Nonsteroidal antiinflammatory drugs (NSAIDs) are used to alleviate osteoarthritis (OA) symptoms and are not considered disease-modifying. We studied NSAID use in subjects with hip or knee OA to investigate the relationship between NSAIDs and OA severity.
Subjects (seen over 2 years at teaching hospitals in London, Ontario) were referred to orthopedic surgeons or rheumatologists, had radiographs available, but had no inflammatory arthritis or evidence of secondary OA. Subjects were divided into mild, moderate (mod) and severe OA by X-ray (worst joint selected). A questionnaire was mailed asking about current/ever use of specific NSAIDs (by generic and trade names) including over the counter and acetylsalicylic acid. Coxibs were not available over the study period. A total of 608 subjects were studied, having mild (71), mod (129) and severe (408) OA, with a disease duration of 10 years in each group (62% were women, who had proportionately more mild disease than men).
Those with severe versus mild OA were older (70 versus 61 years), so age was adjusted for. There was an inverse dose–response with use of > 3 NSAIDs being used by mild versus severe OA (odds ratio 3.7), mild versus mod OA (odds ratio 2.5) and mod versus severe OA (odds ratio 1.5) (P < 0.0001), and the same was found with > 2 NSAIDs (P < 0.0002); and 7% of mild, 15% of mod and 12% of severe OA had received no NSAIDs ever. Only one NSAID (ever used) was increased in severe OA (sulindac, P < 0.008) compared with naproxen, tiaprofenic acid and diclofenac, more frequently used in mild than mod or severe OA (P < 0.01). Other NSAID use ever was not different between the groups (including indomethacin, thought to be chondrodestructive), but numbers using indomethacin were small. Limitations may include inability to study accurately the duration of NSAID use, older age in those with severe OA (which could affect NSAID treatment choice), recall bias and unknown confounders. We did not illicit body mass index or pain level in each group.
There may be biases for referring to different specialists as the rheumatologists prescribed NSAIDs more often and surgeons treated more subjects with severe OA (P < 0.0001). When analyses were stratified by specialist type, there were no differences between mild, mod and severe OA for NSAID use in subjects treated by rheumatologists; however, the trend for increased NSAID use with milder OA was evident in those seen by orthopedic surgeons. When stratified by knee and hip OA, the dose–response remained (P < 0.003 and P < 0.03, respectively). Subjects were followed by general practitioners, so we assumed past NSAIDs should have been prescribed equally in all groups, but this was not the case.
Those with more severe OA are more rapidly progressive (having similar disease duration to the other groups) and could be innate NSAID nonresponders. However, one would assume that they should be exposed to more NSAIDs (looking for one to be helpful). We conclude that NSAID use and the number of NSAIDs used was greater in mild radiographic OA and postulate that lowering prostaglandins in the joint with NSAIDs may be protective for cartilage. Prospective studies are needed to confirm observations. Some in vitro models of NSAIDs in OA support our observations.