Can Indirect comparisons be improved using individual patient data?

Tags: Poster
Tudur C, Williamson P, Marson A, Chadwick D

Abstract: Seven pairwise comparisons of 5 different anti-epileptic drugs (AED) have been examined separately. Individual patient data (IPD) are available for a total of 14 randomised controlled trials (RCT) of these AED's in 3472 patients with epilepsy. As the number of trials included in any single comparison ranges from 2 (178 patients) to 5 (1265 patients), there is a need to improve the precision of the treatment effect estimates. A bayesian approach, incorporating IPD from all relevant trials is explored. The precision of a treatment effect estimate may be further improved by incorporating important prognostic factors using regression modelling. In our example from epilepsy, a high proportion of IPD are available for 5 pre-specified clinically important prognostic factors. Using regression modelling, we will investigate what impact these factors have on precision. The trade-off between including additional parameters in the model and improving the precision of the treatment effect estimate will also be discussed.