Genetic risk factors for knee OA display good predictive power in a high risk group of overweight and obese females

Abstract

Purpose: Over the last few years, several genetic markers have been identified to be robustly associated with knee osteoarthritis (OA). These genetic markers have been identified by gathering all cases and controls available. In general, small effect sizes of these risk factors were found, making OA genetic risk prediction difficult. However, it is known that the association of the genetic marker with the phenotype can be context specific. It could be that in persons with or without major environmental risk factors, different sets of genes are important. We therefore set out to test predictive value of DNA variants previously identified as being associated to knee OA, in a high risk group of overweight and obese, middle-aged women. Methods: Study Sample: This study was performed within the PRevention of knee Osteoarthritis in Overweight Females (PROOF) study. A total of 407 women met the inclusion criteria (BMI ≥ 27, no clinical knee OA (ACR criteria), no contraindications to MRI, no rheumatic diseases), were invited for baseline measurements. Pre-specified primary outcome was incidence of knee OA, defined by incidence of either K&L ≥ 2, joint space narrowing of ≥ 1.0 mm or incident clinical knee OA (ACR criteria). After 2.5 years of follow-up, forty-three women (11%) were lost to follow-up and for a further 72 women no DNA was available. DNA-variants: We studied DNA variants that have been reported as being associated with knee OA at genome-wide significance level (P<5x10-8) in women. These included variants in or near: GDF5 (rs143383), GNL3 (rs11177) and DUS4L (also referred to as the Chr 7 locus, rs4730250) and MCF2L (rs11842874). The assay for the MCF2L variant failed due to technical problems. We subsequently computed a genetic risk score for the 3 variants that were genotyped successfully, where the number of carried risk alleles where added per person. Association between the genetic markers, the genetic risk score and incident knee OA was tested by multivariate logistic regression analysis, adjusting for age and BMI. Results: In total 92 (31%) women developed OA during the 2.5 year follow-up time. When the risk for the separate DNA variants were tested, for all three variants we found increased risk for incident knee OA with the previously identified risk allele (OR(95% CI): GDF5: 1.30 (0.91-1.85); GNL3: 1.23(0.86-1.78); Chr7: 1.75(1.10-2.77)). All these risks were at least twice as large as compared to those previously reported in large meta-analysis. However, only the odds ratio for the chr 7 locus reached nominally significance (p=0.018). The genetic risk score, which was composed of the three SNPs under study, was significantly associated with incident knee OA (p=0.007). With each additional risk allele, the women had 38% more chance for incident knee OA. When we stratified the women into 3 groups according to their genetic risk score, we observed that women with 4 or 5 genetic risk alleles had almost 3.5 times higher risk for incident knee OA compared to women that had 0 or 1 risk allele (OR 3.45 (95%CI 1.41-8.47), p=0.007). Conclusions: The previously identified genetic risk markers for knee OA, were significantly associated with incident knee OA in this high risk group of overweight and obese women. The effect sizes of the studies DNA variants were larger than those reported previously. These results indicate that genetic prediction of OA can be dependent on the presence of other strong clinical risk factors