Associations of metals and micronutrients with functional disability in Chinese older adults: a mixture analysis approach

Abstract

Global aging underscores the public health challenge of disability in activities of daily living (ADL). Nutrition is a modifiable factor for maintaining physical function, yet evidence regarding the joint effects of metals and micronutrients on ADL disability is limited. To evaluate individual and combined associations of seven metals and two micronutrients with ADL disability a Chinese elderly population. This cross-sectional study enrolled 3974 adults aged ≥ 60 years. Blood concentrations of zinc (Zn), cobalt (Co), strontium (Sr), selenium (Se), molybdenum (Mo), manganese (Mn), vanadium (V), folate and vitamin D (VitD) were measured. ADL disability was assessed using the Barthel Index. We utilized logistic regression with restricted cubic splines (RCS) for single-exposure analyses, and weighted quantile sum (WQS) regression, quantile g-computation (QGC), and Bayesian kernel machine regression (BKMR) for mixture analyses. Inverse associations were observed between higher levels of folate (OR = 0.73; 95% CI: 0.63–0.85) and VitD (OR = 0.80; 95% CI: 0.68–0.93) with the risk of ADL disability, with linear dose-response relationships. After covariate adjustment, Mn exhibited a marginal protective effect (OR = 0.91, 95% CI: 0.82–1.00). Co and Sr exhibited U-shaped associations. Mixture analyses consistently indicated an overall protective effect, primarily driven by folate and VitD. Combined exposure to metals and micronutrients is associated with reduced ADL disability risk, with folate and VitD being the most influential components. These findings support holistic nutritional strategies for promoting functional health in aging populations.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-51318-z.

Introduction

Global population aging presents a critical public health challenge, with China undergoing one of the world’s most rapid demographic transitions^1,2. By 2050, adults aged 65 and above are projected to constitute 26.1% of China’s population, representing a total of approximately 365 million people². This shift in the population structure is posing a growing challenge from chronic diseases and age-related functional decline^3,4. As a critical marker of overall health in the elderly, disability in activities of daily living (ADL) embodies the cumulative impact of multiple chronic conditions on functional independence⁵. Data from the China Longitudinal Aging Social Survey (CLASS) reveal that the prevalence of mild-to-moderate and severe ADL disability among older adults is 10.54% and 2.00%, respectively⁶. Given its role in diminishing life quality, exacerbating health issues, and increasing vulnerability to additional diseases^7–9, identifying modifiable risk factors has become an urgent public health priority.

Emerging evidence indicates that adequate intake of metals and micronutrients may help maintain health across diverse populations^10–12. Specific metals and micronutrients, including cobalt (Co), manganese (Mn), molybdenum (Mo), selenium (Se), strontium (Sr), zinc (Zn), vanadium (V), folate, and vitamin D (VitD), play vital roles in biological processes fundamental to physical function, such as skeletal integrity, muscle metabolism, neural function, and antioxidant defense^13–19. For instance, Zn and Mn contribute to bone mineral density by serving as essential cofactors for alkaline phosphatase and other enzymes involved in bone matrix mineralization, while also participating in the antioxidant enzyme superoxide dismutase to protect osteoblasts from oxidative damage^11,16; Supplementation with Mg, Zn, and Se has been associated with improved muscle strength and reduced sarcopenia risk, likely through Se’s incorporation into selenoproteins regulating muscle redox balance, Zn’s activation of the mTOR pathway for muscle protein synthesis, and Mg’s role in ATP metabolism and muscle contraction^17,20; Folate acts as a core cofactor in one-carbon metabolism, regulating homocysteine clearance to prevent endothelial and skeletal muscle dysfunction and mediating epigenetic regulation of genes associated with neuromuscular regeneration^21,22; VitD exerts pleiotropic effects beyond calcium homeostasis, promoting muscle protein synthesis, downregulating pro-inflammatory cytokines (e.g., TNF-α, IL-6) to alleviate chronic low-grade inflammation, and exerting neuroprotective effects for neuromuscular integrity¹⁸. A longitudinal study of Chinese older adults further demonstrated that serum VitD level is a significant predictor for decreased incidence of ADL disability¹⁹.

Despite accumulating evidence on individual micronutrients, comprehensive research evaluating the joint effects of metals and micronutrients on ADL disability remains limited, and few studies have adopted advanced and complementary mixture analysis methodologies to explore their synergistic, antagonistic, and non-linear interactions—an analytical gap that fails to reflect the real-world complex exposure characteristics of older adults, making mixture-based approaches particularly essential for geriatric nutritional epidemiology research. This research gap is especially pronounced for the older adult population in China, as most existing relevant studies focus on Western cohorts, only assess single nutrients or harmful heavy metals, and lack targeted investigations into the combined effects of physiologically relevant metals and core micronutrients on ADL disability in Chinese older adults using multi-model complementary mixture analysis methods^23–25. The potential for suboptimal or imbalanced nutrient status, influenced by dietary patterns and regional soil geochemistry (e.g., for Se, Zn, and VitD), further underscores the need for targeted investigation in this population.

To address these evidence gaps, this study aims to comprehensively investigates the association between joint exposure to seven metals (Co, Mn, Mo, Sr, Se, V, Zn) and two core micronutrients (folate and VitD) and the risk of ADL disability in Chinese older adults. Advanced mixture modeling approaches—including weighted quantile sum (WQS) regression, quantile g-computation (QGC), and Bayesian kernel machine regression (BKMR)—were employed to quantify the joint effect of the nutrient mixture and to identify its most influential constituents. Our findings are expected to provide novel etiological insights and inform integrated nutritional strategies to promote functional health and support healthy aging in China.

Methods

Study population and design

This cross-sectional analysis employed baseline data from the “Older Adults Health and Modifiable Factors” cohort study. The data were collected from July to September 2018 among a community-dwelling older adult population in Fuyang City, Anhui Province, China. A comprehensive account of the cohort profile has been documented in other published works²⁶. This cohort, built by the School of Public Health of Anhui Medical University and the Fuyang Center for Disease Control and Prevention, recruited participants through probability proportional sampling, with enrollment limited to adults aged 60 years or older who had been prevalent residents in their communities for at least six months. Individuals with severe comorbidities, including major mental illnesses, severe neurological disorders (e.g., dementia, Parkinson’s disease), terminal-stage heart disease, or diagnosed malignant tumors, were excluded from the initial recruitment. From the total of 5186 enrolled participants, we excluded those due to missing data on the outcome of ADL disability (n = 159), missing measurements for metals (n = 587), or incomplete data on nutrient levels (folate or VitD, n = 466). The final analytical sample comprised 3974 participants. The Institutional Review Board of Anhui Medical University granted ethical approval (No. 20190288), and all participants provided written informed consent prior to the study.

Laboratory measurements of metals and micronutrients

Fasting venous blood samples were obtained from all participants using standardized protocols, as detailed in the Supplementary Material²⁷. Inductively coupled plasma mass spectrometry (ICP-MS) was employed to quantify the levels of seven metals (Zn, Se, Sr, Co, Mo, Mn, and V), with recovery rates of 78.90–136.54% and coefficients of variation below 10.73% (Table S4). Serum concentrations of folate and VitD levels were measured via automated chemiluminescence immunoassays. All biomarkers were detected in over 99.9% of samples (Table S2). For measurements lower than the limits of detection (LOD), they were imputed to be LOD//√2.

Assessment of ADL disability

ADL disability was evaluated using the Barthel Index (BI), a well-validated instrument for evaluating independence in ten essential activities of daily living, such as feeding, bathing, and dressing^28,29. Participants rated their independence for each activity, with response options ranging from complete independence to requiring substantial assistance. Consistent with established methodology⁵, a participant was classified as having ADL disability if they reported requiring substantial assistance or being completely dependent in at least one activity^5,19,29,30. The complete questionnaire is available in Supplementary Table S1.

Covariates

Covariates were identified based on their known or hypothesized roles as confounders in the relationship between the nutrient panel and ADL disability. They encompassed sociodemographic factors (age, gender, residence, educational attainment, household income, primary lifetime occupation, marital status, religious affiliation), lifestyle-related factors (smoking, physical activity, alcohol consumption, dietary patterns), and health status (physician-diagnosed hypertension, diabetes, COPD, stroke, coronary heart disease (CHD), and chronic kidney disease (CKD), body mass index). Five major patterns were identified (Table S3), Variable definitions are detailed in the Supplementary Material. Missing data were handled using multiple imputation to minimize bias.

Statistical analysis

Descriptive statistics were presented as follows: categorical variables were described by frequency (percentage), and continuous variables by mean ± standard deviation, respectively. Given their skewed distributions, the concentrations of all blood metals and serum micronutrients were log-transformed and standardized before analysis. Their distributions were summarized by geometric means together with the 5th, 25th, 50th, 75th, and 95th percentiles. Spearman’s correlation was used to analyze the bivariate relationships among all metals and micronutrients.

The associations of individual metals and micronutrients with ADL disability were examined using multivariable logistic regression. Three models were constructed: a crude model (Model 1), a model adjusted for sociodemographic, lifestyle, and health-status covariates (Model 2), and a fully adjusted model that additionally controlled for all other metals and micronutrients to isolate independent associations (Model 3). To evaluate potential non-linear relationships, we employed a two-step approach. Initially, each metal and micronutrient was stratified into quartiles, with the median value of each quartile assigned as a continuous variable in the logistic regression models. Subsequently, flexible dose-response relationships were modeled using restricted cubic splines (RCS) with three knots placed at the 10th, 50th, and 90th percentiles.

To assess the joint effect of the mixture, we applied three complementary methods, including WQS, QGC, and BKMR. For WQS, we assumed a unidirectional association and constructed an index from quartiled exposures, with weights derived from 1000 bootstrap samples. QGC was used to estimate the overall mixture effect while allowing individual components to have divergent directions of association (i.e., both protective and harmful). Exposures were quartiled, and the joint effect was defined as the change in the log-odds of ADL disability per one-quantile increase in all exposures. BKMR, a machine learning approach, was implemented to capture potential non-linearities and interactions among exposures without strong parametric assumptions. The model was run with 20,000 Markov chain Monte Carlo (MCMC) iterations, discarding the first 5000 as burn-in. Variable importance was assessed using posterior inclusion probabilities (PIPs).

Subgroup analyses were performed by age (categorized as ≤ 70 or > 70 years) and gender using both logistic and mixture models. Sensitivity analyses included: (1) excluding participants with major chronic conditions (COPD, stroke, CKD, CHD) to minimize confounding; and (2) excluding outliers with metals/micronutrients concentrations beyond three standard deviations from the mean.

All analyses were conducted using the R software (version 4.4.3). We employed the “rms”, “gWQS”, “qgcomp”, and “bkmr” packages for RCS, WQS, QGC, and BKMR analyses, respectively. Statistical significance was defined by a two-sided P-value < 0.05.

Results

Characteristics of the study population

Table 1 presents the baseline characteristics of 3974 study participants, stratified by ADL status. The cohort comprised 1964 males (49.42%) and 2010 females (50.58%), with a mean age of 70.59 ± 5.39 years. Significant differences (P < 0.05) were observed between the ADL-disability and non-disability groups across most sociodemographic, lifestyle, and health-status variables. Specifically, participants with ADL disability were older (mean age 72.46 vs. 70.42 years) and more likely to be female (10.25% vs. 6.92% in males). ADL disability was also more prevalent among those residing in rural areas, with lower educational attainment and income, and among those who were physically inactive, obese, or had comorbid conditions such as coronary heart disease or diabetes (all P < 0.05).

Table 1.

Basic characteristics of participants (n = 3974).

Basic characteristics		n (%)	Non-ADL disability	ADL-disability	t/χ2	P-value
			n = 3632	n = 342
Age (years)					-5.872	< 0.001
		70.59(5.39)	70.42(5.27)	72.46(6.23)
Gender					13.955	< 0.001
	Male	1964(49.42)	1828(93.08)	136(6.92)
	Female	2010(50.58)	1804(89.75)	206(10.25)
Residence					5.169	0.023
	Rural	3351(84.32)	3048(90.96)	303(9.04)
	Urban	623(15.68)	584(93.74)	39(6.26)
Education					35.014	< 0.001
	Illiteracy	2170(54.6)	1936(89.22)	234(10.78)
	Primary school	1059(26.65)	981(92.63)	78(7.37)
	Junior school or above	745(18.75)	715(95.97)	30(4.03)
Annual family income					16.446	< 0.001
	Poor	886(22.29)	780(88.04)	106(11.96)
	Average	3019(75.97)	2789(92.38)	230(7.62)
	Wealthy	69(1.74)	63(91.3)	6(8.7)
Occupation					16.330	0.001
	Technical labor related	137(3.45)	132(96.35)	5(3.65)
	Manual labor related	1825(45.92)	1671(91.56)	154(8.44)
	No regular job	1618(40.71)	1454(89.86)	164(10.14)
	Unemployment	394(9.91)	375(95.18)	19(4.82)
Smoking status					20.227	< 0.001
	Never	2736(68.85)	2466(90.13)	270(9.87)
	Current	781(19.65)	743(95.13)	38(4.87)
	Former	457(11.5)	423(92.56)	34(7.44)
Marital status					2.188	0.139
	Married	3640(91.6)	3334(91.59)	306(8.41)
	Unmarried	334(8.4)	298(89.22)	36(10.78)
Religion					2.216	0.137
	No	3407(85.73)	3123(91.66)	284(8.34)
	Yes	567(14.27)	509(89.77)	58(10.23)
Alcohol drinking					33.451	< 0.001
	Nondrinker	3044(76.6)	2739(89.98)	305(10.02)
	Light drinker	116(2.92)	113(97.41)	3(2.59)
	Moderate drinker	287(7.22)	274(95.47)	13(4.53)
	Excessive drinking	527(13.26)	506(96.02)	21(3.98)
Physical activity					43.854	< 0.001
	No physical activity	2424(61)	2159(89.07)	265(10.93)
	Moderate physical activity	1258(31.66)	1191(94.67)	67(5.33)
	High physical activity	292(7.35)	282(96.58)	10(3.42)
BMI					8.019	0.046
	Under weight	162(4.08)	147(90.74)	15(9.26)
	Normal weight	1694(42.63)	1562(92.21)	132(7.79)
	Overweight	1446(36.39)	1327(91.77)	119(8.23)
	Obesity	672(16.91)	596(88.69)	76(11.31)
Hypertension					7.254	0.007
	No	1536(38.65)	1427(92.9)	109(7.1)
	Yes	2438(61.35)	2205(90.44)	233(9.56)
Diabetes					9.759	0.002
	No	3497(88)	3214(91.91)	283(8.09)
	Yes	477(12)	418(87.63)	59(12.37)
COPD					2.156	0.142
	No	3629(91.32)	3324(91.6)	305(8.4)
	Yes	345(8.68)	308(89.28)	37(10.72)
Stroke					9.648	0.002
	No	3496(87.97)	3213(91.91)	283(8.09)
	Yes	478(12.03)	419(87.66)	59(12.34)
CHD					42.971	< 0.001
	No	3241(81.56)	3007(92.78)	234(7.22)
	Yes	733(18.44)	625(85.27)	108(14.73)
CKD					9.734	0.002
	No	3526(88.73)	3240(91.89)	286(8.11)
	Yes	448(11.27)	392(87.5)	56(12.5)
Five dietary patterns a		Median(Range )
	Dietary pattern1	-0.27(-0.45,0.12)	-0.26(-0.44,0.14)	-0.32(-0.48,-0.01)	1.544	0.123
	Dietary pattern2	-0.22(-0.67,0.38)	-0.21(-0.68,0.39)	-0.36(-0.76,0.21)	2.716	0.007
	Dietary pattern3	-0.22(-0.54,0.29)	-0.23(-0.55,0.29)	-0.23(-0.52,0.32)	-0.954	0.340
	Dietary pattern4	-0.05(-0.53,0.45)	-0.05(-0.54,0.44)	-0.01(-0.43,0.52)	-2.226	0.026
	Dietary pattern5	-0.05(-0.41,0.23)	-0.04(-0.47,0.23)	-0.07(-0.40,0.21)	-0.925	0.355

Abbreviations: ADL, activities of daily living; BMI, body mass index; IQR, interquartile range; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney disease; CHD, coronary heart disease.

^a Dietary habits were captured with a food-frequency questionnaire. Participants first indicated whether, within the past year, they had consumed pork, vegetables, fruit, mushrooms, eggs, milk, or whole grains; a “no” response was coded as zero intake. For any “yes” response, daily, weekly, monthly, or yearly frequencies were recorded. Exploratory factor analysis then condensed these data into five distinct dietary patterns.

A comparison between the included participants (n = 3974) and those excluded from the analysis (n = 1212) showed no significant differences for most variables (Table S5). Significant differences were confined to occupation, hypertension status, residence, and educational attainment.

Distributions of metals/micronutrients

Table S2 presents the detection rates (DR) and detailed concentration distributions of metals (Sr, Mn, Se, Zn, V, Mo, Co) and nutrients (folate and VitD). All metals had detection rates exceeding 99.9%, with Mn, Zn, and V detected in 100% of samples. The ranges between the 5th and 95th percentiles highlight substantial variability in concentrations for all analytes within the population. Fig.S1 presents a Spearman correlation heatmap visualizing pairwise associations between Metals and Micronutrients. The observed correlation coefficients spanned from − 0.09 to 0.44, indicating predominantly weak-to-moderate strength relationships. The strongest observed correlation was between Mo and V (r = 0.44).

Associations between individual metals/micronutrients and ADL disability

In multivariable logistic regression models (Table 2), higher serum levels of VitD and folate were significantly associated with lower odds of ADL disability. Per standard deviation increase in the natural log-transformed concentration, VitD was associated with a 20% reduction in odds (OR = 0.80, 95% CI: 0.68–0.93), while folate was associated with a 27% reduction (OR = 0.73, 95% CI: 0.63–0.85). A significant inverse linear trend was observed across quartiles for both nutrients (P_−trend < 0.05). Mn showed a marginally significant inverse association in the fully adjusted model (OR = 0.91, 95% CI: 0.82–1.00, p = 0.050). No significant association was observed for remaining metals. RCS analysis confirmed significant linear, inverse dose-response relationship for VitD and folate. U-shaped associations were identified for Co (P_−nonlinear = 0.014) and Sr (P_−nonlinear = 0.029), while a linear decreasing trend was observed for Mn (Fig. 1).

Table 2.

Associations between metals/nutrients and ADL-disability using logistic regression.

Metals/nutrients a		Model 1 (unadjusted) b		Model 2 (adjusted) c		Model 3 (adjusted) d
		OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value
Zn	Continuous	0.94(0.83,1.06)	0.297	1.00(0.88,1.14)	0.995	1.08(0.94,1.25)	0.286
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.88(0.65,1.19)	0.392	0.90(0.66,1.24)	0.517	0.92(0.66,1.28)	0.620
	Q3	0.81(0.60,1.10)	0.181	0.96(0.69,1.33)	0.791	1.06(0.75,1.49)	0.765
	Q4	0.76(0.55,1.04)	0.082	0.92(0.65,1.28)	0.609	1.03(0.70,1.52)	0.880
	P for trend	0.069		0.666		0.881
Se	Continuous	0.86(0.75,0.99)	0.033	0.94(0.81,1.08)	0.382	0.95(0.82,1.11)	0.552
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.79(0.58,1.07)	0.121	0.88(0.64,1.21)	0.429	0.90(0.65,1.24)	0.510
	Q3	0.80(0.59,1.08)	0.141	0.89(0.65,1.23)	0.493	0.92(0.66,1.27)	0.600
	Q4	0.71(0.52,0.98)	0.034	0.89(0.64,1.23)	0.473	0.94(0.66,1.33)	0.711
	P for trend	0.038		0.474		0.566
Sr	Continuous	1.05(0.90,1.22)	0.539	0.98(0.83,1.14)	0.754	0.96(0.82,1.13)	0.605
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.93(0.68,1.27)	0.626	0.89(0.64,1.23)	0.469	0.89(0.63,1.24)	0.483
	Q3	0.84(0.61,1.16)	0.285	0.80(0.57,1.12)	0.189	0.79(0.56,1.12)	0.190
	Q4	1.11(0.82,1.51)	0.487	0.96(0.70,1.33)	0.821	0.94(0.67,1.32)	0.708
	P for trend	0.618		0.732		0.607
Co	Continuous	1.01(0.90,1.14)	0.823	1.00(0.88,1.14)	0.954	1.04(0.90,1.20)	0.645
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.77(0.56,1.06)	0.108	0.73(0.52,1.01)	0.056	0.74(0.53,1.04)	0.083
	Q3	0.77(0.56,1.06)	0.108	0.69(0.50,0.96)	0.029	0.74(0.52,1.03)	0.077
	Q4	1.02(0.76,1.38)	0.879	0.95(0.70,1.30)	0.758	1.02(0.72,1.44)	0.924
	P for trend	0.644		0.987		0.676
Mo	Continuous	0.98(0.86,1.12)	0.743	0.97(0.84,1.11)	0.627	1.03(0.87,1.22)	0.726
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	1.00(0.73,1.37)	0.995	0.91(0.65,1.26)	0.554	0.94(0.68,1.31)	0.716
	Q3	1.12(0.82,1.52)	0.485	1.01(0.73,1.39)	0.958	1.14(0.81,1.59)	0.459
	Q4	0.91(0.66,1.26)	0.567	0.86(0.62,1.21)	0.394	0.98(0.66,1.44)	0.899
	P for trend	0.627		0.492		0.897
Mn	Continuous	0.86(0.79,0.94)	0.001	0.91(0.82,1.00)	0.050	0.90(0.81,1.00)	0.047
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.80(0.59,1.08)	0.143	0.88(0.64,1.21)	0.433	0.91(0.66,1.26)	0.583
	Q3	0.82(0.61,1.11)	0.192	0.89(0.65,1.22)	0.463	0.90(0.64,1.26)	0.534
	Q4	0.65(0.47,0.89)	0.007	0.80(0.57,1.11)	0.183	0.78(0.53,1.13)	0.188
	P for trend	0.010		0.195		0.239
V	Continuous	0.90(0.77,1.04)	0.153	0.92(0.79,1.08)	0.320	0.90(0.74,1.09)	0.282
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.86(0.63,1.17)	0.345	0.84(0.61,1.16)	0.287	0.82(0.59,1.15)	0.250
	Q3	0.84(0.62,1.14)	0.269	0.85(0.61,1.18)	0.325	0.80(0.56,1.13)	0.196
	Q4	0.82(0.60,1.12)	0.207	0.86(0.62,1.19)	0.360	0.83(0.56,1.21)	0.325
	P for trend	0.219		0.415		0.383
VitD	Continuous	0.78(0.68,0.90)	< 0.001	0.80(0.68,0.93)	0.004	0.80(0.68,0.94)	0.007
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.79(0.59,1.06)	0.118	0.78(0.57,1.06)	0.113	0.78(0.57,1.07)	0.128
	Q3	0.70(0.52,0.95)	0.023	0.69(0.49,0.95)	0.025	0.68(0.49,0.95)	0.024
	Q4	0.59(0.43,0.81)	0.001	0.60(0.42,0.88)	0.009	0.62(0.43,0.92)	0.016
	P for trend	0.001		0.006		0.008
Folate	Continuous	0.78(0.69,0.90)	< 0.001	0.73(0.63,0.85)	< 0.001	0.74(0.64,0.86)	< 0.001
	Q1	1.00(Reference)		1.00(Reference)		1.00(Reference)
	Q2	0.65(0.48,0.88)	0.005	0.60(0.44,0.83)	0.002	0.59(0.43,0.82)	0.002
	Q3	0.65(0.48,0.88)	0.006	0.58(0.41,0.80)	0.001	0.59(0.42,0.83)	0.002
	Q4	0.68(0.50,0.92)	0.012	0.60(0.43,0.83)	0.002	0.60(0.43,0.84)	0.003
	P for trend	0.005		0.001		0.001

Abbreviations: OR, odds ratio; CI, confidence interval; Sr, strontium; Mn, manganese; Se, selenium; Zn, zinc; V, vanadium; Mo, molybdenum; Co, cobalt; FA, folate acid; VitD, vitamin D, Q1, the first quartile; Q2, the second quartile; Q3, the third quartile; Q4, the fourth quartile.

Bolded values suggest statistically significant.

^a Blood metals/serum nutrients ln-transformed, and then standardized.

^b Unadjusted model.

^c Covariates in adjusted Model 2 included age, gender, education, economic status, occupation, dietary patterns (FAC1_1–FAC5_1), smoking status, marital status, residential region, religion, alcohol-drinking, physical activity, hypertension, diabetes, COPD, stroke, coronary heart disease, body mass index (BMI), and CKD.

^d Covariates in Model 3 included covariates in Model 2 and other metals/nutrients.

Fig. 1.

Associations between blood metals/serum nutrients and ADL-disability in restricted cubic spline (RCS) regression models. All metals/nutrients were ln-transformed and standardized. All models were adjusted for age, gender, education, economic status, occupation, dietary patterns (FAC1_1–FAC5_1), smoking status, marital status, residential region, religion, alcohol-drinking, physical activity, hypertension, diabetes, COPD, stroke, coronary heart disease, body mass index (BMI), and CKD. Abbreviations: Sr, strontium; Mn, manganese; Se, selenium; Zn, zinc; V, vanadium; Mo, molybdenum; Co, cobalt; VitD, vitamin D; CI, confidence interval; solid lines,OR; dashed lines, reference values; blue shades, 95% CI.

Joint associations of the metal/micronutrient mixture with ADL disability

The mixture analysis results consistently indicated a protective joint effect of the nine nutrients against ADL disability. The WQS regression indicated a significant negative association between the metal/micronutrient mixture and the odds of ADL disability (OR = 0.74, 95% CI: 0.63–0.85). folate (weight = 0.36) and VitD (0.35) were identified as the primary contributors to this protective effect, followed by Mn (0.13) (Fig. 2). QGC model also estimated a significant negative joint effect (OR = 0.89, 95% CI: 0.83–0.96). The weights derived from this model similarly highlighted folate (weight: -0.359) and VitD (-0.349) as the strongest protective components within the mixture (Fig. 3).

Fig. 2.

The negative weights of each blood metals/serum nutrients in weighted quantile sum (WQS) regression index for ADL-disability. All models were adjusted for age, gender, education, economic status, occupation, dietary patterns (FAC1_1–FAC5_1), smoking status, marital status, residential region, religion, alcohol-drinking, physical activity, hypertension, diabetes, COPD, stroke, coronary heart disease, body mass index (BMI), and CKD. Abbreviations: Sr, strontium; Mn, manganese; Se, selenium; Zn, zinc; V, vanadium; Mo, molybdenum; Co, cobalt; VitD, vitamin D.

Fig. 3.

Estimated weights of blood metals/serum nutrients on ADL-disability using quantile g-computation (QGC) models. ETEs/nutritions were ln-transformed and standardized. (A) The weight values for individual metals/nutritions in the positive and negative directions and the effect value (95% CI) of joint exposure. (B) Joint effect (95% CIs) plot of the metals/nutritions on ADL-disability. All models were adjusted for age, gender, education, economic status, occupation, dietary patterns (FAC1_1–FAC5_1), smoking status, marital status, residential region, religion, alcohol-drinking, physical activity, hypertension, diabetes, COPD, stroke, coronary heart disease, body mass index (BMI), and CKD. Abbreviations: Sr, strontium; Mn, manganese; Se, selenium; Zn, zinc; V, vanadium; Mo, molybdenum; Co, cobalt; VitD, vitamin D.

The BKMR model provided further visual insights for the association between the metal/micronutrient mixture and ADL disability (Fig. 4). A non-linear exposure-response relationship was observed for the overall mixture, marked by a significant decrease in the risk of ADL disability up to the 70th exposure percentile, beyond which the association plateaued (Fig. 4A). No substantial interaction effect was detected between mixture components (Fig. 4B). The univariate exposure-response curves, conditional on other exposures being fixed at their medians, confirmed the independent, protective associations of folate and VitD (Fig. 4C). Variable importance analysis confirmed folate and Vit D as the most influential factors within the mixture (Fig. 4D). The PIPs derived from the BKMR model identified folate and VitD as the most influential factors within the mixture (Table S6), consistent with the findings from the WQS and QGC models.

Fig. 4.

Associations between overall blood metals/serum nutrients with ADL-disability based on Bayesian kernel machine regression. (A)Overall effect of the mixture (estimates and 95% CI), defined as the differences in ADL-disability when all the metals/nutritions were fixed at a specific quantile (ranging from the 10th to 90th), as compared to when all of the metals/nutritions were fixed at their median values (the 50th). (B) Bivariate exposure-response functions for each of the metals/nutritions present on the right longitudinal axis when the other metals/nutritions presented on the upper coordinate axis holding at different quantiles and the other metals/nutritions were held at the median. (C) Univariate exposure response functions and 95% confidence intervals (CIs) between metals/nutritions and ADL-disability, while the remaining metals were fixed at their median values. (D) Single-exposure effects (95% CIs), defined as the changes in ADL-disability associated with a change in a particular metal from its 25th to its 75th percentile, where all of the remaining metals/nutritions were fixed at a specific quantile (the 25th, 50th, or 75th). All models were adjusted for age, gender, education, economic status, occupation, dietary patterns (FAC1_1–FAC5_1), smoking status, marital status, residential region, religion, alcohol-drinking, physical activity, hypertension, diabetes, COPD, stroke, coronary heart disease, body mass index (BMI), and CKD. Abbreviations: Sr, strontium; Mn, manganese; Se, selenium; Zn, zinc; V, vanadium; Mo, molybdenum; Co, cobalt; VitD, vitamin D.

Subgroup analyses

The results of the stratified analyses by age and gender are detailed in Fig.S2 and Table S7. Specifically, stronger protective effects were observed in the younger age group (60–70 years), although the interaction did not reach statistical significance. No significant effect modification by gender was observed.

Sensitivity analyses

Sensitivity analysis affirmed the robustness of the primary findings. After excluding participants with major chronic diseases (n = 1123 excluded) and further excluding outliers in exposure concentrations (n = 397 excluded), the inverse associations for folate, VitD, and Mn, as well as the overall protective effect of the mixture, remained significant, albeit with slightly attenuated effect sizes (Fig. S3-S6, Tables S8-S9). In all mixture models, folate and VitD consistently accounted for the largest weights.

Discussion

To our knowledge, this is the first study to assess the association of combined exposure to seven metals s and two core nutrients (folate and VitD) with ADL disability in a population of Chinese older adults. Our principal findings are threefold. First, in single-exposure models, higher serum concentrations of VitD and folate were significantly associated with a lower risk of ADL disability, showing linear dose-response relationships. Mn also demonstrated a modest protective association. Second, the three mixture models (WQS, QGC, BKMR) converged to indicate a significant protective joint effect of the nine-micronutrient mixture against ADL disability. Third, variable importance analyses (weight assignments in WQS and QGC, and PIPs in BKMR) unanimously identified folate and VitD as the primary contributors to this overall protective effect. These findings offer novel evidence on the role of combined micronutrient exposure in functional aging and highlight the potential of integrated nutritional interventions for promoting healthy aging.

The most salient finding of our study is the predominant role of folate and VitD in preserving functional independence. This observation is underpinned by several plausible biological pathways. folate serves as an essential cofactor in one-carbon metabolism, essential for nucleotide synthesis, DNA repair, and epigenetic regulation^22,31. Its protective association with ADL disability may be mediated through multiple pathways, including homocysteine metabolism, where it helps prevent hyperhomocysteinemia, a condition linked to endothelial dysfunction and sarcopenia³², and epigenetic regulation of genes involved in neuromuscular integrity and cognitive function³³. Additional evidence indicates that folate enhances peripheral nerve regeneration and muscle strength recovery^33,34. While previous studies have primarily focused on folate ‘s role in cognitive health, few have examined its relationship with physical function. For example, an NHANES study (2011–2018) found that adults in the lowest serum folate quartile faced a 44% elevated risk of sarcopenia compared to their counterparts in higher quartiles³⁵, and a study focusing on Chinese older adults demonstrated an inverse relationship between serum folate and ADL disability³⁶. Our findings extend this evidence by demonstrating a significant, independent, and linear inverse association, further reinforced in mixture models where folate emerged as a major protective component. Future longitudinal and mechanistic studies are warranted to elucidate the causal pathways through which folate may protect against functional decline.

VitD is recognized for its pleiotropic effects beyond calcium homeostasis, including roles in muscle protein synthesis, anti-inflammatory processes, and neuroprotection^37,38. The inverse association observed in our study may be explained by several mechanisms: VitD aids in maintaining muscle strength and balance, thereby lowering the risk of falls^39,40; it modulates inflammatory responses by downregulating cytokines like TNF-α and IL-6, which are implicated in muscle wasting^37,41; and it supports neuronal health and cognitive function^42,43. These mechanisms are consistent with prior observational findings. For instance, A study utilizing the Chinese Longitudinal Healthy Longevity Survey revealed that VitD deficiency was associated with a higher likelihood of ADL disability (OR: 4.08; 95% CI: 2.81–5.92)¹⁹. Our results strengthen these findings by consistently identifying VitD, along with folate, as a key protective factor within the nutrient mixture. Subsequent research should prioritize prospective studies to establish temporality and explore underlying mechanisms.

Mn demonstrated a consistent but non-significant inverse association with ADL disability. This trend may be attributed to Mn’s role as an essential cofactor in key enzymatic reactions involved in neurotransmitter synthesis, energy metabolism, and antioxidant defense^44,45. Previous studies have suggested that adequate Mn status supports skeletal health and mitigates oxidative stress^46,47, yet its benefits may be modulated by other nutritional factors or observable only within a specific physiological range. Alternatively, the relatively narrow variability of Mn concentrations in this population may have limited statistical power. Further investigations with expanded sample sizes are warranted to elucidate Mn’s role in functional aging.

Co and Sr exhibited U-shaped associations with ADL disability, indicating that both deficiency and excess may impair functional health. At lower concentrations, Co supports vitamin B12 synthesis and mitochondrial energy production⁴⁸, while Sr promotes bone density and reduces oxidative stress^49,50. However, beyond certain thresholds, excess Co may induce endoplasmic reticulum stress and disrupt redox homeostasis^51,52, and elevated Sr may dysregulate cellular signaling by substituting for calcium in proteins and receptors^53,54. This biphasic response underscores the need for precision in nutritional management, with both elements being beneficial within narrow ranges but becoming detrimental at higher doses. In contrast, Zn, Se, Mo, and V showed no consistent associations, suggesting their roles in functional decline may be negligible or context-dependent in this cohort.

In real-world settings, individuals are exposed to complex mixtures of elements that may interact synergistically or antagonistically. However, research on the joint effects of metals/ micronutrient mixtures on ADL disability remains limited. Most prior studies have focused on single elements or harmful heavy metals⁵, overlooking the complexity of real-world exposures. By applying advanced mixture models, we established that the metals/ micronutrient mixture confers significant protection against ADL disability, with folate and VitD being the principal drivers. These results emphasize the critical need to transition from single-element approaches to incorporate mixture-based methods for a more comprehensive understanding of nutritional influences on functional aging. Future longitudinal work is required to both confirm these relationships and uncover the biological mechanisms governing these nutrient synergies.

This study has several notable strengths. First, we employed multiple advanced statistical models for mixture analysis (WQS, QGC, and BKMR), which enhanced the robustness and interpretability of the findings regarding the joint effects of multi-element exposures. Second, the study was conducted within a large community-based sample of older adults, providing sufficient statistical power to detect associations. Third, we simultaneously measured a range of metals and key nutrients, allowing a comprehensive evaluation of their combined effects. Finally, this study focused on older adults, an understudied population, which adds valuable evidence to the existing literature on nutritional epidemiology and healthy aging. However, several limitations also merit acknowledgment. First, the cross-sectional design of this study precludes definitive causal inference regarding the associations between the metal/ micronutrient mixture and ADL disability, while also raising the possibility of reverse causation—for instance, disability may lead to reduced sun exposure or poorer dietary habits, thereby lowering nutrient levels. Second, although we adjusted for a broad range of covariates, residual confounding may still exist due to unmeasured factors including objective sun exposure indicators, supplement use, and vitamin B12 and homocysteine levels, among others. Third, blood/serum concentrations constitute only a snapshot of recent exposure and may not accurately represent long-term nutritional status or body burden, in addition, trace element measurements have a relatively wide recovery rate range (78.90–136.54%), which may introduce minor analytical bias despite acceptable quality control performance. Fourth, the definition of ADL disability based on requiring assistance in at least one activity, while consistent with previous studies in community-dwelling older adults, may yield slightly different prevalence estimates compared to alternative Barthel Index cut-offs. Therefore, future prospective studies incorporating repeated biomarker measurements across diverse populations are warranted to validate and generalize these results.

Conclusions

In summary, this study provided novel evidence that combined exposure to a mixture of metals and micronutrients was significantly associated with a reduced risk of ADL disability among Chinese older adults. Serum folate and VitD were consistently identified as the most influential protective components, demonstrating independent, dose-response relationships. While Mn suggested a beneficial trend, Co and Sr exhibited U-shaped associations, highlighting the importance of maintaining optimal concentrations. Overall, these findings underscore the value of a mixture-based approach in nutritional epidemiology and suggest that integrated strategies ensuring adequate folate and VitD status may help preserve functional independence in aging populations. Future prospective and mechanistic studies are warranted to confirm these associations, establish causality, and explore the potential of targeted nutritional interventions for promoting healthy aging.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

Xuqiu Cheng, Hanxiao Yang, Fusheng Lin, Xianglong Liu: Formal analysis, Methodology, Validation, Writing-original draft. Wenyuan Liu, Lei Yu, Ziwei Tian, Yuantao Zhang: Data curation, Investigation, Validation. Bing Hu, Changliu Liang: Data curation, Investigation. Fangbiao Tao: Investigation, Methodology, Validation. Linsheng Yang: Investigation, Data curation, Methodology, Validation, Writing-review & editing. Jun Wang: Investigation, Data curation, Methodology, Validation, Writing-review & editing.

Funding

This study was supported by the Key Scientific Research Fund of Anhui Provincial Education Department [grant number 2023AH050610], Research Funds of Center for Big Data and Population Health of IHM [grant number JKS2022015], and National Natural Science Foundation of China [grant number 81402698].

Data availability

The datasets generated and/or analysed during the current study are not publicly available due to the personal and sensitive nature of the research data collected for this project. However, they are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The study was approved by the biomedical ethics committee of Anhui Medical University (No. 20190288).

Consent to participate

Each participant had provided written informed consent according to the principles of Helsinki Declaration.

Confirm statement

All methods in this study were carried out in accordance with relevant guidelines and regulations.