Algorithms using genome-wide association studies for prediction of effectiveness of biologics in rheumatoid arthritis

Purpose: Two anti-TNF biologics, infliximab and etanercept, are extremely useful for inflammatory diseases such as rheumatoid arthritis (RA). However, approximately 20 to 30% of RA have been described as non-responders. Here, we analyzed the statistical relationships of whole genome single nucleotide polymorphisms (SNPs) with remission or low disease activity (LDA) among RA and developed algorithms using SNPs to predict effectiveness of these biologics. Methods: The total study subjects (first, second and validation population for each infliximab and etanercept) were 260 RA patients for infliximab and another 251 RA patients for etanercept. Effectiveness of infliximab and etanercept was assessed using DAS28-CRP. In first, second and the combined population, relationships of 277,339 SNPs with remission and LDA were analyzed using case-control analyses by Fisher’s exact tests for each infliximab and etanercept. We picked up SNPs (P < 0.05) from each first, second and combined population. Then, 10 SNPs with lower P value in the combined population were selected from common SNPs among the first, second and combined populations. We developed algorithms using the 10 SNPs to predict effectiveness of infliximab and etanercept. In the validation population, availability of the algorithms was evaluated. Results: Using combined infliximab-remission and LDA algorithms in the validation population, the accuracy was 90.4%. Using combined etanercept-remission and LDA algorithms in validation population, the accuracy was 76.8%. Conclusions: The combined use of our two algorithms using SNPs may be very useful in the prediction of remission or LDA before treatment with infliximab or etanercept.