As noted above, a significant association with obesity of deletions of the SH2B1 locus was identified by quantitative analysis of BMI in the NFBC1966 population cohort alone, Thus, the reduced sample size was compensated for by inclusion of subjects with intermediate phenotypes (i.e. overweight) and the increase in statistical power that derives from analysis of quantitative traits compared to case-control approaches to association testing; indeed, this advantage becomes progressively more marked at lower allele frequencies for the genetic marker under test [22]. Therefore, we investigated whether any other putative GSV-obesity associations were replicated when using this approach. For each candidate GSV that was identified in at least 3 NFBC1966 subjects – the SH2B1 locus and 5 other loci (Table 1) – and also the previously-identified 16p11.2 obesity locus, we conducted a 2-way analysis of variance, with gender and GSV status as explanatory factors and log-transformed BMI as the response variable. Since several individuals carried more than one of these GSVs, we also conducted a single, combined, multifactorial analysis of these 7 GSVs; this gave very similar results to those for the separate individual tests. Alternative approaches (e.g. 2-tailed heteroscedastic t-tests using gender-corrected BMI data) also yielded similar results.
