Objective: Randomized trials for uncommon diseases suffer from methodological challenges: difficulty in recruiting sufficient numbers of patients and low power to detect important treatment effects. Using traditional (frequentist) analysis, p values > 0.05 mean investigators are unable to reject the null hypothesis (of no treatment effect). The medical community often labels trials with p values > 0.05 as "negative." Our study demonstrates how Bayesian analysis conveys more relevant information to clinicians - using the example of methotrexate (MTX) in systemic sclerosis (SSc).
Methods: Data from 71 patients with diffuse SSc (n = 35 MTX, n = 36 placebo) in the trial were reanalyzed using Bayesian models. We examined 3 primary outcomes: modified Rodnan skin score (MRSS), University of California Los Angeles (UCLA) skin score, and physician global assessment of overall disease activity. Using noninformative prior probability distributions, the probability of beneficial treatment effects for each outcome and the probability of simultaneous benefit in outcomes were computed.
Results: The probability that treatment with MTX results in better mean outcomes than placebo was 94% for MRSS, 96% for UCLA skin score, and 88% for physician global assessment. There was 96% probability that at least 2 of 3 primary outcomes were better on treatment.
Conclusion: Bayesian analysis of uncommon disease trials allows for more flexible and clinically relevant interpretations of the data. From the trial data, clinicians can infer that MTX has a high probability of beneficial effects on skin score and global assessment.