Performance of the PREVENT heart failure risk score and correlation with microvascular dysfunction and coronary plaque burden in patients with rheumatoid arthritis

1. Introduction

Patients with rheumatoid arthritis (RA) have a 1.5-fold increased risk of heart failure (HF) compared to the general population [1]. The 2023 Predicting Risk of Cardiovascular Disease Events risk score for HF (PREVENT-HF) was developed to identify at-risk patients in the general population earlier for prevention and management [2]. However, there are limited data on the performance of PREVENT-HF in RA to inform use in clinical care and research.

In RA, systemic inflammation is associated with excess HF risk, potentially due to coronary microvascular dysfunction (CMD) and accelerated atherosclerosis [3]. Advanced CV imaging methods that measure CMD and atherosclerotic burden provide an opportunity to understand the correlation between thse imaging markers and predicted HF risk.

The objective of this study was to assess the performance of the PREVENT-HF risk score in RA and to gain mechanistic insight into HF risk captured by PREVENT-HF by assessing measures of coronary microvascular dysfunction and coronary artery plaque volume.

2. Methods

2.1. Participants and study design

We utilized an electronic health record (EHR)-based longitudinal RA cohort in a large, multihospital healthcare system. The cohort was defined using a validated algorithm for RA (positive predictive value (PPV) 90%) [4]. The index date was date of RA diagnosis. We studied patients aged 30 to 79 years with EHR encounters ≥1 year prior to and 90 days after the index date. Subjects with prevalent HF, myocardial infarction, or stroke were excluded. Subjects missing the required PREVENT-HF components available between 1 year prior to and 90 days after the index date were excluded.

Incident HF and HF date were defined using a previously published algorithm using a combination of ICD codes and EHR data extracted using natural language processing (PPV 90% for HF diagnosis) [4]. Subjects were censored at incident HF, loss to follow-up, or 10 years after the index date, whichever occurred first.

The PREVENT-HF score was calculated using the published base model at the index date. The components of PREVENT-HF were extracted from the EHR from 1 year prior to 90 days after the index date; diabetes history was determined from EHR entry to 90 days after index date. Antihypertensive medication use was defined using prescription data. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration 2021 creatinine equation [2]. RA-related covariates including serostatus and RA-specific medications were extracted ± 90 days of the index date.

2.2. Imaging correlates

A subset of RA patients from the same population as the EHR RA cohort were enrolled in a research study, the Lipids, Inflammation and Cardiovascular Risk in RA (LiiRA) study (ClinicalTrials.gov NCT02714881) and a long-term extension study [5]. LiiRA enrolled RA patients about to initiate a tumor necrosis factor inhibitor (TNFi) as directed by their rheumatologist. Inclusion criteria were individuals age >35, without a history of CVD and who were not on a statin. The PREVENT-HF score was calculated for all LiiRA patients at baseline. All LiiRA subjects underwent a myocardial perfusion PET study to calculate their myocardial flow reserve (MFR) and detect presence of CMD (defined by MFR<2.5) [5]. A subset of LiiRA subjects were enrolled in a long-term extension study and underwent an additional cardiac perfusion PET and coronary CT angiogram (CCTA) to quantify coronary plaque volume >2 years after their baseline LiiRA visit and MFR assessment [6]

2.3. Statistical analysis

The clinical characteristics of the RA cohort at baseline were compared for those who eventually developed HF vs no HF using paired t-tests or the Wilcoxon signed rank test. We categorized subjects into 10-year estimated CVD risk categories based on the American Heart Association CVD risk categories: low (<5%), borderline (≥5% to <7.5%), intermediate (≥7.5% to <20%), and high (≥20%).

We assessed discrimination by calculating the area under the curve (AUC). Calibration was assessed by plotting the observed percentage of subjects who had incident HF within 10 years versus the predicted incident HF percentage based on PREVENT-HF scores for each risk category. We additionally assessed calibration with the Hosmer-Lemeshow test and calculated the observed-to-expected (O/E) ratio for each risk category. The O/E ratios in each category were further stratified by CRP above vs below 3 mg/L (a threshold for elevated CV risk) among individuals with available CRP. We used a Cox proportional hazards regression to estimate hazard ratios (HR) for incident HF for each risk category compared with the low-risk category (reference).

For individuals with imaging results from LiiRA, we calculated the Spearman rank correlation between the baseline PREVENT-HF score and baseline MFR, and between baseline PREVENT-HF score and follow-up MFR. For subjects in the follow-up study, we calculated the Spearman rank correlation between the baseline PREVENT-HF score and follow-up plaque volume.

This study was approved by the Mass General Brigham Institutional Review Board.

3. Results

We studied 2064 RA subjects, mean age 55, 75% female, 46% seropositive, with a median 10-year PREVENT-HF score of 2.4%; 125 (6.1%) developed HF within 10 years (Panel A).

3.1. Performance characteristics of the PREVENT-HF equation in RA

The AUC of the PREVENT-HF score was 0.78 (95% CI 0.74–0.82). Based on the calibration plot and O/E ratios, the PREVENT-HF score underestimated HF risk among RA patients in the low, borderline and intermediate risk categories and overestimated risk in the high risk category (Panel B). The Hosmer-Lemeshow test suggested that the PREVENT-HF equations are limited in modeling HF risk in the RA cohort (p-value <0.001). When stratified by CRP levels, the underestimation of risk was driven by RA patients with low or intermediate predicted HF risk and CRP ≥3mg/L (O/E: 1.6 low, 1.7 intermediate). The HR for incident HF was 4.1 (95% CI 2.5, 6.7) for borderline risk, 5.3 (95% CI 3.5, 8.0) for intermediate risk, and 8.8 (95% CI 5.1, 15.1) for high risk.

3.2. Imaging correlates of PREVENT-HF risk

Among subjects in the LiiRA study, which enrolled 73 RA subjects with mean age 55, 82% female, mean BMI 30, and 68% seropositive, median hsCRP was 4.1 mg/L and mean clinical disease activity index [CDAI] was 24 (moderate-high disease activity). The median PREVENT-HF score was 2.1. As previously published, mean baseline MFR was 2.65 (SD 0.48) [5].

At baseline, higher PREVENT-HF scores were cross-sectionally correlated with lower MFR, indicating a higher degree of CMD (r_s −0.31 (95% CI −0.49, −0.10), p = 0.008) (Panel C). In the extension study (>2 years after baseline cardiac PET, n = 18), a higher baseline PREVENT-HF score was correlated with lower (worse) MFR, higher total plaque volume and higher non-calcified plaque volume, but not with total calcified plaque (Panel C).

4. Discussion

In summary, in this RA cohort, the PREVENT-HF score had reasonable discrimination but may underestimate HF risk among individuals with lower estimated risk and higher CRP levels, highlighting a clinical subgroup where additional attention may be required during clinical evaluation or for future efforts to improve risk stratification. A higher PREVENT-HF score was correlated with a higher degree of CMD cross-sectionally and in follow-up, with a signal for higher future non-calcified plaque burden.

The results from this study corroborate findings from a large, predominantly male cohort of RA patients where PREVENT-HF generally underestimated HF [7]. In our study, the PREVENT-HF score was correlated with a higher degree of CMD, consistent with data that CMD is a feature of HF in the general population [3]. These data lend support to the hypothesis that in RA, higher levels of inflammation, correlated with a higher degree of CMD, are possible mechanisms for excess HF risk.

This study should be viewed within the context of several limitations. We used data from a single healthcare system, and HF diagnoses managed outside of the system may have been missed. However, the proportion of RA subjects who developed HF is similar to other studies [1]. Since we used a retrospective cohort for our main analysis, inflammatory markers were not available for about a third of the cohort, and we did not have standardized measures of disease activity. Similarly, most patients did not have the optional components of the PREVENT-HF score measured; 79% did not have a hemoglobin A1c and 94% a urine albumin:creatinine. Thus, we were unable to evaluate if these components might improve HF risk stratification over the base model. In this cohort without prior known CV risk, N-terminal pro B-type natriuretic peptide (NT-proBNP) and cardiac troponin were not available for study, and the role of these markers for improving HF risk stratification is an important area for future study. The PREVENT-HF scores were determined at the time of RA diagnosis and future directions include assessing the performance of HF risk in patients with longer RA duration. The LiiRA study included patients with moderate-high disease activity initiating a TNFi, and the results may not generalize to patients already on biologic therapy or with lower disease activity. Finally, the longitudinal imaging sub-study had a small sample size and the results require replication in a larger cohort.

The PREVENT-HF score is a promising tool for assessing HF risk in RA but should be interpreted with caution among RA patients with low predicted risk, especially in those with elevated CRP. Cardiac imaging biomarkers of microvascular dysfunction and non-calcified plaque burden were correlated with increased HF risk and may improve risk HF risk stratification. Future directions include applying the PREVENT-HF equations to identify individuals with RA at elevated risk for HF, enabling studies of the inflammatory pathways associated with microvascular dysfunction and accelerated atherosclerosis to inform preventative strategies. Fig. 1.

Fig. 1.

Panel A. Baseline characteristics of EHR RA cohort stratified by incident HF within 10 years of RA diagnosis date. Panel B. Calibration plot of predicted HF risk by PREVENT-HF compared to observed incident HF events among patients without history of ASCVD or HF [red squares denote predicted vs observed by PREVENT-HF risk category]. Table shows observed-to-expected ratio (O/E) for each PREVENT-HF risk category stratified by CRP ≥3 mg/L vs < 3 mg/L, among those with available CRP values. Panel C. (i) Cross-sectional Spearman’s correlation between predicted HF risk by PREVENT-HF and Myocardial Flow Reserve (MFR) at baseline (n = 73) and (ii) Spearman’s correlations between PREVENT-HF estimated risk and follow-up (≥2 years) MFR, total plaque volume, total calcified plaque volume, and total non-calcified plaque volume. MFR was measured by cardiac perfusion PET, and plaque volume was measured by coronary CT angiogram.

Funding

KPL:

K24 AR085342

R01 HL127118

P30 AR072577

Llura Gund Award

Harold and DuVal Bowen Fund

BeBrave Fund

JEH:

R01 HL160003, R01 HL168889, K24 HL153669

BW:

NIH K23HL, 2026 Texas Health Scholar Award (THR)

Author declaration

We wish to draw the attention of the Editor to the following facts which may be considered as potential conflicts of interest and to significant financial contributions to this work.

BNW reports consulting for Kinsika, NovoNordisk, BMS, and Horizon, outside of the submitted work.

We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing we confirm that we have followed the regulations of our institutions concerning intellectual property. We further confirm that any aspect of the work covered in this manuscript that has involved either experimental animals or human patients has been conducted with the ethical approval of all relevant bodies and that such approvals are acknowledged within the manuscript. We understand that the Corresponding Author is the sole contact for the Editorial process (including Editorial Manager and direct communications with the office). She is responsible for communicating with the other authors about progress, submissions of revisions and final approval of proofs. We confirm that we have provided a current, correct email address which is accessible by the Corresponding Author.

CRediT authorship contribution statement

Ilana M. Usiskin: Writing – review & editing, Writing – original draft, Visualization, Formal analysis, Conceptualization. Katherine Zhong: Writing – review & editing, Formal analysis. Ying Qi: Writing – review & editing, Formal analysis, Data curation. Lauren Rusnak: Writing – review & editing, Formal analysis, Data curation. Andrew Cagan: Writing – review & editing, Formal analysis, Data curation. Yumeko Kawano: Writing – review & editing, Investigation, Data curation. Gregory C. McDermott: Writing – review & editing, Investigation, Conceptualization. Tianrun Cai: Writing – review & editing, Software, Resources, Methodology, Data curation. Sara M Sweet: Writing – review & editing, Methodology, Data curation. Vidul Panickan: Writing – review & editing, Methodology, Data curation. Alexandra A. Josephson: Writing – review & editing, Methodology, Data curation. Tianxi Cai: Writing – review & editing, Methodology, Data curation. Damini Dey: Writing – review & editing, Resources, Conceptualization. Marcelo Di Carli: Writing – review & editing, Software, Methodology. Daniel H. Solomon: Writing – review & editing, Methodology, Conceptualization. Jennifer E. Ho: Writing – review & editing, Investigation, Conceptualization. Brittany Weber: Writing – review & editing, Methodology, Investigation, Data curation, Conceptualization. Katherine P. Liao: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Katherine Liao reports financial support was provided by National Institutes of Health. Katherine Liao reports a relationship with Merck & Co Inc that includes: consulting or advisory. Katherine Liao reports a relationship with UCB that includes: consulting or advisory. Katherine Liao reports a relationship with Bristol-Myers Squibb Company that includes: consulting or advisory. Brittany Weber reports a relationship with Novo Nordisk Inc that includes: consulting or advisory. Brittany Weber reports a relationship with Cleerly Inc that includes: consulting or advisory. Brittany Weber reports a relationship with Bristol-Myers Squibb Company that includes: consulting or advisory. Brittany Weber reports a relationship with Avalo Therapeutics Inc that includes: consulting or advisory. Brittany Weber reports a relationship with Eli Lilly that includes: consulting or advisory. Gregory McDermott reports a relationship with Boehringer Ingelheim Corp USA that includes: funding grants. Daniel Solomon reports a relationship with Amgen Inc that includes: funding grants. Daniel Solomon reports a relationship with Janssen Pharmaceuticals Inc that includes: funding grants. Daniel Solomon reports a relationship with Artiva Biotherapeutics Inc that includes: consulting or advisory. Daniel Solomon reports a relationship with American College of Rheumatology that includes: board membership. Daniel Solomon reports a relationship with CARRA that includes: board membership. Daniel Solomon reports a relationship with UptoDate Inc that includes: consulting or advisory. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.