Nomogram for prediction of adverse events after lumen-apposing metal stent placement for drainage of pancreatic fluid collections.

Abstract

To generate a prognostic model based on a nomogram for adverse event (AE) prediction after lumen-apposing metal stents (LAMS) placement in patients with pancreatic fluid collections (PFC). Data from a large multicenter series of PFCs treated with LAMS placement were retrieved. AE (overall and excluding mild events) prediction was calculated through a logistic regression model and a nomogram was created and internally validated after bootstrapping. Results were expressed in terms of odds ratio (OR) and 95% confidence interval (CI). Discrimination was assessed by c-statistics and calibrated by comparing deciles of predicted and observed ORs. Overall, 516 patients were included (males 68%, mean age 61.6 ± 15.2 years). PFCs were predominantly walled-off necrosis (52.1%). Independent predictors of AE occurrence were injury of main pancreatic duct (OR in the case of leak 2.51, 95% CI 1.06-5.97, P = 0.03; OR in the case of complete disruption 2.61, 1.53-4.45, P = 0.01), abnormal vessels (OR in the case of perigastric varices 2.90, 1.31-6.42, P = 0.008; OR in the case of pseudoaneurysm 2.99, 1.75-11.93, P = 0.002), using a multigate technique (OR 3.00, 1.28-5.24; P = 0.05), and need of percutaneous drainage (OR 2.81, 1.03-7.65, P = 0.04). By nomogram, a score beyond 200 points corresponded to a 50% probability of AE occurrence. The model was confirmed even when excluding mild AEs and it showed optimal discrimination (c-index 76.8%, 95% CI 74-79), confirmed after internal validation. Patients with preprocedural evidence of pancreatic duct leak/disruption, vessel alteration, requiring percutaneous drainage or a multigate technique are at higher risk for AE.