Physicians' vs patients' global assessments of disease activity in rheumatology and musculoskeletal trials: a meta-research project with focus on reasons for discrepancies
Background: In most rheumatic and musculoskeletal diseases (RMDs), global assessments of disease activity by physicians and patients are 'anchor outcomes' in therapeutic trials evaluating whether a treatment is effective. Objectives: To compare physicians' vs patients' global assessments of disease activity in RMD trials and explore reasons for discrepancies between them. Methods: Eligible trials were sampled from systematic reviews of treatments for RMDs by using the Cochrane database of systematic reviews (i.e., reviews from the Cochrane Musculoskeletal Group, [CMSG]). Randomized controlled trials (RCTs) of interventions for RMDs were eligible if they reported quantitative analyses of both physicians´ and patients´ global assessments at the same time point for the comparison of the same experimental intervention against the same comparator (i.e., placebo, no treatment, or other treatment). We accepted data from trial comparisons for each type of outcome, regardless of the type of intervention and type of RMD within the CMSG. Using mixed-effects meta-regression models, we assigned the dependent variable as the ratio of odds ratios (ROR) of global change with the experimental intervention, versus the control comparator. An ROR>1 would indicate that physicians rated the experimental intervention more favorable than their patients did. Results: We were able to estimate the ROR (data from both physicians' and patients' global assessments) across 70 trials (116 randomized comparisons) in 7 diseases (ankylosing spondylitis, fibromyalgia, psoriatic arthritis, osteoarthritis, rheumatoid arthritis, systemic lupus erythematosus, and gout). The combined ROR across all effectiveness comparisons were rated significantly in favor of the intervention by physicians: ROR=1.15 CI 95% (1.07 to 1.23). This combined ROR was based on a substantial heterogeneity across comparisons (I2=89.1%). Across all the stratified analyses, the type of the RMD was an informative reason for discrepancies, with a statistically significant ROR in rheumatoid arthritis ROR=1.33, CI 95% (1.13 to 1.56), unlike the ROR in all other conditions (ROR=1.04, CI 95% (0.95-1.14). Conclusion: In comparative effectiveness research on rheumatology, physicians' global assessments of disease activity, surprisingly, are more in favor of the experimental interventions than are those of the patients.