Magnetic Resonance Imaging–Detected Tumor Response for ... · For lower-third rectal tumors, the deﬁnition of predicted mrCRM involve-ment was tumor within 1 mm of the levator

Magnetic Resonance Imaging–Detected Tumor Responsefor Locally Advanced Rectal Cancer Predicts SurvivalOutcomes: MERCURY ExperienceUday B. Patel, Fiona Taylor, Lennart Blomqvist, Christopher George, Hywel Evans, Paris Tekkis, Philip Quirke,David Sebag-Montefiore, Brendan Moran, Richard Heald, Ashley Guthrie, Nicola Bees, Ian Swift, Kjell Pennert,and Gina Brown

Uday B. Patel, Paris Tekkis, KjellPennert, Gina Brown, Royal MarsdenHospital, Chelsea and Sutton; FionaTaylor, Nicola Bees, Ian Swift, MaydayUniversity Hospital, Croydon; Christo-pher George, Epsom General Hospital,Epsom; Hywel Evans, Frimley ParkHospital, Frimley; Phillip Quirke, LeedsInstitute of Molecular Medicine, Univer-sity of Leeds; David Sebag-Monegiore,Ashley Guthrie, St James Institute ofOncology, Leeds; Brendan Moran,Richard Heald, Pelican Cancer Founda-tion, North Hampshire Hospital,Basingstoke, United Kingdom; LennartBlomqvist, Karolinska UniversityHospital, Solna, Stockholm, Sweden.

Submitted February 2, 2011; acceptedJune 22, 2011; published online aheadof print at www.jco.org on August 29,2011.

Supported by educational grants fromSiemens Medical UK and the PelicanCancer Foundation (for originalMERCURY study), by a program grantfrom Yorkshire Cancer Research (P.Q.),and by the National Institute for HealthResearch Biomedical research center(G.B. and U.B.P.).

Authors’ disclosures of potential con-flicts of interest and author contribu-tions are found at the end of thisarticle.

Corresponding author: Gina Brown,MBBS, MD, FRCR, Dept of Radiology,The Royal Marsden Hospital NHS Trust,Sutton, United Kingdom, SM2 5PT; e-mail:[email protected].

© 2011 by American Society of ClinicalOncology

0732-183X/11/2999-1/$20.00

DOI: 10.1200/JCO.2011.34.9068

A B S T R A C T

PurposeTo assess magnetic resonance imaging (MRI) and pathologic staging after neoadjuvant therapy forrectal cancer in a prospectively enrolled, multicenter study.MethodsIn a prospective cohort study, 111 patients who had rectal cancer treated by neoadjuvant therapywere assessed for response by MRI and pathology staging by T, N and circumferential resectionmargin (CRM) status. Tumor regression grade (TRG) was also assessed by MRI. Overall survival(OS) was estimated by using the Kaplan-Meier product-limit method, and Cox proportional hazardsmodels were used to determine associations between staging of good and poor responders onMRI or pathology and survival outcomes after controlling for patient characteristics.ResultsOn multivariate analysis, the MRI-assessed TRG (mrTRG) hazard ratios (HRs) were independentlysignificant for survival (HR, 4.40; 95% CI, 1.65 to 11.7) and disease-free survival (DFS; HR, 3.28; 95%CI, 1.22 to 8.80). Five-year survival for poor mrTRG was 27% versus 72% (P � .001), and DFS for poormrTRG was 31% versus 64% (P � .007). Preoperative MRI-predicted CRM independently predictedlocal recurrence (LR; HR, 4.25; 95% CI, 1.45 to 12.51). Five-year survival for poor post-treatmentpathologic T stage (ypT) was 39% versus 76% (P � .001); DFS for the same was 38% versus 84%(P � .001); and LR for the same was 27% versus 6% (P � .018). The 5-year survival for involved pCRMwas 30% versus 59% (P � .001); DFS, 28 versus 62% (P � .02); and LR, 56% versus 10% (P � .001).Pathology node status did not predict outcomes.ConclusionMRI assessment of TRG and CRM are imaging markers that predict survival outcomes for goodand poor responders and provide an opportunity for the multidisciplinary team to offer additionaltreatment options before planning definitive surgery. Postoperative histopathology assessment ofypT and CRM but not post-treatment N status were important postsurgical predictors of outcome.

J Clin Oncol 29. © 2011 by American Society of Clinical Oncology

INTRODUCTION

In rectal cancer, long-course radiotherapy or preop-erative combination chemoradiotherapy togetherwith total mesorectal excision (TME) surgery isassociated with improved curative surgery andreduced local recurrence (LR).1 Cross-sectionalimaging techniques, including magnetic resonanceimaging (MRI), enable appropriate selection of pa-tients on the basis of assessment of locally advanceddisease and the relationship of tumor to the poten-tial resection margin.2

Assessment of treatment efficacy has princi-pally relied on histopathologic assessment of irradi-ated specimens after surgery. These studies have

shown that post-treatment pathologic T and N stage(ypT and ypN, respectively) can predict LR, disease-free survival (DFS), and overall survival (OS).3 It hasalso been observed that qualitative assessment of thedegree of fibrosis in the pathology specimen can beused to derive a tumor regression grading (TRG)system,4,5 which predicts OS and is considered animportant prognostic predictor for DFS.6 High-resolution MRI has been used to assess tumor re-sponse before surgical resection, but the relevanceof post-treatment MRI assessment in predictingsurvival outcomes has not been investigated.7-10

Furthermore, by applying the principles of histo-pathologic grading and by exploiting the charac-teristic MRI low signal intensity appearances of

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

© 2011 by American Society of Clinical Oncology 1

http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2011.34.9068The latest version is at Published Ahead of Print on August 29, 2011 as 10.1200/JCO.2011.34.9068

Copyright 2011 by American Society of Clinical Oncology

Information downloaded from jco.ascopubs.org and provided by at ASCO on February 10, 2016 from 66.102.234.242Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

http://jco.ascopubs.org/cgi/doi/10.1200/JCO.2011.34.9068

fibrosis, it has been possible to develop a similar MRI-basedTRG system.11

In 2002, a prospective study evaluated the diagnostic accuracy ofMRI staging in a consecutive series of patients with biopsy-provenrectal cancer undergoing TME surgery with histopathology as thereference standard. We now report a planned subgroup analysis ofMRI assessment of post-treatment scans in patients enrolled on thisfollow-up study.9 The aims were to evaluate the prognostic relevance,as judged by OS, DFS, and LR, of post–neoadjuvant therapy MRIassessment of tumor stage, nodal status, CRM, and MRI assessment ofTRG (mrTRG) system in patients undergoing neoadjuvant therapyand TME surgery in the MERCURY (Magnetic Resonance Imaging inRectal Cancer European Equivalence Study) trial.

METHODS

Patients

Patients were enrolled onto the original prospective study to evaluate diag-nostic accuracy of MRI staging of rectal cancer between February 2002 and Octo-ber 2003. The local scientific and research ethics committees of each institutionapproved the study. Eleven specialist colorectal units in five European countriesparticipatedinthetrial.Consentwassoughtforcontinuedpatient follow-upwhenthe diagnostic accuracy study was completed. Figure 1 summarizes recruitment.The 5-year follow-up was completed in December 2008.

Inclusion Criteria for This Subgroup Analysis

Biopsy-proven rectal cancer was considered by multidisciplinary team ashigh risk and requiring preoperative therapy for downstaging or downsizing ofthe primary tumor staged on baseline pretreatment scans as: advanced T3c,T3d, or T4, potential CRM involvement on MRI. Patients undergoing preop-erative chemoradiotherapy or long-course radiotherapy preoperative therapy(n � 111 of 374 patients) were included.

Exclusion Criteria for This Subgroup Analysis

Primary surgery or surgery after short-course preoperative radiotherapyand distant metastatic disease on body computed tomography were exclusioncriteria. Contraindications to MRI were also exclusion criteria.

Preoperative Treatment

Preoperative treatment allocated to patients was recorded. All centersoffered long-course preoperative therapy for patients with MRI-identifiedpotential CRM involvement. Some centers also offered preoperative therapy(chemoradiotherapy or radiotherapy) for MRI-predicted T3 tumors withadverse features, such as extramural venous invasion, extramural spreadgreater than 5 mm, or N2 disease.

Radiotherapy consisted of 45 to 54 Gy delivered to the primary tumor andpelviclymphnodesatrisk.Fifty-eightpatientsand53patientsreceivedlong-courseradiotherapy or concurrent chemoradiotherapy with a fluoropyrimidine, respec-tively.MRIscanswereperformedaftercompletionofradiotherapyatanintervalof4 to 6 weeks, and surgery was undertaken at 6 to 8 weeks after chemoradiotherapy.

Surgery

Surgical options included standard TME plane surgery or extended TME(ie, TME with adjacent visceral resection) with or without sphincter preserva-tion according to a standardized technique. The unit policies were to offerpostoperative single-agent chemotherapy for stage III disease. No patientreceived postoperative radiotherapy.

MRI Assessment

MRI TN stage and circumferential resection margin status after chemo-radiotherapy were assessed. A 1-day imaging workshop for specialist gastroin-testinal radiologists participating in the study was held before recruitment in2002 to ensure standardization of scan acquisition techniques (Appendix,online only) and image interpretation before the study, using our publishedcriteria.2,12,13 All scans were single-read by each investigating center radiolo-gist, comprising 5 to 15 years gastrointestinal radiology experience.

MRI-assessed T staging of tumor post treatment (ymrT) was based oninterpretation of local extent of persistent tumor signal intensity relative to thelayers of bowel wall on T2-weighted images. T substaging for both MRI (ymrTstage, after preoperative treatment but before surgery) and pathology (ypTstage assessed after surgery) were standardized (Appendix, online only).

Nodal stage post treatment was based on interpretation of lymph nodeborder characteristics and signal intensity.14 A node was regarded as positive ifeither an irregular border or mixed signal intensity was demonstrated.

Post-treatment MRI scans were evaluated for predicted circumferentialresection margin status (mrCRM). A clear mrCRM was defined if the distanceof tumor to the mesorectal fascia was greater than or equal to 1 mm on MRI.For lower-third rectal tumors, the definition of predicted mrCRM involve-ment was tumor within 1 mm of the levator muscle. If the tumor was presentat or below the level of the puborectalis sling, the mrCRM was predicted asinvolved if there was invasion into the intersphincteric plane or beyond. ThemrCRM status was recorded prospectively in the study by each participatingradiologist. When a post-treatment scan was not performed, the CRM atbaseline was entered as CRM status. All MRI-assessed T, N, and CRM stagingdata were obtained prospectively by workshop-trained radiologists.

mrTRG

Of 111 patients undergoing preoperative therapy, 92 also underwent post-treatmentMRIbeforesurgery.Sixty-six(72%)of92pre-andpost-treatmentscanswere available for central review of mrTRG. MRI scans were made anonymousand were centrally reviewed by a radiologist (G.B.) with 14 years of experience inMRI assessment of rectal cancers by using previously defined criteria.11

To assess the reproducibility of the mrTRG, pre- and post-treatmentscans were reviewed independently by two radiologists (C.D.G. and H.E.) with5 to 15 years experience in MRI staging of rectal cancers and were comparedwith the reviews of the central reviewer.

mrTRG was based on similar principles to the pathologic TRG originallydescribed by Dworak et al.4 Scans were reviewed to determine the degree oftumor replacement by fibrotic stroma (Appendix Fig A1, online only).4

Histopathologic Assessment

After TME, the specimen was axially sectioned into 3- to 5-mm slices, asdescribed by Quirke et al.15 A clear pathologic CRM (pCRM) was defined asgreater than or equal to 1 mm between the tumor and the resection margin.Eighteen pathologists with 5 to 25 years of experience in GI pathology evalu-ated the resected specimens for the post-treatment T and N stages (ie, ypT andypN, respectively) and the circumferential margin status (ie, pCRM).

Consented to follow-up(N = 386)

Lost to follow-up(n = 12)

Short-courseradiotherapy → surgery

(n = 47)

Primary surgery(n = 216)

Post-treatment MRI(n = 90)

Complete pathology and baselineMRI available

(n = 374)

Preoperative chemoradiotherapy/long-course radiotherapy for

locally advanced disease(n = 111)

Fig 1. Trial progress flowchart. MRI, magnetic resonance imaging.

Patel et al

2 © 2011 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY


Table 1. Patient Demographics and Survival Outcomes in 111 Patients Undergoing Preoperative Therapy in the MERCURY Study

Variable No.

OS DFS LR

Rate 95% CI P Rate 95% CI P Rate 95% CI P

Age, years� 65 49 58 43 to 72 51 36 to 66 21 8 to 35� 65 62 49 36 to 62 60 47 to 73 16 6 to 27

SexMale 69 49 37 to 62 48 35 to 61 .04 21 9 to 33Female 42 59 44 to 74 68 53 to 83 .04 16 4 to 28

TreatmentLong course 58 53 39 to 66 53 39 to 67 19 8 to 31Chemotherapy and RT 53 54 40 to 68 58 44 to 72 18 5 to 30

Tumor height, cm� 5 59 49 35 to 62 52 37 to 66 17 5 to 30� 5 52 57 44 to 71 60 46 to 74 21 8 to 33

OperationTME anterior resection 62 50 37 to 63 51 37 to 64 14 4 to 25TME � AP excision 49 56 42 to 70 62 47 to 76 24 10 to 38

Baseline MRI T stageT1-2 10 78 58 to 97 .048 72 51 to 93 13 0 to 31T3a 8T3b 11 49 38 to 60 .048 51 40 to 63 19 10 to 29T3c 29T3d 16T4 37

MRI node stageN0 68 52 36 to 69 56 38 to 73 22 7 to 38N1-2 41 54 42 to 66 54 42 to 66 17 7 to 27Missing 2

ymrT stageT0 6 73 54 to 92 72 52 to 91 20 2 to 38T1-2 13T3a 4T3b 14 48 35 to 60 50 37 to 64 16 6 to 27T3c 22T3d 9T4 22Missing 21

ymrN stageN0 50 61 47 to 76 63 49 to 78 .027 18 5 to 33N1-2 40 45 29 to 61 46 29 to 63 .027 17 3 to 32Missing 21

mrCRMClear 55/64 59 46 to 71 58 46 to 71 12 3 to 22 .013Clear baseline only 9/64Involved 37/47 46 31 to 61 51 35 to 67 28 13 to 44 .013Involved baseline only 10/47

mrTRGGood 32 72 56 to 88 .001 64 47 to 82 .007 14 1 to 27Poor 34 27 8 to 47 .001 31 13 to 49 .007 29 8 to 49Missing 45

ypT stageT0 5 76 62 to 89 .0001 84 72 to 96 .0001 6 0 to 15 .023T1 2T2 25T3a 9T3b 20 39 26 to 51 .0001 38 25 to 51 .0001 27 14 to 40 .023T3c 30T3d 4T4 16

ypN stageN0 67 56 43 to 68 64 52 to 77 16 6 to 26N1-2 44 48 33 to 64 42 26 to 58 77 61 to 93

Pathology CRMClear 88 59 48 to 70 .001 62 51 to 73 .023 10 3 to 18 .0001Involved 23 30 10 to 49 .001 28 7 to 50 .023 56 30 to 83 .0001

Abbreviations: AP, abdominoperineal excision; CRM, circumferential resection margin; DFS, disease-free survival; LR, local recurrence; MERCURY, MagneticResonance Imaging in Rectal Cancer European Equivalence Study; mrCRM, MRI-assessed CRM status; MRI, magnetic resonance imaging; mrTRG, MRIassessment of TRG; OS, overall survival; RT, radiotherapy; TME, total mesorectal excision; TRG, tumor regression grading; ymrN, MRI-assessed N staging posttreatment; ypN, pathologic N stage post treatment; ypT, pathologic T stage post treatment.

MRI Tumor Response in Rectal Cancer After Neoadjuvant Therapy

www.jco.org © 2011 by American Society of Clinical Oncology 3Information downloaded from jco.ascopubs.org and provided by at ASCO on February 10, 2016 from 66.102.234.242

Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

Analysis of Data

Tumors were categorized into good and poor responders to enablebinary comparison by multivariate analysis. On the basis of known histo-pathologic outcomes according to ypT stage, good ypT or ymrT stage wasdefined as stages T0, T1, T2, and T3a; poor was defined as ypT or ymrT stagesT3b, T3c, T3d, or T4. Stages T3a and T2 tumors have similar outcomes andtherefore are classified as good.16,17

MRI-predicted involvement of CRM and pCRM involved were definedas poor, whereas clear pCRM and mrCRM were classified as good. BothmrTRG-4 (predominantly tumor signal intensity with minimal fibrotic low-signal intensity) and -5 (no fibrosis evident; tumor signal visible only) includedpatients who had predominant tumor and minimal/no fibrosis; these weredefined as poor, whereas mrTRG-1 (absence of any tumor signal) to mrTRG-3(mixed areas of low-signal fibrosis and intermediate signal intensity presentbut without predominance of tumor signal) included patients who had 50% orgreater fibrotic stroma and therefore were defined as good. Pathologic TRGwas not undertaken.

Follow-Up

Patients were observed according to local protocols; typically, this com-prised outpatient assessment at intervals of 3 months for a duration of 2 yearsthen intervals of 6 months for a duration of 5 years. Clinical follow-up com-prised physical examination, routine blood tests, and yearly computed tomog-raphy of the thorax, abdomen, and pelvis. Histopathologic confirmation of LRand distant recurrence was sought when feasible. The surgical data, date ofoperation, procedure performed, and quality of the specimen were recordedprospectively. Date of enrollment, last follow-up, date of disease progression,and date and cause of death were also collected. The presence of distantmetastatic disease or LR at the time of death was recorded as rectal cancer–specific death.

Statistical Analysis

LR was measured from date of original trial enrollment until local pro-gression; DFS was measured from date of enrollment until progression at anysite or death as a result of any cause; patients who were alive and disease freewere censored at last follow-up or death.

Cox proportional hazard models were used to analyze the effect on LR,DFS, and OS by preoperative MRI staging variables (ie, ymrT, ymrN, mrCRMand mrTRG) and pathologic variables (ie, ypT, ypN, pCRM). Other potentialfactors—age, sex, type of surgery (TME or abdominoperineal resection withTME approach), tumor height (� 5 cm or � 5 cm), and treatment (chemo-radiotherapy or long-course radiation)—were included in the multivariateanalysis for each MRI/pathology staging variable.

Survival curves were calculated by using the Kaplan-Meier product limitmethod. Differences between survival curves were tested by using the univar-iate log-rank test. Correlation between ymrT and mrTRG with ypT was calcu-lated by using the Spearman correlation coefficient.

Agreement between the two observers who graded mrTRG was determinedbythe� statistic. (��0,pooragreement;��0to0.20, slightagreement;��0.21to 0.40, fair agreement; � � 0.41 to 0.60, moderate agreement; � � 0.61 to 0.80,substantial agreement; and � � 0.81 to 1.00, almost perfect agreement).

P less than .05 were considered significant. Calculations were performedby using Statistical Package for Social Sciences program, version 18.0 (SPSS,Chicago, IL).

RESULTS

Eighty-one (73%) of 111 patients were considered to have potentialmrCRM involvement before neoadjuvant therapy; this was reduced to47 (42%) of 111 after treatment. The number of mrT3c, T3d, and T4

Table 2. Multivariate Analysis of HRs for Outcome According to Preoperative Post-Therapy MRI-Assessed Staging Variables

Variable

Post-TreatmentymrT Stage

Post-TreatmentymrN Stage mrTRG Stage Post-Treatment mrCRM

OS DFS LR OS DFS LR OS DFS LR OS DFS LR

MRIymrT or mrTRG good/ymrN or mrCRM negative 1 1 1 1 1 1 1 1 1 1 1 1ymrT or mrTRG poor/ymrN or mrCRM positive 2.27 1.82 0.91 1.89� 2.09† 1.08 4.40‡ 3.28§ 3.71 1.76 1.08 4.25�

95% CI for HR 1.006 to 3.58 1.06 to 4.15 1.65 to 11.7 1.22 to 8.80 1.45 to 12.51Age, years

� 65 1 1 1 1 1 1 1 1 1 1 1 1� 65 1.48 0.88 0.59 1.58 1.01 0.61 1.46 0.98 0.65 1.57 0.93 1.10

SexMale 1 1 1 1 1 1 1 1 1 1 1 1Female 0.88 0.59 0.931 0.78 0.53 0.92 0.70 0.67 0.80 0.76 0.46¶ 1.16

95% CI 0.24 to 0.89Treatment

Chemotherapy and RT 1 1 1 1 1 1 1 1 1 1 1 1LCRT 0.98 0.93 0.99 1.20 1.07 0.94 0.69 0.69 0.41 0.86 0.89 0.51

OperationAnterior resection 1 1 1 1 1 1 1 1 1 1 1 1AP excision 1.36 0.85 2.69 1.24 0.83 2.75 0.81 0.58 1.69 0.99 0.56 2.05

Height from anal verge, cm� 5 1 1 1 1 1 1 1 1 1 1 1 1� 5 0.92 1.37 0.95 0.88 1.26 0.97 1.05 1.19 0.99 0.73 1.30 1.08

Abbreviations: AP, abdominoperineal excision; DFS, disease-free survival; HR, hazard ratio; LCRT, long-course chemoradiotherapy; LR, local recurrence; mrCRM,MRI-assessed circumferential resection margin status; MRI, magnetic resonance imaging; mrTRG, MRI-assessed tumor regression grading; OS, overall survival; RT,radiotherapy; ymrN stage, MRI assessment of N staging of tumor post treatment; ymrT stage, MRI assessment of T staging of tumor post treatment.

�P � .048.†P � .033.‡P � .003.§P � .019.�P � .009.¶P � .022.

Patel et al



tumors reduced from 80 of 111 to 54 of 111. Table 1 shows demo-graphic characteristics and survival outcomes of 111 patients includedin the analysis. As of December 2008, surviving patients had beenobserved for a median of 50 months (range, 0.6 to 71 months). All 111patients had evaluable post-operative pathology and had completefollow-up data. During the follow-up period, 54 patients died; 40 diedas a result of cancer-related causes, five died as a result of noncancerdeaths, four died as a result of peri- or post-operative causes, and threepatients died as a result of unknown causes. Fifty patients had diseaseprogression; 42 patients experienced recurrence with metastatic dis-ease, and eight experienced relapse with LR only.

Table 2 summarizes the results of multivariate analysis of knownclinical variables—age, sex, height of tumor from anal verge, type ofpreoperative treatment, and type of operation—according to each ofthe MRI preoperative staging variables and survival outcomes. ThemrTRG was significant for OS and DFS; hazard ratios were 4.40 (95%CI, 1.65 to 11.7) and 3.28 (95% CI, 1.22 to 8.80), respectively.

OS at 5 years for patients with poor mrTRG was 27% (95% CI,8% to 47%) compared with 72% (95% CI, 56% to 88%) for patientswith good mrTRG (P � .001 on univariate log-rank analysis). DFS at5 years for patients with poor mrTRG was 31% (95% CI, 13% to 49%)compared with 64% (95% CI, 47% to 82%) for patients with goodmrTRG (P � .007 on univariate log-rank analysis).

On multivariate analysis, post-treatment prediction of mrCRMwas significant for LR. The hazard ratio for predicted involved mr-CRM and LR was 4.25 (95% CI, 1.45 to 12.51). LR rates at 5 years for

patients with predicted involved mrCRM was 28% (95% CI, 13% to44%) compared with 12% (95% CI, 3% to 22%) for patients withpredicted clear mrCRM (P � .013 on univariate log-rank analysis).

On multivariate analysis, post-treatment prediction of mrNshowed borderline statistical significance for OS and DFS, but it wasnot significant for LR. Hazard ratios were 1.89 (95% CI, 1.006 to 3.58)and 2.09 (95% CI, 1.06 to 4.15) for OS and DFS, respectively.

Table 3 summarizes results of multivariate analysis of associationbetween OS, DFS, and time to LR of known clinical variables—age,sex, height of tumor from anal verge, type of preoperative treatment,and type of operation—according to each of the histopathologic stag-ing variables. On multivariate analysis, ypT remained significant forOS, DFS, and LR; hazard ratios for poor ypT and OS, DFS, and LRwere 4.59 (95% CI, 2.16 to 9.78), 5.63 (95% CI, 2.45 to 12.9), and 9.78(95% CI, 2.29 to 41.8), respectively. Kaplan-Meier survival curvesaccording to ypT stage are shown in Figure 2.

The 5-year OS for patients with poor ypT was 39% (95% CI, 26%to 51%) compared with 76% (95% CI, 62% to 89%) for good ypTresponse (P � .0001 on univariate log-rank analysis). The 5-year DFSfor patients with poor ypT was 38% (95% CI, 25% to 51%) comparedwith 84% (95% CI, 72% to 96%) good ypT response (P � .0001 onunivariate log-rank analysis). The 5-year LR rate was 27% (95% CI,14% to 40%) for patients with poor ypT compared with 6% (95% CI,0% to 15%) for good ypT response (P � .018 on univariate log-rank analysis).

Table 3. Multivariate Analysis of HRs for Outcome According to Histopathologically Assessed Variables

Variable

Pathology T Stage (ypT) Pathology N Stage (ypN) pCRM

OS DFS LR OS DFS LR OS DFS LR

PathologyypT good/ypN or pCRM negative 1 1 1 1 1 1 1 1 1ypT poor/ypN or pCRM negative 4.59� 5.63� 9.78� 1.53 1.71 1.49 2.99� 2.17† 8.80�

95% CI 2.16 to 9.78 2.45 to 12.9 2.29 to 41.8 1.64 to 5.47 1.14 to 4.13 2.99 to 25.91Age, years

� 65 1 1 1 1 1 1 1 1 1� 65 1.98† 1.32 1.37 1.49 0.97 0.91 1.74 0.98 1.65

95% CI 1.12 to 3.52Sex

Male 1 1 1 1 1 1 1 1 1Female 0.87 0.58 1.18 0.68 0.44‡ 0.82 0.67 0.44† 0.85

95% CI 0.23 to 0.85 0.23 to 0.85Treatment

Chemotherapy and RT 1 1 1 1 1 1 1 1 1LCRT 0.67 0.49§ 0.31� 1.06 0.99 0.73 0.93 0.87 0.61

95% CI 0.26 to 0.94 0.55 to 6.70Operation

Anterior resection 1 1 1 1 1 1 1 1 1AP excision 1.04 0.63 1.91 1.05 0.62 1.91 0.86 0.51 1.48

Height from anal verge, cm� 5 1 1 1 1 1 1 1 1 1� 5 1.41 2.02 2.62 0.92 1.37 1.33 1.02 1.42 1.65

Abbreviations: AP, abdominoperineal excision; DFS, disease-free survival; HR, hazard ratio; LCRT, long-course chemoradiotherapy; LR, local recurrence; OS, overallsurvival; pCRM, pathology circumferential resection margin status; RT, radiotherapy; ypN, pathologic N stage post treatment; ypT, pathologic T stage post treatment.

�P � .001.†P � .02.‡P � .01.§P � .03.�P � .05.




On multivariate analysis, pCRM was significant for OS, DFS, andLR. Hazard ratios were 2.99 (95% CI, 1.64 to 5.47), 2.17 (95% CI, 1.14to 4.13), and 8.80 (95% CI, 2.99 to 25.91) respectively.

The 5-year OS for patients with involved pCRM was 30% (95% CI,10% to 49%) compared with 59% (95% CI, 49% to 70%) for patientswith clear pCRM (P � .001 on univariate log-rank analysis). The 5-yearDFS for patients with involved pCRM was 28% (95% CI, 7% to 50%)comparedwith62%(95%CI,51%to73%)forpatientswithclearpCRM(P � .023 on univariate log-rank analysis). The 5-year LR rate was 56%(95% CI, 30% to 83%) for patients with involved pCRM compared with10% (95% CI, 3% to 18%) for patients with clear pCRM (P � .0001 onunivariate log-rank analysis). MRI staging compared with pathologyCRM is shown in Appendix Table A1 (online only).

The ypN did not predict for any of the survival outcomes. Thecorrelation coefficient between mrTRG and ypT of 0.65 was greaterthan the correlation between ymrT and ypT of 0.48. The linear �between radiologists G.B. and C.D.G. was 0.65 (95% CI, 0.44 to 0.86),which indicated moderate to substantial agreement. The linear � be-tween radiologists G.B. and H.E. was 0.6 (95% CI, 0.31 to 0.89), whichindicated moderate agreement.

DISCUSSION

MRI assessment of TRG after preoperative therapy predicts DFS andOS, and thus patient prognosis, before definitive surgery. Post-

treatment MRI prediction of CRM involvement also gives importantprognostic information regarding the risk of LR. Our data also showthat both MRI T staging and TRG showed statistical correlation withypT. Results for T staging are consistent with recently published se-ries.7 However, it appears that the degree of tumor replacement byfibrosis correlates with survival at a greater statistical significance thanymrT stage does. Our experience suggests that assessing the degree oftumor replacement by fibrosis is more reliable than attempting todelineate T stage on MRI after treatment.

There are several pathologic TRG systems (eg, those proposed byDworak et al,4 and Mandard et al6), but all are based on the relativeproportion of fibrosis present in the resected specimen. The relation-ship between pathologic TRG and outcome has been considered im-portant in previous studies,18,19 for example, in predicting OS andDFS.6 Applying similar principles with MRI, we have now shown thatit is possible to assess tumor regression before surgery.

It is notable that neither histopathologic nor MRI assessment ofnodal status after radiotherapy or chemoradiotherapy predicted LR,and this differs from previously published studies.20,21 Studies fromthe pre-TME era convincingly demonstrate the strong relationshipbetween node positivity and pelvic recurrence, which has been attrib-uted to mesorectal tissue containing nodes and deposits left behind inthe pelvis.22 However, incomplete TME specimens are also associatedwith 20% pelvic recurrence rates for node-positive patients comparedwith only 6% for complete TME specimens.23 In the MERCURY

BA

0

P = .001

ypTFavorable (n = 41)Unfavorable (n = 70)

Over

all S

urvi

val

(pro

porti

on)

Time (months)

1.0

0.8

0.6

0.4

0.2

12 24 36 48 006

P = .001

ypTFavorable (n = 41)Unfavorable (n = 70)Di

seas

e-Fr

ee S

urvi

val

(pro

porti

on)

Time (months)

1.0

0.8

0.6

0.4

0.2

12 24 36 48 60

DC

0

P = .001

mrTRGFavorable (n = 32)Unfavorable (n = 34)

Over

all S

urvi

val

(pro

porti

on)

Time (months)

1.0

0.8

0.6

0.4

0.2

12 24 36 48 006

P = .007

mrTRGFavorable (n = 32)Unfavorable (n = 34)Di

seas

e-Fr

ee S

urvi

val

(pro

porti

on)

Time (months)

1.0

0.8

0.6

0.4

0.2

12 24 36 48 60

Fig 2. Kaplan-Meier analysis of survival. (A) Post-treatment pathologic T stage (ypT) and overall survival; (B) ypT and disease-free survival; (C) tumor regression gradeby magnetic resonance imaging (mrTRG) and overall survival; and (D) mrTRG and disease-free survival.

Patel et al



study, the TME specimen quality was audited, and an incomplete rateof only 5% was observed, which is significantly better than previouslypublished series.9,23,24 This may explain the observed lack of an effecton outcome by involved lymph nodes influencing pelvic recurrenceafter preoperative therapy.

The histopathologic variables of T staging and CRM involvementboth predicted OS, DFS, and LR. pCRM has previously been describedas a crucial prognostic factor after treatment.25 Our data showed that,of the 23 of 111 patients who had a positive CRM on histopathology,21 (91%) of the 23 also had diseases staged as ypT poor (� ypT3a).Thus, the two variables are highly interdependent, and nearly all thosepatients with involved resection margins were part of the larger poor-prognosis group (n � 70) associated with more advanced pathologicT stage.

The strengths of this study are that it represents high-qualityradiologic, pathologic, and surgical data ensured by training work-shops. The prospective nature of the study also represents a large seriesof patients who have undergone baseline and post-treatment (ie, che-moradiotherapy) MRI imaging as well as outcome data. However,there are potential limitations of this study. First, 66 (72%) of 92post-treatment MRI scans were available for a retrospective review ofmrTRG. Despite this, a representative sample were reviewed, andinterobserver agreement was moderate to substantial (� � 0.6 and0.65, respectively), indicating that this technique is reproducible. Nev-ertheless, prospective validation of TRG in predicting outcomes infuture studies will confirm its value. An additional limitation was lackof standardization of neoadjuvant therapy and use of both chemora-diotherapy and radiotherapy treatment approaches. Clearly, thiscould have influenced outcomes, and it would have been expected thatpatient receiving long-course radiotherapy would have responded lesswell. However, on multivariate analysis, the type of treatment did notmaterially influence the outcomes, and the prognostic importance ofmrTRG and ypT are independent of the type of treatment received.

This is the first time that a prospective study has demonstrated acorrelation between radiologically determined tumor response andlong-term outcomes. That this is achieved before surgical resectionprovides evidence for the design of future studies that could alterpatient therapy on the basis of response assessment. It also allowsidentification of good and poor response groups and differentapproaches that could be adopted after completion of chemora-diotherapy. For example, the role of systemic non–cross-resistantchemotherapy could be tested in patients with poor response, and theevaluation of the timing or even deferral of surgical resection could betested in patients with good response (ie, mrTRG-1 and -2).26 There-fore, post-treatment MRI TRG and CRM evaluation gives the multi-disciplinary team a window of opportunity to refine treatment plansbefore definitive surgical treatment.

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTSOF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Lennart Blomqvist, Hywel Evans, Philip Quirke,David Sebag-Montefiore, Brendan Moran, Richard Heald, AshleyGuthrie, Ian Swift, Gina BrownCollection and assembly of data: Uday B. Patel, Fiona Taylor, PhilipQuirke, David Sebag-Montefiore, Ashley Guthrie, Gina BrownData analysis and interpretation: Uday B. Patel, Fiona Taylor, LennartBlomqvist, Christopher George, Paris Tekkis, Ashley Guthrie, NicolaBees, Kjell Pennert, Gina BrownManuscript writing: All authorsFinal approval of manuscript: All authors

REFERENCES

1. Kapiteijn E, Marijnen CA, Nagtegaal ID, et al:Preoperative radiotherapy combined with total me-sorectal excision for resectable rectal cancer. N EnglJ Med 345:638-646, 2001

2. Brown G, Richards CJ, Newcombe RG, et al:Rectal carcinoma: Thin-section MR imaging for stag-ing in 28 patients. Radiology 211:215-222, 1999

3. Capirci C, Valentini V, Cionini L, et al: Prog-nostic value of pathologic complete response afterneoadjuvant therapy in locally advanced rectal can-cer: Long-term analysis of 566 ypCR patients. Int JRadiat Oncol Biol Phys 72:99-107, 2008

4. Dworak O, Keilholz L, Hoffmann A: Patho-logical features of rectal cancer after preoperativeradiochemotherapy. Int J Colorectal Dis 12:19-23,1997

5. Wheeler JM, Warren BF, Mortensen NJ, et al:Quantification of histologic regression of rectal can-cer after irradiation: A proposal for a modified stag-ing system. Dis Colon Rectum 45:1051-1056, 2002

6. Vecchio FM, Valentini V, Minsky BD, et al: Therelationship of pathologic tumor regression grade(TRG) and outcomes after preoperative therapy inrectal cancer. Int J Radiat Oncol Biol Phys 62:752-760, 2005

7. Barbaro B, Fiorucci C, Tebala C, et al: Locallyadvanced rectal cancer: MR imaging in prediction of

response after preoperative chemotherapy and radi-ation therapy. Radiology 250:730-739, 2009

8. Kim SH, Lee JM, Park HS, et al: Accuracy ofMRI for predicting the circumferential resectionmargin, mesorectal fascia invasion, and tumor re-sponse to neoadjuvant chemoradiotherapy for lo-cally advanced rectal cancer. J Magn Reson Imaging29:1093-1101, 2009

9. MERCURY Study Group: Diagnostic accu-racy of preoperative magnetic resonance imagingin predicting curative resection of rectal cancer:Prospective observational study. BMJ 10.1136/bmj.38937.646400.55 [epub on September 19,2006]

10. Yeo SG, Kim DY, Kim TH, et al: Tumorvolume reduction rate measured by magneticresonance volumetry correlated with pathologictumor response of preoperative chemoradiother-apy for rectal cancer. Int J Radiat Oncol Biol Phys78:164-171, 2010

11. Salerno G, Chandler I, Wotherspoon A, et al:Sites of surgical wasting in the abdominoperinealspecimen. Br J Surg 95:1147-1154, 2008

12. Brown G, Radcliffe AG, Newcombe RG, et al:Preoperative assessment of prognostic factors inrectal cancer using high-resolution magnetic reso-nance imaging. Br J Surg 90:355-364, 2003

13. Brown G, Davies S, Williams GT, et al: Effec-tiveness of preoperative staging in rectal cancer:Digital rectal examination, endoluminal ultrasound or

magnetic resonance imaging? Br J Cancer 91:23-29,2004

14. Koh DM, Chau I, Tait D, et al: Evaluatingmesorectal lymph nodes in rectal cancer before andafter neoadjuvant chemoradiation using thin-sectionT2-weighted magnetic resonance imaging. Int JRadiat Oncol Biol Phys 71:456-461, 2008

15. Quirke P, Durdey P, Dixon MF, et al: Localrecurrence of rectal adenocarcinoma due to inade-quate surgical resection: Histopathological study oflateral tumour spread and surgical excision. Lancet2:996-999, 1986

16. Willett CG, Badizadegan K, Ancukiewicz M, etal: Prognostic factors in stage T3N0 rectal cancer:Do all patients require postoperative pelvic irradia-tion and chemotherapy? Dis Colon Rectum 42:167-173, 1999

17. Hermanek PHD, Hutter RVP, Sobin LH: TNMSupplement. New York, NY, Springer-Verlag, 1993

18. Rodel C, Martus P, Papadoupolos T, et al:Prognostic significance of tumor regression afterpreoperative chemoradiotherapy for rectal cancer.J Clin Oncol 23:8688-8696, 2005

19. Suarez J, Vera R, Balen E, et al: Pathologicresponse assessed by Mandard grade is a betterprognostic factor than down staging for disease-freesurvival after preoperative radiochemotherapy foradvanced rectal cancer. Colorectal Dis 10:563-568,2008




20. Moran MR, James EC, Rothenberger DA, etal: Prognostic value of positive lymph nodes in rectalcancer. Dis Colon Rectum 35:579-581, 1992

21. Bujko K, Michalski W, Kepka L, et al: Associ-ation between pathologic response in metastaticlymph nodes after preoperative chemoradiotherapyand risk of distant metastases in rectal cancer: Ananalysis of outcomes in a randomized trial. Int JRadiat Oncol Biol Phys 67:369-377, 2007

22. Gunderson LL, Jessup JM, Sargent DJ, et al:Revised tumor and node categorization for rectal

cancer based on surveillance, epidemiology, andend results and rectal pooled analysis outcomes.J Clin Oncol 28:256-263, 2010

23. Quirke P, Steele R, Monson J, et al: Effect of theplane of surgery achieved on local recurrence in patientswith operable rectal cancer: A prospective study usingdata from the MRC CR07 and NCIC-CTG CO16 ran-domised clinical trial. Lancet 373:821-828, 2009

24. Nagtegaal ID, van de Velde CJ, van der WorpE, et al: Macroscopic evaluation of rectal cancerresection specimen: Clinical significance of the pa-

thologist in quality control. J Clin Oncol 20:1729-1734, 2002

25. Gosens MJ, Klaassen RA, Tan-Go I, et al:Circumferential margin involvement is the crucialprognostic factor after multimodality treatment inpatients with locally advanced rectal carcinoma. ClinCancer Res 13:6617-6623, 2007

26. O’Neill BD, Brown G, Heald RJ, et al: Non-operative treatment after neoadjuvant chemoradio-therapy for rectal cancer. Lancet Oncol 8:625-633,2007

■ ■ ■

Appendix

Methods. The protocol employed a thin, 3-mm section turbo spin-echo T2-weighted technique with a surface pelvic phased arraycoil and a small field of view. The machines and workstations used by each center are listed in this Appendix (online only). No bowelpreparation, air insufflations, or intravenous antispasmodic agents were routinely used. We did not use intravenous contrast enhance-ment, because this has not been found helpful (Brown G, et al: Br J Radiol 78:245-251, 2005; Vliegen RF, et al: Radiology 234:179-188,2005); nor did we use intrarectal contrast, such as water or gel.

For a 1.5-T MRI scanner, four sequences were used. After a coronal localizer, sagittal turbo (fast) spin-echo sequences from innerpelvic sidewall to sidewall with a 24-cm field of view, 5-mm contiguous/interleaved slices (no gap), time to repetition (TR) greater than2,500 ms and less than 5,000 ms, and time to echo (TE) of 85 ms were obtained. These acquisitions were used to plan thin-section obliqueaxial images.

Axial T2 Turbo fast spin-echo acquisitions of the anatomic pelvis by using a 24 cm field of view, a 5 mm contiguous section thickness,TR 4,000 ms, TE 85 ms, 512 � 256 matrix, an echo train length of eight, no fat saturation, a 32kHz bandwidth, and two signalsacquisitions.

The sagittal T2-weighted images obtained were then used to plan T2-weighted thin-section axial images through the rectal cancer.These images were performed perpendicular to the long axis of the rectum. These were obtained by using a 16-cm field of view, a 3-mmsection thickness, no intersection gap, TR 4,000 ms, TE 85 ms, a 256 � 256 matrix, an echo train length of eight, no fatsaturation, a 32-kHz bandwidth, and four acquisitions. These sequences were repeated with imaging in the coronal plane forall tumors arising at or below the levator muscle origin.

For a 1.0-T MRI scanner, the imaging parameters were modified to obtain an adequate SNR. The high-resolution images wereobtained with 20-cm field of view, 3-mm section thickness, no intersection gap, a 256 � 256 matrix, a TR greater than 2,500 ms but lessthan 5,000 ms, and a TE greater than 80 ms.

MRI definitions for image interpretation. T staging was as follows: T1 represented tumor or tumor signal intensity confined to thesubmucosal layer. The signal intensity was low compared with the high signal intensity of the adjacent submucosa. T2 represented tumoror tumor signal intensity that extended into the muscle layer. T3 represented tumor or tumor signal extending less than 1 mm beyond themuscularis propria, with obliteration of the interface between muscle and perirectal fat; substage mrT3a tumor was � 1 mm beyond themuscularis propria, mrT3b was 1 to 5 mm beyond, mrT3c was greater than 5 mm and � 15 mm beyond, and T3d was greater than 15 mmbeyond. T4 represented tumor or tumor signal that extended through the peritonealized surface of the rectum or into an adjacent organ.

Comparison was made with the pretreatment images. Tumor response manifested as either replacement by tumor signal as low signalintensity fibrosis (dark stroma) or the development of mucin pools representing tumor necrosis. Tumor was interpreted as any area ofpersistent intermediate signal intensity on high-resolution images, and T staging after treatment was denoted by ymrT.

Nodal assessment. Nodal stage both pretreatment and post-treatment was based on the interpretation of lymph node bordercharacteristics and signal intensity.14 A node was regarded as positive if either an irregular border or mixed signal intensity wasdemonstrated. Nodal staging after treatment was denoted by the prefix ymrN.

CRM. The post-treatment MRI scan was evaluated for predicted circumferential resection margin status (mrCRM). A predictedclear CRM was identified if the distance of tumor to the mesorectal fascia was greater than or equal to 1 mm on MRI. For lower-third rectaltumors, the definition of predicted mrCRM involvement was tumor within 1 mm of the levator muscle. If the tumor was present at orbelow the level of the puborectalis sling, the mrCRM was predicted involved if there was invasion into the intersphincteric planeor beyond.

TRG. mrTRG-1 was identified as the absence of any tumor signal. mrTRG-2 was identified as small amounts of residual tumorvisible but with a predominant fibrotic low signal intensity. mrTRG-3 was identified as mixed areas of low signal fibrosis and intermediatesignal intensity present but without predominance of tumor. mrTRG-4 was identified by predominantly tumor signal intensity remainswith minimal fibrotic low signal intensity. mrTRG-5 was identified as no fibrosis evident, tumor signal visible only.

Full acknowledgments. We thank all the assistants, secretaries, and nurse specialists who helped in collating the follow-up data for thisstudy: C. Hughes, Ashford and St Peters Hospital; S. Draheim, Krankenhaus im Freidrichshain; O. Shihab, North Hampshire Hospital;Simon Ambrose, St James’s University Hospital; Colorectal Surgical Department, Oslo data, Norwegian Radium Hospital; Yvette Perston,

Patel et al



Llandough Hospital; Lennart Blomqvist, Karolinska University Institute; Bijal Patel, Royal Marsden and Pelican; B. Bannerman, RoyalMarsden Sutton. We thank all the radiologists, pathologists, surgeons, and nurse specialists who were members of the MERCURYstudy group.

Participating hospitals and clinicians in the MERCURY study group. For original study coordination: I.R. Daniels, Research Fellow;S.E. Fisher, Research Nurse. Study statistician: A.R. Norman, Follow-Up Study Statistician; Kjell Penert. Epsom General Hospital, Epsom,Surrey: P. Toomey, Surgeon; M.A.. Raja, Surgeon; C.D. George, Radiologist, L. Temple, Pathologist; S. Woodward, Nurse Specialist.Mayday University Hospital, Croydon, Surrey: Fiona Taylor, Research Fellow and Follow-Up Study Coordinator; I. Swift, Surgeon; M.Abulafi, Surgeon; N. Bees, Radiologist; H. Blake, Radiologist; N. Jeyadeven, Radiologist; A. Arnaout, Pathologist. Royal Marsden Hospital:G. Brown, Radiologist; A. Wotherspoon, Pathologist; D. Cunningham, Medical Oncologist; M. Hill, Medical Oncologist; P. Ross, MedicalOncologist; D. Tait, Clinical Oncologist; A. Massey, Nurse Specialist. The North Hampshire Hospital, Basingstoke, Hampshire:R.J. Heald, Surgeon; B.J. Moran, Surgeon; T.D. Cecil, Surgeon; D.M. Gold, Surgeon; P.D. Peppercorn, Radiologist; J. Finch, Pathologist;I. Ilesley, Pathologist; G. Sharpe, Oncologist; A. Leppington-Clarke, Nurse Specialist. Leeds Hospitals Teaching Hospitals: St James’sUniversity Hospital, Leeds General Infirmary, Leeds, United Kingdom: P.J. Finan, Surgeon; P. Sagar, Surgeon; D, Burke, Surgeon; K.Sasapu, Research Fellow; A. Chalmers, Radiologist; P. Quirke, Pathologist; A. Cairns, Pathologist; D. Sebag-Montefiore, Oncologist; M.Seymour, Oncologist; N. Chauhan-Lall, YCRN Trials Coordinator; S. Ambrose, Surgeon; I. Botterill, Surgeon; D. Jayne, Surgeon; A.Guthrie, Radiologist; N. Scott, Pathologist; C. Verbeke, Pathologist. The Norwegian Radium Hospital, Oslo, Norway: J. Wiig, Surgeon; K.Kotanska-Groeholt, Pathologist; H. Emblemsvaag, Radiologist; T. Vetrhus, Radiologist; S. Larsen, Surgeon. Frimley Park Hospital,Camberley, Surrey: M. Gudgeon, Surgeon; D. Edwards, Surgeon; S. Mellor, Surgeon; H. Massouh, Radiologist; M. Elmahallawy,Pathologist; K. Bundy, Nurse Specialist; G. Middleton, Oncologist. Ashford St Peters Hospital, Chertsey, Surrey: D.R. Donaldson,Surgeon; H.J. Scott, Surgeon; P. Bearn, Surgeon; K. Galbraith, Surgeon; M. Creagh, Radiologist; S. Dodd, Pathologist; G. Middleton,Oncologist; L. De Snoo, Nurse Specialist. Krankenhaus im Friedrichshain, Berlin, Germany: J. Strassburg, Surgeon; K. Ludwig, Surgeon;A. Weskott, Surgeon; A. Lewin, Surgeon; M. Frei, Research Assistant; P. Knuth, Radiologist; U. Ruhl, Clinical Oncologist; M. Albrecht,Clinical Oncologist; M. Hackenthal, Medical Oncologist; P. Hellriegel, Medical Oncologist; J. Linke, Pathologist; V. Loy, Pathologist; C.Webert, German Cancer Society. Llandough Hospital, Cardiff, United Kingdom: A.G. Radcliffe, Surgeon; J. Torkington, Surgeon; R.Bleehen, Radiologist; N.S. Dallimore, Pathologist; T. Maughan, Oncologist; Y. Perston, Nurse Specialist; M.W. Bourne, Radiologist. TheKarolinska Hospital, Stockholm, Sweden: T. Holm, Surgeon; L. Blomqvist, Radiologist; J. Lindholm, Pathologist; M. Torkzad, Radiolo-gist; B. Glimelius, Oncologist; B. Andersson, Nurse Specialist; Y. Ericsson- Alm, Nurse Specialist.

Hospitals that took part in the study. The Royal Marsden Hospital NHS Foundation Trust; Epsom and St Helier NHS Trust; MaydayUniversity Hospital NHS Trust; The North Hampshire Hospital NHS Trust; Leeds Teaching Hospitals NHS Trust; Norwegian RadiumHospital; Frimley Park Hospital NHS Trust; Ashford St Peter’s Hospital NHS Trust; Llandough Hospital, Krankenhaus im Freidrichshain,Karolinska University Institute.

Table A1. MRI T Stage and CRM Status Before and After Treatment

Variable

Pathology CRM Status

PClear Involved

Preoperative therapyRadiotherapy 39 14 .157With neoadjuvant chemotherapy 49 9

ymrT stageGood 21 1 .062Poor 53 15

mrTRGGood 30 2 .015Poor 24 10

mrCRMClear 54 1 �.001Clear on baseline 9 0Involved 20 17Involved on baseline 5 5

Abbreviations: CRM, circumferential resection margin status; mrCRM, MRI-assessed prediction of CRM status; MRI, magnetic resonance imaging; mrTRG,MRI-assessed tumor regression grading; ymrT stage, MRI-assessed T stage post treatment.




A

C

B

D

FE

Fig A1. T2-weighted images showing magnetic resonance imaging tumor regression grading (TRG). Left-hand images are pretreatment images. Right-hand imagesare post treatment. (A) Axial baseline images showing a semiannular tumor as intermediate signal intensity in the posterior rectal wall between the 3 o’clock and 9o’clock positions. (B) Axial post-treatment image shows a fibrotic low signal scar at 6 o’clock, with high signal intensity submucosal edema; absence of any tumor signalindicates a TRG by magnetic resonance imaging (mrTRG) –1. (C) Coronal baseline images show tumor centered at the 9 o’clock position. (D) Coronal post-treatmentimage shows a fibrotic low signal scar at 9 o’clock. A small amount of residual tumor remains inferior to the scar (white arrow); therefore, there is predominant scarsignal intensity post treatment, with minimal residual tumor indicating mrTRG-2. (E) Axial baseline image shows tumor centered at the 12 o’clock position, infiltratingthrough the anterior rectal wall. (F) Axial post-treatment image show mixed areas of low signal fibrosis and intermediate signal intensity, indicating mrTRG-3. (G) Axialbaseline images show an annular infiltrating tumor (black arrows). (H) Axial post-treatment images show persistent intermediate signal intensity annular tumor withminimal low signal fibrosis (white arrow); therefore, tumor signal intensity predominates and is mrTRG-4. (I) Axial baseline images show an annular tumor, withextramural extension at the 3 o’clock position. (J) Axial post-treatment images show little response compared with the baseline scan, mrTRG-5.

Patel et al



G

I

H

J

Fig A1. continued.




Documents

Magnetic Resonance Imaging–Detected Tumor Response for ... · For lower-third rectal tumors, the deﬁnition of predicted mrCRM involve-ment was tumor within 1 mm of the levator