Clinical prediction model for anastomotic leakage in rectal cancer surgery: a single-center large-sample cohort study.

BACKGROUND

Anastomotic leakage (AL) is a serious complication of surgery for rectal cancer. This study aimed to develop a reliable clinical prediction model based on the confirmed risk factors related to AL during the perioperative period.

METHOD: A retrospective analysis was conducted on 3765 patients. The preoperative, operative, and tumor risk factors affecting AL were first analyzed using univariate and multivariate logistic regression analyses. The results were used to construct clinical prediction models and the nomogram, and the reliability of the models was verified internally.

RESULTS

In total, 241 (6.4%) patients experienced AL. Multivariate logistic regression analysis showed that gender, diabetes mellitus, preoperative chemoradiotherapy, drinking history, distance from the anastomotic site to the anal margin, surgical time, blood loss, distance from the tumor to the anal margin, T stage, N stage, and tumor size were the independent risk factors for AL. In the internal validation training set, the area under the curve (AUC) values of prediction Models 1 and 2 were 0.818 and 0.815, respectively; in the validation set, the AUC values of Models 1 and 2 were 0.767 and 0.759, respectively. Calibration analysis showed Brier values of 0.053 and 0.056 in the training and validation set without statistical difference on the Hosmer-Lemeshow test. Clinical applicability curve results exhibited a high degree of consistency between the predicted results and clinical outcomes.

CONCLUSION

The clinical prediction model constructed in this study, which is based on perioperative risk factors related to AL, has high reliability and practicality, and can be used for perioperative risk assessment of AL in rectal cancer patients.