Bowel preparation for elective colorectal resection: multi-treatment machine learning analysis on 6241 cases from a prospective Italian cohort.

BACKGROUND:Current evidence concerning bowel preparation before elective colorectal surgery is still controversial. This study aimed to compare the incidence of anastomotic leakage (AL), surgical site infections (SSIs), and overall morbidity (any adverse event, OM) after elective colorectal surgery using four different types of bowel preparation. METHODS:A prospective database gathered among 78 Italian surgical centers in two prospective studies, including 6241 patients who underwent elective colorectal resection with anastomosis for malignant or benign disease, was re-analyzed through a multi-treatment machine-learning model considering no bowel preparation (NBP; No. = 3742; 60.0%) as the reference treatment arm, compared to oral antibiotics alone (oA; No. = 406; 6.5%), mechanical bowel preparation alone (MBP; No. = 1486; 23.8%), or in combination with oAB (MoABP; No. = 607; 9.7%). Twenty covariates related to biometric data, surgical procedures, perioperative management, and hospital/center data potentially affecting outcomes were included and balanced into the model. The primary endpoints were AL, SSIs, and OM. All the results were reported as odds ratio (OR) with 95% confidence intervals (95% CI). RESULTS:Compared to NBP, MBP showed significantly higher AL risk (OR 1.82; 95% CI 1.23-2.71; p = .003) and OM risk (OR 1.38; 95% CI 1.10-1.72; p = .005), no significant differences for all the endpoints were recorded in the oA group, whereas MoABP showed a significantly reduced SSI risk (OR 0.45; 95% CI 0.25-0.79; p = .008). CONCLUSIONS:MoABP significantly reduced the SSI risk after elective colorectal surgery, therefore representing a valid alternative to NBP.