Abstract 106: Predicting Pulmonary Hypertension Readmission Using Machine Learning Models

Objectives: Pulmonary hypertension (PH) is increasingly recognized but understudied. We sought to determine predictors of 30-day readmission after PH-related hospitalization using Driverless Artificial Intelligence (DAI) algorithm.

Methods: We utilized the 2012-2014 Healthcare Cost and Utilization Project Nationwide Readmission Database (NRD) that accounts for weighted estimates of roughly 35 million discharges in the United States. Adult patients with primary ICD-9-CM diagnosis codes of 416.0 and 416.8 for primary and secondary PH with an index admission and any readmission within 30-days of the index event were identified. We assessed predictors of all cause 30-day readmission from clinical, hospital and 29 AHRQ comorbidity measures using predictive modeling to provide interpretable risk factors globally at population level and locally associated with each discharge. The risk prediction model was developed using DAI algorithm and results compared to models based on LIME GLM, Random Forest, and Decision Tree. Overall model performance was assessed using concordance statistic.

Results: Data from patients with 14,659 admissions and 2797 30-day readmissions were used to train the model. After data processing, the final model included 3674 variables. The Random Forest model had the best performance with c-statistic of 0.70 and surrogate models compared at AUCs of 0.51(Decision Tree), 0.62 (LIME) and (DAI) 0.59. Global features that mainly important to the overall prediction of DAI model were deficiency anemia, APR-DRG severity risk, APR-DRG mortality risk, renal failure and CHF.

Conclusion: Using predictive models with high predictive power, interpretable risk factors and prediction accuracy may enable health care systems to accurately target high-risk PH patients and prevent recurrent readmissions.

Influence of altitude on hypertension phenotypes and responses to antihypertensive therapy: Review of the literature and design of the INTERVENCION trial

Systemic arterial hypertension constitutes the leading cause of mortality worldwide, and affects people living at different altitudes above sea level (AASL). AASL has a major impact on cardiovascular function and various biologic pathways that regulate blood pressure-related phenotypes, but whether it affects the clinical response to antihypertensive therapy is unknown. The hemodynamic adaptations observed among lowlanders acutely exposed to high altitude (HA) is distinct from those observed among HA dwellers. However, the phenotypic patterns of hypertension and the response to standard antihypertensive agents among adults chronically exposed to different AASL are poorly understood. The authors describe the protocol for the INTERVENCION trial, a randomized clinical trial designed to assess the effects of three first-line antihypertensive monotherapies (a thiazide diuretic, an angiotensin receptor blocker, and a calcium channel blocker) on peripheral and central blood pressure, in-office blood pressure, and ambulatory blood pressure hemodynamics of hypertensive patients living at different AASL (low altitude, intermediate altitude, and high altitude). The primary end point is the reduction in 24-hour brachial systolic blood pressure. The INTERVENCION trial will provide the first clinical trial data regarding the influence of AASL on the response to antihypertensive monotherapy, as well as the hemodynamic characteristics of arterial hypertension at different AASL.

Proteomic Associations of NT-proBNP (N-Terminal Pro-B-Type Natriuretic Peptide) in Heart Failure With Preserved Ejection Fraction

BACKGROUND

NT-proBNP (N-terminal pro-B-type natriuretic peptide) levels are variably elevated in heart failure with preserved ejection fraction (HFpEF), even in the presence of increased left ventricular filling pressures. NT-proBNP levels are prognostic in HFpEF and have been used as an inclusion criterion for several recent randomized clinical trials. However, the underlying biologic differences between HFpEF participants with high and low NT-proBNP levels remain to be fully understood.

METHODS

We measured 4928 proteins using an aptamer-based proteomic assay (SOMAScan) in available plasma samples from 2 cohorts: (1) Participants with HFpEF enrolled in the PHFS (Penn Heart Failure Study; n=253); (2) TOPCAT (Treatment of Preserved Cardiac Function Heart Failure With an Aldosterone Antagonist Trial) participants in the Americas (n=218). We assessed the relationship between SOMAScan-derived plasma NT-proBNP and levels of other proteins available in the SOMAScan assay version 4 using robust linear regression, with correction for multiple comparisons, followed by pathway analysis.

RESULTS

NT-proBNP levels exhibited prominent proteome-wide associations in PHFS and TOPCAT cohorts. Proteins most strongly associated with NT-proBNP in both cohorts included SVEP1 (sushi, von Willebrand factor type-A, epidermal growth factor, and pentraxin domain containing 1; βTOPCAT=0.539; P<0.0001; βPHFS=0.516; P<0.0001) and ANGPT2 (angiopoietin 2; βTOPCAT=0.571; P<0.0001; βPHFS=0.459; P<0.0001). Canonical pathway analysis demonstrated consistent associations with multiple pathways related to fibrosis and inflammation. These included hepatic fibrosis and inhibition of matrix metalloproteases. Analyses using cut points corresponding to estimated quantitative concentrations of 360 pg/mL (and 480 pg/mL in atrial fibrillation) revealed similar proteomic associations.

CONCLUSIONS

Circulating NT-proBNP levels exhibit prominent proteomic associations in HFpEF. Our findings suggest that higher NT-proBNP levels in HFpEF are a marker of fibrosis and inflammation. These findings will aid the interpretation of NT-proBNP levels in HFpEF and may guide the selection of participants in future HFpEF clinical trials.

Urinary Proteomics and Outcomes in Heart Failure With Preserved Ejection Fraction

BACKGROUND

Although several studies have addressed plasma proteomics in heart failure with preserved ejection fraction, limited data are available on the prognostic value of urinary proteomics. The objective of our study was to identify urinary proteins/peptides associated with death and heart failure admission in patients with heart failure with preserved ejection fraction.

METHODS AND RESULTS

The study population included participants enrolled in TOPCAT (Treatment of Preserved Cardiac Function Heart Failure With an Aldosterone Antagonist Trial). The relationship between urine protein levels and the risk of death or heart failure admission was assessed using Cox regression, in both nonadjusted analyses and adjusting for urine creatinine levels, and the MAGGIC (Meta-Analysis Global Group in Chronic Heart Failure) score. A total of 426 (12.4%) TOPCAT participants had urinary protein data and were included. There were 40 urinary proteins/peptides significantly associated with death or heart failure admission in nonadjusted analyses, 21 of which were also significant adjusted analyses. Top proteins in the adjusted analysis included ANGPTL2 (angiopoietin-like protein 2) (hazard ratio [HR], 0.5731 [95% CI, 0.47-0.7]; P=3.13E-05), AMY2A (α amylase 2A) (HR, 0.5496 [95% CI, 0.44-0.69]; P=0.0001), and DNASE1 (deoxyribonuclease-1) (HR, 0.5704 [95% CI, 0.46-0.71]; P=0.0002). Higher urinary levels of proteins involved in fibrosis (collagen VI α-1, collagen XV α-1), metabolism (pancreatic α-amylase 2A/B, mannosidase α class 1A member 1), and inflammation (heat shock protein family D member 1, inducible T cell costimulatory ligand) were associated with a lower risk of death or heart failure admission.

CONCLUSIONS

Our study identifies several novel associations between urinary proteins/peptides and outcomes in heart failure with preserved ejection fraction. Many of these associations are independent of clinical risk scores and may aid in risk stratification in this patient population.

Prevalence of mitral annulus calcification in African Americans: comparison with non-Hispanic whites and Hispanics

BACKGROUND

The association of ethnic ancestry with coronary artery calcifications suggests that mitral annulus calcification may also vary with ethnicity. We sought to compare prevalence and clinical correlates of mitral annulus calcification in non-Hispanic Whites, Hispanics, and African Americans.

DESIGN

This was a retrospective study of 857 patients age 40-75 years that included 217 (25%) African Americans, 349 (41%) Hispanics, and 291 (34%) non-Hispanic Whites referred for echocardiography. Multiple logistic regression was used to determine the interrelationships between mitral annulus calcification, risk factors, and ethnicity.

RESULTS

Mitral annulus calcification was detected in 181 (21.1%) patients including 35 (16.1%) African Americans, 80 (22.9%) Hispanics, and 66 (22.7%) non-Hispanic whites. In univariate analysis, patients with mitral annulus calcification were older and more likely to have hypertension, diabetes, dyslipidemia, smoking history, and two or more risk factors than were those without calcification. In multivariate analysis, age and smoking history were independent predictors of mitral annulus calcification; dyslipidemia and diabetes were borderline significant predictors; and after adjusting for the remaining variables in the model, ethnicity was not an independent significant predictor of mitral annulus calcification.

CONCLUSION

In a retrospective study of middle-aged and elderly African Americans, non-Hispanic Whites, and Hispanics referred for echocardiography, mitral annulus calcification is common in all three major ethnic groups but not significantly associated with ethnic ancestry.

LDL-c Lowering, Ischemic Stroke, and Small Vessel Disease Brain Imaging Biomarkers: A Mendelian Randomization Study

BACKGROUND

The effects of lipid-lowering drug targets on different ischemic stroke subtypes are not fully understood. We aimed to explore the mechanisms by which lipid-lowering drug targets differentially affect the risk of ischemic stroke subtypes and their underlying pathophysiology.

METHODS

Using a 2-sample Mendelian randomization approach, we assessed the effects of genetically proxied low-density lipoprotein cholesterol (LDL-c) and 3 clinically approved LDL-lowering drugs (HMGCR [3-hydroxy-3-methylglutaryl-CoA reductase], PCSK9 [proprotein convertase subtilisin/kexin type 9], and NPC1L1 [Niemann-Pick C1-Like 1]) on stroke subtypes and brain imaging biomarkers associated with small vessel stroke (SVS), including white matter hyperintensity volume and perivascular spaces.

RESULTS

In genome-wide Mendelian randomization analyses, lower genetically predicted LDL-c was significantly associated with a reduced risk of any stroke, ischemic stroke, and large artery stroke, supporting previous findings. Significant associations between genetically predicted LDL-c and cardioembolic stroke, SVS, and biomarkers, perivascular space and white matter hyperintensity volume, were not identified in this study. In drug-target Mendelian randomization analysis, genetically proxied reduced LDL-c through NPC1L1 inhibition was associated with lower odds of perivascular space (odds ratio per 1-mg/dL decrease, 0.79 [95% CI, 0.67-0.93]) and with lower odds of SVS (odds ratio, 0.29 [95% CI, 0.10-0.85]).

CONCLUSIONS

This study provides supporting evidence of a potentially protective effect of LDL-c lowering through NPC1L1 inhibition on perivascular space and SVS risk, highlighting novel therapeutic targets for SVS.

Prognostic Value of Cardio-Ankle Vascular Index for Cardiovascular and Kidney Outcomes: Systematic Review and Meta-Analysis

Background

Arterial stiffness causes cardiovascular disease and target-organ damage. Carotid-femoral pulse wave velocity is regarded as a standard arterial stiffness metric. However, the prognostic value of cardio-ankle vascular index (CAVI), which is mathematically corrected for blood pressure, remains understudied.

Objectives

The purpose of this study was to determine the association of CAVI with cardiovascular and kidney outcomes.

Methods

PubMed, Scopus, and Web of Science were searched until May 6, 2023, for longitudinal studies reporting the association of CAVI with mortality, cardiovascular events (CVEs) (including death, acute coronary syndromes, stroke, coronary revascularization, heart failure hospitalization), and kidney function decline (incidence/progression of chronic kidney disease, glomerular filtration rate decline). Random-effects meta-analysis was performed. Studies were assessed with the "Quality in Prognostic Studies" tool.

Results

Systematic review identified 32 studies (105,845 participants; follow-up range: 12-148 months). Variable cutoffs were reported for CAVI. The risk of CVEs was higher for high vs normal CAVI (HR: 1.46 [95% CI: 1.22-1.75]; P < 0.001; I2 = 41%), and per SD/unit CAVI increase (HR: 1.30 [95% CI: 1.20-1.41]; P < 0.001; I2 = 0%). Among studies including participants without baseline cardiovascular disease (primary prevention), higher CAVI was associated with first-time CVEs (high vs normal: HR: 1.60 [95% CI: 1.15-2.21]; P = 0.005; I2 = 65%; HR per SD/unit increase: 1.28 [95% CI: 1.12-1.47]; P < 0.001; I2 = 18%). There was no association between CAVI and mortality (HR = 1.31 [0.92-1.87]; P = 0.130; I2 = 53%). CAVI was associated with kidney function decline (high vs normal: HR = 1.30 [1.18-1.43]; P < 0.001; I2 = 38%; HR per SD/unit increase: 1.12 [95% CI: 1.07-1.18]; P < 0.001; I2 = 0%).

Conclusions

Higher CAVI is associated with incident CVEs, and this association is present in the primary prevention setting. Elevated CAVI is associated with kidney function decline.

Prognostic Significance and Biologic Associations of Senescence-Associated Secretory Phenotype Biomarkers in Heart Failure

BACKGROUND

The role of cellular senescence in human heart failure (HF) remains unclear. The senescence-associated secretory phenotype (SASP) is composed of proteins released by senescent cells. We assessed the prognostic significance and biologic pathways associated with the SASP in human HF using a plasma proteomics approach.

METHODS AND RESULTS

We measured 25 known SASP proteins among 2248 PHFS (Penn HF Study) participants using the SOMAScan V4 assay. We extracted the common variance in these proteins to generate SASP factor scores and assessed the relationship between these SASP factor scores and (1) all-cause death and (2) the composite of death or HF hospital admission. We also assessed the relationship of each SASP factor to 4746 other proteins, correcting for multiple comparisons, followed by pathway analyses. Two SASP factors were identified. Both factors were associated with older age, lower estimated glomerular filtration rate, and more advanced New York Heart Association class, among other clinical variables. Both SASP factors exhibited a significant positive association with the risk of death independent of the Meta-Analysis of Global-Group in Chronic HF score and NT-proBNP (N-terminal pro-B-type natriuretic peptide) levels. The 2 identified SASP factors were associated with 1201 and 1554 proteins, respectively, belonging to various pathways including the coagulation system, complement system, acute phase response signaling, and retinoid X receptor-related pathways that regulate cell metabolism.

CONCLUSIONS

Increased SASP components are independently associated with adverse outcomes in HF. Biologic pathways associated with SASP are predominantly related to coagulation, inflammation, and cell metabolism.