Abstract
High-resolution manometry (HRM) is increasingly used to assess risk of gastroesophageal reflux disease (GERD). The Milan Score (MS) integrates four HRM variables-EGJ morphology, EGJ contractile integral (EGJ-CI), ineffective esophageal motility (IEM), and straight leg raise (SLR) response-to predict pathological esophageal acid exposure. The inter-observer reproducibility of the MS and its components has never been formally assessed.
A total of 50 consecutive HRM plots from patients undergoing evaluation for suspected GERD at a tertiary referral center were distributed to a multicenter pool of esophageal experts for blind review. Participants assessed EGJ type, EGJ-CI, IEM, and SLR response; the MS and corresponding GERD risk were automatically calculated. Inter-observer variability was evaluated using Fleiss' κ for categorical variables and Kendall's coefficient for continuous variables. MS accuracy was assessed by comparing calculated scores against MII-pH results.
Twenty-three esophageal experts from 16 international institutions reviewed all 50 plots. Agreement was fair for EGJ morphology (κ = 0.321), moderate for SLR (W = 0.531), and good for IEM (κ = 0.638), EGJ-CI (W = 0.675) and the final MS (W = 0.641). Median plot-level concordance between MS and pH-impedance study was 71.74% [IQR 51.08%]; median reader accuracy for Milan Score was 64% [IQR 12%], ranging from 54% to 80%.
Inter-observer variability of the Milan Score is good across a diverse international cohort of esophageal experts, supporting its reproducibility as a clinical tool. However, agreement for individual parameters was heterogeneous, and the assessment of EGJ morphology and the SLR maneuver achieved only fair-to-moderate agreement, highlighting the need for improved standardization.