The SF-36 Health Survey as a generic outcome measure in clinical trials ofpatients with osteoarthritis and rheumatoid arthritis - Tests of data quality, scaling assumptions and score reliability

OBJECTIVE. TO evaluate the psychometric assumptions underlying the construc tion and scoring of SF-36 scales and summary measures among clinical trial participants with arthritis. METHODS. Cross-sectional SF-36 data from the baseline assessment of adult p atients (n = 1,016) participating in four placebo-controlled clinical trial s of treatment for arthritis were analyzed with blinding as to treatment. T ests of the completeness of data, scaling assumptions, internal-consistency reliability, and factor structure of SF-36 scales were performed for the c ombined sample. Eligible participants had at least a 6-month history of mod erate to severe osteoarthritis or rheumatoid arthritis of the knee or hip. Participants meeting inclusion criteria had undergone a washout period of 3 -14 days before baseline assessment to bring about a flare state in osteoar thritis or rheumatoid arthritis symptoms. Baseline sample sizes for the thr ee osteoarthritis trials were n = 121, n = 342 and n = 187. The baseline sa mple size for the rheumatoid arthritis trial was n = 367. The average age o f participants was 60 years, and the majority were females (72%). Measured were functional health and well-being scales and physical and mental health summary measures from the SF-36 Health Survey acute form. RESULTS. Missing responses ranged from 0.0% to 1.5% across SF-36 items, and scale scores could be computed for 96.8% to 100% of participants across tr ials. In all four trials, item internal consistency tests were passed (91.4 %-97.1%) and item discriminant validity tests were passed (96.9%-100.0%). A cross the four trials, internal-consistency reliability coefficients ranged from a low of 0.75 to a high of 0.91 for the eight scales (median = 0.84), exceeding the minimum standards for group comparisons. Ceiling effects wer e minimal for most scales, and floor effects were noteworthy for the role p hysical and role emotional scales. Physical and mental health factors ident ified in previous studies were replicated. CONCLUSION. The SF-36 Health Survey proved to be a psychometrically sound t ool for the assessment of the health status of adult participants in clinic al trials of arthritis.