Background/Purpose: The response to anti-TNF therapy varies widely between patients with rheumatoid arthritis (RA). MicroRNAs (miRNAs) are suggested to influence susceptibility to RA and disease severity. We aimed to investigate the association between miRNAs and treatment response in pre-treatment whole blood from 180 patients with early RA (treatment-naïve, disease duration <6 months) enrolled in the OPERA Study, a prospective, investigator-initiated, double-blinded, placebo-controlled study (1).

Methods: Patients were randomized 1:1 to methotrexate (MTX) 7.5 mg weekly (increased to 20 mg/week within two months) in combination with adalimumab (n=89) or placebo-adalimumab (n=91) 40 mg subcutaneously every other week. Both groups received intra-articular glucocorticoid injections (i.a. GC). Treatment response after 1 year was assessed according to EULAR response, DAS28(CRP) remission, and ACR/EULAR Boolean remission criteria. Expression of miRNAs was determined using TaqMan® Human MicroRNA LDA, A Card v2.0 (Applied Biosystems). Raw Cycle threshold (CT) values of each miRNA were pre-processed using Quantile, 120 most expressed, and Rank normalizations. We performed interaction analyses to identify miRNAs associated differently with response to MTX + adalimumab and MTX + placebo-adalimumab. The overall test for no interaction between miRNAs and treatment was achieved by means of Kolmogorov-Smirnov. Potential predictive miRNAs were included in a multivariate model and backwards eliminated using the Bayesian Information Criterion. All effects were reported for a CT inter-quartile range increase.

Results: Kolmogorov Smirnov tests indicated interactions between miRNAs and treatment using EULAR response as outcome parameter (120 most expressed, p=0.008; Quantile, p=0.003; and Rank normalization, p=0.011). After backwards elimination, miR-22 and miR-886.3p demonstrated interaction with treatment using EULAR response as outcome parameter, Table 1. Nomograms suggested that in patients treated with adalimumab, the combination of low miR-22 and low miR-886.3p expression had the highest probability of EULAR good response (95%), whereas patients with high miR-22 and high miR-886.3p had the lowest probability of EULAR good response (65%).

Conclusion: In a one-year randomized, placebo-controlled, double-blind clinical trial of treatment-naïve patients with early RA, which compared adalimumab+MTX+i.a.GC with placebo+MTX+i.a.GC, we found that expression of miR-22 and miR-886.3p in pre-treatment whole blood was predictive of EULAR good response to adalimumab. Validation studies are needed to investigate the utility of these miRNAs as predictive biomarkers.

¹ Hørslev-Petersen K et al. Ann Rheum Dis Online First 7 mar 2013

Multivariate estimated effects for 120 most expressed miRNAs, Rank, and Quantile normalization.