We used available postnatal weight data from the EFSOCH, Generation R (Discovery 1), Generation R (Discovery 2), LISA, HBCS, NFBC1966 and NFBC1986 (maximum total n = 15,090). Each study analysed weight data at the following time points, where available: birth; 1 (+/− 0.2) month; 2 (+/− 0.2) months; 3 (+/− 0.3) months; 6 (+/− 0.4) months. Within each study, we created weight-for-age z scores for each of the postnatal time points using Growth Analyser 3.0 (http://www.growthanalyser.org; Dutch Growth Research Foundation, Rotterdam, the Netherlands). The reference was a cohort of 475,588 children born between 1977 and 1981 in Sweden53. Birth weight was analysed as described above. At each subsequent time point, we performed linear regression of weight-for-age z score against rs900400 genotype (or designated proxy SNP, see Supplementary Table 12), with gestational age at birth as a covariable. We combined the results across studies using fixed effects inverse variance meta-analysis.
