OBJECTIVES: Individual Patient Data (IPD) from Randomised Control Trials (RCTs) are considered the gold standard for evaluating treatment regimens in a Mixed Treatment Comparison (MTC). However, as the majority of studies do not report IPD, most MTCs are carried out using aggregate data (AD) for at least some, if not all, of the studies. We investigate the benefits of including varying proportions of IPD studies in an MTC.

METHODS: Donegan et al (2013) developed a number of models for including both AD and IPD in the same MTC. We carried out a simulation study of RCTs based on these models to check the effect of additional IPD studies on the accuracy of the estimate of both the treatment effect and the covariate effect. We also compared the Deviance Information Criteria (DIC) between different models to assess model fit. We then applied this to a Hepatitis C network including both RCTs and observational studies.

RESULTS: Our estimate of the covariate effect becomes more accurate as we increase the proportion of IPD studies in the network. However, the estimate of the treatment effect is unaffected, as a well conducted RCT will account for differences in covariates in the study design. The DIC distinguishes between models more often when there is a high proportion of IPD studies. In the Hepatitis C network even one IPD observational study decreases the standard deviation of both covariate effect and treatment effect estimates.

CONCLUSIONS: Inclusion of IPD reduces uncertainty surrounding the covariate effect. As RCTs are considered the gold standard of evidence, including IPD does not improve the accuracy of the treatment effect. However, IPD may be most useful in observational studies where the covariates may not be well balanced between individual treatment arms.