Increasing trend of radiographic features of knee osteoarthritis in rheumatoid arthritis patients before total knee arthroplasty.

To investigate the trend and factors related to the occurrence of osteoarthritis (OA)-like features on knee radiographs of rheumatoid arthritis (RA) patients undergoing total knee arthroplasty (TKA) in the recent decades. To classify antero-posterior knee radiographs into 'RA' and 'OA-like RA' groups, a deep learning model was developed by training the network using knee radiographs of end-stage arthropathy in RA patients obtained during 2002-2005 and in primary OA patients obtained during 2007-2009. We used this model to categorize 796 knee radiographs, which were recorded in RA patients before TKA during 2006-2020, into 'OA-like RA' and 'RA' groups. The annual ratio of 'OA-like RA' was investigated. Moreover, univariate and multivariate analyses were performed to identify the factors associated with the classification as OA-like RA using clinical data from 240 patients. The percentage of 'OA-like RA' had significant increasing trend from 20.9% in 2006 to 67.7% in 2020. Higher body mass index, use of biologics, and lower level of C-reactive protein were identified as independent factors for 'OA-like RA'. An increasing trend of knee radiographs with OA-like features was observed in RA patients in the recent decades, which might be attributed to recent advances in pharmacotherapy.