Score prediction of anastomotic leak in colorectal surgery: a systematic review

Predicting the risk of anastomotic leak (AL) is of importance when defining the optimal surgical strategy in colorectal surgery. Our objective was to perform a systematic review of existing scores in the field.

We followed the PRISMA checklist (S1 Checklist). Medline, Cochrane Central and Embase were searched for observational studies reporting on scores predicting AL after the creation of a colorectal anastomosis. Studies reporting only validation of existing scores and/or scores based on post-operative variables were excluded. PRISMA 2020 recommendations were followed. Qualitative analysis was performed.

Eight hundred articles were identified. Seven hundred and ninety-one articles were excluded after title/abstract and full-text screening, leaving nine studies for analysis. Scores notably included the Colon Leakage Score, the modified Colon Leakage Score, the REAL score, www.anastomoticleak.com and the PROCOLE score. Four studies (44.4%) included more than 1.000 patients and one extracted data from existing studies (meta-analysis of risk factors). Scores included the following pre-operative variables: age (44.4%), sex (77.8%), ASA score (66.6%), BMI (33.3%), diabetes (22.2%), respiratory comorbidity (22.2%), cardiovascular comorbidity (11.1%), liver comorbidity (11.1%), weight loss (11.1%), smoking (33.3%), alcohol consumption (33.3%), steroid consumption (33.3%), neo-adjuvant treatment (44.9%), anticoagulation (11.1%), hematocrit concentration (22.2%), total proteins concentration (11.1%), white blood cell count (11.1%), albumin concentration (11.1%), distance from the anal verge (77.8%), number of hospital beds (11.1%), pre-operative bowel preparation (11.1%) and indication for surgery (11.1%). Scores included the following peri-operative variables: emergency surgery (22.2%), surgical approach (22.2%), duration of surgery (66.6%), blood loss/transfusion (55.6%), additional procedure (33.3%), operative complication (22.2%), wound contamination class (1.11%), mechanical anastomosis (1.11%) and experience of the surgeon (11.1%). Five studies (55.6%) reported the area under the curve (AUC) of the scores, and four (44.4%) included a validation set.

Existing scores are heterogeneous in the identification of pre-operative variables allowing predicting AL. A majority of scores was established from small cohorts of patients which, considering the low incidence of AL, might lead to miss potential predictors of AL. AUC is seldom reported. We recommend that new scores to predict the risk of AL in colorectal surgery to be based on large cohorts of patients, to include a validation set and to report the AUC.