Background/Purpose: Inflammation may contribute to diabetes risk, and some studies indicate that certain DMARDs might be associated with a reduced risk of diabetes. Most studies in RA have not included confirmed diabetes cases, information about disease activity, BMI and DMARD utilization. The objective of the present study was to assess the association between DMARD exposure and incident diabetes in a large multi-center observational cohort of RA patients.

Methods: The study population consisted of RA patients followed in CORRONA, a longitudinal multi-center cohort. Five mutually exclusive DMARD treatment categories were defined: 1) TNFi, incl. combinations with nonbiologic (nb) DMARDs 2) Other biologic (b) DMARDs, incl. combinations with nbDMARDs 3) Methotrexate, MTX without concomitant bDMARDS or hydroxychloroquine (HCQ) 4) HCQ without concomitant bDMARDs or MTX, and 5) Other nbDMARDs (e.g. LEF, CsA, SSZ), without concomitant TNFi, MTX or HCQ. Patients who did not attend at least one follow-up visit, who had a diagnosis of diabetes at index date of the DMARD regimen, did not receive the DMARD treatments of interest or who had other arthritic diagnoses were excluded. Incident cases of diabetes were confirmed through case ascertainment based on medical records and patient interviews. Incidence rates (IR) and incidence rate ratios (IRR) were calculated. Cox regression models were fitted to estimate the risk of incident diabetes, adjusting for propensity score (including variables such as age, BMI, steroid use and CDAI). The comparison group for all analyses was “Other nbDMARDs”.

Results: We identified 21775 eligible treatment regimens – 9880 TNFi regimens, 1756 other bDMARD, 7441 MTX, 1496 HCQ and 1202 other nbDMARD regimens. The mean (SD) age at start of treatment was 58(13) years, mean disease duration 10(10) years and mean CDAI 13.4(12.4). 76% were women and 30% used oral steroids. The mean duration of the treatment regimens ranged from 1.5 – 2.4 years, and 82 incident cases of diabetes of interest were confirmed. The IRs, IRRs and results from Cox regression models are presented in the table. Diabetes development was significantly reduced in patients receiving TNFi compared to patients receiving other nbDMARDs (e.g. LEF, CsA, SSZ). Use of HCQ, MTX, and other bDMARDs also trended toward a reduced risk, but confidence intervals did not exclude the null. Similar results were found in models adjusted directly for BMI, and in models adjusting for propensity score quintiles. Higher steroid dosages were associated with an increased risk of diabetes in separate models.  

Conclusion: After controlling for variables such as BMI, disease activity and steroid use, RA patients treated with TNF inhibitors had a reduced risk for development of diabetes.