Members of the Abdominoperineal Excision Study Group are co-authors of this study and can be found under the heading Collaborators.
Multicentre study of circumferential margin positivity and outcomes following abdominoperineal excision for rectal cancer
Article first published online: 12 NOV 2012
Copyright © 2012 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.
British Journal of Surgery
Volume 100, Issue 1, pages 160–166, January 2013
How to Cite
Kennelly, R. P., Rogers, A. C., Winter, D. C. and on behalf of the Abdominoperineal Excision Study Group (2013), Multicentre study of circumferential margin positivity and outcomes following abdominoperineal excision for rectal cancer. Br J Surg, 100: 160–166. doi: 10.1002/bjs.9001
- Issue published online: 5 DEC 2012
- Article first published online: 12 NOV 2012
- Manuscript Accepted: 3 OCT 2012
Rectal cancer outcomes following abdominoperineal excision (APE) have been inferior to those for anterior resection, including more positive circumferential resection margins (CRMs). An erroneously conservative interpretation of APE (rather than a radical resection termed ‘extralevator’) has been proposed as the cause. In this multicentre study, factors contributing to CRM positivity were examined following APE according to its original description.
Data were collected from five hospital databases up to June 2011 including small- and larger-volume units (3 hospitals had 5 or fewer and 2 hospitals had more than 5 APE procedures per year). Primary outcome measures were CRM status; secondary outcomes were local recurrence and death.
Of 327 patients, 302 patients had complete data for analysis. Some 50·0 per cent of patients had neoadjuvant chemoradiotherapy. Histopathological examination showed that 62·9 per cent had tumour category T3 or T4 cancers, 42·1 per cent had node-positive disease and the CRM positivity rate was 13·9 per cent. Multivariable analysis showed only pathological tumour category pT4 (odds ratio 19·92, 95 per cent confidence interval 6·48 to 68·61) and node positivity (odds ratio 3·04, 1·32 to 8·05) to be risk factors for a positive circumferential margin. CRM positivity was a risk factor for local recurrence (P = 0·022) and decreased overall survival (P = 0·001). Hospital volume had no impact on the likelihood of CRM positivity (P = 0·435).
In patients undergoing APE by appropriately trained surgeons using a standardized approach, margin positivity was dictated by tumour stage, but not by centre or surgeon. Copyright © 2012 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.
The low rectal cancer margin positivity rates achieved with anterior resection1 have not been replicated for abdominoperineal excision (APE)2, 3. A positive circumferential resection margin (CRM) infers a local recurrence risk and poor prognosis4. One explanation for CRM positivity relates to the thinning of the mesorectum (and therefore margin) distally such that a low tumour may abut the levator complex. This is particularly relevant with anterior tumours where the mesorectum is thin and can be virtually non-existent, especially in female patients5, 6.
In the original description of APE, Miles7 described excision of the levator ani with the rectum en bloc to reduce radial margin positivity: ‘After reflecting the skin on either side to the requisite extent … the interval between the levatores ani (is) defined. These muscles should be divided as far outwards as their origin from “the white line” so as to include the lateral zone of spread’. A wide excision of the distal rectum and anus using a perineal approach in either lithotomy or prone position completes the dissection8. Total mesorectal excision (TME) principles may have influenced surgeons to migrate the surgical plane excessively toward the rectum in modern APE, and may be responsible for the CRM positivity rates observed9. Relearning the operation described by Miles (now described as an ‘extralevator’ approach and proposed to be even more radical) may resolve the issue, but there are other factors to be considered. Low tumours may be more aggressive10, and case volume and the use of neoadjuvant chemoradiotherapy alter patient outcomes11–13. Many of the reported variations in the outcome of rectal cancer treatment may relate to these factors.
The present multi-institutional study assessed patient outcome following APE in a relatively homogeneous population with surgeons who were similarly trained in rectal cancer surgery and who practised standard APE.
A request for database information was sent to centres that treat rectal cancer with a catchment population of approximately 1·5 million people. Responses were received from three multiconsultant units and two single-handed units, which were included. All study participants were subspecialty trained colorectal surgeons.
Patients who had APE for recurrent cancer, squamous cell cancer of the anus or melanoma were excluded, whereas all patients who underwent APE for primary adenocarcinoma of the rectum were included. Neoadjuvant therapy was given as a combination of radiation and chemotherapy: 45–55 Gy with bolus or infusional 5-fluorouracil as per local protocol.
At operation, the TME plane was dissected to the prostate (or below the cervix). The perineal approach was used for wide excision of the anus and distal rectum. If there was evidence of macroscopic tumour spread outside the plane of dissection, a wider excision was performed locally (if possible).
Data regarding patient demographics, disease characteristics and treatment were collected from the databases and, where necessary, patient notes and pathology reports were accessed separately. A tumour resection margin was considered positive if cancer cells were found at or within 1 mm of the radial resection margin. The primary outcome of the study was the status of the CRM, and secondary outcomes were local recurrence and death.
Patient follow-up included surveillance by endoscopy and computed tomography in accordance with individual centre protocols, generally on a yearly basis for at least 5 years. Overall survival was defined as the time from the date of surgery to the date of death or the date of last follow-up of patients who were still alive. Time to local recurrence was defined as the time from the date of surgery to the date of any recurrent disease in the pelvis.
Results were analysed using Predictive Analytics SoftWare (PASW version 18·0·2; IBM, Armonk, New York, USA). Comparative analyses of quantitative data were performed using the χ2 test for categorical variables, Student's t test for continuous variables and ANOVA for comparison of more than two means. All tests of significance were two-tailed, with P < 0·050 indicating statistical significance. Univariable logistic regression was used to assess the effects of demographics, co-morbidities and procedural information on CRM status, local recurrence, distant metastasis and mortality rates, expressed as odds ratios (ORs) with 95 per cent confidence intervals (c.i.). Variables with P < 0·200 in univariable analysis were included in multivariable logistic regression analysis. Results were considered significant where an OR did not equal 1·00, the 95 per cent c.i. did not cross 1·00 and the P value was less than 0·050. Kaplan–Meier curves were determined for analysis of survival based on CRM status using the log rank test.
A total of 327 patients were identified. After exclusions (12 patients lost to follow-up, 6 squamous cell carcinoma, 2 melanoma, 5 recurrent cancer), 302 patients were included in the analysis. Demographic data are shown in Table 1. Some 50·0 per cent had long-course neoadjuvant chemoradiotherapy, predominantly those treated in the last 10 years of the study (9·2 per cent in 1990–1999 versus 71·4 per cent in 2000–2011). Some 62·9 per cent of patients had tumour category T3 or T4 disease, and 42·1 per cent had node-positive disease. Distant metastases were present 4·3 per cent of patients at the time of surgical resection (Table 1). Median cancer-specific follow-up (excluding postoperative deaths) was 32 (range 2–192) months.
|No. of patients*|
|Age (years)†||65 (23–93)|
|1 (1990–2011)||174 (57·6)|
|2 (1998–2006)||56 (18·5)|
|3 (2001–2010)||29 (9·6)|
|4 (1996–2006)||28 (9·3)|
|5 (2005–2010)||15 (5·0)|
|Neoadjuvant chemoradiotherapy||151 (50·0)|
|Pathological tumour category|
|Node positivity||127 (42·1)|
|CRM positivity||42 (13·9)|
|Local recurrence||18 (6·0)|
|Distant recurrence||35 (11·6)|
|Cancer-related death||61 (20·2)|
|Disease-free interval (months)†||21 (4–140)|
Neoadjuvant therapy, margin and node positivity, and local recurrence
Forty-two patients had positive resection margins, of whom 24 had undergone neoadjuvant therapy and 18 had not. Of 127 patients with lymph node disease, 62 had received neoadjuvant therapy and 65 had not. Eighteen patients developed local recurrence, of whom nine had had neoadjuvant therapy and nine had not.
Predictors of circumferential resection margin positivity
The overall CRM positivity rate was 13·9 per cent. A comparison of grouped smaller units (5 or fewer patients per year) and larger units (more than 5 patients per year) showed no difference in CRM positivity rates (14 and 13·9 per cent respectively) (Table 2). At least two-thirds of patients with a positive CRM margin had T4 or node-positive tumours. Multivariable analysis showed T4 tumours (OR 19·92, 95 per cent c.i. 6·48 to 68·61; P < 0·001) and positive lymph nodes (OR 3·04, 1·32 to 8·05; P = 0·004) to be independent predictors of CRM positivity. Surgical institution did not impact on CRM status (Table 3).
|CRM-negative (n = 260)||CRM-positive (n = 42)|
|Median (range) age (years)||65 (23–88)||66 (30–93)|
|1||151 (86·8)||23 (13·2)|
|2||47 (84)||9 (16)|
|3||24 (83)||5 (17)|
|4||26 (93)||2 (7)|
|5||12 (80)||3 (20)|
|Provider volume (patients/year)|
|> 5||198 (86·1)||32 (13·9)|
|≤ 5||62 (86)||10 (14)|
|Node positivity||99 (38·1)||28 (67)|
|Metastasis||9 (3·5)||4 (10)|
|Univariable analysis||Multivariable analysis|
|Odds ratio||P||Odds ratio||P|
|Age||1·22 (0·84, 2·35)||0·330|
|Hospital||1·24 (0·45, 6·17)||0·435|
|pT1/ypT1||0·92 (0·88, 0·95)||0·053||0·81 (0·69, 0·99)||0·067|
|pT2/ypT2||0·72 (0·67, 0·78)||< 0·001||0·79 (0·62, 0·94)||0·051|
|pT3/ypT3||1·83 (0·91, 3·67)||0·096||1·71 (0·78, 5·73)||0·435|
|pT4/ypT4||22·86 (7·61, 68·72)||< 0·001||19·92 (6·48, 68·61)||< 0·001|
|Node positivity||3·35 (1·73, 6·58)||0·001||3·04 (1·32, 8·05)||0·004|
|Metastasis||2·94 (0·86, 10·01)||0·090||1·97 (0·33, 5·61)||0·934|
Resection margin status and recurrence
Local recurrence occurred in 18 patients (6·0 per cent), of whom six were originally CRM-positive (Table 4). Tumour and node categories were not significant risk factors for local recurrence. In multivariable analysis the only independent predictor of local recurrence was CRM positivity (OR 3·63, 1·42 to 9·75; P = 0·016) (Table 5). CRM status did not influence distant recurrence (P = 0·796).
|CRM-negative (n = 260)||CRM-positive (n = 42)|
|Local recurrence||12 (4·6)||6 (14)|
|Time to local recurrence (months)*||41(22)||16(4)|
|Distant recurrence||30 (11·5)||5 (12)|
|Time to distant recurrence (months)*||36(17)||18(9)|
|Death||46 (17·7)||15 (36)|
|Time to death (months)*||39(17)||22(16)|
|Univariable analysis||Multivariable analysis|
|Odds ratio||P||Odds ratio||P|
|Age > 65 years||0·89 (0·34, 2·33)||1·000|
|Hospital||1·38 (0·74, 8·24)||0·416|
|pT1/ypT1||0·72 (0·09, 5·67)||1·000|
|pT2/ypT2||0·41 (0·09, 1·82)||0·381|
|pT3/ypT3||1·95 (0·67, 5·62)||0·323|
|pT4/ypT4||2·08 (0·44, 9·87)||0·297|
|CRM positivity||3·63 (1·27, 10·33)||0·022||3·63 (1·42, 9·75)||0·016|
|Node positivity||1·78 (0·68, 4·65)||0·324|
|Metastasis||1·35 (0·17, 11·11)||0·554|
Resection margin status and death
CRM positivity was an independent predictor of death in multivariable analysis (OR 2·94, 1·22 to 9·65; P = 0·017), and was associated with a lower 5-year survival rate than margin-negative tumours (25·4 versus 75·3 per cent respectively) (Fig. 1).
Results of standard abdominoperineal excision in other institutions
|Reference||No. of centres||Total no. of patients||Mean no. of patients/year||pT category||Node positivity (%)||CRT (%)||CRM positivity (%)||5-year local recurrence (%)||5-year overall survival (%)|
|Present study||5||302||14||T0–T2 112 (37·1)||42·1||50·0||13·9||6·4||61·9|
|T3/T4 190 (62·9)|
|Reshef et al.14||1||413||26||T0–T2 186 (45·0)||45||49||7||7||56|
|T3/T4 227 (55·0)|
|Messenger et al.5||1||115||12||T0–T2 56 (48·7)||36·5||58·3||15·7||10·6||67·4|
|T3/T4 59 (51·3)|
|Wibe et al.15||47||821||117||T0–T2 331 (40·3)||31||0||12||15||55|
|T3/T4 490 (59·7)|
|Dehni et al.16||1||106||18||T0–T2 60 (56·6)||NS||44||14||10||73|
|T3/T4 46 (43·4)|
|Matthis et al.17||1||246||18||Stage 1 130 (52·8)*||21·5||0||0·8||5·5||90|
|Stage 2 63 (25·6)*|
This study investigated the factors contributing to CRM positivity and oncological outcome in patients undergoing APE performed according to its original description. During this time, neoadjuvant therapy became commonplace18 and TME was popularized19. Treatment regimens have evolved over time and there is no doubt that some patients from the beginning of the data collection period would have been managed differently had they presented in the latter years of data collection. Notwithstanding alterations in neoadjuvant therapies and surgical technique, patient and tumour factors that affect prognosis have remained remarkably consistent.
Since the introduction of TME, outcomes following anterior resection have become a major focus of surgical research20. Patients who were treated with anterior resection have benefited from standardization of surgery and subsequent uniformly improved outcomes21, even for very low tumours22; however, there has until recently been little improvement in outcomes for patients treated with APE3. Reported high CRM positivity rates following APE have prompted re-examination of the surgical technique. Erroneous ‘coning’ of the excision specimen at the pelvic floor may have been a factor in CRM positivity leading to a redescription of the procedure with en bloc levator excision, similar (but perhaps even more radical) to that originally described by Miles23. This reduced the CRM positivity rate from up to 42 per cent to the more expected figures of 14·8–20 per cent24. Several centres have since reported results showing low CRM positivity rates, with good local recurrence and overall survival rates using ‘standard’ APE5, 14–17. Consequently, whether the improvement attributed to the modern ‘extralevator excision’ derivation of the original operation represents an advance in technique or, rather, highlights the potential drawback of TME-based coned excision of very low tumours is a question of perspective.
The data in the present study reflect standardized techniques applied by subspecialty trained surgeons. The effect of neoadjuvant therapy is difficult to determine from this study. Certainly, all tumours of pathological T0 category after radiotherapy (ypT0) represent complete pathological regression and it must be assumed that many of the ypT1 and ypT2 tumours were significantly downstaged. Owing to the long period of data collection and changing trends in pretreatment radiological staging, it is impossible to quantify the degree of downstaging in this cohort of patients. Preoperative radiotherapy cannot compensate for a positive margin; therefore the data were analysed based on final histological findings irrespective of neoadjuvant therapy. It is now known that response to neoadjuvant therapy correlates with better outcomes25, 26, and a personalized application of neoadjuvant therapy to patients who will benefit would be the ideal scenario. The need for investigations to predict response is clear but, unfortunately, predicting tumour response is not possible at present.
Surgeon and hospital volume have been championed as important determinants of oncological outcome in rectal cancer27. There is little doubt that rectal cancer should be managed in units with multidisciplinary support, although the minimum caseload required to provide adequate outcomes is difficult to calculate28. The present study combined single-handed units and multisurgeon units, but all participants were subspecialty trained colorectal surgeons. The data showed no relationship between CRM positivity and unit APE volume.
Patients with a positive margin had a much lower 5-year survival rate than other patients. It should be noted that two-thirds of patients with a positive margin had node-positive or T4 disease (even following neoadjuvant therapy in many cases). In these patients CRM positivity may represent a surrogate marker for aggressive, resistant disease rather than a comment on surgical technique. Certainly, a lack of response to preoperative therapy predicts poor long-term survival29.
This study has limitations. Preoperative imaging differed between centres over the data collection period. Tumour height was not uniformly reported and therefore could not be included for analysis. Although no difference in oncological outcome was found between units, it could be argued that patient numbers in smaller hospitals may not be sufficiently large to show differences. However, a comparison of the smaller-unit cohort versus the larger-unit cohort failed to show any difference in CRM positivity rates, making a statistical error less likely.
APE is an essential element in the management of low rectal cancer. Clear radial margins are the best predictor of outcome following surgery. In specialist units, low rates of margin positivity are possible with APE as originally described. Even more radical, wider ‘extralevator’ APE, resulting in a gaping musculocutaneous defect in the perineum, is probably required only for very advanced tumours.
A. E. Brannigan, N. Fearon and C. J. Shields (School of Medicine, University College Dublin, Mater Misericordiae University Hospital, Dublin); J. Deasy, D. Gough and D. A. McNamara (Beaumont Hospital, Dublin); J. M. Hyland and P. R. O'Connell (Centre for Colorectal Disease and School of Medicine, University College Dublin, St Vincent's University Hospital, Dublin); K. Mealy (Wexford General Hospital, Wexford); and E. Mulligan (James Connolly Memorial Hospital, Dublin).
The authors declare no conflict of interest.