Combining aggregate and individual level data – opportunities and considerations Session title: High-impact Applications of Model-based Meta-analyses (MBMA) in Drug Development

The combined analysis of aggregate (AD) and individual patient-level data (IPD) can help address weaknesses with analyses focusing on only one data type. AD offer opportunities to compare multiple treatments, but information in such data is scarce about the importance of covariate effects and lack resolution across different variability components. IPD can seldom be accessed to the extent that AD can, but can complement with information about covariate effects and variability components. Therefore, combining AD from many studies, with IPD from a few, can offer unique advantages: e.g. to perform individual level clinical trial or population level simulations across treatments and populations. The challenges of a combined AD and IPD analysis are of both principal, practical and technical nature. For example, (i) covariate relations can differ in strength, and even direction between and within studies, (ii) AD may involve additional error sources, such as digitization error, compared to IPD, and (iii) aggregation bias with respect to variability parameters in AD, impacts the choice of distributional shape. Strategies for handling these and other challenges with combined AD and IPD analysis will be presented and illustrated with continuous (FEV1) and categorical (exacerbation rate) variables based on studies on patients with chronic obstructive pulmonary disease.