Objectives: To determine which classification is more accurate in stratifying severity.
Methods: The study used a retrospective analysis of a prospective acute pancreatitis database (June 2005-December 2007). Acute pancreatitis severity was stratified according to the Atlanta classification (AC) 1992, the revised Atlanta classification (RAC) 2012, and the determinant-based classification (DBC) 2012. Receiver operating characteristic analysis (area under the curve) compared the accuracy of each classification. Logistic regression identified predictors of mortality.
Results: 338 patients were analyzed: 13% had persistent organ failure (POF) (>48 hours), of whom 37% had multisystem POF, and 11% had pancreatic necrosis, of whom 19% had infected necrosis. Mortality was 4.1%. For predicting mortality (area under the curve), the RAC (0.91) and DBC (0.92) were comparable (P = 0.404); both outperformed the AC (0.81) (P < 0.001). For intensive care unit admission, the RAC (0.85) and DBC (0.85) were comparable (P = 0.949); both outperformed the AC (0.79) (P < 0.05). There were 2 patients in the critical category of the DBC. Multisystem POF was an independent predictor of mortality (odds ratio, 75.0; 95% confidence interval, 13.7-410.6; P < 0.001), whereas single-system POF, sterile necrosis, and infected necrosis were not.
Conclusion: The RAC and DBC were generally comparable in stratifying severity. The paucity of patients in the critical category in the DBC limits its utility. Neither classification accounts for the impact of multisystem POF, which was the strongest predictor of mortality.