Reliability of radiologic glenohumeral osteoarthritis classifications.

Abstract

This study was conducted to determine the intraobserver and interobserver reliability of glenohumeral osteoarthritis classification schemes. The plain radiographs of 108 shoulder joints from 96 consecutive patients with glenohumeral osteoarthritis treated at our institution were graded into 6 classification systems by 2 observers on 2 occasions. The intraobserver and interobserver reliability of the classification schemes were determined by using Cohen's κ coefficient and weighted according to Landis and Koch. The intraobserver and interobserver reliabilities were 0.907 (observer 1), 0.965 (observer 2), and 0.851 (interobserver) for the Samilson-Prieto grading system; 0.954, 0.948, and 0.869 for the Allain modification; 0.936, 0.830, and 0.791 for the Gerber modification; 0.887, 0.892, and 0.744 for the Kellgren and Lawrence classification; 0.873, 0.964, and 0.935 for the Weinstein; and 0.854, 0.934, and 0.797 for the Guyette grading system. The classification systems according to Weinstein and Guyette and the modifications of the Samilson-Prieto classification according to Allain and Gerber showed a comparable reliability with the commonly used glenohumeral osteoarthritis grading systems according to Samilson-Prieto and Kellgren-Lawrence. Thus, they are recommended for clinical and especially scientific purposes.