Development and validation of a predictive model for choledocholithiasis.

Background: Gallstone-related conditions affect a significant portion of the population, with varying prevalence among different ethnic groups. Complications such as pancreatitis and cholangitis are associated with the presence of common bile duct (CBD) stones. Existing guidelines for diagnosing choledocholithiasis lack precision, leading to excessive use of invasive procedures like endoscopic retrograde cholangiopancreatography (ERCP).

Methods: A prospective study was conducted at Hospital Central "Dr. Ignacio Morones Prieto," involving 374 patients in the development cohort and 154 patients in the validation cohort. Patients meeting inclusion criteria underwent biochemical testing and ultrasonography. A predictive scoring system was developed using logistic regression and validated in an independent cohort. Clinical and laboratory variables were collected, and model performance was assessed using receiver-operator characteristic (ROC) curves.

Results: The predictive model incorporated variables such as age, pancreatitis, cholangitis, bilirubin levels, and CBD stone presence on ultrasound. The model demonstrated an area under the ROC curve (AUC) of 93.81% in the validation dataset. By adjusting the threshold defining high-risk probability to 40%, the model improved specificity and sensitivity compared to existing guidelines. Notably, the model reclassified patients, leading to a more accurate risk assessment.

Conclusions: The developed algorithm accurately predicts choledocholithiasis non-invasively in patients with symptomatic gallstones. This tool has the potential to reduce reliance on costly or invasive procedures like magnetic resonance cholangiopancreatography and ERCP, offering a more efficient and cost-effective approach to patient management. The user-friendly calculator developed in this study could streamline diagnostic procedures, particularly in resource-limited healthcare settings, ultimately improving patient care.