Gastric myoelectrical activity based AI-derived threshold predicts resolution of gastroparesis post-pyloric balloon dilation.

Abstract

A patient subset with gastroparesis (GP) has normal gastric myoelectrical activity (GMA) and pyloric dysfunction. (1) To determine pyloric balloon dilation (BD) effect on symptoms and gastric emptying in GP patients with normal 3 cycles per minute (cpm) GMA. (2) To demonstrate GMA-based artificial intelligence (AI)-derived formulae predict BD success at 10-12-month follow-up. Cohort subjects completed baseline electrogastrogram w/water load satiety test (WLST), solid-phase nuclear gastric emptying, Gastrointestinal Cardinal Symptom Index (ANMS GCSI-DD) and Leeds questionnaires. Subjects were divided into two groups based on response to the WLST. Group 1 (n = 26) with hypernormal/normal 3cpm GMA and Group 2 (n = 4) hyponormal/normal range 3cpm GMA, compared to healthy normals. All subjects underwent endoscopic pyloric BD. After 10-12months, gastric emptying and dyspepsia questionnaires were repeated to evaluate outcomes. Group 1 ANMS GCSI-DD scores improved from 2 points at baseline (BL) to 0 at follow-up (f/u) (p < 0.001); Group 2 ANMS GSCI-DD scores were 2 at BL and 1.6 at f/u (p = 0.25). Leeds scores improved (p < 0.001) only for Group 1. Group 1 gastric emptying improved (54.5% retained at 2h at BL vs. 12.2% at f/u, p < 0.001) in contrast to Group 2 patients (51.25% at BL vs. 56.25% at f/u, p = 0.252). Percentage 3cpm GMA decreased (41.1% at BL vs. 24.9% at f/u, p ≤ 0.005) in Group 1 versus Group 2 (15.3% at BL vs. 23.4% at f/u, p = 0.114). AI-derived GMA threshold (GMAT) of 0.59 predicted positive pyloric BD outcomes at 10-12months with sensitivity 96%, specificity 75%, and 93% correct classification. Pyloric BD improved symptoms and gastric emptying long term in patients with GP and hypernormal/normal 3cpm GMA. AI-derived GMAT predicted pyloric BD success. GMA post-WLST and GMAT are objective measures for improved selection and outcomes for endoscopic pyloric BD.