NOVEL GENETICALLY DRIVEN OA RISK SCORE REQUIRES ONLY PELVIC RADIOGRAPHS: A LONGITUDINAL MULTICOHORT STUDY

Hip OA is highly heritable, with genetics accounting for up to 60% of hip OA risk, possibly partly mediated by hip morphology. A proxy for genetically-driven morphology could enhance OA risk prediction when an individual's genetic data is unavailable. Previously, the World COACH consortium identified 10 genome-wide significant loci associated with OA-free hip shape using Combined GWAS (C-GWAS). OA-free hip shape was described using statistical shape modeling (SSM), yielding shape modes that quantify individual shape variance within a population. Several of these loci were previously shown to be associated with hip OA using GWAS. These overlapping loci provide a basis for using hip morphology to construct a phenotypic proxy for OA risk.

We hypothesized that a proxy score capturing genetically driven hip morphology patterns on pelvic radiographs can predict OA risk. This study aimed to 1) construct a novel OA risk score, and 2) investigate whether this score is associated with radiographic hip OA (RHOA) development within 4-8 years.

We pooled individual participant data from 12,672 individuals (21,735 hips) within the World COACH consortium with available biological sex, baseline age, and baseline and 4-8 year follow-up RHOA grades. Of these participants 70.5% were female with a mean baseline age of 62.4 ± 8.4 years. RHOA was quantified using Croft, KLG, or modified grades, and harmonized as "No OA" (any score 0), "doubtful OA" (any score 1), or "definite OA" (any score ≥ 2 or THR). Hip morphology was quantified using an SSM of 80 automatically placed landmarks outlining the proximal femur and hemipelvis on the baseline anteroposterior pelvic radiographs. The OA risk score was constructed by merging the SSM shape modes associated with OA-related loci, weighted by their respective C-GWAS effect sizes. This yielded a score per hip that quantifies it's similarity to the genetically defined OA-related shape. Figure 1 visualizes the different steps of creating the OA risk score. A one-stage logistic regression model with generalized mixed effects accounting for hip side, individual, and cohort was used to assess the association between the novel OA risk score and RHOA development within 4-8 years. The score was modelled absolute to reflect increased risk of RHOA development at both extremes with a unit size of 0.01 SD. Models were adjusted for baseline age and biological sex. Statistical significance was set at p<0.05.

Within 4-8 year follow-up, 2.2% of hips developed definite RHOA. The OA risk score was constructed by merging 9 different shape modes. Each 0.01 SD increase in the absolute OA risk score was associated with 1% higher odds of RHOA development (adjusted OR 1.01, 95% CI 1.006-1.017, p < 0.001).

This novel score effectively quantifies a genetically defined hip morphology pattern linked to OA risk. It offers a promising phenotypic proxy for genetically driven hip OA risk without requiring individual genetic data. The score could improve risk stratification only using standard pelvic radiographs.