Introduction

Anastomotic leakage is the most feared complication specific to colorectal surgery, leading to significant morbidity, increased mortality, prolonged hospital stay, and considerable extra cost. Its reported prevalence varies widely from 1 to 39%, but clinically significant leaks probably occur in 3–6% of cases, depending on the definition and the type of resection undertaken [1]. There is no universally accepted definition of a dehiscent colorectal anastomosis, which may present (a) as a generalized peritonitis requiring abdominal reoperation, (b) as fecal discharge from the wound and/or drain, (c) as a localized abscess, which may be amenable to computed tomography (CT) scan-guided percutaneous drainage, and (d) as an extravasation of radiological contrast in an otherwise asymptomatic patient, which may only require surveillance [2].

Several factors have been shown to have independent prognostic significance for anastomotic leakage, including diverticular disease [3], intraoperative septic conditions [4], male gender [5], smoking/alcohol abuse [6], as well as ASA score and emergency surgery [7]. The most significant risk factor for anastomotic leak remains, however, the site of the anastomosis, with leak rates of 2–4% in intra- vs 8–12% in infraperitoneal anastomoses [8]. In fact, Rullier et al. found that the risk of leakage was six times higher for anastomoses situated less than 5 cm from the anal verge than for those situated above 5 cm [9]. The majority of the large (>700 anastomoses) series, however, report either the experience of highly specialized surgeons over a long period of time (>10 years) [10, 11]; the results of multicenter trials [12, 13]; are limited to anastomoses above the pelvic peritoneal reflection [7]; focus on specific surgical techniques [14, 15]; or are retrospective in nature [16].

Therefore, the aim of this prospective study was (1) to identify the risk factors associated with for anastomotic leakage and (2) to evaluate the outcome of this type of complication in a nonselected series of patients who underwent colorectal resection in a single institution over a relatively short period of time.

Material and methods

From November 2002 until February 2006, all consecutive patients undergoing elective or emergency resection of the colon or rectum in our institution were prospectively included in this study and were entered into a prospective database. The University Hospital in Geneva is the only public medical institution in a mainly urban area with a population of approximately 500,000 inhabitants. The results in a larger series of patients in terms of morbidity and mortality have been previously reported [17]. Perioperative intravenous antibiotic prophylaxis was used in all patients. Most of them received a single dose of ceftriaxone 2G IV and metronidazole 500 MG IV at the time of induction. A majority of patients (>85%) did not receive any oral bowel preparation before surgery, in accordance to the results of a randomized controlled trial previously undertaken in our institution [18]. Our current policy is to restrict the use of mechanical bowel preparation in three different clinical situations: (1) patients scheduled for an APR, (2) patients who may need a perioperative colonoscopy, and (3) patients who are scheduled for a low anterior resection and in whom a diverting ileostomy is anticipated.

The structured sheet for data collection included the following items:

  1. 1.

    Patient characteristics: gender, age, American Society of Anaesthesiology (ASA) score, body mass index (BMI), and comorbidities (cardiopulmonary, neurological, hepatic, renal…).

  2. 2.

    Disease features: cancer, polyp, diverticulosis, diverticulitis, inflammatory bowel disease, other, and location of the pathology.

  3. 3.

    Surgical procedure: urgent or elective, open or laparoscopic, type of anesthesia, type of incision, duration of the procedure, blood loss, blood transfusion, type of resection (right, transverse, left, sigmoid, low anterior, abdomino-perineal), and type and location of anastomosis (manual or stapled, end-to-end, side-to-end, side-to-side, ileocolic, colocolic, colorectal, coloanal, ileorectal, ileoanal).

  4. 4.

    Postoperative events: anastomotic leak, mortality, morbidity, the reoperation, postoperative fever, prolonged ileus…

Anastomotic leakage was defined before the beginning of the study as either:

  1. 1.

    Radiological: demonstration of contrast extravasation on abdominal computed tomography scans with triple contrast or by gastrograffin enema

  2. 2.

    Causing diffuse peritonitis: presence of fecal fluid at relaparotomy

  3. 3.

    Causing local sepsis: presence of a localized abscess in the vicinity of the anastomosis

  4. 4.

    Fecal discharge from the drain/wound

In practice, we did not perform any routine contrast enema in asymptomatic patients, but we had a low threshold for abdomen/pelvic imaging with triple contrast CT scan in patients with suspected anastomotic leak, either clinically (pain, fever, abdominal tenderness, prolonged ileus) or biologically (persistently elevated white blood cell or C-reactive protein).

Statistical analysis

Descriptive statistics were summarized as means, standard deviations, medians, and ranges for the continuous variables and as frequencies and ranges for the categorical variables using the SAS V8.2 statistical software. When inference testing was completed, results with a p value less than or equal to 0.05 were considered statistically significant. Both univariable and multivariable logistic regression modeling techniques were used to identify risk factors for anastomotic leak. Parameter estimates, standard errors, odds ratios, and 95% confidence intervals for the odds ratios were reported along with the p values from the logistic regression modeling.

Results

Over a 40-month period, 811 anastomoses were performed in 423 women (52.4%) and 384 men (47.6%) with a median age of 67 (range, 17–98) years. Four patients had two anastomoses (right colectomy + low anterior resection). The median duration of hospital stay was 11 (range, 2–205) days. The characteristics of patients in this series are summarized in Table 1. One half of the operations were performed for tumors and one third for diverticular disease. The median duration of the procedures was 180 (range, 50–640) minutes. Emergency procedures were performed in 128 (16%) patients. A total of 231 patients (29%) presented with an ASA score ≥ 3. We performed 276 sigmoidectomies, 217 right colectomies, 99 low anterior resections, 86 left colectomies, 29 subtotal colectomies, 45 reversal of Hartmann, 40 ileocaecal resections, 11 proctocolectomies, and 8 transverse colectomies. A total of 184 operations (22.8%) were performed by laparoscopy, with a 12.8% conversion rate.

Table 1 Characteristics of patients
Full size table

The overall anastomotic leak rate was 3.8% (31 out of 811). The mortality rate associated with anastomotic leak was 12.9%. The overall mortality rate was 2.7% in the group of patients without a dehiscence. Low anterior resections and subtotal colectomies were the two procedures with the higher incidence of anastomotic leakage (11 and 13%, respectively). Eleven (35.4%) patients with an anastomotic leakage needed a reoperation for fecal diversion. The median delay between the first operation and the return to the operating room for anastomotic leak was six (range, 4–28) days. Twenty patients were conservatively treated with CT scan-guided radiological drainage and/or systemic antibiotics.

In univariate analysis (Table 2), the following parameters were associated with an increased risk for anastomotic dehiscence: (1) ASA score ≥ 3 (p = 0.004), (2) a prolonged (>3 h) operative time (p = 0.02), (3) rectal location of disease (p < 0.001), (4) infraperitoneal anastomosis (p = 0.02), (5) and a BMI > 25 (p = 0.04).

Table 2 Univariate analysis of risk factors for anastomotic leakage
Full size table

In multivariate analysis (Table 3), ASA score ≥ 3 (OR = 2.53; 95% CI 1.5–4.3, p < 0.001), a prolonged operative time (OR = 3.07; 95% CI 1.1–8.0, p = 0.02), and rectal location of the disease [OR = 7.69; 95% CI 2.2–27.3 (p = 0.001) for rectum vs right location, and OR = 3.75; 95% CI 1.5–9.0 (p = 0.003) for rectum vs left location] were factors significantly associated with a higher risk of anastomotic dehiscence.

Table 3 Multivariate analysis of risk factors for anastomotic leakage
Full size table

Discussion

The data presented in this paper indicate that the overall anastomotic leak rate after colorectal surgery in a single teaching institution is inferior to 4% and that two thirds of these cases were managed conservatively. Anastomotic dehiscence, however, remains a life-threatening complication, associated with a significant (13%) mortality and a prolonged (28 days) hospital stay. The risk factors for such a complication included ASA score > 3, prolonged duration of surgery, and an infraperitoneal anastomosis.

The overall anastomotic leak rate of 3.8% reported herein is similar to most other published series (Table 4), and it is interesting to note that the dehiscence rate of low anterior resections (12%) in this study is identical to those reported by the authors of the two most recent randomized trials on adjuvant radiation therapy for rectal cancer [19, 20]. However, the reoperation rate in our study (35%) compares favorably to other series reporting usually reoperation in 50 to 60% of patients with anastomotic dehiscence [21, 22] and up to 95% for leaks occurring after low anterior resections [23]. The mortality rate associated with leakage in this series (13%), although being fivefold higher than in patients without dehiscence (2.7%), is also comparable to those reported in other centers, ranging from 10 to 16% [24, 25, 26]. This relative low reoperation and mortality rates in patients with anastomotic dehiscence may be explained by the routine use of a loop ileostomy in our institution for all patients who underwent either a colo-anal or an ileo-anal anastomosis. It is known that proximal fecal diversion in itself will not reduce the incidence of a leak, but, in association with early recognition of the complication, it represents currently the best strategy to minimize the consequences of severe pelvic sepsis [27].

Table 4 Incidence of anastomotic leaks in several recent series of patients undergoing colorectal surgery
Full size table

Three parameters were identified in multivariate analysis as significant risk factors for anastomotic leakage: ASA score ≥ 3, a prolonged (>3 h) operative time, and anastomosis below the peritoneal reflection. In accordance with most authors, the ASA score ≥ 3 was strongly correlated with an increased risk for anastomotic leak [7, 21, 26, 28]: In our series, for every unit increase in ASA score, there is a 2.5 times increased risk of leak. Prolonged operating time has frequently been identified as a risk factor in univariate [4, 14] but more rarely in multivariate [26, 29] analysis: It is likely that long operation time reflects intraoperative difficulties, particularly, when working low in the pelvis or on account of adhesions from previous operations. Finally, anastomosis location within the rectum is over seven times more likely to leak than an ileocolic (right colon) and almost four times more likely to leak than a colocolic anastomosis (left colon).

Recent studies have well documented that anastomotic dehiscence after surgery for colorectal cancer compromises not only the immediate prognosis but, in addition, is associated with worse long-term survival and/or increased rate of local recurrence after a potentially curative resection [30, 31]. Therefore, it is critical to identify pre- or peroperatively high-risk patients to adapt the surgical strategy. We identified three risk factors for anastomotic leak in multivariate analysis: One is patient-related (ASA score), one is disease-related (infraperitoneal location of the anastomosis), the third being directly related to the surgical procedure (prolonged operating time). These factors should be considered in peroperative decision-making regarding defunctioning stoma formation.