Background/Purpose: Three-quarters of RA patients use glucocorticoids (GC) at some point to manage RA symptoms. Prior work suggests recent GC use is associated with major adverse cardiovascular events (MACE) in RA, but does not adequately account for prior GC exposure or time-varying MACE risk.

Methods: In this retrospective cohort study, we used national VA administrative data to identify RA patients with an initial VA rheumatology visit (index date) in 2010-2018. Exclusions included age < 40 or >90, non-RA rheumatologic disorders, prior MACE, or CHF during a lookback period of up to 5 years. We used pharmacy dispensing data to calculate days’ supply of GC per 6-month period, empirically categorized as 0, 1-7, 8-90, >90 days (exposure). We used VA claims and National Death Index data to identify first MACE, updating every 6 months (outcome). We defined MACE as MI, stroke/TIA, cardiac arrest, coronary revascularization, or death from CV cause. Patients were censored at non-CV death, or 2 years after last VA service use (Fig. 1). We estimated the overall effect of prior period GC on the 6-month lagged MACE outcome using a marginal structural model with stabilized weights. This approach was chosen to account for current and prior time-varying confounders that may affect current GC use. Key time-varying covariates used for model weights included a) cumulative GC duration between index date and exposure period, calculated using pharmacy claims; b) continuous 5-year MACE risk percentage, estimated using the Veteran’s Affairs Risk Score for CV Disease (VARS-CV) (Table 1). VARS-CV uses medical and pharmacy claims, vital signs, and labs to recalibrate the ACC/AHA risk model for the Veteran population. It calculates a continuous 5-year risk percentile that can be categorized as low (< 3%), medium (3-9%) or high ( >9%) to align with ACC/AHA 10-year cutoffs of < 7.5%, 7.5%-20%, and >20%.

Results: Among 19,280 patients (mean age 63 years, 84% male), median 5-year MACE risk at baseline was 5.3%; 3,869 patients (20%) had high MACE risk (Table 1). Mean cumulative GC duration in the 12 months after index date was 63.3 days overall, and 67.0 days for those with high MACE risk. Incident MACE occurred in 4.2% of patients overall, and in 7.1% with high MACE risk. Median time to MACE was 1,079 days. Using a traditional pooled logistic regression model adjusted only for baseline covariates, compared to 0 days of GC use, relative odds of MACE was 1.45 (95% CI 0.97-2.17) for 1-7 days of use, 1.42 (1.14-1.77) for 8-90 days, and 1.99 (1.66-2.39) for >90 days. In the weighted model accounting for time-varying covariates, relative odds of MACE was 1.54 (0.95-2.49) for 1-7 days of use, 1.78 (1.21-2.61) for 8-90 days, and 2.17 (1.70-2.78) for >90 days (Table 2).

Conclusion: In this national RA cohort, we observed a dose-response relationship between days of GC use in the past 6 months and odds of first MACE in the next 6 months, independent of other time-varying CV risk factors and prior GC use. Future work will examine how other MACE risk factors may alter the relative effect of GC use on MACE, the effect of GC dose, the delayed impact of remote GC use, and effects of GC use on subsequent MACE.

« Back to ACR Convergence 2022

ACR Meeting Abstracts - https://acrabstracts.org/abstract/time-dependent-evaluation-of-glucocorticoid-exposure-duration-and-major-adverse-cardiovascular-events-in-a-cohort-of-veterans-with-rheumatoid-arthritis/