Exercise Tests To Detect Early Heart Failure Markers
Overview
Early identification of heart failure (HF) risk is crucial, and current methods require enhancement. This study explored the association between various cardiopulmonary exercise testing (CPET) phenotypes and subclinical markers of HF risk in a large, community-based cohort.
The analysis included 2,532 participants from the Framingham Heart Study, with an average age of 53 years, 52% being women, and a body mass index (BMI) of 28.0 ± 5.3 kg/m². Peak oxygen uptake (VO₂) was measured, showing 21.1 ± 5.9 kg/m² in women and 26.4 ± 6.7 kg/m² in men. Notably, higher peak VO₂ levels were linked to a reduced estimated HF risk, with correlations of -0.60 in men and -0.55 in women (P < 0.0001). However, there was overlap in estimated risk across VO₂ categories.
Through hierarchical clustering of 26 CPET-derived phenotypes (adjusted for age, sex, and BMI), three distinct exercise physiology clusters were identified. These clusters were similar in age, sex distribution, and BMI but had unique associations with subclinical markers. Cluster 1 was characterized by impaired oxygen kinetics, higher levels of subcutaneous and visceral fat, and reduced pulmonary function. Cluster 2 showed impaired vascular function and increased carotid-femoral pulse wave velocity (CFPWV). In contrast, Cluster 3 exhibited favorable exercise responses, lower CFPWV, reduced C-reactive protein and fat volumes, and better lung function (all false discovery rate < 5%).
Cluster membership contributed significant explanatory value (adjusted R² increment of 0.10 in both sexes, P < 0.0001) beyond what was offered by peak VO₂ alone when predicting HF risk. These findings suggest that CPET-derived phenotypes may provide important physiological insights, largely independent of traditional risk factors, potentially uncovering alternative pathways for heart failure prevention.
Introduction
Heart failure (HF) is a significant and growing global public health issue, with an estimated prevalence of 64 million people worldwide, a figure expected to continue rising. New treatments, such as nonsteroidal mineralocorticoid receptor antagonists and sodium-glucose cotransporter 2 inhibitors, have underscored the urgency of identifying HF risk earlier in its progression. While HF is typically described as progressing from risk factors (stage A) to subclinical cardiac abnormalities (stage B) and then to clinical heart failure (stages C/D), current methods for identifying risk are largely based on standard clinical risk assessments at stage A. Although subclinical imaging and blood biomarkers (stage B characteristics) can enhance risk assessment, their clinical application is limited due to uncertainties around screening criteria, the timing of assessments, and challenges in scaling imaging for widespread HF prevention. These methods are also less sensitive in detecting pre-clinical HF with preserved ejection fraction (HFpEF).
Given these limitations, there is a need for additional approaches to identify early HF risk beyond traditional risk factors, improving risk stratification and preventive efforts. Exercise testing, particularly through cardiopulmonary exercise testing (CPET), offers potential in this area by revealing physiological abnormalities during exercise that may not be evident at rest. Higher levels of cardiorespiratory fitness (CRF) have been linked to lower HF risk, but previous studies have largely focused on estimated CRF rather than peak oxygen uptake (VO2) and have not explored other exercise response metrics that could provide further physiological insights.
This study aimed to assess whether CPET response patterns could identify individuals at elevated HF risk by evaluating their association with established Heart failure risk prediction models and multi-organ phenotypes (e.g., vascular, adipose, pulmonary, and cardiac systems) that may indicate impending HF. The results could improve screening and prevention strategies for individuals without clinical symptoms. Utilizing a large community-based sample with comprehensive CPET data, the study tested the relationship between individual and integrated exercise responses with HF risk estimates and subclinical cardiometabolic phenotypes.
Method
The Framingham Heart Study (FHS) enrolled participants in its Third Generation, Omni-2 (a multi-ethnic group), and New Offspring Spouse cohorts, with initial study visits occurring between 2002 and 2005, followed by second visits in 2008-2011. By the third visit (2016-2019), participants underwent cardiopulmonary exercise testing (CPET). From a total of 3,117 participants, 150 were excluded due to insufficient effort (respiratory exchange ratio below 1.05), 188 lacked complete CPET data, and 247 were excluded for taking atrioventricular nodal blocking agents to avoid cluster selection based on reduced heart rates. The final study population consisted of 2,532 individuals, with research protocols approved by Boston University Medical Campus and Massachusetts General Hospital, and informed consent obtained from all participants.
The CPET procedure, which was previously outlined in other FHS publications, was performed on a cycle ergometer and involved real-time gas exchange measurements. Most participants fasted for at least 8 hours before the test. The protocol included 3 minutes of unloaded cycling followed by ramp exercise with either 15 or 25 W/min increments, depending on predicted peak exercise capacity. Peak oxygen consumption (VO2) was calculated using Wasserman-Hansen equations.
In defining covariates, diabetes was identified by fasting blood glucose levels of ≥126 mg/dL, non-fasting glucose levels of ≥200 mg/dL, or use of glucose-lowering medications. Blood pressure (BP) was measured using a mercury column sphygmomanometer, and hypertension was defined as systolic BP ≥130 mmHg, diastolic BP ≥80 mmHg, or use of antihypertensive medications. Cardiovascular disease (CVD) history was also noted, including prior myocardial infarction, stroke, or heart failure (HF). The 30-year risk of HF was calculated based on age, BP, body mass index (BMI), cholesterol, smoking status, diabetes, and hypertension treatment. These risk estimates were validated across White and Black participants, although for other racial/ethnic groups, equations developed for White participants were used.
Subclinical cardiometabolic assessments included left ventricular mass (measured via echocardiogram), coronary artery calcium (CAC), abdominal aortic calcification (AAC), and adipose tissue measurements from computed tomography (CT) scans. Pulmonary function, including forced expiratory volume (FEV1) and diffusing capacity (DLCO), as well as carotid-femoral pulse wave velocity (CFPWV), were also analyzed, though some assessments were done in earlier study visits (2002-2005 or 2008-2011).
Statistical Analysis
The statistical analysis calculated HF risk for 2,361 participants, comparing peak VO2 levels across gender-based tertiles. Using a range of cardiopulmonary and phenotypic variables, standardized residuals were calculated to account for age, sex, and BMI influences before clustering. Euclidean distance and Ward’s minimum variance criterion were applied, resulting in three optimal clusters identified through R software. Phenotypic traits were compared across clusters, and multivariable logistic regression was used to assess individual characteristics. HF risk differences across clusters were analyzed using ANOVA, and nested linear models were evaluated for predictive performance regarding HF risk. All analyses were conducted using R 4.0.3.
Result
This study included 2,532 participants from the Framingham Heart Study (FHS), with an average age of 53 ± 9 years, 52% being women, and a mean BMI of 28.0 ± 5.3 kg/m². The overall median 30-year estimated heart failure (HF) risk was 13.2% (15.3% in men and 11.2% in women). Peak VO₂ averaged 21.1 ± 5.9 mL/kg/min in women and 26.4 ± 6.7 mL/kg/min in men, with an overall predicted value of 96.6 ± 19.9%. The inverse relationship between peak VO₂ and estimated HF risk was statistically significant (P < 0.001) for both sexes, though there was notable overlap among different VO₂ categories. Spearman’s correlation between peak VO₂ and HF risk was 0.60 for men and 0.55 for women, explaining 30–36% of the variance.
Upon examining cardiopulmonary exercise test (CPET) metrics and organ-specific subclinical phenotypes, moderate correlations were found between oxygen uptake measures (e.g., VO₂ efficiency slope, peak VO₂ pulse) and pulmonary (higher FEV1 and DLCO) and cardiometabolic profiles (lower CRP, reduced fat volumes, and increased left ventricular mass). Adverse blood pressure responses were associated with higher resting carotid-femoral pulse wave velocity (CFPWV).
Three exercise response clusters were identified using hierarchical clustering: Cluster 1 exhibited impaired oxygen kinetics with the lowest oxygen uptake and ventilatory efficiency. Cluster 2 had impaired vascular responses, characterized by elevated blood pressure and reduced oxygen pulse. Cluster 3 represented a favorable exercise response, showing higher peak VO₂, better oxygen uptake kinetics, and moderate blood pressure responses. Although age, sex, and BMI were similar across the clusters, hypertension was more prevalent in the impaired vascular group, while diabetes was more common in the impaired oxygen kinetics group (P < 0.001).
In terms of organ-specific associations, the impaired oxygen kinetics group had lower lung function and higher fat volumes, while the impaired vascular group was linked with higher CFPWV. The favorable exercise response group displayed lower abdominal aortic calcium, CFPWV, and fat measures, alongside better lung function.
Adding CPET cluster information to models predicting HF risk enhanced the explanatory power, increasing the adjusted R² from 0.30 to 0.40 in women and from 0.36 to 0.46 in men (P < 0.0001), indicating that cluster membership provides additional predictive value beyond peak VO₂ alone.
Conclusion
This study assessed the relationship between cardiorespiratory fitness (CRF), as measured by peak VO2, and subclinical disease indicators, as well as the estimated risk of heart failure (HF) in a large community-based cohort. While peak VO2 showed a statistically significant association with increased HF risk, there was considerable overlap in HF risk across different peak VO2 categories. This prompted the researchers to explore whether integrated cardiopulmonary exercise testing (CPET) phenotypes could offer more insight into exercise responses and HF risk prediction.
Using an unbiased clustering analysis, three distinct exercise response patterns were identified: (1) impaired VO2 kinetics, (2) altered hemodynamic-vascular response, and (3) favorable exercise response. Although age, sex, and BMI were consistent across the clusters, each group exhibited unique exercise responses. Clusters 1 and 2 both demonstrated impaired CRF through different mechanisms, while Cluster 3 exhibited favorable exercise responses. These clusters also correlated with specific subclinical cardiometabolic phenotypes, providing additional physiological insights beyond standard HF risk prediction models.
This study highlights the importance of using comprehensive CPET data, including hemodynamic and gas exchange measurements, to better assess HF risk. The hierarchical clustering revealed distinct exercise response patterns, suggesting that CRF impairments occur through different physiological mechanisms. Importantly, this additional information provided by integrated CPET phenotypes enhances the accuracy of HF risk prediction beyond what is achieved through traditional clinical risk factors and peak VO2 alone.
The clinical implications of these findings are significant, especially in light of emerging therapies for HF prevention. Integrating CPET phenotypes into HF risk assessment could offer more individualized predictions and help identify individuals who may benefit from targeted interventions. However, widespread implementation of CPET may be challenging due to the specialized equipment and expertise required. Future research should explore whether simpler exercise measures or biomarkers can replicate these findings for broader population screening.
In conclusion, this study demonstrates that hierarchical clustering of multidimensional CPET responses reveals distinct exercise phenotypes that correlate with subclinical cardiometabolic profiles. These phenotypes provide additional insights beyond traditional peak VO2 assessments and may prove valuable in refining HF risk prediction and guiding preventive strategies.
Oncology Related Tools
Other
Latest Research
