Single patient trials with a binomial measure of treatment success: a hierarchical Bayesian model
by
Philip Schluter
University of Queensland

Physicians are often faced with the difficulty of applying trial results of some treatment, invariably reported at the level of the group, to individuals who each may exhibit considerable variability in their response to that treatment. Physicians and their patients are thus compelled to seek more pragmatic solutions by performing their own experiments to assess treatment effectiveness. Exploiting these individual investigations, using a more rigorous design, the single patient trial (SPT) methodology has been developed and utilised for the study of individual patients with a chronic condition. A SPT uses a randomised controlled multi-crossover study design. Based on the normal likelihood assumption, Zucker et al. (1997) derive a hierarchical Bayesian model to combine SPT studies to obtain an estimate of treatment effectiveness for the population and to use this population information to aid in the evaluation of an individual patient's trial results. However, given that most SPTs consist of between two and four crossover periods and outcome variables are frequently measured on discrete scales, such as the visual analogue scale, the assumption of normality may not always be appropriate. To circumvent this difficulty, a hierarchical Bayesian model that is based on a binomial measure of treatment success is derived. This model shares the same salient features of the model proposed by Zucker et al. (1997). Using published data from 23 SPT results comparing amitriptyline and placebo for the treatment of fibromyalgia, we demonstrate the proposed method and compare the derived numerical results to those presented in Zucker et al. (1997).

Reference

Zucker DR, Schmid CH, McIntosh MW, D'Agostino RB, Selker HP and Lau J (1997). Combining single patient (N-of-1) trials to estimate population treatment effects and to evaluate individual patient responses to treatment. J Clin Epdemiol, 50(4):401-410.

Date received: April 2, 2002