine the point assignment, a vertical line ismade between the appropriate variable value and thePOINTS line. For example, a pathological size of 1 cm(PATHSIZE, 2) confers about 10 points.

The assigned points for all eight variables aresummed, and the total is found in row 10 (TOTALPOINTS). Once the total is located in row 10 (TOTALPOINTS), a vertical line is made between it and thecorresponding value in the final row, row 11 (PredictedProbability of �non-SLN). The version of the nomo-gram in Figure 1 is for use when information on frozen-

section analysis is available; that in Figure 3 is for thosecases where frozen-section information is not available.

In addition to the graphic nomograms, to facilitateease of use in the clinical setting, we have made apersonal digital assistant (PDA)–compatible applicationfor use with hand-held Palm-type devices (Palm, Milpi-tas, CA). We will make these applications available atour Web site, www.mskcc.org/nomograms.

DISCUSSION

With the adoption of SLN biopsy, a new clinicalconundrum has become commonplace: should a comple-tion ALND be done for a patient with a positive SLNbiopsy? This question is particularly difficult with regardto patients with micrometastatic disease, disease whichwas undetectable in the era prior to SLN biopsy. Otherinvestigators have attempted to address this question andhave identified risk factors for the presence of additional,non-SLN disease, but all such attempts are limited by thepractical difficulty of simultaneously including several

FIG. 1. Nomogram to predict likelihood of additional, non–sentinel lymph node (non-SLN) metastases in a patient with a positive SLN.NUCGRADE, tumor type and nuclear grade (ductal, nuclear grade I; ductal, nuclear grade II; ductal, nuclear grade III; lobular); LVI, lymphovascularinvasion; MULTIFOCAL, multifocality of primary tumor; ER, estrogen-receptor status; NUMNEGSLN, number of negative SLNs; NUMSLNPOS,number of positive SLNs; PATHSIZE, pathological size, defined in centimeters; and METHDETECT, method of detection of SLN metastases (frozen,routine H&E, serial HE, IHC). The first row (POINTS) is the point assignment for each variable. Rows 2–9 represent the variables included in themodel. For an individual patient, each variable is assigned a point value (uppermost scale, POINTS) based on the histopathological characteristics.A vertical line is made between the appropriate variable value and the POINTS line. The assigned points for all eight variables are summed, and thetotal is found in row 10 (TOTAL POINTS). Once the total is located, a vertical line is made between TOTAL POINTS and the final row, Row 11(Predicted Probability of �non-SLN).



variables in the risk estimate. Here, simultaneously usingseveral variables in a large population, we have devel-oped nomograms to predict the likelihood of additionalnodal metastases after a positive SLN biopsy. We haveprospectively tested them, demonstrating that they per-form well in the prospective population.

Our nomograms utilize readily available clinical infor-mation and allow quick calculation. This approach mayallow identification of extremely low-risk individuals forwhom the risks associated with completion ALND arejudged to outweigh the benefits. Conversely, our nomo-gram may allow identification of women at sufficientrisk of additional nodal disease that they and their sur-geon elect to proceed with completion ALND eventhough clinical “guesstimates” would suggest that theyare at low risk.

The nomograms provide risk estimates that will haveto be judged on an individual basis. A woman with a1.8-cm, ER-positive, high-nuclear-grade ductal carci-

noma with no LVI who has a single IHC-positive SLNmight be considered to be at low risk. Our nomogramsuggests that she has a 12% risk of having non-SLNmetastases. Should she undergo completion ALND?Given this scenario, some will judge that a 12% risk ofadditional, non-SLN metastases justifies further ALND;others will not. The nomogram itself makes no actualtreatment recommendations.

There are several limitations to our model. The nodesretrieved at completion ALND were examined by routinepathological analysis only. Other investigators25,35 haveshown that if non-SLNs are examined with serial sec-tioning and IHC, a higher proportion of patients withadditional, non-SLN disease at completion ALND areidentified. Evaluation by enhanced pathological analysiswould clearly alter our model.

Furthermore, the clinical relevance of resecting addi-tional nodal disease (even that detected by routine anal-ysis) remains unknown. Although some argue that sur-

FIG. 2. Calibration plot for nomogram with frozen section information. The nomogram developed with use of the retrospective group of patients(n � 702) was applied to the prospective group (n � 373). A histogram of the calculated probabilities for the prospective population is shown alongthe horizontal axis. These 373 patients are grouped in deciles of their predicted probabilities, and the actual incidence of additional, non-SLNmetastases was calculated for each decile. The vertical axis represents the actual, observed incidence (Actual Probability), and the horizontal axisrepresents the probability calculated by the nomogram (Predicted Probability). For each decile of the prospective group, a triangle is plotted to showactual probability. If the model were perfect, all triangles would lie on the dotted line, with a slope of 1.



gical removal of subclinical nodal disease is associatedwith a small but nonzero survival benefit, others arguethat current adjuvant systemic therapy and radiation ther-apy would likely treat the majority of patients ade-quately. This study does not address this issue but pro-vides accurate and individualized estimates of thelikelihood of finding additional disease at completionALND. The American College of Surgeons OncologyGroup Protocol Z0011 (ACOSOG Z0011), currently un-der way and randomizing women with a positive SLN toundergo ALND or no ALND, is designed to address thisquestion directly.

In addition, the prognostic significance of micrometa-static nodal disease is a subject of debate. In his 1997review of the published literature, Dowlatshahi36 con-cluded that all but one of the large (N �147) and long-term (�6-year) studies demonstrated a statistically sig-nificant decrement in survival associated withmicrometastatic disease. At MSKCC, Tan et al.37 re-cently re-examined all axillary nodes from 373 patientstreated in the 1970s who were deemed to be node-negative by routine histopathological analysis. Nodeswere examined by serial sectioning and IHC, and thepresence of any detectable micrometastatic disease wasassociated with worse disease-free and overall survival.

Another limitation of our data is the absence of sizedetermination for the nodal metastases. The AJCC Can-cer Staging Manual, 6th edition, now includes size of

metastasis as an important determination of stage (and,therefore, of prognosis). However, we have had diffi-culty assigning a size to many cases because of thedifference in pattern of distribution of malignant cellswithin the node. For example, some nodes may havescattered single cells or multiple small clusters of cells.How should these be measured? Ideally, an accurateestimate of volume could be assigned to each SLN me-tastasis. However, this is extremely time-consuming andsomewhat impractical.

Nevertheless, it is clear that IHC is more sensitive thanH&E in detecting micrometastases, that routine H&Eanalysis is more sensitive than frozen-section analysis,and that there is a correlation between method of detec-tion and volume of disease. Others10,24 have demon-strated quantitatively that method of detection is corre-lated with measured size of the SLN metastasis.Therefore, in order to have a consistent, practical, andreproducible methodology of estimation, the method ofdetection of the nodal metastasis was used. This providesa general estimate of the amount of nodal disease andallows categorizing into four distinct groups.

Another potential weakness in our data is that some ofour patients, especially those with a perceived low risk ofadditional, non-SLN metastases, did not undergo a com-pletion ALND and therefore were not included in ourmodel. However, as demonstrated in the histograms(Figs. 2 and 4), the patients in the prospective population

FIG. 3. Nomogram without frozen-section information, for use when frozen-section analysis is not done at the time of sentinel lymph node (SLN)biopsy. See Figure 1 legend for instructions on nomogram use. Routine, routine H&E.



are distributed quite evenly across the range of predictedrisk. Furthermore, as demonstrated on the calibrationcurves, the models do predict well in the prospectivepopulation, even for those in the lowest decile of pre-dicted risk.

Some might express surprise at the effect of ER statusin the nomogram. Intuitively, one might expect ER pos-itivity to be associated with lower risk of additional,non-SLN metastases. In our model, the effect of ERstatus is of only borderline statistical significance (P �.08), but it was included to improve the overall predictiveability of the model. Furthermore, although the finding iscounterintuitive, others have reported similar findings.Nationwide data from the American College of Sur-geons38 also indicated that cases negative for ER had alower risk of lymph node metastases, after adjustment forall other factors.

Last, our models are imperfect. For the first model, thearea under the ROC curve was 0.77 for the prospective

population. This means that if we randomly select twowomen, of whom one has at least one positive non-SLNand the other has negative non-SLNs, there is a 77%chance that the nomogram will predict a higher proba-bility for the positive woman. This is a scale that rangesfrom 0.5, which would be achieved by tossing a coin, to1.0, which would require perfect ability to tell the posi-tive woman from the negative one.

Nevertheless, these models represent a significant im-provement over estimates based on one or two variablesin smaller populations. We have used our large, prospec-tive database to develop the models and have proventheir validity by testing them prospectively on a subse-quent population. The calibration errors of our modelsare small (see Figs. 2 and 4), generally �10% across thespectrum of predictors. Other investigators have shownthat removing statistically insignificant predictors actu-ally worsens the predictive ability of the model.39 Here,we have incorporated all statistically significant vari-

FIG. 4. Calibration plot for nomogram without frozen section data. The nomogram developed with use of the retrospective group of patients (n �702) was applied to the prospective group (n � 373). A histogram of the calculated probabilities for the prospective population is shown along thehorizontal axis. These 373 patients are grouped in deciles of their predicted probabilities, and the actual incidence of additional, non-SLN metastaseswas calculated for each decile. The vertical axis represents the actual, observed incidence (Actual Probability), and the horizontal axis represents theprobability calculated by the nomogram (Predicted Probability). For each decile of the prospective group, a triangle is plotted to show actualprobability. If the model were perfect, all triangles would lie on the dotted line, with a slope of 1.



ables, as well as other clinically available and relevantvariables, to provide improved prediction capability. No-mograms provide improved predictive ability in compar-ison with the crude counting of risk factors, and inaddition, nomograms usually outperform clinical judg-ment, according to numerous studies conducted in otherareas of medicine.40

With the important clinical question of whether toperform a completion ALND in a patient with a positiveSLN biopsy arising more and more frequently, our no-mograms provide an easy-to-use tool with which to si-multaneously incorporate several important variablesinto the estimate of risk of additional, non-SLN metas-tases. Further validation and follow-up studies such asACOSOG Z0011 will ultimately provide additionalguidance to the clinician and patient. These nomogramsprovide a risk estimate that can help in weighing the prosand cons of completion ALND for an individual patientwith SLN metastases.

ACKNOWLEDGMENTS

The acknowledgments are available online in the full-text version at www.annalssurgicaloncology.org. They

are not available in the PDF version.

REFERENCES

1. Giuliano AE, Jones RC, Brennan M, Statman R. Sentinel lymph-adenectomy in breast cancer. J Clin Oncol 1997;15:2345–50.

2. Albertini JJ, Lyman GH, Cox C, et al. Lymphatic mapping andsentinel node biopsy in the patient with breast cancer. JAMA1996;276:1818–22.

3. Veronesi U, Paganelli G, Galimberti V, et al. Sentinel node biopsyto avoid axillary dissection in breast cancer with clinically negativelymph-nodes. Lancet 1997;349:1864–7.

4. O’Hea BJ, Hill AD, El-Shirbiny AM, et al. Sentinel lymph nodebiopsy in breast cancer: initial experience at Memorial Sloan-Kettering Cancer Center. J Am Coll Surg 1998;186:423–7.

5. Krag D, Weaver D, Ashikaga T, et al. The sentinel node in breastcancer: a multicenter validation study. N Engl J Med 1998;339:941–6.

6. Veronesi U, Paganelli G, Viale G, et al. Sentinel lymph nodebiopsy and axillary dissection in breast cancer: results in a largeseries. J Natl Cancer Inst 1999;91:368–73.

7. Giuliano AE, Haigh PI, Brennan MB, et al. Prospective observa-tional study of sentinel lymphadenectomy without further axillarydissection in patients with sentinel node-negative breast cancer.J Clin Oncol 2000;18:2553–9.

8. Turner RR, Ollila DW, Krasne DL, Giuliano AE. Histopathologicvalidation of the sentinel lymph node hypothesis for breast carci-noma. Ann Surg 1997;226:271–8.

9. Chu KU, Turner RR, Hansen NM, Brennan MB, Bilchik A, Giu-liano AE. Do all patients with sentinel node metastasis from breastcarcinoma need complete axillary node dissection? Ann Surg 1999;229:536–41.

10. Kamath VJ, Giuliano R, Dauway EL, et al. Characteristics of thesentinel lymph node in breast cancer predict further involvement ofhigher-echelon nodes in the axilla. Arch Surg 2001;136:688–92.

11. Cabanes PA, Salmon RJ, Vilcoq JR, et al. Value of axillarydissection in addition to lumpectomy and radiotherapy in earlybreast cancer. Lancet 1992;339:1245–8.

12. Moore MP, Kinne DW. Axillary lymphadenectomy: a diagnosticand therapeutic procedure. J Surg Oncol 1997;66:2–6.

13. Carter CL, Allen C, Henson DE. Relation of tumor size, lymphnode status and survival in 24,740 breast cancer cases. Cancer1989;63:181–7.

14. American Joint Committee on Cancer. Breast. AJCC Cancer Stag-ing Manual. 6th ed. New York: Springer, 2002:221–40.

15. Sosa JA, Diener-West M, Gusev Y, et al. Association betweenextent of axillary lymph node dissection and survival in patientswith stage I breast cancer. Ann Surg Oncol 1998;5:140–9.

16. Hayward J, Caleffi M. The significance of local control in theprimary treatment of breast cancer. Arch Surg 1987;122:1244–7.

17. Osteen RT, Harris JR. Patients with early breast cancer benefitfrom effective axillary treatment. Breast Cancer Res Treat 1985;5:17–21.

18. Orr RK. The impact of prophylactic axillary node dissection onbreast cancer survival: a Bayesian meta-analysis. Ann Surg Oncol1999;6:109–16.

19. Cady B. Case against axillary lymphadenectomy for most patientswith infiltrating breast cancer. J Surg Oncol 1997;66:7–10.

20. Giuliano AE, Kirgan DM, Guenther JM, Morton DL. Lymphaticmapping and sentinel lymphadenectomy for breast cancer. AnnSurg 1994;220:391–401.

21. Reynolds C, Rosemarie M, Donohue JH, et al. Sentinel lymphnode biopsy with metastasis: can axillary dissection be avoided insome patients with breast cancer? J Clin Oncol 1999;17:1720–6.

22. Teng S, Dupont E, McCann C, et al. Do cytokeratin-positive-onlysentinel lymph nodes warrant complete axillary lymph node dis-section in patients with invasive breast cancer? Am Surg 2000;66:574–8.

23. Abdessalam SF, Zervos EE, Prasad M, et al. Predictors of positiveaxillary lymph nodes after sentinel lymph node biopsy in breastcancer. Am J Surg 2001;182:316–20.

24. Rahusen FD, Torrenga H, van Diest PJ, et al. Predictive factors formetastatic involvement of nonsentinel nodes in patients with breastcancer. Arch Surg 2001;136:1059–63.

25. Turner RR, Chu KU, Qi K, et al. Pathologic features associatedwith nonsentinel lymph node metastases in patients with metastaticbreast carcinoma in a sentinel lymph node. Cancer 2000;89:574–81.

26. Weiser MR, Montgomery LL, Tan LK, et al. Lymphovascularinvasion enhances the prediction of non-sentinel node metastasesin breast cancer patients with positive sentinel nodes. Ann SurgOncol 2001;8:145–9.

27. Wong SL, Edwards MJ, Chao C, et al. Predicting the status of thenonsentinel axillary nodes. Arch Surg 2001;136:563–8.

28. Viale G, Maiorano E, Mazzarol G, et al. Histologic detection andclinical implications of micrometastases in axillary sentinel lymphnodes for patients with breast carcinoma. Cancer 2001;92:1378–84.

29. Sachdev U, Murphy K, Derzie A, Jaffer S, Bleiweiss IJ, Brower S.Predictors of nonsentinel lymph node metastasis in breast cancerpatients. Am J Surg 2002;183:213–7.

30. Mignotte H, Treilleux I, Faure C, Nessah K, Bremond A. Axillarylymph-node dissection for positive sentinel nodes in breast cancerpatients. Eur J Surg Oncol 2002;28:623–6.

31. Jakub JW, Diaz NM, Ebert MD, et al. Completion axillary lymphnode dissection minimizes the likelihood of false negatives forpatients with invasive breast carcinoma and cytokeratin positiveonly sentinel lymph nodes. Am J Surg 2002;184:302–6.

32. Czerniecki BJ, Scheff AM, Callans LS, et al. Immunohistochem-istry with pancytokeratins improves the sensitivity of sentinellymph node biopsy in patients with breast carcinoma. Cancer1999;85:1098–103.



33. Cody HS III. Clinical aspects of sentinel node biopsy. BreastCancer Res 2001;3:104–8.

34. Harrell FE Jr. Regression Modeling Strategies With Applications toLinear Models, Logistic Regression, and Survival Analysis. SpringerSeries in Statistics. New York: Springer-Verlag, 2001:568.

35. Chu KU, Turner RR, Hansen NM, Brennan MB, Giuliano AE.Sentinel node metastasis in patients with breast carcinoma accu-rately predicts immunohistochemically detectable nonsentinelnode metastasis. Ann Surg Oncol 1999;6:756–61.

36. Dowlatshahi K, Fan M, Snider HC, Habib FA. Lymph nodemicrometastases from breast carcinoma: reviewing the dilemma.Cancer 1997;80:1188–97.

37. Tan LK, Giri D, Panageas K, et al. Occult/micrometastases in

axillary lymph nodes of breast cancer patients are significant: aretrospective study with long-term follow-up [abstract]. Proc AmSoc Clin Oncol 2002;21:37a.

38. Gann PH, Colilla SA, Gapstur SM, Winchester DJ, WinchesterDP. Factors associated with axillary lymph node metastasis frombreast carcinoma: descriptive and predictive analyses. Cancer1999;86:1511–9.

39. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models:issues in developing models, evaluating assumptions and adequacy,and measuring and reducing errors. Stat Med 1996;15:361–87.

40. Ross PL, Gerigk C, Gonen M, et al. Comparisons of nomogramand urologists’ predictions in prostate cancer. Semin Urol Oncol2002;20:82–8.



Editorial

The Eternally Enigmatic Axilla: Further Controversy AboutAxillary Lymph Nodes in Breast Cancer

Kelly M. McMasters, MD, PhD

“It is a capital mistake to theorize before one has data.Insensibly one begins to twist facts to suit theories,instead of theories to suit facts.” Sherlock Holmes to Dr.Watson, A Scandal in Bohemia. Sir Arthur Conan Doyle.

Few topics have remained as durably controversial asthe management of the axillary lymph nodes in breastcancer. In this issue, two reports allow us to reflect onceagain on the value of axillary lymph node dissection,particularly in light of ongoing clinical trials of sentinellymph node (SLN) biopsy.

By performing SLN biopsy, about two-thirds of pa-tients avoid having more extensive axillary dissection—because they have negative SLN. For the other one-thirdof patients with positive SLN, standard treatment is toperform completion lymph node dissection. Althoughthere is no proof that removing these non-sentinel nodeswill improve survival, most patients and physicians areloath to leave behind easily resectable disease, especiallygiven the fact that about 37% of patients will haveresidual axillary nodal metastases detected by routinehistopathology.1 Van Zee and colleagues2 from Memo-rial Sloan-Kettering Cancer Center have constructed anomogram to determine the likelihood of finding non-sentinel node metastases for patients with positive SLN.After creating the nomogram based on data from 702patients, they then applied it prospectively to 373 pa-tients to predict the status of the remaining axillarynodes. The implication of this nomogram is obvious—ifwe can predict the status of the non-sentinel nodes ac-

curately, we can identify patients who may not requirecompletion axillary dissection.

I am old-fashioned about this issue. I still think that,outside of a clinical trial, axillary dissection should beperformed for patients with a positive SLN. Althoughthere is little evidence to suggest a survival advantage foraxillary dissection, I do not discount completely thepossibility that axillary dissection benefits some node-positive patients. Furthermore, the additional staging in-formation (number of positive nodes) obtained from ax-illary dissection for node-positive patients sometimesalters the adjuvant therapy plan. Finally, I believe thatlocoregional disease control in the axilla is a laudableaim that is often overlooked. The development of axil-lary nodal recurrence, an otherwise catastrophic event forpatients, is almost completely preventable if axillarydissection is performed for the SLN-positive patients.

The nomogram has good predictive ability across theentire range of patients with positive nodes. It likely canbe improved, however, with further prospective testingand more data. As the authors point out, from the exist-ing literature, the two best predictors of non-sentinelnode status are the size of the primary tumor and the sizeof the nodal micrometastases. However, the nomogramdoes not incorporate the crucial factor of size of nodalmetastasis, but rather relies on an indirect measure of thisparameter, i.e. the method by which the nodal metastasesare detected (routine histology, serial sections, or immu-nohistochemistry). I think that size of nodal metastasis,which is now included in the new American Joint Com-mittee on Cancer (AJCC) staging system, needs to beincorporated specifically in the nomogram. Furthermore,I do not think it is as important for the nomogram todiscriminate between a between a 15% risk and a 50%risk of residual nodal metastases as it is to predict accu-rately those patients with a very low risk (0 to 5%).These low-risk patients are the ones who reasonablymight choose to avoid further axillary surgery. From the

Received October 15, 2003; accepted October 21, 2003.From the Division of Surgical Oncology, University of Louisville–J.

Graham Brown Cancer Center, 315 East Broadway, Louisville,Kentucky.

Address correspondence to Kelly M. McMasters, MD, PhD, Divi-sion of Surgical Oncology, University of Louisville–J. Graham BrownCancer Center, 315 East Broadway, Louisville, KY 40202; Fax: 502-629-3183; E-mail: [email protected].

Published by Lippincott Williams & Wilkins © 2003 The Society of SurgicalOncology, Inc.

Annals of Surgical Oncology, 10(10):1128–1130

DOI: 10.1245/ASO.2003.10.917

1128

data presented, we do not know how good the nomogramis for predicting this very-low-risk population. I suspectthere are some loopholes. For instance, according to thenomogram, a patient with a 1 cm, nuclear grade II,estrogen receptor positive tumor, with only a singlepositive SLN removed that contains a 2 mm metastasisdetected by frozen section, has nearly a 40% risk ofresidual nodal metastasis. Based on other available data,I suspect that the true risk is �5%.3,4

The Memorial nomogram represents significantprogress in our understanding of factors predicting moreextensive axillary nodal metastasis. My only concern isthat this nomogram, which soon will be available tosurgeons everywhere in easy-to-use PDA and internetformat, may make it far too easy for some surgeons toskip axillary dissection for patients with positive SLN. Ido not think we are at the point that the nomograminformation should be used for this purpose. Sure, someSLN-positive patients will refuse axillary dissection, andthat is their right. The nomogram at least will helpquantify their risk of non-sentinel node metastasis sothey can make a more informed decision. Until we haveresults from the American College of Surgeons Oncol-ogy Group (ACOSOG) trial Z0011 to guide us in thisregard, axillary dissection should be the rule, not theexception, for patients with a positive SLN.

Krag and Single5 would no doubt have little use for thenomogram, because their analysis of the Surveillance,Epidemiology and End Results (SEER) database leadsthem to believe that axillary dissection improves survivalfor all subgroups of patients. They divided patients intocategories: age groups 40–49 or 50–79, node-negativeor node-positive (1 to 3 positive nodes only), and ana-lyzed survival by the number of axillary nodes removed.In these subgroups, survival improves with increasingnumber of lymph nodes removed. According to the au-thors, not only does removing a greater number of axil-lary nodes improve survival for patients with positivelymph nodes, but for patients with negative nodes aswell! For node-negative patients with �10 nodes re-moved, there was an absolute survival benefit of 3%compared to those with fewer nodes removed. This dif-ference was statistically significant only in the post-menopausal age group. For patients with 1–3 positivenodes, removing �10 nodes resulted in an absolute in-crease in survival of 10% and 11% for the premeno-pausal and postmenopausal age groups, respectively. Be-cause the database includes over 72,000 patients, theauthors are careful to state that statistical significancemay not equate to clinical relevance.

How do we interpret these results? Krag and Singlebelieve that they have unearthed new evidence that ax-

illary dissection improves survival. I believe that theyhave discovered new proof that stage migration is aconcept that is poorly understood, and that retrospectivesubgroup analysis of large databases is fraught withpotential error.

In the node-negative patients, a 3% survival advantagefor removing more lymph nodes is undoubtedly ac-counted for by stage migration. The authors do not tellus, but I suspect there is a striking correlation betweenthe number of nodes removed and the presence andnumber of positive nodes found! The more nodes exam-ined, the more patients will be thrown into the positivenode category, and the stages will migrate (much like themigration of the Okies to California in Will Rogers’time, which improved the average I.Q. of both states).The same explanation likely holds true for the node-positive population. The authors limited their analysis topatients with 1 to 3 positive nodes. Why limit the anal-ysis to this subgroup? All this does is create an artificialcategory that allows stage migration into the �3 positivenode group. It makes more sense to include all patientswith positive nodes, or all patients regardless of nodalstatus. If you look at the survival curves, it is clear thatmost of the apparent survival decrease is accounted forby the patients who had 1–5 nodes removed. If only 5lymph nodes are examined, how likely is it that 4 or morewill be positive? Because the number of positive nodes issuch a strong predictor of survival, this strongly arguesfor stage migration.

To bolster their case for the therapeutic value of axil-lary dissection, the authors point to the 10-year follow-updata of the NSABP trial B-04, which demonstrated a 4%survival difference favoring patients who had axillarydissection but which did not reach clinical significance.They fail to mention the more current 25-year follow-updata from this study that show, unequivocally, no sur-vival advantage for axillary dissection.6 The authorsclaim that the sample size in B-04 is too small to detectsmall differences in survival. Yet there is no difference todetect–the curves for total mastectomy and radical mas-tectomy actually intersect at about 20 years follow-up.Increasing the sample size and statistical power will notmake differences appear when they do not exist. Kragand Single cite a meta-analysis that suggests a 5.4%survival advantage for axillary dissection when B-04 andfive other studies are combined.7 No doubt a repeatmeta-analysis, taking into account the longer follow-upof B-04, would result in a lower estimate of potentialbenefit for axillary dissection.

It is difficult to gaze upon the 25-year survival datafrom the landmark NSABP B-04 study, then come upwith the hypothesis that axillary dissection improves

1129EDITORIAL


survival—even when the node-positive patients are in-cluded. Still, I do not completely discount the possibilitythat axillary dissection might improve survival in a smallfraction of patients. But for a therapy to be effective, theappropriate risk group must be singled out to demon-strate efficacy. If axillary dissection can benefit anypatients, they are the ones with positive nodes. There canbe no advantage to removing normal lymph nodes. Soeven if one believes that there is a 5.4% survival advan-tage for axillary dissection, who believes that this sur-vival advantage would persist if the node-positive pa-tients were excluded? Yet that is exactly what Drs. Kragand Single propose that the SEER data demonstrate.They would have us believe that, through some mysticalprocess, patients with negative SLN have submicro-scopic nodal disease, not detectable by pathologists, thatspreads systemically and kills them, even though theynever develop clinically evident axillary nodal metasta-ses. This is despite the facts that: (1) intensive immuno-histochemical analysis of the non-sentinel nodes fails toreveal microscopic nodal metastases when the SLN isnegative for tumor,8 (2) SLN biopsy detects at least thesame proportion of node-positive patients as axillarydissection,9 and (3) axillary nodal recurrence after anegative SLN biopsy appears to be a very rare event.10–12

This article from Krag and Single is a paper-thinattempt to justify the implausible hypothesis of the Na-tional Surgical Adjuvant Breast and Bowel Project(NSABP) Trial B-32. This trial is purportedly designedto determine “whether sentinel-node biopsy alone actu-ally reduces the rate of survival when compared withaxillary dissection.”13 What it really tests, in a convo-luted way, is the hypothesis that failure to perform axil-lary dissection for patients with negative SLN decreasessurvival. Usually, large randomized trials are based onsignificant hypothesis-generating data. But where are thedata to suggest that axillary dissection benefits the node-negative patients? The authors reach back to retrospec-tive data from 1970 to find support for their hypothesis.14

Even if axillary dissection is associated with a smallsurvival advantage for the node-positive patients, there isno clear justification, without resorting to analysis ofdata that are explained by stage migration, for the ideathat axillary dissection will save the lives of patients withnegative nodes. Therefore, the only way there could be alegitimate difference between the randomized treatmentarms in the NSABP B-32 study is for the SLN false-negative rate to be so high that it affects survival—andeven this is mathematically impossible. While this studywill provide important information regarding the prog-nostic significance of nodal micrometastases, and hasbeen an excellent mechanism for training surgeons to

perform SLN biopsy under the auspices of a cooperativegroup trial, I fear that its ultimate impact will be limited.

So, I hope that clears things up about the axillarylymph nodes in breast cancer. Van Zee and colleagues2

are cautiously optimistic that we can identify some pa-tients with positive SLN who do not need axillary dis-section. This issue is being addressed, if accrual can becompleted, in ACOSOG trial Z0011. Krag and Single5

tell us that we should perform axillary dissection evenfor the patients with negative nodes. This issue is beingaddressed, notwithstanding my concerns expressedabove, in NSABP trial B-32. Despite a century of debate,we still are left wondering about the therapeutic value ofaxillary lymph node dissection in breast cancer.

REFERENCES

1. Wong SL, Edwards MJ, Chao C, Tuttle TM, Noyes RD, Woo C,Cerrito PB, McMasters KM Predicting the status of non-sentinelaxillary nodes: a multicenter study. Arch Surg 2001;136:563–8.

2. Van Zee KJ, Manasseh D-M E, Bevilacqua JLB, et al. A nomo-gram for predicting the likelihood of additional nodal metastases inbreast cancer patients with a positive sentinel node biopsy. AnnSurg Oncol 2003;10:1140–51.

3. Chu KU, Turner RR, Hansen NM, Brennan MB, Bilchik A, Giu-liano AE. Do all patients with sentinel node metastasis from breastcarcinoma need axillary node dissection? Ann Surg 1999;229:536–541.

4. Reynolds C, Donohue MR, Grant CS, et al. Sentinel lymph nodebiopsy with metastasis: can axillary dissection be avoided in somepatients with breast cancer? J Clin Oncol 1999;17:1720–6.

5. Krag DN, Single RM. Breast cancer survival according to numberof nodes removed. Ann Surg Oncol 2003;10:1152–9.

6. Fisher B, Jeong J-H, Anderson S, Bryant J, Fisher ER, Wolmark N.Twenty-five-year follow-up of a randomized trial comparing rad-ical mastectomy, total mastectomy, and total mastectomy followedby irradiation. N Engl J Med 2002;347:567–75.

7. Orr RK. The impact of prophylactic axillary node dissection onbreast cancer survival—a Bayesian meta-analysis. Ann Surg Oncol1999;6:109–16.

8. Turner RR, Ollila DW, Krasne DL, Giuliano AE. Histopathologicvalidation of the sentinel lymph node hypothesis for breast carci-noma. Ann Surg 1997;226:271–6.

9. Giuliano AE, Dale PS, Turner RR, Morton DL, Evans SW, KrasneDL. Improved axillary staging of breast cancer with sentinellymphadenectomy. Ann Surg 1995;222:394–9.

10. Veronesi U, Paganelli G, Viale G, et al. A randomized comparisonof sentinel-node biopsy with routine axillary dissection in breastcancer. N Engl J Med 2003;349:546–53.

11. Blanchard DK, Donohue JH, Reynolds C, Grant CS. Relapse andmorbidity in patients undergoing sentinel lymph node biopsy aloneor with axillary dissection for breast cancer. Arch Surg 2003;138:482–7.

12. Giuliano AE, Haigh PI, Brennan MB, et al. Prospective observa-tional study of sentinel lymphadenectomy without further axillarydissection in patients with sentinel node-negative breast cancer.J Clin Oncol 2000;18:2553–9.

13. Krag D, Ashikaga T. The design of trials comparing sentinel-nodesurgery and axillary resection. N Engl J Med 2003;349:603–6.

14. Fisher B, Slack JH. Number of lymph nodes examined and theprognosis of breast carcinoma. Surg Gynecol Obstet 1970;131:79–88.

1130 K. M. MCMASTERS


Proc Annu Meet Am Soc Clin Oncol 2002; 21:37A.

146 OCCULT/MICROMETASTASES IN AXILLARY LYMPH NODES OF BREAST CANCER PATIENTS ARE SIGNIFICANT: A RETROSPECTIVE STUDY WITH LONG-TERM FOLLOW-UP. L K Tan, D Giri, K Panageas, M Cranor, J Bevilacqua, A Hummer, E Brogi, L Norton, C Hudis, P Borgen, Memorial Sloan-Kettering Cancer Center, New York, NY. Background: In breast cancer, sentinel lymph node biopsy (SLNB) as an alternative to axillary lymph node (ALN) dissection has ushered in an era of enhanced pathologic nodal analysis [H&E serial sections and cytokeratin immunohistochemistry (IHC)] raising questions concerning the clinical significance of occult/micrometastatic (O/MM) disease. IHC increases detection of O/MM, and thus decreases the false negative rate of SLNB. We investigated the prognostic significance of O/MM in a cohort of node-negative breast cancer (NNBC) patients with long-term follow-up. Methods: From 1976 to 1978, 373 unselected patients with NNBC were treated with mastectomy and ALN dissection at MSKCC. Median follow-up among survivors was 17.6 years (range, 2.6-24.9 years). Enhanced pathologic analysis was performed on a total of 2,470 ALN blocks by sectioning tissue blocks at 2 deeper levels, 50 mm apart, and staining with H&E and IHC (AE1:AE3, Ventana Medical Systems, Inc., Tucson, AZ). We estimated disease-free survival (DFS) using Kaplan-Meier methodology and breast cancer specific death (BCSD) by a competing risk model. The log-rank test was used to compare the IHC/H&E subgroups. Results: 84/373 (23%) were found to be node-positive by either H&E or IHC; 28/84 (33%) were positive by IHC only, while 56/84 (67%) were positive by H&E. 60/84 (71%) had only one positive ALN. Estimates of disease-free survival (DFS) and BCSD at twenty years are given below. [TABLE] Both DFS and BCSD were associated with positive nodes (p<0.01). Among the IHC(-)/H&E(-), DFS in patients with tumors ≤2 cm was significantly better than those with tumors >2 cm (p<0.05) and equaled the survival of historic series of patients with tumors ≤1 cm. Conclusions: The presence of O/MM in ALN detected by IHC is univariately associated with worsened breast cancer-specific outcomes. IHC analysis of ALN identifies a subset of NNBC patients at increased risk for recurrence and death from breast cancer.

Rerefence: Tan LK, Giri D, Pangeas K, Cranor M, Bevilacqua JL, Hummer A, Brogi E, Norton L, Hudis C, Borgen PI. Occult micrometastases in axillary lymph nodes of breast cancer patients are significant: a retrospective study with long term follow up. (Meeting Abstract #146). Proc Annu Meet Am Soc Clin Oncol 2002; 21:37A.

EJSO 2002; 28: 490±500

doi:10.1053/ejso.2002.1268, available online at http://www.idealibrary.com on1

A model for predicting axillary node metastasesbased on 2000 sentinel node procedures andtumour positionJ. L. B. Bevilacqua*, H. S. Cody III², K. A. MacDonald², L. K. Tan³,P. I. Borgen² and K. J. Van Zee²

*Hospital SõÂrio Libanes, Rua Adma Jafet 91, SaÄo Paulo, SP 01308-050, Brazil, Instituto Brasileiro de Controle do CaÃncer, Avenida AlcaÃntara

Machado 2576, SaÄo Paulo, SP 03102-002, Brazil, and Disciplina de Cirurgia Geral, Departamento de Cirurgia, Faculdade de Medicina da

Univerdidade de SaÄo Paulo, Av. Dr Arnaldo, 455 SaÄo Paulo, SP 01246-903, Brazil. Departments of ² Surgery and ³Pathology,

Memorial Sloan±Kettering Cancer Center, 1275 York Avenue, New York, NY 10021, USA

Aims: The purpose was to identify the independent predictive factors of axillary lymph-node metastases (ALNM) ininfiltrating ductal carcinoma (IFDC) and to create a prospective, validated statistical model to predict the likelihood ofALNM in patients in the present era of sentinel lymph-node (SLN) biopsy and enhanced histopathology.Methods: Univariate and multivariate analyses of 13 clinicopathological variables (including tumour location) wereperformed to determine predictors of ALNM in 1659 eligible SLN biopsy procedures. A logistic regression model wasdeveloped and then prospectively validated on a second population of 187 subsequent consecutive procedures.Results: Age, pathological tumour size, palpability, lymphovascular invasion (LVI), histological grade, nuclear grade,ductal histological subtype, tumour location (quadrant) and multifocality were associated with ALNM in univariateanalyses (P , 0.001). Of these, only palpability and histological grade were not statistically associated with ALNM inthe multivariate analysis (P . 0.05). The frequency of ALNM in upper-inner-quadrant (UIQ) tumours was 20.6%,compared with 33.2% for all other quadrants (P , 0.0005). There was no statistical difference between UIQ andother-quadrant tumours in any clinicopathological variables analysed. The logistic regression model, developed basedon the population of 1659, had the same accuracy, sensitivity, specificity, positive predictive value and negativepredictive value when applied prospectively to the second population.Conclusion: Tumour size, LVI, age, nuclear grade, histological subtype, multifocality and location in the breast wereindependent predictive factors for ALNM in IFDC. ALNM is less frequent in UIQ tumours than in other-quadranttumours. Our prospectively validated predictive model could be valuable in pre-operative patient discussions,although staging of the axilla in the individual patient remains necessary.

# 2002 Elsevier Science Ltd. All rights reserved.

Key words: breast carcinoma; axillary lymph-node metastases; sentinel lymph nodes; tumour location; medialtumours; predictors.

INTRODUCTION

Axillary lymph-node status is the most significantprognostic factor in patients with invasive breastcancer.1±5 Axillary sentinel lymph-node (SLN) biopsy is

0748±7983/02/$35.00

Presented in part at the 23rd Annual San Antonio Breast CancerSymposium, San Antonio, TX, 6±9 December 2000.Correspondence to: Kimberly J. Van Zee MS, MD, FACS, 1275 YorkAvenue, MRI 1026, New York, NY 10021, USA. Tel: (212) 639-6997;Fax: (212) 794-5812; E-mail: [email protected]

rapidly being adopted as a standard treatment option forpatients with early-stage breast cancer, replacing trad-itional axillary lymph-node dissection (ALND).6,7 SLNbiopsy is highly accurate for staging the axilla,6,7 with theaccuracy of SLN biopsy surpassing that of standardALND when multiple sectioning or immunohisto-chemical (IHC) staining of the SLNs is used.8 In theera of SLN biopsy, enhanced histopathological analysis(serial sectioning and IHC) of the lymph nodes isfeasible, because of the smaller number of axillary nodesrequiring analysis per patient.

# 2002 Elsevier Science Ltd. All rights reserved.

AXILLARY METASTASES ARE LESS FREQUENT IN UIQ CANCERS 491

The frequency of axillary lymph-node metastases(ALNM) by T stage (AJCC-1997) varies among pre-SLN-era studies. In the larger studies, the frequency is10% in T1mic,9 9±13% in T1a,10±14 13±19% in T1b,10±15

26±29% in T1c,10,11,14,15 39±50% in T2 less than or equalto 3 cm,10,11,14,15 48±59% in T2 larger than 3 cm,14,15 and71±80% in T3.11,14

We identified 32 studies published between 1989and 2000 that address the prediction of ALNM.9±40

The great majority of these studies collected the dataretrospectively, did not mention how the data were col-lected, or had no central pathology review. In the32 studies identified, tumour size,9,10,12,14±21,23±39

age,9,10,12±15,19,22,23,30 lymphovascular invasion (LVI)10,14,

17,18,20,21,34 and histological subtype12,14,15,20,23,37,38

were the variables most commonly described asindependently associated with ALNM. Only one studyvalidated a statistical model in another data set.23

In the SLN literature, we identified only one smallstudy of 110 cases that addressed the prediction ofALNM in patients who underwent SLN biopsy withenhanced histopathological analysis.25 Tumour size andHER-2/neu positivity were the only factors independ-ently associated with ALNM in this study.

The goal of the present study is to identify thefrequency and independent predictive factors of ALNMand to create a prospective, validated statistical modelto predict the likelihood of lymph-node metastases inpatients in this new era of SLN biopsy and enhancedhistopathology.

PATIENTS AND METHODS

Our retrospective study population consisted of the first2000 breast cancer cases undergoing axillary SLN biopsyat Memorial Sloan±Kettering Cancer Center (MSKCC)between 12 September 1996 and 24 March 2000. These2000 cases represented a total of 1955 individualpatients. Data were prospectively collected for eachprocedure.

We excluded 33 cases in which no exact pathologicaltumour size could be determined, one case of sarcoma,117 cases of ductal carcinoma in situ (DCIS) and 190cases of infiltrating lobular carcinoma (IFLC). A total of1659 SLN cases of infiltrating ductal carcinoma (IFDC)were eligible for our study. Of these, 1556 were IFDCnot otherwise specified (NOS), 103 were tubular, colloidor medullary carcinomas, two were metaplastic carci-nomas and one was secretory carcinoma.

We included the two cases of metaplastic carcinomawith the NOS group and the single case of secretorycarcinoma with the classic special subtypes (tubular,colloid, and medullary). We excluded the 190 cases ofIFLC because this tumour type is not classified byhistological and nuclear grade, and these missing values

would exclude these cases in the multivariate statisticalanalysis.

The 1659 remaining cases constituted our total studypopulation. Their clinicopathologic characteristics aredescribed in Table 1. Patient selection, technique ofSLN biopsy and histopathological analyses have beendescribed elsewhere.41

Positive lymph nodes were categorized as IHC only(lymph node metastases found exclusively by IHC),H&E+IHC (lymph node metastases found on serialsections by haematoxylin and eosin (H&E) or H&E as aresult of IHC) and conventional pathology (CP) (lymphnode metastases detected by routine single-sectionH&E).

First, we performed a retrospective analysis of the 1659eligible cases in the prospective SLN database. In order todetermine factors that were statistically significant inpredicting ALNM, we evaluated each of the followingvariables by univariate and multivariate analyses: age,clinical and pathological tumour size, laterality, palpabil-ity, ductal subtype (NOS and special subtypes), tumourlocation within the breast (quadrant), LVI, multifocality,ER status, PR status, HER-2/neu status, tumour (T)stage, histological grade and nuclear grade. We classifiedT stage according to the 1997 tumour, node andmetastasis (TNM) classification system.42

Secondly, we prospectively validated our logisticregression model of prediction of ALNM, developedbased on the 1659 retrospective cases. In this secondpart, we studied 187 consecutive patients with IFDCbetween 24 March 2000 and 14 July 2000.

Statistical methods

The variables analysed for the prediction of ALNM arelisted in Table 1. Statistical analyses were done using thestatistical software SPSS, version 10.0.5 (SPSS Inc.,Chicago, IL), and StatXact, version 4.0.1 (Cytel SoftwareCorporation, Cambridge, MA).

Univariate analysis

We determined correlations among categorical vari-ables in the univariate analysis using the Pearson w2 test.We also analysed age, pathological and clinical tumoursize as continuous variables using the Wilcoxon rank-sum test, because these variables lacked normaldistribution.

Multivariate analysis

The variables that demonstrated statistical correlationswith ALNM on univariate analysis (defined as P , 0.05)were analysed with binary logistic regression usingboth stepwise and enter methods. We used theHosmer±Lemeshow test to test the goodness of fit ofthe observed and predictive results of the logistic

Table 1 Patient and tumor characteristics

Number (%) Number (%)

Number of cases 1659 (100.0) Histological gradeI 214 (12.9)

Mean age (�SE) 56.36 ((0.30) II 524 (31.6)III 845 (50.9)

Age distribution Unknown 76 (4.6)�40 years 162 (9.8)41±69 years 1216 (73.3) Nuclear grade� 70 years 281 (16.9) I 179 (10.8)

II 860 (51.8)Tumour size (T stage) III 509 (30.7)

T1mic 31 (1.9) Unknown 111 (6.7)T1a 195 (11.7)T1b 458 (27.6) LVIT1c 716 (43.2) No 1255 (75.6)T2�3.0 cm 203 (12.2) Yes 404 (24.4)T2� 3.0 cm 49 (3.0)T3 7 (0.4) Multifocality

No 1404 (84.6)Laterality Yes 255 (15.4)

Right 816 (49.2)Left 843 (50.8) Oestrogen receptor (ER) status

Negative 262 (15.8)Bilaterality Positive 940 (56.7)

No 1585 (95.5) Unknown 457 (27.5)Yes 74 (4.5)

Progesterone receptor (PR) statusPalpability Negative 430 (25.9)

No 744 (44.8) Positive 763 (46.0)Yes 863 (52.0) Unknown 466 (28.1)Unknown 52 (3.1)

HER-2/neuTumour location Negative 284 (17.1)

UOQ 897 (54.1) 1� positive 177 (10.7)LOQ 206 (12.4) 2� positive 130 (7.8)UIQ 247 (14.9) 3� positive 73 (4.4)LIQ 113 (6.8) � intensity unknown 11 (0.7)Central 192 (11.6) Unknown 984 (59.3)Unknown 4 (0.2)

Axillary metastasesOperation type No 1139 (68.7)

Conservation 1308 (78.8) Yes 520 (31.3)Mastectomy (MRM or TM) 351 (21.2)

Number of positive nodesTumour type No positive nodes 1139 (68.7)

IFDC NOS 1556 (93.8) 1±3 positive nodes 428 (25.8)Colloid/medullary/tubular 103 (6.2) 4±9 positive nodes 63 (3.8)

� 10 nodes 29 (1.7)

SE, standard error; UIQ, upper-inner quadrant; UOQ, upper-outer quadrant; LOQ, lower-outer quadrant; LIQ,lower-inner quadrant.

492 J. L. B. BEVILACQUA ET AL.

regression model. A minimum value (cut-off value) of30% was used to determine positive predictive risk ofALNM.

In the first part of the study (retrospective), theresults are presented using two models: the complete

and simplified models. In the complete model, thevariables are expressed as categorical; in the simplifiedmodel, some variables are expressed as continuous.

In the second part of the study (prospective), wetested the simplified logistic regression model on the


next 187 consecutive patients with IFDC. We usedFisher's exact test to compare the accuracy, specificity,sensitivity and predictive positive and negative values ofthe model in the retrospective and prospectivepopulations.

RESULTS

The overall frequency of ALNM was 31.1%. Thefrequency of ALNM by T stage and method of detectionof metastases is demonstrated in Figure 1.

Univariate analysis

Age, clinical and pathological tumour size (includingT stage), palpability, tumour subtype, tumour location,LVI, multifocality, and histological and nuclear gradewere statistically correlated with ALNM (Tables 2 and 3).

T stage IHC-only

(%) H&E±IHC

(%)Conventional

Pathology (%)Overall (%)

T1mic 9.7 0 0 9.7

T1a 5.1 3.6 6.7 15.4

T1b 4.4 3.5 13.3 21.2

T1c 4.2 5 24.9 34.1

T2 (≤≤3cm) 3.4 3.9 44.9 52.2

T2 (>3cm) 4.1 0 63.2 67.3

T3 0 0 100 100

0

25

50

75

100

Ax

illa

ry M

eta

sta

se

s (

%)

T1a

N=195

T1b

N=458

T3

N=7

T1mic

N=31

T1c

N=716

T2( 3cm)

N=203

T2(>3cm)

N=49

9.7% 15.4%21.2%

34.1%

52.2%

67.3%

100.0%

Figure 1 Frequency of ALNM in ductal carcinomas byT stage and method of detection.

Table 2 Ductal carcinoma ± univariate a

N

Mean (�SE)Age (years) 57.4Clinical tumour size (cm) 0.60Pathological tumour size (cm) 1.18

n� 1659.* Wilcoxon rank-sum test.

Tumour location as a predictor of axillarylymph-node metastases

ALNM were seen in roughly one-third of all cases withtumours located in quadrants other than the UIQ; theincidence of ALNM in the UIQ was 20.6% (P , 0.0005)(Fig. 2).

We performed multiple univariate comparisonsbetween the UIQ and each of the other quadrants. Toavoid a type 1 statistical error, we corrected thesignificance of the P value to P� 0.01. UIQ tumours hadstatistically fewer ALNM than tumours in each of theother quadrants (P , 0.01). This difference became onlymarginally significant (P� 0.012), however, when com-pared with ALNM incidence in LIQ tumours (Table 4).

To be certain that there was no difference betweenthe tumours of the UIQ and those at other locationswithin the breast which might bias our results, wecompared all other variables for these two groups. Wefound no statistical difference between these two groupsin any of the variables analysed (Table 5). The lowerfrequency of metastases from the UIQ tumours was theonly difference between these tumours and thoselocated elsewhere in the breast.

Multivariate analysis

Variables that were independently associated withALNM in ductal carcinomas in the final model wereductal subtype, LVI, T stage, tumour location, nucleargrade, age and multifocality (Tables 6 and 7).

Predicting axillary lymph-node metastases

The logistic regression model estimates the probabilityof an event; in our case, the probability of ALNM.Models with multiple variables, such as age, T stage andnuclear grade, and subcategories within each variable aredifficult to apply in clinical settings. To simplify the appli-cation of this model and to calculate the probability ofALNM for a specific patient, we formulated a simplifiedmodel (Tables 7 and 8) that could be used easily with ascientific calculator, a computer program such asMicrosoft ExcelTM or a personal digital assistant.

nalysis

egativenodes

Positivenodes

P *

(�12.3) 54.2 (�11.8) ,0.0005(�0.94) 1.13 (�1.26) ,0.0005(�0.71) 1.73 (�1.02) ,0.0005

Table 3 Ductal carcinoma ± univariate analysis

Number(n)

Positivenodes n (%)

P * Number(n)

Positivenodes n (%)

P *

Age distribution ER�40 years 162 69 (42.6) Negative 262 82 (31.3) 0.35341±69 years 1216 387 (31.8) ,0.0005 Positive 940 323 (34.4)� 70 years 281 64 (22.8) Unknown ² 457 115 (25.2)

Laterality PRRight 816 252 (30.9) 0.690 Negative 430 130 (30.2) 0.070Left 843 268 (31.8) Positive 763 270 (35.4)

Unknown ² 466 120 (25.8)Palpability

No 744 171 (23.0) ,0.0005 HER-2/neu (clinical)YesUnknown²

86352

329 (38.1) Negative(negative and 1� )

460 158 (34.3) 0.316

Positive (2� and 3� ) 214 82 (38.3)Unknown² 985 280 (28.4)

Tumour typeIFDC-NOS 1556 515 (33.1) ,0.0005 T stageColloid/medullary/

tubular103 5 (4.9) T1mic

T1aT1b

31195458

3 (9.7)30 (15.4)97 (21.2)

Tumour locationUOQ 897 299 (33.3) T1c 716 244 (34.1) ,0.0005LOQ 206 69 (33.5) T2 (�3 cm) 203 106 (52.2)UIQ 247 51 (20.6) 0.004 T2 (> 3 cm) 49 33 (67.3)LIQ 113 37 (32.7) T3 7 7 (100.0)Central 192 62 (32.3)Unknown ² 4 2 (50.0) Histological grade

I 214 22 (10.3)Tumour location II 524 162 (30.9) ,0.0005

UIQ 247 51 (20.6) ,0.0005 III 845 327 (38.7)Other 1408 467 (33.2) Unknown ² 76 9 (11.8)Unknown ² 4 2 (50.0)

Nuclear gradeLVI I 179 18 (10.1)

No 1255 290 (23.1) ,0.0005 II 860 264 (30.7) ,0.0005Yes 404 230 (56.9) III 509 216 (42.4)

Unknown ² 111 22 (19.8)Multifocality

No 1404 417 (29.7) 0.001Yes 255 103 (40.4)

* 2-sided Pearson �2 test.² Excluded from statistical analysis.


Using the general equation (formula) of binary logisticregression43 and applying the b values and patient andtumour parameter values as listed in Tables 7 and 8,respectively, we developed the following formula tocalculate the probability of ALNM for individualpatients:

P � 1=�1� exp�4:043� 0:015Aÿ 0:576T ÿ 0:229N

ÿ 1:715Sÿ 1:131Lÿ 0:837Qÿ 0:359M��

where is A age, T tumour size, N nuclear grade, S ductalsubtype, L LVI, Q quadrant and M multifocality.

Of 1659 IFDC cases, 1544 (93.1%) were included inthe logistic regression analysis. The model excludedfrom the analysis 115 (6.9%) cases that had missingvalues for either nuclear grade (111 cases) or tumourlocation (four cases).

The simplified logistic regression model achieved adiagnostic accuracy of 69.2%, a sensitivity of 67.1%, aspecificity of 70.1%, a positive predictive value of 51.5%and a negative predictive value of 81.8%. The Hosmer±Lemeshow test observed a significance of P� 0.888 forthe �2 value (�2� 3.641, 8 degrees of freedom). Basedon this, we accepted the null hypothesis that there was

Figure 2 Frequency of ALNM in ductal carcinomas by tumourlocation (quadrant) (P-value, two-sided Pearson w2 test).

Table 4 ALNM ± univariate analysis: multiplecomparisons of UIQ vs other-quadrant tumours

Comparison P (�2)*

UIQ vs UOQ ,0.0005UIQ vs LOQ 0.002UIQ vs central 0.006UIQ vs LIQ 0.012UOQ vs LOQ 0.965UOQ vs LIQ 0.900UOQ vs central 0.781LOQ vs LIQ 0.892LOQ vs central 0.798LIQ vs central 0.935

* Level of significance adjusted ± statistically significantwhen P , 0.01.

Table 5 Tumour location ± univariate analysis: UIQ vsother-quadrant tumours

Variable UIQ Other

P *Mean age (years) 56.2 56.4 0.834Mean clinical size (cm) 0.77 0.76 0.967Mean pathological size (cm) 1.31 1.35 0.466

P ²Palpability (�) 53.9 53.7 0.954LVI (�) 23.9 24.5 0.835Multifocal (�) 15.4 15.3 0.983ER (�) 78.2 78.2 0.995PR (�) 65.2 63.7 0.713HER-2/neu (�) 28.8 32.5 0.468

T stageT1mic 1.6 1.9T1a 11.3 11.9T1b 29.6 27.3T1c 45.3 42.8 0.457T2 (�3 cm) 8.5 12.8T2 (.3 cm) 3.6 2.8T3 0 0.5

SubtypeIFDC-NOS 93.1 93.9 0.642Colloid/medullary/tubular 6.9 6.1

Nuclear gradeI 11.9 11.5II 54.6 55.8 0.945III 33.5 32.7

Histological gradeI 12.3 13.7II 33.5 33.1 0.842III 54.2 53.2

* Wilcoxon rank-sum test.² 2-sided Pearson w2 test.


no difference between the observed and predictivevalues. Therefore, the model fitted the data.

Prospective validation of the logisticregression model

Our simplified logistic regression model was testedprospectively in 187 consecutive patients with IFDC.There was no statistical difference in the prediction ofnodal metastases between the retrospective andprospective groups for model accuracy, sensitivity,

specificity, positive predictive value and negative pre-dictive value (Table 9).

DISCUSSION

Frequency of axillary lymph-nodemetastases

The frequency of ALNM in each T stage in the SLN eraappears higher than the corresponding frequency fromthe pre-SLN era. This is probably due to the use ofenhanced histopathological analysis in the SLN era,which allows detection of so-called `occult metastases',not usually seen in the past with conventional patho-logical analyses. Although the clinical relevance of thesemicrometastatic deposits is the subject of controversy,data from large studies with long follow-up periodssuggest that occult metastases may be associatedwith a small but significant decrease in overall and/or

Table 6 Multivariate analysis: ductal carcinoma ± complete logistic regression model

Variable b OR (eb) 95.0% CI (lower±upper) P

Tumour typeColloid/medullary/tubular 1IFDC-NOS or metaplastic 1.498 4.473 1.547±12.934 ,0.0005

LVINo 1Yes 1.154 3.171 2.436±4.129 ,0.0005

T StageT3�T2 (.3 cm) 1T3�T2 (.3 cm) vs T1mic ÿ3.061 0.047 0.012±0.186 ,0.0005T3�T2 (.3 cm) vs T1a ÿ2.193 0.112 0.050±0.248 ,0.0005T3�T2 (.3 cm) vs T1b ÿ1.841 0.159 0.079±0.319 ,0.0005T3�T2 (.3 cm) vs T1c ÿ1.441 0. 237 0.121±0.464 ,0.0005T3�T2(.3 cm) vs T2 (�3 cm) ÿ0.983 0.374 0.183±0.766 0.007

Tumour locationUIQ 1Other 0.854 2.349 1.612±3.424 ,0.0005

Nuclear gradeI 1I vs II 0.667 1.948 1.116±3.403 0.019I vs III 0.870 2.387 1.336±4.265 0.003

Age (years)�70 1�70 vs �40 0.590 1.804 1.116±2.915 0.019�70 vs 41±69 0.464 1.590 1.129±2.239 0.008

MultifocalityNo 1Yes 0.411 1.508 1.092±2.083 0.013

Constant ÿ2.926 0.054 ,0.0005

CI, confidence interval; OR, odds ratio.

Table 7 Multivariate analysis: ductal carcinoma ± simplified logistic regression model

Variable b OR (eb) 95.0% CI (lower±upper) P

Age (years) ÿ0.015 0.985 0.976±0.995 0.003Tumour pathological size (cm) 0.576 1.779 1.527±2.073 ,0.0005Nuclear grade (1 or 2 or 3) 0.229 1.257 1.015±1.557 0.036

Tumour subtypeColloid/medullary/tubular 0 1IFDC-NOS or metaplastic 1.715 5.557 1.970±15.676 0.001

LVINo 0 1Yes 1.131 3.099 2.385±4.027 ,0.0005

Tumour location (quadrant)UIQ 0 1Other 0.837 2.309 1.584±3.368 ,0.0005

MultifocalityNo 0 1Yes 0.359 1.432 1.432±1.970 0.027

Constant ÿ4.043 ,0.0005


Table 8 Values to be applied in the equation of thesimplified logistic regression model

Variable Value to be appliedin the equation

A Age YearsT Tumour size (pathological) cmN Nuclear grade 1, 2 or 3S Tumour subtype

Colloid/medullary/tubular 0IFDC-NOS or metaplastic 1

L LVINo 0Yes 1

Q Tumour location (quadrant)UIQ 0Other 1

M MultifocalityNo 0Yes 1

Table 9 Application of the model in the retrospectiveand prospective populations

Retrospective Prospective P-value*

Accuracy 69.2% 68.9% 1.000Sensitivity 67.1% 67.2% 1.000Specificity 70.1% 70.0% 1.000Positive

predictive value51.5% 55.6% 0.555

Negativepredictive value

81.8% 79.2% 0.508

* Fisher's exact test: exact P-value.


disease-free survival.44,45 Large, ongoing prospectivestudies such as the National Surgical Adjuvant Breast andBowel Project trial B32 (NSABP-B-32), the AmericanCollege of Surgeons Oncology Group trial Z0010(ACOSOG-Z0010) and the International Breast CancerStudy Group (IBCSG) trial 23-01 will better define thesignificance of micrometastatic lymph-node disease.

Predictors of axillary lymph-nodemetastases

Using SLN biopsy and enhanced pathological analysis, wehave demonstrated that size, LVI, ductal histologicalsubtype, nuclear grade, age, multifocality and tumourlocation within the breast are independent predictivefactors for nodal metastases in patients with invasiveductal carcinomas. Other, non-SLN studies have alsodemonstrated that size,9±21,23,24,27,30±40 age,9±15,18,19,

21±23,27,30 LVI,10,11,14,17,18,20,21,29,34,36,40 histological sub-type,12,13,15,20,23,27,37,38 nuclear grade,11,13,17,18,32,34,38

multifocality10,20 and location12,14,15,21,23,26,28,38 are pre-dictive factors of ALNM in univariate and/or multivariateanalysis. A list of the variables independently associatedwith ALNM, identified by multivariate analysis in studiesdone over the past decade, is presented in Table 10.The factors associated with ALNM have not changeddramatically with the advent of SLN biopsy, a fact thatserves to reinforce general confidence in the SLN biopsyprocedure.

Tumour size, LVI, age and tumour subtype (ductalNOS subtypes vs others) are the most commonlyreported independent factors for ALNM among thesestudies; not surprisingly, we also identified these four

variables as independent predictors of ALNM in ourown study.

Only a minority of published studies have identifiednuclear grade and histological grade (or tumour grade)as predictive of ALNM (Table 10). Nuclear grading is thecytological evaluation of tumour nuclei in comparisonwith the nuclei of normal mammary epithelial cells,whereas histological or tumour grading describes themicroscopic growth patterns of invasive ductal carci-nomas as well as cytologic features of differentiation.46

These variables may be absent from other studiesbecause IFLC is not usually classified by grade andmissing values are usually excluded from the final modelin multivariate analysis. For this reason, we excluded allcases of IFLC in our study population. In this analysis, inwhich we intentionally include only IFDC, we identifiednuclear grade as an independent predictive factor forALNM.

Our data demonstrate multifocality to be an inde-pendent predictive factor for ALNM. Although weidentified only two studies in which multifocality hasbeen found to be a predictor of ALNM,10,20 thisobservation is consistent with the fact that multifocalcarcinomas generally have a larger volume of tumourthan similarly staged unifocal tumours and tumour size ispredictor of ALNM. Evidence also exists that multi-focality is an independent predictor of worse disease-free survival.47

Tumour location

Recent studies have demonstrated a worse prognosisassociated with medial-quadrant tumours than withlateral-quadrant tumours in terms of disease-free andoverall survival.28,48,49 Other studies have reporteddifferences in patterns of distant metastasis for thesetumour locations.50 The authors of these studiesexplained that these differences may be due to thegreater likelihood of occult spread to internal mammarylymph nodes seen in medial tumours. We hypothesizedthat, if this likelihood is real, the frequency of ALNMshould be different for medial and lateral tumours and

Table 10 Studies of predictors of axillary lymph-node metastases by multivariate analysis

Variable Number of studies References

Size 20 9,10,12,14±21,23±25,30,34,36±39Age 10 9,10,12±15,19,22,23,30LVI 7 10,14,17,18,20,21,34Type 7 12,14,15,20,23,37,38Location 4 12,15,21,23Tumour palpability 3 14,17,21Nuclear grade 3 13,17,34Histological grade 3 9,12,40Tumour grade 2 23,39Multifocality 2 10,20PR 2 23,30Clinical nodal status 2 10,14Menopausal status 1 10ER 1 23HER-2/neu 1 25Margin status 1 14Border shape (spiculated) 1 20S-phase 1 30Ploidy 1 23Race 1 23Microcalcification 1 39Number of nodes examined 1 15Nipple invasion 1 20Parity 1 10


therefore included primary tumour location as a variablein our analyses.

Our data demonstrate that tumour location is anindependent predictive factor for ALNM. This finding issupported by other published studies,12,14,15,21,23,26,38,50

four of which also found tumour location to be anindependent predictive factor,12,15,21,23 and reported ahigher frequency of ALNM associated with lateraltumours. Indeed, in the two largest studies on predictionof ALNM, the first by Gann et al.23 with 18 000 patients(using the Patient Care Evaluation Survey Database ofthe American College of Surgeons) and the other byMaibenco et al.12 with 13 000 patients (using NationalCancer Institute±SEER database), UIQ tumours had alower frequency and lower OR for ALNM than lateraltumours. In the study by Gann et al.,23 the ALNM rate ofUIQ tumours was 0.54 times that of UOQ tumours.This is consistent with our finding that the ALNM rate ofUIQ tumours was 0.62 that of UOQ tumours (20.6% vs33.3%, Table 3). Maibenco et al.12 demonstrated thattumours in the UIQ measuring 1 cm or less had anALNM incidence rate of 9.2%, compared with a rate of13.6% for similar size tumours located in otherquadrants (P , 0.0005). They report that the ALNMrate of medial-quadrant tumours was 0.76 times that ofnon-medial-quadrant tumours. Again, this is in agree-ment with our data, in which this ratio was 0.73.

The lower frequency of ALNM in medial-locationtumours observed in the current study and by others

raises the issue of alternative routes of lymphatic flow.Are these tumours draining ± and metastasizing ± to theinternal mammary chain? In a companion review, weconsider the clinical relevance of internal mammarylymph-node metastases and propose an algorithm forselecting patients for internal mammary SLN biopsy.52 Itis possible that, in a small subset of patients withsubcentimetre tumours and negative axillary lymphnodes, information obtained by internal mammary SLNbiopsy could markedly change their prognosis andtreatment.

Simplified logistic regression model

Our logistic regression model for predicting ALNM wasdeveloped following the completion of the first 2000SLN biopsy procedures at our institution (MSKCC) inMarch 2000 and was prospectively tested in thesubsequent 187 consecutive patients with IFDC. Therewere no statistical differences in model specificity,sensitivity, accuracy and positive and negative predictivevalues between the two groups (retrospective andprospective). The simplified logistic regression model(Tables 7 and 8) can be used easily with a scientificcalculator or a personal digital assistant to calculate theprobability of finding an axillary lymph-node metastasisin an individual patient. Because the result is given as aprobability rather than as an OR, the model can convey


meaningful information to both patient and physicianduring pre-operative discussions.

ACKNOWLEDGEMENTS

The authors would like to express their gratitude toProfessor R. G. Bevilacqua MD, PhD of Faculdade deMedicina da Universidade de SaÄo Paulo, Brazil, for hisgenerous statistical support. We would also like tothank Dr M. F. Brennan of MSKCC, NY, for his insightsand encouragement and Drs A. Heerdt, L. Montgomery,M. Paglia and J. Petrek, also of MSKCC, for contributingpatients to the series. Lastly, we are grateful to J. Feyand J. Winawer for their assistance with databasemanagement.

REFERENCES

1. Rosen PP, Groshen S. Factors influencing survival and prognosis inearly breast carcinoma (T1N0M0±T1N1M0). Assessment of 644patients with median follow-up of 18 years. Surg Clin North Am1990; 70(4): 937±62.

2. Rosen PP, Groshen S, Kinne DW, Norton L. Factors influencingprognosis in node-negative breast carcinoma: analysis of 767T1N0M0/T2N0M0 patients with long-term follow-up. J Clin Oncol1993; 11(11): 2090±100.

3. Rosen PP, Groshen S, Saigo PE, Kinne DW, Hellman S. Pathologicalprognostic factors in stage I (T1N0M0) and stage II (T1N1M0)breast carcinoma: a study of 644 patients with median follow-up of18 years. J Clin Oncol 1989; 7(9): 1239±51.

4. Mustafa IA, Cole B, Wanebo HJ, Bland KI, Chang HR. Prognosticanalysis of survival in small breast cancers. J Am Coll Surg 1998;186(5): 562±9.

5. NIH Consensus Conference. Treatment of early-stage breastcancer. J Am Med Assoc 1991; 265(3): 391±5.

6. Cody HS III. Sentinel lymph-node mapping in breast cancer.Oncology (Huntingt) 1999; 13(1): 25±34.

7. Cody HS III. Management of the axilla in early stage breast cancer:will sentinel node biopsy end the debate? J Surg Oncol 1999; 71(3):137±9 (editorial).

8. Giuliano AE, Dale PS, Turner RR, Morton DL, Evans SW,Krasne DL. Improved axillary staging of breast cancer withsentinel lymphadenectomy. Ann Surg 1995; 222(3): 394±9.

9. Rivadeneira DE, Simmons RM, Christos PJ, Hanna K, Daly JM,Osborne MP. Predictive factors associated with axillary lymph-node metastases in T1a and T1b breast carcinomas: analysis inmore than 900 patients. J Am Coll Surg 2000; 191(1): 1±6.

10. Fehr MK, Kochli OR, Helfenstein U, Schar G, Bussmann C, Fink D,Haller U. Multivariate analysis of clinico-pathological predictors ofaxillary lymph-node metastasis in invasive breast carcinoma.Geburtshilfe Frauenheilkd 1995; 55(4): 182±8 (in German).

11. Jackson JS, Olivotto IA, Wai MDE, Grau C, Mates D, Ragaz J. Adecision analysis of the effect of avoiding axillary lymph-nodedissection in low risk women with invasive breast carcinoma.Cancer 2000; 88(8): 1852±62.

12. Maibenco DC, Weiss LK, Pawlish KS, Severson RK. Axillary lymph-node metastases associated with small invasive breast carcinomas.Cancer 1999; 85(7): 1530±6.

13. Mustafa IA, Cole B, Wanebo HJ, Bland KI, Chang HR. The impactof histopathology on nodal metastases in minimal breast cancer.Arch Surg 1997; 132(4): 384±90.

14. Olivotto IA, Jackson JS, Mates D et al. Prediction of axillary lymph-node involvement of women with invasive breast carcinoma: amultivariate analysis. Cancer 1998; 83(5): 948±55.

15. Voogd AC, Coebergh JW, Repelaer van Driel OJ et al. The risk ofnodal metastases in breast cancer patients with clinically negativelymph nodes: a population-based analysis. Breast Cancer Res Treat2000; 62(1): 63±9.

16. Ahlgren J, Stal O, Westman G, Arnesson LG. Prediction of axillarylymph-node metastases in a screened breast cancer population.South-East Sweden Breast Cancer Group. Acta Oncol 1994; 33(6):603±8.

17. Barth A, Craig PH, Silverstein MJ. Predictors of axillary lymph-nodemetastases in patients with T1 breast carcinoma. Cancer 1997;79(10): 1918±22.

18. Chadha M, Chabon AB, Friedmann P, Vikram B. Predictors ofaxillary lymph-node metastases in patients with T1 breast cancer.A multivariate analysis. Cancer 1994; 73(2): 350±3.

19. Choong PL, deSilva CJ, Dawkins HJ et al. Predicting axillary lymph-node metastases in breast carcinoma patients. Breast CancerRes Treat 1996; 37(2): 135±49.

20. De Laurentiis M, Gallo C, De Placido S, et al. A predictive index ofaxillary nodal involvement in operable breast cancer. Br J Cancer1996; 73(10): 1241±7.

21. Fein DA, Fowble BL, Hanlon AL, et al. Identification of women withT1±T2 breast cancer at low risk of positive axillary nodes. J SurgOncol 1997; 65(1): 34±9.

22. Gajdos C, Tartter PI, Bleiweiss IJ et al. Stage 0 to stage III breastcancer in young women. J Am Coll Surg 2000; 190(5): 523±9.

23. Gann PH, Colilla SA, Gapstur SM et al. Factors associated withaxillary lymph-node metastasis from breast carcinoma: descriptiveand predictive analyses. Cancer 1999; 86(8): 1511±9.

24. Giuliano AE, Barth AM, Spivack B et al. Incidence and predictors ofaxillary metastasis in T1 carcinoma of the breast. J Am Coll Surg1996; 183(3): 185±9.

25. Giuliano AE, Jones RC, Brennan M, Statman R. Sentinellymphadenectomy in breast cancer. J Clin Oncol 1997; 15(6):2345±50.

26. Halverson KJ, Taylor ME, Perez CA et al. Management of the axillain patients with breast cancers one centimeter or smaller. Am J ClinOncol 1994; 17(6): 461±6.

27. Iwasaki Y, Fukutomi T, Akashi-Tanaka S et al. Axillary nodemetastasis from T1N0M0 breast cancer: possible avoidance ofdissection in a subgroup. Jpn J Clin Oncol 1998; 28(10): 601±3.

28. Lohrisch C, Jackson J, Jones A et al. Relationship between tumourlocation and relapse in 6781 women with early invasive breastcancer. J Clin Oncol 2000; 18(15): 2828±35.

29. Port ER, Tan LK, Borgen PI, Van Zee KJ. Incidence of axillarylymph-node metastases in T1a and T1b breast carcinoma. Ann SurgOncol 1998; 5(1): 23±7.

30. Ravdin PM, De Laurentiis M, Vendely T, Clark GM. Prediction ofaxillary lymph-node status in breast cancer patients by use ofprognostic indicators. J Natl Cancer Inst 1994; 86(23): 1771±5.

31. Reger V, Beito G, Jolly PC. Factors affecting the incidence oflymph-node metastases in small cancers of the breast. Am J Surg1989; 157(5): 501±2.

32. Saiz E, Toonkel R, Poppiti RJ Jr, Robinson MJ. Infiltrating breastcarcinoma smaller than 0.5 centimeters: is lymph-node dissectionnecessary? Cancer 1999; 85(10): 2206±11.

33. Seidman JD, Schnaper LA, Aisner SC. Relationship of the size ofthe invasive component of the primary breast carcinoma to axillarylymph-node metastasis. Cancer 1995; 75(1): 65±71.

34. Shoup M, Malinzak L, Weisenberger J, Aranha GV. Predictors ofaxillary lymph-node metastasis in T1 breast carcinoma. Am Surg1999; 65(8): 748±52.

35. Silverstein MJ, Gierson ED, Waisman JR et al. Predicting axillarynode positivity in patients with invasive carcinoma of the breast byusing a combination of T category and palpability. J Am Coll Surg1995; 180(6): 700±4.

36. Van Lancker M, Goor C, Sacre R, et al. Patterns of axillarylymph-node metastasis in breast cancer. Am J Clin Oncol 1995;18(3): 267±72.

37. Velanovich V. Predictors of lymph-node metastasis for breastcancer. In: Proceedings of the Annual Meeting of the AmericanSociety for Clinical Oncology. Vol. 16. 1997: A687, (meetingabstract).

38. Velanovich V, Szymanski W. Lymph-node metastasis in breastcancer: common prognostic markers lack predictive value. AnnSurg Oncol 1998; 5(7): 613±9.

39. Walls J, Boggis CR, Wilson M, et al. Treatment of the axilla inpatients with screen-detected breast cancer. Br J Surg 1993; 80(4):436±8.


40. Yiangou C, Shousha S, Sinnett HD. Primary tumour characteristicsand axillary lymph-node status in breast cancer. Br J Cancer 1999;80(12): 1974±8.

41. Cody HS III, Borgen PI. State-of-the-art approaches to sentinelnode biopsy for breast cancer: study design, patient selection,technique and quality control at Memorial Sloan±Kettering CancerCenter. Surg Oncol 1999; 8(2): 85±91.

42. Breast. In: American Joint Committee on Cancer: AJCC Cancer StagingManual (5th edn). Philadelphia, PA: Lippincott-Raven, 1997:171±80.

43. Dawson-Sounders B, Trapp RG. Basic and Clinical Biostatistics (2ndedn). Norwalk, CT: Appleton & Lange, 1994: 222.

44. Liberman L. Pathologic analysis of sentinel lymph nodes in breastcarcinoma. Cancer 2000; 88(5): 971±7 (editorial; comment).

45. Dowlatshahi K, Fan M, Snider HC, Habib FA. Lymph-nodemicrometastases from breast carcinoma: reviewing the dilemma.Cancer 1997; 80(7): 1188±97.

46. Rosen PP. Invasive duct carcinoma and morphological prognosticmarkers. In: Rosen's Breast Pathology (1st edn). Philadelphia, PA:Lippincott-Raven, 1997: 279±80.

47. Giri D, Thaler H, Bevilacqua JLB, Cranor M, Tan LK: Multifocalversus unifocal node-negative breast carcinomas: Are they

prognostically different? 90th Annual Meeting of United Statesand Canadian Academy of Pathology, Atlanta, Georgia, March2001, Abstract 132.

48. Asaga T, Masuzawa T, Yoshida A, Kawahara S. Location of breastcancer as a prognostic factor. Breast Cancer Res Treat 1994; 32: S48(meeting abstract).

49. Zucali R, Mariani L, Marubini E, et al. Early breast cancer: evaluationof the prognostic role of the site of the primary tumour. J ClinOncol 1998; 16(4): 1363±6.

50. Israel L, Gogeul E, Morere JF, Breau JL. Distinct metastatic patternin breast tumour of inner and central areas. In: Proceedings of theAnnual Meeting of the American Society for Clinical Oncology.Vol. 14. 1995: A110 (meeting abstract).

51. Urban JA. Clinical experience and results of excision of the internalmammary lymph node chain in the primary operable breast cancer.Cancer 1959; 12(1): 14±22.

52. Bevilacqua JLB, Gucciardo G, Cody HS III, et al. Proposedpatient selection algorithm for internal mammary sentinellymph-node biopsy in breast cancer. Eur J Surg Oncol 2002; 28:000±00. doi:10.1053/ejso.2002.1269.

Accepted for publication 1 February 2002

EJSO 2002; 28: 603±614


REVIEW

A selection algorithm for internal mammarysentinel lymph node biopsy in breast cancerJ. L. B. Bevilacqua*, G. Gucciardo², H. S. Cody III³, K. A. MacDonald³,V. Sacchini³, P. I. Borgen³ and K. J. Van Zee³

*Hospital SõÂrio Libanes, Rua Adma Jafet 91, SaÄo Paulo, SP 01308-050, Brazil, Instituto Brasileiro de Controle do CaÃncer, Avenida AlcaÃntara

Machado 2576, SaÄo Paulo, SP 03102-002, Brazil, and Disciplina de Cirurgia Geral, Departamento de Cirurgia, Faculdade de Medicina da

Univerdidade de SaÄo Paulo, Av. Dr Arnaldo, 455 SaÄo Paulo, SP 01246-903, Brazil, ²Department of Surgery, Azienda Ospedaliera

S. Camillo ± Forlanini, Via Portuense, 332, 00149 Rome, Italy, and ³Department of Surgery, Memorial Sloan±Kettering Cancer Center,

1275 York Avenue, New York, NY 10021, USA

Internal mammary lymph-node (IMN) metastases in breast carcinomas have a major influence on survival, comparablewith the influence of axillary lymph-node metastases (ALNM). Prospective, randomized trials have demonstratedthat complete IMN dissection as part of extended radical mastectomy does not improve overall or disease-freesurvival. In the subset of patients with tumours 1 cm or less in size and no ALNM, information on IMN status wouldprovide important information. In these cases, the presence of IMN metastases would change the staging from stage Ito stage IIIB, according to the current tumour, node and metastasis classification. More importantly, it would influencethese patients' adjuvant treatment. Lymphatic mapping for sentinel lymph-node (SLN) biopsy has demonstrated extra-axillary drainage in up to 35% of patients. Recent reports have demonstrated the feasibility of internal mammarysentinel lymph-node (IM-SLN) biopsy.Here we review the general prognostic and clinical significance of tumor location and lymph-node metastases in breastcancer and discuss the specific factors associated with IMN identification, metastases and treatment in the pre-SLNand SLN eras. Based on our review, we propose an algorithm for a selective approach to IM-SLN in breast cancer.

# 2002 Published by Elsevier Science Ltd

Key words: breast carcinoma; internal mammary; lymph nodes; sentinel lymph nodes; tumour location; axillarymetastases; prognosis.

INTRODUCTION

The importance of the influence of internal mammarynode (IMN) metastases on survival is well known, havingbeen demonstrated in numerous studies.1±4 IMN dissec-tion for the staging and treatment of breast cancer hasbeen largely abandoned in current practice, however,primarily because of the failure of past studies to demon-strate any significant survival benefit associated with theremoval of IMNs.5 Furthermore, although studies of

1748±7983/02/$35.00

Correspondence to: Kimberly J. Van Zee MS, MD, FACS, 1275 YorkAvenue, MRI 1026, New York, NY 10021, USA. Tel: (212) 639 6997;Fax: (212) 794 5812; E-mail: [email protected]

sentinel lymph-node (SLN) biopsy with lymphatic mappingusing a radiolabelled isotope report the detection ofextra-axillary drainage in up to 35% of patients,6 atpresent extra-axillary nodes are not usually consideredin the management of early-stage breast cancer.7

A number of recent reports have addressed thefeasibility of internal mammary sentinel lymph-node (IM-SLN) biopsy as a staging procedure for early-stage breastcancer.3,8±10 IM-SLN biopsy is not routinely performed,however, and does involve potential risks includingpneumothorax8,9 and bleeding8,9 and, in some cases,additional scarring resulting in poorer cosmetic results.

Staging the IMNs may be particularly important inpatients with subcentimetre tumours (1 cm or less insize) and negative axillary nodes (stage I), because these

# 2002 Published by Elsevier Science Ltd


patients are generally not treated with adjuvanttreatment. According to current tumour, node andmetastasis (TNM) staging guidelines,11 the presence ofIMN metastases in stage I patients would result in theirbeing upstaged to stage III. Therefore, in the absence ofIM-SLN biopsy, some stage I patients may in fact beunderstaged. These patients may constitute a subsetwith a worse outcome among patients with subcenti-metre tumours.

In this paper, we address the factors associated withIMN identification and metastases, their clinical rele-vance and the feasibility of and considerations for IMNbiopsy in the pre-SLN and SLN eras. We also attempt toidentify the subset of patients in whom IMN metastasesare most likely to occur, using the predictive modelpresented in our previous study.12 The presence of IMNmetastases could alter adjuvant treatment in a veryselect subset of cases. Through the identification ofthis subset of patients by means of the predictive model,IM-SLN biopsy could be performed with far greaterselectivity, thereby improving the procedure's overallfeasibility and value.

TUMOUR LOCATION,LYMPH-NODE METASTASESAND PROGNOSIS

Tumour location and lymph-nodemetastases

In a previous study, we demonstrated the lowerfrequency of axillary lymph node metastases (ALNM)in upper-inner-quadrant (UIQ) tumours and the inde-pendent predictive value of tumour location forALNM.12 These findings are supported by the findingsof other studies.13±20 Four studies have demonstratedtumour location to be an independent predictivefactor13,14,16,20 and reported a higher frequency ofALNM associated with lateral tumours and a lowerfrequency associated with medial tumours. Thesefindings raise a question regarding the anatomy oflymphatic drainage in the breast: are medial tumorsmore likely to have an alternative drainage route, namelyto the internal mammary chain?

In 1627, Gaspard Aselius described the lymphaticdrainage pathway of the breast to the internal mammarychain using a dog model.21 Cruikshank, in 1786, describedthis pathway in humans.21 This concept was subsequentlyconfirmed by Stibbe22 in 1918 and Turner-Warwick21

in 1956. In the recent SLN era, breast lymphoscintigraphyshows mapping to IMNs in 10±35% of cases.6,23±26

There seems to be a direct correlation between thetiming of injection of the radioisotope and the findingson the scan image obtained. Specifically, the longer theinterval between injection and scanning, the greaterthe number of IMNs seen on the lymphoscintigram. This

may explain the range of percentages seen amongthe various studies. According to Guenther et al.24 thefrequency of positive (`hot') IMNs on lymphoscintigra-phy in medial tumours is 26.7% (8/30). According toBorgstein et al.23 this frequency is 27.6% (8/29) in medialtumours and 31.3% (5/16) in UIQ tumours, comparedwith 10.9% (9/82) in lateral and 8.2% (5/61) in upper-outer-quadrant (UOQ) tumours (medial vs lateral,P� 0.0675). Uren et al.6 found that 28% (5/18) of lateraltumours had drainage to IMNs, compared with 50% (3/6)of medial tumours (P� 1.000). Johnson et al.9 foundIM-SLNs in 12.5% (4/32) of medial/central tumours andin 12.5% (6/48) of outer-quadrant tumours (P� 1.000).(All P values discussed above were calculated by theauthors and do not appear in the original studies.)

In 1951, Urban18 reviewed 1000 consecutive operablebreast cancer cases treated at Memorial Sloan±KetteringCancer Center between 1940 and 1945 by radicalmastectomy (RM) and adjuvant radiation therapy. Hefound a higher incidence of parasternal chest-wallrecurrences in patients with medial tumours than inpatients with lateral tumours: 17±20% vs 2±7% respect-ively. He described these parasternal chest-wall recur-rences as local extensions of metastatic cancer from theIMNs.18 In 1952, based on these findings, he describedthe extended radical mastectomy (ERM), which includescomplete IM chain dissection, for the treatment ofprimary operable breast cancers.27

The overall frequency of IMN metastases in breastcancer is approximately 22.5% (Table 1). The frequencyin axillary node-positive patients is approximately 36.1%,and in node-negative patients 9.3% (Table 1). In a popula-tion pre-selected for medial location (79% medial), Codyand Urban1 reported a frequency of IMN metastases of33% (4/12) in axillary node-negative patients withtumours 1 cm or less in size. This is the only study forwhich we were able to identify the frequency of IMNmetastases in this node-negative, subcentimetre subsetof patients, although the high frequency reported seemsinfluenced by selection criteria for the study population(medial location). Yet, it is this node-negative, subcenti-metre population in whom the presence of IMNmetastases would markedly change adjuvant systemictreatment. To estimate better the frequency of IMNmetastases in this subgroup, we have made a series ofcalculations, as outlined in the following paragraph.

In our previous study,12 the frequency of ALNM forT1 and T2 tumours was 31%. For tumours less than orequal to 1 cm, the frequency was 19%. Therefore, theratio of frequencies of ALNMs in tumours 1 cm or less(�T1b) and in T1 and T2 tumours was 0.613 (19%(T1mic�T1a�T1b) divided by 31% (T1�T2)). Assum-ing the ratio of the frequencies of IMN metastases forthese tumours is the same as that for ALNMs (0.613),we can estimate the frequency of IMN metastases intumours 1 cm or less by multiplying this ratio by theoverall frequency of IMN metastases in T1 and T2

Table 1 Frequency of IMN metastases in patients who underwent pathological examination of both axillary and IM nodes in the pre-SLN era (the largest series from each group orinstitution was selected)

Study IMN metastases in axillary node-negative patients, % (n) IMN metastases in axillary node-positive patients, % (n) All patients,

% (n), overall

Overall IMN

metastases

axillary ÿ

Tumour location P value,

medial vs

lateral

Overall IMN

metastases

axillary �


medial vs

lateral

IMN

metastases

Medial Central Lateral Medial Central Lateral

Unselected population

Caceres28 6.1 (6/98) 9.8 (4/41) 0 (0/15) 4.8 (2/42) 0.4326 31.6 (48/152) 45.8 (22/48) 38.5 (10/26) 20.5 (16/78) 0.0047 21.6 (54/250)

Sugarbaker29 9.6 (7/73) Ð Ð Ð Ð 38.8 (33/85) Ð Ð Ð Ð 25.3 (40/158)

Dahl-Iversen and

Tobiassen30

6.1 (18/294) Ð Ð Ð Ð 36.8 (67/182) Ð Ð Ð Ð 17.9 (85/476)

Livingston and

Arlen31

7.8 (28/358) 14.3 (10/70) 8.1 (10/123) 4.8 (8/165) 0.0282 34.2 (77/225) 59.3 (16/27) 46.3 (31/67) 22.9 (30/131) 0.0003 18.0 (105/583)

Lacour et al.32,33 9.8 (32/327) 11.4 (19/166)a Ð 8.1 (13/161) 0.3543 27.9 (105/376) 36.9 (58/157)a Ð 21.5 (47/219) 0.0011 20.1 (141/703)

Rai et al.34 Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð 32.0 (8/25)

Veronesi and

Valagussa35 b

9.3 (15/161) 10.3 (7/68)a Ð 8.6 (8/93) 0.7869 30.4 (55/181) 36.4 (28/77)a Ð 25.9 (27/104) 0.1441 20.5 (70/342)

Li and Shen2 4.9 (30/607) 9.2 (16/174)c 5.8 (6/104)c 2.1 (6/291)c 0.0010 30.1 (191/635) 34.9 (52/149)c 36.8 (49/133)c 25.2 (80/317)c 0.0362 17.8 (221/1242)

Morimoto

et al.36

0 (0/45) 0 (0/27)a Ð 0 (0/18) 1.000 17.6 (9/51) 25.0 (5/20)a Ð 12.9 (4/31) 0.2893 9.4 (9/96)

Subtotald 6.7 (121/1802) 10.3 (49/478) 6.6 (16/242) 4.3 (29/677) 0.0001 31.1 (530/1706) 38.2 (153/401) 39.8 (90/226) 22.8 (177/776) ,0.0001 18.8 (663/3533)

Selected population

McDonald

et al.37

Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð 33.8 (35/104)

Wyatt et al.38 24.1 (7/29) 62.5 (5/8) 28.6 (2/7) 0 (0/14) 0.0021 38.7 (12/31) 60.0 (3/5) 71.4 (5/7) 21.1 (4/19) 0.1265 31.7 (19/60)

Hutchinson and

Kiriluk39

Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð 23.4 (33/141)

Pavrovsky

et al.40

15.8 (6/38) Ð Ð Ð Ð 40.6 (13/32) Ð Ð Ð Ð 27.1 (19/70)

Bucalossi

et al.41 e

8.2 (50/610) Ð Ð Ð Ð 35.9 (217/603) Ð Ð Ð Ð 22.0 (267/1213)

Urban and

Castro42

15.6 (60/384) 15.9 (58/364) 8.3 (1/12) 12.5 (1/8) 1.0000 51.9 (177/341) 54.8 (143/261) 42.6 (23/54) 42.3 (11/26) 0.3025 32.7 (237/725)

Richardson

et al.43

6.7 (1/15) Ð Ð Ð Ð 59.1 (13/22) Ð Ð Ð Ð 37.8 (14/37)

Caceres44 f 6.8 (20/296) 7.7 (8/104)c 9.1 (4/44)c 5.4 (8/148)c 0.6010 28.9 (120/414) 45.7 (58/127)c 30.1 (25/83)c 18.5 (37/200)c , 0.0001 19.7 (140/710)

ASE

LEC

TIO

NA

LG

OR

ITH

MFO

RIN

TER

NA

LM

AM

MA

RY

SLN

BIO

PSY

605

Study IMN metastases in axillary node-negative patients, % (n) IMN metastases in axillary node-positive patients, % (n) All patients,

% (n), overall

Overall IMN

metastases

axillary ÿ


medial vs

lateral

Overall IMN

metastases

axillary �


medial vs

lateral

IMN

metastases

Medial Central Lateral Medial Central Lateral

Selected population

Handley45 7.7 (36/465) 12.3 (22/179) 6.6 (5/76) 4.3 (9/210) 0.0045 34.9 (187/535) 50.0 (65/130) 46.4 (65/140) 21.5 (57/265) , 0.0001 22.5 (225/1000)

Donegan46 6.2 (3/48) 11.8 (2/17)c 0 (0/8)c 4.3 (1/23)c 0.5647 33.8 (22/65) 54.5 (6/11)c 33.3 (2/6)c 31.8 (14/44)c 0.1810 22.1 (25/113)

Deemarski and

Seleznev47

17.2 (43/250) Ð Ð Ð Ð 46.5 (106/228) Ð Ð Ð Ð 31.2 (149/478)

Veronesi et al.4 9.1 (51/563) Ð Ð Ð Ð 29.1 (162/556) Ð Ð Ð Ð 19.0 (213/1119)

Horino et al.48 14.3 (7/49) 15.9 (7/44)a Ð 0 (0/5) 1.0000 Ð Ð Ð Ð Ð 19.2 (76/396)g

Noguchi et al.49 4.8 (4/84) Ð Ð Ð Ð 35.0 (21/60) Ð Ð Ð Ð 17.7 (25/141)

Cody and

Urban1

18.3 (22/120) Ð Ð Ð Ð 33.3 (25/75) Ð Ð Ð Ð 24.1 (47/195)

Sugg et al.3 5.1 (7/137) Ð Ð Ð Ð 43.6 (65/149) Ð Ð Ð Ð 25.2 (72/286)

Subtotalh 10.8 (267/2478) 14.2 (102/716) 8.2 (12/147) 4.7 (19/408) ,0.0001 36.8 (923/2508) 51.5 (275/534) 41.4 (120/290) 22.2 (123/554) , 0.0001 23.8 (1329/5575)

Totali 9.3 (317/3421) 13.1 (143/1090) 6.9 (24/345) 4.3 (40/937) ,0.0001 36.1 (1171/3244) 45.8 (370/808) 42.7 (185/433) 23.3 (263/1130) , 0.0001 22.5 (1639/7279)

Unselected population, we considered the population unselected when the methodology of study described no selection based on tumour location; selected population, series withselection based on tumour location. n, ratio between number of patients with IMN metastases and total number of patients. P value, Fisher's exact test.a Medial-central tumours.b Patients in this series are also included in the International Cooperative Trial (Lacour et al.'s series 1976, 198332,33) and Veronesi's 1985 series.4c Unknown location excluded.d For the calculation of the subtotals, we excluded Veronesi and Valagussa's 1981 series35 because of patient overlap with the International Cooperative Trial (Lacour et al.'s series1976, series 198332,33).

e Some patients in this series were included in the International Cooperative Trial (Lacour et al.'s series 1976, 198332,33), Veronesi and Valagussa's 1981 series35 and Veronesi's 1985series.f Some patients in this series were included in the International Cooperative Trial (Lacour et al.'s series 1976, 198332,33).g Authors presented the data by clinical stage, not axillary node status.h For the calculation of the subtotals, we excluded Bucalossi et al.'s 1971 series41 because of patient overlap with Veronesi et al.'s 1985 series4 and the International Cooperative Trial(Lacour et al.'s series 1976, 198332,33).

i For the calculation of the totals, we excluded all the series that had overlapped patients to the International Cooperative Trial (Lacour et al.'s series 1976, 198332,33).

606

J.L.B

.B

EV

ILA

CQ

UA

ET

AL.

0%

20%

40%

60%

80%

100%

0 1 2 3 4 5 6 7 8 9 10

Years

(A) Without lymph node metastases

(B) Axillary lymph node metastases only

(C) Internal mammary lymph node metastases only

(D) Axillary lymph and internal mammary lymph node metastases

A

B

C

Dp = 10–9

N = 1119

Figure 1 Ten year overall survival in patients submitted toextended radical mastectomy.4

A SELECTION ALGORITHM FOR INTERNAL MAMMARY SLN BIOPSY 607

axillary node-negative patients described by Veronesiand Valagussa35 (8.2%; 12/146). By this calculation, thefrequency of IMN metastases in subcentimetre node-negative tumours should be low, approximately 5%(0.082� 0.613).

Of course, these assumptions can be criticized. In theera when ERMs were performed, the average tumoursize was substantially larger than it is today. If weassume Veronesi and Valagussa's population of T1 andT2 axillary node-negative patients to have a higherproportion of T2 cancers than we have today, the ratiocalculated above would be lower than 0.613, so theresultant estimation of the frequency of IMN metastasesin subcentimetre node-negative tumours would be lessthan 5%. Curiously, Margottini50 in 1952, in a series of227 patients submitted to ERM, observed IMN metas-tases in about 5%, when the tumour was small (size notspecified in the original report) and there were noaxillary metastases.

Careful analysis of the results of the majority of serieson the frequency of IMN metastases by location isrequired. The majority of the studies (series) pre-selected patients with medial tumours or large lateraltumours, based on a presumed higher frequency of IMNmetastases in these patients. In unselected-populationseries, in patients with medial tumours and axillarymetastases, the frequency of IMN metastases was 38.2%compared with 22.8% in lateral tumours (P� 0.0001)(Table 1). In the same group of series, in axillary node-negative patients, the frequency of IMN metastases formedial tumours was 10.3% compared with 4.3% inlateral tumours (P� 0.0001) (Table 1). These frequen-cies suggest the importance of tumour location as apredictive factor for IMN metastases. Indeed, they alsosuggest the likelihood of an alternative route of lympha-tic flow for medial tumours. This is concordant with thefindings of our previous study,12 which demonstrated alower frequency of ALNM in medial-location tumours.

Tumour location and prognosis

Two large studies, one from Veronesi's group (Italy)with 2396 patients, the other from British Columbia(Canada) with 6781 patients, have demonstrated aworse prognosis for patients with medial-quadranttumours than for those with lateral-quadrant tumoursin terms of disease-free and overall survival.51,52 This istrue despite the fact that, in both studies, lateraltumours had more ALNM, and, in the Canadian study,the lateral tumours were larger.

A Japanese study with 1327 patients demonstratedthat medial T1, N0 tumours had a worse prognosis thanlateral T1, N0 tumours.53 Another study, by Israel et al.54

demonstrated different patterns of tumour metastasisfor medial and lateral tumours. According to this studyof 490 patients, medial tumours have more brain,subclavicular and pulmonary lymphangitis metastases

than lateral tumours.54 Both studies attributed thereported differences between medial and lateraltumours to the probable higher frequency of IMNmetastases seen in tumours in medial locations.

IMN metastases and prognosis

Numerous studies have demonstrated the prognosticvalue of IMN metastasis in breast cancer.1±4 Veronesiet al.4 demonstrated that the 10 year survival rate variedfrom 80.4% in patients with negative axillary nodes andIMNS to 30.0% in patients with involvement in bothnodal basins. Intermediate survival rates (54.6% and53.%) were found when only one or the other of thenodal stations (axillary or IMN chain) was affected (seeFig. 1). Similar results to those reported by Veronesiet al.4 were found in a large Chinese study.2 Finally, Codyand Urban1 demonstrated that the influence of IMNmetastases on disease-free survival was second only tothat of axillary lymph-node involvement and surpassedthe influence of tumour size.

While the current fifth AJCC TNM classificationsystem11 classifies IMN metastases as N3, correspondingto stage IIIB, the observation that survival rates forpatients with IM metastases only and those with axillarymetastases only are in fact similar2,4 may lead to changesin the next version of the AJCC TNM classification.

IMN AND EXTENDED RADICALMASTECTOMY

In the 1960s, a large prospective, randomized trial of theInternational Cooperative Group comparing ERM andRM demonstrated no statistical difference in overallsurvival, relapse-free survival or local-regional recur-rence between the two treatment groups at 10 yearsfollow-up.33 In the light of these results and similar


results from smaller trials,36,55 surgical oncologistsabandoned the ERM. Other factors also may havecontributed to the changes that have occurred since theearly 1980s in the surgical treatment of breast cancer:

(1) a decrease in the number of advanced breast cancercases now being seen owing to the development ofscreening programmes and a widespread increasein breast cancer education among the generalpopulation;

(2) the results of randomized trials of breast-conservation therapy;56,57

(3) the increasing acceptance of the potentially systemicnature of operable breast cancer;

(4) the advent of effective adjuvant systemic therapy;(5) decreased acceptance of morbidity associated with

the extended radical procedure.

It is worth noting, however, that despite the fact thatcomplete IMN chain dissection does not improve overallsurvival, selective IMN biopsy may be a treatment optionworth considering for a very specific subset of patients,namely those who are not considered candidates forsystemic adjuvant therapy but who may be IMN positive.For this group of patients, the knowledge of IMNpositivity obtained on IM-SLN biopsy could dramaticallychange treatment.

In our previous study, 41.2% of cases were patients withtumors less than or equal to 1 cm.12 This consider-ablepercentage reflects the increasing number of early-stagecases (patients with subcentimetre tumours) routinelyseen by breast surgeons as the result of the widespreadincrease in breast cancer education and screeningamong the general population, mentioned above. In1995, Cody and Urban 1 in fact suggested that the failureto consider IMN status in the steadily enlarging cohort ofT1, N0 breast cancers may result in the undertreatmentof a significant proportion of stage I breast patients. In thesame study, Cody and Urban1 also suggested thatsystemic adjuvant therapy be considered for T1, N0patients with central or medial tumours, as routine IMNbiopsies may be both technically cumbersome and subjectto sampling error. In the present era of SLN biopsy,however, these considerations no longer present anobstacle, as IMNs can now be localized and excised withconsiderable precision through the use of pre-operativelymphoscintigraphy and radio-guided surgery.

IMN IRRADIATION

Internal mammary node irradiation is somewhatcontroversial. The efficacy of postoperative regionalnodal irradiation (supraclavicular, axillary, and internalmammary node irradiation; without chest-wall irradi-ation) in patients with breast cancer has been examinedin a number of randomized trials.58±62 The results of

these trials demonstrated no difference in overallsurvival with the addition of regional irradiation.58±62

Some of these studies, however, demonstrated areduction on local-regional recurrences.58,59,62

Two recent randomized trials demonstrated a sig-nificant survival advantage of post-mastectomy radiationtherapy (regional lymph nodes and chest wall) in patientswith stage II and stage III breast cancer.63,64 In bothtrials, patients received adjuvant chemotherapy andwere randomized to receive radiation therapy orno therapy. Both studies included irradiation of thechest wall and supraclavicular, axillary and internalmammary nodes. Whelan et al.65 recently performed ameta-analysis of trials examining the effect of post-mastectomy adjuvant radiation. In the trials included inthis meta-analysis, the fields, arrangements or tech-niques varied, and 15 out of 18 trials included IMNirradiation.65 All included adjuvant systemic treatmentand demonstrated that adjuvant radiation therapyreduces recurrences and mortality in breast cancer.

None of the above prospective, randomizedstudies58±64 specifically addressed IMN irradiation as anendpoint. Therefore, whether IMN irradiation speci-fically and significantly contributes any survival advantageis unclear.66

In a retrospective study, Obedian and Haffty66

addressed the significance of IMN radiation therapy ina cohort of breast cancer patients treated with con-servative surgery and adjuvant radiation therapy. Themajority of the patients received adjuvant chemother-apy. They compared two groups of patients, the firsttreated with intentional IMN irradiation and the otherwithout intentional IMN irradiation. The group treatedwith IMN irradiation had statistically more ALNM, largertumours, more medial-central tumours and moreunknown margins and received less adjuvant chemo-therapy (all, P� 0.05) than the control (non-IMNirradiation) group. At median follow-up of 13 years,there were no statistical differences between the groupsin terms of overall and distant disease-free survival.66

In a similar retrospective study, Fowble et al.67

addressed the significance of IMN radiation therapy instage I±II breast cancer. In this study, of 1383 patientswho underwent wide-local excision, axillary lymph nodedissection (ALND) and radiation therapy, 114 womenhad radiation targeted to the IMNs, while 1269 did not.The 114 patients in the IMN irradiation group had moreALNM, larger tumours and a greater percentage ofmedial tumours (all, P , 0.002) than the control (non-IMN irradiation) group, indicating the former group'sworse prognosis and higher risk for IMN metastases. Atmedian follow-up of 6 years, there was no statisticaldifference in disease-specific or overall survival betweenthe IMN radiation group and the control group.67

In these two retrospective studies,66,67 the grouptreated with IMN irradiation had a worse prognosis thanthe group receiving no radiation to the IMNs. In both


studies, however, there were no statistical differences interms of overall and disease-free survival, suggestinga benefit of IMN irradiation.

In a recent review, Freedman et al.68 have suggestedthat irradiation of the IMN chain should be consideredonly for those patients with pathologically provenpositive IMNs (or IMN metastases), with the goal ofimproving regional tumour control. The present era ofSLN biopsy, and the ability to perform IM-SLN biopsy,would allow the surgeon and radiation therapist toidentify these patients better. Based on the publisheddata, however, there is no clear evidence that IMNirradiation improves survival. Therefore, further pros-pective studies are necessary to evaluate specificallywhether IMN irradiation might benefit patients with, orat high risk for, IMN metastases. In fact, the influence ofradiation therapy to the IMN chain is the subject of anongoing European Organization for Research andTreatment of Cancer Trial (EORTC-22922) and aSouthwest Oncology Group trial (SWOG-9927).

IMN BIOPSY IN THE SENTINELNODE ERA

SLN biopsy can accurately identify axillary node-positivepatients who require ALND and spare node-negativepatients an operation from which they will not benefit. Atotal of 41 series to date (all validated by a back-upALND) collectively demonstrate that the axillary SLNcan be found over 90% of the time and that the SLNaccurately predicts axillary node status in 98% of allcases and in 94% of all node-positive cases.69,70 Theadvent of SLN biopsy and, in some cases, findings ofIMN on lymphatic mapping (lymphoscintigraphy) haveprompted a recent resurgence of interest in IMNmetastases. The ability to stage the IMN with precisionwould enable us to determine those patients who wouldreceive the greatest benefit from adjuvant treatment.

Several studies have evaluated IMNs with lymphos-cintigraphy and SLN biopsy.6,8,10,23,26,71±78 All of thesestudies injected the radioisotope intraparenchymally.The lymphoscintigram demonstrated mapping to theIMNs in 0±35% of patients. In eight of these studies, anIM-SLN biopsy was performed based on the results oflymphoscintigraphy, and IMN metastases were identifiedin 0±33% of cases.8,10,23,72,75±78 When the results ofthese eight studies are combined, IM-SLNs containedmetastases in 17.7% (20/113) of the patients whounderwent IM-SLN biopsy. Curiously, the overallfrequency of IMN metastases in the SLN era (17.7%) isalmost the same as the overall frequency of IMNmetastases in the unselected-population series of thepre-SLN era (18.8%). This correspondence serves toreinforce confidence in the IM-SLN procedure (Table 1).

The length of the interval between injection of theradioisotope and performance of pre-operative

lymphoscintigraphy appears to be a crucial factor inthe detection of non-axillary SLNs. Precise identificationof extra-axillary SLNs requires that the isotope beinjected many hours prior to the lymphoscintigraphy andsurgery.77 Furthermore, use of radioisotope (ratherthan blue dye alone) is mandatory for identification ofIMNs. The use of lymphoscintigraphy and an intraopera-tive probe allows the surgeon to identify and open thecorrect parasternal intercostal space. This same ease ofidentifying the correct space is not possible with bluedye alone.

Intradermal injection is a highly accurate method ofisotope delivery for lymphatic mapping of the breast. Itmay in fact be preferable to intraparenchymal injectionowing to its greater ease of use and high rate of successfor identification of axillary SLNs.79 The preferentialpathway of dermal lymphatics is not to the internalmammary chain, however, and the identification of IMNswith intradermal injection is low.80 Therefore, inpatients at risk for IMN metastases and in whomidentification of any IM-SLN is desired, a combination ofinjection sites (intradermal and intraparenchymal) wouldlikely be more appropriate.

STATISTICAL MODEL FORPREDICTING NODALMETASTASES

In our previous, prospective study, we demonstratedthat size, lymphovascular invasion (LVI), ductal histo-logical subtype, nuclear grade, age, multifocality andtumour location within the breast were predictors ofALNM.12 We also found that UIQ tumours wereassociated with a lower frequency of ALNM thanother-quadrant tumours.12 This difference was statis-tically significant for UOQ, lower-outer quadrant andcentral tumours but of only borderline significance forlower-inner quadrant tumours. This suggests that UIQ(or medial) tumours may have alternative routes oflymphatic flow to the IMN chain. Numerous otherstudies support this hypothesis, that medial tumoursmetastasize more frequently to the IMN chain (Table 1).

If we regard these predictors of ALNM as markers ofnodal metastases in general, we can identify patients athigh risk for IMN metastases (or other extra-axillarynodal metastases) and select those patients for whomthis information will change adjuvant therapy andprognosis. By applying our prospectively validatedlogistic regression model (Table 2), it is possible todetermine the specific group of patients for whomIM-SLN biopsy might be of value. However, becauseUIQ (or medial) tumours were associated with a lowerrisk of axillary metastases in the model, the model mayunderestimate the risk of IMN metastases in UIQ(or medial) tumours. Therefore we include these medialtumours when defining `high risk' for IMN metastases.

Table 2 Simplified model: ductal carcinoma ± multivariate analysis ± logistic regression (valuesto be applied in the following equation)12

P � 1

1� exp�4:043� 0:015Aÿ 0:576T ÿ 0:229Nÿ 1:715Sÿ 1:131Lÿ 0:837Qÿ 0:359M�

Variable Value to be applied in the equation

A Age yearsT Tumour size (pathological) cmN Nuclear grade 1, 2, or 3S Tumour subtype

Colloid/medullary/tubular 0IFDC-NOS or metaplastic 1

L LVINo 0Yes 1

Q Tumour location (quadrant)UIQ 0Other 1

M MultifocalityNo 0Yes 1

IFDC-NOS, infiltrating ductal carcinoma not otherwise specified.


One limitation of the proposed model is its applica-bility to infiltrating ductal carcinomas only. The authorshave not yet developed a robust statistical model for theprediction of lymph node metastases for other types ofbreast cancer, because of the relative infrequency ofthese other types.

PROPOSED SURGICAL APPROACHFOR IM-SLN BIOPSY

In Figure 2, we propose a rationale for selecting patientsfor IM-SLN biopsy (or non-axillary SLN biopsy). Thisalgorithm allows the selection of patients who would notroutinely receive systemic adjuvant treatment but whoare at high risk for IMN metastases. This could includesubcentimetre tumours at relatively high risk of nodalmetastases according to our statistical model. Asmentioned above, this model may underestimate therisk of IMN metastases in UIQ (or medial) tumours. Inview of this potential underestimation of risk, and inview of the literature suggesting that medial tumoursmay be at higher risk of IMN metastases than tumours atother locations in the breast, we include subcentimetremedial tumours in addition to tumours calculated to be`high risk'. This flowchart could also be applied in caseswhere an SLN in another area is demonstrated on thelymphoscintigraphy, such as in the case of a sentinelsupraclavicular lymph node.

As expected, when we apply the flowchart to ourstudy population, only a very small number of patients

are candidates for IM-SLN biopsy. The calculation of theexpected frequency of IM-SLN biopsy involves threesteps, as set out in Figure 3 and explained below.

(1) Frequency calculation: step A (Fig. 3). We multiply thepercentage of patients with tumours less than orequal to 1 cm (41.2%) by the sum of the frequencyof the medial tumours (21.7%) and the proportion ofpatients at high risk for nodal metastases, defined byour statistical model for prediction of nodal metas-tases (as shown in Table 2). In the statistical modelfor prediction of nodal metastases, UIQ (or medial)tumours are associated with a lower risk for ALNM;however, because medial tumours may be morelikely to metastasize to the IMN chain, as wedescribed earlier, we forced the inclusion of medialtumours into the selection criteria for IM-SLNbiopsy. Therefore, for the calculation of theexpected frequency of IM-SLN biopsy, we haveadded the frequency of medial tumours to theproportion of patients at high risk for nodal metas-tases.

The proportion of high risk patients in our popu-lation will depend on our definition of high risk. If weconsider a 30% probability of nodal metastases to behigh risk, 41.8% of our population will be classified ashigh risk. If we consider a 50% probability as thedefinition of high risk, 20.8% of our study will beclassified as high risk. If we consider this definition tobe 70%, only 6% of our population will be classifiedas high risk.


(2) Frequency calculation: step B (Fig. 3). Once a definitionof high risk in our study has been decided on, andstep A completed, we multiply the result of step Aby 20%, which is the approximate frequency of posi-tive findings of IMNs on lymphoscintigraphy.6,26

Step A = Frequency tumor ≤ 1cm x (frequency media

Step B = Result of Step A x frequency of pos

Step C = Result Step B x proportion of ne

Expected frequency of IM-SLN

Figure 3 Algorithm for the calculation of the expected frequen

Axillary SLN intraoperativeexamination

Metastasesdetected

(Positive )

Axillary SLN biop sy / ALND Nonaxillary SLN biopsy(IM-SLN biop sy)

Tumor ≤ 1cmand

Medial tumors or high risk of nodalmetastases (based on the logistic

regression model [Table 2])

Nodes seen at extra-axillarysites

on lymphoscintigram(‘Hot’ nodes è Positive

lymphoscinti gram)

No nodes seen at extra-axillary sites

on lymphoscintigram(No ‘hot’ nodes è Negative

lymphoscinti gram)

No metastasesdetected

(Negative )

Lymphoscinti graphy study forextra -axillary sites

(i.e., IMN or supraclavicular)Intraparenchymal + intradermal injection

of Tc-99m

Figure 2 Proposed rationale for IM-SLN biopsy in patientsat high risk for IMN metastasis.

(3) Frequency calculation: step C (Fig. 3). Next, this num-ber (result of step B) is multiplied by 85%, which isthe proportion of negative frozen sections on SLNbiopsy.81 The result of step C equals the expectedfrequency of IM-SLN biopsy.

Therefore, when a high risk definition of 30%probability of nodal metastases is used, the expectedfrequency of IM-SLN biopsy in our population is 4.4%.At a definition of 50% the frequency is 3%, and ata definition of 70% the frequency of IM-SLN biopsy is 2%.

At present, we have no data on the proportion ofpositive IMNs in the group that was selected using oursuggested approach. We can assume, however, that itwould be relatively high, based on the highly selectiveprocess by which these patients were identified ascandidates for biopsy.

The IM-SLN biopsy rationale outlined above has thepotential to change the adjuvant therapy in 2±4.4% of allpatients in our study. It is interesting to note that thispercentage range is approximately the same as theexpected incidence of IMN metastases in axillary node-negative patients with tumours less than or equal to1 cm, described earlier (�5%).

One could argue that the incidence of IMN metastasesin axillary node-negative patients with tumours lessthan or equal to 1 cm is very low and that this procedurecould therefore be avoided. This may be a reasonableargument; however, although the number of patientswho would undergo IM-SLN biopsy seems small, it isin fact of the same order as the number of axillary node-negative patients who benefit from adjuvant tamoxifenand chemotherapy (5%82 and 7%83 respectively).

Similarly, one could argue that IM-SLN biopsy shouldbe considered routinely in all subcentimetre tumours(41% in our population) which demonstrate a `hot' IMNon pre-operative lymphoscintigram (*20%), constitut-ing approximately 8% of our total population. With thisless selective approach, however, a smaller proportion

l tumors + frequency high risk of nodal metastases)

itive findings of IMN on lymphoscintigraphy

gative frozen sections on SLN biopsy

bio psy = Resul t of Step C

cy of IM-SLN biopsy: based on Figure 2; see text for details.


of patients undergoing IM-SLN biopsy would have IMNmetastases and a resultant change in managementrecommendations. The proposed algorithm (Fig. 2),would allow us to define the specific, small population ofpatients most likely to benefit from an IM-SLNprocedure. We realize that this more selective approachwould not result in absolutely all IMN metastases beingdetected. However, this approach seems to be areasonable compromise between identifying the rela-tively rare patient whose IMN metastases would changeher care and minimizing unnecessary surgery.

Furthermore, it is possible that some axillary node-negative patients who have a worse outcome are in factIMN positive. In the future, we may be able to delineateadjuvant systemic treatment more effectively, in whichcase systemic treatment in true node-negative patientsmay be revised.

According to the National Comprehensive CancerNetwork,7 axillary node-negative patients with tumoursless than 1 cm in size are usually not candidates foradjuvant systemic therapy. A recent report fromNational Surgical Adjuvant Breast and Bowel Project(NSABP), however, suggests that adjuvant systemictreatment may be considered for the treatment ofsubcentimetre axillary node-negative breast cancerpatients.84 To identify this subgroup of patients, thisstudy executed a subset analysis combining five NSABPtrials. Further prospective, randomized studies aresuggested to confirm these results. If adjuvant systemictreatment did in fact prove to be valuable for all patientswith invasive carcinomas of the breast despite theiraxillary node status, the importance of IMN status wouldbe decreased.

Applying our suggested algorithm to non-axillary SLNbiopsy and identifying patients with non-axillary lymphnode metastases would also help the radiation therapistto determine better which patients might be candidatesfor IMN and supraclavicular irradiation.

RISKS, COMPLICATIONS ANDCONSIDERATIONS FORIM-SLN BIOPSY

Internal mammary SLN biopsy is a simple procedure butnot one routinely done. It should be performed by asurgeon with expertise or, initially, with assistance of athoracic surgeon. As with any surgical procedure,IM-SLN biopsy does have certain risks. Some importantconsiderations and potential complications that must beconsidered in relation to IM-SLN biopsy include thefollowing:

(1) Pneumothorax secondary to violation of the pleura canoccur in IM-SLN biopsy cases in which the IMN liesin close proximity to the pleura. In one recentstudy,9 pneumothorax occurred in one out of six

patients who underwent IM-SLN biopsy; in this case,it was evacuated intraoperatively, and the patienthad no post-operative complications.

(2) Bleeding can occur if the internal mammary vesselretracts secondary to accidental transection withoutpre-clamping. This situation can result in haemor-rhaging that may be difficult to control.

(3) Additional scarring is possible owing to the necessity ofa second incision to access the intercostal space ofthe IMN chain. This is particularly likely in cases inwhich the tumour is located other than mediallywithin the breast.

(4) Resection of costal cartilage may be necessary in somecases where the IM-SLN is located behind the costalcartilage.22 This can increase the difficulty of theprocedure and negatively affect the aesthetic result.

(5) Exclusion on the basis of prior coronary bypass surgerymay be due to resultant scarring from the priorsternotomy, which would increase the difficulty ofidentifying the IMN. In this situation, there is also arisk of injuring the IM artery during IM-SLN biopsy.Also, injury of the IM artery could preclude its use infuture cardiac surgery.

(6) Increased cost and time of pre-operative lymphoscinti-graphy can be incurred when institutional protocolrequires that a physician perform the intraparen-chymal injection of radiolabelled dyes, as is the casein our own institution.

CONCLUSIONS

IMN metastasis is a prognostic factor in breast cancer.The value of IMN status is potentially greatest forpatients who are not considered candidates foradjuvant systemic treatment by the current treatmentrecommendations7,9,24 (patients with subcentimetretumours and negative axillary nodes). Not all patientswith these characteristics would benefit from IM-SLNbiopsy, however. In this report the authors suggest aselective approach to IM-SLN biopsy that involvesidentifying and selecting for this procedure thosepatients most likely to have IMN metastases, whilesimultaneously identifying and excluding patients withlower risk of IMN metastases. Through the identificationof this particular subset of patients by means of thepredictive model, we could perform IM-SLN biopsy withfar greater selectivity, thereby improving the proce-dure's overall value and feasibility.

In the future, we may be able to delineate adjuvantsystemic and radiation treatment more effectively.Further prospective studies are needed to addressspecifically the frequency and predictive factors of IMNmetastases, and their prognostic implications, in the SLNera. Finally, the risk±benefits of IM-SLN biopsy should beconsidered for each patient, and an individualizedapproach to this procedure is suggested.


REFERENCES

1. Cody HS III, Urban JA. Internal mammary node status: a majorprognosticator in axillary node-negative breast cancer. Ann SurgOncol 1995; 2: 32±7.

2. Li YYK, Shen ZZ. An analysis of 1,242 cases of extended radicalmastectomy. Breast, Dis Breast 1984; 10: 10±19.

3. Sugg SL, Ferguson DJ, Posner MC. Should internal mammary nodesbe sampled in the sentinel lymph node era? Ann Surg Oncol 2000; 7:188±92.

4. Veronesi U, Cascinelli N, Greco M et al. Prognosis of breast cancerpatients after mastectomy and dissection of internal mammarynodes. Ann Surg 1985; 202: 702±7.

5. Veronesi U, Marubini E, Mariani L et al. The dissection of internalmammary nodes does not improve the survival of breast cancerpatients. 30-Year results of a randomised trial. Eur J Cancer 1999;35: 1320±5.

6. Uren RF, Howman-Giles RB, Thompson JF et al. Mammarylymphoscintigraphy in breast cancer. J Nucl Med 1995; 36:1775±80.

7. Anderson BO, Bensinger W, Cox CE et al. NCCN practiceguidelines for breast cancer. Oncology (Huntingt) 2000; 14: 33±49.

8. Jansen L, Doting MH, Rutgers EJ et al. Clinical relevance of sentinellymph nodes outside the axilla in patients with breast cancer.Br J Surg 2000; 87: 920±5.

9. Johnson N, Soot L, Nelson J et al. Sentinel node biopsy and internalmammary lymphatic mapping in breast cancer. Am J Surg 2000;179: 386±8.

10. Noguchi M, Tsugawa K, Miwa K. Internal mammary chain sentinellymph node identification in breast cancer. J Surg Oncol 2000; 73:75±80.

11. American Joint Committee on Cancer. AJCC Cancer Staging Manual(5th edn). Philadelphia, PA: Lippincott-Raven, 1997: 171±80.

12. Bevilacqua JLB, Cody HS III, MacDonald KA et al. A model forpredicting axillary node metastases based on 2000 sentinelnode procedures and tumour location. Eur J Surg Oncol 2002; 28:490±500.

13. Fein DA, Fowble BL, Hanlon AL et al. Identification of women withT1±T2 breast cancer at low risk of positive axillary nodes. J SurgOncol 1997; 65: 34±9.

14. Gann PH, Colilla SA, Gapstur SM et al. Factors associated withaxillary lymph node metastasis from breast carcinoma: descriptiveand predictive analyses. Cancer 1999; 86: 1511±19.

15. Halverson KJ, Taylor ME, Perez CA et al. Management of the axillain patients with breast cancers one centimeter or smaller. Am J ClinOncol 1994; 17: 461±6.

16. Maibenco DC, Weiss LK, Pawlish KS et al. Axillary lymph nodemetastases associated with small invasive breast carcinomas.Cancer 1999; 85: 1530±6.

17. Olivotto IA, Jackson JS, Mates D et al. Prediction of axillary lymphnode involvement of women with invasive breast carcinoma:a multivariate analysis. Cancer 1998; 83: 948±55.

18. Urban JA. Clinical experience and results of excision of the internalmammary lymph node chain in the primary operable breast cancer.Cancer 1959; 12: 14±22.

19. Velanovich V, Szymanski W. Lymph node metastasis in breastcancer: common prognostic markers lack predictive value.Ann Surg Oncol 1998; 5: 613±19.

20. Voogd AC, Coebergh JW, Repelaer van Driel OJ et al. The risk ofnodal metastases in breast cancer patients with clinically negativelymph nodes: a population-based analysis. Breast Cancer Res Treat2000; 62: 63±9.

21. Turner-Warwick RT. The lymphatics of the breast. Br J Surg 1959;46: 574±82.

22. Stibbe EP. The internal mammary lymphatic glands. J Anat 1918; 52:257±64.

23. Borgstein PJ, Pijpers R, Comans EF et al.Sentinel lymph node biopsyin breast cancer: guidelines and pitfalls of lymphoscintigraphy andgamma probe detection. J Am Coll Surg 1998; 186: 275±83.

24. Guenther JM, Collins JC, Barnes G Jr et al. Selectivelymphoscintigraphy: a necessary adjunct to dye-directed sentinelnode biopsy for breast cancer? Arch Surg 2000; 135: 1101±5.

25. Kamath V, Ramnath EM, Shivers S et al. Role of lymphoscintigraphyfor lymphatic mapping in women with breast cancer (abstract 1).

In: 53rd Annual Cancer Symposium of Society of SurgicalOncology, New Orleans, LA, 16±19 March 2000 (abstract).

26. O'Hea BJ, Hill AD, El Shirbiny AM et al. Sentinel lymph nodebiopsy in breast cancer: initial experience at MemorialSloan±Kettering Cancer Center. J Am Coll Surg 1998; 186:423±7.

27. Urban JA, Baker HW. Radical mastectomy in continuity withen bloc resection of the internal mammary lymph node chain ± anew procedure for primary operable cancer of the breast. Cancer1952; 5: 992±1008.

28. Caceres E. Incidence of metastases in the internal mammary chainin operable cancer of the breast. Surg Gynecol Obstet 1959; 108:715±20.

29. Sugarbaker ED. Extended radical mastectomy. J AM Med Assoc1964; 187: 96±8.

30. Dahl-Iversen E, Tobiassen T. Radical mastectomy with parasternaland supraclavicular dissection for mammary carcinoma. Ann Surg1969; 170: 889±91.

31. Livingston SF, Arlen M. The extended extrapleural radicalmastectomy: its role in the treatment of carcinoma of thebreast. Ann Surg 1974; 179: 260±5.

32. Lacour J, Bucalossi P, Caceres E et al. Radical mastectomy versusradical mastectomy plus internal mammary dissection. Five-yearresults of an international cooperative study. Cancer 1976; 37:206±14.

33. Lacour J, Le M, Caceres E et al. Radical mastectomy versus radicalmastectomy plus internal mammary dissection. Ten year results ofan international cooperative trial in breast cancer. Cancer 1983; 51:1941±3.

34. Rai S, Jindal RP, Nijjer SS. Internal mammary lymph node metastasisin carcinoma. Indian J Cancer 1980; 17: 169±71.

35. Veronesi U, Valagussa P. Inefficacy of internal mammary nodesdissection in breast cancer surgery. Cancer 1981; 47: 170±5.

36. Morimoto T, Monden Y, Takashima S et al. Five-year results of arandomized clinical trial comparing modified radical mastectomyand extended radical mastectomy for stage II breast cancer. SurgToday 1994; 24: 210±14.

37. McDonald JJ, Haagensen CD, Stout AP. Metastasis from mammarycarcinoma to the supraclavicular and internal mammary lymphnodes. Surgery 1953; 34: 521±42.

38. Wyatt JP, Sugarbaker ED, Stanton MF. Involvement of internalmammary lymph nodes in carcinoma of the breast. Am J Pathol1955; 31: 143±50.

39. Hutchinson WB, Kiriluk LB. Internal mammary node investigationin carcinoma of the breast. Am J Surg 1956; 92: 151±5.

40. Pavrovsky J, Tersip K, Palecek L. Results of radical operation formammary gland carcinoma with revision of the parasternal space.Int Surg 1969; 51: 509±14.

41. Bucalossi P, Veronesi U, Zingo L et al. Enlarged mastectomy forbreast cancer. Review of 1,213 cases. Am J Roentgenol Radium TherNucl Med 1971; 111: 119±22.

42. Urban JA, Castro EB. Selecting variations in extent of surgicalprocedure for breast cancer. Cancer 1971; 28: 1615±23.

43. Richardson JD, Meeker WR, Jewell WR. The morbidity of theextended radical versus the standard radical mastectomy. J Ky MedAssoc 1972; 70: 537±9.

44. Caceres E. Radical and extended surgery for cancer. Surg Annu1973; 5: 55±67.

45. Handley RS. Carcinoma of the breast. Ann R Coll Surg Engl 1975; 57:59±66.

46. Donegan WL. The influence of untreated internal mammarymetastases upon the course of mammary cancer. Cancer 1977; 39:533±8.

47. Deemarski LY, Seleznev IK. Extended radical operations on breastcancer of medial or central location. Surgery 1984; 96: 73±7.

48. Horino T, Fujita M, Ueda N et al. Efficacy of internal mammarynode dissection in the treatment of breast cancer. Jpn J Clin Oncol1991; 21: 422±7.

49. Noguchi M, Ohta N, Koyasaki N et al. Reappraisal of internalmammary node metastases as a prognostic factor in patients withbreast cancer. Cancer 1991; 68: 1918±25.

50. Margottini M. Recent developments in the surgical treatmentof breast carcinoma. Acta Unio Internat Contra Cancrum 1952; 8:176±8.


51. Lohrisch C, Jackson J, Jones A et al. Relationship between tumorlocation and relapse in 6,781 women with early invasive breastcancer. J Clin Oncol 2000; 18: 2828±35.

52. Zucali R, Mariani L, Marubini E et al. Early breast cancer: evaluationof the prognostic role of the site of the primary tumor. J Clin Oncol1998; 16: 1363±6.

53. Asaga T, Masuzawa T, Yoshida A et al. Location of breast cancer asa prognostic factor. Breast Cancer Res Treat 1994; 32: S48 (abstract).

54. Israel L, Gogeul E, Morere JF et al: Distinct metastatic pattern inbreast tumor of inner and central areas. In: Proceedings of theAnnual Meeting of the American Society of Clinical Oncology. Vol.14. 1995; A110 (abstract).

55. Meyer W. An improved method of the radical operation forcarcinoma of the breast. Med Rec 1894; 46: 746±9.

56. Fisher B, Bauer M, Margolese R et al. Five-year results of arandomized clinical trial comparing total mastectomy andsegmental mastectomy with or without radiation in thetreatment of breast cancer. N Engl J Med 1985; 312: 665±73.

57. Veronesi U, Saccozzi R, Del Vecchio M et al. Comparing radicalmastectomy with quadrantectomy, axillary dissection, andradiotherapy in patients with small cancers of the breast. N EnglJ Med 1981; 305: 6±11.

58. Fisher B, Slack NH, Cavanaugh PJ et al. Postoperative radiotherapyin the treatment of breast cancer: results of the NSABP clinicaltrial. Ann Surg 1970; 172: 711±32.

59. Host H, Brennhovd IO, Loeb M. Postoperative radiotherapy inbreast cancer ± long-term results from the Oslo study. Int J RadiatOncol Biol Phys 1986; 12: 727±32.

60. Palmer MK, Ribeiro GG. Thirty-four year follow up of patientswith breast cancer in clinical trial of postoperative radiotherapy.Br Med J (Clin Res Edn) 1985; 291: 1088±91.

61. Veronesi U, Zucali R, Luini A. Local control and survival in earlybreast cancer: the Milan trial. Int J Radiat Oncol Biol Phys 1986; 12:717±20.

62. Yamashita T, Hurukawa M, Sekiguchi K et al. Efficacy of loco-regional lymphnodes irradiation after mastectomy for breastcancer with biopsy proven parasternal lymph node metastases ±a randomized study (abstract 1066). In: 38th Annual AmericanSociety for Therapeutic Radiology and Oncology Meeting, LosAngeles, CA, 27±30 October 1996 (abstract).

63. Overgaard M, Hansen PS, Overgaard J et al. Postoperativeradiotherapy in high-risk premenopausal women with breastcancer who receive adjuvant chemotherapy. Danish BreastCancer Cooperative Group 82b. N Engl J Med 1997; 337: 949±55.

64. Ragaz J, Jackson SM, Le N et al. Adjuvant radiotherapy andchemotherapy in node-positive premenopausal women with breastcancer. N Engl J Med 1997; 337: 956±62.

65. Whelan TJ, Julian J, Wright J et al. Does locoregional radiationtherapy improve survival in breast cancer? A meta-analysis. J ClinOncol 2000; 18: 1220±9.

66. Obedian E, Haffty BG. Internal mammary nodal irradiation inconservatively-managed breast cancer patients: is there a benefit?Int J Radiat Oncol Biol Phys 1999; 44: 997±1003.

67. Fowble B, Hanlon A, Freedman G et al. Internal mammary nodeirradiation neither decreases distant metastases nor improvessurvival in stage I and II breast cancer. Int J Radiat Oncol Biol Phys2000; 47: 883±94.

68. Freedman GM, Fowble BL, Nicolaou N et al. Should internalmammary lymph nodes in breast cancer be a target for theradiation oncologist? Int J Radiat Oncol Biol Phys 2000; 46: 805±14.

69. Cody HS III, Borgen PI. State-of-the-art approaches to sentinelnode biopsy for breast cancer: study design, patient selection,technique, and quality control at Memorial Sloan±KetteringCancer Center. Surg Oncol 1999; 8: 85±91.

70. Cody HS III. Clinical aspects of sentinel node biopsy. Breast CancerRes 2001; 3: 104±8.

71. Burak WE Jr, Walker MJ, Yee LD et al. Routine preoperativelymphoscintigraphy is not necessary prior to sentinel node biopsyfor breast cancer. Am J Surg 1999; 177: 445±9.

72. Cox CE, Dupont E, Peltz ES et al. Preoperative detection ofinternal mammary lymph nodes using lymphoscintigraphy andintraoperative lymphatic mapping for breast cancer mandatesexcision (abstract 67). In: 53rd Annual Cancer Symposium of theSociety of Surgical Oncology, Washington, DC, 15±18 March 2001(abstract).

73. Krag D, Weaver D, Ashikaga T et al. The sentinel node in breastcancer ± a multicenter validation study. N Engl J Med 1998; 339:941±6.

74. Roumen RM, Geuskens LM, Valkenburg JG. In search of the truesentinel node by different injection techniques in breast cancerpatients. Eur J Surg Oncol 1999; 25: 347±51.

75. Snider H, Dowlatshahi K, Fan M et al. Sentinel node biopsy in thestaging of breast cancer. Am J Surg 1998; 176: 305±10.

76. van der Ent FW, Kengen RA, van der Pol HA et al. Sentinel nodebiopsy in 70 unselected patients with breast cancer: increasedfeasibility by using 10 mCi radiocolloid in combination with a bluedye tracer. Eur J Surg Oncol 1999; 25: 24±9.

77. Winchester DJ, Sener SF, Winchester DP et al. Sentinellymphadenectomy for breast cancer: experience with 180consecutive patients: efficacy of filtered technetium 99m sulphurcolloid with overnight migration time. J Am Coll Surg 1999; 188:597±603.

78. Zurrida S, Galimberti V, Orvieto E et al. Radioguided sentinel nodebiopsy to avoid axillary dissection in breast cancer. Ann Surg Oncol2000; 7: 28±31.

79. Boolbol SK, Fey JV, Borgen PI et al. Intradermal isotope injection: ahighly accurate method of lymphatic mapping in breast carcinoma.Ann Surg Oncol 2001; 8: 20±4.

80. Linehan DC, Hill AD, Akhurst T et al. Intradermal radiocolloid andintraparenchymal blue dye injection optimize sentinel nodeidentification in breast cancer patients. Ann Surg Oncol 1999; 6:450±4.

81. Weiser MR, Montgomery LL, Susnik B et al. Is routineintraoperative frozen-section examination of sentinel lymphnodes in breast cancer worthwhile? Ann Surg Oncol 2000; 7:651±5.

82. Tamoxifen for early breast cancer: an overview of the randomisedtrials. Early Breast Cancer Trialists' Collaborative Group. Lancet1998; 351: 1451±67.

83. Polychemotherapy for early breast cancer: an overview of therandomised trials. Early Breast Cancer Trialists' CollaborativeGroup. Lancet 1998; 352: 930±42.

84. Fisher B, Dignam J, Tan-Chiu E et al. Prognosis and treatmentof patients with breast tumours of one centimeter or lessand negative axillary lymph nodes. J Natl Cancer Inst 2001; 93:112±20.

Accepted for publication 1 Feburary 2002

EJSO 2002; 28: 899


CORRIGENDA

Corrigendum for Eur J Surg Oncol 2002;28: 603±14

Bevilacqua J, Gucciardo G, Cody H, MacDonald K, Sacchini V, Borgen P, Van Zee K. A selection algorithm for internalmammary sentinel lymph node biopsy in breast cancer. Eur J Surg Oncol 2002; 28: 603±14.

All research presented in this article was done by the Departments of Surgery and Pathology at Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021. This paper is the result of research carried outsolely at the aforementioned institution.

JoseÂ Luiz B. Bevilacqua, MD is currently affiliated with Hospital SõÂrio Libanes, Rua Adma Jafet 91, SaÄo Paulo, SP Brazil01308-050; Instituto Brasileiro de Controle do CaÃncer, Avenida AlcaÃntara Machado 2576, SaÄo Paulo, SP Brazil 03102-002; and Disciplina de Cirurgia Geral, Departamento de Cirurgia, Faculdade de Medicina da Univerdidade de SaÄoPaulo, Av. Dr Arnaldo, 455 SaÄo Paulo, SP Brazil 01246-903.

doi:10.1053/ejso.2002.1388

Corrigendum for Eur J Surg Oncol 2002;28: 490±500

Bevilacqua J, Cody H, MacDonald K, Tan L, Borgen P, Van Zee K. A model for predicting axillary node metastasesbased on 2000 sentinel node procedures and tumour position. Eur J Surg Oncol 2002; 28: 490±500.

The authors wish it stated that the correct title of this article is as follows:

`A prospective validated model for predicting axillary node metastases based on 2,000 sentinel node procedures: the role oftumour location.'

All research presented in this article was done by the Departments of Surgery and Pathology at Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021. This paper is the result of research carried outsolely at the aforementioned institution.

JoseÂ Luiz B. Bevilacqua, MD is currently affiliated with Hospital SõÂrio Libanes, Rua Adma Jafet 91, SaÄo Paulo, SP Brazil01308-050; Instituto Brasileiro de Controle do CaÃncer, Avenida AlcaÃntara Machado 2576, SaÄo Paulo, SP Brazil 03102-002; and Disciplina de Cirurgia Geral, Departamento de Cirurgia, Faculdade de Medicina da Univerdidade de SaÄoPaulo, Av. Dr Arnaldo, 455 SaÄo Paulo, SP Brazil 01246-903.

0748±7983/02/$35.00 # 2002 Elsevier Science Ltd. All rights reserved.

Comprehensive Review of the Management ofInternal Mammary Lymph Node Metastases inBreast CancerNancy Klauber-DeMore, MD, Jose Luiz B Bevilacqua, MD, Kimberly J Van Zee, MD, FACS,Patrick Borgen, MD, Hiram S Cody III, MD, FACS

At the suggestion of Sampson Handley, Philip Stibbe,1

in 1918, characterized the spread of breast cancer to theinternal mammary lymph node (IMN) chain as a pat-tern of metastasis. In 1952, Wangensteen2 published thetechnique of extended radical mastectomy. This tech-nique, developed in an attempt to increase local controland survival in breast cancer, combined the classic radi-cal mastectomy with resection of the IMN chain fromthe fifth intercostal space to the base of the neck. Earlyinvestigations in the 1960s and 1970s focused on thesignificance of IMN metastases, but the extended radicalmastectomy was subsequently abandoned when pro-spective, randomized trials failed to demonstrate a sig-nificant increase in survival.

The resurgence in interest in IMN metastases seenover the last several years is a result of the advent oflymphatic mapping and sentinel lymph node (SLN) bi-opsy. Lymphoscintigrams show mapping to internalmammary lymph nodes in up to 35%3 of patients, afinding that raises questions about the role for internalmammary sentinel lymph node (IM SLN) biopsy. In anattempt to formulate guidelines for the management ofIM SLNs, we critically reviewed the historical and newevidence concerning the significance of IMN metasta-ses. In this article, we evaluate the incidence of IMNmetastases, the effect of IMN metastases on survival, theeffect of IMN dissection and radiation on survival, andrecent results of IMN lymphoscintigraphy and SLNbiopsy.

Incidence of internal mammary and axillary lymphnode metastasesClinical review based on the results of reports of ex-tended radical mastectomy confirms the importance ofthe IMN chain as a major pathway of lymphatic drain-age. Among seven studies of extended radical mastec-tomy (ERM)4-10 (totaling 4,172 breast cancer patients),the estimated incidence of IMN metastasis at the time ofsurgery ranged from 18% to 33% (Table 1). Fourteenpercent to 24% of all patients had both IMN and axil-lary node (AN) metastases, whereas only 2% to 11% ofpatients had IMN metastases alone (Table 1). Of the2,107 AN-positive patients, 29% to 52% had IMN me-tastases (Table 2). Of the 2,065 AN-negative patients,4% to 18% had IMN metastases (Table 2).

Incidence of internal mammary node metastases byprimary tumor location and axillary node status:Axilla negativeThe incidence of IMN metastases in AN-negative pa-tients based on location of the primary tumor was eval-uated in three studies. In two reports, there was no dif-ference in the incidence of IMN metastases betweenmediocentral tumors and lateral tumors.11,12 In contrast,in the third review, Li and Shen7 reported a higher inci-dence for mediocentral tumors. Among the total of 634AN-negative patients with mediocentral tumors fromthe three studies, the incidence of IMN metastases was8% to 10%; among the total 701 AN-negative patientswith lateral tumors, the overall incidence was 2% to13% (Table 3).

Incidence of internal mammary node metastases byprimary tumor location and axillary node status:Axilla positiveEvaluation of IMN metastases in AN-positive patientsby tumor location showed a higher incidence of IMNmetastases for mediocentral tumors compared with lat-

No competing interests declared.

Received January 24, 2001; Revised April 20, 2001; Accepted May 18, 2001.From the Department of Surgery, Memorial Sloan-Kettering Cancer Center,New York, NY.Correspondence address: Hiram S Cody III, MD, FACS, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, MRI-1026 New York, NY 10021.

547© 2001 by the American College of Surgeons ISSN 1072-7515/01/$21.00Published by Elsevier Science Inc. PII S1072-7515(01)01040-7

eral tumors in four separate studies. Among the total of958 AN-positive patients with mediocentral tumors fromthe four studies, the incidence of IMN metastases was 36%to 49%; among the 1,143 AN-positive patients with lateraltumors, the incidence was 22% to 26% (Table 4).5,7,11,12

Overall survival by status of lymph nodeinvolvement after extended radical mastectomyAnalysis of survival based on lymph node status in pa-tients who underwent extended radical mastectomyshowed that the greatest survival was seen in patientswho had no AN or IMN metastases4-10,13 (Table 5). Thesurvival of IMN-positive/AN-negative patients was gen-erally similar to that of AN-positive only (IMN nega-tive) patients. Patients with both AN and IMN metas-tases consistently had the worst survival (Table 5).Veronesi and colleagues13 recently reported the results at30 years’ followup of their randomized trial of radicalmastectomy versus extended radical mastectomy.Among 342 patients, the annual death rate (ADR) waslowest for AN-negative/IMN-negative patients (ADR�0.031). Axillary node-negative/IMN-positive patientshad a higher annual death rate (ADR�0.055), whichwas not statistically significantly different from that of

AN-positive/IMN-negative patients (ADR�0.077).The highest death rate was seen with AN-positive/IMN-positive patients (ADR�0.163).

Extended radical mastectomy versusradical mastectomyTo address whether dissection of IMNs prolongs sur-vival, a prospective, randomized trial comparing ex-tended radical mastectomy to radical mastectomy wasconducted from 1963 to 1968 by five participating in-stitutions that, together, formed the International Co-operative Study.11 The trial, which analyzed 1,443 pa-tients, showed no significant difference between the twogroups in overall 5-year survival or local-regional recur-rence (Table 6). After 5 years, the International Cooper-ative Study ceased to follow the patients collectively, andeach center subsequently reported their individual 10-or 15-year followup results. Lacour and associates14,15

found no difference in the 10- and 15-year overall sur-vival between the groups, but there was a statisticallysignificant difference in local-regional recurrence (24%for radical mastectomy versus 14% for extended radicalmastectomy) at 10 years.14 This result was not seen at

Table 1. Frequency of Metastases to Internal Mammary and Axillary Lymph Node Chains in Patients Treated by ExtendedRadical Mastectomy

First author Year n

Total IMN� AN� IMN� AN� IMN� AN� IMN� AN� IMN�

n % n % n % n % n %

Urban4 1971 725 237 33 177 24 60 8 324 45 164 23Handley5 1975 500 140 28 117 23 23 5 185 37 175 35Donegan6 1977 105 20 19 18 17 2 2 45 43 40 38Li7 1984 1,242 221 18 191 15 30 2 577 46 444 36Veronesi8 1985 1,119 213 19 162 14 51 5 512 46 394 35Cody9 1995 195 49 25 27 14 22 11 98 50 48 25Sugg10 2000 286 72 25 66 23 6 2 130 45 84 29

AN, axillary node; IMN, internal mammary node.

Table 2. Incidence of Internal Mammary Lymph Node Metastases Based on Axillary Lymph Node Status in Patients Treatedwith Extended Radical Mastectomy

First author YearTotal

AN�, n

IMN�/AN�Total

AN�, n

IMN�/AN�

n % n %

Urban4 1971 341 177 52 384 60 16Handley5 1975 292 117 40 208 23 11Donegan6 1977 58 18 31 47 2 4Li7 1984 635 191 30 607 30 5Veronesi8 1985 556 162 29 563 51 9Cody9 1995 75 27 36 120 22 18Sugg10 2000 150 66 44 136 6 4

AN, axillary node; IMN, internal mammary node.

548 Klauber-DeMore et al Internal Mammary Lymph Node Metastases J Am Coll Surg

Lacour’s 15-year followup,15 nor was it reproduced inother trials.12,16 Koszarowski16 and Veronesi and col-leagues12,13 (all of whom had participated in the Interna-tional Cooperative Trial) also reported no difference inoverall survival and local regional recurrence at 5-year,16

10-year,12 and 30-year13 followup, respectively.Meier and associates17 performed a prospective, ran-

domized trial from 1973 to 1982 that showed no differ-ence in 10-year overall survival between radical mastec-tomy and extended radical mastectomy.

Collectively, these prospective, randomized trials havedemonstrated that surgical resection of IMNs does notimprove overall survival and local-regional recurrence(Table 6).

Extended radical mastectomy versus radicalmastectomy: Results for mediocentral tumorsSubset analyses performed by the International Cooper-ative Study,11 Lacour and colleagues,15 and Meier andcoworkers17 showed that extended radical mastectomyimproved the results for the subgroups of patients withmediocentral tumors. Criticism of this conclusion hasbeen the increased chance of a type one error with thesesubset analyses. In contrast, Veronesi and Valagussa12

and Koszarowski16 saw no difference in survival for pa-tients with mediocentrally located cancers who weretreated with extended radical mastectomy comparedwith those treated with radical mastectomy (Table 7).

Trials of prophylactic internal mammarynode irradiationTwo recent randomized trials have demonstrated a sig-nificant benefit to postmastectomy radiation (which in-cluded four different radiation fields: chest wall and ax-illary, supraclavicular, and IMN chains) in high-riskbreast cancer patients.18,19 Additionally, a recent meta-analysis has shown that radiation (either postmastec-tomy or postlumpectomy) reduces both recurrence andmortality.20 These studies have renewed interest in IMNirradiation and in the relationship between local controland survival. But since the four-field technique of chest-wall irradiation was used in the recent randomized tri-als,18,19 and a variety of radiation fields in the metaanaly-ses,20 the contribution of IMN irradiation alone isunclear and remains controversial.

There have been several prospective, randomized tri-als of postmastectomy radiation that excluded coverageof the chest wall.21-25 These studies are useful to isolatethe benefit of regional IMN and supraclavicular nodeirradiation. Veronesi and colleagues23 reported 56 pa-tients with positive ANs who had undergone eitherquadrantectomy with AN dissection (with radiation tothe ipsilateral breast) or Halsted’s mastectomy. Patientswere randomized to receive either postoperative radia-tion therapy to the supraclavicular and IMNs or no fur-ther treatment. Survival curves at 10 years showed noadvantage with adjuvant radiation therapy to these sites.

Palmer and Ribeiro25 reported a prospective, random

Table 3. Incidence of Internal Mammary Node Metastases by Primary Tumor Location and Axillary Node Status:Axilla Negative

First author Year MediocentralMediocentral

IMN� Lateral Lateral IMN�

n % n n % n

International Coop Study11 1976 288 10 30 317 9 25Veronesi12 1981 68 10 7 93 13 9Li7 1984 278 8 22 291 2 6

IMN, internal mammary node.

Table 4. Incidence of Internal Mammary Node Metastases by Primary Tumor Location and Axillary Node Status: Axilla Positive

First author Year MediocentralMediocentral

IMN� Lateral Lateral IMN�

n % n n % n

Handley5 1975 270 49 131 265 22 57International Coop Study11 1976 329 36 119 457 22 100Veronesi12 1981 77 36 28 104 26 27Li7 1984 282 36 101 317 25 80

IMN, internal mammary node.

549Vol. 193, No. 5, November 2001 Klauber-DeMore et al Internal Mammary Lymph Node Metastases

ized trial of 741 patients with operable breast cancer whounderwent Halsted’s mastectomy between 1952 and1955. These patients were stratified for AN status andrandomized to receive either postoperative radiation tothe axilla, supraclavicular region, and parasternal area orradiation only if the patient developed local recurrence.With 30 years’ followup, there was no difference in sur-vival between the two groups.

Fisher and associates21 reported the National SurgicalAdjuvant Breast Project (NSABP) B-02, where patientswere randomized to receive radical mastectomy with orwithout radiation. Radiation was limited to the internal

mammary, apical axillary, and supraclavicular nodes.The 5-year survival was 56% without radiation, versus62% for radical mastectomy with radiation (p�NS).Fisher and coworkers26 subsequently reported theNSABP B-04 trial, in which, again, there was no survivalbenefit for patients randomized to postmastectomy ra-diation that included the IMNs, even in the subgroupwith inner-quadrant tumors.

Host and associates24 reported longterm results of theOslo Study, in which women who had undergone Hal-sted’s radical mastectomy were randomized to postoper-ative radiation (n�562) versus no further treatment

Table 5. Overall Survival by Status of Lymph Node Involvement after Extended Radical Mastectomy

First author Year Survival measurement Statistical method FollowupAN� IMN�

(%)AN� IMN�

(%)AN� IMN�

(%)AN� IMN�

(%)

Urban4 1971 10-y OS ND ND 75 53 52 21Handley5 1975 10-y OS ND ND 64 36 39 7Donegan6 1977 10-y OS ND 10 y, 97%,

mean 13 y40 21 0 6

Li7 1984 10-y OS ND 10 y, 94% 85 44 61 2720-y OS 75 37 41 24

Veronesi8 1985 10-y OS ND ND 80 55 53 30Veronesi13 1999 30-y OS Kaplan-Meier median

10.5 y38 15 20 2

Cody9 1995 10-y OS Life-table method 10 y, 93%;median10.3 y

75 59 54 42

Sugg10 2000 10-y DFS Kaplan-Meier Median15.5 y

84 61 28 26

20-y DFS 76 46 28 23

AN, axillary node; DFS, disease-free survival; IMN, internal mammary node; ND, not described; OS, overall survival.

Table 6. Extended Radical Mastectomy Versus Radical MastectomyFirst Author Year Operation n % Survival % Recurrence

International Coop Study11 1976 RM 746 69 (5 y) 12 (5 y)ERM 697 72 (ns) 9 (ns)

Koszarowski16 1976 RM 746 69 (5 y) 8 (5 y)ERM 697 72 (ns) 5 (ns)

Veronesi12 1981 RM 374 61 (10 y) 22 (10 y)ERM 342 57 (ns) 19 (ns)

Lacour14 1983 RM 750 53 (10 y) 24 (10 y)ERM 703 56 (ns) 14 (p�0.05)

Lacour15 1987 RM 117 51 (15 y) 21 (15 y)ERM 126 51 (ns) 20 (ns)

Meier17 1989 RM 56 61 (10 y) n/aERM 56 74 (ns)

Veronesi13 1999 RM 374 20 (30 y) n/aERM 342 20 (ns)

ERM, extended radical mastectomy; n/a, not available; ns, not significant (statistical comparisons are within each study, and not across studies); RM, radicalmastectomy.


(n�533). Radiation treatment encompassed the axil-lary, supraclavicular, and IMN chains. Radiation low-ered the incidence of local-regional recurrence, but hadno influence on overall survival.

Lymphoscintigraphy and sentinel lymph node biopsyof internal mammary nodesEvaluation of lymphatic mapping and SLN biopsy as analternative to axillary lymph node dissection in clinicallynode-negative breast cancer patients is currently under-way. An SLN is the first lymph node in a given lymphaticdrainage group to receive lymphatic flow from a primarytumor site.27 This node should be the first node to be-come involved by metastasis from the tumor. It has beennoted in several series that the SLN may be located inextraaxillary sites, including the IMN chain. Fourteen re-cent studies have evaluated IMN lymphatic mapping;3,28-40

13 of the 14 used lymphoscintigraphy (Table 8). In theselatter 13 studies, patients were injected with radionu-clide by the intraparenchymal route, and in 1 of the 13,intraparenchymal injection of radionuclide was com-pared with intradermal injection.32 Technetium 99m-sulfur colloid was used as the radionuclide in 9 reports;technetium 99m-human albumin colloid was used in 4reports; and radionuclide in combination with blue dyewas used in 2 of 12 reports (Table 8). Study size rangedfrom 24 to 443 patients, with a median of 80 patients.The lymphoscintigram demonstrated mapping to theIMNs in 0% to 35% of patients. In 7 of the 13 reports,biopsy of IM SLNs was performed based on lympho-scintigraphy; IM SLNs contained metastases from 0%to 33% of the time. When these seven studies are com-bined, IM SLNs contained metastases in 18% of pa-tients (15 of 83) who underwent IM SLN biopsy.

Bobin and associates39 performed IM SLN biopsy in 33patients using 5mL of Evan’s blue dye intraparenchymallywithout radionuclide. For tumors located in the inner orcentral quadrant, conservative IMN chain dissection wasperformed through the incision of the breast-conservingtreatment. Blue IM SLNs were identified after opening theintercostal muscles. This technique yielded identificationof SLNs in 79% of patients (26 of 33). Eight patients, ofwhom three were AN negative, had IMN metastases.

All four quadrants of the breast have been shown inrecent studies using lymphoscintigraphy to have lym-phatic pathways to IMNs. Uren and colleagues3 found28% of outer-quadrant tumors (5 of 18) had drainage toIMNs, and 37.5% of inner/central tumors (3 of 8) haddrainage to IMNs. Johnson and coworkers36 found IMSLNs in 12.5% of outer-quadrant tumors (6 of 48) andin 12.5% of mediocentral tumors (4 of 32). Byrd andassociates40 found IMN drainage in 15% of outer-quadrant tumors (24 of 159) and 21% of mediocentraltumors (13 of 61). There was a statistically significantdifference (p�0.01) when comparing the rate of IMNdrainage in the upper-outer quadrant (10% versus 24%in all other quadrants), but substantial drainage was seenwith all quadrants of the breast.

Byrd and coworkers40 also showed that there was noassociation between the frequency of IMN drainage andtumor size. In fact 44% of patients (4 of 9) with high-grade ductal carcinoma in situ had IMN drainage. Fif-teen percent of T1 tumors (22 of 145), 17% of T2tumors (9 of 52), 8% of T3 tumors (1 of 12), and 0% ofT4 tumors (0 of 2) mapped to IMNs. The authors didnot biopsy the IMN sentinel nodes, so the incidence ofIMN metastases based on tumor size was not reported.

Table 7. Extended Radical Mastectomy Versus Radical Mastectomy: Results for Mediocentral TumorsFirst author Year Operation n % Survival p-Value

Intl Coop Study11* 1976 RM 192 52 (5y) 0.01ERM 71

Koszarowski16 1976 RM 310 68 (5y) nsERM 320 75

Veronesi12 1981 RM 141 62(10y) nsERM 145 56

Lacour15 1987 RM 36 28(15y) 0.02ERM 34 53

Meier17 1989 RM 56 60(10y) 0.025ERM 56 86

*Axillary lymph node positive, T1 and T2.ERM, extended radical mastectomy; ns, not significant; RM, radical mastectomy.


Roumen and coworkers32 compared lymphatic map-ping to extraaxillary sites using radionuclide injected in-traparenchymally versus intradermally in the same pa-tients. There was no difference in effectiveness of axillarylymphatic mapping between the two injection tech-niques, but lymphatic mapping to IMNs was seen onlywith intraparenchymal injection.

DISCUSSIONThe recent renewal in interest in IMN metastases is at-tributable to the advent of lymphatic mapping and SLNbiopsy, with which extraaxillary drainage has been dem-onstrated. A frequently asked question is “Should IMSLNs be biopsied when seen on lymphoscintigraphy?”This article provides insight into the prognostic signifi-cance of IMN metastases and the results of various treat-ment options, for the purpose of providing guidelinesfor the management of IM SLNs.

The IMN chain is a major pathway of lymphaticdrainage, with 18% to 33% of patients with operablebreast cancer demonstrating IMN metastases (Table 1).The incidence of IMN metastases is higher in patientswith positive ANs (29% to 52%) than in those with

negative ANs (4% to 18%) (Table 2). Additionally, inpatients with positive ANs, although mediocentral tu-mors have a higher incidence of IMN metastases thanlateral tumors, a substantial rate of metastasis (22% to26%) is nevertheless seen with lateral tumors (Table 4).

In AN-negative patients, lateral tumors have beendemonstrated to metastasize to IMNs in 2% to 13% ofcases; by comparison, mediocentral tumors have an in-cidence of IMN metastases of 8% to 10% in AN-negative patients (Table 3).

Recent studies with lymphatic mapping have dem-onstrated drainage to IMNs from all quadrants of thebreast. Uren and colleagues3 found drainage to IMNsin 28% of outer-quadrant tumors and 37.5% ofinner/central tumors. Johnson and associates36 foundIM SLNs in 12.5% of both outer- and inner-quadranttumors. Byrd and coworkers40 found IMN drainage in15% of outer-quadrant tumors and 21% of medio-central tumors. So, because substantial IMN drainageoccurs regardless of tumor location, tumor locationshould not be the sole factor in determining whetherto use SLN biopsy for the purpose of detecting IMNmetastases.

Table 8. Lymphoscintigraphy and Sentinel Lymph Node Biopsy of Internal Mammary Nodes

First author Year Type of injection Time to lymphoscintigram n% IMN�

lymphoscintigram (n) IMN� IMN�/AN�

Uren3 1995 2.5–7 MBq TSC IP 2.5–4 h 34 35 (12) n/a n/aKrag28 1998 1 mCi TSC IP 30 min–8 h 443 4 (19) n/a n/aO’Hea29 1998 0.3 mCi TSC IP Morning of surgery 60 10 (6) n/a n/aSnider30 1998 1 mCi TSC IP 45–310 min* 48 2 (1) 0% (0/1) n/aBurak31 1999 0.4 mCi TSC IP 2 h 24 0 (0) n/a n/aRoumen32 1999 60 MBq TAC IP

�15 MBq TACID or periareolar

18 h 99 9 (9) IP0 (0) ID

n/a n/a

van der Ent33 1999 10 mCi TACIP� blue dye

n/a 70 34 (24) 33% (5/15) n/a

Winchester34 1999 1 mCi TSC IP 1–4 h, 80 patients; 16–20h, 100 patients

180 11 (20) 15% (3/20) 0/3

Jansen35 2000 40–60 MBq TAC IP 30 min and 4 h 113 15 (17) 18% (3/17) 2/3Johnson36 2000 1 mCi TSC IP 2 h 80 12.5 (10) 30% (3/10) 0/3Noguchi37* 2000 3 mCi TSC IP�

4ml blue dyen/a 41 12 (5) 0% (0/5) n/a

Zurrida38 2000 5–10 MBq TACIP (ID if tumorsuperficial)

20 min and 3 h 376 4 (15) 7% (1/15) 0/1

Byrd40 2001 1 mCi TSC IP 2–3 h 220 17 (37) n/a n/aTotal 18% (15/83) 2/15

*No lymphoscintigram done, time to gamma probe identification.ID, intradermal; IMN, internal mammary node; IP, intraparenchymal; n/a, not available; TAC, technetium 99m-human albumin colloid; TSC, technetium99m-sulfur colloid.


Prospective, randomized trials have shown no benefit insurvival or local-regional recurrence from extended radicalmastectomy (which includes dissection of the IMN chain)when compared with radical mastectomy.11,12,14-17 It is notclear that the studies of extended radical mastectomy hadthe statistical power to show a survival benefit in the groupof patients most likely to benefit. One would anticipate thatthe group most likely to have a survival benefit would be theIM-positive/AN-negative women, since those patientswith metastases to both node groups are at very high risk forsystemic micrometastases. Because the IM-positive/AN-negative subgroup is very small, it is unlikely that even theInternational Cooperative trial would have the power todemonstrate a survival benefit. Nevertheless, given the po-tential morbidity of extended radical mastectomy and thelack of data showing a survival benefit, we would not rec-ommend biopsy of sentinel IMNs for the purpose of pur-suing an IMN dissection in the case of positive biopsy out-side the setting of a randomized trial.

Two recent randomized trials18,19 and a metaanaly-sis20 of postmastectomy radiation have demonstrateda significant survival benefit in high-risk breast cancerpatients. In each trial, the radiation ports included theIMN chain, so the results of these trials have fosteredrenewed interest in IMN irradiation. The studies donot clarify the possible role of IMN irradiation be-cause both trials included four different fields. Pro-spective, randomized trials and retrospective reviewsof IMN irradiation alone have shown no benefit insurvival with IMN irradiation. As with the data onextended radical mastectomy, there is a lack of data onthe potential survival benefit of radiation in the groupof patients most likely to benefit, namely IM-positive/AN-negative patients. A randomized trial ofinternal mammary node radiation for IM-positive/AN-negative patients would be reasonable, and cer-tainly preferable to a trial with higher morbidity usingextended radical mastectomy. Of note, the influenceof radiation therapy to the IMN chain is the subject ofan ongoing European Organization for Research andTreatment of Cancer Trial (No. 22922).

It is important to make the distinction between theincidence of clinically detected IMN recurrence and thepresence of subclinical metastases. The relatively highincidence of metastases in the IM nodes when the axil-lary nodes contain tumor (29%–52%, Table 2) is said bysome to justify biopsy. But patients treated by modifiedradical mastectomy and adjuvant systemic therapy alone

will have a less than 1% incidence of IMN failure as acomponent of first failure.41-45

Despite there being no proved benefit to dissection orirradiation of the IMNs, the prognostic value of theIMNs is great in select circumstances, namely for AN-negative patients with tumors less than 1cm. Patientswith negative ANs and positive IMNs have a decreasedoverall survival, as illustrated in Table 5, suggesting thepossible benefit to them of IM SLN biopsy if it wouldprovide information useful in determining whether toadd chemotherapy to their management. Because che-motherapy is already routinely recommended for AN-negative tumors greater than 1cm and for patients withpositive ANs,46 knowledge of IMN metastases wouldalter management only in AN-negative patients withtumors less than 1cm.

Fourteen recent studies have evaluated IMNs withlymphoscintigraphy.3,28-40 Of the 13 studies in which ra-dionuclide was injected intraparenchymally, the lym-phoscintigram demonstrated mapping to the IMNs infrom 0% to 35% of patients. In 7 of the 13 reports,biopsy of IM SLNs was performed based on lympho-scintigraphy. When these 7 studies are combined, IMSLNs contained metastases in 18% of patients (15 of 83)who underwent IM SLN biopsy; this finding is identicalwith observations in the earlier literature (21% IMNmetastases, Table 1). Of the 15 patients with IMN me-tastases, only 2 patients were AN negative, and tumorsize was not stated. Assuming that these tumors were lessthan 1cm in size, systemic therapy would have beenchanged in only 2.4% of patients (2 of 83) who under-went IM SLN biopsy. This demonstrates that only asmall number of patients will benefit from IM SLNbiopsy.

In the present American Joint Committee on Cancerstaging system for breast cancer,47 a patient with a posi-tive IMN is designated N3 and Stage IIIB. As illustratedin Table 5, if a patient has a positive IMN and negativeANs, her survival is the same as that of a patient who isAN positive and IMN negative. Veronesi and col-leagues13 recent 30-year followup of 342 patients whohad extended radical mastectomies also showed similarannual death rates (ADRs) between AN-negative/IMN-positive patients and AN-positive/IMN-negative pa-tients. A higher ADR was seen when both ANs andIMNs were positive; this may simply reflect a greatertumor burden and not a unique biology of the IMN. Sowe propose that the present American Joint Committee


on Cancer staging system be changed to better reflect theprognostic equivalence of isolated IMN and ANmetastases.

In conclusion, in the era of lymphatic mapping andSLN biopsy, we recommend consideration of sentinelIMN biopsy only when the information obtained couldchange therapy. At present, this pertains primarily topatients who have tumors less than 1cm with negativeANs. This group of patients in general would otherwisenot receive chemotherapy, and knowledge of positiveIMNs could change their management through the ad-dition of chemotherapy.

We emphasize that, with this recommendation, wewould expect only a small percentage of patients ulti-mately to benefit from IMN biopsy. We know fromhistorical studies that 4% to 18% of patients have IMmetastases with negative ANs, but the proportion ofthese patients with tumors less than 1cm is not known.We would expect this number to be small because largestudies have documented longterm survival rates of 90%or more in axillary node-negative patients with tumorsless than 1cm.48 So although it would be reasonable toperform sentinel IMN biopsy in this group of patients,we do not consider this to be mandatory in standardpractice because only a very small percentage of patientswould potentially benefit from this technique. Clinicaltrials that evaluate the yield of internal mammary nodemetastases found on sentinel lymph node biopsy in thisselect group of patients are warranted before changingthe standard of care.

REFERENCES

1. Stibbe E. The internal mammary lymphatic glands. J Anat1918;52:257–259.

2. Wangensteen OH. Super-radical operation for breast cancer inpatients with lymph node involvement. Proceedings of the Sec-ond National Cancer Conference, Cincinnati, OH. 1952;230–242.

3. Uren RF, Howman-Giles RB, Thompson JF, et al. Mammarylymphoscintigraphy in breast cancer. J Nucl Med 1995;36:1775–1780.

4. Urban JA, Marjani MA. Significance of internal mammarylymph node metastases in breast cancer. Am J Roentgenol Ra-dium Ther Nucl Med 1971;111:130–136.

5. Handley RS. Carcinoma of the breast. Ann R Coll Surg Engl1975;57:59–66.

6. Donegan WL. The influence of untreated internal mammarymetastases upon the course of mammary cancer. Cancer1977;39:533–538.

7. Li KY, Shen Z. An analysis of 1,242 cases of extended radicalmastectomy. Breast, Diseases of the Breast 1984;10:10–19.

8. Veronesi U, Cascinelli N, Greco M, et al. Prognosis of breastcancer patients after mastectomy and dissection of internalmammary nodes. Ann Surg 1985;202:702–707.

9. Cody HS III, Urban JA. Internal mammary node status: a majorprognosticator in axillary node-negative breast cancer. Ann SurgOncol 1995;2:32–37.

10. Sugg SL, Ferguson DJ, Posner MC, Heimann R. Should inter-nal mammary nodes be sampled in the sentinel lymph node era?Ann Surg Oncol 2000;7:188–192.

11. Lacour J, Bucalossi P, Cacers E, et al. Radical mastectomy versusradical mastectomy plus internal mammary dissection. Five-yearresults of an international cooperative study. Cancer 1976;37:206–214.

12. Veronesi U, Valagussa P. Inefficacy of internal mammary nodesdissection in breast cancer surgery. Cancer 1981;47:170–175.

13. Veronesi U, Marubini E, Mariani L, et al. The dissection ofinternal mammary nodes does not improve the survival of breastcancer patients. 30-year results of a randomised trial. Eur J Can-cer 1999;35:1320–1325.

14. Lacour J, Le M, Caceres E, et al. Radical mastectomy versusradical mastectomy plus internal mammary dissection. Ten yearresults of an international cooperative trial in breast cancer. Can-cer 1983;51:1941–1943.

15. Lacour J, Le MG, Hill C, et al. Is it useful to remove internalmammary nodes in operable breast cancer? Eur J Surg Oncol1987;13:309–314.

16. Koszarowski T. Rational approach to the efficiency of internalmammary lymph node dissection in breast cancer cure. Breast,Diseases of the Breast 1976;2:44–47.

17. Meier P, Ferguson DJ, Karrison T. A controlled trial of extendedradical versus radical mastectomy. Ten-year results. Cancer1989;63:188–195.

18. Overgaard M, Jensen MB, Overgaard J, et al. Postoperative ra-diotherapy in high-risk postmenopausal breast-cancer patientsgiven adjuvant tamoxifen: Danish Breast Cancer CooperativeGroup DBCG 82c randomised trial [see comments]. Lancet1999;353:1641–1648.

19. Ragaz J, Jackson SM, Le N, et al. Adjuvant radiotherapy andchemotherapy in node-positive premenopausal women withbreast cancer [see comments]. N Engl J Med 1997;337:956–962.

20. Whelan TJ, Julian J, Wright J, et al. Does locoregional radiationtherapy improve survival in breast cancer? A meta-analysis.J Clin Oncol 2000;18:1220–1229.

21. Fisher B, Slack NH, Cavanaugh PJ, et al. Postoperative radio-therapy in the treatment of breast cancer: results of the NSABPclinical trial. Ann Surg 1970;172:711–732.

22. Arriagada R, Le MG, Rochard F, Contesso G. Conservativetreatment versus mastectomy in early breast cancer: patterns offailure with 15 years of follow-up data. Institut Gustave-RoussyBreast Cancer Group. J Clin Oncol 1996;14:1558–1564.

23. Veronesi U, Zucali R, Luini A. Local control and survival inearly breast cancer: the Milan trial. Int J Radiat Oncol Biol Phys1986;12:717–720.

24. Host H, Brennhovd IO, Loeb M. Postoperative radiotherapy inbreast cancer—long-term results from the Oslo study. Int J Ra-diat Oncol Biol Phys 1986;12:727–732.

25. Palmer MK, Ribeiro GG. Thirty-four year follow up of patientswith breast cancer in clinical trial of postoperative radiotherapy.Br Med J (Clin Res Ed) 1985;291:1088–1091.

26. Fisher B, Redmond C, Fisher ER, et al. Ten-year results of arandomized clinical trial comparing radical mastectomy and to-


tal mastectomy with or without radiation. N Engl J Med1985;312:674–681.

27. Cox CE, Bass SS, Ku NN, et al. Sentinel lymphadenectomy: asafe answer to less axillary surgery? Recent results. Cancer Res1998;152:170–179.

28. Krag D, Weaver D, Ashikaga T, et al. The sentinel node in breastcancer—a multicenter validation study [see comments]. N EnglJ Med 1998;339:941–946.

29. O’Hea BJ, Hill AD, El Shirbiny AM, et al. Sentinel lymph nodebiopsy in breast cancer: initial experience at Memorial Sloan-Kettering Cancer Center [see comments]. J Am Coll Surg1998;186:423–427.

30. Snider H, Dowlatshahi K, Fan M, et al. Sentinel node biopsy inthe staging of breast cancer. Am J Surg 1998;176:305–310.

31. Burak WE Jr, Walker MJ, Yee LD, et al. Routine preoperativelymphoscintigraphy is not necessary prior to sentinel node bi-opsy for breast cancer. Am J Surg 1999;177:445–449.

32. Roumen RM, Geuskens LM, Valkenburg JG. In search of thetrue sentinel node by different injection techniques in breastcancer patients [see comments]. Eur J Surg Oncol 1999;25:347–351.

33. van der Ent FW, Kengen RA, van der Pol HA, Hoofwijk AG.Sentinel node biopsy in 70 unselected patients with breast can-cer: increased feasibility by using 10 mCi radiocolloid in com-bination with a blue dye tracer. Eur J Surg Oncol 1999;25:24–29.

34. Winchester DJ, Sener SF, Winchester DP, et al. Sentinel lymph-adenectomy for breast cancer: experience with 180 consecutivepatients: efficacy of filtered technetium 99m sulphur colloidwith overnight migration time. J Am Coll Surg 1999;188:597–603.

35. Jansen L, Doting MH, Rutgers EJ, et al. Clinical relevance ofsentinel lymph nodes outside the axilla in patients with breastcancer. Br J Surg 2000;87:920–925.

36. Johnson N, Soot L, Nelson J, et al. Sentinel node biopsy andinternal mammary lymphatic mapping in breast cancer. AmJ Surg 2000;179:386–388.

37. Noguchi M. Internal mammary chain sentinel lymph nodeidentification in breast cancer. J Surg Oncol 2000;73:75–80.

38. Zurrida S, Galimberti V, Orvieto E, et al. Radioguided sentinelnode biopsy to avoid axillary dissection in breast cancer. AnnSurg Oncol 2000;7:28–31.

39. Bobin JY, Zinzindohoue C, Isaac S, et al. Tagging sentinellymph nodes: a study of 100 patients with breast cancer [seecomments]. Eur J Cancer 1999;35:569–573.

40. Byrd DR, Dunnwald LK, Mankoff DA, et al. Internal mam-mary lymph node drainage patterns in patients with breast can-cer documented by breast lymphoscintigraphy. Ann Surg Oncol2001;8:234–240.

41. Freedman GM, Fowble BL, Nicolaou N, et al. Should internalmammary lymph nodes in breast cancer be a target for theradiation oncologist? Int J Radiat Oncol Biol Phys 2000;46:801–803.

42. Fowble B, Gray R, Gilchrist K, et al. Identification of a subgroupof patients with breast cancer and histologically positive axillarynodes receiving adjuvant chemotherapy who may benefitfrom postoperative radiotherapy. J Clin Oncol 1988;6:1107–1117.

43. Marks LB, Halperin EC, Prosnitz LR, et al. Post-mastectomyradiotherapy following adjuvant chemotherapy and autologousbone marrow transplantation for breast cancer patients with $10positive axillary lymph nodes. Int J Radiat Oncol Biol Phys1992;23:1021–1026.

44. Buzdar AU, McNeese MD, Hortobagyi GN, et al. Is chemother-apy effective in reducing the local failure rate in patients withoperable breast cancer? Cancer 1990;65:394–399.

45. Stefanik D, Goldberg R, Byrne P, et al. Local-regional failure inpatients treated with adjuvant chemotherapy for breast cancer.J Clin Oncol 1985;3:660–665.

46. Early Breast Cancer Trialists’ Collaborative Group. Polychemo-therapy for early breast cancer: an overview of the randomisedtrials. Lancet 1998;352:930–942.

47. American Joint Committee on Cancer. Breast. AJCC cancerstaging manual. 5th ed. Philadelphia: Lippincott-Raven; 1997.

48. Rosen PR, Groshen S, Saigo PE, et al. A long-term follow-upstudy of survival in stage I (T1N0M0) and stage II (T1N1M0)breast carcinoma. J Clin Oncol 1989;7:355–366.


Bevilacqua, JLB; Bevilacqua, RGB; Fey, J; Borgen, PI; Van Zee, KJ. Doctor what is my chance of having a positive sentinel node? (Meeting Abstract OP-032). 13th International Congress on Senology (Breast Diseases), Recife-PE, Brazil, 11-14th November 2004

Title: DOCTOR, WHAT IS MY CHANCE OF HAVING A POSITIVE SENTINEL NODE?

Author: BEVILACQUA,J.L.B.

Co-author(s): BEVILACQUA,R.G.; FEY,J.; BORGEN,P.I.; VAN ZEE,K.J.

Institutions: Memorial Sloan-Kettering Cancer Center, NY, USA; Depto. Cirurgia FMUSP, São Paulo,SP, Brazil; Hospital Sírio Libanês, São Paulo,SP, Brazil Introduction: Lymph node metastasis (LNM) is a multifactorial event. Several

variables have been described as predictors of LNM. However, it is difficult to

apply these data – usually expressed as odds ratio – to calculate the probability of

sentinel LNM (SLNM) for a specific patient. We developed a user-friendly

computerized prediction model based on a large dataset to assist in predicting the

presence of SLNM.

Methods: Clinical and pathological features of 4115 sequential sentinel lymph

node biopsy (SLNB) procedures were assessed with multivariable logistic

regression to predict the presence SLNM. A computer program based on the

logistic regression model (below) was created using age, pathological size, tumor

type, lymphovascular invasion, tumor location and multifocality. The model was

subsequently applied prospectively to 1792 sequential SLNBs.

Results: The program for the retrospective population was accurate and

discriminating, with an area under the receiver operating characteristic (ROC)

curve of 0.76. When applied to the prospective group, the model accurately

predicted likelihood of SLNM (ROC, 0.76).

Conclusions: Newly diagnosed breast cancer patients are increasingly interested

in information about their disease. This program is a useful tool that helps

physicians and patients to accurately predict the likelihood of SLNM. It should

not be used to avoid SLNB.

Probability of SLNM = [1+e (1.144 + 0.016xA – 0.669xS + 0.314xQ + 1,727xT – 1.142xL – 0.571xM)] –1

Values in the equation: (A) Age: years; (S) Pathological tumor size: centimeter

(cm); (Q) Tumor Location: Medial = 0(zero), Other quadrants = 1.

(T) Tumor Type: Colloid/Medullary/Tubular = yes/no; (L) Lymphovascular

Invasion = yes/no; (M) Multifocality = yes/no. Yes = 1, No = 0(zero)

sum ofall points is used to read the predicted probability ofhaving additionalnon-SLN metastases from the bottom scale.

Conc1usions: We have developed a nomogram to allow easy ca1culation ofthe likelihood of additional axillary metastases after a positive SLN biopsy.Thus, with information commonly available, the chance that additional non-SLNs will contain metastases can be estimated with a reasonable degree ofaccuracy for an individual patient. Using this tool, the patient can more fullyparticipate in the decision-making processo

85Chest Wall Recurrence Following Mastectomy Does Not Always Por-tend Dismal Prognosis A. Chagpar, 1*F. Meric, IK. Hunt, IM. ROSS,IM. Cristofanilli,1 S.E. Singletary, IT. Buchholz, I F. Ames, I S. Marcy, IG. Babiera,1 B. Feig, I G. Hortobagyi, I H. Kuerer.1 The University ofTexas M.D. Anderson Cancer Center.

lntroduction: Chest wall recurrence (CWR) in breast cancer patients (pts)is viewed as a harbinger of grim prognosis. Predictors of outcome after CWRare not well elucidated. While it is felt that characteristics of the CWR influ-ence outcome, little is known about the impact of primary tumor and treat-ment factors.

Methods: From 1988 to 1998, 130 pts with isolated CWRs were seen atour center, and retrospectively reviewed. Clinicopathologic factors, inclusiveof primary tumor features, and type of treatment were studied by univariateand multivariate analyses for distant metastasis free survival (DMFS) post-CWR. Median time to CWR from primary diagnosis was 25 months and medianfollow-up post-CWR was 37 months.

Results: lnirial nodal (N) status was the strongest predictor of Dutcome byunivariate analysis (median DMFS 112 months vs. 20 months for N negative(-) and N positive pts respectively). Other significant factors by this analysis'included initial T4 disease (p=0.006), primary lymphovascular invasion(p=O.024), treatment ofthe primary (neoadjuvant therapy: p=O.OOI, radiation(XRT): p=0.013), time to CWR> 24 months (p<O.OOI),and treatrnentforCWR(surgery: p=0.002, XRT: p<O.OOI,multimodality therapy: p<O.OOI). Featuresof the CWR itself were not significant. Multivariate analysis revealed that ini-tial N status hadthe greatest impact on survival (odds ratio=3.18, 95% CI:1.60-6.29, p=O.OOI); time to CWR (p=0.006) and the use ofXRT for CWR(p=0.005) were also independent predictors. Using these 3 factors, 3 groupsofpts were identified. Low risk was defined by initial N- disease, time to CWR>24 months, and XRT for CWR. lntermediate risk had 1-2 favorable features.High risk had none ofthese factors. The median DMFS post-CWR for the entirecohort was 20 months, and was significantly different for each of these groups(see figure).

Conclusions: Although CWR is thought to forecast a bleak outcome, thesepatients are truly a heterogeneous group, and survival post-CWR is largelydependent on primary tumor biology. Pts with initial N- disease who developa CWR after 24 months have an optimistic prognosis, especially ifthey can betreated with XRT.

an adjacent 5mm-l Ommofsonographically normal appearing breasttissue. Oncethe desired IB diameter was achieved, a maintenance (Iow freeze) system settinglimited further IB growth during excision. The cryoprobe can be used to elevatethe specimen into the incision and is automatícally warmed at the end ofthe pro-cedure to facilitate rapid removal (see photo). Results: CAL procedures: 14.Tumorhistology: 8 mc, 2 ILC, 4 both. Mean maxímum tumor diameter: 11 ± 4 mm.Mean tumor distance from skin: 10 ± 2 mm. Mean IB dimensions prior to exci-sion: 23 ± 6 mm (both width and depth) and 35 ± 7 mm length. Mean proceduretime (probe insertion to specimen remova!): 23 ±4 mins. Nuclear grade and recep-tor status evaluation were impacted by cryoartifact, but margins remained com-pletely unaffected in all cases. Condusions: CAL may be a useful altemative toWLB in the treatment of small breast cancers. Real-time U/S visualization forprobe placement and IB monitoring allows the surgeon to accurately tailor desiredmargins and facilitates lumpectomy by creating a firm, palpable target. A ran-domized trial to establish whether CAL reduces positive margin rates is in order.

84A multivariate nomogram to predict Iikelihood of non-sentinel nodemetastases in breast cancer patients with positive sentinel nodesJ.L. Bevilacqua,1 D.E. Manasseh,l* M.W. Kattan,1 L.K. Tan,1 p.I. Bor-gen,1 H.S. Cody, I KJ. Van Zee.1 Department of Surgery, Breast Service,Mernorial Sloan-Kettering Cancer Center, New York. NY.

lntroduction: ln breast cancer patients with positive sentinellymph nodes(SLN), controversy surrounds management ofthe axilla and the need for axil-lary dissecrion. Accurate estimates of the likelihood of additional disease inthe axilla would contribute greatly to this decision-making processo Currentpredictive models are limited by small numbers of SLN-positive cases and alimited set of predictor variables. We developed a user-friendly prediction modeland nomogram based on a large dataset to assist in predicting the presence offurther nodal metastases.

Methods: 702 patients with SLN metastases and completion ALND wereprospectively collected. Clinical data included pathologic size, presence oflym-phovascular invasion, number of positive SLNs, number of negative SLNs,method of detection ofSLN metastasis (routine histopathology [Routine l, serialsectioning with H&E [SS], or immunohistochemistry [lHC]), nuclear grade(including a separate category for lobular carcinoma), multifocality, estrogenand progesterone receptor status.

Results: By multivariate analysis, all but ERlPR and Grade were signifi-cantly predictive of additional nodal disease. Anomogram was created thatwas accurate and discriminating in predicting the likelihood of additional nodaldisease (area under the ROC curve was 0.74). Bootstrapping was used to achievegreater predictive accuracy and estimate discriminative ability with smallervariance. To use the nomogram, the appropriate category is selected for eachof the nine variables and the assigned points are read from the top scale. The

Polnls

PathSlze

Grade

LVI

Muni

Meth Detect

PosSLN

NegSLN

ERPRTotal Polnls

Probo of .••.•onSLN

"li 111~I ~u.

YH~

IHConly' ,

14 12 10 a 1 e !5

Posililrt~N~ll'IW'

N~"".Posililr.

534 550 56TH ANNUAL CANCER SYMP051UM

Ann Surg Onco12003; 10 (1): S34

36 Abstracts – Poster Session I

101 The Almanac Trial: Initial Experience.Mansel RE, on Behalf of ALMANAC Collaborators Group. Cardiff, UnitedKingdom.Introduction: In recent years sentinel node technique has been used increasingly inearly breast cancer. However, there are no published randomised controlled trials tovalidate this technique in breast cancer. The ALMANAC (Axillary LymphaticMapping Against Nodal Axillary Clearance) trial is a multi-centre randomisedcontrolled trial comparing sentinel node biopsy with standard axillary therapy inearly breast cancer.Methods: The trial, comprises an audit phase to assess the learning curve and arandomised phase. In the audit phase all patients had a sentinel node biopsy performedprior to a definitive axillary procedure. The primary objective of the audit phase is tomeasure the learning curve and standardise this new surgical technique. In order toproceed to the randomised phase, a localisation rate of >90% and a false negativerate of 5% or less are required in 40 consecutive patients per surgeon.The sentinel node was localised using a combination of blue dye and radioactiveisotope with a mandatory pre operative scintiscan.Results: There are currently 14 centres across the UK with 24 surgeons participatingin the audit phase of the trial. The early results in 292 patients show that 30% (89/292) of cancers were screen detected and the tumour size ranged from 2 to 100 mm( mean 21.2 mm, sd 11.8).Axillary drainage was seen in 75% of cases and internal mammary drainage in 10%.A total of 631 sentinel nodes were removed, an average of 2.2 per patient (range 1-8). Of these nodes there were 483 (78%) blue nodes and 516 (86%) hot nodes. Asentinel node was identified in 277 (95.5%) patients. The mean time for performinga sentinel node biopsy was 17.5 minutes (range 2-75 minutes).37.3% of patients were node positive. The false negative rate was 7%.Conclusions: The initial results of the ALMANAC trial confirm the feasibility ofsentinel node biopsy as an axillary staging procedure in the management of breastcancer. The randomised phase is underway and will demonstrate if sentinel nodebiopsy has any advantage over conventional axillary treatment in terms of localcontrol, morbidity, quality of life and health economics.

102 Physiologic and Socioeconomic Consequences of Breast CancerSurgery: A Comparison between Sentinel Lymph Node Biopsyand Axillary Lymph Node Dissection.Burak, Jr. WE, Hollenbeck ST, Zervos EE, Young DC. Departments of Surgeryand Biostatistics, The Ohio State University, Arthur G. James Cancer Hospitaland Richard J. Solove Research Institute, Columbus, OH.Background: Sentinel lymph node biopsy (SLNB) is an emerging technology in thetreatment of breast cancer. This study was designed to compare the postoperativemorbidity and socioeconomic impact of SLNB versus routine axillary lymph nodedissection (ALND) in patients with early stage breast cancer.Methods: A prospective, non-randomized, controlled study was designed to includeall patients who underwent breast conservation surgery and SLNB±ALND. GroupA consisted of patients who had a negative SLNB and did not go on to completionALND. Group B (control group) consisted of patients who underwent a SLNBfollowed by a completion ALND (level I and II) because either, 1) their sentinelnode contained cancer, or 2) they were within the validation phase of our sentinellymph node protocol. All patients were assessed at a minimum of 6 months from thecompletion of breast irradiation with a questionnaire addressing subjective complaintsof arm symptoms and recovery time. Patients then underwent a standardized physicalexamination to determine arm circumference.Results: Data was obtained from 96 patients with a mean follow-up of 15.3 months(range 8-29 months). Groups A (n=48) and B (n=48) were comparable with respectto age and comorbidity. The difference in mid-bicep circumference for Group A andGroup B was significant when comparing the ratio of procedure arm to the non-procedure arm and when subtracting the non-procedure arm from the procedure arm(p<0.003 and p<0.016, respectively). Return to normal activity in less than 4 dayswas reported in 70.7% of Group A, compared with 7.1% of Group B (p<0.001).Other objective comparisons are listed in the table below.Conclusion: SLNB results in less postoperative morbidity in terms of subjective armnumbness, objective mid-arm swelling and the requirements for a wound drain.Expeditious return to work or normal activity following SLNB has potentiallysignificant socioeconomic consequences.

Arm Complaints Arm Numbness Drain Days/pt Outpatient ProcedureGroup A 41.7% 16.7% 0.5 87.5%Group B 87.5% 81.2% 13.2 14.6%p value <0.001 <0.001 <0.001 <0.001

103 Quantitative Real-Time RT-PCR Detection of Breast CancerMicrometastasis Using a Multi-Gene Marker Panel.Mitas M,1 Mikhitarian K,1 Walters C,1 Baron P,1 Elliot B,1 Brothers T,1 RobisonJ,1 Metcalf JS,2 Gillanders WE,1 Cole DJ.1 1Surgery; 2Pathology, MedicalUniversity of South Carolina, Charleston, SC.The presence of metastatic disease in the ipsilateral axillary lymph nodes (ALN)remains the most valuable predictor for breast cancer recurrence. However, stagingby ALN dissection followed by H&E staining apparently lacks sensitivity since upto 30% of patients with pathologically negative nodes develop recurrent disease.Real-time RT-PCR is a relatively new technology that uses an on-line fluorescencedetection system to determine gene expression levels. It has the potential tosignificantly impact detection of breast cancer metastasis by virtue of its exquisitesensitivity, high-throughput capacity and quantitative readout system. To assess theutility of this technology in breast cancer staging, we determined the relative levelsof expression of 11 cancer-associated genes (MAM, PIP, CEA, CK19, VEGF, Erb2B,MUC1, c-myc, p97, VIM, and Ki67) in 42 negative control normal lymph nodes andin 16 histopathology-positive ALNs. We then performed a receiver operatingcharacteristic (ROC) curve analysis to determine the sensitivity and specificity levelsof each gene. At a 98% level of specificity, the following levels of sensitivity wereobserved: MAM, 100%; PIP, 68%; CK19, 67%; MUC1, 51%; CEA, 41%; and Erb2B,38%. Areas under the ROC curves indicated that the most accurate diagnostic markerswere MAM (100%), CK19 (91%), PIP (87%), and CEA (87%). MAM wasoverexpressed in the pathology-positive ALNs at levels ranging from 14- to 1.5 x105-fold above normal mean expression, whereas PIP was overexpressed from 52-to 1.3 x 106-fold above normal. Analysis of 21 pathology-negative lymph nodesrevealed overexpression of at least one of three genes (MAM (3 nodes), PIP (4 nodes),and CEA (1 node); overexpression range = 25- to 5.5 x 104-fold above normal) in 6of 21 nodes (29%). These results provide evidence that the most accurate diagnosticmarkers for detection of breast cancer metastatic disease in ALNs are MAM, PIP,CK19, and CEA, and that these markers should be applied as a panel for detection ofmicrometastatic disease. A multi-institutional trial is currently ongoing to correlategene overexpression with clinical outcomes.

104 Detection of Axillary Node Positivity Is Increased with SentinelLymph Node Biopsy in Early Stage Breast Carcinoma.Bevilacqua JLB, Van Zee KJ, Mann GB, Tan LK, Cody III HS, Heerdt AS,Montgomery LL, Petrek J, Borgen PI, Port ER. Breast Service, Department ofSurgery, Memorial Sloan-Kettering Cancer Center, New York, NY.Axillary lymph node status is the single most important prognostic factor for patientswith breast carcinoma. The detection of positive nodes significantly impacts patienttreatment, particularly for patients with tumors ≤ 1 cm who otherwise would notreceive systemic therapy. Sentinel node biopsy (SNB) with its enhanced pathologyhas been reported to increase detection of node positivity over axillary dissection(AD). We compared the incidence of axillary node positivity found with AD to thatof SNB for patients with microinvasive, T1a, and T1b breast carcinoma. BetweenJanuary 1, 1989 and December 31, 1991 258 patients were identified with tumors≤1cm who underwent standard AD obtaining a minimum of 10 nodes. Between June30, 1997 and November 19, 1999 596 patients were identified with 603 tumors ≤1cm who underwent SNB. For each tumor category, there was no statisticallysignificant difference between the SNB group and the AD group with respect topatient age, tumor location, laterality, type, histologic grade, or the presence oflymphovascular invasion. However, there was a statistically significant differencebetween the two groups with respect to multifocality, palpability, estrogen receptor(ER) status and frequency of breast conservation surgery. The table below lists theincidence of node positivity for each tumor category comparing the two differenttechniques for axillary evaluation.

Microinvasive T1a T1b AllAD 0/13 (0.0%) 8/68 (11.8%) 26/177 (14.7%) 34/258 (13.2%)SNB 2/28 (7.1%) 30/180 (16.7%) 91/395 (23.0%) 123/603(20.4%)p value p= 1.0 p= 0.43 p=0.025 p=0.012The detection of node positivity was increased with SNB for all categories, and wasstatistically significantly increased overall and specifically for those with T1b tumors.Among patients with positive nodes who underwent SNB, nodal positivity wasdetected by immunohistochemistry alone in both patients with microinvasive tumors(100%), 13/30 with T1a tumors (43.3%), and 26/91 with T1b tumors (28.6%). SNBsignificantly increases the detection of nodal positivity in patients with the earliestbreast cancers, and should be offered to these patients by surgeons experienced withthe technique.

Breast Cancer Research and Treatment - Vol. 64, No. 1, November (1) 2000

Abstracts – Poster Session I 37

105 The First 2,000 Cases - Axillary Node Metastases in the SentinelLymph Node Era.Bevilacqua JLB, Tan LK, Borgen PI, Cody III HS, Heerdt AS, MontgomeryLL, Petrek JA, Van Zee KJ. Breast Service, Departments of Surgery andPathology, Memorial Sloan-Kettering Cancer Center, New York, NY.The presence of axillary lymph node metastases (ALNM) remains the most importantprognostic indicator for a woman with breast carcinoma (ca). The adoption of sentinellymph node (SLN) biopsy and enhanced pathologic analysis of the SLN marks anew era in the treatment of breast cancer. We undertook this analysis to examine theclinical and pathological predictive factors of ALNM in this new era.METHODS: From 9/1996 to 3/2000, we performed 2,000 consecutive SLNprocedures (cases) in 1,819 patients. 1,849 cases had measurable invasive ca. Ofthese, 1,556 were ductal not otherwise specified (NOS), 103 were special ductalsubtypes (mucinous, tubular, or medullary), and 190 were lobular ca. Univariate (T-Test and χ2) and multivariate analyses (binary logistic regression) were performed.RESULTS: By T stage (AJCC-1997), the frequency of ALNM was: T1mic= 3/32(9.4%), T1a= 34/223 (15.2%), T1b= 106/501 (21.2%), T1c= 272/791 (34.4%),T2≤3cm= 126/231 (54.5%), T2>3cm= 40/58 (69%), T3= 13/13 (100%). The p-valuesfor variables significantly associated with ALNM are listed in the table.Variable Univariate Univariate Multivariate Multivariate

Ductal Lobular Ductal LobularAge <0.0005 <0.0005 0.003 NSSize <0.0005 <0.0005 <0.0005 <0.0005Palpability <0.0005 0.001 NS NSLymphovascular Invasion (LVI) <0.0005 0.035 <0.0005 NSHistologic Grade <0.0005 NA NS NANuclear Grade <0.0005 NA 0.036 NADuctal NOS vs. Special Subtype <0.0005 NA 0.001 NAUIQ vs. Other Quadrant <0.0005 NS <0.0005 NSMultifocality 0.001 0.001 0.027 <0.0005Table Notes: NA=not applicable, NS=non-significant.CONCLUSIONS: Tumor size, LVI, age, nuclear grade, histologic subtype,multifocality, and location in the breast were independent predictive factors for ALNMfor ductal ca. For upper-inner quadrant (UIQ) ductal ca, ALNM were less frequentthan for other quadrant tumors. Only tumor size and multifocality were independentpredictive factors for ALNM in lobular ca.

106 RT-PCR Analysis for Mammaglobin and CarcinoembryonicAntigen Detects Metastases in Histology-Negative SentinelLymph Nodes.Verbanac KM,1 Min CJ,1 Purser SM,1 Swanson MS,1 Lo K,2 Albrecht JA,2

Tafra L.3 1Surgery, East Carolina University, Greenville, NC; 2National GeneticsInstitute, Los Angeles, CA; 3Anne Arundel Medical Center, Annapolis, MD.Axillary staging is critical for predicting breast cancer recurrence and for guidingtreatment, yet lymph node (LN) metastases are missed in many patients by currentmethods. To improve staging accuracy, we have combined sentinel lymphadenectomywith the sensitive technique of reverse transcriptase-polymerase chain reaction (RT-PCR). As the first investigators to identify mammaglobin (MG) and carcinoembryonicantigen (CEA) as specific dual markers for the RT-PCR detection of SLN metastases,our objective was to determine if this molecular analysis does in fact detect occultmetastases. Here we present analysis of 154 SLN from 87 patients enrolled in amulti-center trial: 37% node-positive by histology (Histo+ = H&E ± IHC) and 63%node-negative. As single markers, PCR for MG and CEA identified 94% and 84%of patients with Histo+ LN. Only one patient with histologically involved LN failedto express either marker (3% false negative rate) and 78% had LN that expressedboth markers. Each marker increased the detection of metastases in Histo- patients(n = 55):MG+ CEA+ MG+ CEA+ MG+ or CEA+ MG- CEA-15 pts (27%) 14 pts (25%) 5 pts (9%) 24 pts (44%) 31 pts (56%)Thus, PCR detection of MG or CEA in SLN would upstage 44% of Histo- women,consistent with clinical recurrence rates. Marker expression correlates with tumorsize and estrogen receptor negativity. Clinical recurrence wil l determine thesignificance of molecular staging: no node-negative women have recurred to date(mean follow-up 31 months). These results show that MG and CEA remain promisingdual clinical markers for RT-PCR detection of SLN micrometastases for improvedstaging and management of breast cancer patients.

107 Variations of Success Rates for Sentinel Node Mapping in BreastCancer: Results of a German Multicenter Study.Kuehn T, Kotzerke J, Santjohanser C, Schirrmeister H, Grimm S, Koretz K,Rebstock AB, Reske SN, Kreienberg R. University Medical Center of Ulm,Germany.Objective: Promising results have been reported for sentinel node mapping in breastcancer and numerous institutions have already abandoned routine axillary dissectionfor small breast cancers. The success rates reported in literature, however, vary andmany questions concerning the technical proceeding are still not standardized. Inorder to analyze whether sentinel node biopsy was a reliable procedure, we performeda multicenter validation study and evaluated the detection rate and the sensitivity ofeach participating center as well as the results of different surgeons.Material and Methods: The data of 709 patients were prospectively collected. Onlypatients with unicentric disease and clinically negative axillary status were included.All patients received full axillary dissection after sentinel node biopsy had beenperformed. The surgeon was free to chose the lymphography, as well as the injectiontechnique.Results: 16 centers participated in the trial and performed 709 sentinel node biopsies.The blue dye technique was used in 139 patients, lymphoscintigraphy in 127 and acombined procedure in 442 cases. The overall detection rate was 84.3% and thesensitivity 91.3%. An interinstitutional variation of success rates for sentinel nodeidentification was found between 57.1% (Tc only and combined) and 94.1% (Tconly), the false negative rate varied between 22.2% (Tc only and combined) and 0%(all techniques). The center with the lowest detection rate had also the highest rate offalse negative findings. When the data of all surgeons with less than 10 procedureswere summarized, an overall detection rate of 75.4% and a sensitivity of 91.4% wasfound. For surgeons with more than 20 procedures the detection rate varied between71.4% and 100% and the sensitivity between 71.4% and 100%.Conclusion: Sentinel node mapping seems a promising tool for axillary managementin breast cancer. Success rates, however, vary strongly between different institutionsand surgeons. In order to obtain reliable results for the procedure, a standardized anddetailed protocol for the technical proceeding is necessary, as well a defined trainingprogram.

108 A New Score for the Assessment of Postsurgery MorbidityFollowing Axillary Surgery.Kuehn T, Klauss W, Darsow M, Regele S, Flock F, Wendt I, Rebstock AB,Kreienberg R. University Medical Center of Ulm, Germany.Introduction: One endpoint of numerous trials is the assessment of arm symptomsfollowing new techniques of axillary surgery. Shoulder arm morbidity, however, isa complicated polysymptomatic disease and diverging techniques are currently usedto assess subjective and objective symptoms. In order to optimize the assessment ofthis complex syndrome, we developed a specific score that summarizes the occurrenceas well as the intensity of subjective and objective symptoms.Material and Methods: 396 patients who had undergone conventional axillarysurgery were thoroughly examined for postsurgical morbidity. Objective andsubjective symptom occurrence and symptom intensity were evaluated using 5-pointscales. Reproducible techniques were applied for clinical evaluation. According totheir significance and their degree of intensity, different symptoms were summarizedon a scale where subjectively and objectively assessed symptoms were each given50% consideration. A maximum of 5 pts was assigned if no sensitivity problem wasregistered, 25 pts for absence of pain and 20 pts for unimpaired strength, mobilityand edema respectively. According to the intensity of symptoms within the 5-pointscales, score points were subtracted. A patient with 100 score points was regarded ascompletely free of disease.Results: When all clinical measurements and the results from a 5-point-scaledquestionnaire were summarized in a specific score, only 12 patients achieved 100score points. 82 patients had > 90 pts, 80 women achieved between 80 and 90 pts, 73between 70 and 80 pts, 73 between 60 and 70 pts, 34 between 50 and 60 pts and 27had <50 pts. The median for all patients was of 77.5 score points. For differentsurgeons a variation of the median between 70.8 and 82.3 was registeredDiscussion: Shoulder arm morbidity following axillary surgery is a polysymptomaticdisease that can bes t be expressed on a score that summarizes the occurrence as wellas the intensity of several symptoms. The use of a specific score facilitates theassessment of postsurgical outcome so that the comparability of clinical trials maybe improved. The effect of rehabilitation measures can equally be verified.

Breast Cancer Research and Treatment - Vol. 64, No. 1, November (1) 2000

Documents

Modelos estatísticos para previsão de metástases axilares ... · mama em pacientes submetidos à biópsia de linfonodo sentinela [tese]. São Paulo: Faculdade de Medicina, Universidade