Introduction

Current consensus in the management of locally advanced rectal cancer advocates the use of a multi-modal approach to treatment, with pre-operative conventionally fractionated radiotherapy and concurrent chemotherapy1–3 (CRT) followed by high quality mesorectal excision (TME) surgery4 to achieve the lowest possible local recurrence rates. Anastomotic leak after rectal cancer surgery is one of the most important potential complications of anterior resection, with the consequent risk of early and late mortality, morbidity, and implications for maintenance of intestinal continuity and quality of life.5–8 Previous studies have identified risk factors for anastomotic leak such as male gender, malnutrition, pre-operative weight loss, steroid use, peri-operative transfusion, and importantly pelvic radiotherapy. Studies in experimental animals have demonstrated reduced colonic anastomotic integrity following the administration of high dose radiotherapy,9,10 and reduced blood flow to colorectal anastomoses.11 It is also recognized that leak rates after anterior resection are increased following the formation of low pelvic anastomoses.12–14

Defunctioning ileostomy is widely used in rectal cancer surgery,12,15,16 with the intention of ameliorating the consequences of anastomotic leak, should this occur. In a meta-analysis of 4 randomized studies, comprising a total of 358 patients, the presence of defunctioning stoma significantly reduced clinical anastomotic leak (i.e., need for intervention should leak occur) after rectal cancer surgery with an odds ratio of 0.32 [0.17, 0.59]. It should be noted that the rate of clinically relevant anastomotic leaks (up to 28 % in non-defunctioned patients) in the only large study included in the meta-analysis was substantially higher than that generally reported by large centers.12–14 This raises questions about the results of the meta-analysis potentially being skewed by the findings of the larger study. The meta-analysis also identified a benefit for stoma creation in terms of re-operation, but no benefit in terms of improved mortality was observed.17

Due to the risk of complications that can occur in the creation and closure of defunctioning stomas and the need for readmission,18–22 most units have adopted a selective approach to defunctioning anastomoses in anterior resections, with recommendations for the creation of a defunctioning stoma for anastomoses within 6 cm of the anal verge.16,23 Given the current emphasis on enhanced recovery, minimally invasive colorectal surgery24,25 and patients’ desire for cosmesis and stoma avoidance, it is necessary to re-examine the strategy of defunctioning rectal resections and the indications for defunctioning patients who receive multi-modal treatment for rectal cancer. As stated above, it is likely that more patients will now receive neoadjuvant treatment for rectal cancer; however, there is no clear consensus for the routine use of defunctioning stomas in these patients. While there is evidence to suggest that radiotherapy may reduce anastomotic integrity it is not clear whether radiotherapy is an independent risk factor for leak following rectal resection, and how radiotherapy interplays with other potential risk factors such as gender, site of lesion and patient age. The aim of this study is to investigate the risk of anastomotic leak in patients undergoing restorative rectal cancer resection in the presence of radiotherapy and to determine if the creation of a defunctioning ostomy protects against this potential risk.

Methods

All patients who underwent rectal cancer resection at the Cleveland Clinic Foundation from 1980 to 2010 were identified from the prospectively maintained Cleveland Clinic cancer database. Patients were grouped according to whether they received neoadjuvant radiotherapy (+RT) or no neoadjuvant radiotherapy (−RT), prior to rectal cancer resection. This study was approved by the Institutional Review Board and is HIPPA compliant.

Outcome Measurement

The primary outcome of interest was the occurrence of anastomotic leak following surgery. Anastomotic leak was defined by imaging, clinical findings, or operative findings, and therefore any type of leak (clinical and subclinical leak) was considered as the primary outcome. The following criteria were used to define leak: (a) the presence of anastomotic separation or pelvic abscess on post-operative imaging with contrast enema or CT scan, (b) postoperative physical signs of abdominal sepsis with a staff surgeon documenting a diagnosis of leak, or (c) intra-operative diagnosis made by a staff surgeon during abdominal exploration or examination under anesthesia.

Inclusion and Exclusion Criteria

Patients were included if they underwent transabdominal restorative rectal cancer resection with tumor site at 15 cm or less from the anal verge. All included patients underwent resection with curative intent. Patients with pathologic diagnoses other than adenocarcinoma were excluded. The usual practice at our institution is to perform a high ligation of the inferior mesenteric artery with the utilization of well vascularized colon (usually left) for the colorectal or colon-anal anastomosis. Splenic flexure mobilization is performed for additional colon length when indicated. Patients were excluded if they underwent transanal local excision of tumor, abdominoperineal excision of rectum or ileal pouch anal or rectal anastomosis. Patients undergoing resection for recurrent cancer were also excluded. A long-course neoadjuvant radiotherapy regime is used at the Cleveland Clinic, utilizing 45 Gy over 5 weeks, with surgery performed at 4–6 weeks after completion of treatment.

Study Variables

Data on patient demographics, ASA grade, height of tumor, clinical tumor stage as determined by a combination of clinical, endorectal ultrasound and/or magnetic resonance imaging, use of chemotherapy, use of adjuvant radiotherapy, type and level of anastomosis, type of stoma, and surgeon volume were collected. In the cases with a stapled anastomosis, an anastomosis located 3 cm or less from the anal verge was defined as a colo-anal anastomosis. Pathological tumor variables included size of tumor, TNM stage, and distal margin. Considering the duration of the study period over which patients underwent surgery, details pertaining to the quality of mesorectal excision were not uniformly available for all patients. Secondary endpoints were length of inpatient stay, 30-day mortality and type of intervention required for anastomotic leak (CT guided drainage, examination under anesthesia, exploratory laparotomy). Quality of life was assessed prospectively using the SF-36 questionnaire and fecal incontinence severity index (FISI).

Statistical Analysis

Statistical analysis was performed using R software (www.r-project.org). Categorical variables are summarized as frequencies and percentages. Quantitative variables are summarized as mean ± standard deviation. Tests for association between −RT and +RT groups and categorical variables were performed using the Chi-square method. For quantitative variables the Wilcoxon rank sum method was used. A p value <0.05 was accepted as statistically significant. Multivariate models were generated using the forward stepwise method, with a univariate entry cutoff of p = 0.05 for inclusion.

Results

From 1980 to 2010, a total of 1,862 patients underwent restorative rectal cancer resection and were included in the study. 1,338 (71.9 %) patients did not receive neoadjuvant radiotherapy (−RT group) and 524 (28.1 %) patients received neoadjuvant radiotherapy (+RT group). In the +RT group 448 (85.5 %) patients received neoadjuvant chemoradiotherapy. Five surgeons contributed >100 cases each, comprising 73 % of the sample. Baseline characteristics and details of anastomotic configuration in −RT and +RT groups stratified by the creation or otherwise of a defunctioning ostomy are given in Table 1. In the +RT group, there were significantly more males, patients were younger at the time of surgery, more patients had a lower third rectal cancer site and more patients had a colo-anal anastomosis (including straight, colopouch or coloplasty configuration). In the −RT group, a diverting colostomy was performed in 80 (6 %) patients and a diverting ileostomy in 514 (38.4 %). Post-operative (adjuvant) radiotherapy was given to 155 (11.5 %) of patients in the −RT group. In the +RT group, a diverting colostomy was performed in 25 (4.8 %) patients and a diverting ileostomy in 430 (82.1 %).

Table 1 Baseline characteristics and anastomotic configuration in −RT and + RT groups
Full size table

Four hundred and twenty-five patients in the cohort underwent surgery prior to 1989, 629 between 1990 and 1999 and 808 between 2000 and 2010. The cumulative occurrence of leak during these time periods was 7.3 %, 5.9 %, and 6.2 %, respectively (p = 0.64). In 1,623 patients, a stapled anastomosis was performed and in 118 patients a handsewn anastomosis with leak rates of 4.4 % and 4.3 %, respectively (p = 0.94). Overall, 118 (6.3 %) patients had an anastomotic leak, 76 (5.7 %) in the −RT group vs. 42 (8 %) in the +RT group (p = 0.06). Outcomes for the two groups are provided in Table 2. There was no statistically significant difference in occurrence of anastomotic leak between −RT and +RT groups when patients were subcategorized according to presence of diverting ostomy, tumor location, and colo-anal vs colorectal anastomosis. There was no significant difference in 30-day mortality in −RT vs +RT patients, however +RT patients did have a shorter length of inpatient stay. In the −RT group, 31 (5.2 %) of undiverted patients developed anastomotic leak, in the +RT group, 8 (11.6 %) of undiverted patients developed a leak. Table 3 describes the management of leaks in undiverted patients in −RT and +RT groups. The overall frequency of exploratory laparotomy for leak patients that were undiverted was 77.8 % in the −RT group and 37.5 % in the +RT group. Table 4 describes the management of leaks in diverted patients in −RT and +RT groups. The overall frequency of exploratory laparotomy for leak patients that were diverted was 12 % in the −RT group and 9.1 % in the +RT group.

Table 2 Anastomotic leak and outcomes in −RT and + RT groups
Full size table
Table 3 Management of anastomotic leaks in undiverted patients in −RT and + RT groups
Full size table
Table 4 Management of anastomotic leaks in diverted patients in −RT and +RT groups
Full size table

The 10-year local recurrence rates in the −RT group were 4.2 % in defunctioned and 6.4 % in non-defunctioned patients. In the +RT group, it was 5.1 % defunctioned and 4.8 % in non-defunctioned patients (p = 0.94). The risk of local recurrence was significantly increased in patients who sustained anastomotic leak; OR = 3.29 (1.71–6.32, p = 0.001). In +RT patients with a leak, the increased risk of local recurrence was greater; OR = 6.01 (2.16–16.76, p = 0.002) than in −RT patients with a leak; OR = 2.32 (0.96–5.62, p = 0.06). There was no statistical difference in distal margin between patients in the −RT vs +RT groups (1.16 ± 0.87 cm vs 1.00 ± 0.68 cm, p = 0.52) or in patients who sustained a leak vs those without a leak (1.10 ± 0.99 cm vs 1.08 ± 0.76 cm, p = 0.94).

Predictors of Anastomotic Leak

Univariate factors associated with anastomotic leak are presented in Table 5. Using forward stepwise inclusion, the following factors were identified as independent predictors of anastomotic leak (given as OR [95 % CI]): ASA grade 4 4.70 (2.07–10.7), p < 0.001; tumor in the lower third of the rectum 2.49 (1.37–4.52), p = 0.003; tumor size at surgery over 4 cm 1.75 (1.15–2.65), p = 0.009; and male gender 1.64 (1.03–2.62), p = 0.038. Neoadjuvant radiotherapy 1.44 (0.85–2.46), p = 0.18; pathologic T4 disease 0.60 (0.18–2.01), p = 0.4; M1 disease 1.54 (0.89–2.68), p = 0.13 were not independently associated with anastomotic leak. Creation of a diverting ostomy 0.75 (0.44–1.28), p = 0.29 was also not independently associated with a reduced risk of anastomotic leak.

Table 5 Univariate analysis of factors associated with anastomotic leak
Full size table

Quality of Life and Bowel Function

Table 6 summarizes quality of life and bowel function in undiverted vs diverted patients in the −RT and +RT groups respectively. Questionnaires were returned at median 3.9 (1.9–4.9) years in −RT group and 2.9 (0.9–2.8) in the +RT group. There were no significant differences in the physical and mental SF-36 summary scores between undiverted and diverted patients in the −RT and +RT groups. In the −RT group, an increased nocturnal stool frequency and increased frequency of pad use was observed in diverted patients. In the +RT group, an increased frequency of major stool incontinence was observed in diverted patients

Table 6 Quality of life and bowel function in patients diverted vs undiverted −RT patients and diverted vs undiverted + RT patient
Full size table

Discussion

In this large single institution series, we have analyzed the risk of anastomotic leak in patients undergoing restorative rectal cancer resection with curative intent. The objective of this study was to examine the effect of neoadjuvant radiotherapy on anastomotic leak rate in these patients. While there is a sound scientific basis for anticipating an increased anastomotic leak rate in patients receiving neoadjuvant radiotherapy, this finding has not been firmly established by clinical studies, with conflicting reports in the literature. In a large series, Enker and colleagues did not observe an increased leak rate in 150 patients undergoing low anterior resection with neoadjuvant radiotherapy.26 In a large randomized study, Sauer and colleagues also did not observe an increased postoperative leak rate in patients receiving neoadjuvant vs adjuvant chemoradiotherapy who underwent sphincter preserving surgery.1 However other reports containing smaller numbers (n = 37–50) of patients receiving neoadjuvant radiotherapy have described an increased risk of leak in these patients.27,28

The findings of the current study therefore agree with the larger studies mentioned, as we have also not been able to demonstrate an association with neoadjuvant radiotherapy and leak. Whilst there was a very close statistical trend towards increased crude leak rate in +RT patients, on multivariate analysis radiotherapy was not a predictor for anastomotic leak. Furthermore in subgroup analysis, neoadjuvant radiotherapy did not significantly increase leak rate after low (colo-anal) anastomosis, after surgery for lower third rectal cancer or in patients who were not diverted. In addition we did not observe an increased requirement for exploratory laparotomy in +RT patients who developed a leak, although the pelvic abscess rather than peritonitis in this group may be due to increased frequency of low defunctioned anastomoses. The use of radiotherapy did not appear to adversely affect the management of patients who developed a leak and is unlikely to influence the degree of sepsis in these patients. As stated, it is the absence of diversion in patients who sustain a leak that increases the necessity for major surgery.17 In diverted +RT patients that did have a leak, we observed a much lower requirement for major abdominal surgery (9.1 %) compared with undiverted patients (37.5 %). The requirement of exploratory laparotomy for undiverted −RT patients was however even higher than those who were diverted (77.8 vs 12 %).

The current study does have limitations, in that it retrospectively analyzes data in pre-selected patient groups, with the decision to divert left to the operating surgeon, and therefore only 13.4 % of patients in the +RT group were undiverted. A lower threshold for the diversion of +RT patients may also have been responsible for the higher need for exploratory laparotomy for the −RT patients who were not diverted and suffered a leak. Ideally, a prospective randomized study of diversion in +RT patients would provide a sound scientific method of determining the role of diversion in this setting. However, it would be inadvisable to conduct such a study as +RT patients frequently possess adverse tumor characteristics which would influence the decision to divert. In view of the fact that the vast majority of +RT patients in our tumor registry were diverted, we chose to include all anastomotic leaks as our principal outcome measure, rather than the rate of symptomatic or clinically relevant leak that has been reported in studies investigating the role of diverting ileostomy.29,30 A case-matched design for factors found to be significant between the +RT and −RT groups was also considered. However, the large numbers of variables did not allow for sufficient numbers in the groups to allow for an adequate comparison. We hence conducted a multivariable analysis adjusting for the creation of diverting ostomy so as to identify predictors of outcome in this large cohort as an alternative design. Omission of diverting ostomy and use of neoadjuvant radiotherapy were not identified as predictors of leak in this study.

We have previously described the morbidity associated with defunctioning loop ileostomy in patients with ileo-anal pouch31 and the published morbidity of loop ileostomy creation and closure in studies on unselected patients ranges from 18 % to 33 %, with mortality ranging from 0 % to 2.5 %.32–35 In colorectal cancer patients the published morbidity associated with loop ileostomy is 19.2 % and mortality is 1.7 %.36 In this study, we also observed an adverse association between fecal diversion and bowel function after restorative rectal resection for specific variables. Pad usage was increased in diverted patients in the −RT group and stool incontinence was increased in diverted +RT patients. While continence after rectal cancer surgery is dependent on multiple factors, this may be related to the fact that patients with more adverse factors and lower anastomoses were selected out for diversion. We were not able to conclusively demonstrate the effect of radiotherapy on bowel function after restorative surgery, as 11.5 % of −RT patients did receive adjuvant radiotherapy.

Consistent with previous reports,13,28 including one from our unit,14 height of tumor (i.e., in the lower rectum) is a central factor in the decision to divert. We have also identified ASA grade 4, tumor size at surgery over 4 cm and male gender as predictors for anastomotic leak. Therefore, the decision to divert in restorative rectal cancer surgery should be made on a case by case basis taking into account patient and tumor characteristics. Our data support the use of a selective policy of diversion based on these factors, rather than a blanket approach towards patients receiving multimodal treatment for rectal cancer.