Radiological glenohumeral osteoarthritis in long-term type 1 diabetes. Prevalence and reliability of three classification systems. The Dialong shoulder study

Objective In the present study, we evaluate the intra- and interrater agreement of radiological glenohumeral OA using three different classification systems and estimate the prevalence of radiological and clinical glenohumeral OA in patients with type 1 diabetes mellitus (DM1), for over 45 years and controls (The Dialong study). Materials and methods We included 102 patients with DM1 (49% women, mean age, 61.9 years) and 73 controls (57% women, mean age, 62.6 years). Anterior-posterior shoulder radiographs were interpreted by two observers applying the Kellgren-Lawrence (K-L), Samilson-Prieto (S-P) and Samilson-Prieto Allain (S-PA) classifications. Results The interrater agreement was moderate (weighted kappa, 0.46 to 0.48) for all classifications and the intrarater agreement mainly substantial (0.48-0.86) for both observers. The agreed prevalence of radiological OA was 26 and 18% (OR 1.6 (0.8 to 3.3), p = 0.22, 44 and 26% (OR 2.2 (1.2 to 4.2), p = 0.02) and 30 and 17% (OR 2.1 (1.0 to 4.5), p = 0.05) for the K-L, S-P and S-PA classifications respectively in the diabetes and control groups. The prevalence of moderate or severe radiological OA was 1 to 6% and clinical OA 1 to 2% with no difference between the groups. Conclusions The prevalence of radiological glenohumeral OA was higher in the diabetes group with the Samilson-Prieto classification systems, but not associated with clinical OA. The interrater agreement was moderate. We recommend the Samilson-Prieto Allain classification for glenohumeral OA to avoid interpretation of osteophytes <1 mm as OA in patient groups with a low pre-test likelihood of glenohumeral OA.