Development and Validation of Machine Learning-Based Race-Specific Models to Predict 10-Year Risk of Heart Failure: A Multicohort Analysis.

BACKGROUND: Heart failure (HF) risk and the underlying risk factors vary by race. Traditional models for HF risk prediction treat race as a covariate in risk prediction and do not account for significant parameters such as cardiac biomarkers. Machine learning (ML) may offer advantages over traditional modeling techniques to develop race-specific HF risk prediction models and to elucidate important contributors of HF development across races. METHODS: We performed a retrospective analysis of 4 large, community cohort studies (ARIC [Atherosclerosis Risk in Communities], DHS [Dallas Heart Study], JHS [Jackson Heart Study], and MESA [Multi-Ethnic Study of Atherosclerosis]) with adjudicated HF events. The study included participants who were >40 years of age and free of HF at baseline. Race-specific ML models for HF risk prediction were developed in the JHS cohort (for Black race-specific model) and White adults from ARIC (for White race-specific model). The models included 39 candidate variables across demographic, anthropometric, medical history, laboratory, and electrocardiographic domains. The ML models were externally validated and compared with prior established traditional and non-race-specific ML models in race-specific subgroups of the pooled MESA/DHS cohort and Black participants of ARIC. The Harrell C-index and Greenwood-Nam-D'Agostino χ2 tests were used to assess discrimination and calibration, respectively. RESULTS: The ML models had excellent discrimination in the derivation cohorts for Black (n=4141 in JHS, C-index=0.88) and White (n=7858 in ARIC, C-index=0.89) participants. In the external validation cohorts, the race-specific ML model demonstrated adequate calibration and superior discrimination (Black individuals, C-index=0.80-0.83; White individuals, C-index=0.82) compared with established HF risk models or with non-race-specific ML models derived with race included as a covariate. Among the risk factors, natriuretic peptide levels were the most important predictor of HF risk across both races, followed by troponin levels in Black and ECG-based Cornell voltage in White individuals. Other key predictors of HF risk among Black individuals were glycemic parameters and socioeconomic factors. In contrast, prevalent cardiovascular disease and traditional cardiovascular risk factors were stronger predictors of HF risk in White adults. CONCLUSIONS: Race-specific and ML-based HF risk models that integrate clinical, laboratory, and biomarker data demonstrated superior performance compared with traditional HF risk and non-race-specific ML models. This approach identifies distinct race-specific contributors of HF.

Racial Differences in Malignant Left Ventricular Hypertrophy and Incidence of Heart Failure: A Multicohort Study.

BACKGROUND: A malignant subphenotype of left ventricular hypertrophy (LVH) has been described, in which minimal elevations in cardiac biomarkers identify individuals with LVH at high risk for developing heart failure (HF). We tested the hypothesis that a higher prevalence of malignant LVH among blacks may contribute to racial disparities in HF risk. METHODS: Participants (n=15 710) without prevalent cardiovascular disease were pooled from 3 population-based cohort studies, the ARIC Study (Atherosclerosis Risk in Communities), the DHS (Dallas Heart Study), and the MESA (Multi-Ethnic Study of Atherosclerosis). Participants were classified into 3 groups: those without ECG-LVH, those with ECG-LVH and normal biomarkers (hs-cTnT (high sensitivity cardiac troponin-T) <6 ng/L and NT-proBNP (N-terminal pro-B-type natriuretic peptide) <100 pg/mL), and those with ECG-LVH and abnormal levels of either biomarker (malignant LVH). The outcome was incident HF. RESULTS: Over the 10-year follow-up period, HF occurred in 512 (3.3%) participants, with 5.2% in black men, 3.8% in white men, 3.2% in black women, and 2.2% in white women. The prevalence of malignant LVH was 3-fold higher among black men and women versus white men and women. Compared with participants without LVH, the adjusted hazard ratio for HF was 2.8 (95% CI, 2.1-3.5) in those with malignant LVH and 0.9 (95% CI, 0.6-1.5) in those with LVH and normal biomarkers, with similar findings in each race/sex subgroup. Mediation analyses indicated that 33% of excess hazard for HF among black men and 11% of the excess hazard among black women was explained by the higher prevalence of malignant LVH in blacks. Of black men who developed HF, 30.8% had malignant LVH at baseline, with a corresponding population attributable fraction of 0.21. The proportion of HF cases occurring among those with malignant LVH, and the corresponding population attributable fraction, were intermediate and similar among black women and white men and lowest among white women. CONCLUSIONS: A higher prevalence of malignant LVH may in part explain the higher risk of HF among blacks versus whites. Strategies to prevent development or attenuate risk associated with malignant LVH should be investigated as a strategy to lower HF risk and mitigate racial disparities.

Differences in statin utilization and lipid lowering by race, ethnicity, and HIV status in a real-world cohort of persons with human immunodeficiency virus and uninfected persons.

BACKGROUND: Risks for cardiovascular diseases, including myocardial infarction and stroke, are elevated in people with HIV infection (PWH). However, no trials of statin utilization with clinical cardiovascular disease (CVD) end points have been completed in PWH, and there are sparse real-world data regarding statin use and lipid-lowering effectiveness. We therefore used a unique cohort of PWH and uninfected controls to evaluate (1) differences in statin types used for PWH versus uninfected persons; (2) lipid lowering achieved by statin use for PWH versus uninfected persons; and (3) racial and ethnic disparities in appropriate statin use among PWH and uninfected persons. METHODS: We analyzed a cohort of 5,039 PWH and 10,011 uninfected demographically matched controls who received care at a large urban medical center between January 1, 2000, and May 17, 2017. Medication administration records, prescription data, and validated natural language processing algorithms were used to determine statin utilization. Statins were categorized by generic active ingredient name and intensity (high, moderate, or low). Lipid values collected in routine clinical care were available for analysis. The first set of analyses was restricted to PWH and uninfected matched controls taking statins and compared (1) differences in statin type and (2) difference in cholesterol levels after versus before statin initiation by HIV status. For the second set of analyses, we first used prevalent CVD risk factors to determine participants with statin indications and then determined how many of these participants were taking statins. We then compared statin utilization among persons with indications for statins by race/ethnic group for PWH and uninfected matched controls using multivariable-adjusted logistic regression. RESULTS: Among people prescribed statins, PWH were more likely than controls to have ever taken pravastatin (34.8% vs 12.3%, P < .001) or atorvastatin (72.2% vs 65.6%, P = .002) and less likely to have ever taken simvastatin (14.2% vs 39.5%, P < .001). Among PWH with indications for statin utilization, 55.7% of whites, 39.4% of blacks, and 45.8% of Hispanics were prescribed statins (P < .001). These differences in statin prescription by race/ethnicity remained significant after adjustment for demographics (including insurance status), cardiovascular risk factors, antiretroviral therapy use, HIV viremia, and CD4 count. These racial/ethnic disparities in statin utilization were less pronounced among uninfected persons. CONCLUSIONS: Among PWH with statin indication(s), blacks and Hispanics were less likely than whites to have been prescribed a statin. These racial/ethnic disparities were less pronounced among uninfected persons. There were significant differences in type of statin used for PWH compared to uninfected matched controls. Future efforts addressing disparities in CVD prevention among PWH are warranted.

Comparative outcomes of transvenous extraction of sprint fidelis and riata defibrillator leads: a single center experience.

INTRODUCTION: The FDA has issued class I advisories for Medtronic Sprint Fidelis(®) and St. Jude Medical Riata(TM) ICD lead families. Transvenous Riata(TM) ICD lead extraction is typically considered higher risk than Fidelis(®) extraction, based on longer duration from implant, presence of externalized conductors and lack of silicone backfill in the SVC and RV coils. However, published data comparing procedural outcomes between these leads are limited. METHODS: Records were reviewed for all patients undergoing transvenous extraction of Sprint Fidelis(®) or Riata(TM) ICD leads at the Vanderbilt Heart and Vascular Institute from July 2006 to April 2013 to ascertain indication for extraction, procedural details, complications, and 30-day mortality. RESULTS: There were significant differences between those undergoing extraction of a Sprint Fidelis(®) (n = 145) or Riata(TM) lead (n = 47). In the Riata(TM) group, device-related endocarditis was a more common indication for extraction, the mean duration of implant was longer, and larger excimer laser sheaths were required. Lead malfunction was a more common indication in the Fidelis(®) group. There were no statistically significant differences in median procedure duration, procedural success (97.9% vs 95.7%, P = 0.41), median length of hospital stay (1 day vs 1 day, P = 0.23), procedural complication rate (5.5% vs 10.6%, P = 0.23) or 30-day mortality (2.1% vs 2.1%, P = 0.98). Analyses excluding patients with device infection revealed similar results. CONCLUSION: Despite differences in baseline characteristics, this study indicates that Medtronic Sprint Fidelis(®) and St. Jude Riata(TM) ICD leads have similar procedural outcomes with transvenous lead extraction.

Corrigendum: Rationale and Design of a Pharmacist-led Intervention for the Risk-Based Prevention of Heart Failure: The FIT-HF Pilot Study.

[This corrects the article DOI: 10.3389/fcvm.2021.785109.].

Factors associated with baseline and serial changes in circulating NT-proBNP and high-sensitivity cardiac troponin T in a population-based cohort (Dallas Heart Study).

Aim: N-terminal pro-B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin T (hs-cTnT) associate with structural heart disease and heart failure risk in individuals without known cardiovascular disease (CVD). However, few data are available regarding whether factors influencing levels of these two biomarkers are similar or distinct. We performed serial measurement of NT-proBNP and hs-cTnT in a contemporary multiethnic cohort with extensive phenotyping, with the goal of identifying their respective biological determinants in a population without known or suspected CVD. Methods: We evaluated 1877 participants of the Dallas Heart Study who had NT-proBNP and hs-cTnT measured and were free from clinical CVD at the each of its two examinations (2000-2002 and 2007-2009). Variables collected included demographic and risk factors, high-sensitivity C-reactive protein, body composition via dual-energy x-ray absorptiometry, coronary artery calcium by computed tomography, and cardiac dimensions and function by cardiac MRI. Linear regression was used to identify associations of these factors with each biomarker at baseline and with changes in biomarkers over follow-up. Results: NT-proBNP and hs-cTnT were poorly correlated at baseline (Spearman rho 0.083, p = 0.015), with only moderate correlation between change values (rho 0.18, p < 0.001). hs-cTnT positively associated and NT-proBNP inversely associated with male gender and black race. At baseline, both NT-proBNP and hs-cTnT associated with left ventricular end-diastolic volume and wall thickness, but only NT-proBNP associated with left atrial size. Changes in cardiac dimensions between phases were more strongly associated with changes in NT-proBNP than hs-cTnT. NT-proBNP was more strongly associated with high-sensitivity C-reactive protein and measures of body composition than hs-cTnT. Conclusion: Among individuals without CVD in the general population, NT-proBNP and hs-cTnT are nonredundant biomarkers that are differentially associated with demographic and cardiac factors. These findings indicate that hs-cTnT and NT-proBNP may reflect different pathophysiological pathways.

Rationale and Design of a Pharmacist-led Intervention for the Risk-Based Prevention of Heart Failure: The FIT-HF Pilot Study.

Background: Given rising morbidity, mortality, and costs due to heart failure (HF), new approaches for prevention are needed. A quantitative risk-based strategy, in line with established guidelines for atherosclerotic cardiovascular disease prevention, may efficiently select patients most likely to benefit from intensification of preventive care, but a risk-based strategy has not yet been applied to HF prevention. Methods and Results: The Feasibility of the Implementation of Tools for Heart Failure Risk Prediction (FIT-HF) pilot study will enroll 100 participants free of cardiovascular disease who receive primary care at a single integrated health system and have a 10-year predicted risk of HF of ≥5% based on the previously validated Pooled Cohort equations to Prevent Heart Failure. All participants will complete a health and lifestyle questionnaire and undergo cardiac biomarker (B-type natriuretic peptide [BNP] and high-sensitivity cardiac troponin I [hs-cTn]) and echocardiography screening at baseline and 1-year follow-up. Participants will be randomized 1:1 to either a pharmacist-led intervention or usual care for 1 year. Participants in the intervention arm will undergo consultation with a pharmacist operating under a collaborative practice agreement with a supervising cardiologist. The pharmacist will perform lifestyle counseling and recommend initiation or intensification of therapies to optimize risk factor (hypertension, diabetes, and cholesterol) management according to the most recent clinical practice guidelines. The primary outcome is change in BNP at 1-year, and secondary and exploratory outcomes include changes in hs-cTn, risk factor levels, and cardiac mechanics at follow-up. Feasibility will be examined by monitoring retention rates. Conclusions: The FIT-HF pilot study will offer insight into the feasibility of a strategy of quantitative risk-based enrollment into a pharmacist-led prevention program to reduce heart failure risk. Clinical Trial Registration: https://clinicaltrials.gov/ct2/show/NCT04684264.