Pediatric Rheumatology Online Journal → June 2003 → Health-Related Quality of Life, Disease Activity and Outcome Assessments→ Abstract #37
A META-ANALYSIS TO ESTIMATE THE
N. Ruperto,1 A. Pistorio,1 A. Martini,1 S. Simianer,2 R. Sigmund,2 G. Hanft.2
1Pediatria II, IRCCS G.Gaslini, Genova, Italy; 2Boehringer Ingelheim Pharma KG, Biberach, Germany
Background: The ethics of placebo-controlled trial, even in adults, has recently come under intense debate. Regulatory and health economy demands are best met by direct comparison of active drug against placebo. An adequate meta-analysis of previous placebo data in JRA could be a very helpful tool for designing ethically acceptable trials and optimising sample size
calculations.
Methods: All trials in the literature including placebo treatment of JRA were eligible. Key features were compared, evaluated and analysed. Selection criteria for inclusion in the meta-analysis were developed. Response rates and confidence intervals (CI) from included trials were calculated. Based on the variance in each trial, a weighted calculation was performed for the total
population.
Results: Response rates across trials showed two clusters: recent trials used composite scores (CS) and had lower response rates while older trials, used single one-dimensional (OD) Likert scale. Re-analysis of trials using OD criteria with CS had yielded much lower response rates; for example in a trial using auranofin the OD placebo response rate dropped from 61.1% to 32.9 % with a CS. Inherent multi-dimensionality made CS generally more preferable. Therefore, the current meta-analysis included all trials in JRA using CS. Six placebo controlled trials with a total of 85/323 placebo responders were included. Weighted mean observation time was 5.85 months. Placebo response rates ranged from 8.8% to 35.9%. The pooled estimate of the response rate was 28.9% (CI: 24.0; 34.2).
Conclusions: Trials using OD had much higher placebo response rates. The range of response rates across trials is acceptable if only trials using CS are accepted for the analysis. The calculated placebo effect of 28.9% over 5.85 months is similar to data from adult rheumatoid arthritis. Despite the diversity of the trials, the resulting figure appears realistic and could be used as supporting information to assess the efficacy of active drugs in JRA.