Endoscopic retrograde cholangiopancreatography (ERCP) has been used in patients referred to cholecystectomy when clinical information, biochemical values, or ultrasonography (clinical characterization) have indicated possible presence of common bile duct stones. A retrospective study of 599 patients treated for gallstone disease was used to develop a characterization procedure for predicting common bile duct stones by a discriminant analysis procedure. The variables selected by the analysis as the best combination for CBDS prediction were age (years), the values of bilirubin (micromol/l), ALAT (U/l) and gamma GT (U/l). The characterization was false positive in 22 cases (3.7%) and false negative in 11 cases (1.8%), compared to 198 false positive cases (33.1%) and three false negative cases (0.5%) by the clinical characterization. A leaving-one-out correction did not change the results. In a test set of 157 cholecystectomy patients, clinical characterization was false positive in 44.6% of the patients, compared to 4.5% false positive results when using the discriminant analysis procedure. The discriminant analysis procedure would have missed one patient with common bile duct stones. Selection by the discriminant analysis characterization procedure seems to reduce the frequency of preoperative ERCP significantly without an increase in undetected common bile duct stones.