THE ROLE OF RADIOGRAPHIC ENDOPHENOTYPING IN SIGNAL DETECTION IN GENOME-WIDE ASSOCIATION ANALYSES OF OSTEOARTHRITIS

Introduction: Despite OA heritability estimated at 40-65%, fewer than 20 robustly-associated genetic loci explaining ~10% of heritability have been identified. Genome-wide association studies (GWAS) have been limited by modest sample sizes and the employment of simple dichotomous descriptions of joint involvement. Risk loci associated with hip shape variation, a heritable risk factor for the development of OA, is also largely unknown. Aims: We aimed to determine whether case stratification into precise phenotypic sub-categories is a critical parameter in signal detection in GWAS of hip and knee OA, and hip shape variation. This has not previously been tested systematically. Method: This thesis describes several nested studies from the arcOGEN resource. We initially describe several GWAS in 2,118 cases with radiographic hip OA, stratified by site of maximal joint space narrowing and bone remodelling response, versus 6,500 population-based controls (PBC). We also performed several GWAS’ of 2,010 cases with radiographic knee OA, stratified by dominant compartment involvement, versus 6,500 PBC. Finally, we describe several GWAS with radiographically-derived indices of acetabular and proximal femoral shape. Results: We identified suggestive evidence (p<9.9x10-6) of associations at 6 and 10 variants with hip and knee OA endophenotypes respectively. All would have been missed by a broad phenotype definition approach. The signal yield was increased beyond that associated with chance. Compartmental stratification at the knee also increased the disease heritability estimate. We identified 4 loci (p<9.9x10-6) associated with hip shape variation, including associations with genes regulating myogenesis and skeletal development. Conclusion: We show that increased phenotypic definition in nested case/control cohorts of OA leads to substantial increases in signal detection for susceptibility loci. We report novel associations that were entirely obscured by the heterogeneity of the disease, indicating that the use of narrower OA phenotype definitions can lead to a comprehensive account of variants affecting risk.