Improved genetic prediction of the risk of knee osteoarthritis using the risk factor-based polygenic score

Table 2 PRS analysis summary of traits that showed association with knee OA

Traits	p-value	Adjusted OR (95% CI)	R² (%)	AUC	AUC test p-value	NRI (p-value)
OA PRS analysis
Multi-population GWAS
Knee OA	6.70E-05*	1.19 (1.09–1.29)	0.520	0.540	0.034	0.081 (0.036)
Knee and/or hip OA	9.90E-04*	1.15 (1.06–1.25)	0.421	0.534	0.059	0.054 (0.161)
All OA	0.014	1.11 (1.02–1.21)	0.330	0.530	0.067	0.027 (0.486)
Total knee replacement	0.008	1.12 (1.03–1.22)	0.165	0.523	0.120	0.067 (0.085)
Total knee and/or hip replacement	0.025	1.10 (1.01–1.19)	0.144	0.524	0.197	0.068 (0.079)
Knee pain	0.026	1.10 (1.01–1.19)	0.211	0.527	0.254	0.092 (0.017)
Risk trait PRS analysis
Biobank Japan GWAS
Body mass index	4.57E-06*	1.22 (1.12–1.32)	0.689	0.548	0.272	0.054 (0.162)
Calcium	0.021	0.91 (0.83–0.99)	0.072	0.514	0.264	0.086 (0.027)
High-density-lipoprotein cholesterol	6.94E-04*	0.87 (0.80–0.94)	0.434	0.539	0.118	0.116 (0.003)
Multi-population GWAS
UKB-body mass index	0.002*	1.14 (1.05–1.24)	0.523	0.540	0.361	-0.017 (0.658)

PRS Polygenic risk score, OA Osteoarthritis, OR Odds ratio, AUC Area under the curve, NRI Net reclassification improvement, GWAS Genome-wide association study, UKB UK Biobank
P-values were evaluated using a logistic regression model. Adjusted OR per standard deviation and 95% confidence intervals of risk of knee OA were calculated using a logistic regression model including age, sex, 10 principal components, and a dummy variable for the array type used in genotyping as covariates. Disease liability explained by the PRS was estimated by the conversion of observed PVE to R2 on the liability scale using a linear model. The AUC test (Delong method) and net reclassification improvement analysis were performed to compare MODEL I (only clinical information, including sex, age, and BMI) and MODEL II (incorporating PRS into MODEL I)
^*Significant after Bonferroni correction

ISSN: 1478-6362