In each sample, we analysed the association between CNVs and (extreme) obesity. All primary analyses of the GWAS discovery samples were based on fluorescent intensities which represent the copy numbers. For the family-based GWAS sample, we applied the CNV-FBAT methodology (37) to test the raw intensities for association with the phenotype obesity. We based our association tests on the raw intensities in order to avoid the known loss of information and statistical power resulting from classifying raw copy number measurements into discrete classes, i.e. CNV calls (36,37). To reduce the number of statistical tests and to minimize the risk of false-positive findings, the CNV-FBAT association tests were limited to the common CNVRs derived as described above. Additionally, we performed multi-marker FBATs (35) incorporating all possible marker combinations in a specific CNVR. The case–control GWAS sample was analysed by logistic regression with the predictors normalized intensities, sex and age.
