Meta-analysis of rare variants

Meta-analysis is the use of statistical methods to synthesize results of individual studies examining the same trait. A genome-wide meta-analysis primarily serves the purpose of combining data to increase power to obtain statistical evidence of association between disease and variants that would have otherwise escaped detection, for example because of their small effect sizes. For example, the power to attain a p-value of genome-wide significance (5 × 10^-8) for a common variant with 0.20 MAF and a small effect size (odds ratio 1.15) in a GWAS of 2,000 cases and 3,000 controls is 0.45 %, assuming disease prevalence of 1 %, a multiplicative disease model and that the causal variant is typed itself. In contrast, a GWAS meta-analysis of five similar homogeneous studies across 10,000 cases and 15,000 controls has 80 % power to identify risk variants at the genome-wide significance level. Chapman et al. (2011) investigated the way sample size affects the power of GWAS meta-analyses, in the presence and absence of modest levels of heterogeneity and across a range of different allelic architectures.