Phenotypic age acceleration and all-cause and cardiovascular mortality among U.S. adults at risk for heart failure

Abstract

Objective

To examine the associations of phenotypic age acceleration (PAA) with all-cause and cardiovascular mortality among U.S. adults at risk for heart failure.

Methods

We analyzed 19,665 participants aged ≥20 years at risk for heart failure from the National Health and Nutrition Examination Survey (NHANES) 1999–2010 and 2015–2018. PAA was defined as the residual from regressing phenotypic age on chronological age. Kaplan-Meier analysis, weighted Cox regression, Fine-Gray competing-risk models, and restricted cubic spline analyses were used to evaluate these associations.

Results

Compared with the lowest quartile, participants in the highest quartile of PAA had higher risks of all-cause mortality (HR, 2.63; 95% CI, 2.31–2.98) and cardiovascular mortality (HR, 2.55; 95% CI, 2.00–3.25). Restricted cubic spline analysis showed a nonlinear association between PAA and all-cause mortality (P for nonlinearity = 0.005), with a threshold at PAA = −8.26, whereas the association with cardiovascular mortality was linear (P for nonlinearity = 0.881). In competing-risk analysis, the highest PAA quartile remained significantly associated with increased cardiovascular mortality (SHR, 1.39; 95% CI, 1.15–1.68).

Conclusion

Higher PAA was significantly associated with increased risks of all-cause and cardiovascular mortality among adults at risk for heart failure, suggesting that PAA may be a potential marker of mortality risk.

1. Introduction

Heart failure (HF) is a common and fatal chronic condition, with its prevalence steadily increasing due to the aging global population [1]. According to the 2022 AHA/ACC/HFSA Heart Failure Management Guidelines, individuals at risk for heart failure are those who do not yet show symptoms of HF or structural heart abnormalities but are at high risk due to the presence of cardiovascular risk factors such as hypertension, diabetes, obesity, and coronary artery disease [2]. While these individuals are asymptomatic, the accumulation of cardiovascular risk factors over time significantly increases their likelihood of developing heart failure [3]. As the number of individuals with high cardiovascular risk continues to rise, early identification and intervention in these groups are critical for preventing heart failure [4].

Phenotypic age acceleration (PAA) is a tool that estimates biological age using clinical biochemical markers. It combines markers related to inflammation, metabolism, immune function, and organ health to assess the biological aging process [5]. These markers include C-reactive protein, glucose, immune cell counts, albumin, creatinine, and alkaline phosphatase, which are commonly used to evaluate the health status of individuals with chronic diseases such as diabetes. Compared to traditional methods that rely on tissue-based data to measure biological age, PAA is more practical and cost-effective [6]. Research shows that PAA is a strong predictor of cardiovascular disease risk and all-cause mortality, and it helps reveal the differences in aging within populations. Its prediction model has shown high reliability and accuracy in risk stratification [7].

This study uses long-term follow-up data to investigate the relationship between PAA and both all-cause and cardiovascular mortality. The goal is to provide new insights into the aging process for the precise management of individuals at risk for heart failure.

2. Methods

2.1. Study population and design

Data for this study were sourced from the National Health and Nutrition Examination Survey (NHANES), a cross-sectional study conducted by the National Center for Health Statistics (NCHS). Each NHANES cycle includes a new, nationally representative sample of the U.S. population. All participants provided written informed consent, which was approved by the Institutional Review Board of the Centers for Disease Control and Prevention (CDC). Detailed descriptions of the survey design, sampling techniques, and data collection protocols are publicly available on the CDC's official website (https://www.cdc.gov/nchs/nhanes/) [8].

This analysis included data from eight NHANES cycles (1999–2010 and 2015–2018). Among 49,363 individuals aged ≥20 years, participants were excluded if they had no cardiovascular risk factors or self-reported atherosclerotic cardiovascular disease (ASCVD) (n = 25,943), had missing phenotypic age data (n = 134), had missing follow-up data (n = 40), or had missing data on angina (n = 30), body mass index (BMI) (n = 154), coronary heart disease (n = 33), diabetes mellitus (n = 97), alcohol drinking status (n = 1173), education level (n = 19), marital status (n = 252), family income (n = 1794), smoking status (n = 11), or stroke (n = 18). Finally, 19,665 participants were included in the study (Fig. 1).

Fig. 1.

Flow diagram of the screening and enrolment of study participants.

Abbreviations: NHANES, National Health and Nutrition Examination Survey.

2.2. Definition of adults at risk for heart failure

Participants were classified as adults at risk for heart failure based on the Stage A concept in the 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure if they had at least one major HF risk factor identifiable in NHANES, including hypertension, diabetes mellitus, obesity, or established ASCVD [2], Hypertension was defined as systolic blood pressure ≥ 130 mmHg, diastolic blood pressure ≥ 80 mmHg, or self-reported use of antihypertensive medication [9]. Obesity was defined as a body mass index (BMI) ≥30 kg/m² [10]. Diabetes mellitus was defined as fasting glucose ≥126 mg/dL, glycated hemoglobin (HbA1c) ≥6.5% [11], or self-reported use of glucose-lowering medication. ASCVD was defined as self-reported physician diagnosis of coronary heart disease, angina, or stroke.

2.3. Assessment of PAA

Biological aging was assessed using phenotypic age. Phenotypic age was calculated based on chronological age and 9 biomarkers, including albumin, creatinine, glucose, the natural logarithm of C-reactive protein, lymphocyte percentage, mean corpuscular volume, red cell distribution width, alkaline phosphatase, and white blood cell count [12]. Chronological age was obtained from the NHANES demographic data, and the remaining biomarkers were obtained from laboratory data. PAA was defined as the residual from regressing phenotypic age on chronological age and was expressed in years, with positive values indicating accelerated biological aging relative to chronological age. Phenotypic age was calculated according to the published formula [13]:

$$\text{phenotypic}\mathrm{age}=141.50+\frac{\ln \left[-0.00553\times \ln \left(1-\mathrm{M}\right)\right]}{0.09165}$$

where

$$M=1-\exp \left(\frac{-1.51714\times \exp \left(\mathit{xb}\right)}{0.0076927}\right)$$

and, xb = −19.907–0.0336 × albumin (g/L) + 0.0095 × creatinine (umol/L) + 0.1953 × glucose (mmol/L) + 0.0954 × ln (CRP (mg/dl)) − 0.0120 × lymphocyte percent (%) + 0.0268 × mean cell volume (fL) + 0.3306 × red cell distribution width (%) + 0.00188 × alkaline phosphatase (U/L) + 0.0554 × white blood cell count (1000 cell/μL) + 0.0804 × chronological age (years).

2.4. Mortality assessment

Death information was obtained from the NHANES public-use linked mortality files, with follow-up through December 31, 2019. Causes of death were classified according to the International Classification of Diseases, 10th Revision (ICD-10). The primary outcomes were all-cause mortality and cardiovascular mortality. Cardiovascular mortality was defined using ICD-10 codes I00–I09, I11, I13, I20–I25, I26–I28, I30–I51, and I60–I69 [14].

2.5. Covariates

Covariates were selected based on prior literature and clinical relevance. The adjusted variables included demographic, socioeconomic, lifestyle, and health-related factors. Demographic variables included age, sex, and race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, and others). Socioeconomic variables included education level (<9 years, 9–12 years, and >12 years), marital status (married/cohabiting vs. unmarried, including widowed, divorced, separated, or never married), and poverty-to-income ratio (PIR; low, ≤1.3; middle, 1.3–3.5; high, >3.5). Lifestyle factors included physical activity, smoking status, and alcohol consumption. Physical activity was classified as inactive (<150 min/week) or active (≥150 min/week) according to the World Health Organization (WHO) recommendation of at least 150 min/week of moderate-intensity physical activity. Smoking status was categorized as never, former, or current. Alcohol consumption was defined as drinking ≥12 alcoholic drinks in the past year. Health-related factors included body mass index (BMI), hypertension, hyperlipidemia, diabetes mellitus, and established ASCVD. Hyperlipidemia was defined as the use of lipid-lowering medications, triglycerides ≥150 mg/dL, total cholesterol ≥200 mg/dL, low-density lipoprotein cholesterol (LDL-C) ≥130 mg/dL, or high-density lipoprotein cholesterol (HDL—C) <40 mg/dL [15].

2.6. Statistical analysis

Due to the complex NHANES sampling design, appropriate sample weights were incorporated into all analyses to obtain nationally representative estimates, as recommended by NHANES. Specifically, MEC examination weights from each survey cycle (WTMEC4YR, WTMEC2YR, and WTMEPRP) were combined according to NHANES analytic guidelines. For baseline characterization, weighted means (standard errors [SEs]) were used for continuous variables, and weighted numbers (weighted percentages) were used for categorical variables. The study population was categorized into four groups according to quartiles of PAA (PAA < −7.05, −7.05 ≤ PAA < −2.64, −2.64 ≤ PAA < 3.02, and PAA ≥ 3.02). Differences in baseline characteristics across PAA quartiles were compared using ANOVA for continuous variables and the Rao-Scott χ² test for categorical variables. Survival analyses were performed using the Kaplan-Meier method to estimate all-cause mortality and cardiovascular mortality across PAA groups, with group differences assessed using the log-rank test. The associations between PAA and mortality outcomes were evaluated using weighted multivariable Cox proportional hazards regression models, with results reported as hazard ratios (HRs) and 95% confidence intervals (CIs). In addition, Fine–Gray competing risks models were used to further assess the association between PAA and cardiovascular mortality, with non-cardiovascular death treated as a competing event. The proportional hazards assumption was evaluated using Schoenfeld residuals, and no violations were detected. Covariates were selected based on prior literature and clinical relevance. Four sequential models were constructed: Model 1 was unadjusted; Model 2 was adjusted for age, sex, race/ethnicity, marital status, poverty-to-income ratio, and education level; Model 3 was further adjusted for smoking status, physical activity, alcohol consumption, and BMI; and Model 4 was additionally adjusted for established ASCVD, hypertension, hyperlipidemia, and diabetes mellitus. To further assess the potential impact of unequal follow-up duration across PAA groups, time-dependent Cox regression analyses were performed using 24-month follow-up intervals. HRs and 95% CIs were estimated within each interval using the same covariates as in Model 4. Potential nonlinear associations were examined using restricted cubic spline (RCS) models, with nonlinearity considered statistically significant at P < 0.05. The spline models were constructed using four knots placed at the 5th, 35th, 65th, and 95th percentiles of the PAA distribution. In addition, smoothed piecewise Cox proportional hazards regression models were used to explore potential threshold effects after full adjustment for covariates. Subgroup analyses were performed stratified by age, sex, BMI, physical activity, hypertension, diabetes mellitus, hyperlipidemia, and angina. Within each subgroup, HRs and 95% CIs were estimated using fully adjusted survey-weighted Cox proportional hazards models, with the lowest quartile of PAA as the reference group. P values for interaction were also calculated. Sensitivity analyses were performed to evaluate the robustness of the findings, including: (1) excluding participants who died within the first 2 years of follow-up to reduce potential reverse causality; (2) excluding individuals with obesity alone to further refine the population at risk for heart failure; (3) excluding participants with self-reported cardiovascular diseases; (4) handling missing covariate values using multiple imputation by chained equations, with 5 imputed datasets generated based on variables included in the final statistical model and pooled estimates obtained across imputations [16]; (5) conducting stratified analyses by NHANES survey wave (1999–2002, 2003–2006, 2007–2010, and 2015–2018) to assess potential temporal trends; (6) additionally assessing biological aging using KDM (Klemera-Doubal method) biological age, calculated from 8 clinical biomarkers using the BioAge R package. KDM-based age acceleration was defined as the residual from regressing KDM biological age on chronological age [12]. All statistical analyses were performed using R software (version 4.3.2) and FreeStatistics software (version 2.0), and a two-sided P value <0.05 was considered statistically significant.

3. Results

3.1. Participant characteristics

The weighted study population comprised 10,334.9 thousand participants and was categorized into quartiles of PAA. The mean age of the study population was 50.34 ± 16.12 years, and age differed significantly across quartiles (P < 0.001). Participants in Q4 were older than those in the other quartiles (51.23 ± 16.50 years). The proportion of males was higher in Q4 than in Q1 (51.0% vs. 44.1%, P < 0.001), whereas the proportion of non-Hispanic Whites was lower (62.8% vs. 71.2%, P < 0.001). In addition, participants in Q4 had lower proportions of married/cohabiting individuals and lower income and education levels (all P < 0.001), as well as higher proportions of smoking and alcohol consumption (all P < 0.001). In terms of health-related characteristics, Q4 had higher prevalences of obesity (41.2% vs. 28.6%, P < 0.001), hypertension (58.3% vs. 53.5%, P < 0.001), and diabetes mellitus (37.0% vs. 6.4%, P < 0.001). Baseline characteristics of the study population are shown in Table 1. The baseline characteristics of excluded and included participants are shown in Table S1.

Table 1.

Weighted baseline characteristics of U.S. adults aged ≥20 years at risk for heart failure according to quartiles of PAA, NHANES 1999–2010 and 2015–2018.

Variables	Total	Q1 (PAA < −7.05)	Q2 (−7.05 ≤ PAA < −2.64)	Q3 (−2.64 ≤ PAA < 3.02)	Q4 (PAA ≥ 3.02)	P
Weighted population, n (thousands)	10,334.9	2682.93	2689.08	2644.68	2318.21
Age, mean (SD), years	50.34 (16.12)	51.26 (15.58)	49.09 (15.81)	49.91 (16.55)	51.23 (16.50)	<0.001
Sex, n (thousands), %
Male	5206.42 (50.4)	1184.44 (44.1)	1442.57 (53.6)	1396.82 (52.8)	1182.59 (51.0)	<0.001
Female	5128.49 (49.6)	1498.50 (55.9)	1246.52 (46.4)	1247.86 (47.2)	1135.61 (49.0)
Race/ethnicity, n (thousands), %
Non-Hispanic White	7355.65 (71.2)	2046.30 (76.3)	1999.66 (74.4)	1852.92 (70.1)	1456.77 (62.8)	<0.001
Non-Hispanic Black	1126.47 (10.9)	216.15 (8.1)	250.54 (9.3)	305.37 (11.5)	354.41 (15.3)
Mexican American	781.04 (7.6)	168.01 (6.3)	192.59 (7.2)	202.87 (7.7)	217.57 (9.4)
Other Hispanic	490.06 (4.7)	113.18 (4.2)	112.64 (4.2)	135.84 (5.1)	128.39 (5.5)
Other racea	581.68 (5.6)	139.29 (5.2)	133.65 (5.0)	147.67 (5.6)	161.07 (6.9)
Marital status, n (thousands), %
Married or living with a partner	6856.32 (66.3)	1868.27 (69.6)	1867.30 (69.4)	1705.22 (64.5)	1415.52 (61.1)	<0.001
Unmarriedb	3478.59 (33.7)	814.66 (30.4)	821.78 (30.6)	939.46 (35.5)	902.69 (38.9)
PIRc, n (thousands), %
≤1.30	2002.51 (19.4)	402.19 (15.0)	491.59 (18.3)	516.85 (19.5)	591.88 (25.5)	<0.001
1.31–3.50	3798.67 (36.8)	950.31 (35.4)	977.75 (36.4)	968.99 (36.6)	901.61 (38.9)
>3.50	4533.72 (43.9)	1330.43 (49.6)	1219.74 (45.4)	1158.83 (43.8)	824.72 (35.6)
Education level, n (thousands), %
Less than 9th grade	584.44 (5.7)	155.98 (5.8)	153.26 (5.7)	137.27 (5.2)	137.93 (5.9)	<0.001
9–11th grade (includes 12th grade with no diploma)	1147.36 (11.1)	265.13 (9.9)	319.55 (11.9)	289.18 (10.9)	273.50 (11.8)
High school grad/GED or equivalent	2740.32 (26.5)	666.32 (24.8)	714.12 (26.6)	698.93 (26.4)	660.96 (28.5)
Some college or AA degree	3251.43 (31.5)	827.95 (30.9)	795.74 (29.6)	866.55 (32.8)	761.20 (32.8)
College graduate or above	2611.36 (25.3)	767.55 (28.6)	706.43 (26.3)	652.75 (24.7)	484.62 (20.9)
Smoking status, n (thousands), %
Never	5368.84 (51.9)	1510.46 (56.3)	1386.19 (51.5)	1349.66 (51.0)	1122.53 (48.4)	<0.001
Former	2858.73 (27.7)	812.81 (30.3)	711.49 (26.5)	696.17 (26.3)	638.25 (27.5)
Current	2107.33 (20.4)	359.66 (13.4)	591.40 (22.0)	598.84 (22.6)	557.43 (24.0)
Drinking status, n (thousands), %
Never	2459.61 (23.8)	762.07 (28.4)	636.29 (23.7)	567.62 (21.5)	493.63 (21.3)	<0.001
Former	1174.32 (11.4)	231.10 (8.6)	259.94 (9.7)	330.35 (12.5)	352.93 (15.2)
Current	6700.97 (64.8)	1689.77 (63.0)	1792.85 (66.7)	1746.71 (66.0)	1471.65 (63.5)
BMI, n (thousands), %
<25	1863.37 (18.0)	768.10 (28.6)	441.37 (16.4)	379.67 (14.4)	274.23 (11.8)	<0.001
25–30	3004.65 (29.1)	953.49 (35.5)	829.42 (30.8)	735.28 (27.8)	486.47 (21.0)
>30	5466.88 (52.9)	961.34 (35.8)	1418.30 (52.7)	1529.73 (57.8)	1557.51 (67.2)
Physical activityd, n (thousands), %
Inactive	5343.02 (51.7)	1510.60 (56.3)	1476.84 (54.9)	1265.69 (47.9)	1089.89 (47.0)	<0.001
Active	4991.89 (48.3)	1172.34 (43.7)	1212.24 (45.1)	1378.99 (52.1)	1228.32 (53.0)
Coronary heart disease, n (thousands), %
No	10,105.82 (97.8)	2631.14 (98.1)	2637.53 (98.1)	2583.57 (97.7)	2253.57 (97.2)	0.085
Yes	229.09 (2.2)	51.80 (1.9)	51.55 (1.9)	61.11 (2.3)	64.64 (2.8)
Stroke, n (thousands), %
No	10,015.92 (96.9)	2620.74 (97.7)	2622.99 (97.5)	2559.30 (96.8)	2212.88 (95.5)	<0.001
Yes	318.99 (3.1)	62.19 (2.3)	66.08 (2.5)	85.37 (3.2)	105.35 (4.5)
Angina, n (thousands), %
No	10,149.17 (98.2)	2642.34 (98.5)	2641.72 (98.2)	2599.99 (98.3)	2265.11 (97.7)	0.167
Yes	185.74 (1.8)	40.59 (1.5)	47.36 (1.8)	44.68 (1.7)	53.10 (2.3)
Hyperlipidemia, n (thousands), %
No	2102.26 (20.3)	536.02 (20.0)	549.70 (20.4)	550.57 (20.8)	465.98 (20.1)	0.875
Yes	8232.64 (79.7)	2146.92 (80.0)	2139.39 (79.6)	2094.11 (79.2)	1852.23 (79.9)
Hypertension, n (thousands), %
No	4835.34 (46.8)	1248.69 (46.5)	1340.98 (49.9)	1278.57 (48.3)	967.10 (41.7)	<0.001
Yes	5499.56 (53.2)	1434.25 (53.5)	1348.10 (50.1)	1366.11 (51.7)	1351.11 (58.3)
Diabetes mellitus, n (thousands), %
No	8608.01 (83.3)	2511.37 (93.6)	2420.32 (90.0)	2215.35 (83.8)	1460.97 (63.0)	<0.001
Yes	1726.90 (16.7)	171.56 (6.4)	268.76 (10.0)	429.33 (16.2)	857.24 (37.0)

Abbreviations: SD, standard deviation; PIR, poverty-income ratio; NHANES, National Health and Nutrition Examination Survey; PAA, phenotypic age acceleration; BMI, body mass index.

Notes: All estimates were derived using survey weights and are nationally representative of the U.S. adult population. Continuous variables are presented as weighted means (SDs), and categorical variables as weighted population estimates (thousands) with weighted percentages.

^aOther race includes racial/ethnic groups other than non-Hispanic White, non-Hispanic Black, Mexican American, and other Hispanic; NHANES did not provide further details for this category.

^bUnmarried includes widowed, divorced, separated, and never married participants.

^cPIR indicates poverty-income ratio.

^dPhysical activity was classified according to the WHO recommendation of at least 150 min/week of moderate-intensity physical activity. Participants were categorized as inactive (<150 min/week) or active (≥150 min/week).

3.2. Survival analysis

Weighted Kaplan-Meier survival analyses showed significant differences in all-cause mortality and cardiovascular mortality across PAA quartiles. Participants in the highest quartile (Q4) of PAA had the lowest survival probability, whereas those in the lowest quartile (Q1) had the highest survival probability. In the overall population, mortality increased progressively across increasing PAA quartiles (all log-rank P < 0.001). Detailed Kaplan-Meier curves are shown in Fig. 2.

Fig. 2.

Kaplan–Meier survival curves for all-cause mortality (A) and cardiovascular mortality (B) across quartiles of PAA among U.S. adults aged ≥20 years at risk for heart failure.

Notes: Survival probabilities were compared using the log-rank test. Participants were categorized into four groups based on PAA quartiles: PAA ≤ −7.05, −7.05 < PAA ≤ −2.64, −2.64 < PAA ≤ 3.02, and PAA > 3.02. The number at risk at each time point is shown below the plots.

Abbreviations: PAA, phenotypic age acceleration.

3.3. Association between PAA and mortality

Among the 19,665 participants at risk for heart failure, 3545 all-cause deaths (18.03%) and 1075 cardiovascular deaths (5.47%) occurred during a mean follow-up of 10.36 years. As shown in Table 2, higher PAA quartiles were associated with progressively increased risks of both all-cause and cardiovascular mortality in the Cox regression analyses. In the fully adjusted model, compared with Q1, the HRs (95% CIs) for all-cause mortality were 1.20 (1.08–1.32) for Q2, 1.50 (1.35–1.68) for Q3, and 2.63 (2.31–2.98) for Q4. The corresponding HRs (95% CIs) for cardiovascular mortality were 1.31 (1.07–1.61), 1.48 (1.22–1.80), and 2.55 (2.00–3.25), respectively. To account for the competing risk of non-cardiovascular death, Fine-Gray competing risk regression was additionally performed for cardiovascular mortality. In the fully adjusted model, PAA remained positively associated with cardiovascular mortality as a continuous variable (SHR = 1.01, 95% CI: 1.01–1.02, P < 0.001). In the quartile analysis, compared with Q1, only Q4 remained significantly associated with cardiovascular mortality after accounting for competing risks (SHR = 1.39, 95% CI: 1.15–1.68, P = 0.001). To further assess whether unequal follow-up duration across PAA groups influenced the observed associations, time-dependent Cox regression analyses were performed using Model 4 (Table S2). The association between higher PAA and all-cause mortality remained robust, whereas a broadly similar temporal pattern was observed for cardiovascular mortality, although the magnitude and statistical significance of the associations varied across follow-up intervals.

Table 2.

Weighted Cox and Fine-Gray regression analyses of the associations of PAA with all-cause and cardiovascular mortality among U.S. adults at risk for heart failure.

Characteristics	Participants, n	Events, n (%)	Person-years	Model1	P value	Model2	P value	Model3	P value	Model4	P value
				HR/SHR (95% CI)		HR/SHR (95% CI)		HR/SHR (95% CI)		HR/SHR (95% CI)
All-cause mortality
PAA (per 1-unit increase)	19,665.	3545 (18)	198,653	1.03 (1.03–1.04)	<0.001	1.03 (1.03–1.04)	<0.001	1.04 (1.03–1.04)	<0.001	1.04 (1.03–1.04)	<0.001
Q1	4916	945 (19.2)	66,168	1.00		1.00		1.00		1.00
Q2	4916	880 (17.9)	59,551	1.09 (0.97–1.21)	0.136	1.25 (1.13–1.38)	<0.001	1.21 (1.09–1.34)	0.001	1.20 (1.08–1.32)	<0.001
Q3	4916	810 (16.5)	43,532	1.52 (1.36–1.71)	<0.001	1.65 (1.48–1.84)	<0.001	1.58 (1.41–1.76)	<0.001	1.50 (1.35–1.68)	<0.001
Q4	4917	910 (18.5)	29,401	3.16 (2.81–3.56)	<0.001	3.00 (2.68–3.37)	<0.001	2.87 (2.55–3.23)	<0.001	2.63 (2.31–2.98)	<0.001
Cardiovascular mortality
PAA (per 1-unit increase)	19,665	1075 (5.5)	198,653	1.03 (1.03–1.04)	<0.001	1.03 (1.03–1.04)	<0.001	1.04 (1.03–1.05)	<0.001	1.03 (1.03–1.04)	<0.001
Q1	4916	282 (5.7)	66,168	1.00		1.00		1.00		1.00
Q2	4916	282 (5.7)	59,551	1.22 (1.00–1.50)	0.052	1.41 (1.14–1.74)	0.001	1.34 (1.09–1.66)	0.006	1.31 (1.07–1.61)	0.01
Q3	4916	247 (5)	43,532	1.63 (1.33–2.01)	<0.001	1.74 (1.41–2.14)	<0.001	1.61 (1.32–1.97)	<0.001	1.48 (1.22–1.80)	<0.001
Q4	4917	264 (5.4)	29,401	3.44 (2.81–4.22)	<0.001	3.20 (2.58–3.97)	<0.001	2.96 (2.36–3.71)	<0.001	2.55 (2.00–3.25)	<0.001
Competing-risk analysis for cardiovascular mortality
PAA (per 1-unit increase)	19,665	1075 (5.5)	198,653	1.03 (1.02–1.03)	<0.001	1.02 (1.01–1.02)	<0.001	1.02 (1.01–1.02)	<0.001	1.01 (1.01–1.02)	<0.001
Q1	4916	282 (5.7)	66,168	1.00		1.00		1.00		1.00
Q2	4916	282 (5.7)	59,551	1.13 (0.96–1.34)	0.134	1.21 (1.03–1.43)	0.021	1.17 (0.99–1.38)	0.064	1.16 (0.99–1.37)	0.075
Q3	4916	247 (5)	43,532	1.39 (1.18–1.65)	<0.001	1.34 (1.13–1.60)	0.001	1.27 (1.06–1.51)	0.01	1.19 (1.00–1.43)	0.056
Q4	4917	264 (5.4)	29,401	2.21 (1.86–2.61)	<0.001	1.71 (1.43–2.04)	<0.001	1.58 (1.32–1.91)	<0.001	1.39 (1.15–1.68)	0.001

Abbreviations: PAA, phenotypic age acceleration; HR, hazard ratio; SHR, subdistribution hazard ratio; CI, confidence interval; NHANES, National Health and Nutrition Examination Survey.

Notes: Weighted Cox proportional hazards regression models were used to estimate HRs and 95% CIs for all-cause mortality and cardiovascular mortality. For the competing-risk analysis, effect estimates were obtained using Fine-Gray competing risk regression models, with non-cardiovascular death treated as the competing event for cardiovascular mortality. The effect size is represented by SHR with 95% CIs Two-sided P values <0.05 were considered statistically significant.

3.4. RCS analysis

RCS analysis showed a significant nonlinear association between PAA and all-cause mortality (Fig. 3A, P for non-linearity = 0.005), whereas the association between PAA and cardiovascular mortality was linear (Fig. 3B, P for non-linearity = 0.881). Threshold effect analysis (Table 3) identified an inflection point at PAA = −8.26 for all-cause mortality. Below this threshold, the adjusted hazard ratio (HR) was 1.02 (95% CI: 0.99–1.06), whereas above this threshold, the adjusted HR increased to 1.05 (95% CI: 1.04–1.06). The likelihood ratio test indicated that the two-piecewise model provided a significantly better fit than the single-line model (P = 0.048).

Fig. 3.

Restricted cubic spline curves for the associations of PAA with all-cause mortality (A) and cardiovascular mortality (B) among U.S. adults aged ≥20 years at risk for heart failure.

Abbreviations: PAA, phenotypic age acceleration; HR, hazard ratio; CI, confidence interval.

Table 3.

Threshold effect analysis of the association between PAA and all-cause mortality.

Variable	Adjusteda HR (95% CI)
PAA (per 1-unit increase)
<−8.26	1.03 (0.99, 1.06)
≥−8.26	1.05 (1.04, 1.06)
Likelihood ratio test p value	0.048

Abbreviations: PAA, phenotypic age acceleration; BMI, body mass index.

^aAdjusted for age, sex, race, marital status, poverty income ratio, education, smoking, physical activity, alcohol use, BMI, ASCVD, hypertension, hyperlipidemia and diabetes mellitus.

3.5. Subgroup analyses

Subgroup analyses showed that higher PAA was generally associated with increased risks of both all-cause and cardiovascular mortality across most prespecified subgroups. Significant interactions were observed for age and stroke in the all-cause mortality analysis (P for interaction = 0.001 and 0.010, respectively), and for age and diabetes mellitus in the cardiovascular mortality analysis (P for interaction = 0.015 and 0.040, respectively). No significant interactions were identified for the other stratification variables. Because some subgroups, particularly those with stroke or angina, had relatively small sample sizes and limited numbers of events, these findings should be interpreted with caution (Fig. 4).

Fig. 4.

Subgroup analyses of the associations between quartiles of PAA and all-cause mortality (A) and cardiovascular mortality (B) among U.S. adults aged ≥20 years at risk for heart failure.

Notes: HRs and 95% CIs were estimated using fully adjusted Cox proportional hazards regression models, with Q1 as the reference group. P for interaction was calculated by including the cross-product term between PAA quartiles and each stratification variable.

3.6. Sensitivity analysis

Sensitivity analyses confirmed the robustness of the main findings. Excluding participants with follow-up <2 years (Table S3) did not materially change the results, with Q4 remaining significantly associated with a higher risk of all-cause mortality compared with Q1 (HR = 2.42, 95% CI: 2.16–2.70, P < 0.001). Excluding individuals with obesity alone (Table S4) also did not substantially alter the findings, and Q4 remained significantly associated with a higher risk of all-cause mortality (HR = 2.37, 95% CI: 2.12–2.65). After excluding participants with self-reported cardiovascular disease (Table S5), the association was slightly attenuated but remained significant in Q4 (HR = 2.08, 95% CI: 1.85–2.34). Results from multiple imputation (Table S6) were consistent with the main analysis. Stratified analyses by NHANES survey wave (Tables S7–S10) showed that the positive association between higher PAA and all-cause mortality remained generally consistent across survey cycles. For cardiovascular mortality, the direction of association was similar, although the magnitude and statistical significance varied across cycles. In the competing-risk analysis for cardiovascular mortality, the overall direction remained positive, but statistical significance was observed only in some survey cycles. Additionally, when biological aging was assessed using the KDM-based metric (Table S11), the association with all-cause mortality remained significant in Q4 (HR = 1.44, 95% CI: 1.30–1.59).

4. Discussion

Using NHANES 1999–2010 and 2015–2018 data, we found that higher PAA was independently associated with increased risks of all-cause and cardiovascular mortality among U.S. adults at risk for heart failure. Mortality risk increased progressively across PAA quartiles, and participants in the highest quartile had the poorest survival. Restricted cubic spline analysis further suggested a nonlinear association with all-cause mortality but a more linear association with cardiovascular mortality. Together, these findings indicate that PAA, as an integrative marker of biological aging, may capture important heterogeneity in long-term prognosis within this at-risk population. Competing-risk analyses yielded generally consistent results, although the effect estimates were attenuated, suggesting that conventional Cox models may somewhat overestimate the association by not accounting for competing non-cardiovascular death.

The findings of this study align with previous research on biological aging and mortality risk, while providing new evidence specifically for adults at risk for heart failure. Prior studies have shown that accelerated biological aging is strongly associated with all-cause mortality and adverse cardiovascular outcomes in the general population, as well as in individuals with diabetes, coronary artery disease, and cancer survivors [7], [17], [18]. However, unlike prior studies that primarily focused on the general population or single chronic diseases, this study specifically examined adults at risk for heart failure, those without overt heart failure but with key cardiovascular risk factors, including hypertension, diabetes, obesity, or atherosclerotic cardiovascular disease [2]. This population is large and carries a significant disease burden, yet traditional risk factors alone do not fully account for the marked heterogeneity in long-term outcomes among individuals with similar risk profiles [3]. Therefore, our findings build on previous evidence by demonstrating a stable association between PAA and mortality risk in this high-risk population, suggesting that biological aging could serve as an important complementary dimension in understanding the prognostic heterogeneity in adults at risk for heart failure.

The biological mechanisms linking accelerated phenotypic aging to increased mortality risk are likely multifactorial. PAA integrates biomarkers related to inflammation, metabolic dysfunction, renal function, nutritional status, and immune regulation, and therefore may reflect declines in physiological reserve and multisystem homeostasis rather than dysfunction of a single organ system [19], [20], [21]. Chronic low-grade inflammation, immune senescence, and metabolic imbalance may jointly promote endothelial dysfunction, impaired stress responses, reduced tissue repair capacity, and greater susceptibility to disease burden [22], [23]. In this context, elevated PAA may represent a broader aging-related vulnerability that predisposes individuals to adverse outcomes and ultimately manifests as increased risks of all-cause and cardiovascular mortality at the population level [24], [25].

From a population health perspective, these findings highlight the potential value of biological aging markers as a complement to traditional cardiovascular risk assessment [26]. Established risk factors help identify individuals at higher cardiovascular risk, while aging-related markers such as PAA may further reveal heterogeneity in long-term outcomes among those already classified as high risk [18], [27]. Therefore, PAA may have value in population-level risk stratification and hypothesis generation. However, its clinical utility for individual risk prediction and decision-making still requires validation in prospective and interventional studies.

This study has several strengths worth emphasizing. First, it is based on a nationally representative sample with long-term mortality follow-up, which enhances the generalizability of the findings to the population at risk for heart failure. Second, the study made comprehensive adjustments for demographic, lifestyle, and clinical covariates, and employed various statistical methods to reduce the impact of major confounding factors. Additionally, phenotypic age acceleration, as an integrative biological aging marker, reflects changes in multisystem physiological status rather than being limited to a single cardiovascular risk factor.

This study has several limitations. First, due to its observational design, causal inferences cannot be established, and reverse causality cannot be completely excluded. Second, the biomarkers used to derive phenotypic age acceleration were measured at a single time point and may not capture long-term trajectories of biological aging, which could have resulted in exposure misclassification. Third, the study population was operationally defined as adults at risk for heart failure based on major heart failure risk factors identifiable in NHANES, rather than structural or functional cardiac abnormalities; therefore, the findings should be interpreted within this specific population framework. Fourth, established atherosclerotic cardiovascular disease was identified based on self-reported diagnoses and may be subject to recall bias. Fifth, missing data and differences between included and excluded participants may have introduced selection bias, and the observed group differences suggest that missingness may not have been completely random; although multiple imputation yielded broadly consistent results, residual bias cannot be fully excluded. Sixth, although additional Fine-Gray competing-risk analyses and time-dependent Cox regression analyses were performed, the estimates for cardiovascular mortality should still be interpreted cautiously, as they may be influenced by competing events and unequal follow-up duration across groups. Finally, because NHANES is based on the U.S. population, the generalizability of these findings to other geographic or ethnic populations may be limited.

5. Conclusion

In conclusion, PAA was significantly associated with all-cause and cardiovascular mortality among U.S. adults at risk for heart failure and may serve as a potential independent biomarker for prognostic assessment in this population. Further studies are needed to investigate its longitudinal changes, underlying mechanisms, and potential intervention value.

CRediT authorship contribution statement

Huifang Su: Writing – original draft, Software, Resources, Methodology, Data curation, Conceptualization. Xiandong Liu: Writing – review & editing, Supervision, Investigation, Formal analysis, Data curation, Conceptualization.

Clinical trial number

Not applicable.

Consent for publication

Not applicable.

Ethical statement

This study was conducted using data from a publicly available database. The data were fully de-identified prior to release, and no identifiable private information was accessed. According to national regulations and institutional policies, analyses of publicly available, anonymized data do not constitute human subjects research and therefore do not require additional ethical approval or informed consent.

Ethics approval and consent to participate

Study protocols for NHANES were approved by the National Center for Health Statistics Ethics Review Board (Protocol #98-12, Protocol #2005-06, Protocol #2011-17, Protocol #2018-01, https://www.cdc.gov/nchs/nhanes/irba98.htm). All the participants signed the informed consent before participating in the study. All methods were carried out in accordance with relevant guidelines and regulations.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

These survey data are free and publicly available, and can be downloaded directly from the NHANES website (http://www.cdc.gov/nchs/nhanes/) by users and researchers worldwide.