GASTROINTESTINAL

Magnetic Resonance Imaging-Detected ExtramuralVenous Invasion in Rectal Cancer before and after PreoperativeChemoradiotherapy: Diagnostic Performance and PrognosticSignificance

Eun Sun Lee1,2,3 &Min Ju Kim3& Sung Chan Park4

& Bo Yun Hur3 & Jong Hee Hyun4&

Hee Jin Chang4 & Ji Yeon Baek4& Sun Young Kim4,5

& Dae Yong Kim4& Jae Hwan Oh4

Received: 12 January 2017 /Revised: 25 May 2017 /Accepted: 29 June 2017# European Society of Radiology 2017

AbstractObjectives We evaluated the diagnostic performance of mag-netic resonance imaging (MRI) in terms of identifying extra-mural venous invasion (EMVI) in rectal cancer patients withpreoperative chemoradiotherapy (CRT) and its prognosticsignificance.Methods During 2008–2010, 200 patients underwent surgeryfollowing preoperative CRT for rectal cancer. Two radiolo-gists independently reviewed all pre- and post-CRT MRI ret-rospectively. We investigated diagnostic performance of pre-CRT MR-EMVI (MR-EMVI) and post-CRT MR-EMVI(yMR-EMVI), based on pathological EMVI as the standardof reference. We assessed correlation between MRI findingsand patients’ prognosis, such as disease-free survival (DFS)and overall survival (OS). Additionally, subgroup analysis inMR- or yMR-EMVI-positive patients was performed to con-firm the significance of the severity of EMVI in MRI on pa-tient’s prognosis.

Results The sensitivity and specificity of yMR-EMVI were76.19% and 79.75% (area under the curve: 0.830), respective-ly. In univariate analysis, yMR-EMVI was the only significantMRI factor in DFS (P = 0.027). The mean DFS for yMR-EMVI (+) patients was significantly less than for yMR-EMVI (−) patients: 57.56 months versus 72.46 months.Conclusion yMR-EMVI demonstrated good diagnostic per-formance. yMR-EMVI was the only significant EMVI-related MRI factor that correlated with patients’ DFS in uni-variate analysis; however, it was not significant in multivariateanalysis.Key Points•Diagnostic performance of MRI for EMVI after preoperativechemoradiotherapy is good.

• The mean DFS was lower in yMR-EMVI-positive than yMR-EMVI-negative patients.

• MRI can facilitate prognosis prediction of rectal cancer pa-tients with preoperative chemoradiotherapy.

Keywords Rectal neoplasms .Magnetic resonance imaging .

Disease-free survival . Chemoradiotherapy . Prognosis

AbbreviationsAUC Area under the curveCRT ChemoradiotherapyDFS Disease-free survivalMR-vTRG

EMVI tumour regression grading assessed bymagnetic resonance

EMVI Extramural venous invasionMR-EMVI

Initial MRI-detected EMVI

MRI Magnetic resonance imagingOS Overall survival

* Min Ju Kimkimmj@ncc.re.kr

1 Department of Radiology, Chung-Ang University Hospital,Seoul, Korea

2 College of Medicine and Graduate School of Medicine, Chung-AngUniversity, Seoul, Korea

3 Department of Radiology, National Cancer Centre, 323 Ilsan-ro,Ilsandong-gu, Goyang-si, Gyeonggi-do 410-769, Korea

4 Centre for Colorectal Cancer, National Cancer Centre,Goyang, Gyeonggi-do, Korea

5 Department of Oncology, Asan Medical Centre, University of UlsanCollege of Medicine, Seoul, Korea

Eur RadiolDOI 10.1007/s00330-017-4978-6

(y)pEMVI Pathologically proven EMVIyMR-EMVI

Post-chemoradiotherapy MRI-detected EMVI

ROC Receiver operating characteristic

Introduction

Extramural venous invasion (EMVI) has been proposed tobe an important and independent pathological factor asso-ciated with poor prognosis in rectal cancer [1–4].Histopathologically, EMVI is defined as the presence ofmalignant cells within endothelial cell-lined blood vesselsbeyond the muscularis propria [5]. Pathological EMVIcan only be confirmed postoperatively. Advances in mag-netic resonance imaging (MRI) have enabled detection ofEMVI in rectal cancer patients preoperatively through theidentification of expanded vessels or tumour signals in thevenous lumen [6, 7].

Several studies have revealed the relevance between MRI-detected EMVI (MR-EMVI) and prognosis of the disease[8–11]. These studies indicate that MR-EMVI correlates sig-nificantly with the 3-year disease-free survival (DFS) or riskof metachronous metastases within 1 year. Given that mostresectable stage II or III rectal cancers usually undergo preop-erative chemoradiotherapy (CRT) and EMVI occurs at stage IIor III disease, by definition—malignant cells beyond themuscularis propria— assessing the correlations amongEMVI on post-CRTMRI (yMR-EMVI), pathology, and prog-nosis seems to bemore important [12]. However, in post-CRT,accuracy of determining tumour extension usually decreasesbecause of the presence of increased amounts of fibroticstrands and reactive oedema [13]. Therefore, additional vali-dation of the diagnostic performance and prognostic impor-tance of yMR-EMVI has become necessary.

Even though there have been few recent reports showingthe clinical significance of MR- or yMR-EMVI [14, 15], wecannot assure their diagnostic performance based on histo-pathological results, especially in post-CRT patients. To ourknowledge, the correlation between MRI findings regardingEMVI and overall survival (OS) still remains unknown.Therefore, this study aimed to evaluate the diagnostic perfor-mance of MRI in terms of EMVI in rectal cancer patients withpreoperative CRT and its prognostic significance in terms ofDFS and OS.

Material and methods

This retrospective study was approved by the relevant institu-tional review board, and the requirement for informed consentwas waived (IRB No. NCC2015-0138).

Patients

From January 2008 to December 2010, a total of 312 consec-utive patients underwent surgery after preoperative CRT forrectal cancer with curative intent in our institute. We excludedpatients who had not undergone any pre- or post-CRTMRI (n= 99). Most of these patients had not undergone post-CRTMRI due to a shortage of MRI facilities in the early stagesof our institution and MRI studies had been replaced by com-puted tomography (CT) for some of those patients. In addi-tion, we also excluded patients with distant metastasis, preop-eratively (n = 13). Finally, we included 200 patients (males:139, females: 61, mean age: 59.57 ± 10.76 years, range: 23–87) in this study.

Treatment

Preoperative chemoradiotherapy regimen All patents hadundergone long-course preoperative CRT for rectal cancerbecause of potential circumferential resection margin (CRM)involvement, T3 or T4 stage in terms of the depth of thetumour, or suspected nodal disease, based on the initialMRI. An initial dose of 45 Gy, in 25 fractions, was deliveredto the entire pelvis, including the gross tumour andmesorectum, presacral space, whole of the sacral hollow,and regional lymphatics (i.e., the perirectal, internal iliac,presacral, and distal common iliac lymphatics). This wasfollowed by a boost of 5.4 Gy, in three fractions, to the grosstumour, and mesorectum or tumour bed, for 5 days a week.Chemotherapy was administered concurrently with radiother-apy in all patients using fluoropyrimidine (n = 189) oririnotecan (n = 11). Because we had not routinely investigatedEMVI in our MRI reports from 2008 to 2010, we were unableto identify whether EMVI on pre-CRT MRI had influencedthe preoperative treatment.

Postoperative medical treatment One hundred patients re-ceived adjuvant chemotherapy using the following regimen:fluoropyrimidine (n = 180), oxaliplatin-based regimen (n =11), irinotecan (n = 5), or other (n = 4). In two patients, post-operative radiation therapy was performed because ofmetachronous, solitary distant metastases, including to theliver and lung.

Surgery The included patients underwent surgery after pre-operative CRT as follows, according to the tumour locationand severity: low anterior resection (n = 105), ultra-low ante-rior resection (n = 72), Miles’ operation (n = 20), Hartmann’sprocedure (n = 2), and trans-anal excision (n = 1). The medianinterval between CRT and an operation was 6 weeks (range:4–8 weeks).

Eur Radiol

MRI acquisition

All included patients underwent MRI examinations both pre-CRT and post-CRT. The rectal MRI protocol employed a thin,−3-mm section, 4-mm spacing, T2-weighted turbo spin echo(TSE) sequence with a 32-channel-phased array body coil anda field of view of 240mm× 240mm, 300 × 300matrix.We used3.0-T MRmachines in 360 examinations (Achieva 3.0T, PhilipsHealthcare, Amsterdam, Netherlands; n = 327; HDx 3.0T, GEMedical Systems, Boston, MA, USA; n = 33) and 1.5-T MRImachine in 30 examinations (Signa 1.5T, GEMedical Systems).The other scans (n = 10) were performed at outside hospitalsusing different MRI systems. We routinely obtained MRI sagit-tal, coronal, and axial planes in all patients, including outsideMRIs. Among the 200 included patients, 164 patients werescanned using MR scanners with the same magnetic field (3 T:n = 162; 1.5 T: n = 2). Thirty-six patients underwent MRI usingscanners with different magnetic fields (3 T vs. 1.5 T).

Histopathologic examination

After radical surgery, all tumour specimens were examined byone pathologist (C.H.J). The entire tumour with surroundingmesorectal fat was serially sliced into 4-mm-thick sectionsand embedded in paraffin. Then, tumours were assessed andstaged microscopically using the TNM staging system as theguidelines of the American Joint Committee on Cancer (6thEdition) [16]. In addition, our pathologist assessed and reportedthe presence of intramural or extramural venous invasion of thetumour, angiolymphatic invasion , perineural invasion andperitonealisation of the tumour, which are known as patholog-ical factors associated with clinical prognosis [17], in additionto Dworak’s rectal cancer regression grade [18]. The presenceor absence of EMVI beyond the muscularis propria, accordingto the standard definition (the presence of a rounded mass oftumour tissue within an endothelium-lined space, eithersurrounded by a rim of smooth muscle or containing red blood

Fig. 1 Sagittal scan of rectalmagnetic resonance imaging(MRI) in a patient that was MR-EMVI and yMR-EMVI-positive.(A) Pre-chemoradiotherapy(CRT) MRI. In the expandedvessel, there is a tumour with lossof normal signal void, grade 4MR-EMVI at baseline. (B) Post-CRT MRI showing extensiveextramural venous invasion(EMVI) with tumour expansionand destruction of a vessel. (C) Aresected surgical specimenrevealing the invasion byneoplasm. An elongated structure(arrow) is identified an adjacentvein filled with tumour (ypEMVI-positive). (D) Scan view of EMVI(arrow) in mesorectum. (E) Ahigher magnification view (x100)of EMVI showing residualtumour cells within degeneratedvenous wall (arrows)

Eur Radiol

cells) was noted. The presence of vascular invasion in the sub-mucosal and/or muscular layer was documented as intramuralinvasion. When we observed both intramural and extramuralvenous invasion, we recorded the patient as EMVI-positive [5].

Assessment of MR-EMVI, yMR-EMVI, and EMVItumour regression grading by MR (MR-vTRG)

Two board certified abdominal radiologists (H.B.Y, K.M.J;10- and 15-year-experienced) independently reviewed allpre- and post-CRT MRI scans (Figs. 1 & 2) and scored thepossibility of EMVI according to the 5-scale EMVI scoringsystem suggested by Smith et al. [10] (Fig. 3), according toradiological features. They were blinded to the clinical data,including cancer stage and patients’ prognoses. After indepen-dent scoring, comparison between the results of the two radi-ologists was performed and inter-observer agreement was

calculated. In the next step, a second consensus reading wasperformed to adjust for discrepancies. Finally, we investigatedthe diagnostic performance ofMR-EMVI, yMR-EMVI, basedon pathologic EMVI (ypEMVI) as the standard of referencewith an agreed result. In terms of EMVI tumour regressiongrading by MRI (MR-vTRG), we used a 5-scale scoring sys-tem suggested by Chand et al. [15]; 1: tumour signal replacedby vessel fibrosis, 2: 50−70% fibrosis of tumour signal, 3: 25−49% fibrosis of tumour signal, 4: less than 25% fibrosis oftumour signal, and 5: minimal fibrosis of tumour signal, with-in the lumen. For statistical analysis, we regarded grades 4 and5 as poor responders.

Correlation between MRI findings and clinical outcomes

We assessed the correlation between MRI findings and pa-tients’ prognoses, such as DFS and OS. We obtained survival

Fig. 2 Coronal rectal magneticresonance imaging (MRI) scanthat is MR-EMVI-positive andyMR-EMVI-negative. (A) A pre-chemoradiotherapy (CRT) MRIscan revealing tumour signalwithin the vessels’ (arrow)perpendicular outer margin of therectal tumour, grade 3 MR-EMVI. (B) Post-CRT MRIdemonstrating no definite tumoursignal within the vessels. (C)Fibrotic whitish tissue in themesorectum in the vicinity of therectal cancer (arrow) on a cutsection of the tissue. (D) Scanview of post-CRT regressivechange (arrow) in mesorectum.(E) On a high magnification view,venous wall shows luminalocclusion with infiltration ofinflammatory cells andmacrophage, suggesting post-irradiational regressive change.Residual tumour cells are not seenin the wall

Eur Radiol

and clinical data from the electronic medical record system ofour institute.

Subgroup analysis ofMR-EMVI and yMR-EMVI positivepatients

We performed subgroup analysis to confirm the significanceof the severity of EMVI in MRI on patients’ prognoses. Weregarded score B3^ of the Smith scoring system [10] as mildEMVI, whereas score B4^ was considered severe EMVI.Comparison of DFS and OS was performed according to theseverity of EMVI on MRI.

Statistical analysis

Diagnostic performance of an agreed result on MRIfindings was investigated using receiver operatingcharacteristic curves and the relevant interobserveragreement was assessed with weighted kappa analysis.Regarding correlation with clinical outcomes, we usedthe Cox hazard regression test and Kaplan−Meiercurves. The median observation was 50 months(range: 1–80 months). For all statistical analyses inthis study, Medcalc® software, version 14.8.1. wasused.

Fig. 3 Summary of MR- andyMR-EMVI scoring system

Eur Radiol

Results

Histopathological results

Of the 200 obtained surgical specimens, 172 were confirmedas well to moderately differentiated adenocarcinoma, fivewere poorly differentiated adenocarcinoma, and 23 had noresidual tumours, respectively. The rate of complete remission

(CR) after CRTwas 11.5% (23/200) in this study. Among theremaining 177 non-CRs, we found 46 stage I cancers (ypT1 or2, N0, M0), 61 stage II cancers (ypT3 or 4, N0, M0), 65 stageIII cancers (ypT1-4, N1-2, M0), and two stage IV cancers (ypanyT, anyN, M1). Pathological examination revealed thatthere were 42 ypEMVI (+) cases among the 200 patients inthis study.

Interobserver agreement of MRI findings and diagnosticperformance

In the first independent reading session of two radiologists, theobtained quadratic weighted kappa values revealed 0.801 (SE0.028) for MR-EMVI and 0.736 (SE 0.039) for yMR-EMVI,representing good agreement. In terms of tumour regression inEMVI, good agreement was also seen between two reviewers(k = 0.699, SE 0.045).

ROC curve analysis using consensus results showed fairto good diagnostic performance of yMR-EMVI and MR-EMVI (AUC, 0.830 vs. 0.778) (Fig. 4). When we used ≥3as a positive criterion, the sensitivity and specificity ofMR-EMVI were 90.48% and 41.14%, respectively. Onthe other hand, sensitivity and specificity of yMR-EMVIwere 76.19% and 79.75% when using the same criterion.

Fig. 4 The receiver operating characteristic (ROC) curve analysis for thediagnostic performance of MR-EMVI and yMR-EMVI (p = 0.076)

Table 1 Univariate and multivariate analysis (Cox proportional Hazards for DFS) by clinical, pre-CRT MR, post-CRT MR and postoperativepathologic characteristics

Variables Group Patient Numbers Univariate analysis Multivariate analysis

RR 95% CI P RR 95% CI P

Patientcharacteristics

Sex Male 139 1.122 0.5469 - 2.3019 0.7535 _ _ _Female 61

Height Upper/mid 144 0.8215 0.4005 - 1.6858 0.5921 _ _ _Low (<5 cm) 56

Initial CEA WNL (≤5 ng/mL) 136 2.2154 1.1310 - 4.3398 0.0204* _ _ _Elevated 64

Pre-CRT MR MR-EMVI Negative 69 1.3484 0.6448 - 2.8199 0.4271 _ _ _Positive 131

Post-CRT MR yMR-EMVI Negative 136 2.6891 1.3697 - 5.2714 0.004* _ _ _Positive 64

MR-vTRG None or Regressed 182 2.089 0.8081 - 5.4002 0.1284* _ _ _Poor response 18

Final pathology Stage Stage 0-II 133 3.6393 1.8211 - 7.2728 0.0003* 2.3939 1.0994 - 5.2124 0.0279**Stage III-IV 67

Grade Low risk 195 5.595 1.6986 - 18.4294 0.0046* _ _ _High risk 5

ypEMVI Negative 158 2.4474 1.2079 - 4.9588 0.013* 2.4021 1.1066 - 5.2140 0.0267**Positive 42

ypCRM Negative 186 4.4034 1.9153 - 10.1237 0.0005* 2.9362 1.2358 - 6.9764 0.0147**Positive 14

*- Variables included multivariate analysis based on P < 0.15, **- independently significant variables

Eur Radiol

Correlation between MRI findings and clinical outcomes

In univariate analysis, yMR-EMVI was the only significantMRI factor in DFS (P = 0.027, Table 1 and Fig. 5). The meanDFS for yMR-EMVI-positive patients was significantly lower

than yMR-EMVI-negative patients; 57.6 months (95% CI:50.7−64.5) versus 72.5 months (95% CI: 69.0−76.0). Amongthe other clinical and histopathological factors, elevation ofinitial serum carcinoembryonic antigen (CEA > 5 ng/mL),stage, grade, ypEMVI, and ypCRM involvement were

Fig. 5 Kaplan−Meier curvesshowing the relationship betweenyMR-EMVI and disease-freesurvival (DFS)

Table 2 Univariate and multivariate analysis (Cox Proportional Hazards for OS) by clinical, pre-CRT MR, post-CRT MR and postoperative patho-logic characteristics

Variables Group PatientNumbers

Univariate analysis Multivariate analysis

RR 95% CI P RR 95% CI P

Patient characteristics Sex Male 139 0.9369 0.3886 - 2.2590 0.8846 _ _ _Female 61

Height Upper/mid 144 0.8029 0.3436 - 1.8765 0.6123 _ _ _Low (<5 cm) 56

Initial CEA WNL (≤5 ng/mL) 136 1.0021 0.4283 - 2.3448 - 0.9962 _ _ _Elevated 64

Pre-CRT MR MR-EMVI Negative 69 1.1833 0.5056 - 2.7691 0.6981 _ _ _Positive 131

Post-CRT MR yMR-EMVI Negative 136 1.976 0.8838 - 4.4178 0.0971* _ _ _Positive 64

MR-vTRG None or Regressed 182 1.5345 0.4572 - 5.1500 0.4882 _ _ _Poor response 18

Final pathology Stage Stage 0-II 133 3.7134 1.6217 - 8.5029 0.0019* _ _ _Stage III-IV 67

Grade Low risk 195 13.6077 4.5442 - 40.7485 <0.0001* 6.0831 1.8341 - 20.1757 0.0032**High risk 5

ypEMVI Negative 158 4.5169 2.2041 - 10.0796 0.0002* 3.3403 1.3779 - 8.0972 0.0076**Positive 42

ypCRM Negative 186 4.2139 1.5702 - 11.3092 0.0043* _ _ _Positive 14

*- Variables included multivariate analysis based on P < 0.15, **- independently significant variables

Eur Radiol

significantly related to DFS. However, multivariate analysisrevealed that stage, ypEMVI, and ypCRM were the only inde-pendent variables statistically significantly associated withDFS (Table 1). In terms of OS, no MRI findings for EMVIwere significantly correlated with patients’ survival time inunivariate or multivariate Cox proportional hazard regressionanalysis. In contrast, tumour grade and ypEMVI were con-firmed as independent and significant factors for survival time(Table 2).

Subgroup analysis of MR-EMVI-and yMR-EMVI-positive patients

In terms of MR-EMVI, patients with severe EMVI had statis-tically significantly shorter DFS (Log-rank test, P = 0.018) aswell as OS (P = 0.046) than patients with mild EMVI. Incontrast, the DFS of yMR-EMVI-positive patients was signif-icantly different according to the severity of EMVI (P = 0.001;Fig. 6), whereas the OS was not (P = 0.065).

Discussion

In this study, the diagnostic performance of yMR-EMVI (areaunder the receiver operating characteristic curve [AUROC]0.830) was better than that of MR-EMVI (AUROC, 0.778)for predicting ypEMVI, even though the difference was notstatistically significant (p = 0.076). This is likely because therewas a time interval between pre-CRTMRI and histopatholog-ical examination of surgical specimens, during which treat-ment was administered. Moreover, treatment response toCRT can vary patient among patients, and it is generally dif-ficult to predict the responsiveness before treatment. The

diagnostic performance ofMR or yMR-EMVI, based on path-ological results as a standard of reference, has not been report-ed previously. In previous literature, histopathological corre-lation was performed only in terms of TNM-staging and CRM[15]. The good performance of yMR-EMVI in predictingypEMVI observed in this study may alleviate concerns aboutevaluating EMVI in post-CRT cases, where there may betreatment-related oedema and inflammation. In addition,MRI examinations were performed from July 2007 toNovember 2010 in this study. Given that MRI technologyhas developed markedly over the past several years in termsof temporal and spatial resolutions, detecting tumour signalswithin the vessel or smaller venous structure, as well as otheranatomical structures, has become easier and more precise[19, 20]; thus our result is reasonable.

In terms of prognosis, yMR-EMVI (P = 0.004) wasthe only significant factor that correlated with DFSamong the EMVI-related MRI findings in univariateanalysis. In contrast, yMR-EMVI did not remain anindependent factor in multivariate analysis that includ-ed clinical and pathological findings, such as tumourstaging, ypEMVI, or CRM involvement in our study.This differs slightly from the recent literature, whichshowed that yMR-EMVI was an independent prognos-tic factor (P = 0.044) [14]. However, EMVI mostlyoccurs in T3 or T4 cancer, because EMVI is definedas the presence of malignant cells within endothelialcell-lined blood vessels beyond the muscularis propria[5]. yMR-EMVI more frequently develops in high-stage cancer, and is an MRI finding for predictingypEMVI; thus, it may be difficult to avoid the con-founding effect from tumour staging or presence ofypEMVI.

Fig. 6 Differences in disease-freesurvival (DFS) according to theseverity of yMR-EMVI in yMR-EMVI-positive patients

Eur Radiol

When we performed subgroup analysis in MR- oryMR-EMVI-positive patients, the severity of EMVIseemed to be an important prognostic factor, related toprognosis. Currently, we routinely evaluate the presenceor absence of EMVI on MRI according to previous liter-ature or guidelines [7, 10, 21]. The severity of EMVI canbe considered for predicting prognosis of rectal cancerpatients and deciding optimal postoperative treatment.However, further study and more evidence will be re-quired to establish this kind of policy [22].

In contrast to the previous study by Chand et al. [15], EMVItumour regression grading assessed by MRI (MR-vTRG) wasnot a significant factor predicting patients’ DFS (P = 0.13,Table 1) andOS (P= 0.49, Table 2) in our study. In their previousstudy [15], MR-vTRG was the only significant factor in im-proved 3-year DFS, confirmed by multivariate analysis usingCox regression (P = 0.013). Given that there were only 18 pa-tients (18/200, 9%) who showed a poor response in EMVI re-gression in our study, the statistical power may have been affect-ed by the small sample size. Therefore, future studies with largerstudy populations are necessary to identify the prognostic valueof MR-vTRG in terms of long-term DFS as well as OS.

The study had several limitations. First, the study popula-tion was heterogeneous, because we enrolled patients retro-spectively. Asmentioned above, the chemotherapy regimen orpostoperative management protocol varied. However, the pro-cess of MRI review was performed independently by twoexperienced radiologists under blindness of clinical data.Second, the objective imaging quality of MRI was heteroge-neous, given that several MRI systems were used, and do notconform to state-of-the-art imaging techniques. However, itwas adequate to assess venous invasion into large vessels.Thirdly, we could not exclude the effects of confounding fac-tors that were perfectly associated with prognosis, such astumour location, staging, performance status, etc., becauseof limitations imposed by the clinical data and number ofpatients.

In conclusion, the diagnostic performance of yMR-EMVIwas good, based on pathological results as a standard of refer-ence. The yMR-EMVI was the only significant EMVI-relatedMRI finding, correlated with patients’ DFS in univariateanalysis.

Compliance with ethical standards

Guarantor The scientific guarantor of this publication is Dr.Min JuKim

Conflict of interest The authors of this manuscript declare no relation-ships with any companies, whose products or services may be related tothe subject matter of the article.

Funding This study has received funding by a grant from the NationalCancer Centre of Korea (NCC-1710880)

Statistics and biometry No complex statistical methods were neces-sary for this paper.

Informed consent Written informed consent was waived by theInstitutional Review Board.

Ethical approval Institutional Review Board approval was obtained.

Methodology• retrospective• diagnostic or prognostic study• performed at one institution

References

1. Bokey EL, Chapuis PH, Dent OF et al (1997) Factors affectingsurvival after excision of the rectum for cancer: a multivariate anal-ysis. Dis Colon Rectum 40:3–10

2. Dresen RC, Peters EE, Rutten HJ et al (2009) Local recurrence inrectal cancer can be predicted by histopathological factors. Eur JSurg Oncol 35:1071–1077

3. Freedman LS, Macaskill P, Smith AN (1984) Multivariate analysisof prognostic factors for operable rectal cancer. Lancet 2:733–736

4. Horn A, Dahl O, Morild I (1991) Venous and neural invasion aspredictors of recurrence in rectal adenocarcinoma. Dis ColonRectum 34:798–804

5. Talbot IC, Ritchie S, Leighton MH, Hughes AO, Bussey HJ,Morson BC (1980) The clinical significance of invasion of veinsby rectal cancer. Br J Surg 67:439–442

6. Brown G, Radcliffe AG, Newcombe RG, Dallimore NS, BourneMW, Williams GT (2003) Preoperative assessment of prognosticfactors in rectal cancer using high-resolution magnetic resonanceimaging. Br J Surg 90:355–364

7. Smith NJ, Shihab O,Arnaout A, Swift RI, BrownG (2008)MRI fordetection of extramural vascular invasion in rectal cancer. AJR AmJ Roentgenol 191:1517–1522

8. BuggWG, Andreou AK, Biswas D, Toms AP,Williams SM (2014)The prognostic significance of MRI-detected extramural venousinvasion in rectal carcinoma. Clin Radiol 69:619–623

9. Kim YC, Kim JK, Kim MJ, Lee JH, Kim YB, Shin SJ (2016)Feasibility of mesorectal vascular invasion in predicting early dis-tant metastasis in patients with stage T3 rectal cancer based onrectal MRI. Eur Radiol 26:297–305

10. SmithNJ, BarbachanoY,NormanAR, Swift RI,AbulafiAM,BrownG(2008) Prognostic significance of magnetic resonance imaging-detectedextramural vascular invasion in rectal cancer. Br J Surg 95:229–236

11. Sohn B, Lim JS, Kim H et al (2015) MRI-detected extramural vas-cular invasion is an independent prognostic factor for synchronousmetastasis in patients with rectal cancer. Eur Radiol 25:1347–1355

12. Group MS (2006) Diagnostic accuracy of preoperative magneticresonance imaging in predicting curative resection of rectal cancer:prospective observational study. BMJ 333:779

13. Kim YE, Park MS, Hong HS et al (2009) Effects of neoadjuvantcombined chemotherapy and radiation therapy on the CTevaluationof resectability and staging in patients with pancreatic head cancer.Radiology 250:758–765

14. Chand M, Evans J, Swift RI et al (2015) The prognostic signifi-cance of postchemoradiotherapy high-resolution MRI and histopa-thology detected extramural venous invasion in rectal cancer. AnnSurg 261:473–479

15. ChandM, Swift RI, Tekkis PP, Chau I, Brown G (2014) Extramuralvenous invasion is a potential imaging predictive biomarker of neo-adjuvant treatment in rectal cancer. Br J Cancer 110:19–25

Eur Radiol

16. Nelson H, Petrelli N, Carlin A et al (2001) Guidelines 2000 forcolon and rectal cancer surgery. J Natl Cancer Inst 93:583–596

17. Compton CC, Fielding LP, Burgart LJ et al (2000) Prognostic fac-tors in colorectal cancer. College of American PathologistsConsensus Statement 1999. Arch Pathol Lab Med 124:979–994

18. Dworak O, Keilholz L, HoffmannA (1997) Pathological features ofrectal cancer after preoperative radiochemotherapy. Int J ColorectalDis 12:19–23

19. Kaur H, Choi H, You YN et al (2012) MR imaging for preoperativeevaluation of primary rectal cancer: practical considerations.Radiographics 32:389–409

20. Kluza E, Kleijnen JP, Martens MH et al (2016) Non-invasive MRassessment of macroscopic and microscopic vascular abnormalities inthe rectal tumour-surrounding mesorectum. Eur Radiol 26:1311–1319

21. Nougaret S, Reinhold C,Mikhael HW, Rouanet P, Bibeau F, BrownG (2013) The use of MR imaging in treatment planning for patientswith rectal carcinoma: have you checked the "DISTANCE"?Radiology 268:330–344

22. Sclafani F, Brown G, Cunningham D et al (2016) PAN-EX: apooled analysis of two trials of neoadjuvant chemotherapy follow-ed by chemoradiotherapy in MRI-defined, locally advanced rectalcancer. Ann Oncol 27:1557–1565

Eur Radiol