Results from both studies were combined using a fixed effects inverse variance meta-analysis, using the software package METAL (55). This approach weights effect size estimates according to the inverse of their standard errors. Variance explained by each SNP was calculated as 1 − var(res.full)/var(res.null)*100 of the model (proportional odds logistic regression/survival regression) with age at measurement, sex and gestational age (var, variance; res, residuals). To correct for over-fitting each individuals, phenotype was estimated from a model that did not include that individual (6).
