Exploring the PREVENT HF Score and Myocardial Function among Persons with HIV

Summary

Objective:

Persons with HIV(PWH) are at risk for myocardial structural changes, which can progress to diastolic dysfunction and heart failure with preserved ejection fraction(HFpEF). We explored the AHA PREVENT HF(Predicting Risk of cardiovascular disease EVENTs for Heart Failure) risk score in relation to cardiac magnetic resonance(CMR) imaging.

Design:

This cross-sectional study included 37 PWH on ART, ages 40–65, without known CVD who underwent CMR.

Methods:

The risk score was assessed using the AHA PREVENT HF calculator. Scores were correlated to variables on CMR that are known indicators of subclinical myocardial dysfunction [left atrial volume index(LAVI), global longitudinal strain(GLS), and left ventricular mass index(LVMI)] and inflammation[extracellular volume(ECV) and longitudinal relaxation (T1)].

Results:

PWH were age 55(6)years[mean (SD)], predominantly male(76%) and white(57%) with BMI in the obese(≥30kg/m²) range: 31(5)kg/m². Median PREVENT HF score was 2.6(1.4,4.1)% [median(25^th, 75^th)]. The PREVENT HF score correlated to LAVI(ρ=0.35, P=.04), T1(ρ=0.35, P=.04), IL-6(ρ=0.36, P=.03) and NT-proBNP(ρ=0.42, P=.01). Risk scores were higher for those meeting clinical cutoffs LAVI>34 mL/m² and T1≥1250 ms. For predicting LAVI >34 mL/m², a PREVENT HF score 2.5 was the optimal cutoff[sensitivity 85%, specificity 65%, AUROC 0.769 (P<.05)]. In predicting T1≥1250 ms, a PREVENT HF score 3.6 was the optimal cutoff[71% sensitivity, 95% specificity, AUROC 0.727 (P<.05)].

Conclusion:

The PREVENT HF score related to indices of altered myocardial structure and inflammation among asymptomatic PWH with subclinical disease. These data begin to inform us about the utility of PREVENT HF score using radiographic findings, though more studies are needed among PWH to validate its use as a prediction tool.

Introduction

Similar to atherosclerotic disease risk, the relative risk of diastolic dysfunction(DD) and heart failure(HF) are increased approximately 1.2 to 2-fold among persons with HIV(PWH).^{[1, 2]} Studies have shown PWH to be at higher risk for HF with preserved ejection(HFpEF) compared to persons without HIV(PWOH).^[2] While HIV and HIV-specific factors have been associated with HFpEF risk,^{[2, 3]} metabolic co-morbidities are being increasingly recognized as important mediators of HFpEF.^[4] Obesity, hypertension, and diabetes can be associated with changes in cardiac function and structure due to myocardial inflammation, intramyocellular lipid accumulation, or fibrosis.^[4–7] The newly developed American Heart Association (AHA) PREVENT HF risk score may be a relevant application, as it takes into account key metabolic parameters which tend to be abnormal in many PWH, such as body mass index(BMI), HbA1c, and estimated glomerular filtration rate(eGFR). Currently, there are no validated tools in HIV to predict who may clinically progress to HFpEF, though we know structural changes in the left atrium and left ventricle are indicative of future progression to DD and HFpEF.^[8] In the current study, we explored the PREVENT HF score in relation to myocardial indices on cardiac magnetic resonance(CMR) imaging among PWH with no known history of HF.

Methods

Study participants

The current study was leveraged from baseline data of the Mineralocorticoid Receptor Antagonism for Cardiovascular Health in HIV(MIRACLE HIV) Study,^[9] which investigated a strategy to improve subclinical myocardial disease in HIV and was conducted at the Massachusetts General Hospital and Brigham and Women’s Hospital from February 2017-March 2022. Participants were included for ages 40–65, continuous ART use>12 months, viral load<100 copies/mL and increased visceral adiposity(VAT>110cm²) on screening CT. Thirty-seven participants had baseline CMR data available. Participants were excluded for known CVD or HF. In addition, HbA1c≥7.5%, BP>140/90mmHg, and kidney impairment(Cr>1.5 mg/dL or eGFR<60 mL/min/1.73m²) were not allowed. This study was approved by the Mass General Brigham Human Research Committee. All participants provided informed consent.

AHA PREVENT HF Score

The HF risk score was calculated using the AHA PREVENT(Predicting Risk of CVD EVENTs) calculator.^[10] Baseline fasting data were used as inputs. The following variables were required: sex(male/female), age, total cholesterol, high density lipoprotein(HDL) cholesterol, systolic blood pressure(SBP), BMI, eGFR as well as diabetes(yes/no), current smoking(yes/no), anti-hypertensive medication(yes/no), lipid-lowering medication(yes/no). The optional variables were not included in this analysis.

CMR Imaging

CMR was performed using a 3-T magnet(Magnetom Tim Trio or Prisma-Fit; Siemens Healthcare) with an 18-element body phased-array surfacecoil as previously described in our studies.^[9] (see also Supplemental Methods)

Statistical analysis

Data are reported as mean(standard deviation) for normally distributed variables and median(Q1:25^th percentile, Q3:75^th percentile) for non-normally distributed variables. Categorical variables are reported as percentages. Spearman’s correlations(ρ) were used to measure associations between all continuous variables. CMR indices were converted to dichotomous variables based on clinically relevant cutoffs for subclinical myocardial dysfunction which are relevant to the development of HFpEF[LAVI>34 ml/m², GLS≤18%, LV mass index(LVMI)≥95 g/m² for females or ≥115 g/m² for males]^[11–16] and inflammation[T1≥1250ms performed at 3T, ECV≥0.25]^{[17, 18]}. Cutoffs for cardiac abnormalities are based on values set by the American Society of Echocardiography and the European Association of Cardiovascular Imaging^[14]. We utilized these cutoffs for diastolic function, as have been shown to be comparable between echocardiography and CMR.^{[11, 13]} Between group comparisons were made using the Kruskal-Wallis test. ROC curves were generated for cutoffs of LAVI, GLS, T1, and ECV to understand the performance of the PREVENT HF score. No individuals had evidence of increased LVMI, so an ROC curve could not be assessed. Analyses were performed using JMP version 17.

Results

Clinical Characteristics

Participants were age 55(6)years[mean(SD)] and predominantly male(76%) and white(57%). 19% of the 37 participants were of Hispanic ethnicity. Median ASCVD 10-year risk was borderline 6.2%(3.4,10.1) [median (25^th, 75^th)]. PREVENT HF risk score was 2.6%(1.4,4.1). 19% of PWH were current tobacco users or had hypertension. Current statin use was 22%. 100% of PWH were on nucleoside reverse transcriptase inhibitors(NRTIs), 46% on non-nucleoside reverse transcriptase inhibitors(NNRTIs), and 64% on integrase strand transfer inhibitors(InSTIs).(Supplemental Table 1) Combinations of current regimens are shown in Supplemental Table 2. 92% of PWH had been on more than one ART regimen following HIV diagnosis. CD4+ T-cell count was 792(644,933)cells/μL and 78% had undetectable HIV viral load. Duration of HIV infection was 20(7)years. Based on the parent study, there was no evidence of DM with HbA1c 5.5(5.3,6.0)%. Mean BMI was in the obese range 31(5)kg/m². Lipid panel showed total cholesterol 186(37)mg/dL and HDL 47(44,54)mg/dL. Median SBP was 126(121,135)mmHg and eGFR was 87(13) mL/min/1.73m².(Supplemental Table 1)

Relationship of PREVENT HF Score to Myocardial Indices and Inflammation

The PREVENT HF score related to CMR parameters: LAVI(ρ=0.35, p=.04) and T1 relaxation time(ρ=0.35, P=.04).(Supplemental Table 3) No significant correlation was observed with LVMI, GLS, or ECV. PREVENT HF scores were significantly higher among those with LAVI>34 mL/m² [3.6(2.6,6.6)] vs. LAVI≤34 mL/m²[1.8(1.3,3.3), P=.008). Similarly, participants with T1 relaxation time≥1250 ms[4.1(1.4,6.1)] had significantly higher PREVENT HF scores vs. those with T1<1250 ms[1.8(1.4,2.6), P=.03]. There tended to be higher PREVENT HF among those with ECV≥0.25 compared to those with ECV<0.25.(Table 1) The PREVENT HF score correlated with IL-6(ρ=0.36, P=.03) and NT-proBNP (ρ=0.42, P=.01). The PREVENT HF score did not relate to HIV-related parameters (Supplemental Table 3) and did not differ by unsuppressed vs. suppressed viral load [2.0%(1.0,3.9) vs. 2.6%(1.5,4.4), P=0.43].

Table 1.

Comparison of PREVENT HF scores by clinically significant CMR parameters

		P Value
LAVI≤34ml/m2 (n=18)	LAVI >34ml/m2 (n=18)
1.8(1.3, 3.3)	3.6(2.6, 6.6)	.008
GLS>18% (n=19)	GLS≤18% (n=15)
3.4(1.4, 4.8)	2.1(1.5, 3.6)	.61
T1<1250ms (n=19)	T1≥1250ms (n=14)
1.8(1.4, 2.6)	4.1(1.4, 6.1)	.03
ECV<0.25 (n=9)	ECV≥0.25 (n=24)
1.5(1.3, 2.4)	3.0(1.5, 4.9)	.07

Data are presented as median(Q1: quartile 1 or 25th percentile,Q3: quartile 3 or 75th percentile).

Abbreviations: CMR, cardiac magnetic resonance; LAVI, left atrial volume index; GLS, global longitudinal strain; T1, T1 relaxation time; ECV, extracellular volume

PREVENT HF Score to Predict CMR Parameters

For predicting LAVI>34 mL/m², the best cutoff value for the PREVENT HF score was 2.5[sensitivity 85%, specificity 65%, AUROC 0.769(p<.05)]. In predicting T1 relaxation time≥1250 ms, the optimal PREVENT HF score cutoff was 3.6[71% sensitivity, 95% specificity, AUROC 0.727(P<.05)]. For ECV ≥0.25, a PREVENT HF score cutoff of 1.7 was best[75% sensitivity, 67% specificity, AUROC 0.713 (P<0.05)]. The PREVENT HF was not predictive of GLS.(Table 2, Supplemental Figure 1)

Table 2.

Efficacy of PREVENT HF score in predicting clinically significant CMR parameters

	Best Cutoff Values	Sensitivity (%)	Specificity (%)	AUROC
LAVI>34 mL/m2	2.5	85	65	0.769*
GLS≤18%	3.3	73	53	0.553
T1≥1250ms	3.6	71	95	0.727*
ECV≥0.25	1.7	75	67	0.713*

Abbreviations: CMR, cardiac magnetic resonance; LAVI, left atrial volume index; GLS, global longitudinal strain; T1, T1 relaxation time; ECV, extracellular volume; AUROC; area under the receiver operating characteristic curve

^*Indicates model is significant P<.05

Discussion

In this study, we saw that the AHA PREVENT HF risk score correlated to myocardial structure(LAVI) and inflammation(T1) on CMR and appeared to be higher among those with abnormal LAVI and T1. We were able to detect PREVENT HF scores that may predict abnormal LAVI, T1, and ECV based on clinically accepted ranges.^{[12–15, 18, 19]} While the derived PREVENT HF cutoffs are not validated, they begin to help us understand the utility of the PREVENT HF score in predicting early changes in the myocardium before symptoms are present.

Studies of PWH have shown increased LAVI compared to PWOH on CMR^[19]. Abnormal LAVI is a marker of early diastolic dysfunction and mortality. In this regard, findings of abnormal LAVI on cardiac imaging predicted a 6-fold increase in mortality compared to normal cardiac imaging among community-dwelling individuals suspected to have HF^[20]. LAVI>34, the cutoff applied in the current study, has been associated with increased mortality compared to those with LAVI≤34^[21] with preserved EF. The LAVI>34 cutoff has been used for both echocardiography and CMR and shown to be comparable across imaging techniques.^[11–13] Higher LAVI has also been associated with proteins important to TNF signaling and extracellular matrix organization among PWH, and these proteins were further predictive of incident HF.^[22] To that end, not only did the PREVENT HF score correlate with LAVI in this study, but also systemic inflammation and CMR indices relevant to inflammation and fibrosis. T1 and ECV are non-specific indicators of inflammation and fibrosis, and similar to LAVI, are increased among PWH.^{[23, 24]} In the current study, none of the PWH met the criteria for increased LVMI, so we were unable to look at clinical cutoffs.

While the PREVENT HF calculator does not account for HIV-specific information, it does account for additional variables that may be critical metabolic drivers of heart disease in HIV. In that regard, the PREVENT HF risk score uniquely requires BMI as one of the inputs. Excess BMI is associated with a chronic inflammatory milieu and alterations in adipokines with increases in TNF-α, IL-6, IL-1β, MCP-1, and a reduction in adiponectin. This inflammation can be a driver of endothelial dysfunction, coronary microvascular dysfunction and myocardial fibrosis, potential mechanisms of which may lead to a HFpEF phenotype in the absence of atherosclerotic disease.^{[25, 26]} Moreover, there is a growing prevalence of obesity among PWH, which may be related to initiation of weight-promoting ART(INSTIs) or withdrawal of weight-suppressive ART(TDF or efavirenz).^[27] In the MESA study of community dwelling individuals without known CVD, body compositions, such as BMI, waist circumference, and visceral fat, were higher among those who developed incident HFpEF.^[28]

There are few other calculators used for HF risk prediction. The H2FPEF score requires that input of variables from echocardiography, which was not available in our study, and was validated to discern cardiac vs. non-cardiac causes of dyspnea among symptomatic patients.^[29] Other available HF calculators are primarily used to predict prognosis after a known diagnosis of HF(MAGGIC, BCN Bio-HF)^{[30, 31]} and require input of clinical signs and symptoms, and cannot be directly compared to the PREVENT HF whose purpose is for primary prevention in asymptomatic individuals.

A strength of this study was that we were able to ascertain objective measures and phenotype subclinical myocardial disease using imaging techniques. Furthermore, we excluded individuals with known symptomatic CVD or HF, as the PREVENT HF score was intended to be used as a prevention tool before progression to clinical disease. In addition, by study design ACEi, ARB, and mineralocorticoid receptors blockers were not allowed, which could have otherwise confounded myocardial measures as these classes of medications are indicated in clinical HF.

There were limitations of the study. These exploratory data were leveraged from a completed study which had cardiac MRI available, whose primary design was to understand subclinical cardiovascular disease among those PWH on ART.^[9] Individuals were recruited for specific body composition, glycemic control, and kidney function, as the inclusion/exclusion criteria were designed to address the aims of the parent study. These data are not representative of all PWH and may underpredict or overpredict PREVENT HF scores. A larger prospective study enrolling a more diverse cohort with more generalizable characteristics, inclusive of a broader range of age, BMI, metabolic disease, and immunologic control, for example, will inform us as to how the AHA PREVENT HF risk score performs among PWH and should be used to generate clinically significant cutoffs. Our findings are preliminary and need to be replicated in multicenter studies to enhance validation.

The AHA PREVENT HF calculator is the one of the first primary prevention tools and differs from traditional CVD calculators by focusing on HF risk. While this clinical tool is not tailored to HIV, it considers important metabolic variables highly relevant to PWH. If we can validate cutoffs for the PREVENT HF score in HIV, this may provide a risk stratification tool to potentially help identify important early structural and inflammatory changes in the myocardium among PWH to direct preventative care before symptomatic HF develops. More studies are needed to inform us as to the performance and utility of the PREVENT HF score in clinical practice.

Data Sharing Statement

Datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.