Changes in Clinical Practice Reduce the Rate of Anastomotic Leakage After Colorectal Resections

This is a cohort study that included all consecutive patients treated with colon or rectal resection and primary anastomosis at the Karolinska University Hospital, a tertiary referral center, between January 2006 and December 2013. Colorectal resections were performed with standardized mainly open surgery during the study. Surgical principles, preoperative and postoperative care remained unchanged during the entire study period. Structured changes in clinical practice aiming at reducing the anastomotic leakage rate were introduced in January 2010. Cumulative incidence of anastomotic leakage was the primary outcome and compared between patients treated before (period 1: 2006–2009) and after (period 2: 2010–2013) implementing the changes in clinical practice.

Changes in clinical practice aiming at reducing the risk of anastomotic leakage included three key elements:

The same research nurse recorded data prospectively throughout the whole study period on sex, age, diagnosis, surgical procedure, length of stay, elective/emergency surgery, type of complications (including anastomotic leakage) and surgical re-interventions in a dedicated database. Computed tomography with rectal contrast enema was the standard examination in patients with a deviant postoperative course and was liberally performed. Occurrence of anastomotic leakage was registered during index hospital stay or 30 days postoperatively. Visible signs of an anastomotic defect during reoperation, leakage of luminal contrast on radiological assessment or need for percutaneous drainage of any intraabdominal fluid collection near the anastomotic site was regarded as anastomotic leakage. Anastomotic leakage was thus defined according to the International Study Group of Rectal Cancer as a defect of the intestinal wall at the anastomotic site (including suture and staple lines of neorectal reservoirs) leading to a communication between the intra- and extraluminal compartments, or as an abscess adjacent to the anastomosis [18]. Early postoperative radiology with rectal contrast enema was not routinely performed for all patients. This was only performed in case of a deviant postoperative course, e.g., leukocytosis or fever. Hence, only grade B and C leakages requiring any therapeutic intervention were included in this trial.

Anterior resection implied TME surgery with transection of the inferior mesenteric artery (IMA) [19, 20] with ligation of the inferior mesenteric vein and the left colic artery at the same level. The marginal artery and terminal bifurcation of the ascending left colic artery were spared in order to preserve adequate blood supply to the descending colon in case of an insufficient marginal artery at the splenic flexure [21]. The left colon was divided after complete mobilization of the splenic flexure in order to resect the sigmoid together with the specimen. The rationale for a complete resection of the sigmoid colon was to obtain a better anastomotic blood perfusion, since dividing the IMA results in a decreased arterial perfusion in the sigmoid colon which may decrease tissue oxygenation in the proximal limb and hence increase the risk of anastomotic leakage [22‐24]. Before transecting the colon the marginal artery and vasa recta supplying the proximal staple line were indentified and preserved. A stapled side-to-end colorectal anastomosis was performed [25, 26]. Except for non-irradiated women without other risk factors a defunctioning loop ileostomy was recommended. In sigmoid resections the IMA was divided at its origin, the bowel transected at the level of the promontory and the anastomosis was stapled. In right-sided colectomies the ileocolic vessels were divided at their origin together with the right branch of the middle colic artery after identification of the superior mesenteric vein. Anastomoses were hand sewn or stapled. The splenic flexure was resected and the middle colic artery divided at its origin in extended right-sided colectomies or resections of the transverse colon.

There were no changes in operative techniques in period 2 compared to period 1.

Symptomatic anastomotic leakages with no sign of fecal peritonitis were managed by transanal or percutaneous drainage and antibiotics. If there were signs of fecal peritonitis a laparotomy was performed. In severe cases the anastomosis was taken down and a descending colostomy created. In cases of limited pelvic sepsis with only a minor anastomotic defect a protective ileostomy was created if not already present. Patients with anastomotic leakage after colonic resections were managed similarly, but the anastomosis was more likely taken down with a creation of a deviating ostomy. However, if bowel continuity was present and sepsis well confined to an abscess adjacent to the anastomosis the patient could be managed more conservatively with percutaneous drainage and antibiotics. The management of anastomotic leakage was the same during the whole study period.

All electively treated study participants were included in the ERP, implemented in January 2006. The ERP included preoperative carbohydrate-rich beverage, intra-operative buffered glucose solution (25 mg ml⁻¹) at a rate of 2–3 ml kg⁻¹ h⁻¹ to replace insensible loss and norepinephrine to achieve a mean arterial pressure of above 60–70 mm Hg. Oral bowel preparation was restricted to patients scheduled for anterior resection. For analgesia, a low thoracic epidural catheter was inserted before induction of anesthesia and celecoxib, a selective cyclooxygenase (COX) 2 inhibitor, was routinely used for postoperative pain management.

The reason for surgery did not change between the two time periods and were colorectal neoplasia in 85%, inflammatory bowel disease (IBD) in 13% and other pathology in 2%. Age, sex, type of resection, the proportion of multivisceral resections and protective ostomies were similar in both groups (Table 1). The proportion of emergency resections decreased significantly from 14 to 4% (p < 0.01) in colon resections and from 4% to zero (p = 0.01) in rectal resections (Table 1).

The cumulative incidence of anastomotic leakage was 10.0% (41 of 409) before the change in clinical practice (period 1) and decreased to 4.5% (22 of 485) after the change in clinical practice (period 2) with a relative risk (RR) for anastomotic leakage of 0.45 (95% CI 0.27–0.75, p = 0.002). For colon resections the cumulative incidence decreased from 9.1% (23 of 252) to 4.5% (14 of 310) and for rectal resections from 11.5% (18 of 157) to 4.6% (8 of 175). The relative risk was 0.49 (95% CI 0.26–0.94, p = 0.039) for colon and 0.40 (95% CI 0.18–0.89, p = 0.024) for rectal resections, respectively (Table 2).

The effect of the changes in clinical practice was similar in the analysis restricted to elective colorectal resections with RR 0.47 (95% CI 0.28–0.79, p = 0.004). Figure 1 displays the cumulative incidence of anastomotic leakage per year. The incidence decreased after the introduction of the changes in clinical practice, and the highest value in period 2 was consistently lower than the lowest value in period 1.

The logistic regression resulted in an unadjusted OR for anastomotic leakage of 0.43 (95% CI 0.25–0.73, p = 0.002) for all patients operated with colorectal resections in period 2 versus period 1 (Table 3). The crude effect of age, sex, type of resection, type of pathology, multivisceral resection, acute surgery and protecting ostomy on anastomotic leakage was statistically not significant. The multivariable logistic regression indicated no confounding effect of the predictors above, nor were their interaction terms significant. The model adjusted for age, sex, type of resection and acute surgery resulted in an OR of 0.45 (95% CI 0.26–0.78, p = 0.004). The multivariable analysis restricted to elective colorectal procedures yielded an OR of 0.47 (95% CI 0.27–0.82, p = 0.007) after adjustment for age, sex, type of resection and acute surgery.

The median length of stay for participants with an anastomotic leakage was longer; 25 (6–190) days versus 9 (2–167) days, (p < 0.001) in the entire study population. The median length of stay was shorter after the introduction of changes in clinical practice; 9 (3–119) days versus 10 (2–190) days (p = 0.043) (Table 4).

The majority of patients referred to the Karolinska University Hospital have advanced colorectal cancer or complex inflammatory bowel disease, which is reflected by a high proportion of simultaneous multivisceral resections. The risk of selection bias was minimized by inclusion of all consecutive patients undergoing colorectal resection with primary anastomosis. During the entire study period the referral pattern did not change and patients treated before and after the changes in clinical practice were comparable regarding age, sex, type of diagnosis, type of resection, multivisceral resection and protective ostomy. Principles of care, except from the described changes in clinical practice, and senior medical staff remained unchanged during the study period. An overestimation of NSAIDs use during period 1 and goal-directed fluid therapy during period 2 cannot be excluded but would bias the presented estimates toward no effect. Preoperative irradiation has been identified in some studies as an independent risk factor for anastomotic leakage after low anterior resection [7, 36], while other studies have not [37]. The present study is based on a local quality database without information about neo-adjuvant treatment. A majority of the patients operated with low anterior resection were operated for rectal cancer (85%). Approximately 75% of these patients, irrespectively of period, received preoperative radiotherapy according to the national Swedish colorectal cancer register. Increased attention of involved surgeons to the problem of anastomotic leakage, especially at the time of the introduction of changes in clinical practice, may have improved the results in period 2 but cannot be quantified. Preoperative intervention for smoking cessation seems to reduce the risk for postoperative complications, but the policy regarding smoking cessation did not change during the study period [38]. Age, sex, type of resection, type of pathology, multivisceral resection, acute surgery and protecting ostomy had no important confounding effect in this study. It is not likely that other patient-related risk factors for anastomotic leakage, not assessed in this study, changed between period 1 and 2 to an extent that would explain the difference in leakage rate.

A subgroup of asymptomatic grade A leakages, commonly associated with late leakage, were not registered in this trial. They represent a minority of leakages that do not alter the clinical course during the index hospitalization, but are relevant as they may prevent or delay the subsequent closure of a protective ostomy, or impair functional outcome after low anterior resection.

The internal validity of the presented findings is robust as the limitations discussed above do not suggest that the observed reduction in anastomotic leakage can be attributed to other factors than the described changes in clinical practice. The external validity might be restricted to some extent as the study population had more multivisceral resections than the general population undergoing colorectal resections and the proportion of minimally invasive procedures was very small. However, from a global perspective open surgery is still the dominating surgical method.