Development of an online app to predict post-endoscopic retrograde cholangiopancreatography adverse events using a single center retrospective cohort.

2021 
INTRODUCTION Endoscopic retrograde cholangiopancreatography (ERCP) is a technically demanding procedure with a high risk for adverse events (AEs). AIM Evaluate patient- and procedure-related risk factors for ERCP-related AEs and develop an online app to estimate risk of AEs. METHODS Retrospective study of 1491 consecutive patients who underwent 1991 ERCPs between 2012 and 2017. AEs definition and severity were classified according to most recent ESGE guidelines. Each variable was tested for association with occurrence of overall AEs, post-ERCP pancreatitis (PEP) and cholangitis. For each outcome, two regression models were built, from which an online Shiny-based app was created. RESULTS Overall AE rate was 15.3%; in 19 procedures, more than one AE occurred. Main post-ERCP AE was PEP (7.5%), followed by cholangitis (4.9%), bleeding (1.3%), perforation (1%), cardiopulmonary events (0.9%) and cholecystitis (0.3%). Seventy-eight percent of AEs were mild/moderate; of severe (n=55) and fatal (n=20) AEs more than half were related to infection, cardiac/pulmonary AEs and perforation. AE-related mortality rate was 1%. When testing pre-cannulation, procedural co-variates and ERCP findings, AE occurrence was associated with age (OR 0.991), previous PEP (OR 2.198), ERCP complexity grade III/IV (OR 1.924), standard bile duct cannulation (OR 0.501), sphincterotomy (OR 1.441), metal biliary stent placement (OR 2.014), periprocedural bleeding (OR 3.024) and biliary duct lithiasis (OR 0.673). CONCLUSION Our app may allow an optimization of the patients' care, by helping in the process of decision making, not only regarding patient or endoscopist's selection but also definition of an adequate and tailored surveillance plan after the procedure.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    18
    References
    0
    Citations
    NaN
    KQI
    []