Does Adding Single-Nucleotide Polymorphisms to Risk Algorithms Improve Cardiovascular Disease Risk Prediction in Rheumatoid Arthritis? An Internal and External Validation of a Clinical Risk Score

Objective

Current risk algorithms do not accurately predict cardiovascular disease (CVD) risk in rheumatoid arthritis (RA). An area of interest is that of single-nucleotide polymorphisms (SNPs), of which several have been associated with CVD in the general population. We investigated whether these SNPs are associated with CVD in RA and whether SNPs could improve CVD risk prediction in RA.

Methods

Sixty SNPs were genotyped in 353 patients with RA. Logistic and Cox regression analyses were performed to identify SNPs that were associated with CVD (n = 99). A prediction model with clinical variables was made. SNPs were added to investigate the additional predictive value. Both models were internally validated. External validation was done in a separate cohort (n = 297).

Results

rs3184504, rs4773144, rs12190287, and rs445925 were significantly associated with new CVD. The clinical prediction model consisted of age, sex, body mass index, systolic blood pressure, high-density lipoprotein cholesterol (HDLc), and creatinine, with an area under the curve (AUC) of 0.74 (P = 0.03). Internal validation resulted in an AUC of 0.76 (P < 0.01). A new model was made including SNPs and resulted in a model with rs17011666 and rs801426, age, total cholesterol, and HDLc, which performed slightly better with an AUC of 0.77 (P < 0.01). External validation resulted in a good fit for the clinical model, but a poor fit for the SNP model.

Conclusion

Several SNPs were associated with CVD in RA. Risk prediction slightly improved after adding SNPs to the models, but the clinical relevance is debatable. However, larger studies are needed to determine more accurately the additional value of these SNPs to CVD risk prediction algorithms.