CAMARADES, University of Edinburgh
Abstract: Meta-Analysis of Preclinical Data in Drug Discovery Research
Meta-analysis of preclinical data is used to evaluate internal and external validity of research findings to inform future studies. Unlike to clinical, preclinical meta-analysis encounters different methodological challenges. We review these assuming aggregate level data and focus on two topics: (1) estimation of heterogeneity using method of moments (DL), maximum likelihood (REML) and a Bayesian approach; (2) comparison of univariable versus multivariable meta-regression for adjusting heterogeneity in treatment effects between studies. Based on statistical theory on hierarchical data, we predict: i) DL will overestimate between study variance; ii) REML will reduce the bias in estimating heterogeneity; iii) Bayesian will more precisely estimate the heterogeneity than the former two as it does not assume known within-study variances. Our findings highlight the importance of heterogeneity estimation method chosen and the use of multivariable meta-regression to best quantify heterogeneity.
Co-Authors: Christel Faes (Hasselt University, Centre for Statistics), Tom Van De Casteele (Janssen Pharmaceutica NV, Translational Medicine and Early Development Statistics), Sarah K. McCann (Charité Universitätsmedizin Berlin, QUEST BIH Center for Transforming Biomedical Research), Malcolm R. Macleod (University of Edinburgh, CAMARADES, Centre for Clinical Brain Sciences)
Poster: Meta-Analysis of Preclinical Data in Drug Discovery Research