Partitioning knee osteoarthritis genetic variants based on BMI association to improve polygenic risk prediction

Introduction: Polygenic risk scores (PRSs) show promise for improving knee osteoarthritis (KOA) risk prediction. However, many KOA-associated variants are also linked to body mass index (BMI)—a major KOA risk factor—complicating the combined use of PRS and BMI in predictive models. This study aimed to disentangle BMI-mediated from BMI-independent genetic contributions to KOA by partitioning the KOA-PRS. Method: Using data from 345,080 white European participants in the UK Biobank, we refined a KOA-PRS based on 146 genome-wide significant variants identified by the GO consortium. KOA variants were partitioned into BMI-associated (KOA-BMI PRS), and BMI-independent (KOA-nonBMI PRS) groups based on linkage disequilibrium with BMI-associated variants from the GIANT consortium. We assessed associations of each PRS with prevalent and incident KOA (with and without BMI adjustment), with BMI in cases and controls, and with KOA across BMI strata. Results: Of the 146 KOA-associated variants, 73 were classified as BMI-associated. The KOA-BMI PRS was associated with incident KOA (OR per SD:1.23;95%CI:1.20–1.26), attenuated after BMI adjustment (1.18;1.15–1.21). In contrast, the KOA-nonBMI PRS remained robust to BMI adjustment (1.24 before vs.1.23 after-adjustment). KOA-nonBMI PRS was negatively associated with BMI in KOA cases (β=–0.09;95%CI:–0.17to–0.01), but not in controls. Stratified analyses showed its effect on KOA diminished in higher BMI strata, while KOA-BMI PRS effects remained stable. Conclusion: Half of KOA-associated variants overlap with BMI loci. Partitioning the KOA-PRS improves interpretation by distinguishing BMI-driven from independent genetic effects—enhancing prediction and informing tailored prevention. Validation in diverse populations is warranted.