A landscape of metabolic burden and its cardiovascular impact across the adult lifespan: a prospective cohort study of 7.5 million people in Northwestern China

Summary

Background

Metabolic abnormalities are increasingly prevalent in young adults and commonly cluster in the elderly. We aimed to map the landscape of metabolic burden and quantify the related cardiovascular impact across the adulthood in Chinese population.

Methods

The prospective cohort included 7,513,828 adults (3,907,602 women and 3,606,226 men) from northwestern China who were free of cardiovascular disease (CVD) at baseline. Participants were grouped by age (18–<40, 40–<60, 60–<80, and ≥80 years). Age-specific distribution and time-series trajectories were depicted for 6 major metabolic risk factors, including obesity, hepatic steatosis, hypertension, hyperlipidemia, diabetes, and renal dysfunction. Their age-specific associations with incident CVD and population attributable fractions (PAFs) were assessed, considering potential sex difference.

Findings

Metabolic states exhibited obvious age-dependent sexual dimorphism, characterized by an initial female advantage in young adults which reversed around age 45. Over a median follow-up of 4.5 years, the metabolic factors exhibited strong associations with incident CVD in younger adults which gradually attenuated with advancing age. In particular, the HR for hypertension declined from 3.60 in the youngest group to 2.12 in the middle-aged group, 1.41 in the older group, and 1.21 in the very old (≥80 years). Regardless of age or sex, obesity demonstrated the highest PAF to CVD. For combined metabolic burden, the PAF was highest in middle-aged group (37.89%), followed by people aged 18–<40 years (29.78%) and 60–<80 years (25.24%), and lowest in the very old (15.85%). The PAF was higher in men than women only in those under 40 and the sex disparity reversed in the older groups. Among metabolically abnormal participants, transitioning to a healthier metabolic state, within an average of 1 year, was associated with a substantially reduced risk of CVD, particularly in the younger groups.

Interpretation

The metabolic burden and its cardiovascular effects are heterogeneous across demographic segments, supporting the rationale of implementing age-specific and gender-sensitive strategies for metabolic management and CVD prevention.

Funding

Key Research and Development Project of Xinjiang Uyghur Autonomous Region; Suzhou Applied Basic Research (Medical and Health) Science and Technology Innovation Project; Jiangsu Frontier Technology Research and Development Program.

Research in context.

Evidence before this study

Evidence on the cardiovascular impact of metabolic abnormalities largely comes from middle-aged and older populations, with scanty of data covering the entire adult lifespan and simultaneously considering age and sex heterogeneity. We searched PubMed for English-language publications up to Dec 31, 2025, using the terms (“metabolic abnormalities” OR “metabolic risk factors”) AND (“cardiovascular disease” OR “CVD”) AND (“age-specific” OR “lifespan” OR “young adults” OR “older adults”) AND (“sex difference” OR “gender”). While several studies reported age- or sex-specific associations for individual metabolic factors, none provided a landscape of multiple metabolic abnormalities and quantified the associated cardiovascular risk across the full adult age spectrum.

Added value of this study

Using longitudinal health data from 7.5 million adults in northwestern China, this study provides the first comprehensive, population-based assessment of 6 major metabolic risk factors and their cardiovascular impacts across the entire adult lifespan. We found that metabolic states exhibited an age-dependent sexual dimorphism, with an initial female advantage in young adulthood that reversed around age 45, followed by a second reversal in multimorbidity prevalence around age 50. Regardless of sex, the associations of metabolic abnormalities with CVD risk were most evident among adults under 40 and generally attenuated with advancing age. The population attributable fraction of combined metabolic burden exhibited an inverted U-shape pattern peaking in midlife (40–<60 years old), with obesity being the top contributor across all age groups. For metabolically abnormal individuals, transitioning to a healthier metabolic state, over on average of 1 year, was associated with a substantial reduction in CVD risk, particularly in younger adults. These findings provide critical evidence for personalized metabolic management and CVD prevention strategies tailored to major demographic segments.

Implications of all the available evidence

Despite the high and persistent metabolic burden in older adults, our findings highlight early and middle adulthood as a critical window for achieving maximum cardiovascular benefit from metabolic management. The age-dependent reversal of metabolic advantage in women underscores the need for age-specific and gender-sensitive strategies for metabolic management, particularly considering midlife transitions in women. The dominant and sustained contribution of obesity to CVD burden across all ages calls for prioritization of obesity prevention and control in public health policy. Moreover, the rapid metabolic deterioration observed in young adults, together with the strong risk reduction associated with early metabolic improvement, suggest that timely intervention in this commonly-underrecognized population could yield substantial long-term cardiovascular benefits.

Introduction

Cardiovascular disease (CVD) remains the leading cause of death and a major contributor to disability worldwide, posing a substantial and persistent challenge to global public health.¹ Each year, CVD is responsible for an estimated 19.2 million deaths and 437 million disability-adjusted life years, with China bearing one of the heaviest national burdens.² As CVD is largely preventable, the cornerstone of primary prevention is targeting modifiable risk factors, among which metabolic abnormalities have been shown to make the greatest contribution to cardiovascular burden.3, 4, 5, 6 It is estimated that 42% of incident cases of CVD in China are attributable to 4 major metabolic risk factors, including hypertension, diabetes, hyperlipidemia, and abdominal obesity.³

Over recent decades, amid rapid socioeconomic and lifestyle transitions, obesity and metabolic abnormalities have become increasingly prevalent among young populations, globally but particularly in developing regions.7, 8, 9 Early adulthood constitutes a critical window in which lifelong cardiovascular trajectories can be decisively shaped, but metabolic abnormalities in this age group are often unrecognized and inadequately managed for years.10, 11, 12 On the other hand, advances in healthcare have significantly extended the human lifespan, leading to a growing older demographic for whom multimorbidity is common and complex to manage.¹³ Given the dual health trend which poses increasing socioeconomic challenges, there is an urgent need to characterize the life-course metabolic burden and the related cardiovascular impacts to advance a precision public health paradigm. Besides, East Asians including Chinese have unique metabolic features compared with Western populations, such as a higher amount of visceral fat and greater susceptibility to metabolic dysfunctions at a lower body mass index, younger age at diagnosis of type 2 diabetes, and a poorer lipid profile featured by elevated triglycerides and lower levels of high-density lipoprotein cholesterol.^14,15 Thus, it is essential to expand local cohort studies to better inform public health strategies in China.

In the presented study, leveraging longitudinal, repeatedly collected health data from 7.5 million Chinese adults spanning the entire adult lifespan, we aimed to map the landscape of metabolic burden and quantify the cardiovascular impacts from young adults to the very old. Given the differences in men and women regarding physiological and psychosocial trajectories across the adulthood, sex difference in metabolic features and the associated cardiovascular burden were also assessed. We hypothesized a life-course pattern in which metabolic abnormalities exhibit age- and sex-specific prevalence, have stronger cardiovascular impacts in younger adults, and yield greater benefit from metabolic improvement in early adulthood.

Methods

Study design and population

The present study was conducted and reported in accordance with‌ the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.¹⁶ The Tianshan Community Cohort (originally termed the Northwest China Real-World and Population-based Cohort) is an ongoing, dynamic, population-based prospective cohort built on routinely collected real-world data (https://clinicaltrials.gov/study/NCT07207863).17, 18, 19 This region-wide, multi-center cohort spans the entire Xinjiang Uyghur Autonomous Region (approximately 1.66 million km²), enrolls participants from 12,520 primary healthcare institutions across all prefectures, and includes an ethnically diverse population. Baseline recruitment was conducted from Jan 2019 to Dec 2023 based on local resident health profiles, applying predefined enrollment criteria.¹⁷ Follow-up consists of annual health examinations mandated by the National Basic Public Health Services Program in China and real-time event-driven ascertainment of inpatient records, migration status, and vital status. Annual health examinations comprise questionnaires (general health and lifestyle), laboratory tests, and device-based examinations (e.g., electrocardiography, ultrasonography [abdomen, prostate, bladder, and breast], and chest radiography), with all procedures standardized per national protocols to guarantee uniformity across sites. As described elsewhere, the overall loss-to-follow-up rate was 14.47% during 2019–2023, with no systematic differences in baseline demographic, lifestyle, or health characteristics between participants lost to follow-up and those retained.²⁰ The study received ethical approvals from the Tsinghua University Science and Technology Ethics Committee (Medicine) (THU01-20240123) and the Ethics Committee of the People's Hospital of Xinjiang Uyghur Autonomous Region (KY2023042009). All participants provided written informed consent.

The present study included 10,047,820 participants who had attended at least 1 health examination since 2019, with their initial examination serving as baseline, yielding 3 main analytic samples (Figure S1). From the initial cohort, we excluded those with prevalent angina, coronary heart disease (CHD), stroke, or heart failure at baseline, leaving 9,825,399 CVD-free participants. Among them, missing data for each metabolic measurement and covariate were generally low (<3% for most variables) and broadly similar across age groups (Table S1), consistent with a missing-at-random assumption. After further excluding those with missing metabolic factors or covariates, 7,513,828 participants remained for the primary analysis (Sample 1). Among these, 4,646,653 had at least 1 metabolic risk factor at baseline. After further excluding those who¹ did not attend the second examination within 2 years²; experienced incident CVD before their second examination; and³ had missing data on the metabolic risk factors or covariates at the second examination, 3,712,053 participants were included in the analysis on transition in metabolic status and incident CVD (Sample 2). In addition, a subset of 1,649,758 participants (Sample 3), for which all 5-wave metabolic data were available, was included in the descriptive analysis of time-series trajectories in metabolic states from 2019 to 2023.

Definition of metabolic risk factors

We assessed 6 common metabolic risk factors, including obesity, hepatic steatosis, hypertension, hyperlipidemia, diabetes, and renal dysfunction. These metabolic factors were selected a priori given their well-established pathophysiological links to CVD, and are increasingly prevalent in China leading to substantial public health impacts.^15,21,22 Beyond the classical risk factors emphasized in prior literature,^23,24 we additionally included hepatic steatosis and renal dysfunction, which share core pathophysiological drivers with other metabolic conditions. This integration is now formally endorsed by contemporary clinical frameworks.^25,26 We assessed these factors using both objective measurements from health examinations and clinical diagnoses from hospital records (Table S2). Hepatic steatosis was defined via the Hepatic Steatosis Index, based on established thresholds incorporating liver enzyme levels and anthropometric measures.²⁷ Estimated glomerular filtration rate, a biomarker to assess renal function, was calculated from serum creatinine using the modified CKD-EPI equation.²⁸ Detailed clinical and laboratory criteria are provided in prior reports.27, 28, 29, 30, 31, 32

Ascertainment of outcomes

Event-driven follow-up information was extracted from all tertiary hospitals in the region, covering 163 hospitals and 63 traditional Chinese medicine hospitals.¹⁷ On the basis of hospital inpatient records from Jan 2019 to Dec 2023, cardiovascular events were objectively assessed and coded according to the International Classification of Diseases, 10th Revision (ICD-10). The primary endpoint in this study was major CVD, including CHD (ICD code: I21–I25) and stroke (I60–I64).

Assessment of covariates

A set of covariates was selected based on prior literature and their potential role as confounders in the association between metabolic risk factors and incident CVD^23,24:¹ sociodemographic characteristics including age, sex, ethnicity (categorized as Han, Uyghur, or other ethnic minorities), marital status, residence area, health insurance, and education level; and² lifestyle factors, including smoking status, drinking frequency, exercise frequency, and eating habits. Sex was ascertained from national identification records and refers to biological sex. Other sociodemographic data were extracted from health profiles, whereas lifestyle factors were self-reported via standardized questionnaires during health examinations.¹⁷ Eating habits were assessed by a single self-reported question: “Which of the following dietary patterns best describes your usual diet over the past year?” with response options of balanced omnivorous, animal-based, or plant-based diet.

Statistical analysis

Participants were divided into 4 age groups: 18–<40, 40–<60, 60–<80, and ≥80 years. Baseline participant characteristics were summarized across the age groups, as means with standard deviations (SD) or numbers with percentages, as appropriate. Normality testing was not performed due to the large sample size, under which even trivial deviations from normality would be statistically significant. Different combinations of metabolic abnormalities were ranked according to their frequencies within each age group.

For the 1,649,758 participants with metabolic measurements over the 5 health examinations, we mapped time-series trajectories for each metabolic factor and for overall metabolic burden (0, 1, 2, and ≥3 metabolic comorbidities) from 2019 to 2023 across the age groups. This analysis aimed to explore the natural changes in metabolic states with advancing age in the same population, particularly considering sex-specific patterns. In order to better identify any crossover points between sexes, we estimated the 2019 prevalence data directly at single-age resolution, and smoothed the resulting curves using locally weighted scatterplot smoothing (LOWESS) with pointwise 95% confidence intervals (CIs).

Across the age groups, we used multivariable Cox proportional hazards models to calculate hazard ratios (HRs) and 95% CIs of CVD associated with each metabolic risk factor or the overall metabolic burden. Results were adjusted for age (continuous within strata), sex, ethnicity, marital status, residence area, health insurance, education level, smoking status, drinking frequency, exercise frequency, and eating habits. Follow-up time was counted as the time interval between inception date (date when attending the first health examination) and the date of first diagnosis of CVD, death, loss to follow-up (i.e., out-of-region migration) or the most recent follow-up (Dec 2023), whichever occurred first. The proportional hazards assumption was evaluated using Schoenfeld residual tests. Given the large sample size, we focused on absolute rho (residual–time correlation) rather than P-values, as trivial deviations would be significant.³³ Meanwhile, a series of landmark analyses were performed with time cutoffs at 2, 2.5, 3, and 3.5 years, with baseline reset and inclusion restricted to event-free participants at each time point.

Several sensitivity analyses were performed to test the robustness of the findings. First, we excluded incident cases diagnosed within 2 years since enrollment, a time window that reduces reverse causation while preserving power, particularly for young adults with a low event rate. Second, we excluded participants with a parental history of CVD to reduce the impact of genetic susceptibility. Third, to account for intercorrelations among the metabolic factors, we re-assessed the associations after excluding 1 of the other 5 metabolic comorbidities at each turn. Fourth, to account for the competing risk of death from causes other than CVD, we additionally performed cause-specific competing-risks regression using the Fine–Gray subdistribution hazard model,³⁴ treating deaths from non-CVD causes as competing events. Finally, to account for potential selection bias due to missing covariates, we performed multiple imputation by chained equations (MICE) and re-examined the main associations using the imputed datasets.

To assess the potential sex-specific difference in the examined associations, apart from formal sex-interaction tests, we further compared the risk estimates in women and those in men by calculating the ratio of HR (RHR) with the corresponding 95% CIs.³⁵ Given the potential for our large sample size to yield statistically significant yet trivial findings, we prespecified RHR <0.8 or >1.2 as an indicator for a clinically meaningful sex-specific difference.

We calculated the average population attributable fractions (PAFs) and 95% CIs of CVD attributable to each metabolic factor following a well-established approach by Eide and Gefeller,³⁶ implemented via the ‘averisk’ package in R.³⁷ The average PAFs for all 6 metabolic factors combined were estimated from a single multivariable-adjusted model that included all metabolic factors simultaneously and was adjusted for age (continuous within strata), sex (except stratification by sex), ethnicity, marital status, residence area, health insurance, education level, smoking status, drinking frequency, exercise frequency, and eating habits. Each risk factor was permuted through every possible position in the model hierarchy, and the final average PAFs were calculated by aggregating estimates across all permutations. By averaging sequential PAFs over all possible removal orders, the method accounts for correlations and shared causal pathways (including potential mediation) without requiring a prespecified causal ordering, while ensuring that the sum of the individual average PAFs equals the joint PAF and does not exceed 100%. PAFs were also estimated in the MICE-imputed datasets to address missingness.

Finally, among metabolically abnormal participants (≥1 metabolic risk factor at baseline), we investigated the association between longitudinal transition in metabolic states and incident CVD. Within each age stratum, participants were categorized into 7 groups based on their metabolic states (defined by the number of coexisting metabolic disorders) during the first and the second health examinations: ≥3 risk factors controlled, 2 risk factors controlled, 1 risk factor controlled, no change (referent group), 1 risk factor increased, 2 risk factors increased, and ≥3 risk factors increased. For this analysis, baseline was set as the later health examination, and the multivariable model was adjusted for the covariates updated at the second health examination in addition to the number of metabolic risk factors at the first examination. We also repeated the transition analysis in the dataset with imputed covariates.

R software version 4.4.1 was used for statistical analyses and figure preparation. The significance threshold was set at a P < 0.05 (two-sided).

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Participant characteristics

Overall and age-stratified participant characteristics are shown in Table 1. Mean (SD) age of the 7,513,828 participants was 44.4 (15.3) years and 52.0% (3,907,602) were female. Ethnically, there were 32.8% Han, 52.9% Uyghur, and 14.3% other minorities. Compared with participants aged 18–<40 years, older participants were less likely to be female (48.9% vs. 52.1% in the oldest vs. youngest) or rural residents (39.2% vs. 58.1%), had lower educational attainment, were less likely to be current smokers (4.6% vs. 14.4%) or alcohol drinkers (2.8% vs. 11.4%), and were more likely to be regular exercisers (14.8% vs. 4.6% exercised almost daily). Older participants were also more likely to have metabolic risk factors including hypertension, hyperlipidemia, diabetes, and renal dysfunction, whereas obesity was most prevalent among those aged 40–<80 years.

Table 1.

Baseline characteristics of the study participants by age.

Characteristics	Overall n = 7,513,828	18–<40 y n = 3,149,523	40–<60 y n = 3,053,630	60–<80 y n = 1,201,914	≥80 y n = 108,761
Age, year	44.4 (15.3)	29.9 (5.6)	49.0 (5.3)	67.4 (5.3)	83.6 (3.5)
Female	3,907,602 (52.0)	1,641,505 (52.1)	1,593,363 (52.2)	619,574 (51.5)	53,160 (48.9)
Ethnicity
Han	2,464,463 (32.8)	667,054 (21.2)	1,185,353 (38.8)	548,341 (45.6)	63,715 (58.6)
Uyghur	3,972,889 (52.9)	2,019,138 (64.1)	1,415,719 (46.4)	505,003 (42.0)	33,029 (30.4)
Other ethnic minorities	1,076,476 (14.3)	463,331 (14.7)	452,558 (14.8)	148,570 (12.4)	12,017 (11.0)
Marital status
Unmarried	813,746 (10.8)	761,468 (24.2)	44,634 (1.5)	7047 (0.6)	597 (0.5)
Married	6,264,461 (83.4)	2,316,598 (73.6)	2,865,151 (93.8)	1,011,293 (84.1)	71,419 (65.7)
Divorced	154,054 (2.1)	61,912 (2.0)	74,638 (2.4)	16,605 (1.4)	899 (0.8)
Widowed	281,567 (3.7)	9545 (0.3)	69,207 (2.3)	166,969 (13.9)	35,846 (33.0)
Residence area
Urban	3,630,241 (48.3)	1,320,834 (41.9)	1,590,252 (52.1)	652,997 (54.3)	66,158 (60.8)
Rural	3,883,587 (51.7)	1,828,689 (58.1)	1,463,378 (47.9)	548,917 (45.7)	42,603 (39.2)
Health insurance
Underinsured	123,678 (1.6)	53,733 (1.7)	53,003 (1.7)	15,446 (1.3)	1496 (1.4)
Insured	7,390,150 (98.4)	3,095,790 (98.3)	3,000,627 (98.3)	1,186,468 (98.7)	107,265 (98.6)
Education level, year
<9	5,444,430 (72.5)	2,045,421 (64.9)	2,295,391 (75.2)	1,009,326 (84.0)	94,292 (86.7)
9–12	1,109,778 (14.8)	529,245 (16.8)	445,351 (14.6)	126,362 (10.5)	8820 (8.1)
>12	959,620 (12.8)	574,857 (18.3)	312,888 (10.2)	66,226 (5.5)	5649 (5.2)
Parental history of CVD	149,832 (2.0)	62,001 (2.0)	72,175 (2.4)	15,148 (1.3)	508 (0.5)
Smoking status
Never	6,405,483 (85.3)	2,673,945 (84.9)	2,555,888 (83.7)	1,073,309 (89.3)	102,341 (94.1)
Former	72,338 (1.0)	22,047 (0.7)	30,537 (1.0)	18,333 (1.5)	1421 (1.4)
Current	1,036,007 (13.8)	453,531 (14.4)	467,205 (15.3)	110,272 (9.2)	4999 (4.6)
Drinking frequency
Never	6,752,989 (89.9)	2,791,750 (88.6)	2,719,303 (89.1)	1,136,213 (94.5)	105,723 (97.2)
<1 time/week	686,548 (9.1)	329,232 (10.5)	299,796 (9.8)	55,190 (4.6)	2330 (2.1)
1–4 times/week	60,157 (0.8)	25,546 (0.8)	28,638 (0.9)	5735 (0.5)	238 (0.2)
Almost daily	14,134 (0.2)	2995 (0.1)	5893 (0.2)	4776 (0.4)	470 (0.4)
Exercise frequency
Never	6,572,275 (87.5)	2,876,247 (91.3)	2,666,841 (87.3)	943,751 (78.5)	85,436 (78.6)
<1 time/week	232,883 (3.1)	87,850 (2.8)	97,652 (3.2)	43,489 (3.6)	3892 (3.6)
1–4 times/week	132,804 (1.8)	39,237 (1.2)	57,677 (1.9)	32,561 (2.7)	3329 (3.1)
Almost daily	575,866 (7.7)	146,189 (4.6)	231,460 (7.6)	182,113 (15.2)	16,104 (14.8)
Eating habits
Balanced omnivorous diet	7,026,279 (93.5)	2,927,409 (92.9)	2,871,685 (94.0)	1,124,234 (93.5)	102,951 (94.7)
Animal-based diet	246,250 (3.3)	109,068 (3.5)	91,382 (3.0)	42,446 (3.5)	3354 (3.1)
Plant-based diet	241,299 (3.2)	113,046 (3.6)	90,563 (3.0)	35,234 (2.9)	2456 (2.3)
Metabolic risk factors
Obesity	3,399,000 (45.2)	1,075,232 (34.1)	1,617,208 (53.0)	658,335 (54.8)	48,225 (44.3)
Hepatic steatosis	2,657,872 (35.4)	928,438 (29.5)	1,288,013 (42.2)	417,679 (34.8)	23,742 (21.8)
Hypertension	1,042,583 (13.9)	99,211 (3.2)	506,674 (16.6)	398,358 (33.1)	38,340 (35.3)
Hyperlipidemia	1,078,215 (14.3)	311,370 (9.9)	534,161 (17.5)	217,380 (18.1)	15,304 (14.1)
Diabetes	436,981 (5.8)	49,403 (1.6)	214,121 (7.0)	159,917 (13.3)	13,540 (12.4)
Renal dysfunction	247,750 (3.3)	36,737 (1.2)	90,717 (3.0)	97,730 (8.1)	22,566 (20.7)

Data are shown as mean (standard deviation) for continuous variables and n (%) for categorical variables.

Pattern of metabolic comorbidities

The metabolic risk factors may exist alone or concurrently. Figure S2 ranks the top 15 metabolic comorbidities for each age group. Obesity combined with hepatic steatosis was the most frequent combination among participants younger than 60. Renal dysfunction ranked higher while hyperlipidemia ranked lower in the very old group (≥80 years old) as compared with the rankings in the younger groups.

Time-series trajectories in metabolic abnormalities

From 2019 to 2023, the prevalence of obesity rose gradually in the youngest group from 36.0% to 48.5%, stabilized in those who were aged 40–<80 years at baseline, and declined in the very old group (from 45.7% to 39.7%), similarly in women and men (Figure S3). Over the years, the youngest group consistently had the lowest prevalence of hypertension, diabetes, and renal dysfunction, whereas the middle-aged group (40–<60 years old in 2019) consistently had the highest prevalence of obesity and hepatic steatosis. The highest prevalence of diabetes was among those aged 60–<80 years at baseline, followed by the very old group who had a particularly high burden of renal dysfunction.

Over the 5 years, the proportion of metabolically normal individuals gradually declined from 50.9% to 38.8% in the youngest group, whereas the declines were more modest in older groups (Fig. 1). Notably, in the youngest group, women exhibited a more favorable metabolic profile than men, characterized by a higher proportion of metabolically normal individuals and a lower prevalence of multimorbidity (≥3 metabolic abnormalities). However, such a sex advantage reversed in the older groups, with women being less likely to be metabolically normal after the age of 40 and more likely to have multimorbidity with increasing age (Fig. 1).

Fig. 1.

Time-series trajectories of metabolic states across the adult lifespan. Data were derived from 5 waves of repeated measurements (2019–2023) (n = 1,649,758).

For women and men aged 35–55 years, the age range where the observed sex reversals likely occurred, we mapped the prevalence of each metabolic risk factor, multimorbidity, as well as the proportion of metabolically normal individuals at every age and smoothed these curves using LOWESS. Both obesity and hepatic steatosis were initially more frequent in men but showed a sex reversal around age 40, whereas hypertension and hyperlipidemia did so at a later age of ∼50 and ∼55, respectively (Fig. 2A). Meanwhile, the age when women became less likely to be metabolically normal was around 45, and starting from the age of ∼50, the prevalence of multimorbidity in women surpassed that in men (Fig. 2B). These two critical crossover points were confirmed by our more granular time-series analyses (Figure S4).

Fig. 2.

The midlife crossing over of metabolic burden by sex. (A) Prevalence of each metabolic factor among women and men aged 35–55 years. (B) Proportions of metabolically normal individuals (left panel) and prevalence of multimorbidity (≥3 conditions) (right panel) in 2019 among women and men aged 35–55 years (n = 870,854). Smoothing lines were fitted using locally weighted scatterplot smoothing (LOWESS) with a span of 0.9. Shaded areas represent pointwise 95% confidence intervals for the fitted LOESS trends.

Incidence of cardiovascular disease

During a median follow-up of 4.5 years, 669,995 incident cases of CVD were identified (352,138 in women and 317,857 in men), including 442,120 cases of CHD and 398,300 cases of stroke. As expected, the incidence of CVD increased dramatically with advancing age, ranging from 1.8 per 1000 person-years for young adults (<40 years old) to 107.5 per 1000 person-years for the very old (Figure S5). The incidence was markedly higher in men than women (2.3 vs. 1.5 per 1000 person-years) among those under 40 and became comparable in the group aged 40–<60 years (22.9 vs. 22.4 per 1000 person-years).

Metabolic abnormalities and incidence of cardiovascular disease

The metabolic risk factors showed substantial age differences in the association with incident CVD (Fig. 3A and Table S3). Formal tests for interaction with age (treated as a continuous variable) reached nominal statistical significance for all metabolic factors (all P values for interaction <3.50E-15). For hypertension, hyperlipidemia, diabetes, and renal dysfunction, the associations were most pronounced in the group under 40 and progressively attenuated with increasing age. Specifically, the HR for hypertension declined from 3.60 in the youngest group to 1.21 in the oldest old (≥80 years), with similar attenuation patterns observed for diabetes (2.48–1.31), hyperlipidemia (1.65–1.05), and renal dysfunction (1.69–1.12). For both obesity and hepatic steatosis, the associations with incident CVD were comparable between the youngest and middle-aged groups, with moderate declines after age 60. Moreover, the heterogeneity in the risk estimates across different metabolic factors was evident in those under 60 but narrowed markedly in the older groups.

Fig. 3.

Associations between metabolic risk factors and incident cardiovascular disease across the adult lifespan. (A) Hazard ratio (HR) for cardiovascular disease associated with metabolic risk factors by age. (B) Ratio of HRs (RHR; women vs. men) for cardiovascular disease associated with metabolic risk factors by age. Results were adjusted for age (continuous within strata), sex (except stratification by sex), ethnicity, marital status, residence area, health insurance, education level, smoking status, drinking frequency, exercise frequency, and eating habits. An RHR <0.8 or >1.2 was considered indicative of a clinically meaningful sex difference in the strength of association.

In the youngest group, hypertension demonstrated the strongest association with risk of CVD (HR = 3.60; 95% CI: 3.47, 3.73), whereas in older groups diabetes showed more evident associations than other metabolic risk factors (Fig. 3A and Table S3). Such association patterns were largely consistent for CHD and stroke, although in the youngest group hypertension showed an exceptionally potent association with stroke (HR = 4.84; 95% CI: 4.58, 5.11). When the metabolic risk factors were jointly considered, an increasing number of abnormalities was associated with a progressively higher risk of CVD (Table S4). HRs of CVD comparing ≥3 with 0 risk factors were 3.16 (95% CI: 3.04, 3.28), 2.88 (95% CI: 2.85, 2.92), 1.94 (95% CI: 1.92, 1.96), and 1.50 (95% CI: 1.45, 1.55) across the youngest to the oldest groups, respectively. The maximum |rho| from Schoenfeld tests was 0.043 and the vast majority were <0.03, well below the 0.05 threshold (Figure S6), indicating that any time-varying effect was clinically negligible. In the landmark analyses, the age-specific HR estimates were highly consistent with those from the main analysis across all time cutoffs (Table S5).

Moreover, these findings were consistent after excluding cases diagnosed within the first 2 years (Table S6) or excluding participants with a parental history of CVD (Table S7). In the analyses excluding participants with another specific metabolic risk factor, alterations in the risk estimates were largely modest (Table S8). There were exceptions for hepatic steatosis and obesity, for which the associations with incident CVD were moderately attenuated after excluding one for another. Meanwhile, excluding individuals with hypertension moderately attenuated the risk estimates for renal dysfunction. After accounting for the competing risk of death from non-CVD causes, the subdistribution HRs for each metabolic factor were virtually identical to the corresponding Cox regression estimates, regardless of age groups (Table S9). The results from the imputed data were consistent with those from the complete-case analysis (Table S10), further supporting the robustness of our primary findings.

Sex-specific associations

For hypertension and renal dysfunction, there were notable sex differences in the associations with incident CVD, particularly in the youngest group, per our predefined RHR threshold (i.e., RHR <0.8 or >1.2) (Fig. 3B and Tables S11 and S12). In the youngest group, the association between hypertension and risk of CVD was stronger in women than men (Table S11), with an RHR of 1.22 (95% CI: 1.15, 1.31). This sex disparity was especially evident for stroke, for which the HR was 6.41 in young women compared to 4.22 in young men (RHR = 1.52; 95% CI: 1.38, 1.67). Conversely, among the 2 groups under 60, the association of renal dysfunction with CVD was more pronounced in men than women (Table S11), with RHRs of 0.59 (95% CI: 0.51, 0.68) in the youngest group and 0.72 (95% CI: 0.69, 0.74) in the middle-aged group. For most metabolic factors, P-values for interaction with sex were commonly significant in the groups under 80, but were largely non-significant in the oldest group (Table S12).

Population attributable fractions

As shown in Fig. 4A and Table S13, the PAF of CVD for combined metabolic burden exhibited an inverted U-shape pattern across the adult lifespan, highest in the group aged 40–<60 years (37.89%), followed by people aged 18–<40 years (29.78%) and 60–<80 years (25.24%), and lowest in the very old group (15.85%). Regardless of age, obesity was the largest contributor to total CVD, followed by hypertension and hepatic steatosis. The PAF for hyperlipidemia demonstrated a steep, age-dependent decline, from 3.24% in the youngest group to a negligible level in the very old (0.42%). Conversely, the PAF for renal dysfunction, while small in magnitude, increased with advancing age (0.61%–1.62%). The top 3 contributors held true for CHD (Table S13), for which obesity demonstrated a higher PAF than that observed for total CVD, particularly in the middle-aged group (40–<60 years) (PAF = 13.91%). For stroke, however, hypertension was the predominant contributor across the age groups (Table S13). The PAFs estimated from the covariate-imputed dataset were consistent with the complete-case estimates (Table S14).

Fig. 4.

Population attributable fractions (PAFs) of metabolic risk factors for total CVD across the adult lifespan in the overall population (A) and by sex (B). Results were estimated from a single multivariable-adjusted model that included all six metabolic factors simultaneously and was adjusted for age (continuous within strata), sex (except stratification by sex), ethnicity, marital status, residence area, health insurance, education level, smoking status, drinking frequency, exercise frequency, and eating habits. Each risk factor was permuted through every possible position in the model hierarchy, and the final average PAF was calculated by aggregating estimates across all permutations. Risk factors were ranked according to their average PAF within each age group.

Across different age groups, the PAFs showed different sex-specific patterns (Fig. 4B and Table S15). In the group under 40, PAF of CVD for combined metabolic burden was higher in men than women (33.99% vs. 22.41%), especially so for CHD (39.90% vs. 23.49%). With increasing age, the PAFs became comparable between sexes and clearly reversed in the oldest group (13.45% in men vs. 18.20% in women).

Transition of metabolic states and incidence of cardiovascular disease

Among the 3,712,053 participants who were metabolically abnormal at baseline, we examined the relationship of transition in metabolic states, over an average of 1 year, with subsequent risk of CVD. The distribution of time intervals between the examinations is shown in Figure S7, with a median of 365 (interquartile range: 318, 420) days. Among these participants, 1,480,409 (39.88%) maintained a stable metabolic state, 1,417,325 (38.18%) had a decrease in the number of metabolic risk factors, whereas 814,319 (21.94%) experienced an increasing number. Over a median follow-up of 3.5 years, compared to the stable state, the risk of CVD increased with an increasing number of metabolic abnormalities, and decreased with a reducing number, with a clear dose–response relationship (Fig. 5). Regardless of sex, these associations were most potent in young adults and attenuated remarkably with advancing age. For example, achieving control of ≥3 risk factors was associated with a 55% lower risk of CVD in the youngest group, compared to only a 17% risk reduction in the very old group. These results were similar in the analysis considering missing covariates (Figure S8).

Fig. 5.

Transition of metabolic states and total CVD among participants with one or more metabolic risk factors at baselinein the overall population (A), women (B) and men (C). The median interval between the two health examinations was 365 (interquartile range: 318, 420) days. Results were adjusted for age (continuous within strata), sex (except stratification by sex), ethnicity, residence area, health insurance, education level, smoking status, drinking frequency, exercise frequency, and eating habits at the second health examination, in addition to the number of metabolic risk factors at the first examination.

Discussion

In this prospective cohort study including >7.5 million Chinese covering the whole adult lifespan, we depicted the landscape of 6 major metabolic risk factors and quantified their cardiovascular impacts across demographic segments (i.e., age and sex). There are several major findings. First, metabolic states exhibited obvious age-dependent sexual dimorphism, with an initial female advantage in young adults which reversed during early midlife. Second, the cardiovascular effects of metabolic abnormalities attenuated with advancing age, with notable sex-specific impacts observed for certain metabolic factors, especially among young adults. Third, the contribution of aggregate metabolic burden to CVD escalates through early adulthood, peaks in midlife, and diminishes in later life, with obesity being the top contributor across the whole adulthood. In parallel with the observed sexual dimorphism in metabolic burden, a reversal in PAF estimates was found between sexes. Finally, among metabolically abnormal participants, transiting to a healthier metabolic state within an average of 1 year was associated with a substantially reduced risk of CVD, whereas shifting to a poorer metabolic profile was associated with a higher risk, particularly in the young age groups.

Given the rising epidemic of metabolic abnormalities in young populations coupled with accelerated societal aging in China, our study offers a crucial and timely overview of the metabolic burden panorama. We found that the metabolic factors except renal dysfunction deviated from a linear trajectory with age, with a plateau or decline in the very old, a pattern indicative of survival bias.³⁸ Among adults under 40, we observed a higher proportion of metabolically normal individuals in women than men, a sex advantage that reversed at approximately 45 years old. Such a reversal could be a result of metabolic transitions specific to women of childbearing age, such as pregnancy-related hormone fluctuation and complications, and child raising-related lifestyle or psychosocial changes following changes in familial responsibilities and socioeconomic status.^39,40 Furthermore, we observed another reversal around age of 50, where women began to demonstrate a higher proportion of multimorbidity than men. This is likely a result of the perimenopausal transition, which often triggers a reshaping of body composition and various deleterious metabolic alterations in women.^40,41 The heightened metabolic burden faced by midlife women is likely rooted in the intersection of social, behavioral, physiological, and psychological aspects, constituting a multidimensional public health challenge.

We observed age- and sex-specific patterns of CVD incidence. The incidence was higher among men in the young group, comparable between sexes in middle age, higher among women at ages 60–80 years, and higher among men after age of 80 years. This somewhat parallels the above age- and sex-specific cardiovascular impacts of metabolic risk factors. For example, the potential female metabolic advantage noted above may have partially contributed to the lower rate of CVD in young women, whereas adverse transitions in metabolic and other risk factors across middle and older ages in women may eliminate and even reverse the sex difference in CVD incidence.

We found that the cardiovascular impact of metabolic abnormalities generally attenuated with advancing age, a pattern consistent with findings from prior studies of middle-aged and older participants.^23,24,42 The declining effect in older groups may also be partly explained by survival bias, as individuals who are less susceptible to metabolic dysfunction or CVD are more likely to survive to an advanced age. Additionally, it may reflect a transition in dominant risk from metabolic abnormalities per se to irreversible downstream sequelae, such as atherosclerotic plaques induced by chronic hyperlipidemia.⁴³ Leveraging the large population sample, our study included more than 3 million participants under 40, a stage where lifelong cardiovascular trajectories can be decisively shaped.44, 45, 46 For this significantly underrepresented population group, we found hypertension to be the strongest cardiovascular risk factor compared with other metabolic abnormalities. With advancing age, other metabolic factors (especially diabetes) became increasingly important and showed more comparable risk estimates with that for hypertension. This shift likely mirrors an underlying pathophysiological evolution from early hemodynamic dominance to a complex, multi-metabolic etiology in later life. Other previous studies have suggested a stronger association between hypertension and cardiovascular risk in women than in men,^47,48 whereas in the present study such a sex difference was primarily observed among young adults. Our findings provide critical, population-based evidence highlighting the vulnerability of younger women, refining the discourse advocating for age-specific and gender-informed hypertension management.⁴⁹

Unlike a monotonic decline reported in previous research conducted largely in middle-aged and older populations,^23,24 our study revealed a more nuanced, inverted U-shaped pattern for the contribution of total metabolic burden to CVD risk. Specifically, the population impact of metabolic risk factors escalates through early adulthood, peaks in midlife, and diminishes in later life, supporting the rationale of strategic reallocation of resources towards middle-aged and younger populations in resource-limited settings. Additionally, due to the age-dependent sexual dimorphism in their prevalence, metabolic risk factors exhibited a greater contribution to CVD in young men than young women, a sex-specific pattern that reversed in midlife.

As the most prominent metabolic issue, obesity is more prevalent in northwestern China than the majority of national areas, which may be partially attributed to the unique climate, ethnic composition, and dietary habits in the region.^50,51 Despite the moderate risk estimate, obesity emerged as the leading contributor to CVD burden throughout the adulthood, potentially explaining 6.07%–11.86% of incident cases of CVD across age groups. Additionally, we observed that the prevalence of obesity in women exceeded that for men at the age of 40, at least 10 years earlier compared to the age where hypertension and hyperlipidemia became more prevalent in women than men. This temporal precedence likely reflects the pathogenic cascade, implicating obesity as a key upstream etiological factor. Given its dual role as a principal upstream driver of metabolic comorbidities and its extensive population impact, a concerted public health strategy targeting obesity may achieve the most widespread reduction in cardiovascular burden.

Moving beyond static metabolic profiling, we presented the time-series trajectories of metabolic abnormalities with increasing age and revealed a particularly alarming decline in the metabolic health of young adults, coupled with a persistently high metabolic burden in older groups. In the absence of prompt and effective management, existing metabolic abnormalities can undergo rapid progression, further worsening the metabolic state and thereby triggering a vicious cycle.^52,53 As pointed out elsewhere,10, 11, 12 the accelerating metabolic deterioration observed in young adults may point to a greater shortfall in adequate treatment for this population. In our subsequent analysis of metabolically abnormal participants, CVD risk declined with a favorable transition in metabolic states, supporting the notion that the individual cardiovascular health trajectory is fundamentally modifiable. Notably, such benefits were strikingly age-dependent. Metabolic derangements in the young are typically nascent and largely reversible, with minimal cumulative end-organ damage, such that early intervention can yield substantial relative gains. By contrast, in the very old, metabolic control often represents long-term maintenance therapy rather than true reversion, and is frequently accompanied by multimorbidity and polypharmacy. Collectively, our finding reveal that the window of opportunity for maximal benefit is likely confined to early and middle adulthood, reinforcing the urgent need to address the treatment shortfall in younger populations.

Our study has several notable strengths. The unprecedented sample size enabled detailed age and sex stratification that addressed key evidence gaps at both ends of the adult lifespan. The repeated metabolic measurements, spanning up to 5 annual health examination waves for over 1.6 million participants, yielded high-resolution insights into the sex-specific metabolic trajectories across adulthood. The availability of longitudinal health examination data further provided a dynamic perspective to understand the cardiovascular impact of metabolic transitions across the demographic segments. Moreover, given the representativeness of the general people in northwest China, our findings provide valuable insights for public health planning in this and other similar settings.

Despite its merits, several limitations should be acknowledged. First, hepatic steatosis was defined by an established index rather than direct imaging or histological examination. Despite an indirect measure, the HSI was originally developed and validated in East Asians and has since been widely validated as a robust non-invasive screening tool for metabolic dysfunction-associated fatty liver disease.^27,54 Second, given the 5-year study period, which led to a scarcity of fatal CVD events among younger participants, we defined our primary outcome as a composite of fatal and non-fatal CVD events. Extended follow-up is warranted to separately evaluate fatal versus non-fatal events and to fully elucidate the long-term effects of metabolic abnormalities. Third, information on medication use was not systematically recorded, which could lead to misclassification of exposures and bias the associations toward the null, especially for the older groups. Nonetheless, the attenuation of the associations with advancing age was also observed for obesity and hepatic steatosis, which are less prone to confounding by medications, and is consistent with prior findings from randomized controlled trials^55,56 and cohort studies^57,58 with well-documented medication data. Regardless, our findings should be interpreted in the context of actual physiological burden carried by an individual beyond medical treatment. A further limitation is that several covariates were based on self-reported questionnaires, which may be subject to reporting bias. Finally, our primary analysis adopted a complete-case strategy and excluded individuals with missing covariates, which may introduce potential selection bias. Nonetheless, sensitivity analyses confirmed that it had limited impact on the results.

In summary, metabolic burden and the associated cardiovascular impacts are heterogeneous across demographic segments, supporting the rationale of implementing age-specific and gender-sensitive strategies for metabolic management and CVD prevention. Despite a persistently high metabolic burden in older adults, our findings highlight early and middle adulthood as a critical window for metabolic management.

Contributors

G-CC, Z-YL, and KG designed the research. Z-YL and KG had access to raw data. Z-YL developed the analytical plan, performed the statistical analysis, and had primary responsibility for writing the manuscript. KG directly accessed and verified the underlying data reported in the manuscript. G-CC, KG, and YY had the final responsibility for submission. All authors contributed to the interpretation of data, critically reviewed and revised the manuscript, and approved the final manuscript.

Data sharing statement

For permission requests to access the Tianshan Community Cohort data, please contact Yining Yang (yangyn5126@163.com) & Qian Di (qiandi@tsinghua.edu.cn).

Declaration of interests

The authors declare that they have no competing interests.