2.4. The association of adult hip shape variants with adolescent hip shape

Genome-wide meta-analysis of adult hip shape (based on five cohorts: ALSPAC mothers, Framingham Osteoporosis Study, Osteoporotic Fractures in Men, Study of Osteoporotic Fractures and TwinsUK) identified nine SNPs across the 10 HSMs tested: five were associated with HSM1, three with HSM2 and one with HSM5 [10]. As previously described [10,19], genotypes were imputed to Haplotype Reference Consortium panel version 1 (ALSPAC, FOS, TwinsUK) or 1000 Genomes Project phase 1 version 3 (MrOS, SOF). Of participating adult cohorts Twins, SOF and MrOS DXA scans were acquired with Hologic scanners whereas FOS and ALSPAC (mothers and offspring) were acquired using Lunar scanner. These differences were reflected in skewed HSM1 scores, thus all genetic estimates and SDs were rescaled before the meta-analysis was conducted to standardize the results across the modes, as previously described [10].

Associations between adult SNPs and corresponding HSMs in adolescents were investigated separately at age 14 and 18 years. Linear regression analyses were performed using SNPTEST v2.5 assuming an additive genetic model, adjusting for age (in years) and sex. Additional analyses (adjusted for height and BMI) were also carried out. To adjust the level of statistical evidence, we applied a Bonferroni adjusted p value of 0.006 (α 0.05 divided by 8 SNPs tested), to account for the number of tests carried out. Given the units of the HSM variables, regression coefficients are expressed as change in the adult-standardised deviations from the mean hip shape per allele. Such effects are not that meaningful in themselves but do allow for comparison of effect estimates between the adult and adolescent analyses. In order to address the potential non-independence from the adult study (due the relationship between ALSPAC mothers and offspring), adult meta-analysis effects were re-calculated excluding ALSPAC mothers.