The genetic architecture of hip shape and its role in the development of hip osteoarthritis and fracture

Abstract

Objectives: Hip shape is thought to be an important causal risk factor for hip osteoarthritis and fracture. We aimed to identify genetic determinants of hip shape and use these to assess causal relationships with hip osteoarthritis.

Methods: Statistical hip shape modelling was used to derive 10 hip shape modes (HSMs) from DXA images in UK Biobank and Shanghai Changfeng cohorts (ntotal = 43 485). Genome-wide association study meta-analyses were conducted for each HSM. Two-sample Mendelian randomisation (MR) was used to estimate causal effects between HSM and hip osteoarthritis using hip fracture as a positive control.

Results: Analysis of the first 10 HSMs identified 203 independent association signals (P < 5 × 10-9). Hip shape SNPs were also associated (P < 2.5 × 10-4) with hip osteoarthritis (n = 26) and hip fracture (n = 4). Fine mapping implicated SMAD3 and PLEC as candidate genes that may be involved in the development of hip shape and hip osteoarthritis. MR analyses suggested there was no causal effect between any HSM and hip osteoarthritis, however there was evidence that HSM2 (more obtuse neck-shaft angle) and HSM4 (wider femoral neck) have a causal effect on hip fracture (ORIVW method 1.27 [95% CI 1.12-1.44], P = 1.79 × 10-4 and ORIVW 0.74 [0.65-0.84], P = 7.60 × 10-6 respectively).

Conclusions: We report the largest hip shape GWAS meta-analysis that identifies hundreds of novel loci, some of which are also associated with hip osteoarthritis and hip fracture. MR analyses suggest hip shape may not cause hip osteoarthritis but is implicated in hip fractures. Consequently, interventions targeting hip shape in older adults to prevent hip osteoarthritis may prove ineffective.

Keywords: Mendelian randomisation; genome wide association study; hip fracture; hip shape; osteoarthritis.

Graphical Abstract

Figure 1

Visual representation of the first 10 hip shape modes. Each hip shape mode (HSM) represents statistically independent (orthogonal) variation in hip shape captured by hip DXA. HSM1 explains the most variation and each HSM there after explains less variation. The HSMs are plotted here where the dotted line represents the hip shape comprised of - 2 standard deviation variation away from the mean and solid line represents + 2 standard deviation hip shape away from the mean. Each participant in this study is given a score for each HSM which represents where their hip shape lies within each mode.

Figure 2

Genetic associations for each hip shape mode. Independent single nucleotide polymorphisms (SNPs) are those that are conditionally independent as obtained by GCTA-COJO. Independent loci are defined by the closest gene. Novel hip shape loci are those further than 1 Mb from previously published SNPs associated with hip shape.

Figure 3

Mendelian randomisation results.Bi-directional causal analyses between hip shape, and hip osteoarthritis and hip fracture.