Development and evaluation of a Japanese prediction model for low anterior resection syndrome after rectal cancer surgery

Abstract

BACKGROUND: Low anterior resection syndrome (LARS) is the most common complication after rectal cancer resection. We aimed to identify LARS' predictive factors and construct and evaluate a predictive model for LARS. METHODS: This retrospective study included patients with rectal cancer more than 1 year after laparoscopic or robotic-assisted surgery. We administered a questionnaire to evaluate the degree of LARS. In addition, we examined clinical characteristics with univariate and multivariate analysis to identify predictive factors for major LARS. Finally, we divided the obtained data into a learning set and a validation set. We constructed a predictive model for major LARS using the learning set and assessed the predictive accuracy of the validation set. RESULTS: We reviewed 160 patients with rectal cancer and divided them into a learning set (n = 115) and a validation set (n = 45). Univariate and multivariate analyses in the learning set showed that male (odds ratio [OR]: 2.88, 95% confidence interval [95%CI] 1.11-8.09, p = 0.03), age < 75 years (OR: 5.87, 95%CI 1.14-47.25, p = 0.03) and tumors located < 8.5 cm from the AV (OR: 7.20, 95%CI 2.86-19.49, p < 0.01) were significantly related to major LARS. A prediction model based on the patients in the learning set was well-calibrated. CONCLUSIONS: We found that sex, age, and tumor location were independent predictors of major LARS in Japanese patients that underwent rectal cancer surgery. Our predictive model for major LARS could aid medical staff in educating and treating patients with rectal cancer before and after surgery.