Accelerated Aging Mediates the Associations of Unhealthy Lifestyles with Cardiovascular Disease, Cancer, and Mortality

Abstract

BACKGROUND:

With two well-validated aging measures capturing mortality and morbidity risk, this study examined whether and to what extent aging mediates the associations of unhealthy lifestyles with adverse health outcomes.

METHODS:

Data were from 405,944 adults (40–69 years) from UK Biobank (UKB) and 9,972 adults (20–84 years) from US National Health and Nutrition Examination Survey (NHANES). An unhealthy lifestyles score (range: 0–5) was constructed based on five factors (smoking, drinking, physical inactivity, unhealthy body mass index, and unhealthy diet). Two aging measures, Phenotypic Age Acceleration (PhenoAgeAccel) and Biological Age Acceleration (BioAgeAccel) were calculated using nine and seven blood biomarkers, respectively, with a higher value indicating the acceleration of aging. The outcomes included incident cardiovascular disease (CVD), incident cancer, and all-cause mortality in UKB; CVD mortality, cancer mortality, and all-cause mortality in NHANES. A general linear regression model, Cox proportional hazards model, and formal mediation analysis were performed.

RESULTS:

The unhealthy lifestyles score was positively associated with PhenoAgeAccel (UKB: β=0.741; NHANES: β=0.874, all P<0.001). We further confirmed the respective associations of PhenoAgeAccel and unhealthy lifestyles with the outcomes in UKB and NHANES. The mediation proportion of PhenoAgeAccel in associations of unhealthy lifestyles with incident CVD, incident cancer, and all-cause mortality were 20.0%, 17.8%, and 26.6% (all P <0.001) in UKB, respectively. Similar results were found in NHANES. The findings were robust when using another aging measure—BioAgeAccel.

CONCLUSIONS:

Accelerated aging partially mediated the associations of lifestyles with cardiovascular disease, cancer, and mortality in UK and US populations. The findings reveal a novel pathway and the potential of geroprotective programs in mitigating health inequality in late life beyond lifestyle interventions.

INTRODUCTION

Unhealthy lifestyle factors, such as lack of exercise ¹, unhealthy diet ², and smoking ³, are independently associated with multiple adverse health outcomes ^{4, 5}, such as cardiovascular disease (CVD) and all-cause mortality. Furthermore, such factors are associated with each other and tend to have synergistic effects on health ^{6, 7}. A few randomized controlled trials have shown that multi-domain and intensive lifestyle interventions as a whole could improve cognitive function and reduce the prevalence of frailty, a geriatric syndrome ^{8, 9}. Despite the observed benefit of lifestyle interventions, the potential mechanisms explaining how healthy lifestyles contribute to the reduction of adverse health outcomes remain poorly understood.

Accelerated aging is the primary risk factor for chronic diseases and death ¹⁰. Geroscience researchers hypothesize that therapies or preventive programs targeting aging would alleviate the incidence or severity of most chronic diseases ¹⁰. Interestingly, emerging evidence suggests that unhealthy lifestyles are associated with the acceleration of aging in various populations ^{9, 11–13}. For instance, smoking and higher body mass index (BMI) are found to be strongly associated with biological aging ¹⁴. Thus, we hypothesize that aging could play a mediating role in the associations of unhealthy lifestyles with adverse health outcomes, which, however, has rarely been investigated in previous literature.

To test this hypothesis, we used two well-validated aging measures, Phenotypic Age Acceleration (PhenoAgeAccel)^{15, 16} and Biological Age Acceleration (BioAgeAccel)¹⁷ that were based on multi-system blood biomarkers. We have demonstrated that these two comprehensive aging measures could predict mortality and morbidity risk across different subpopulations, even among those who are disease-free ¹⁶. These two aging measures provide useful indicators for differentiating at-risk individuals and evaluating the efficacy of interventions, and it facilitates the investigation of mechanisms of aging ¹⁵. Here, with these two validated aging measures, we tested the hypothesis that lifestyle interventions impact health outcomes (CVD, cancer, and mortality) via their effects on aging rates. We used data from the UK Biobank (UKB) and the US National Health and Nutrition Examination Survey (NHANES), which comprised nationally large samples of participants in the UK and the US. More specifically, this study aimed to examine: (1) the associations of unhealthy lifestyles with aging; (2) the associations of aging with adverse health outcomes; and (3) the mediating role of aging in the association of unhealthy lifestyles with adverse health outcomes.

METHODS

The detailed descriptions of methods are presented in Supplementary files: Supplementary Text S1.

Study Population

In UKB, we included 405,944 participants for all-cause mortality (Analytic sample 1). We further excluded participants with CVD at baseline (N= 27,236) and cancer at baseline (N= 33,741). We included 378,708, and 372,203 participants for the analyses of incident CVD (Analytic sample 2), and incident cancer (Analytic sample 3), respectively. In NHANES, we included 9,972 participants for the analysis of CVD mortality, cancer mortality, and all-cause mortality. The flow chart is shown in Figure 1. The detailed descriptions are presented in Supplementary Text S1: Method of Study Population.

Figure 1. Flow chart of two study populations (A for UK Biobank and B for NHANES).

Abbreviations: UK Biobank: UKB; US National Health and Nutrition Examination Survey: NHANES; CVD, cardiovascular diseases.

Assessments of Lifestyles

Consistent with the recommendations from World Health Organization¹⁸, we constructed an unhealthy lifestyles score (range: 0–5), including five lifestyle factors (smoking, drinking, physical inactivity, unhealthy BMI, and unhealthy diet) assessed at baseline by a touchscreen questionnaire in UKB and NHANES. Refer to a previous study¹⁹, we assigned the same weight to each lifestyle factor in constructing the unhealthy lifestyles score to evaluate the potential for improvement in health with aging through lifestyle modification. The unhealthy lifestyles score was subsequently categorized as favorable (score: 0 to 1), intermediate (score: 2 to 3), and unfavorable (score: 4 to 5) lifestyles. The detailed descriptions are presented in Supplementary Text S1: Method of Assessments of Lifestyles.

Aging Measures

PhenoAgeAccel was previously developed and validated using data from NHANES ^{16, 20} and has been applied to UKB participants in our recent work ¹⁵. Briefly, nine biomarkers (i.e., albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean corpuscular volume (NHANES: mean cell volume), red blood cell distribution width, alkaline phosphatase, white blood cell count) and chronological age collected at baseline were used to calculate PhenoAge (in years), with the equation as follows:

$$\mathrm{P}\mathrm{h}\mathrm{e}\mathrm{n}\mathrm{o}\mathrm{A}\mathrm{g}\mathrm{e}=141.50225+\frac{\ln \left\{-0.00553\times \ln \left(1-\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{y}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{k}\right)\right\}}{0.090165}$$

Where

$$\text{mortality risk}=1-e^{{-e}^{xb}(\exp \left(120\mathrm{*}\gamma \right)-1)/\gamma }$$

$$\mathrm{\gamma }=0.0076927$$

and

$$\begin{gathered} \text{xb}=-19.907-0.0336\times \mathrm{a}\mathrm{l}\mathrm{b}\mathrm{u}\mathrm{m}\mathrm{i}\mathrm{n}+0.0095\times \mathrm{c}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{e}+0.1953\times \mathrm{g}\mathrm{l}\mathrm{u}\mathrm{c}\mathrm{o}\mathrm{s}\mathrm{e} \\ +0.0954\times \mathrm{l}\mathrm{n}(\mathrm{C}\text{-reactive protein)}-0.012\times \mathrm{l}\mathrm{y}\mathrm{m}\mathrm{p}\mathrm{h}\mathrm{o}\mathrm{c}\mathrm{y}\mathrm{t}\mathrm{e}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{g}\mathrm{e} \\ +0.0268\times \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{u}\mathrm{s}\mathrm{c}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{r}\mathrm{v}\mathrm{o}\mathrm{l}\mathrm{u}\mathrm{m}\mathrm{e}+0.3306\times \mathrm{r}\mathrm{e}\mathrm{d}\mathrm{c}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{d}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{r}\mathrm{i}\mathrm{b}\mathrm{u}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{w}\mathrm{i}\mathrm{d}\mathrm{t}\mathrm{h} \\ +0.00188\times \mathrm{a}\mathrm{l}\mathrm{k}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{p}\mathrm{h}\mathrm{o}\mathrm{s}\mathrm{p}\mathrm{h}\mathrm{a}\mathrm{t}\mathrm{a}\mathrm{s}\mathrm{e}+0.0554\times \mathrm{w}\mathrm{h}\mathrm{i}\mathrm{t}\mathrm{e}\mathrm{b}\mathrm{l}\mathrm{o}\mathrm{o}\mathrm{d}\mathrm{c}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{c}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t} \\ +0.0804\times \mathrm{c}\mathrm{h}\mathrm{r}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{g}\mathrm{e} \end{gathered}$$

PhenoAgeAccel was calculated as a residual from a linear regression of PhenoAge against chronological age ^{15, 16}.

BioAge (in years) was calculated using seven biomarkers (i.e., albumin, creatinine, C-reactive protein, alkaline phosphatase, glycated hemoglobin A1c, systolic blood pressure, and total cholesterol) and chronological age at baseline with the algorithm proposed by Klenmera and Doubal (see the equation below) ²¹.

$$\mathrm{B}\mathrm{i}\mathrm{o}\mathrm{A}\mathrm{g}\mathrm{e}=\frac{{\sum }_{j=1}^7\left(x_j-q_j\right)\left(\frac{k_j}{s_j^2}\right)+\frac{CA}{s_{BA}^2}}{{\sum }_j^7{\left(\frac{k_j}{s_j}\right)}^2+\frac{1}{s_{BA}^2}}$$

In the above equation, x represents the value of the biomarker j; CA represents chronological age; q, k, and s represent parameters when regressed the biomarker j on CA; sBA represents a scaling factor equal to the square root of the variance in chronological age explained by the biomarker set.

BioAgeAccel was calculated as a residual from a linear regression of BioAge against chronological age ^{15, 17}.

Ascertainment of Adverse Health Outcome

In UKB, the outcomes – incident CVD, incident cancer, and all-cause mortality – were ascertained through linked hospital admissions data through Aug 29, 2021. In NHANES, CVD mortality, cancer mortality, and all-cause mortality during follow-up were based on linked data from records taken from the National Death Index through December 31, 2015, provided through the Centers for Disease Control and Prevention.

The detailed descriptions are presented in Supplementary Text S1: Method of Ascertainment of Adverse Health Outcome.

Covariates

We considered the following covariates through questionnaires, i.e., age, sex, race/ ethnicity, education level, occupational status (UKB only), Townsend deprivation index (TDI) (UKB only), marital status (NHANES only), and family poverty income ratio (PIR) (NHANES only). The detailed descriptions are presented in Supplementary Text S1: Method of Covariates.

Statistical Analysis

The basic characteristics of the study participants were summarized by sex using mean and standard deviation (SD) for continuous variables and number (percentage) for categorical variables. Basic characteristics by sex were compared using the analysis of variance (ANOVA) or χ² test.

To evaluate the potential mediating role of aging (measured by PhenoAgeAccel and BioAgeAccel) in the association of unhealthy lifestyles with adverse health outcomes, we considered three pathways (a: unhealthy lifestyles → aging; b: aging → adverse health outcomes; c: unhealthy lifestyles → adverse health outcomes. Figure 2). For path a, general linear regression models were used to investigate the associations of unhealthy lifestyle scores with aging. The coefficient (β) and standard error (SE) were documented in model 1. According to previous studies ^{16, 22}, model 1 included chronological age, gender, education level, race, occupational status (UKB only), TDI (UKB only), marital status (NHANES only), and PIR (NHANES only). For path b, Cox proportional hazards regression models were performed to investigate the associations of aging with adverse health outcomes. The hazard ratios (HR) and 95% confidence intervals (CIs) were documented in model 1. For path c, Cox proportional hazards regression models were used to estimate the HR and 95% CIs of adverse health outcomes associated with unhealthy lifestyles and aging in model 1, model 2 (additionally including PhenoAgeAccel based on model 1), and model 3 (additionally including BioAgeAccel based on model 1). To determine whether aging mediated the associations of unhealthy lifestyles with adverse health outcomes, the formal mediation analyses were performed to estimate the mediation proportions and 95% CIs in model 1 ²³.

Figure 2. The mediating role of aging in the association of unhealthy lifestyles and adverse health outcomes.

We performed several sensitivity analyses to test the robustness of our results. First, given the sex difference in aging and health, we further conducted a stratified analysis by sex to investigate the associations of unhealthy lifestyles with adverse health outcomes and the mediation proportion of aging in adverse health outcomes attributed to unhealthy lifestyles. Second, to reduce the potential reverse causation, we repeated the main analyses after excluding outcomes that occurred within the first two years of follow-up. Third, we repeated the main analyses using the difference method ²⁴ to evaluate the mediating role of PhenoAgeAccel. Forth, to account for the varying magnitudes of association between different lifestyle factors and outcomes, we constructed a weighted unhealthy lifestyle score based on each lifestyle factor’s association with the outcomes. Since this weighted score was not a continuous variable, participants were grouped by values close to the tertiles of this weighted score ²⁵. Fifth, we repeated the mediating analyses by using the individual lifestyle factors to evaluate the mediation proportion of aging, and the factors were mutually included in models.

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC), and R version 3.6.3 (2020–02-29). We considered a two-sided P value <0.05 to be statistically significant.

RESULTS

Population Characteristics

Table 1 shows sex-specific characteristics of UKB (N= 405,944) and NHANES participants (N= 9,972), respectively. In UKB, the chronological age and PhenoAge of participants were 57.0 years (SD= 8.0) and 53.2 years (SD= 10.0); In NHANES, the chronological age and PhenoAge of participants were 49.6 years (SD= 17.6) and 43.3 years (SD= 19.8), respectively. In UKB, we recorded 29,726 incident CVD cases, 34,413 incident cancer cases, and 27,471 deaths during a mean follow-up of 11.9 years. In NHANES, we recorded 258 CVD deaths, 325 cancer deaths, and 1,374 all-cause deaths during a mean follow-up of 9.8 years. The proportion of participants with favorable, intermediate, and unfavorable lifestyles were 23.0%, 59.9%, and 17.1% in UKB, and 24.4%, 66.4%, and 9.2% in NHANES, respectively. The distributions of unhealthy lifestyle scores of UKB and NHANES were shown in Supplementary Figure S1 and Supplementary Figure S2.

Table 1.

Baseline characteristics of participants.

Characteristics	Total	Women	Men	P value
UKB	(n= 405,944)	(n= 218,528)	(n= 187,416)
Chronological age in years, Mean± SD	57.0±8.0	56.8±7.9	57.1±8.1	<0.001
Race/ethnicity, n (%)				<0.001
White	385,421 (94.9)	207,622 (95.0)	177,799 (94.9)
Mixed	2,364 (0.6)	1,463 (0.7)	901 (0.5)
South Asian	7,367 (1.8)	3,290 (1.5)	4,077 (2.2)
Black	6,087 (1.5)	3,441 (1.6)	2,646 (1.4)
Chinese	1,197 (0.3)	750 (0.3)	447 (0.2)
Others	3,508 (0.9)	1,962 (0.9)	1,546 (0.8)
TDI, Mean± SD	−1.3±3.1	−1.4±3.0	−1.3±3.1	0.010
Education level, n (%)				<0.001
High	140,280 (34.6)	71,734 (32.8)	68,546 (36.6)
Intermediate	132,828 (32.7)	78,021 (35.7)	54,807 (29.2)
Low	132,836 (32.7)	68,773 (31.5)	64,063 (34.2)
Occupation status, n (%)				<0.001
Working	235,655 (58.1)	120,882 (55.3)	114,773 (61.2)
Retired	134,133 (33.0)	76,294 (34.9)	57,839 (30.9)
Other	36,156 (8.9)	21,352 (9.8)	14,804 (7.9)
Smoking, yes, n (%)	179,506 (44.9)	83,547 (38.9)	95,959 (51.8)	<0.001
Drinking, yes, n (%)	147,559 (42.6)	80,280 (44.9)	67,279 (40.1)	<0.001
Physical inactivity, yes, n (%)	106,264 (26.7)	60,562 (28.3)	45,702 (24.8)	<0.001
Unhealthy diet, yes, n (%)	255,648 (63.0)	128,229 (58.7)	127,419 (68.0)	<0.001
Unhealthy BMI, yes, n (%)	276,778 (68.4)	134,697 (61.8)	142,081 (76.1)	<0.001
Unhealthy lifestylesa				<0.001
Favorable	93,491 (23.0)	59,754 (27.3)	33,737 (18.0)
Intermediate	243,161 (59.9)	128,617 (58.9)	114,544 (61.1)
Unfavorable	692,92 (17.1)	30,157 (13.8)	39,135 (20.9)
PhenoAge, Mean±SD	53.3±10.0	52.2±9.5	54.4±10.5	<0.001
NHANES	(n= 9,972)	(n= 5,099)	(n= 4,873)
Chronological age in years, Mean±SD	49.6±17.6	49.5±17.7	49.7±17.5	0.567
Race/ethnicity, n (%)				0.165
Non-Hispanic white	5,162 (51.8)	2,664 (52.2)	2,498 (51.3)
Non-Hispanic black	1,800 (18.1)	895 (17.6)	905 (18.6)
Mexican-American	2,021 (20.3)	1,057 (20.7)	964 (19.8)
Other	989 (9.9)	483 (9.5)	506 (10.4)
PIR, n (%)				<0.001
≤ 1.30	2,801 (28.1)	1,318 (25.8)	1,483 (30.4)
1.31~3.50	3,884 (38.9)	1,999 (39.2)	1,885 (38.7)
≥ 3.51	3,287 (33.0)	1,782 (34.9)	1,505 (30.9)
Education level, n (%)				<0.001
Less than HS	2,841 (28.5)	1,521 (29.8)	1,320 (27.1)
HS/GED	2,380 (23.9)	1,216 (23.8)	1,164 (23.9)
Some college	2,751 (27.6)	1,285 (25.2)	1,466 (30.1)
College	2,000 (20.1)	1,077 (21.1)	923 (18.9)
Marital status, n (%)				<0.001
Married or living with partner	6,333 (63.5)	3,539 (69.4)	2,794 (57.3)
Other	3,639 (36.5)	1,560 (30.6)	2,079 (42.7)
Smoking, yes, n (%)	4,829 (48.5)	2,903 (57.0)	1,926 (39.5)	<0.001
Drinking, yes, n (%)	711 (7.4)	436 (8.8)	275 (6.0)	<0.001
Physical inactivity, n (%)	3,410 (52.3)	1,631 (46.8)	1,779 (58.7)	<0.001
Unhealthy diet, yes, n (%)	5,883 (60.1)	3,224 (64.4)	2,659 (55.7)	<0.001
Unhealthy BMI, yes, n (%)	6,774 (71.2)	3,533 (72.8)	3,241 (69.5)	<0.001
Unhealthy lifestyles				<0.001
Favorable	2,436 (24.4)	987 (19.4)	1,449 (29.7)
Intermediate	6,619 (66.4)	3,557 (69.8)	3,062 (62.8)
Unfavorable	917 (9.2)	555 (10.9)	362 (7.4)
PhenoAge, Mean± SD	43.6±19.8	44.3±20.3	42.9±19.2	0.005

Abbreviations: SD, standard deviation; TDI, Townsend deprivation index; BMI, body mass index; PIR, poverty income ratio; HS/GED, less than high school (HS) /general educational development (GED); PhenoAge, Phenotypic Age.

^aFavorable (unhealthy lifestyle score ranged from 0 to 1), Intermediate (unhealthy lifestyle score ranged from 2 to 3), and Unfavorable (unhealthy lifestyle score ranged from 4 to 5) lifestyles.

Associations of Unhealthy Lifestyles with Aging (Path a)

The unhealthy lifestyles score was positively associated with PhenoAgeAccel (UKB: β=0.741± 0.008 SE, P<0.001; NHANES: β=0.874± 0.064 SE, P<0.001) (Supplementary Table S1) and BioAgeAccel (UKB: β=0.213± 0.003 SE, P<0.001; NHANES: β=0.106± 0.032 SE, P<0.001) (Supplementary Table S2).

Associations of Aging with Adverse Health Outcomes (Path b)

PhenoAgeAccel and BioAgeAccel were positively associated with the risk of adverse health outcomes in both UKB and NHANES, as we observed in the previous study ¹⁶. For instance, for each one-year increase in PhenoAgeAccel, the risk of incident CVD, incident cancer, and all-cause mortality increased by 4% (HR:1.04, 95%CI: 1.04, 1.05), 2% (HR: 1.02, 95%CI: 1.02, 1.02), and 7% (HR: 1.07, 95%CI: 1.07, 1.07), respectively in UKB (Supplementary Table S3, Supplementary Table S4).

Mediations Analysis of Aging on Associations of Unhealthy Lifestyles with Adverse Health Outcomes (Path c)

Table 2 and Table 3 presented the associations of unhealthy lifestyles with adverse health outcomes and the mediation proportion of aging (measured by PhenoAgeAccel and BioAgeAccel, respectively) in adverse health outcomes attributed to unhealthy lifestyles. As shown in table 2, in UKB, unhealthy lifestyles were significantly associated with adverse health outcomes. When unfavorable lifestyles were compared with favorable lifestyles, the risks of incident CVD, incident cancer, and all-cause mortality were increased by 59% (HR: 1.59, 95%CI: 1.53, 1.65), 33% (HR: 1.33, 95%CI: 1.29, 1.38), and 84% (HR: 1.84, 95%CI: 1.77, 1.92) in model 1, respectively. After further including PhenoAgeAccel in model 2, the magnitude of increased risk of incident CVD, incident cancer, and all-cause mortality was reduced. The mediation proportion of PhenoAgeAccel in associations of unhealthy lifestyles with risk of incident CVD, incident cancer, and all-cause mortality were 20.0%, 17.8%, and 26.6% (all P values < 0.001), respectively. In NHANES, unhealthy lifestyles were significantly associated with adverse health outcomes as well, except for CVD mortality. Although the association of unhealthy lifestyles with the risk of CVD mortality was not statistically significant, effect sizes were broadly consistent with the findings from NHANES and the analyses appear to be underpowered due to a small number of cases in the favorable lifestyle group. Similarly, after further including PhenoAgeAccel in model 2, the magnitude of increased risk of cancer mortality and all-cause mortality was reduced. The mediation proportion of PhenoAgeAccel in associations of lifestyles with risk of cancer mortality and all-cause mortality were 25.7% and 35.2% (all P values < 0.05), respectively.

Table 2.

Associations of unhealthy lifestyles with outcomes and mediation proportion of aging (measured by PhenoAgeAccel).

Unhealthy lifestyle	No. of events/No. participants	Model 1a	Model 2b	Mediation proportion of PhenoAgeAccel (95% CI) a	P value
		HR (95% CI)	HR (95% CI)
	UKB
	Incident CVD
	29,726/378,708
Favorable c	4,967/89,538	Ref.	Ref.
Intermediate	18,234/226,657	1.33 (1.29, 1.38)	1.27 (1.23, 1.31)
Unfavorable	6,525/62,513	1.59 (1.53, 1.65)	1.45 (1.39, 1.51)	0.200 (0.182, 0.220)	<0.001
	P for trend	<0.001	<0.001
	Incident cancer
	34,413/372,203
Favorable	6,695/ 85,776	Ref.	Ref.
Intermediate	20,700/223,155	1.15 (1.12, 1.18)	1.12 (1.09, 1.15)
Unfavorable	7,018/63,272	1.33 (1.29, 1.38)	1.27 (1.22, 1.31)	0.178 (0.155, 0.200)	<0.001
	P for trend	<0.001	<0.001
	All-cause mortality
	27,471/405,944
Favorable	4,146/93,491	Ref.	Ref.
Intermediate	16,174/243,161	1.34 (1.29, 1.39)	1.22 (1.18, 1.26)
Unfavorable	7,151/69,292	1.84 (1.77, 1.92)	1.55 (1.49, 1.61)	0.266 (0.250, 0.290)	<0.001
	P for trend	<0.001	<0.001
	NHANES
	CVD mortality
	258/9,972
Favorable	54/2,436	Ref.	Ref.
Intermediate	181/6,619	1.23 (0.90, 1.67)	1.14 (0.83, 1.55)
Unfavorable	23/917	1.37 (0.82, 2.28)	1.23 (0.74, 2.05)	-d	-
	P for trend	0.175	0.373
	Cancer mortality
	325/9,972
Favorable	69/2,436	Ref.	Ref.
Intermediate	223/6,619	1.22 (0.92, 1.61)	1.15 (0.87, 1.52)
Unfavorable	33/917	1.60 (1.04, 2.47)	1.47 (0.95, 2.27)	0.257 (0.135, 1.010)	0.004
	P for trend	0.036	0.101
	All-cause mortality
	1,374/9,972
Favorable	317/2,436	Ref.	Ref.
Intermediate	921/6,619	1.11 (0.98, 1.27)	1.04 (0.92, 1.18)
Unfavorable	136/917	1.48 (1.20, 1.82)	1.34 (1.09, 1.65)	0.352 (0.245, 0.610)	<0.001
	P for trend	<0.001	0.023

Abbreviations: PhenoAgeAccel, Phenotypic Age Acceleration; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease;

^aModel 1 included age; sex; education level; self-reported race/ethnicity; occupational status (UKB only); Townsend deprivation index (TDI) (UKB only); marital status (NHANES only); poverty income ratio (PIR) (NHANES only);

^bModel 2 additional included PhenoAgeAccel based on model 1;

^cFavorable (unhealthy lifestyle score ranged from 0 to 1), Intermediate (unhealthy lifestyle score ranged from 2 to 3), and Unfavorable (unhealthy lifestyle score ranged from 4 to 5) lifestyle;

^d“-” The results were not significant.

Table 3.

Associations of unhealthy lifestyles with adverse health outcomes and mediation proportion of aging (measured by BioAgeAccel).

Unhealthy lifestyle	No. of events/No. participants	Model 1a	Model 3b	Mediation proportion of BioAgeAccel (95% CI) a	P value
		HR (95% CI)	HR (95% CI)
	UKB
	Incident CVD
	29,726/378,708
Favorable c	4,967/89,538	Ref.	Ref.
Intermediate	18,234/226,657	1.33 (1.29, 1.38)	1.27 (1.23, 1.32)
Unfavorable	6,525/62,513	1.59 (1.53, 1.65)	1.46 (1.41, 1.52)	0.187 (0.174, 0.210)	<0.001
	P for trend	<0.001	<0.001
	Incident cancer
	34,413/372,203
Favorable	6,695/85,776	Ref.	Ref.	-d	-
Intermediate	20,700/223,155	1.15 (1.12, 1.18)	1.15 (1.11, 1.18)
Unfavorable	7018/63,272	1.33 (1.29, 1.38)	1.33 (1.28, 1.38)
	P for trend	<0.001	<0.001
	All-cause mortality
	27,471/405,944
Favorable	4,146/93,491	Ref.	Ref.
Intermediate	16,174/243,161	1.34 (1.29, 1.39)	1.31 (1.27, 1.36)
Unfavorable	7,151/69,292	1.84 (1.77, 1.92)	1.78 (1.71, 1.85)	0.065 (0.057, 0.070)	<0.001
	P for trend	<0.001	<0.001
	NHANES
	CVD mortality
	258/9,972
Favorable	54/2,436	Ref.	Ref.
Intermediate	181/6,619	1.23 (0.90, 1.67)	1.19 (0.87, 1.63)
Unfavorable	23/917	1.37 (0.82, 2.28)	1.36 (0.82, 2.26)	-	-
	P for trend	0.175	0.192
	Cancer mortality
	325/9,972
Favorable	69/2,436	Ref.	Ref.
Intermediate	223/6,619	1.22 (0.92, 1.61)	1.21 (0.92, 1.60)
Unfavorable	33/917	1.60 (1.04, 2.47)	1.60 (1.04, 2.46)	-	-
	P for trend	0.036	0.038
	All-cause mortality
	1,374/9,972
Favorable	317/2,436	Ref.	Ref.
Intermediate	921/6,619	1.11 (0.98, 1.27)	1.09 (0.96, 1.24)
Unfavorable	136/917	1.48 (1.20, 1.82)	1.46 (1.19, 1.80)	0.066 (0.019, 0.160)	0.004
	P for trend	<0.001	0.002

Abbreviations: BioAgeAccel, Biological Age Acceleration; HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease;

^aModel 1 included age; sex; education level; self-reported race/ethnicity; occupational status (UKB only); Townsend deprivation index (TDI) (UKB only); marital status (NHANES only); poverty income ratio (PIR) (NHANES only);

^bModel 3 additional included BioAgeAccel based on model 1;

^cFavorable (unhealthy lifestyle score ranged from 0 to 1), Intermediate (unhealthy lifestyle score ranged from 2 to 3), and Unfavorable (unhealthy lifestyle score ranged from 4 to 5) lifestyles.

^d“-” The results were not significant.

Using another aging measure—BioAgeAccel, we observed similar results, but the mediation proportion of BioAgeAccel (ranging from 6.5% in association of unhealthy lifestyles with all-cause mortality to 18.7% in association of unhealthy lifestyles with incident CVD in UKB and 6.6% in association of unhealthy lifestyles with all-cause mortality in NHANES) was smaller than that of PhenoAgeAccel (Table 3).

Sensitivity Analyses

First, the mediation proportion of PhenoAgeAccel in associations of lifestyles with incident CVD was slightly smaller in women than that in men (P for interaction< 0.001). The above sex-specific results were not observed in NHANES (Supplementary Table S5). Second, after excluding outcomes that occurred within the first two years of follow-up, the association between unhealthy lifestyles and adverse health outcomes was maintained and the mediation proportion of PhenoAgeAccel and BioAgeAccel in associations of lifestyles with adverse health outcomes did not change substantially (Supplementary Table S6). Third, the difference method showed similar results (Supplementary Table S7). Forth, using weighted unhealthy lifestyle scores, the results were largely consistent, except that the mediation proportion in the association of unhealthy lifestyles with all-cause mortality became not statistically significant. In UKB, the mediation proportion ranged from 14.7% to 19.8% (for PhenoAgeAccel) and from 3.1% to 20.3% (for BioAgeAccel); in NHANES, it ranged from 21.8% to 45.4% (for PhenoAgeAccel) (Supplementary Table S8). Fifth, overall, using the individual lifestyle factors to evaluate the mediation proportion of aging, we observed that the associations of different lifestyle factors with outcomes and the mediation proportion of aging in the above associations varied a lot in UKB and NHANES (Supplementary Table S9) (The detailed descriptions are presented in Supplementary Text S2: Result of Sensitivity Analyses).

DISCUSSION

Based on the two large population-based cohort studies, this study found that accelerated aging partially mediated the associations of unhealthy lifestyles with adverse health outcomes (CVD, cancer, and mortality) in the UK and US populations. The mediation proportion ranged from 17.8 % to 26.6% in UKB and 25.7% to 35.2% in NHANES. The findings reveal a novel pathway linking unhealthy lifestyles to adverse health outcomes and the potential for lifestyle intervention.

To our knowledge, studies investigating the mediating role of aging in the association of lifestyles with adverse health outcomes are scarce, with one exception by our group ²⁶. In this previous study of older Chinese adults, aging is suggested to partially mediate the association of lifestyles with mortality. In the present study, we enlarged the age range by including a large sample size of participants aged 40–69 years old in UKB and 20–84 years old in NHANES. We further included incident CVD and cancer in UKB, and CVD mortality and cancer mortality in NHANES for adverse health outcomes, because CVD and cancer are the leading causes of increased disability-adjusted life-years and mortality globally ²⁷. We extended our previous observation to more general populations and more adverse health outcomes, enhancing the significance of the findings in disease prevention and intervention.

The explanation of the mediating role of aging in the associations of unhealthy lifestyles with CVD, cancer, and mortality might be related to inflammation/immunity. First, unhealthy lifestyles, such as smoking, and obesity, are likely to contribute to accelerated aging by triggering low-grade chronic inflammation ^{28, 29}, which increases individuals’ susceptibility to diverse adverse outcomes. Second, PhenoAgeAccel is found to be associated with increased activation of pro-inflammatory pathways ²⁰. Additionally, as blood biomarkers involved in PhenoAgeAccel could reflect individuals’ status of inflammation (i.e., C-reactive protein) and immune system (i.e., lymphocyte percent, mean (red) cell volume, red cell distribution width, and white blood cell count), PhenoAgeAccel could capture risks of multiple diseases from the preclinical perspective. Genes linked to PhenoAgeAccel are overrepresented in the immune system, carbohydrate homeostasis pathways, and cell function, suggesting that the effect of unhealthy lifestyles on accelerated aging may be through multiple pathways including inflammation/immune pathways¹⁵. Third, inflammation and dysfunctional immune function contribute to the pathogenesis of multiple diseases, such as CVD and cancer, which eventually contribute to death ^{30, 31}. Based on the evidence above, we speculate that unhealthy lifestyles lead to the pathogenesis of disease through accelerated aging (unhealthy lifestyles → accelerated aging (inflammation and immune dysfunction) → adverse health outcomes).

Using another aging measure–— BioAgeAccel, we observed similar results though with weaker effect sizes, which enhanced our speculation that aging partially mediated the associations of unhealthy lifestyles with CVD, cancer, and mortality. Researchers revealed that genes linked to BioAgeAccel were enriched in lipid-related pathways ³² which were reported to be associated with the regulation of aging and longevity ¹⁵. Since dysregulation in lipid metabolism is another prominent pathway in the pathogenesis of CVD and cancer ^{33, 34}, there may be another pathway that unhealthy lifestyles lead to the pathogenesis of disease through aging (unhealthy lifestyles → accelerated aging (lipid metabolism) → adverse health outcomes). The mutual verification of the mediating role of aging by two aging measures further tells us that if more aging measures are available, we could capture aging more comprehensively. Many aging measures are under development, involving a single hallmark of aging (eg, DNA methylation) ³⁵ or composite aging measures by integrating multi-omics data ³⁶. Hence, we have the opportunity to further understand the important role of aging, a complex process, in the pathway linking unhealthy lifestyles to adverse health outcomes. Future studies are needed to validate our findings using more aging measures.

The associations of unhealthy lifestyles with CVD, cancer, and mortality observed in this study highlight the potential of multi-domain lifestyle interventions (i.e., healthy diet, no smoking, no alcohol consumption, regular exercise, and healthy BMI) in disease prevention. As the implementation of multi-domain interventions in preventing cognitive decline and frailty has been suggested to be effective ^{8, 9, 37}, it is meaningful to extend such interventions to CVD, cancer, and even aging itself. Besides, from the point of view of pharmacological intervention, many anti-aging drugs have been developed such as caloric restriction mimetics (resveratrol, rapamycin, metformin), senolytics, and synthetic sirtuin activators, which are promising to alleviate or delay age-related conditions ^38–42. With a more comprehensive understanding of aging hallmarks, more aging interventions are expected to flourish. Likewise, advances in biomedical research may bring us the possibility of anti-aging and delaying the onset of diseases in the future.

We caution that it would premature to interpret our results as showing that lifestyle interventions reduce adverse health outcomes by slowing aging. PhenoAgeAccel and BioAgeAccel are currently well-validated and widely used aging measures in terms of their mimicking of the aging process and predictive utility in adults without chronic diseases ¹⁶. However, the complexity of aging and the unclear definition of normal aging ^{43, 44} make it difficult to explicitly develop a perfect aging measure and relate these aging measures to mechanisms of “normal aging”. For example, it is possible that many of the adverse health outcomes observed today are a result of the cumulative impact of industrialized lifestyles over decades, and not of the basic human biology of aging. In that case, PhenoAgeAccel and BioAgeAccel might at least partially reflect these aging-independent lifestyle impacts, and we could still see the relationships observed here. At a practical level, such nuances could be important to disentangle mechanisms and help tailor interventions to specific aspects of aging and/or health that are most important for preventing adverse outcomes.

The strengths of our study included the large sample sizes of middle-aged and older adults from the UK and the US and the use of two well-validated aging measures. Nevertheless, several limitations should be noted. First, the assessment of lifestyle was based on self-reported and relatively simple information and thus, could be subject to measurement and misclassification bias. Second, the UKB participants are healthier than the general population and are mostly Caucasian. Hence, our findings may not be generalizable to other populations. Third, there was likely residual confounding even though we controlled for various covariates. Forth, we used aging measures that were constructed by combining multi-system blood biomarkers instead of considering the blood biomarkers as individual variables. Our approach may lose some information; however, given the complexity of aging, composite measures like PhenoAgeAccel and BioAgeAccel based on multi-system blood biomarkers may better reflect multiple systems as a whole ⁴⁵. Future studies are needed to further validate our findings in other populations and to develop appropriate lifestyle interventions and geroprotective programs.

CONCLUSION

This study, for the first time, demonstrated that accelerated aging partially mediated the associations of unhealthy lifestyles with CVD, cancer, and mortality in UK and US populations. The findings reveal a novel pathway and the potential of geroprotective programs in mitigating health inequality in late-life beyond lifestyle interventions.

Supplementary Material

Supinfo

Supplementary Text S1.

Supplementary Text S2.

Supplementary Table S1. Associations of unhealthy lifestyles score with PhenoAgeAccel in UKB and NHANES

Supplementary Table S2. Associations of PhenoAgeAccel with adverse health outcome in UKB and NHANES

Supplementary Table S3. Associations of unhealthy lifestyle score with BioAgeAccel in UKB and NHANES

Supplementary Table S4. Associations of BioAgeAccel with adverse health outcome in UKB and NHANES

Supplementary Table S5. Associations of unhealthy lifestyles with adverse health outcomes and mediation proportion of aging (measured by PhenoAgeAccel) in adverse health outcomes attributed to different lifestyles stratified by sex

Supplementary Table S6. Associations of unhealthy lifestyle with adverse health outcomes and mediation proportion of aging (measured by PhenoAgeAccel and BioAgeAccel) in adverse health outcomes attributed to different lifestyles after excluded outcome occurred within the first two years of follow-up

Supplementary Table S7. The mediation proportion of PhenoAgeAccel on the associations of unhealthy lifestyles with adverse health outcomes (the difference method)

Supplementary Table S8. Associations of weighted unhealthy lifestyles scores with outcomes and mediation proportion of aging (measured by PhenoAgeAccel and BioAgeAccel)

Supplementary Table S9. Associations of individual unhealthy lifestyles factors with outcomes and mediation proportion of aging (measured by PhenoAgeAccel and BioAgeAccel)

Supplementary Figure S1. Distribution of unhealthy lifestyle score in UKB

Supplementary Figure S2. Distribution of unhealthy lifestyle score in NHANES

KEY POINTS:

• Accelerated aging partially mediated the associations of lifestyles with adverse health outcomes in UK and US populations.

• This study revealed a novel pathway and geroprotective programs in mitigating health inequality beyond lifestyle interventions.

Why does this matter?

Our findings highlight the potential of multi-domain lifestyle interventions in chronic disease prevention and even in slowing aging per se. Aging is complicated by the lack of consensus on what normal aging is. In this case, Phenotypic Age Acceleration and Biological Age Acceleration might at least partially reflect these aging-independent lifestyle impacts, and we could still see the relationships observed here. At a practical level, such nuances could be important to disentangle mechanisms and help tailor interventions to specific aspects of aging and/or health that are most important for preventing adverse outcomes. The findings reveal a novel pathway and the potential of geroprotective programs in mitigating health inequality in late life beyond lifestyle interventions.