DEVELOPMENT AND PRELIMINARY VALIDATION OF A SIMPLIFIED ENDOSCOPIC ULTRASOUND (EUS) SCORE FOR DIAGNOSIS OF CHRONIC PANCREATITIS (CP)

Abstract

Introduction: The current EUS CP scoring system includes 9 equally-weighted criteria. However, some criteria have higher interobserver variability (IOV) or carry more diagnostic value than others. Aim: To develop and validate a simplified EUS CP scoring system. Methods: Consecutive pts evaluated from 4/06 to 3/08 for CP underwent a combined EUS and secretin endoscopic pancreatic function test (ePFT). Clinical EUS scoring (9 criteria) was performed before ePFT results were available. Logistic regression was used to develop a simplified EUS score with abnl ePFT as the outcome (peak bicarbonate <80 mEq/L). In model development, some criteria were eliminated due to high IOV or a low diagnostic significance ranking based on expert consensus (Wallace, GIE 1999). A variable selection algorithm was also employed to generate a parsimonious model. The area under the receiver operating characteristics (ROC) curve was used to compare models. From 4/08 to the present, more pts are being recruited for validation of the simplified score. Results: In the development group (n = 200), 73 (37%) had an abnl ePFT. The final prediction model contained 4 criteria: calcifications, lobularity, duct irregularity, and visible side-branches. The 4-criteria score had similar discrimination as the 9-criteria score (area under ROC curve 0.80 for both). Weighted and non-weighted 4-criteria scores had similar discrimination. A cutpoint of ≥2 criteria optimized the non-weighted 4-criteria score. In the validation group (n = 20), 6 (30%) had an abnl ePFT. The 4-criteria score (cutpoint ≥2 criteria) was 83% sensitive and 93% specific. The 9-criteria score (cutpoint ≥4 criteria) was 67% sensitive and 93% specific. Conclusions: A simplified 4-criteria score may have similar diagnostic performance as the standard 9-criteria system. More patients are being recruited into the validation group.