Preoperative prediction model for 30 days biliary events after laparoscopic cholecystectomy

Abstract

BACKGROUND

Laparoscopic cholecystectomy (LC) is the standard treatment for symptomatic cholelithiasis, but a subset of patients still develop clinically relevant biliary complications within 30 days, including retained common bile duct (CBD) stones, bile leak, cholangitis, or biliary pancreatitis that often require therapeutic endoscopic retrograde cholangiopancreatography (ERCP), drainage, or unplanned readmission. Quantitative, preoperative tools to estimate individual short-term biliary risk and support selective imaging or intervention remain limited.

AIM

To develop and internally validate a concise preoperative nomogram based on total bilirubin (TBil) and the CBD/height index for predicting 30-day clinically relevant biliary events after LC.

METHODS

In this single-center retrospective cohort, consecutive adult patients undergoing LC between 2022 and 2024 (n = 287) were included. Patients were randomly allocated to a training cohort (n = 200) and a validation cohort (n = 87) using stratified randomization by outcome (7:3 ratio). Preoperative candidate variables (demographics, clinical history, laboratory tests, and ultrasound findings) were entered into least absolute shrinkage and selection operator (LASSO) logistic regression for feature selection. Variables retained at the λ₁se penalty were further assessed by multivariable logistic regression to identify independent predictors and construct a two-variable prediction model, which was visualized as a nomogram. Model performance was evaluated in both cohorts using the area under the receiver operating characteristic curve (AUC) with 95%CI, Brier score, calibration-in-the-large, calibration slope, calibration plots, and decision curve analysis.

RESULTS

Clinically relevant biliary events within 30 days occurred in 29 of 287 patients (10.1%). At the λ₁se penalty, LASSO retained four variables with non-zero coefficients (TBil, gamma-glutamyl transferase, CBD/height index, and ultrasound-suspected CBD stones/obstruction). In multivariable analysis, only preoperative TBil and the CBD/height index remained independent predictors and were used to construct the final model. In the original random-split analysis, the AUC was 0.881 (95%CI: 0.777-0.986) in the training cohort and 0.902 (95%CI: 0.818-0.985) in the validation cohort. Bootstrap-based internal validation of the full-cohort two-variable model showed an apparent AUC of 0.892 and an optimism-corrected AUC of 0.890. In a sensitivity analysis excluding 54 patients who underwent ERCP within 30 days before surgery, the associations of TBil and the CBD/height index with the primary outcome remained materially similar, and model discrimination remained stable (AUC = 0.895).

CONCLUSION

A simple two-variable preoperative nomogram based on TBil and the CBD/height index can effectively predict 30-day clinically relevant biliary events after LC, with favorable discrimination, calibration, and clinical applicability. This tool may facilitate individualized risk communication and optimization of perioperative strategies, but requires prospective multicenter external validation.

Keywords: Laparoscopic cholecystectomy; Total bilirubin; Common bile duct/height index; Nomogram; Least absolute shrinkage and selection operator; Biliary events; Endoscopic retrograde cholangiopancreatography; Decision curve analysis; Prediction model

Core Tip: Clinically relevant biliary events within 30 days after laparoscopic cholecystectomy, such as retained common bile duct (CBD) stones, bile leak, cholangitis, and biliary pancreatitis, may necessitate endoscopic retrograde cholangiopancreatography (ERCP), drainage, or unplanned readmission. In this single-center retrospective cohort of 287 patients, 29 patients (10.1%) experienced the primary outcome. A parsimonious preoperative nomogram based on total bilirubin and a height-normalized CBD/height index showed robust discrimination in the original split-sample analysis, supportive bootstrap-based internal validation, and stable performance in a sensitivity analysis excluding patients who underwent preoperative ERCP.