OBJECTIVE

To validate and assess clinical efficacy of a prognostic model for predicting severe acute pancreatitis (SAP) based on inflammatory markers (IL-6, ΔIL-22), thromboelastography parameters (K-time) and the BISAP score.

MATERIAL AND METHODS

A prospective observational cohort study enrolled 181 patients with acute pancreatitis. Serum IL-6 and IL-22 were measured in 24 and 48 hours after clinical manifestation, respectively. Clotting status was evaluated using the TEG 5000 system. Clinical risk was assessed using the BISAP score, and the previously developed prognostic model was applied to estimate the risk of SAP. Statistical analysis included ROC curves, AUC, calibration assessment with isotonic regression, bootstrap analysis for confidence interval estimation and Decision Curve Analysis (DCA) to evaluate clinical utility.

RESULTS

SAP developed in 30 (16.6%) patients. The model demonstrated high discriminative power with AUC 0.914. Bootstrap analysis yielded AUC 0.935 (95% CI: 0.890—0.970) and Brier Score 0.065 (95% CI: 0.044—0.086). At baseline risk threshold, accuracy was 84.5%, sensitivity 93.3%, specificity 82.8% and positive predictive value 51.9%. After threshold optimization, accuracy increased to 87.8%, specificity to 86.8% and PPV to 58.3%, while sensitivity remained the same (93.3%). Calibration curves after isotonic regression indicated better concordance between the predicted probabilities and actual outcomes. DCA identified an optimal risk threshold range 0.3—0.6, where the model provided maximum clinical benefit.

CONCLUSION

Our data indicate significant value of the model for early detection of patients with high risk of SAP. Integration of biomarkers, hemostasis parameters and clinical assessments significantly increases the accuracy of risk stratification. This can contribute to optimized treatment and reduction of complications.