Definitive Risk Factors for Anastomotic Leaks in Elective Open Colorectal Resection

Hypothesis  Anastomotic leaks following elective colorectal resections increase morbidity, mortality, and the need for additional interventions. An accurate understanding of risk factors would potentially reduce anastomotic leaks and/or allow appropriate selection of patients for diverting stomas.

Design  Prospective review of patient and operative characteristics that contribute to anastomotic leaks.

Setting  Fifty-one sites within the United States (May 2002-March 2005).

Patients  Six hundred seventy-two patients who participated in a trial comparing preoperative antimicrobials in elective open colorectal surgery.

Main Outcome Measures  Anastomotic leaks were diagnosed using clinical findings and were confirmed with imaging. We examined 20 variables possibly affecting anastomotic healing in univariate and multivariate analyses.

Results  There were 24 anastomotic leaks in 672 patients (3.6%) undergoing elective colorectal resection. There were 10 deaths (1.5%). A baseline albumin level of less than 3.5 g/dL (to convert to grams per liter, multiply by 10) (P = .04) and male sex (P = .03) were associated with anastomotic leaks in both univariate and multivariate analyses (adjusted odds ratios, 2.56 and 3.12, respectively). Increased duration of surgery (SD, 60 minutes; odds ratio, 1.53; 95% confidence interval, 1.06-2.22; P = .03) and steroid use at the time of surgery (odds ratio, 3.85; 95% confidence interval, 1.24-11.93; P = .02) were significant in univariate analysis. Surgical procedure with rectal resection; prophylaxis with ertapenem (vs cefotetan); or history of obesity, tobacco use, or diabetes was not associated with anastomotic leaks.

Conclusions  Significant risk factors for anastomotic leaks include low preoperative serum albumin level, steroid use, male sex, and increased duration of surgery. Appreciation of risk factors provides a rational basis for temporary diversion.

Anastomotic leaks are recognized as the primary morbidity in colorectal surgery, often requiring diversion of the gastrointestinal tract and causing longer hospitalizations and increased mortality.^1,2 The more frequent recurrence of colorectal carcinoma associated with anastomotic leaks is less appreciated.^3,4 We report an analysis of risk factors associated with leaks after colorectal surgery based on outcome data collected from a large, multicenter, prospective clinical trial. Understanding patient and operative characteristics may lead to reduction in the frequency of anastomotic leaks through improvement in risk status or more appropriate selection of patients for diverting stomas.

Anastomotic leaks after elective colorectal resection reportedly occur in 4% to 9% of patients.^5-7 Few patient series, however, have had prospective collection of detailed clinical data and some rely on retrospective analysis of commercially oriented databases that may not provide leak rates.⁸ With the increasing prevalence of laparoscopic colectomy, it may be useful to establish a contemporary benchmark for leak rates in open surgery against which the occurrence of leaks after laparoscopic colectomy could be compared.

We prospectively analyzed risk factors for anastomotic leaks in elective colon resections. Patients were recruited from a large, multicenter study that was designed as a trial of perioperative antimicrobial drugs in elective open colorectal resections. The trial was a randomized, double-blind comparison of intravenous antimicrobial prophylaxis (2g of cefotetan vs 1 g of ertapenem) given as a single dose within 1 hour before surgery. No oral antimicrobials were given preoperatively. A detailed description of the methods and results of this study with regard to infection outcomes has been published previously.⁹ The study demonstrated a decreased incidence of postoperative surgical infections after 4 weeks of follow-up in patients who received ertapenem.

Of the 1002 randomized patients, 901 qualified for the modified intention-to-treat population and 672 were included in the evaluable population. To qualify, the patient was required to have undergone electively scheduled open colorectal surgery with completion of standard bowel preparation (sodium phosphate or polyethylene glycol) and received a complete dose of study medication. The antimicrobial drug must have been administered within 2 hours of incision and 6 hours of surgical closure. Patients were considered unevaluable if they were treated with delayed primary closure, developed a distant site infection, violated the 4-week follow-up, received prior antibiotic therapy, or had a study therapy dosing window violation. Patients were excluded because of a failure to meet surgical definition (n = 40), a bowel preparation violation (n = 8), a failure to meet follow-up guidelines (n = 27), a distant site infection (n = 67), a prior antibiotic violation (n = 63), a study therapy dose violation (n = 69), and/or other causes (n = 14). Patients were recruited from 51 sites within the United States from May 2002 through March 2005. Each site obtained local institutional review board approval and informed consent from each patient. Anastomotic leaks were analyzed only in the 672 patients who were in the evaluable population. Patients were assessed for multiple clinical and laboratory parameters during their hospital stay and were observed for 4 weeks after treatment. Patients were examined for signs of infection at the surgical and other sites. Anastomotic leaks were diagnosed if sepsis was present. Evidence of the leak was confirmed radiologically.¹⁰ Subsequent treatment of anastomotic leaks was left to the judgment of the operating surgeon. Patients were older than 18 years. All operations were elective open colorectal resections. Laparoscopic and emergency colorectal operations were excluded. Elective colorectal procedures for revision of previous surgery (eg, colostomy closure or revision) and isolated rectal procedures were also excluded from the analysis.

We examined 20 variables in univariate and multivariate analyses. Patient characteristics (Table 1) included age, race, sex, tobacco use, obesity (body mass index [calculated as weight in kilograms divided by height in meters squared]>30), renal function (creatinine clearance <30 mL/min/1.73 m² [to convert to milliliters per second per meters squared, multiply by 0.0167]), history of diabetes, history of chronic obstructive pulmonary disease, preoperative albumin level (<3.0 g/dL [to convert to grams per liter, multiply by 10]), hematocrit level, and steroid use at time of surgery (0-40 mg/d). Operative variables (Table 2) comprised whether the procedure included rectal resection, indications for surgery, type of bowel preparation, whether the procedure required protective stoma or placement of drain, surgical duration, and whether gross spillage was encountered. Resections were not divided between right and left but whether or not they included the rectum. All patients received mechanical bowel preparation. No oral antibiotics were given with the bowel preparation. The decision to construct a protective colostomy or ileostomy or to place a drain was at the surgeon's judgment.

Anastomotic leaks were suspected in patients who exhibited clinical signs of infection, including increasing pain, fever, tachycardia, and a distended abdomen. Anastomotic leaks were diagnosed by using evidence of fecal contamination draining externally or with imaging. Computed tomography was most frequently used, though a water-soluble contrast enema could also be used. A fluid collection adjacent to the anastomoses associated with extraluminal contrast confirmed anastomotic leaks on imaging studies.

Demographic characteristics were typical for patients undergoing colorectal surgery in North America. Mean patient age was 61 (SD, 14) years; 54% were male; and 77% were white. Half of the patients were tobacco users. Obesity was evident in 29% of our patients. Patients had evidence of malnutrition and chronic disease, demonstrated by albumin levels less than 3.5 g/dL (21%). Five percent of our patients were taking steroids at the time of surgery. Concomitant medical conditions included diabetes (18%), renal insufficiency (1.3%), and chronic obstructive pulmonary disease (6%).

Rectal resection was required in 23% of our patients. However, we did not differentiate between left and right colon resections. Colon and rectal cancers were the most frequent indications for operation (47% and 17%, respectively), followed by diverticulitis (11%) and benign colonic neoplasm (9%). All patients underwent mechanical bowel preparation, with 45% receiving polyethylene glycol and the remainder receiving sodium phosphate. Patients did not receive oral antibiotics as part of their preparation. Protective colostomy or ileostomy was constructed in 17% of our patients and drains were placed in 27%. Inadvertent spillage occurred in 2.7% of the operations. Mean operative duration was 133 minutes.

There were 24 anastomotic leaks in 672 patients (3.6%). Of the 10 deaths (1.5%), 1 was associated with an anastomotic leak (10%). Patients who received ertapenem had a lower, but not statistically significant, leak rate compared with those who received cefotetan (3% and 4.2%, respectively) (Table 3). In a multivariate analysis, a baseline albumin level less than 3.5 g/dL (odds ratio [OR], 2.56; 95% confidence interval [CI], 1.07-6.16; P = .04) was significantly associated with developing a leak (Table 4). Female sex was protective (OR, 0.32; 95% CI, 0.12-0.88; P = .03). Chronic steroid therapy was significantly associated with leaks in univariate analysis (OR, 3.85;95% CI, 1.24-11.93; P = .02) and showed a strong trend in multivariate analysis (OR, 3.18; 95% CI, 0.97-10.43; P = .06). Increased duration of surgery (SD, 60 minutes) was also significantly associated with leaks in a univariate analysis (OR, 1.53; 95% CI, 1.06-2.22; P = .03).

Anastomotic leaks occurred in 3.6% of 672 patients after elective colorectal resection. This is within the range of prior series (2%-8%).^11-15 Our patients were prospectively studied, and surgeons were asked to report evidence of anastomotic leaks based on clinical and radiological findings. Specific patient characteristics were collected, and follow-up was standardized so that all patients were examined at regular intervals during their hospital stay until 4 weeks after their operation. Since patients were recruited in 51 sites within the United States, this incidence of anastomotic leaks likely represents a reliable benchmark for this outcome in North America.

In the multivariate analysis, low serum albumin level (<3.5g/dL) and male sex were significantly associated with anastomotic leaks. New Zealander surgeons have also confirmed male sex as a risk factor.¹⁶ We did not find other significant characteristics that led to leaks, eg, diabetes, tobacco use, alcohol use, duration of surgery, intraoperative fecal spillage, duration of operation, blood transfusion, or rectal anastomoses. Several of these characteristics, however, are associated with postoperative infection.^17,18 Such patient and operative characteristics may serve as a guide to the surgeon as to whether he or she should construct a protective stoma. Construction of a protective stoma relies on the surgeon's experience, particularly his or her ability to judge blood supply adequacy and perform a sound anastomosis in an adverse anatomic situation. We present our findings as additional data not only to guide the surgeon in operative decisions but also to aid in counseling patients regarding the likelihood of undergoing protective diversion. Men who were malnourished (serum albumin level <3.5 g/dL) or were taking steroids long-term had a higher risk for leaks in our series; they should be told of this risk preoperatively and would benefit most from construction of a protective stoma.

Contrary to prior studies, we did not find anastomoses to the rectum to be more significantly associated with postoperative leaks. Several authors report an increased risk as high as 11%, particularly with low anastomoses (<8 cm from anal verge) and with preoperative chemoradiation.¹⁹ Our data did not specify exact level of rectal anastomoses, possibly leading us to miss this association. Furthermore, some studies have described different leak rates in right and left colon resections. Right colon resections are usually less demanding technically; therefore, lower associated leak rates would be expected. For example, Veyrie et al⁵ report leak rates for right colectomies of 1.35% vs 5.20% for left colon resections. Subclinical leaks have been described in colorectal surgery, exist in a higher percentage of patients, and have been found in 14% of patients who are followed up for evidence of radiological leakage.²⁰ Minor clinical anastomotic leaks can also affect patient outcomes, though our study was not designed to detect minor leaks.²¹

The anastomotic leak rate in this study may serve as a benchmark against which the rate in laparoscopic colorectal resections may be compared. Large series of laparoscopic colorectal resections and their associations with anastomotic leaks have been published, but there are few compared with studies of open surgical resections.^22-24 We expect the same patient and operative characteristics that are significantly associated with a leak after an open operation to be similarly associated with leaks in laparoscopic surgery. The laparoscopic approach to colorectal surgery has not been as universally adopted as operations like cholecystectomy and fundoplication.

Although each patient received a polyethylene glycol or sodium phosphate mechanical bowel preparation, preoperative nonabsorbable oral antibiotics were not given in this trial. Despite absence of oral antibiotics, our anastomotic leak rate was not different from those published in prior work. Also, there was no association between bowel preparation and leak rates. Prior studies indicate that there may be no significant benefits of using bowel preparation; however, we have recently published data that suggest that bowel preparation may reduce surgical site infections.^25-27

Correspondence: Samuel Eric Wilson, MD, University of California–Irvine, 333 City Blvd W, Ste 810, Orange, CA 92868 (sewilson@uci.edu).

Accepted for Publication: May 22, 2008.

Author Contributions:Study concept and design: Jensen, Abramson, Itani, and Wilson. Acquisition of data: Jensen, Abramson, Itani, and Wilson. Analysis and interpretation of data: Suding, Abramson, Itani, and Wilson. Drafting of the manuscript: Suding, Jensen, Abramson, Itani, and Wilson. Critical revision of the manuscript for important intellectual content: Suding and Wilson. Statistical analysis: Jensen and Abramson. Obtained funding: Abramson and Wilson. Administrative, technical, and material support: Abramson, Itani, and Wilson. Study supervision: Wilson.

Financial Disclosure: Data on anastomotic leaks were obtained from a study supported by Merck, which examined ertapenem compared with cefotetan for colorectal surgery prophylaxis. Drs Itani and Wilson receive consulting fees from Merck, and Dr Wilson receives lecture fees from Merck. At the time of manuscript preparation, Ms Jensen and Dr Abramson were employees of Merck and reported having equity in the company. No other potential conflicts of interest relevant to this article exist.

Previous Presentations: This paper was presented at the 2008 Annual Meeting of the Pacific Coast Surgical Association; February 17, 2008; San Diego, California; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.

Financial Disclosure: None reported.