Background/Purpose: Patients with rheumatoid arthritis (RA) are at increased risk of heart failure (HF) compared to the general population, and HF with preserved ejection fraction (HFpEF) is the predominant subtype in RA. The 2023 Predicting Risk of Cardiovascular Disease Events (PREVENT) risk equations include estimates of 10-year risk for HF (PREVENT-HF) in the general population. As inflammation is important in HF, particularly HFpEF, the ability to estimate HF risk in the RA population may open opportunities for earlier intervention with anti-inflammatory therapy. We aimed to study the performance of the PREVENT-HF risk equation for incident HF among patients with RA.

Methods: We studied a multi-institution electronic health record (EHR)-based longitudinal RA cohort. The study applied validated definitions for RA diagnosis, and patients were followed for incident HF, and HF subtype [HFpEF (EF≥50%) vs HF with reduced EF (HFrEF, EF < 50%)] using a combination of ICD codes and clinical information extracted using natural language processing. RA diagnosis date was the index date, and subjects with prevalent HF were excluded. Covariates for PREVENT-HF and other known HF risk factors were extracted from the EHR 1 year prior and 3 months after the index date. The PREVENT-HF score was calculated at the index date. Subjects were followed until their 1st HF diagnosis, loss to follow-up, or through 10 years, whichever occurred earlier. Discrimination of PREVENT-HF was evaluated by calculating the area under the curve (AUC), and calibration was tested using a calibration plot and the Hosmer-Lemeshow test.

Results: We studied 2,209 patients with RA, and 150 (6.8%) developed HF within 10 years after the index date. RA patients with HF were older, more likely to be male, with higher BMI, lower eGFR, and a higher burden of cardiovascular risk factors than patients without incident HF (Table 1). Those with incident HF also had higher C-reactive protein (CRP) and were less likely to be on methotrexate or TNF inhibitor than patients without incident HF. Of those with incident HF, 72% had HFpEF (EF≥50%) and 28% HFrEF (EF < 50%). The PREVENT-HF estimated median 10-year HF risk was 8.2% (IQR 4.5-16.1) among those who developed HF and 2.4% (IQR 0.9-5.4) among those who did not. The AUC of PREVENT-HF was 0.78, and the calibration plot showed overall underestimation of HF risk (Hosmer-Lemeshow p-value < 0.001; Figure 1). Underestimation was more prominent in the borderline and intermediate risk categories, which were predominantly HFpEF (Table 2).

Conclusion: The PREVENT-HF equation is a promising tool to identify RA patients at increased HF risk, however it generally underestimates HF risk particularly among subjects with borderline and intermediate risk. Further work is needed to determine factors that may enhance prediction for HF, particularly HFpEF, in RA.

Table 1. Baseline characteristics of RA cohort stratified by incident heart failure within 10 years of RA diagnosis date

Figure 1. Calibration plot of predicted HF risk by PREVENT-HF compared to observed incident HF events [red squares denote predicted vs observed by PREVENT-HF risk category].

Table 2. Comparison of the predicted HF risk by PREVENT-HF to observed incidence of HF, categorized by risk category.

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/performance-of-the-prevent-heart-failure-general-population-risk-score-in-patients-with-rheumatoid-arthritis/