Predicting response to topical non-steroidal anti-inflammatory drugs in osteoarthritis: an individual patient data meta-analysis of randomized controlled trials.

Objectives: To identify predictors of the specific (difference between treatment and placebo) and overall (change from baseline in treatment arm) treatment effects of topical NSAIDs in OA.

Methods: Randomized controlled trials (RCTs) of topical NSAIDs in OA were identified through systematic literature searching and inquiry to pharmaceutical companies. The raw, de-identified data were analysed in one-stage individual patient data meta-analysis (IPD-MA). Negative values for treatment effects (0-100 scale) indicate pain reduction.

Results: Of 63 eligible RCTs, 15 provided IPD (n = 1951 on topical NSAID), including 11 placebo-controlled RCTs (n = 1587 on topical NSAIDs, 1553 on placebo). Seven potential predictors of response were examined. Topical NSAIDs were superior to placebo [-6 (95% CI -9, -4)], with a small, but statistically significant greater effect in women than men [difference -4 (95% CI -8, -1)]. The overall treatment effect was 4-fold larger than the specific effect [-25 (95% CI -31, -19)] and increased with greater baseline pain severity (P < 0.001). No differences in efficacy were observed for age, BMI, features of inflammation, duration of complaints or radiographic OA severity.

Conclusion: Topical NSAIDs are effective for OA pain relief. Greater overall pain relief in individuals with more baseline pain might be due to contextual and non-specific effects, including regression to the mean. Additional factors that have been linked either mechanistically or through empirical evidence to outcomes should be selected for inclusion across future RCTs in order to facilitate the identification of response predictors through IPD-MA.