ORIGINAL ARTICLE
The Prognostic Significance of Postchemoradiotherapy High-resolution MRI and Histopathology Detected Extramural Venous Invasion in Rectal Cancer Manish Chand, MRCS,∗ †‡ Jessica Evans, MRCS,∗ †‡ Robert I. Swift, FRCS,† Paris P. Tekkis, FRCS,∗ ‡ Nicholas P. West, FRCPath,§ Gordon Stamp, FRCPath,∗ ‡ Richard J. Heald, FRCS,¶ and Gina Brown, FRCR∗ Objective: This study aimed to determine the prognostic significance of extramural venous invasion (EMVI) after chemoradiotherapy (CRT) by both magnetic resonance imaging (MRI) (ymrEMVI) and histopathology (ypEMVI). Background: EMVI is a prognostic factor in rectal cancer but whether this remains so after CRT preoperative is unknown. Histopathological definitions of EMVI are variable and lead to underreporting particularly after CRT. Methods: All consecutive patients staged on initial MRI as EMVI-positive undergoing preoperative CRT and curative surgery between Jan 2006 and Jan 2012 were included. Posttreatment EMVI status (yEMVI) was reevaluated for both MRI and pathology. The primary endpoint of disease-free survival (DFS) for ymrEMVI and ypEMVI was calculated using the Kaplan-Meier product limit and compared with a Mantel-Cox log-rank test. A P < 0.05 was considered significant. Hazard ratios (HRs) for disease recurrence were generated using Cox proportional hazard regression for MRI and histopathology tumor characteristics. Results: A total of 188 patients who had evidence of EMVI on initial baseline MRI staging were included. MRI detected significantly more patients with persistent EMVI than histopathology (53% vs 19%) but both were prognostic for worse survival—ymrEMVI (HR 1.97) and ypEMVI (HR 2.39). Patients with persistent ymrEMVI-positivity had significantly worse DFS at 3 years (42.7%) compared with ymrEMVI-negative tumors (79.8%); DFS for was 36.9% versus 65.9% positive and negative ypEMVI, respectively. Conclusions: Detection of EMVI post-CRT is prognostically significant whether detected by MRI or histopathology. EMVI status after treatment may be used to counsel patients regarding ongoing risks of metastatic disease, implications for surveillance, and systemic chemotherapy. (Ann Surg 2015;261:473–479)
V
ascular invasion has long been recognized as a prognostic factor in rectal cancer since Brown and Warren’s seminal report in 1938.1 Several other studies have confirmed its importance as an independent marker of poor prognosis in terms of both survival and disease recurrence.2–6 The modern term “extramural venous invasion” (EMVI) more accurately describes tumor cells actively invading the veins beyond the outer limits of the muscularis propria. Because this is tumor spread beyond the bowel wall, it is exclusively found in more locally advanced tumors—T3 and T4.7
Magnetic resonance imaging (MRI) has been shown to be highly accurate in identifying adverse prognostic features preoperatively8 including the visualization of tumor signal expanding extramural vessels, which we define as mrEMVI.9 MRI has also been shown to be accurate in identifying EMVI.7 However, the prognostic importance of EMVI after preoperative chemoradiotherapy (CRT) on either imaging or pathology is not fully known. EMVI status after preoperative treatment is denoted ymrEMVI and ypEMVI for MRI and histopathology, respectively. Traditionally, EMVI has been detected on histopathological analysis of the tumor resection specimens. However, detection rates are variable ranging from as low as 9% to more than 50%1,2,6,10 with the worst results seen in preoperatively treated rectal cancers.11–13 This study aimed to evaluate the prognostic importance of EMVI status after preoperative CRT in both MRI and histopathologydetected disease. We further compared the survival outcomes between ymrEMVI and ypEMVI to determine whether they can be complimentary in prognosis-based treatment decisions.
METHODS Patients Patients were identified from the prospectively maintained Royal Marsden Hospital Rectal Cancer database, which included images or data from the local referring network of 6 hospitals. Data were extracted on consecutive patients undergoing potentially curative resectional surgery for rectal cancer that had undergone long-course CRT for locally advanced rectal cancer with no evidence of metastatic disease between January 2006 and January 2013. Inclusion criteria were all adult patients, who had undergone long-course CRT for locally advanced rectal cancer with no evidence of metastatic disease. Patients must have had evidence of EMVI on baseline MRI scans to be included. All patients must have undergone total mesorectal excision (TME) surgery.14 Exclusion criteria were patients with synchronous tumors, incomplete neoadjuvant treatment, noncurative surgery, and emergency presentation. There was central review of all pathology and radiology by specialized gastrointestinal pathologists and radiologists, respectively. Local ethical approval was obtained from the institution’s Research and Development committee.
Preoperative Staging From the ∗ Royal Marsden Hospital, Fulham Road, London; †Croydon University Hospital, London Road, Croydon; ‡Imperial College, Department of Cancer and Surgery, London; §Pathology & Tumour Biology, Leeds Institute of Cancer and Pathology, University of Leeds, Leeds; and ¶Basingstoke Hospital, Basingstoke, UK. Disclosures: This research was funded by NIHR BRC Royal Marsden and Pelican Cancer Foundation, and the authors declare no conflicts of interest. Reprints: Gina Brown, FRCR, Royal Marsden Hospital, Downs Road, Sutton, London SM2 5PT, UK. E-mail: [email protected]. C 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright ISSN: 0003-4932/14/26103-0473 DOI: 10.1097/SLA.0000000000000848
Annals of Surgery r Volume 261, Number 3, March 2015
All patients were staged by clinical examination and imaging. MRI scans of the pelvis and high-resolution computed tomographic scan of the thorax, abdomen, and pelvis are routinely conducted as part of baseline staging in our institution. All treatment decisions were made as part of a multidisciplinary team meeting. Patients with MRI-detected tumor penetration into the mesorectum of more than 5 mm (T3c), suspected N2 nodal disease, or a potentially threatened mesorectal fascia/distal TME plane (defined as tumor within 1 mm of the mesorectal fascia) were considered for long-course preoperative CRT (45 Gy in fractions with concomitant 5-fluorouracil-based chemotherapy). Patients were fully restaged at 6 weeks after www.annalsofsurgery.com | 473
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Chand et al
completion of neoadjuvant therapy. Surgery was performed within 6 weeks of the last set of complete radiological staging. Patients underwent a minimum of 2 scans (before and after neoadjuvant treatment), which were made available for review. If a patient had undergone more than 1 posttreatment MRI scan, then the most recent MRI to subsequent surgery was used for the purposes of assessment. The MRI scans were reviewed independently by a senior radiologist (G.B.), who was blinded to the pathology reports.
MRI Technique and Image Acquisition The protocol employed a thin, 3-mm section turbo spin-echo T2-weighted technique with a surface pelvic phased array coil and a small field of view (160 mm × 160 mm, 256 × 256 matrix) with a minimum of 4 signal averages to ensure a 0.6 × 0.6 × 3 mm3 high-resolution image (1 mm3 voxel size). For a 1.5-T MRI scanner, 4 sequences were used: sagittal, coronal and at least 2 acquisitions in the oblique axial planes. We ensured that oblique axial scans were always perpendicular to the long axis of the rectal wall.15,16
Histopathology Technique The standard data sets for reporting colorectal cancers as issued by the Royal College of Pathologists (UK) were used for all examinations at our institution.17 All slides had been initially assessed by the local hospital before undergoing further review by our institution, blinded to the results of the pre- or posttreatment MRI.
Data Collection Data were collected on patient demographics, preoperative staging [tumor depth—mrT-stage, nodal disease—mrN-stage, proximity to the circumferential resection margin (mrCRM) and mrEMVI status], and survival outcomes. Final tumor stage was documented from the pathology reports, which included ypT-stage, ypN-stage, ypCRM involvement, and ypEMVI status.
Definitions EMVI—Histopathological (ypEMVI) and MRI Criteria (ymrEMVI) Evidence of EMVI was confirmed on histopathology if adherent tumor was present within an extramural, well-defined tubular or rounded structure accompanying an artery. This would have a diameter of at least 100 μ. There should be at least one of these other criteria: (i) a wall containing smooth muscle; (ii) presence of erythrocytes ± fibrin or thrombus at the advancing edge within the lumen; (iii) no associated nerve fibers; (iv) no infiltration beyond the smoothmuscle/fibrosis of the wall into the surrounding adipose tissue; (v) demonstration of continuity with an endothelial lined vein not containing tumor. Specifically excluded would be detached tumor cells floating “free” within the lumen not adherent to endothelium (which may be artefact). mrEMVI was considered positive if certain morphological features were seen on 3 mm slices—a serpiginous extension of tumor signal within a vascular structure leading to vessel expansion and irregular contouring of the vessel border (vessel was seen as a tubu-
lar structure containing signal void on T2-weighted images shown in continuity on adjacent slices). This corresponded to a score of 3 or 4 on the mrEMVI classification (Table 1).7,9
T, N-Stage, and CRM Involvement T-stage was classified as either “good” or “poor” for purposes of analysis. Tumor stage of T3b or less (5 mm or less of extramural spread) was considered “good,” whereas T-stage of T3c or more was classified as “poor.” Nodal disease was recorded as either positive or negative on MRI assessment but recorded as pN0-2 on pathology. The predicted mrCRM involvement was documented as either positive or negative depending on whether the tumor margin was within 1 mm of the mesorectal fascia or bordering the distal TME plane at or below the level of the puborectalis sling.
Height of Tumor The height of the tumor from the anal verge was grouped into lower (≤5 cm), middle (6–10 cm), and upper thirds (11–15 cm) of the rectum using height measurements from the distal edge of the tumor on the sagittal MRI.
Analysis Methods The primary outcome measure was 3-year disease-free survival (DFS). Secondary endpoints included recurrence rates, final yT-stage, yN-stage, and yCRM status. Statistical analysis of differences between the groups was performed using χ 2 test or Fisher exact test where appropriate. Survival analysis was undertaken between ymrEMVI and ypEMVI including rate of recurrence, time to recurrence, and 3-year DFS, which was obtained using the KaplanMeier product-limit and compared using the Mantel-Cox log-rank test. A P < 0.05 was considered significant. A positive event was histology-proven recurrence (local, distant, or both) or death from any cause. Cox’s proportional hazard models were built to test the impact of confounding variables on survival. This allows the effect of predictive factors on outcome to be assessed, accounting for censored outcome, differing time of follow-up, and the interval between surgery and the adverse event of interest. Hazard ratios (HRs) and 95% confidence intervals (CIs) were generated. To provide clinically and meaningful risk adjustment, several strategies were used to create multivariable models: 1. Fully adjusted model-–All predictive risk factors that were judged to be clinically relevant (irrespective of statistical significance) were entered into a fixed model to adjust the impact of EMVI/N on survival. 2. Parsimonious bidirectional stepwise model based on Akaike information criterion (AIC)—AIC measures the relative quality of a statistical model, trading off goodness of fit for complexity. It thus provides an accurate means for model selection and is not driven by significant P values alone, which can lead to poorly fitted models. A smaller AIC indicates a better model. The step AIC package in R with bidirectional selection was used. 3. Parsimonious forward stepwise model—a final model was built using selection based on P values alone to compare to previous
TABLE 1. MRI Classification for EMVI MRI Score 0 1 2 3 4
Morphology Features on MRI
MRI Status
Pattern of tumour extension through the rectal wall is not nodular; no adjacent vessels Minimal extramural stranding; no adjacent vessels Stranding in proximity of vessels but no tumour signal in vessels which have a normal calibre lumen Intermediate tumor signal in lumen of vessels; vessel expansion by tumor signal Irregular vessel contour with tumor signal in large anatomical vessels such as: superior, middle, or inferior rectal veins
474 | www.annalsofsurgery.com
Negative Negative Negative Positive Positive
C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Extramural Venous Invasion After Chemoradiotherapy
models. Variables significant at P < 0.1 at univariable level were entered, and remained if P < 0.05. Data were analyzed using SPSS 19 (IBM Corp., Armonk, NY) and R 3.0.0 (R Foundation for Statistical Computing).
RESULTS
detected EMVI negative or positive—ypEMVI and ymrEMVI. For each group, the final pT and pN staging was recorded from histopathological analysis of the surgical resection specimen. The results are shown in Table 2. There were more advanced pT-stage and pN-stage cases when comparing pEMVI-negative and -positive patients—P < 0.05. There was no statistical difference in CRM status between the patients with and without pEMVI but the overall prevalence of CRM involvement in the study was low. Comparison of the posttreatment staging characteristics showed similar results between mrEMVI negative and positive; there were more locally advanced pT-stage and pN-stage tumors—P < 0.05.
A database search revealed 188 patients who met the inclusion/exclusion criteria detailed earlier from a total of 580 patients who had been diagnosed and treated for rectal cancer between 2006 and 2009 at our institution. Therefore, the prevalence of EMVI at baseline MRI stage was 32.4%. All patients had completed preoperative long-course CRT before potentially curative surgery. Each patient had undergone a minimum of 2 MRI scans either side of CRT treatment. In the case of serial MRI scans post-CRT, the scan immediately before surgery was used of the purposes of this study. Subsequent curative TME surgery was carried out on all patients. Of the total number of patients, 121 were male and 67 were female. The median age was 74 ± 13 years (range: 30–87 years). All patients were offered adjuvant 5-fluorouracil chemotherapy postoperatively.
There were 99 patients (52.7%) who had evidence of persistent ymrEMVI after neoadjuvant CRT, and 89 patients (47.3%) had become mrEMVI negative after treatment. Histopathology detected 36 cases (19.1%) of ypEMVI after CRT. There was a significant difference in detection rates of EMVI between the 2 techniques.
Baseline Staging
Recurrence Rates
A total of 137 of 188 (73%) patients were staged as mrTpoor (>mrT3c); 123 of 188 (65%) had MR nodal disease; and 81 of 188 (43%) were defined as potentially positive mrCRM. Therefore, majority of the patients included in the analysis had locally advanced tumors. There were 69 low-rectal (37%), 62 mid-rectal (33%), and 57 upper-rectal (30%).
The distant and local recurrence rates are shown in Table 3. When comparing the recurrence rates of EMVI negative and positive patients whether by MRI or pathology, there was a statistically significant difference. Furthermore, the recurrence rates for ypEMVI- and ymrEMVI-negative patients of 23.9% and 25.8%, respectively, were similar as was the case comparing ypEMVI- and ymrEMVI-positive groups 41.7% and 40.4%, respectively. Of the discrepant cases between MRI and histopathology, the subgroup in which patients were ymrEMVI positive but ypEMVI negative, the recurrence rate was 39.4%. Patients who were mrEMVI negative on baseline staging had a recurrence rate of 22.2%. If MRI was overstaging EMVI, then
Posttreatment Staging At the end of treatment, 5.3% of cases were pathologically CRM positive compared with 43% with threatened CRM on MRI before treatment. Patients were categorized into pathology and MRI
Concordance Between MRI and Histopathology EMVI Status
TABLE 2. Final Staging Characteristics After CRT and Surgery Post-CRT Subgroups Staging Characteristic ypT stage
ypN stage ypCRM status ypEMVI status
ypEMVI Negative (n = 142)
ypEMVI Positive (n = 46)
ymrEMVI Negative (n = 89)
ymrEMVI Positive (n = 99)
14 7 31 83 7 101 37 4 136 6 142 0
0 1 2 34 9 17 19 10 42 4 0 46
11 7 20 45 6 70 17 2 87 2 79 10
3 1 13 72 10 48 39 12 91 8 63 36
PCR (T0N0) T1 (n = 8) T2 (n = 33) T3 (n = 117) T4 (n = 16) 0 (n = 118) 1 (n = 56) 2 (n = 14) Negative (n = 178) Positive (n = 10) Negative (n = 142) Positive (n = 46)
TABLE 3. Recurrence Rates of EMVI Negative and Positive Tumors After CRT ypEMVI Negative (n = 142) ymrEMVI negative (n = 89) ymrEMVI positive (n = 99) Odds ratio for local or distant recurrence and ypEMVI status Odds ratio for local or distant recurrence and mrEMVI status
76 66
C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Local 2 1 2.38 2.99
Distant
Total
11 23
13 24
ypEMVI Positive (n = 46)
Local
Distant
13 1 3 33 4 13 (95% CI: 1.20–4.74), P < 0.01
Total 4 17
(95% CI: 1.55–5.78), P < 0.01
www.annalsofsurgery.com | 475
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Chand et al
the observed worsening of recurrence rates of mrEMVI-positive but ypEMVI-negative patients would not be observed.
Survival Analysis The DFS for ymrEMVI-positive patients was significantly worse than ymrEMVI-negative patients—42.7% (95% CI: 16.8%– 68.6%) versus 79.2% (95% CI: 70.0%–88.4%); Mantel Cox log-rank test, P < 0.05 (Fig. 1). For those patients with ypEMVI, there was also worsened DFS of 36.9% (95% CI: 15.7%–48.1%) compared with 65.9% (95% CI: 56.1%–75.7%) for ypEMVI negative (Fig. 2); Mantel Cox log-rank test, P < 0.05. A comparison of MRI-detected or
pathology-detected EMVI showed no significant difference between the groups. This was also shown to be the case comparing ymrEMVIand ypEMVI-negative patients. In the subgroup of discrepant cases where patients were ymrEMVI positive but ypEMVI negative DFS was 46.7% (95% CI: 33.5%–54.5%), indicating that MRI is unlikely to have “overstaged” patients with respect to EMVI status. Factors found to be significant on univariate analysis were tested further in multivariate regression (Table 4). Using the Cox proportional hazard regression model, both mrEMVI positive and pEMVI positive were independently significant for predicting disease recurrence with HRs of 1.97 (95% CI: 1.01–3.90) and 2.39 (95% CI: 1.11–5.14), respectively.
DISCUSSION
FIGURE 1. Comparison of survival outcome of 3-year DFS between ymrEMVI negative and ymrEMVI positive patients.
FIGURE 2. Comparison of 3-year DFS between ypEMVI negative and ypEMVI positive. 476 | www.annalsofsurgery.com
The main finding of this study was that DFS was significantly reduced for patients who had either MRI or histopathological evidence of EMVI after preoperative CRT, and this was independent of yT and yN stage. DFS for ymrEMVI-positive patients was significantly worse than for ymrEMVI-negative patients—42.7% versus 79.2%. Equally, patients with ypEMVI had a DFS of 36.9% compared with 65.9% for ypEMVI negative. The difference in survival between patients who have converted from EMVI positive to negative on MRI or histopathology versus those who showed persistent EMVI implies that this feature is of predictive as well as prognostic relevance. This is the first study to examine the independent clinical relevance of post-CRT yEMVI status. Baseline EMVI status has been previously shown to be a prognostic factor.7 In a previous retrospective series of 142 patients, the prevalence of MRI-detected EMVI (mrEMVI) on initial staging was approximately 35%, which predicted survival outcomes. In this study, mrEMVI status was not shown to be an independent variable, only the initial T-stage was significant. This study differs as we have only evaluated locally advanced tumors undergoing preoperative CRT; these were almost exclusively T3 and above—hence EMVI status was independent of T stage in our series. The overall prevalence of EMVI in this study at baseline was similar; however, after CRT ymrEMVI was shown to be an independent variable for disease recurrence. The importance of possible post-CRT prognostic factors and their role in disease recurrence are still not well understood and there is still no current randomized trial evidence confirming the benefit of adjuvant chemotherapy after preoperative therapy.18 We have shown that disease relapse rates for persistent mrEMVI at 3 years are 40.4% with HR 1.97 compared with mrEMVI-negative patients where relapse rates are 25.8%. This information may be of relevance when counseling patients regarding the ongoing risks and potential gains of adjuvant chemotherapy after CRT in the absence of randomized trials that have specifically addressed this question. This study suggests that patients could be risk-stratified before and after CRT on the basis of their mrEMVI status and advised accordingly. Examining the agreement between ymrEMVI and ypEMVI reveals a striking difference in detection. Of the initial 188 mrEMVIpositive patients, 99 remained EMVI positive on MRI, only 36 of whom (36.4%) were detected on histopathology using standard methods, yet the prognostic outcomes for mrEMVI-positive tumors that were pathologically ypEMVI negative was as poor and significantly worse than patients with either ymrEMVI- or ypEMVI-negative status. Furthermore, both the survival outcomes and recurrence rates of patients with no evidence of mrEMVI on baseline staging before CRT are similar of those patients who have become EMVI negative after CRT. It is likely therefore that, rather than MRI overdiagnosing EMVI at the end of treatment, these cases are hard to detect by conventional pathology. MRI has recently been shown to be able to grade the degree of change in EMVI after CRT and validated against survival outcomes.19 C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Extramural Venous Invasion After Chemoradiotherapy
TABLE 4. Univariate and Multivariate Analysis (Cox Proportional Hazards for DFS) by Clinical, Preoperative MRI, and Postoperative Histopathology Characteristics Variables Patient characteristics
Sex Height
Baseline MR staging
mrT stage mrN stage mrEMVI mrCRM
Post-CRT preoperative MR staging
ymrT stage ymrN stage ymrEMVI ymrCRM
Final pathology staging
ypT ypN ypEMVI ypCRM
Univariate Analysis
Group
Patient Numbers
HR
Female Male Upper/mid Low Good Poor Negative Positive Negative Positive Negative Positive Good Poor Negative Positive Negative Positive Clear Involved/threatened Good Poor Negative Positive Negative Positive Negative Positive
67 121 119 69 51 137 65 123 0 188 107 81 116 72 104 84 89 99 148 40 64 124 118 70 142 46 178 10
Ref 1.093 Ref 1.369 Ref 1.187 Ref 1.196 Ref 0.902 Ref 0.846 Ref 1.218 Ref 1.179 Ref 1.987 Ref 1.26 Ref 1.125 Ref 2.912 Ref 3.889 Ref 3.352
CRT is known to lead to fibrosis and this can distort the normal microarchitecture, which guides pathologists in identifying venous invasion leading to a known false-negative rate during histopathology analysis.20,21 This can be seen in Figure 3. Whereas a post-
95% CI
Multivariate Analysis P
0.625–1.912
0.756
0.815–2.298
0.235
0.638–2.206
0.588
0.691–2.071
0.523
0.527–1.544
0.706
0.497–1.441
0.539
0.723–2.052
0.459
0.701–1.982
0.534
1.237–4.323
0.004
0.674–2.354
0.469
0.695–1.279
0.534
1.724–4.878
The Prognostic Significance of Postchemoradiotherapy High-resolution MRI and Histopathology Detected Extramural Venous Invasion in Rectal Cancer Manish Chand, MRCS,∗ †‡ Jessica Evans, MRCS,∗ †‡ Robert I. Swift, FRCS,† Paris P. Tekkis, FRCS,∗ ‡ Nicholas P. West, FRCPath,§ Gordon Stamp, FRCPath,∗ ‡ Richard J. Heald, FRCS,¶ and Gina Brown, FRCR∗ Objective: This study aimed to determine the prognostic significance of extramural venous invasion (EMVI) after chemoradiotherapy (CRT) by both magnetic resonance imaging (MRI) (ymrEMVI) and histopathology (ypEMVI). Background: EMVI is a prognostic factor in rectal cancer but whether this remains so after CRT preoperative is unknown. Histopathological definitions of EMVI are variable and lead to underreporting particularly after CRT. Methods: All consecutive patients staged on initial MRI as EMVI-positive undergoing preoperative CRT and curative surgery between Jan 2006 and Jan 2012 were included. Posttreatment EMVI status (yEMVI) was reevaluated for both MRI and pathology. The primary endpoint of disease-free survival (DFS) for ymrEMVI and ypEMVI was calculated using the Kaplan-Meier product limit and compared with a Mantel-Cox log-rank test. A P < 0.05 was considered significant. Hazard ratios (HRs) for disease recurrence were generated using Cox proportional hazard regression for MRI and histopathology tumor characteristics. Results: A total of 188 patients who had evidence of EMVI on initial baseline MRI staging were included. MRI detected significantly more patients with persistent EMVI than histopathology (53% vs 19%) but both were prognostic for worse survival—ymrEMVI (HR 1.97) and ypEMVI (HR 2.39). Patients with persistent ymrEMVI-positivity had significantly worse DFS at 3 years (42.7%) compared with ymrEMVI-negative tumors (79.8%); DFS for was 36.9% versus 65.9% positive and negative ypEMVI, respectively. Conclusions: Detection of EMVI post-CRT is prognostically significant whether detected by MRI or histopathology. EMVI status after treatment may be used to counsel patients regarding ongoing risks of metastatic disease, implications for surveillance, and systemic chemotherapy. (Ann Surg 2015;261:473–479)
V
ascular invasion has long been recognized as a prognostic factor in rectal cancer since Brown and Warren’s seminal report in 1938.1 Several other studies have confirmed its importance as an independent marker of poor prognosis in terms of both survival and disease recurrence.2–6 The modern term “extramural venous invasion” (EMVI) more accurately describes tumor cells actively invading the veins beyond the outer limits of the muscularis propria. Because this is tumor spread beyond the bowel wall, it is exclusively found in more locally advanced tumors—T3 and T4.7
Magnetic resonance imaging (MRI) has been shown to be highly accurate in identifying adverse prognostic features preoperatively8 including the visualization of tumor signal expanding extramural vessels, which we define as mrEMVI.9 MRI has also been shown to be accurate in identifying EMVI.7 However, the prognostic importance of EMVI after preoperative chemoradiotherapy (CRT) on either imaging or pathology is not fully known. EMVI status after preoperative treatment is denoted ymrEMVI and ypEMVI for MRI and histopathology, respectively. Traditionally, EMVI has been detected on histopathological analysis of the tumor resection specimens. However, detection rates are variable ranging from as low as 9% to more than 50%1,2,6,10 with the worst results seen in preoperatively treated rectal cancers.11–13 This study aimed to evaluate the prognostic importance of EMVI status after preoperative CRT in both MRI and histopathologydetected disease. We further compared the survival outcomes between ymrEMVI and ypEMVI to determine whether they can be complimentary in prognosis-based treatment decisions.
METHODS Patients Patients were identified from the prospectively maintained Royal Marsden Hospital Rectal Cancer database, which included images or data from the local referring network of 6 hospitals. Data were extracted on consecutive patients undergoing potentially curative resectional surgery for rectal cancer that had undergone long-course CRT for locally advanced rectal cancer with no evidence of metastatic disease between January 2006 and January 2013. Inclusion criteria were all adult patients, who had undergone long-course CRT for locally advanced rectal cancer with no evidence of metastatic disease. Patients must have had evidence of EMVI on baseline MRI scans to be included. All patients must have undergone total mesorectal excision (TME) surgery.14 Exclusion criteria were patients with synchronous tumors, incomplete neoadjuvant treatment, noncurative surgery, and emergency presentation. There was central review of all pathology and radiology by specialized gastrointestinal pathologists and radiologists, respectively. Local ethical approval was obtained from the institution’s Research and Development committee.
Preoperative Staging From the ∗ Royal Marsden Hospital, Fulham Road, London; †Croydon University Hospital, London Road, Croydon; ‡Imperial College, Department of Cancer and Surgery, London; §Pathology & Tumour Biology, Leeds Institute of Cancer and Pathology, University of Leeds, Leeds; and ¶Basingstoke Hospital, Basingstoke, UK. Disclosures: This research was funded by NIHR BRC Royal Marsden and Pelican Cancer Foundation, and the authors declare no conflicts of interest. Reprints: Gina Brown, FRCR, Royal Marsden Hospital, Downs Road, Sutton, London SM2 5PT, UK. E-mail: [email protected]. C 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright ISSN: 0003-4932/14/26103-0473 DOI: 10.1097/SLA.0000000000000848
Annals of Surgery r Volume 261, Number 3, March 2015
All patients were staged by clinical examination and imaging. MRI scans of the pelvis and high-resolution computed tomographic scan of the thorax, abdomen, and pelvis are routinely conducted as part of baseline staging in our institution. All treatment decisions were made as part of a multidisciplinary team meeting. Patients with MRI-detected tumor penetration into the mesorectum of more than 5 mm (T3c), suspected N2 nodal disease, or a potentially threatened mesorectal fascia/distal TME plane (defined as tumor within 1 mm of the mesorectal fascia) were considered for long-course preoperative CRT (45 Gy in fractions with concomitant 5-fluorouracil-based chemotherapy). Patients were fully restaged at 6 weeks after www.annalsofsurgery.com | 473
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Chand et al
completion of neoadjuvant therapy. Surgery was performed within 6 weeks of the last set of complete radiological staging. Patients underwent a minimum of 2 scans (before and after neoadjuvant treatment), which were made available for review. If a patient had undergone more than 1 posttreatment MRI scan, then the most recent MRI to subsequent surgery was used for the purposes of assessment. The MRI scans were reviewed independently by a senior radiologist (G.B.), who was blinded to the pathology reports.
MRI Technique and Image Acquisition The protocol employed a thin, 3-mm section turbo spin-echo T2-weighted technique with a surface pelvic phased array coil and a small field of view (160 mm × 160 mm, 256 × 256 matrix) with a minimum of 4 signal averages to ensure a 0.6 × 0.6 × 3 mm3 high-resolution image (1 mm3 voxel size). For a 1.5-T MRI scanner, 4 sequences were used: sagittal, coronal and at least 2 acquisitions in the oblique axial planes. We ensured that oblique axial scans were always perpendicular to the long axis of the rectal wall.15,16
Histopathology Technique The standard data sets for reporting colorectal cancers as issued by the Royal College of Pathologists (UK) were used for all examinations at our institution.17 All slides had been initially assessed by the local hospital before undergoing further review by our institution, blinded to the results of the pre- or posttreatment MRI.
Data Collection Data were collected on patient demographics, preoperative staging [tumor depth—mrT-stage, nodal disease—mrN-stage, proximity to the circumferential resection margin (mrCRM) and mrEMVI status], and survival outcomes. Final tumor stage was documented from the pathology reports, which included ypT-stage, ypN-stage, ypCRM involvement, and ypEMVI status.
Definitions EMVI—Histopathological (ypEMVI) and MRI Criteria (ymrEMVI) Evidence of EMVI was confirmed on histopathology if adherent tumor was present within an extramural, well-defined tubular or rounded structure accompanying an artery. This would have a diameter of at least 100 μ. There should be at least one of these other criteria: (i) a wall containing smooth muscle; (ii) presence of erythrocytes ± fibrin or thrombus at the advancing edge within the lumen; (iii) no associated nerve fibers; (iv) no infiltration beyond the smoothmuscle/fibrosis of the wall into the surrounding adipose tissue; (v) demonstration of continuity with an endothelial lined vein not containing tumor. Specifically excluded would be detached tumor cells floating “free” within the lumen not adherent to endothelium (which may be artefact). mrEMVI was considered positive if certain morphological features were seen on 3 mm slices—a serpiginous extension of tumor signal within a vascular structure leading to vessel expansion and irregular contouring of the vessel border (vessel was seen as a tubu-
lar structure containing signal void on T2-weighted images shown in continuity on adjacent slices). This corresponded to a score of 3 or 4 on the mrEMVI classification (Table 1).7,9
T, N-Stage, and CRM Involvement T-stage was classified as either “good” or “poor” for purposes of analysis. Tumor stage of T3b or less (5 mm or less of extramural spread) was considered “good,” whereas T-stage of T3c or more was classified as “poor.” Nodal disease was recorded as either positive or negative on MRI assessment but recorded as pN0-2 on pathology. The predicted mrCRM involvement was documented as either positive or negative depending on whether the tumor margin was within 1 mm of the mesorectal fascia or bordering the distal TME plane at or below the level of the puborectalis sling.
Height of Tumor The height of the tumor from the anal verge was grouped into lower (≤5 cm), middle (6–10 cm), and upper thirds (11–15 cm) of the rectum using height measurements from the distal edge of the tumor on the sagittal MRI.
Analysis Methods The primary outcome measure was 3-year disease-free survival (DFS). Secondary endpoints included recurrence rates, final yT-stage, yN-stage, and yCRM status. Statistical analysis of differences between the groups was performed using χ 2 test or Fisher exact test where appropriate. Survival analysis was undertaken between ymrEMVI and ypEMVI including rate of recurrence, time to recurrence, and 3-year DFS, which was obtained using the KaplanMeier product-limit and compared using the Mantel-Cox log-rank test. A P < 0.05 was considered significant. A positive event was histology-proven recurrence (local, distant, or both) or death from any cause. Cox’s proportional hazard models were built to test the impact of confounding variables on survival. This allows the effect of predictive factors on outcome to be assessed, accounting for censored outcome, differing time of follow-up, and the interval between surgery and the adverse event of interest. Hazard ratios (HRs) and 95% confidence intervals (CIs) were generated. To provide clinically and meaningful risk adjustment, several strategies were used to create multivariable models: 1. Fully adjusted model-–All predictive risk factors that were judged to be clinically relevant (irrespective of statistical significance) were entered into a fixed model to adjust the impact of EMVI/N on survival. 2. Parsimonious bidirectional stepwise model based on Akaike information criterion (AIC)—AIC measures the relative quality of a statistical model, trading off goodness of fit for complexity. It thus provides an accurate means for model selection and is not driven by significant P values alone, which can lead to poorly fitted models. A smaller AIC indicates a better model. The step AIC package in R with bidirectional selection was used. 3. Parsimonious forward stepwise model—a final model was built using selection based on P values alone to compare to previous
TABLE 1. MRI Classification for EMVI MRI Score 0 1 2 3 4
Morphology Features on MRI
MRI Status
Pattern of tumour extension through the rectal wall is not nodular; no adjacent vessels Minimal extramural stranding; no adjacent vessels Stranding in proximity of vessels but no tumour signal in vessels which have a normal calibre lumen Intermediate tumor signal in lumen of vessels; vessel expansion by tumor signal Irregular vessel contour with tumor signal in large anatomical vessels such as: superior, middle, or inferior rectal veins
474 | www.annalsofsurgery.com
Negative Negative Negative Positive Positive
C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Extramural Venous Invasion After Chemoradiotherapy
models. Variables significant at P < 0.1 at univariable level were entered, and remained if P < 0.05. Data were analyzed using SPSS 19 (IBM Corp., Armonk, NY) and R 3.0.0 (R Foundation for Statistical Computing).
RESULTS
detected EMVI negative or positive—ypEMVI and ymrEMVI. For each group, the final pT and pN staging was recorded from histopathological analysis of the surgical resection specimen. The results are shown in Table 2. There were more advanced pT-stage and pN-stage cases when comparing pEMVI-negative and -positive patients—P < 0.05. There was no statistical difference in CRM status between the patients with and without pEMVI but the overall prevalence of CRM involvement in the study was low. Comparison of the posttreatment staging characteristics showed similar results between mrEMVI negative and positive; there were more locally advanced pT-stage and pN-stage tumors—P < 0.05.
A database search revealed 188 patients who met the inclusion/exclusion criteria detailed earlier from a total of 580 patients who had been diagnosed and treated for rectal cancer between 2006 and 2009 at our institution. Therefore, the prevalence of EMVI at baseline MRI stage was 32.4%. All patients had completed preoperative long-course CRT before potentially curative surgery. Each patient had undergone a minimum of 2 MRI scans either side of CRT treatment. In the case of serial MRI scans post-CRT, the scan immediately before surgery was used of the purposes of this study. Subsequent curative TME surgery was carried out on all patients. Of the total number of patients, 121 were male and 67 were female. The median age was 74 ± 13 years (range: 30–87 years). All patients were offered adjuvant 5-fluorouracil chemotherapy postoperatively.
There were 99 patients (52.7%) who had evidence of persistent ymrEMVI after neoadjuvant CRT, and 89 patients (47.3%) had become mrEMVI negative after treatment. Histopathology detected 36 cases (19.1%) of ypEMVI after CRT. There was a significant difference in detection rates of EMVI between the 2 techniques.
Baseline Staging
Recurrence Rates
A total of 137 of 188 (73%) patients were staged as mrTpoor (>mrT3c); 123 of 188 (65%) had MR nodal disease; and 81 of 188 (43%) were defined as potentially positive mrCRM. Therefore, majority of the patients included in the analysis had locally advanced tumors. There were 69 low-rectal (37%), 62 mid-rectal (33%), and 57 upper-rectal (30%).
The distant and local recurrence rates are shown in Table 3. When comparing the recurrence rates of EMVI negative and positive patients whether by MRI or pathology, there was a statistically significant difference. Furthermore, the recurrence rates for ypEMVI- and ymrEMVI-negative patients of 23.9% and 25.8%, respectively, were similar as was the case comparing ypEMVI- and ymrEMVI-positive groups 41.7% and 40.4%, respectively. Of the discrepant cases between MRI and histopathology, the subgroup in which patients were ymrEMVI positive but ypEMVI negative, the recurrence rate was 39.4%. Patients who were mrEMVI negative on baseline staging had a recurrence rate of 22.2%. If MRI was overstaging EMVI, then
Posttreatment Staging At the end of treatment, 5.3% of cases were pathologically CRM positive compared with 43% with threatened CRM on MRI before treatment. Patients were categorized into pathology and MRI
Concordance Between MRI and Histopathology EMVI Status
TABLE 2. Final Staging Characteristics After CRT and Surgery Post-CRT Subgroups Staging Characteristic ypT stage
ypN stage ypCRM status ypEMVI status
ypEMVI Negative (n = 142)
ypEMVI Positive (n = 46)
ymrEMVI Negative (n = 89)
ymrEMVI Positive (n = 99)
14 7 31 83 7 101 37 4 136 6 142 0
0 1 2 34 9 17 19 10 42 4 0 46
11 7 20 45 6 70 17 2 87 2 79 10
3 1 13 72 10 48 39 12 91 8 63 36
PCR (T0N0) T1 (n = 8) T2 (n = 33) T3 (n = 117) T4 (n = 16) 0 (n = 118) 1 (n = 56) 2 (n = 14) Negative (n = 178) Positive (n = 10) Negative (n = 142) Positive (n = 46)
TABLE 3. Recurrence Rates of EMVI Negative and Positive Tumors After CRT ypEMVI Negative (n = 142) ymrEMVI negative (n = 89) ymrEMVI positive (n = 99) Odds ratio for local or distant recurrence and ypEMVI status Odds ratio for local or distant recurrence and mrEMVI status
76 66
C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Local 2 1 2.38 2.99
Distant
Total
11 23
13 24
ypEMVI Positive (n = 46)
Local
Distant
13 1 3 33 4 13 (95% CI: 1.20–4.74), P < 0.01
Total 4 17
(95% CI: 1.55–5.78), P < 0.01
www.annalsofsurgery.com | 475
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Chand et al
the observed worsening of recurrence rates of mrEMVI-positive but ypEMVI-negative patients would not be observed.
Survival Analysis The DFS for ymrEMVI-positive patients was significantly worse than ymrEMVI-negative patients—42.7% (95% CI: 16.8%– 68.6%) versus 79.2% (95% CI: 70.0%–88.4%); Mantel Cox log-rank test, P < 0.05 (Fig. 1). For those patients with ypEMVI, there was also worsened DFS of 36.9% (95% CI: 15.7%–48.1%) compared with 65.9% (95% CI: 56.1%–75.7%) for ypEMVI negative (Fig. 2); Mantel Cox log-rank test, P < 0.05. A comparison of MRI-detected or
pathology-detected EMVI showed no significant difference between the groups. This was also shown to be the case comparing ymrEMVIand ypEMVI-negative patients. In the subgroup of discrepant cases where patients were ymrEMVI positive but ypEMVI negative DFS was 46.7% (95% CI: 33.5%–54.5%), indicating that MRI is unlikely to have “overstaged” patients with respect to EMVI status. Factors found to be significant on univariate analysis were tested further in multivariate regression (Table 4). Using the Cox proportional hazard regression model, both mrEMVI positive and pEMVI positive were independently significant for predicting disease recurrence with HRs of 1.97 (95% CI: 1.01–3.90) and 2.39 (95% CI: 1.11–5.14), respectively.
DISCUSSION
FIGURE 1. Comparison of survival outcome of 3-year DFS between ymrEMVI negative and ymrEMVI positive patients.
FIGURE 2. Comparison of 3-year DFS between ypEMVI negative and ypEMVI positive. 476 | www.annalsofsurgery.com
The main finding of this study was that DFS was significantly reduced for patients who had either MRI or histopathological evidence of EMVI after preoperative CRT, and this was independent of yT and yN stage. DFS for ymrEMVI-positive patients was significantly worse than for ymrEMVI-negative patients—42.7% versus 79.2%. Equally, patients with ypEMVI had a DFS of 36.9% compared with 65.9% for ypEMVI negative. The difference in survival between patients who have converted from EMVI positive to negative on MRI or histopathology versus those who showed persistent EMVI implies that this feature is of predictive as well as prognostic relevance. This is the first study to examine the independent clinical relevance of post-CRT yEMVI status. Baseline EMVI status has been previously shown to be a prognostic factor.7 In a previous retrospective series of 142 patients, the prevalence of MRI-detected EMVI (mrEMVI) on initial staging was approximately 35%, which predicted survival outcomes. In this study, mrEMVI status was not shown to be an independent variable, only the initial T-stage was significant. This study differs as we have only evaluated locally advanced tumors undergoing preoperative CRT; these were almost exclusively T3 and above—hence EMVI status was independent of T stage in our series. The overall prevalence of EMVI in this study at baseline was similar; however, after CRT ymrEMVI was shown to be an independent variable for disease recurrence. The importance of possible post-CRT prognostic factors and their role in disease recurrence are still not well understood and there is still no current randomized trial evidence confirming the benefit of adjuvant chemotherapy after preoperative therapy.18 We have shown that disease relapse rates for persistent mrEMVI at 3 years are 40.4% with HR 1.97 compared with mrEMVI-negative patients where relapse rates are 25.8%. This information may be of relevance when counseling patients regarding the ongoing risks and potential gains of adjuvant chemotherapy after CRT in the absence of randomized trials that have specifically addressed this question. This study suggests that patients could be risk-stratified before and after CRT on the basis of their mrEMVI status and advised accordingly. Examining the agreement between ymrEMVI and ypEMVI reveals a striking difference in detection. Of the initial 188 mrEMVIpositive patients, 99 remained EMVI positive on MRI, only 36 of whom (36.4%) were detected on histopathology using standard methods, yet the prognostic outcomes for mrEMVI-positive tumors that were pathologically ypEMVI negative was as poor and significantly worse than patients with either ymrEMVI- or ypEMVI-negative status. Furthermore, both the survival outcomes and recurrence rates of patients with no evidence of mrEMVI on baseline staging before CRT are similar of those patients who have become EMVI negative after CRT. It is likely therefore that, rather than MRI overdiagnosing EMVI at the end of treatment, these cases are hard to detect by conventional pathology. MRI has recently been shown to be able to grade the degree of change in EMVI after CRT and validated against survival outcomes.19 C 2014 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2014 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Annals of Surgery r Volume 261, Number 3, March 2015
Extramural Venous Invasion After Chemoradiotherapy
TABLE 4. Univariate and Multivariate Analysis (Cox Proportional Hazards for DFS) by Clinical, Preoperative MRI, and Postoperative Histopathology Characteristics Variables Patient characteristics
Sex Height
Baseline MR staging
mrT stage mrN stage mrEMVI mrCRM
Post-CRT preoperative MR staging
ymrT stage ymrN stage ymrEMVI ymrCRM
Final pathology staging
ypT ypN ypEMVI ypCRM
Univariate Analysis
Group
Patient Numbers
HR
Female Male Upper/mid Low Good Poor Negative Positive Negative Positive Negative Positive Good Poor Negative Positive Negative Positive Clear Involved/threatened Good Poor Negative Positive Negative Positive Negative Positive
67 121 119 69 51 137 65 123 0 188 107 81 116 72 104 84 89 99 148 40 64 124 118 70 142 46 178 10
Ref 1.093 Ref 1.369 Ref 1.187 Ref 1.196 Ref 0.902 Ref 0.846 Ref 1.218 Ref 1.179 Ref 1.987 Ref 1.26 Ref 1.125 Ref 2.912 Ref 3.889 Ref 3.352
CRT is known to lead to fibrosis and this can distort the normal microarchitecture, which guides pathologists in identifying venous invasion leading to a known false-negative rate during histopathology analysis.20,21 This can be seen in Figure 3. Whereas a post-
95% CI
Multivariate Analysis P
0.625–1.912
0.756
0.815–2.298
0.235
0.638–2.206
0.588
0.691–2.071
0.523
0.527–1.544
0.706
0.497–1.441
0.539
0.723–2.052
0.459
0.701–1.982
0.534
1.237–4.323
0.004
0.674–2.354
0.469
0.695–1.279
0.534
1.724–4.878
