RESUMO
AIMS: Gallbladders resected for non-neoplastic diseases are systemically examined microscopically to rule out incidental dysplasia and carcinoma. The main aim of this study was to test whether a pre-grossing algorithm can detect incidental gallbladder carcinoma. The secondary aim was to test whether the algorithm can detect high-grade dysplasia. METHODS AND RESULTS: A retrospective study of clinical, pathological and radiological findings in cholecystectomy recipients was performed on a test set to develop a classification and regression tree algorithm. Cholecystectomy cases were included; exclusion criteria were age <18 years, missing pathology reports, preoperative suspicion of neoplastic disease, and cholecystectomy for non-gallbladder oncological disease. Five thousand nine hundred and eighty-two cholecystectomies from 2006 to 2018 were included in the study, with 18 cases of incidental gallbladder carcinoma and 11 cases of high-grade dysplasia. Three hundred and ninety controls were randomly selected for the testing set. Patient age, surgical approach, operation duration, dilatation of the biliary tract and gallbladder gross anomalies were statistically significant distinguishing factors in multivariate analysis (P < 0.00-0.026). Unsupervised testing with a conditional inference tree suggested that age, procedure type and operation duration can be used to identify incidental gallbladder carcinoma from controls, whereas high-grade dysplasia also requires grossing parameters to identify half of the cases (5/11). CONCLUSION: Readily available clinical parameters and postoperative data can be used to detect incidental gallbladder carcinoma. High-grade dysplasia mostly requires grossing and microscopic examination.