Background/Purpose:  Pharmacogenetics studies of anti-tumor necrosis factor (TNF) response among patients with rheumatoid arthritis (RA) have routinely used assessment of response at a single time-point, ranging between 3 and 12 months after initial treatment, to define the phenotype. Given the general lack of consistency in the findings from published studies, we have examined whether differences in the timing of anti-TNF response assessment in the same patients can influence the results of a genome-wide association study (GWAS) investigating associations between anti-TNF response and single nucleotide polymorphisms (SNPs).

Methods:  Disease Activity Scores based on 28 joint counts and C-reactive protein (DAS28) were assessed at baseline (before initial therapy), and after 3 and 6 months in 487 Japanese RA patients starting anti-TNF therapy for the first time or switching to a new anti-TNF agent. DNA samples from these patients were genotyped using a genome-wide panel of 1,133,484 SNPs and additional SNPs were imputed. Change in DAS28 scores from baseline were evaluated at 3 months (ΔDAS28-3) and at 6 months (ΔDAS28-6). Association analyses between anti-TNF response at one follow up and each SNP were performed using 2 separate multivariate linear regression models, with ΔDAS28-3 (model 1) and ΔDAS28-6 (model 2) as the outcome (response) variable. A longitudinal Generalized Estimating Equations model (model 3) was also used to allow inclusion of response at both 3 and 6 months in the outcome variable. All models adjusted for baseline DAS28, type of anti-TNF agent and concomitant methotrexate. In model 3, time since initiation of therapy was also adjusted for.

Results:  At 3 months (model 1), five SNPs were moderately associated with ΔDAS28-3 (p<1×10^-5). At 6 months (model 2), a SNP at 6q15 was associated with ΔDAS28-6 at a genome-wide significance level (p=2.5×10^-8). This SNP mapped close to the MAP3K7 gene which is a key player in TNFα-mediated inflammatory pathway signaling. Four other chromosomal regions showed evidence of moderate association with ΔDAS28-6 (p<1×10^-5), including a SNP at 10q25.3 which mapped to the GFRA1 gene (p=5.3×10^-6). There was no overlap between the regions of association identified when using as phenotype response at 3 months (model 1) vs. response at 6 months (model 2). On the other hand, when response at both 3 and 6 months were included in the phenotype (model 3), suggestive association with anti-TNF response was observed at two of the same loci identified in model 2: 6q15 (p=6.6×10^-7) and 10q25.3 (p=8.1×10^-7).  A third region mapping to the WDR27 gene at 6q27 showed suggestive evidence of association in model 3 (p=6.3×10^-7), but not in models 1 and 2 (ΔDAS28-3: p=7×10^-4; ΔDAS28-6: p=5.5×10^-5).

Conclusion: The timing of anti-TNF response assessment significantly influences the results of genetic association studies, especially since clinical response is known to fluctuate over time. Using more than one assessment of response provides a more reliable clinical phenotype and, in our dataset, enhanced the power to detect associations.

« Back to 2015 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/does-the-timing-of-phenotype-assessment-influence-association-results-in-pharmacogenetics-studies-of-anti-tnf-response-in-rheumatoid-arthritis/