Introduction: Treatment response in randomized clinical trials (RCT) of osteoarthritis (OA) has been assessed by multiple primary and secondary outcomes, including pain, function, patient and clinician global measures of status and response to treatment, and various composite and responder measures. Identifying outcome measures with greater responsiveness to treatment is important to increase the assay sensitivity of RCTs.

Objective: To assess and compare the responsiveness of different outcome measures used in placebo-controlled RCTs of OA.

Search strategy: The Resource for Evaluating Procedures and Outcomes of Randomized Trials database includes placebo-controlled clinical trials of pharmacologic treatments (oral, topical, or transdermal) for OA identified from a systematic literature search of RCTs published or publicly available before August 5, 2009, which was conducted using PubMed, the Cochrane collaboration, publicly-available websites, and reference lists of retrieved publications.

Data collection and analysis: Data collected included: (1) pain assessed with single-item ratings and the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain subscale; (2) patient and clinician global measures of status, improvement, and treatment response; (3) function assessed by the WOMAC function subscale; (4) stiffness assessed by the WOMAC stiffness subscale; and (5) the WOMAC and Lequesne Algofunctional Index composite outcomes. Measures were grouped according to the total number of response categories (i.e., <10 categories or ≥10 categories). The treatment effect (difference in mean change from baseline between the placebo and active therapy arms) and standardized effect size (SES) were estimated for each measure in a meta-analysis using a random effects model.

Results: There were 125 RCTs with data to compute the treatment effect for at least one measure; the majority evaluated non-steroidal anti-inflammatory drugs (NSAIDs), followed by opioids, glucosamine and/or chondroitin, and acetaminophen. In general, the patient-reported pain outcome measures had comparable responsiveness to treatment as shown by the estimates of treatment effects and SES. Treatment effects and SESs were generally higher for patient-reported global measures compared with clinician-rated global measures but generally similar for the WOMAC and Lequesne composite measures.

Conclusions: Comparing different outcome measures using meta-analysis and selecting those that have the greatest ability to identify efficacious treatments may increase the efficiency of clinical trials of treatments for OA. Improvements in the quality of the reporting of clinical trial results are needed to facilitate meta-analyses to evaluate the responsiveness of outcome measures and to also address other issues related to assay sensitivity.