Inter-observer agreement for multichannel intraluminal impedance-pH testing

Twenty-four-hour ambulatory multichannel intraluminal impedance (MII)-pH detects both acid and nonacid reflux (NAR). A computer-based program (Autoscan™, Sandhill Scientific, Highlands Ranch, CO, USA) automates the detection of reflux episodes, increasing the ease of study interpretation. Inter-observer agreement between multiple reviewers and with Autoscan™ for the evaluation of significant NAR with MII-pH has not been studied in the adult population. Twenty MII-pH studies on patients taking a proton pump inhibitor twice daily were randomly selected. Autoscan™ analyzed all studies using the same pre-programmed parameters. Four reviewers interpreted the MII-pH studies, adding or deleting reflux episodes detected by Autoscan™. Positive studies for NAR and total reflux episodes were based on published criteria. Cohen's kappa statistic (κ) evaluated inter-observer agreement between reviewers and Autoscan™ analysis. The average κ for pathologic NAR between reviewers was 0.57 (0.47-0.70), and between reviewers and Autoscan™ was 0.56 (0.4-0.8). When using the total reflux episode number as a marker for pathologic reflux (acid and NAR), the κ score was 0.72 (0.61-0.89) between reviewers, and 0.74 (0.53-0.9) when evaluating total reflux episodes. Two reviewers agreed more often with each other and with Autoscan™ on the number of NAR episodes, while the other two reviewers agreed with each other, but did not agree with either Autoscan™ or the first two reviewers. Inter-observer agreement between reviewers and Autoscan™ for detecting pathologic NAR is moderate, with reviewers either excluding more of the Autoscan™-defined events or excluding fewer events and therefore agreeing with Autoscan™.