这是indexloc提供的服务，不要输入任何密码
Skip to main content
Download PDF
Download ePub

Causal association of modifiable factors with cardiometabolic multimorbidity: an exposome-wide Mendelian randomization investigation

Cardiovascular Diabetology volume 24, Article number: 241 (2025) Cite this article

Abstract

Background

Cardiometabolic multimorbidity (CMM), characterized by the co-existence of two or more cardiometabolic diseases (CMDs) including type 2 diabetes (T2D), coronary artery disease (CAD), and stroke, persists as a global health challenge. However, the causal associations of modifiable factors with CMDs and CMM remains to be systematically investigated.

Methods

In this study, a three-stage design Mendelian randomization (MR) investigation was conducted, using two-sample MR with potential sample overlap correction and multiple testing, multivariable MR analysis, and multi-response MR, with modifiable factors covering domains of socioeconomic factors, behavioral factors, biochemical factors, and physical measures as exposures, and CMM and CMDs as outcomes. Updated large-scale genome-wide association study (GWAS) data based on systematic collection from GWAS Catalog were applied.

Results

Our major findings suggested that, 13 of 23 modifiable factors across four domains, including educational attainment (odds ratio: 0.858, 95% confidence interval: 0.834–0.883), household income (0.794, 0.720–0.875), lifetime smoking behavior (1.201, 1.145–1.260), leisure screen time (1.255, 1.186–1.327), low-density lipoprotein cholesterol levels (1.062, 1.046–1.079), total cholesterol levels (1.045, 1.029–1.062), Apolipoprotein B (1.035, 1.019–1.051), fasting glucose (1.096, 1.038–1.157), glycated hemoglobin (HbA1c) (1.062, 1.037–1.089), systolic blood pressure (1.125, 1.104–1.146), diastolic blood pressure (1.104, 1.083–1.126), forced expiratory volume in 1 s (FEV1) (0.953, 0.931–0.975), and body mass index (BMI) (1.243, 1.210–1.277) were evaluated with relatively robust effects on CMM. Furthermore, household income, lifetime smoking behavior, HbA1c, systolic blood pressure, and FEV1 with CMM were detected as independent associations within a single domain. Similar results were observed in each CMD. Moreover, the multi-response MR provided reinforcing evidence for the associations of educational attainment, serum urate, and BMI with CMM, and lifetime smoking behavior, moderate to vigorous intensity physical activity, and leisure screen time with diverse CMDs.

Conclusions

Promoting educational attainment, maintaining favorable serum urate, and controlling obesity are specifically prioritized for CMM prevention. Furthermore, avoiding smoking and sedentary behavior, and strengthening physical activity held prominent protective impacts on CMDs. Additionally, improving dyslipidemia and dysglycemia, maintaining favorable blood pressure, and enhancing lung function, would contribute to the co-management of CMDs and preventing the long-term CMM condition. Our investigation provided causality-oriented evidence to establish the risk profile of CMM.

Research insights

What is currently known about this topic?

  • Cardiometabolic multimorbidity mainly include type 2 diabetes, coronary artery disease, and stroke. The multimorbidity is common in individuals with cardiometabolic diseases. Targeted management on modifiable factors contributes to alleviating the m ultimorbidity burden.

What is the key research question?

  • Which modifiable factors are causally associated with cardiometabolic diseases and multimorbidity.

What is new?

  • This study focused on multiple modifiable factors within four major domains. Three-stage design Mendelian randomization investigation strengthens causal inferences. Six major factors are prioritized for the prevention of cardiometabolic diseases and multimorbidity.

How might this study influence clinical practice?

  • Targeted strategy on specific causal evidence facilitates cardiometabolic multimorbidity prevention.

Background

In the context of population aging, the co-occurrence of at least two chronic diseases in individuals, defined as multimorbidity, has rapidly prevailed as a global health challenge [1]. In particular, cardiometabolic multimorbidity (CMM), characterized by the co-existence of two or more cardiometabolic diseases (CMDs), typically including type 2 diabetes (T2D), coronary artery disease (CAD), and stroke, has gradually emerged as a major concern [1,2,3]. Evidence suggested that an estimated 30% of older adults are influenced by CMDs [4], and that at the age of 60 years, individuals with any two of CMDs lead a 12-year reduction in life expectancy, while individuals with all three resulted in a 15-year life expectancy reduction [1], which would be a tremendous impediment to achieving healthy aging.

Despite advances in understanding the pathophysiology and the increasing availability of multiple intervention strategies for CMDs, the clinical management of CMM remains suboptimal. Thus, early intervention targeting modifiable factors is a promising prevention strategy for CMM that could provide valuable insights into the co-prevention and co-management of CMDs. Several previous observational studies have indicated that social inequality, smoking, physical inactivity, cholesterol, triglycerides, abnormal blood pressure, lung function, and obesity, are potentially associated with CMM [3, 5,6,7,8,9,10,11]. However, due to the potential influence of confounders and reverse causation, the findings are elusive to be interpreted as robust causal associations, and the association of more additional modifiable factors with CMM remains ambiguous, which needs to be supported by further investigations.

Mendelian randomization (MR) is a robust study framework located at a relatively higher hierarchy of the scientific evidence pyramid. It typically encompasses two-sample MR (TSMR) and multivariable MR (MVMR) analysis methods, and uses genetic variants randomly allocated at conception as proxies for exposures from genome-wide association study (GWAS) data. This strengthens causal inferences on exposure-outcome associations, and helps overcome certain limitations of observational studies, such as susceptibility to confounders and reverse causation [12,13,14]. However, existing typical MR models consider only one outcome in isolation, therefore, applying multi-response MR, a newly developed MR method designed specifically for multiple outcomes, to identify exposures that cause multiple responses or, conversely, exposures that exert their effects on distinct responses, contributes to providing enhanced evidence for causal inference in CMM, based on typical MR framework [15]. Thus, systematically disentangling the linkages underlying extensive modifiable factors with CMM with MR is desirable for establishing accurate risk evidence and the primary prevention strategies.

Therefore, in this three-stage MR investigation, we adopted a framework of modifiable factors covering domains of socioeconomic factors, behavioral factors, biochemical factors, and physical measures as exposures, and CMM and CMDs including T2D, CAD, and stroke as outcomes, with the aims of (1) systematically investigating the causal associations of modifiable factors with the outcomes using TSMR, followed by potential sample overlap correction and multiple testing, (2) extensively identifying the independent causal effects of modifiable factors within an individual domain on the outcomes using MVMR, and (3) comprehensively detecting shared and distinct causal associations of modifiable factors with CMD outcomes in a joint model using multi-response MR that strengthen causal inferences, to provide insights for establishing risk evidence for CMM.

Methods

Systematic collection of summary statistics from genome-wide association study

Summary statistics were systematically collected from the GWAS Catalog with the criteria that (1) GWAS data with largest sample size, (2) GWAS data were updated after January 1, 2018, (3) GWAS participants were of European ancestry, and (4) GWAS data with unadjusted for heritable covariates, as single nucleotide polymorphisms (SNPs) remain statistically independent from potential confounders in GWAS with unadjusted for heritable covariates, which fulfills MR’s independence assumption, conversely, when GWAS adjusted for heritable covariates, it opens the path between SNPs with potential confounders through collider bias, destroying the independence and introducing bias to MR analysis [16,17,18].

Overall, the 35 modifiable factors that met the inclusion criteria for the dataset were systematically categorized into four major domains of socioeconomic factors (four factors) [19,20,21,22], behavioral factors (12 factors) [23,24,25,26,27,28,29], biochemical factors (12 factors) [30,31,32,33,34,35,36], and physical measures (seven factors) [37,38,39], as exposures (Fig. 1A). The newly published CMM GWAS dataset from the UK Biobank included 367,147 European ancestry individuals, and individuals with CMM were defined as the coexistence of two or three CMDs, including T2D, CAD, and stroke, while those with no or only one of the above conditions as controls [40]. The T2D GWAS dataset from the UK Biobank included 468,298 European ancestry individuals [38]. The CAD GWAS dataset was obtained from a large-scale meta-GWAS based on 181,522 cases among 1,165,690 European ancestry participants [41]. The stroke GWAS dataset was derived from a multi-source meta-GWAS including 1,308,460 European ancestry participants [42] (Fig. 1A).

Fig. 1
figure 1

Overview of the study design. A Summary of GWAS data classifications involved in this MR investigation. B Flow diagram depicts this three-stage design MR investigation on the causal associations of modifiable factors on CMDs and CMM, using TSMR, potential sample overlap correction and multiple testing, MVMR analysis, and multi-response MR. BMI, Body mass index; CAD, Coronary artery disease; CMM, Cardiometabolic multimorbidity; CRP, C-reactive protein; GWAS, Genome-wide association study; IV, Instrumental variable; MR, Mendelian randomization; MVMR, Multivariable MR; TSMR, Two-sample MR; T2D, Type 2 diabetes; 25OHD, 25-hydroxyvitamin D

Full size image

All GWAS data involved were adjusted only for age, sex, ancestry, and necessary study-specific covariates. The details of each GWAS data were summarized in Additional file 1: Table S1. The study design of each GWAS above, such as sample collection, quality control procedures, and imputation methods were described in the original publications, and ethical approval was obtained by each GWAS [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42].

Instrument selection

The instrument selection process was performed independently for each exposure. Preliminarily, SNPs with a P-value < 5×10−8 and minor allele frequency > 0.01 were extracted. To obtain independent SNPs, linkage disequilibrium clumping was further performed by retaining SNPs that have an r2 < 0.001 within a 10 MB window from the European panel in the 1000 genomes project using PLINK clumping method [43].

Then, the F-statistic for each retained instrumental variables (IVs) was calculated to evaluate the SNPs’ power. The F-statistic < 10 reflects a low instrument validity. The F-statistic was calculated as follows [14]:

$${F}_{\text{statistic}}=\frac{{BETA}^{2}}{{SE}^{2}}$$
(1)

The R2 for the proportion of the phenotype variance explained by each instrument was calculated as follows [14, 44]:

$$ R^{2} = \frac{{\frac{{\left( {2 \times EAF \times \left( {1 - EAF} \right)} \right.}}{{\left[ {\left( {2 \times EAF \times \left( {1 - EAF} \right) \times BETA^{2} } \right) + } \right.}}}}{{\frac{{\left. { \times BETA^{2} } \right)}}{{\left. {\left( {2 \times EAF \times \left( {1 - EAF} \right) \times N \times SE^{2} } \right)} \right]}}}} $$
(2)

In the above equations, BETA is the estimated genetic effects on modifiable factors, SE is the standard error of the genetic effect, EAF is the effect allele frequency, and N is the sample size of the GWAS for the modifiable factors. The details of each SNP chosen as IV in the respective exposure-outcome analysis were summarized in Additional file 2: Table S2.

The GWAS datasets of each set analysis of exposure and outcome were harmonized so that the effect size of both corresponded to the identical effect alleles. IVs that were palindromic with ambiguous allele frequencies or that had incompatible alleles were removed.

Mendelian randomization study design

This MR investigation was performed in accordance with three basic assumptions: (1) the IVs are vigorously associated with the modifiable factors in the TSMR analysis or at least one of the modifiable factors in MVMR analysis; (2) the IVs are independent of any potential confounders; and (3) the IVs influence each outcome only through each exposure rather than other ways. The MR analysis included three stages (Fig. 1B). In stage 1, the causal associations of modifiable factors on CMM and CMDs were systematically identified using TSMR analysis, then the potential sample overlap correction was comprehensively performed using effect difference test and causal effect correction analysis, followed by multiple testing. In stage 2, the independent causal associations of modifiable factors within an individual domain on CMM and CMDs were integrally estimated using MVMR analysis based on multiple testing. In stage 3, multi-response MR was used to model multiple CMDs in a joint MR model that simulated CMM, enhancing the robustness of causal inference for CMM, followed by multiple testing. In particular, the strength of multi-response MR is to infer the shared or distinct effects of single exposures on multiple outcomes through the joint model structure, and its relative robustness in the scenario of partially or completely overlapping samples of exposure and outcome GWAS data.

Statistical analysis

In the TSMR analysis, the Wald ratio was adopted to estimate the effect of each set of exposure on outcome for each IV, and then the inverse variance weighted (IVW) method was used to combine the effect size of each IV to generate an overall estimation of the causal effect, which served as the main method [14]. Moreover, weighted median (WM) and MR-Egger were performed as supplements to the IVW method. The WM method effectively pools the effects of individual variants under the principle that over 50% of the weight is derived from valid IVs, and the MR-Egger method allows all variants to be directly influenced, biasing the estimate in the same direction [14]. The causal effect sizes (β) and corresponding standard error (se) were converted into odds ratio (OR) with 95% confidence interval (CI) through natural logarithm e, as follows:

$$OR={e}^{\beta }$$
(3)
$${CI}_{95\%}=\left[{e}^{\beta -1.96\times se},{e}^{\beta +1.96\times se}\right]$$
(4)

As a further sensitivity analysis to evaluate the causal association, the leave-one-out analysis was performed to detect the outliers, the Cochrane’s Q values were applied to assess the heterogeneity, and the MR-Egger intercept method was tested for horizontal pleiotropy [14]. The MR Pleiotropy Residual Sum and Outlier (MR-PRESSO) test was performed to detect and correct for outliers with horizontal pleiotropy, and when significant outliers were detected, the outliers were removed and MR causal estimation was reassessed [45]. Furthermore, effect difference test and causal effect correction analysis were performed to minimize the bias in the estimation due to the potential sample overlap of GWASs for exposure and outcome, and multiple testing based on the Bonferroni procedure was conducted to detect more robust associations within the significant causal effect of the identified exposures on outcomes (threshold for Bonferroni P-adjust < 0.00143 [0.05/35]). In the MVMR analysis, the IVW served also as the main estimation method, and median, MR-Lasso, and MR-Egger were used as sensitivity analysis methods. In the multi-response MR, genetic variants that have been preprocessed in the IVW estimation for association with any of the modifiable factors that passed the multiple tests are used as IVs, and the results were derived by removing outliers or high-leverage and influential observations using scaled conditional predictive ordinate and fitting the remaining IVs. The significantly correlated pairs of responses is based on marginal posterior probability of inclusion (mPPI), which measures the strength of the direct causal association between each exposure-outcome combination and the corresponding direct causal effect, and the edge posterior probability of inclusion (ePPI), which describes the strength of the residual dependence between each pair of summary-level responses and the corresponding residual partial correlation, respectively. A non-parametric false discovery rate (FDR) strategy based on two-component mixture models, which clusters low and high levels of mPPIs and low and high levels of ePPIs, is applied to select important modifiable factors for each outcome and significant dependence patterns among outcomes at a fixed FDR level [15].

All the statistical analyses and visualization were conducted using R version 4.2.1. The TSMR and MVMR analyses were performed using the “TwoSampleMR” [46], “MendelianRandomization” [47], and “MVMR” [48] packages, and the multi-response MR were performed using the “MR2” package [15]. The effect difference test and causal effect correction analysis were performed using the “MRlap” package [49], and the MR-PRESSO was conducted using the “MRPRESSO” package [45].

Results

The number of SNPs involved in the modifiable factors ranged from 3 to 424, and the R2 varied from 0.031 to 9.493%. The F-statistic of each retained SNP was above the empirical threshold of 10, indicating that all the SNPs participating in the MR investigation had sufficient validity (Additional file 1: Table S1; Additional file 2: Table S2). The forest plot, funnel plot, scatter plot, and results of leave-one-out sensitivity analysis of TSMR analysis were summarized in Additional file 5: Fig. S1, and all the results of TSMR analysis, proportion of SNPs removed during harmonization, sample overlap correction, and sensitivity analysis were summarized in Additional file 3: Table S3.

Causal effect estimates of modifiable factors on CMM and CMDs

In stage 1, by performing the potential sample overlap correction analysis based on effect difference test statistic, the observed causal effect estimates were mostly retained and directionally consistent over the correction process across CMM and CMDs, other than the several associations, which were neutralized or converted to significant (Figs. 2, 3A, B, C, D; Additional file 3: Table S3). Based on rigorous screening thresholds for SNPs, outlier SNPs detection, and leave-one-out sensitivity analysis, the IVs used for causal association analysis of modifiable factors with CMM (Fig. 3A) and CMDs (Fig. 3B, C, D) demonstrated no horizontal pleiotropy.

Fig. 2
figure 2

TSMR analysis estimates of causal associations in modifiable factors with CMM and CMDs after potential sample overlap correction analysis. Forest plots depict the causal associations of socioeconomic factors, behavioral factors, biochemical factors, and physical measures with A CMM, B T2D, C CAD, and D stroke after potential sample overlap correction analysis. The effect sizes are converted to OR with 95% CI. Significant causal associations are designated with asterisk (*P-value < 0.05; **P-value < 0.01; ***P-value < 0.001). BMI, Body mass index; CAD, Coronary artery disease; CI, Confidence interval; CMDs, Cardiometabolic diseases; CMM, Cardiometabolic multimorbidity; CRP, C-reactive protein; FEV1, Forced expiratory volume in 1 s; FVC, Forced vital capacity; HbA1c, Glycated hemoglobin; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; MR, Mendelian randomization; NSNP, The number of single nucleotide polymorphisms; OR, Odds ratio; TSMR, Two-sample MR; T2D, Type 2 diabetes; 25OHD, 25-hydroxyvitamin D

Full size image
Fig. 3
figure 3

Detection of sample overlap correction difference, heterogeneity, and pleiotropy for TSMR analysis. Heatmap integrating the significance of effect difference test, heterogeneity, and pleiotropy in the causal inference on four major modifiable factor domains of socioeconomic factors, behavioral factors, biochemical factors, and physical measures with A CMM, B T2D, C CAD, and D stroke. Significant differences are designated with asterisk (*P-value < 0.05; **P-value < 0.01; ***P-value < 0.001). BMI, Body mass index; CAD, Coronary artery disease; CMM, Cardiometabolic multimorbidity; CRP, C-reactive protein; FEV1, Forced expiratory volume in 1 s; FVC, Forced vital capacity; HbA1c, Glycated hemoglobin; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; MR, Mendelian randomization; TSMR, Two-sample MR; T2D, Type 2 diabetes; 25OHD, 25-hydroxyvitamin D

Full size image

Among the socioeconomic factors, the TSMR analysis suggested that higher educational attainment was predicted to be genetically associated with lower risks of CMM (OR: 0.858; 95% CI: 0.834–0.883, P-adjusted: 2.100×10−24) (Fig. 2A), T2D (OR: 0.868; 95% CI: 0.843–0.893, P-adjusted: 9.030×10−21) (Fig. 2B), CAD (OR: 0.878; 95% CI 0.862–0.894, P-adjusted: 9.030×10−45) (Fig. 2C), and stroke (OR: 0.941; 95% CI 0.928–0.954, P-adjusted: 8.680×10−16) (Fig. 2D), and higher household income was identified to be associated with lower risks of CMM (OR: 0.794; 95% CI 0.720–0.875, P-adjusted: 1.133×10−4) (Fig. 2A), T2D (OR: 0.817; 95% CI 0.734–0.908, P-adjusted: 6.779×10−3) (Fig. 2B), and CAD (OR: 0.878; 95% CI 0.828–0.931, P-adjusted: 5.355×10−4) (Fig. 2C).

Among the behavioral factors, moderate to vigorous intensity physical activity was found to be consistently associated with lower risks of CMM (OR: 0.690; 95% CI 0.543–0.876, P-adjusted: 8.099×10−2) (Fig. 2A), T2D (OR: 0.666; 95% CI 0.515–0.861, P-adjusted: 6.822×10−2) (Fig. 2B), and CAD (OR: 0.717; 95% CI 0.608–0.845, P-adjusted: 2.538×10−3) (Fig. 2C). In contrast, leisure screen time was shown to be the most influential factor that associated with higher risks of CMM (OR: 1.255; 95% CI 1.186–1.327, P-adjusted: 8.085×10−14) (Fig. 2A) and T2D (OR: 1.239; 95% CI 1.165–1.317, P-adjusted: 2.716×10−10) (Fig. 2B), and lifetime smoking behavior was identified to be the most influential factor being associated with higher risks of CAD (OR: 1.186; 95% CI 1.152–1.221, P-adjusted: 3.364×10−29) (Fig. 2C) and stroke (OR: 1.060; 95% CI 1.037–1.083, P-adjusted: 8.470×10−6) (Fig. 2D).

Among the biochemical factors, higher apolipoprotein A-I (OR: 0.979; 95% CI 0.962–0.996, P-adjusted: 5.322×10−1), total cholesterol levels (OR: 0.981; 95% CI 0.965–0.996, P-adjusted: 4.906×10−1), and high-density lipoprotein (HDL) cholesterol levels (OR: 0.916; 95% CI 0.899–0.935, P-adjusted: 1.201×10−16) were predicted to be the most influential factors being genetically associated with lower risks of CMM (Fig. 2A), T2D (Fig. 2B), and CAD (Fig. 2C), respectively. While higher triglyceride levels (OR: 1.214; 95% CI 1.042–1.415, P-adjusted: 4.570×10−1), triglyceride levels (OR: 1.317; 95% CI 1.167–1.487, P-adjusted: 2.930×10−4), apolipoprotein B (OR: 1.163; 95% CI 1.148–1.178, P-adjusted: 1.012×10−111), and serum urate (OR: 1.028; 95% CI 1.017–1.040, P-adjusted: 5.530×10−5) were identified to be the most influential factors being associated with higher risks of CMM (Fig. 2A), T2D (Fig. 2B), CAD (Fig. 2C), and stroke (Fig. 2D), respectively.

Among the physical measures, higher forced expiratory volume in 1s (FEV1) was found to be the most influential factors that associated with lower risks of CMM (OR: 0.953; 95% CI 0.931–0.975, P-adjusted: 1.740×10−3) (Fig. 2A) and stroke (OR: 0.985; 95% CI 0.974–0.996, P-adjusted: 3.567×10−1) (Fig. 2D), and better forced vital capacity (FVC) with lower risks of T2D (OR: 0.927; 95% CI 0.901–0.954, P-adjusted: 9.800×10−6) (Fig. 2B) and CAD (OR: 0.941; 95% CI 0.928–0.954, P-adjusted: 1.691×10−16) (Fig. 2C). Additionally, higher body mass index (BMI) was shown to be the most influential factors that associated with higher risks of CMM (OR: 1.243; 95% CI 1.210–1.277, P-adjusted: 2.720×10−54) (Fig. 2A) and T2D (OR: 1.319; 95% CI 1.289–1.350, P-adjusted: 2.832×10−120) (Fig. 2B), and higher systolic blood pressure with higher risks of CAD (OR: 1.171; 95% CI 1.154–1.188, P-adjusted: 9.240×10−100) (Fig. 2C) and stroke (OR: 1.082; 95% CI 1.070–1.095, P-adjusted: 9.730×10−42) (Fig. 2D).

Independent causal effect estimates of modifiable factors on CMM and CMDs

In stage 2, there were 13, 13, 19, and 8 of the 23, 16, 25, and 12 causal effect estimates-corrected associations across CMM, T2D, CAD, and stroke remained significant after multiple testing based on Bonferroni procedure, respectively (Fig. 4). Further, the MVMR analysis was performed to identify the independence within each domain from the significant associations of modifiable factors with CMM and CMDs after multiple testing (Fig. 5A, B, C, D).

Fig. 4
figure 4

Multiple testing for the corrected causal effect estimates from TSMR analysis. Bidirectional Manhattan plot depicts more robust causal associations with CMM, T2D, CAD, and stroke risks before and after multiple testing based on Bonferroni procedure for four major modifiable factor domains of socioeconomic factors, behavioral factors, biochemical factors, and physical measures. Significant causal associations after multiple testing based on Bonferroni procedure are designated with asterisk (*P-value < 0.05; **P-value < 0.01; ***P-value < 0.001). BMI, Body mass index; CAD, Coronary artery disease; CMM, Cardiometabolic multimorbidity; FEV1, Forced expiratory volume in 1 s; FVC, Forced vital capacity; HbA1c, Glycated hemoglobin; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; MR, Mendelian randomization; TSMR, Two-sample MR; T2D, Type 2 diabetes

Full size image
Fig. 5
figure 5

MVMR analysis estimates of independent causal associations in modifiable factors with CMM and CMDs after adjusting exposures for one another within a single domain. Forest plot depicts the causal associations of socioeconomic factors, behavioral factors, biochemical factors, and physical measures with A CMM, B T2D, C CAD, and D stroke. The effect sizes are converted to OR with 95% CI. Significant causal associations are designated with asterisk (*P-value < 0.05; **P-value < 0.01; ***P-value < 0.001). To avoid multicollinearity, for MVMR analysis estimates of multiple exposures with CAD within the biochemical factor domain: a factors in lipid traits were adjusted for one another, and b glycemic traits and serum urate were adjusted for one another. BMI, Body mass index; CAD, Coronary artery disease; CI, Confidence interval; CMDs, Cardiometabolic diseases; CMM, Cardiometabolic multimorbidity; FEV1, Forced expiratory volume in 1 s; FVC, Forced vital capacity; HbA1c, Glycated hemoglobin; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; MR, Mendelian randomization; MVMR, Multivariable MR; NSNP, The number of single nucleotide polymorphisms; OR, Odds ratio; T2D, Type 2 diabetes

Full size image

After adjusting exposures for one another within a single domain, among the socioeconomic factors, MVMR analysis suggested that household income was predicted to be independently associated with CMM (Fig. 5A) and T2D (Fig. 5B) from educational attainment, and conversely, educational attainment was found to be independently associated with CAD (Fig. 5C) from household income.

Among the behavioral factors, lifetime smoking behavior were found to be independently associated with CMM from leisure screen time (Fig. 5A), daytime napping and lifetime smoking behavior were identified to be independently associated with T2D from leisure screen time (Fig. 5B), moderate to vigorous intensity physical activity and leisure screen time were shown to be independently associated with CAD from lifetime smoking behavior (Fig. 5C), and leisure screen time was also found to be independently associated with stroke from lifetime smoking behavior (Fig. 5D).

Among the biochemical factors, glycated hemoglobin (HbA1c) was predicted to be independently associated with CMM from low-density lipoprotein (LDL) cholesterol levels, total cholesterol levels, apolipoprotein B, and fasting glucose (Fig. 5A), HbA1c was also identified to be independently associated with T2D from triglyceride levels and fasting glucose (Fig. 5B), and serum urate was found to be independently associated with CAD (Fig. 5C) and stroke (Fig. 5D) from fasting glucose and HbA1c, and LDL cholesterol levels, respectively.

Among the physical measures, systolic blood pressure and FEV1, and systolic blood pressure and FVC were identified to be independently associated with CMM (Fig. 5A) and T2D (Fig. 5B) from diastolic blood pressure and BMI, and peak expiratory flow, FEV1, and BMI, and conversely, BMI was shown to be independently associated with CAD (Fig. 5C) and stroke(Fig. 5D) from systolic blood pressure, diastolic blood pressure, FVC, and FEV1, and systolic blood pressure and diastolic blood pressure, respectively.

Shared and distinct causal associations of modifiable factors across CMM and CMDs

Overall, the causal associations of multiple modifiable factors across CMM and CMDs overlapped (Fig. 6A). Specifically, after multiple testing, total cholesterol levels and apolipoprotein B were the causal modifiable factors shared by CMM and CAD, triglyceride levels and FVC by T2D and CAD, and serum urate by CAD and stroke, respectively. Further, household income, fasting glucose, HbA1c, and FEV1 were shared by CMM, T2D, and CAD, and LDL cholesterol levels and diastolic blood pressure by CMM, CAD, and stroke. Moreover, educational attainment, lifetime smoking behavior, leisure screen time, systolic blood pressure, and BMI were shared by CMM, T2D, CAD, and stroke jointly. Additionally, the effects of daytime napping and peak expiratory flow on T2D, and moderate to vigorous intensity physical activity, HDL cholesterol levels, and apolipoprotein A-I on CAD were distinct.

Fig. 6
figure 6

Shared and distinct causal associations of modifiable factors with CMM and CMDs. A Venn diagram depicts the number and term of shared and distinct causal associations of modifiable factors on CMM, T2D, CAD, and stroke in observed causal effect estimation, corrected causal effect estimation, and multiple testing causal effect estimation. B Heatmap depicts the mPPI of each modifiable factor against each CMD in multi-response MR, from the perspective of all domains of modifiable factors and CMDs. The mPPIs above the 5% FDR are highlighted with asterisks. In multi-response MR analysis, associations that are also significant in two-sample MR are highlighted with blank triangles, and associations that are still significant in multivariable MR are further highlighted with black triangles. The network depicts the ePPI among CMDs. C Heatmap depicts the mPPI of each modifiable factor against each CMD in multi-response MR, within each single domain of modifiable factors and CMDs. Domains of socioeconomic factors, behavioral factors, biochemical factors, physical measures are presented in blue, red, yellow, and green in order. The mPPIs above the 5% FDR are highlighted with asterisks. In multi-response MR analysis, associations that are also significant in two-sample MR are highlighted with blank triangles, and associations that are still significant in multivariable MR are further highlighted with black triangles. The network depicts the ePPI among CMDs. BMI, Body mass index; CAD, Coronary artery disease; CMD, Cardiometabolic disease; CMM, Cardiometabolic multimorbidity; ePPI, Edge posterior probability of inclusion; FDR, False discovery rate; FEV1, Forced expiratory volume in 1 s; FVC, Forced vital capacity; HbA1c, Glycated hemoglobin; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; mPPI, Marginal posterior probability of inclusion; MR, Mendelian randomization; T2D, Type 2 diabetes

Full size image

Further, multi-response MR, which enables comparison of causal associations across multiple modifiable factor domains and outcomes, was applied to enhance the findings from TSMR and MVMR. When focusing on the intersections of the TSMR findings after multiple testing (Figs. 4, 6A) with multi-response MR analysis, as marked by blank triangles (Fig. 6B, C), the results suggesting that the associations of educational attainment with all CMDs, triglyceride levels and fasting glucose with T2D and CAD, LDL cholesterol levels, serum urate, and BMI with CAD and stroke, lifetime smoking behavior with T2D, moderate to vigorous intensity physical activity, HDL cholesterol levels, total cholesterol levels, apolipoprotein B, HbA1c, systolic blood pressure, and FVC with CAD, and leisure screen time and diastolic blood pressure with stroke are substantially robust.

Moreover, multi-response MR specifically takes into account both the conditional independence among modifiable factors and across outcomes, reinforcing the evidence from the MVMR. From the perspective of all domains of modifiable factors and CMDs, when focusing on the intersections of the MVMR findings (Fig. 5B, C, D) with multi-response MR analysis, as marked by black triangles (Fig. 6B), the results suggesting that educational attainment holds a prominent and independent influence on CAD. In contrast, when further focusing on modifiable factors within each single domain (Fig. 6C), educational attainment, serum urate, and BMI were the modifiable factors shared by CAD and stroke independent of other exposure-outcome pair associations, and lifetime smoking behavior with T2D, moderate to vigorous intensity physical activity with CAD, and leisure screen time with stroke were distinct and independent associations within the behavioral factor domain.

Therefore, based on the findings of TSMR (Figs. 4, 6A) and MVMR (Fig. 5B, C, D), with the enhancement of multi-response MR (Fig. 6B, C), educational attainment, serum urate, and BMI presented as the specific priority for the prevention and control of CMM, and lifetime smoking behavior, moderate to vigorous intensity physical activity, and leisure screen time, held prominent impacts on diverse CMDs. Furthermore, glycemic and lipid profile, blood pressure, and lung functions require additional attention in terms of their significant effects on the risk of CMDs and CMM.

Discussion

In this three-stage design MR investigation, we applied newly published CMM and large-scale CMDs data to systematically identify the causal associations of modifiable factors with CMM and CMDs. Overall, our findings indicated that 1) 23, 16, 25, and 12 of 24, 17, 25, and 12 modifiable factors across the four domains were associated with CMM, T2D, CAD, and stroke under the support of potential sample overlap correction analysis in the TSMR analysis, respectively, and 13, 13, 19, and 8 modifiable factors were evaluated to have robust associations with CMM, T2D, CAD, and stroke under the filter of multiple testing, respectively, 2) furthermore, household income, lifetime smoking behavior, HbA1c, systolic blood pressure, and FEV1 with CMM, household income, lifetime smoking behavior, daytime napping, HbA1c, systolic blood pressure, and FVC with T2D, educational attainment, moderate to vigorous intensity physical activity, leisure screen time, serum urate, and BMI with CAD, and leisure screen time, serum urate, and BMI with stroke, were detected as independent associations in the MVMR analysis after adjusting exposures for one another within a single domain, and 3) moreover, the multi-response MR provided reinforcing evidence for the associations of educational attainment, lifetime smoking behavior, moderate to vigorous intensity physical activity, leisure screen time, serum urate, and BMI with diverse CMDs, which are prioritized for CMM and CMDs prevention. Our findings emphasized that a targeted approach to promoting healthy modifiable factors contributes to the co-prevention and co-management of CMDs, mitigating the tremendous burden of CMM on healthy aging.

Several previous scattered findings from observational have addressed the links of some common modifiable factors with CMM and CMDs. A 20-year cohort study based on Australian women provided several key findings. Over 3 years, the age-adjusted odds of developing two or more CMDs were approximately twice that of developing one new condition relative to women without any new conditions. Additionally, this study found that social inequality, obesity, hypertension, physical inactivity, and smoking were prominently associated with increased odds of accumulating multimorbidity [5]. An investigation based on British population demonstrated that, at the age of 45 years, physical activity contributed to an increase in life expectancy of 2.34 years for individuals with diabetes relative to physical inactivity, while the corresponding estimates for individuals with cardiovascular disease (CVD) and CMM were 2.28 and 2.15 years, respectively [9]. Our previous analysis based on multistate models also showed that baseline cardiorespiratory fitness was associated with a lower risk of progressing from healthy state to incident first CMD and transitioning to CMM, suggesting that strengthening physical activity to improve cardiorespiratory fitness is a potential strategy for preventing the development of CMM [10]. Additionally, our previous large prospective cohort study found that participants with more than 5 h of leisure screen time per day had a significantly higher risk of stroke compared to participants with less than 1 h [11]. Furthermore, a pooled analysis was conducted across 16 cohort studies from the USA and Europe. The analysis suggested that compared to individuals with healthy BMI, the risk of CMM was twice as high in overweight individuals, and more than ten times in severely obese individuals [7]. And an investigation on lung function assessed by FVC and FEV1 based on British population indicated that better lung function was associated with a lower risk of new-onset CMDs [8]. In addition, one previous research found that elevated blood remnant cholesterol and triglycerides were statistically associated with gradually higher CMM risk. The risk increase was especially notable for the progression of CAD to CAD and T2D, and the corresponding causal associations were detected using one-sample MR [3].

Compared to previous MR studies, the novelty of our study is, firstly, the three-stage MR design, which promotes evidence robustness. Secondly, our study uniquely applies multimorbidity data, rather than data of a single CMD in TSMR and MVMR analyses, to provide new insights for causal inference of modifiable factors in multimorbidity scenarios. Thirdly, we further applied a newly developed multi-response MR method based on the classical MR approach, which provides an additional practice pathway for the validation of causal effects of multimorbidity by integrating data from a single CMD to simulate the CMM state. Our results provided consistent and strengthened evidence for previous wide-angled MR studies, such as the influences of educational attainment, smoking and BMI with T2D [50], most of lipid and glycemic traits with CAD [51], educational attainment and blood pressure with stroke [52]. Additionally, this study also disambiguated several causal effects of exposures on CMDs using large-scale updated GWAS data that were overestimated or underestimated previously. In previous MR studies based on prior data from different sources than this study, conflicts among limited evidence exist in the associations of serum urate with CMDs [53, 54]. And our results support the potential adverse effects of serum urate on CMDs and CMM. Moreover, two studies have revealed negative associations between FVC and FEV1 with CMDs [55, 56], and our study further extends the lung function indices, with novel findings of robust associations of peak expiratory flow with T2D and FEV1 with CMM after multiple testing of TSMR, and further, although not significant after multiple testing, FEV1/FVC was suggestively associated with CAD and CMM, and FVC and peak expiratory flow with CMM. Different from previous MR studies, interestingly, we also identified some novel associations at diverse stages in this MR investigation. After multiple testing in TSMR, daytime napping was a potential risk factor for T2D. In addition, although not supported by multiple testing, short sleep, daytime sleepiness with CMM risk, short sleep, daytime napping, and daytime sleepiness with CAD risk showed suggestive positive associations in TSMR, whereas champagne or white wine consumption presented a suggestive protective effect for CAD. However, some unexpected results exist. Inconsistent with conventional perceptions, current data failed to support the associations of alcohol consumption with T2D and stroke, and serum C-reactive protein with CMDs and CMM. For complex lipid traits, based on updated large-scale GWAS data, our findings further support previous extensive MR analyses that included the associations of triglyceride levels with CAD, HDL cholesterol levels with CAD, LDL cholesterol levels with CAD, total cholesterol levels with T2D and CAD [50, 51, 57], and additionally found potential associations of triglyceride levels with T2D, LDL cholesterol levels with stroke, apolipoprotein A-I with CAD, and apolipoprotein B with CAD and stroke at diverse MR identification stages. Moreover, based on the results of TSMR after multiple testing, multi-response MR particularly emphasized the associations of triglyceride levels and LDL cholesterol levels with CMM.

CMDs tend to share multiple risk factors, which not only contribute to prolonged accumulation of multiple cardiometabolic risks in individuals through some common biological mechanisms, but also facilitate the emergence of future CMM conditions. Our proteomic signature analysis of CMM revealed that specific proteins, including PCSK9, NRP1, and CD27, are critical for mediating CMM development. These proteins influence vascular homeostasis and angiogenesis by participating in multiple biological mechanisms, including lipid metabolism and vascular regulation, leading to the transition from CMDs to CMM in individuals [58]. Previous evidence suggested that education, as a critical factor in early adulthood, persistently impacts a wide range of downstream factors that potentially linger throughout the lifespan in individuals, from shaping healthy behaviors, favorable metabolic indicators, to better physical measures tied to subsequent health outcomes [59, 60]. Mounting evidence also underscored the essential role of epigenetic mechanisms, typically DNA methylation, in mediating the influence of socioeconomic factors, mainly educational attainment and household income, on the development and progression of CMM. Investigations revealed that low childhood socioeconomic status (SES) was associated with CMM-related DNA methylation in three stress-related genes (AVP, FKBP5, OXTR) and two inflammation-related genes (CCL1, CD1D). In addition, low adult SES was associated with CMM-related DNA methylation in one stress-related gene (AVP) and five inflammation-related genes (CD1D, F8, KLRG1, NLRP12, and TLR3) [61]. Moreover, individuals with low SES tended to be more dependent on ultra-processed foods with high sugar and fat, which triggered an imbalance in the gut microbiome, affecting the DNA methylation patterns, adversely influencing metabolic processes that lead to the production of cardiometabolic risk metabolites like trimethylamine N-oxide [61,62,63]. Accumulating experimental evidence demonstrated that the main toxicants in tobacco, such as transition metals, carbon monoxide, aldehydes, and nicotine, severely impact endothelial and vascular function, elevating thiobarbituric acid reactive substances in heart, lipid peroxidation, interleukin-1β, tumor necrosis factor-α, fibrinogen, and DNA strand breaks. These effects could trigger oxidative stress, inflammation, thrombosis, and DNA damage, leading to the glycolipid metabolism disorder and progression of CMDs [64, 65]. The reduced energy expenditure resulting from sedentary behavior limits skeletal muscle activation, decreases blood flow, and raises the risk of metabolic impairment. The resulting metabolic impairment drives insulin resistance, promotes endothelial dysfunction and reactive oxygen species production. In contrast, physical activity contributes to the triggering of protective mechanisms that shield individuals from CMDs risks [66]. Previous evidence suggested that soluble uric acid could have multiple pro-inflammatory effects, and activates the renin-angiotensin system by stimulating plasma renin activity and renal renin expression, along with the activation of the intra-renal angiotensin system. Uric acid also induces insulin resistance through the inhibition of hepatic AMP-activated protein kinase, which exacerbates the risk of CMM in individuals [67]. While impaired lung functions similarly lead to systemic inflammation and excessive oxidative stress, which induces vascular endothelial dysfunction and consequently the formation of atherosclerotic plaques [68, 69]. And obesity could also trigger glucose metabolism disorders and lipotoxicity through insulin resistance, prompting the release of inflammatory factors from macrophages and adipocytes, inactivation of nitric oxide, activation of the sympathetic nervous system and the renin–angiotensin–aldosterone system, and the prolonged presence of this series of biological mechanisms would result in cardiac dysfunction and myocardial damage, facilitating the transition of health states to CMDs and subsequent CMM [70, 71]. Overall, as risk factors across multiple domains continuously accumulate, multiple detrimental biological mechanisms in individuals would be triggered and all segments of the vasculature would experience inflammation, remodeling, and dysfunction leading to a vicious circulatory and metabolic burden, resulting in the gradual superimposition of multiple CMDs, thus developing CMM condition.

Our findings underscored some modifiable factors that represent opportunities for CMM early intervention and integrated care. Translating these findings into clinical practice could begin with enhanced screening and risk stratification strategies. Our results emphasized that, among the multiple associations with CMM identified in the current multi-domain framework, improving educational attainment is essential for reducing the risk of CMM for populations in general. For individuals with limited educational attainment, prioritizing improvements in lifetime smoking behavior, sedentary behavior, physical activity, serum urate, and unfavorable BMI is beneficial for the co-prevention in the CMDs risk and the detrimental effects of CMM in the long run. By incorporating these modifiable factors into routine risk assessments, clinicians could more effectively tailor lifestyle interventions and educational initiatives for individuals at high risk for CMDs and CMM. Our results also highlighted that, strengthening the control of indicators of glycemia and lipids, maintaining favorable blood pressure, and enhancing lung functions would contribute to the co-management of CMDs and preventing the development of long-term CMM condition. Based on the findings from our study and the integration of evidence from other studies, firstly, regarding clinical pathways and risk stratification, promoting multi-domain risk assessment focusing on educational attainment, smoking history, physical activity, physical measures (blood pressure, BMI, lung function), and critical biomarkers (lipids, glucose, and serum urate), to identify high-risk individuals, conducting multidisciplinary care such as coordinated referrals to dietitians, exercise specialists, and smoking-cessation programs, and personalizing targets for each modifiable factor, are essential for improving individual CMDs and CMM risk. Secondly, regarding community health promotion and education, efforts in advancing healthy behavior education programs for individuals with limited education attainment and providing self-management skills such as smoking cessation, self-directed exercise, and nutritional choices would contribute to mitigating the risk of CMDs and long-term CMM for high-risk populations. Thirdly, regarding public health policies, advocating the establishment of smoke-free zones in public areas, expanding accessible exercise facilities, and strengthening the implementation of education equality policies are valuable in alleviating the burden of CMDs and CMM on the populations. The integration of these potential modifiable factors into the health management of CMM is valuable, and future guidelines might broaden the focus through continuous monitoring from more research to establish causal evidence.

However, several limitations of this MR investigation need to be addressed. First, the possibility of violating independence and exclusion restriction assumptions cannot be completely avoided, but we applied multiple sensitivity analysis methods to infer robust causal estimates to minimize the bias. Second, causality should be interpreted with caution due to the potential for residual confounders, horizontal pleiotropy, unmeasured confounding, and exposure misclassification in MR analysis. But we adhered to rigorous screening procedure to ensure that the analysis was based on high-quality GWAS data, capturing IVs through a series of rigorous threshold screening processes and further testing the robustness of the IVs using MR-PRESSO outlier test and leave-one-out test, detecting and correcting horizontal pleiotropy issues through MR-Egger intercept test and MR-PRESSO test, and minimizing the false-positive rate of the result by multiple testing. Third, limited by the availability of exposure GWAS data, some exposures potentially associated with CMM and CMDs, such as environmental factors, cannot be evaluated due to the absence of available IVs or corresponding GWAS data adjusted for heritable covariates that may introduce collider bias. And specifically, due to the lack of dietary pattern GWAS data or their failure to meet the data inclusion criteria, we did not include dietary factors in the scope of this study, and future high-quality dietary pattern data would provide insights into the associations of dietary factors with CMDs and CMM. Fourth, our study relied strongly on European ancestry data, thus, generalizing our findings to other populations should be with caution, and sample overlap between exposures and outcomes data may exist, but we used potential sample overlap correction analysis to ameliorate measurement bias. In addition, MR contributes to providing one-sided insights that converge on the causal associations, however, establishing solid causal evidence requires the integration of findings from diverse research paradigms through triangulation.

Despite these limitations, our MR investigation holds several major strengths. First, this investigation applied newly published CMM and large-scale CMDs data, to address the complex links of multiple domains of modifiable factors with CMM and CMDs through a comprehensive MR analysis strategy, and MR analysis strengthened the causal inference by diminishing confounding factors and reverse causality. Second, for all GWAS data, participants were of European ancestry, indicating that the population stratification is unlikely to bias our findings, rendering the investigation relatively robust. Third, we used a series of rigorous testing and correction procedures including potential sample overlap correction analysis and multiple testing in causal effect estimation to minimize bias due to sample overlap of the GWAS data and filter out more robust causal associations. Fourth, based on updated GWAS data and relatively rigorous and robust methods combining TSMR, MVMR, and multi-response MR, we disambiguated or identified some causal effects of exposures on CMDs that were overestimated or underestimated in previous MR studies.

Conclusions

In conclusion, this three-stage design MR investigation comprehensively identified causal associations of multi-domain modifiable factors with CMM and CMDs. Our major findings underscored pivotal role of educational attainment in reducing the risk of CMM. For individuals with limited educational attainment, targeted strategies in smoking cessation, reducing sedentary behavior, strengthening physical activity, maintaining favorable serum urate, and controlling obesity are prioritized for CMDs prevention. Furthermore, improving dyslipidemia and dysglycemia, maintaining favorable blood pressure, and enhancing lung function, held prominent protective impacts on CMDs and long-term CMM condition. These findings carry significant clinical implications, suggesting that incorporating these modifiable factors could facilitate the construction of individualized intervention strategies for CMDs and CMM, and inform future guidelines aimed at multimorbidity risk management. From the perspective of the triangulation, the integration of findings across multiple study types to further minimize study bias is of significant public health implication in establishing robust evidence of causal associations between multiple modifiable factors and CMDs, providing informative support for primary prevention policies of CMM.

Availability of data and materials

All GWAS summary statistics that support the findings of this study are publicly downloadable on the websites of the GWAS Catalog (https://www.ebi.ac.uk/gwas/home). All of these data are de-identified and with public availability. All the R-packages used are open-source for academic use. All other data are available within the article. All the analyses used in this MR investigation were conducted using TwoSampleMR, MendelianRandomization, MVMR, MRlap, MR2, and MRPRESSO of R packages. The codes that support the findings of this MR investigation are available from the corresponding author upon reasonable request.

Abbreviations

BMI:

Body mass index

CAD:

Coronary artery disease

CI:

Confidence interval

CMDs:

Cardiometabolic diseases

CMM:

Cardiometabolic multimorbidity

CVD:

Cardiovascular disease

ePPI:

Edge posterior probability of inclusion

FDR:

False discovery rate

FEV1 :

Forced expiratory volume in 1 s

FVC:

Forced vital capacity

GWAS:

Genome-wide association study

HbA1c:

Glycated hemoglobin

HDL:

High-density lipoprotein

IVs:

Instrumental variables

IVW:

Inverse variance weighted

LDL:

Low-density lipoprotein

mPPI:

Marginal posterior probability of inclusion

MR:

Mendelian randomization

MR-PRESSO:

MR pleiotropy residual sum and outlier

MVMR:

Multivariable MR

OR:

Odds ratio

SES:

Socioeconomic status

SNPs:

Single nucleotide polymorphisms

TSMR:

Two-sample MR

T2D:

Type 2 diabetes

WM:

Weighted median

References

  1. Di Angelantonio E, Kaptoge S, Wormser D, Willeit P, Butterworth AS, et al. Association of cardiometabolic multimorbidity with mortality. JAMA. 2015;314(1):52–60.

    Article  PubMed  Google Scholar 

  2. Jin Y, Liang J, Hong C, Liang R, Luo Y. Cardiometabolic multimorbidity, lifestyle behaviours, and cognitive function: a multicohort study. Lancet Healthy Longev. 2023;4(6):e265–73.

    Article  PubMed  Google Scholar 

  3. Zhao Y, Zhuang Z, Li Y, Xiao W, Song Z, Huang N, et al. Elevated blood remnant cholesterol and triglycerides are causally related to the risks of cardiometabolic multimorbidity. Nat Commun. 2024;15(1):2451.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Gerdts E, Regitz-Zagrosek V. Sex differences in cardiometabolic disorders. Nat Med. 2019;25(11):1657–66.

    Article  CAS  PubMed  Google Scholar 

  5. Xu X, Mishra GD, Dobson AJ, Jones M. Progression of diabetes, heart disease, and stroke multimorbidity in middle-aged women: a 20-year cohort study. PLoS Med. 2018;15(3): e1002516.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Han Y, Hu Y, Yu C, Guo Y, Pei P, Yang L, et al. Lifestyle, cardiometabolic disease, and multimorbidity in a prospective Chinese study. Eur Heart J. 2021;42(34):3374–84.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Kivimaki M, Kuosma E, Ferrie JE, Luukkonen R, Nyberg ST, Alfredsson L, et al. Overweight, obesity, and risk of cardiometabolic multimorbidity: pooled analysis of individual-level data for 120 813 adults from 16 cohort studies from the USA and Europe. Lancet Public Health. 2017;2(6):e277–85.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Li G, Lu Y, Qiao Y, Hu D, Ke C. Role of pulmonary function in predicting new-onset cardiometabolic diseases and cardiometabolic multimorbidity. Chest. 2022;162(2):421–32.

    Article  PubMed  Google Scholar 

  9. Chudasama YV, Zaccardi F, Gillies CL, Dhalwani NN, Yates T, Rowlands AV, et al. Leisure-time physical activity and life expectancy in people with cardiometabolic multimorbidity and depression. J Intern Med. 2020;287(1):87–99.

    Article  CAS  PubMed  Google Scholar 

  10. Chen Y, Yang H, Li D, Zhou L, Lin J, Yin X, et al. Association of cardiorespiratory fitness with the incidence and progression trajectory of cardiometabolic multimorbidity. Br J Sports Med. 2025;59(5):306–15.

    Article  PubMed  Google Scholar 

  11. Xu C, Cao Z, Lu Z, Hou Y, Wang Y, Zhang X. Associations between recreational screen time and brain health in middle-aged and older adults: a large prospective cohort study. J Am Med Dir Assoc. 2024;25(8): 104990.

    Article  PubMed  Google Scholar 

  12. Davey Smith G, Ebrahim S. What can Mendelian randomisation tell us about modifiable behavioural and environmental exposures? BMJ. 2005;330(7499):1076–9.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. 2014;23(R1):R89-98.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Burgess S, Thompson SG. Mendelian randomization: methods for using genetic variants in causal estimation. Boca Raton: Chapman & Hall; 2015.

    Book  Google Scholar 

  15. Zuber V, Lewin A, Levin MG, Haglund A, Ben-Aicha S, Emanueli C, et al. Multi-response Mendelian randomization: identification of shared and distinct exposures for multimorbidity and multiple related disease outcomes. Am J Hum Genet. 2023;110(7):1177–99.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Burgess S. “C-reactive protein levels and risk of dementia”: subgroup analyses in Mendelian randomization are likely to be misleading. Alzheimers Dement. 2022;18(12):2732–3.

    Article  CAS  PubMed  Google Scholar 

  17. Sollis E, Mosaku A, Abid A, Buniello A, Cerezo M, Gil L, et al. The NHGRI-EBI GWAS catalog: knowledgebase and deposition resource. Nucleic Acids Res. 2023;51(D1):D977–85.

    Article  CAS  PubMed  Google Scholar 

  18. Hartwig FP, Tilling K, Davey Smith G, Lawlor DA, Borges MC. Bias in two-sample Mendelian randomization when using heritable covariable-adjusted summary associations. Int J Epidemiol. 2021;50(5):1639–50.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Day FR, Ong KK, Perry JRB. Elucidating the genetic basis of social interaction and isolation. Nat Commun. 2018;9(1):2457.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Hill WD, Davies NM, Ritchie SJ, Skene NG, Bryois J, Bell S, et al. Genome-wide analysis identifies molecular systems and 149 genetic loci associated with income. Nat Commun. 2019;10(1):5741.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Kim S, Kim K, Hwang MY, Ko H, Jung SH, Shim I, et al. Shared genetic architectures of subjective well-being in East Asian and European ancestry populations. Nat Hum Behav. 2022;6(7):1014–26.

    Article  PubMed  Google Scholar 

  22. Okbay A, Wu Y, Wang N, Jayashankar H, Bennett M, Nehzati SM, et al. Polygenic prediction of educational attainment within and between families from genome-wide association analyses in 3 million individuals. Nat Genet. 2022;54(4):437–49.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Cole JB, Florez JC, Hirschhorn JN. Comprehensive genomic analysis of dietary habits in UK Biobank identifies hundreds of genetic associations. Nat Commun. 2020;11(1):1467.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Dashti HS, Daghlas I, Lane JM, Huang Y, Udler MS, Wang H, et al. Genetic determinants of daytime napping and effects on cardiometabolic health. Nat Commun. 2021;12(1):900.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Dashti HS, Jones SE, Wood AR, Lane JM, van Hees VT, Wang H, et al. Genome-wide association study identifies genetic loci for self-reported habitual sleep duration supported by accelerometer-derived estimates. Nat Commun. 2019;10(1):1100.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Saunders GRB, Wang X, Chen F, Jang SK, Liu M, Wang C, et al. Genetic diversity fuels gene discovery for tobacco and alcohol use. Nature. 2022;612(7941):720–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Wang H, Lane JM, Jones SE, Dashti HS, Ollila HM, Wood AR, et al. Genome-wide association analysis of self-reported daytime sleepiness identifies 42 loci that suggest biological subtypes. Nat Commun. 2019;10(1):3503.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Wang Z, Emmerich A, Pillon NJ, Moore T, Hemerich D, Cornelis MC, et al. Genome-wide association analyses of physical activity and sedentary behavior provide insights into underlying mechanisms and roles in disease prevention. Nat Genet. 2022;54(9):1332–44.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Wootton RE, Richmond RC, Stuijfzand BG, Lawn RB, Sallis HM, Taylor GMJ, et al. Evidence for causal effects of lifetime smoking on risk for depression and schizophrenia: a Mendelian randomisation study. Psychol Med. 2020;50(14):2435–43.

    Article  PubMed  Google Scholar 

  30. Cho C, Kim B, Kim DS, Hwang MY, Shim I, Song M, et al. Large-scale cross-ancestry genome-wide meta-analysis of serum urate. Nat Commun. 2024;15(1):3441.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Graham SE, Clarke SL, Wu KH, Kanoni S, Zajac GJM, Ramdas S, et al. The power of genetic diversity in genome-wide association studies of lipids. Nature. 2021;600(7890):675–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Jung H, Jung HU, Baek EJ, Kwon SY, Kang JO, Lim JE, et al. Integration of risk factor polygenic risk score with disease polygenic risk score for disease prediction. Commun Biol. 2024;7(1):180.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Lagou V, Magi R, Hottenga JJ, Grallert H, Perry JRB, Bouatia-Naji N, et al. Sex-dimorphic genetic effects and novel loci for fasting glucose and insulin variability. Nat Commun. 2021;12(1):24.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Revez JA, Lin T, Qiao Z, Xue A, Holtz Y, Zhu Z, et al. Genome-wide association study identifies 143 loci associated with 25 hydroxyvitamin D concentration. Nat Commun. 2020;11(1):1647.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Richardson TG, Sanderson E, Palmer TM, Ala-Korpela M, Ference BA, Davey Smith G, et al. Evaluating the relationship between circulating lipoprotein lipids and apolipoproteins with risk of coronary heart disease: a multivariable Mendelian randomisation analysis. PLoS Med. 2020;17(3): e1003062.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Said S, Pazoki R, Karhunen V, Vosa U, Ligthart S, Bodinier B, et al. Genetic analysis of over half a million people characterises C-reactive protein loci. Nat Commun. 2022;13(1):2198.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Koskeridis F, Evangelou E, Said S, Boyle JJ, Elliott P, Dehghan A, et al. Pleiotropic genetic architecture and novel loci for C-reactive protein levels. Nat Commun. 2022;13(1):6939.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Loh PR, Kichaev G, Gazal S, Schoech AP, Price AL. Mixed-model association for biobank-scale datasets. Nat Genet. 2018;50(7):906–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Sinkala M, Elsheikh SSM, Mbiyavanga M, Cullinan J, Mulder NJ. A genome-wide association study identifies distinct variants associated with pulmonary function among European and African ancestries from the UK Biobank. Commun Biol. 2023;6(1):49.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Zhao C, Ma T, Cheng X, Zhang G, Bai Y. Genome-wide association study of cardiometabolic multimorbidity in the UK Biobank. Clin Genet. 2024;106(1):72–81.

    Article  CAS  PubMed  Google Scholar 

  41. Aragam KG, Jiang T, Goel A, Kanoni S, Wolford BN, Atri DS, et al. Discovery and systematic characterization of risk variants and genes for coronary artery disease in over a million participants. Nat Genet. 2022;54(12):1803–15.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Mishra A, Malik R, Hachiya T, Jurgenson T, Namba S, Posner DC, et al. Stroke genetics informs drug discovery and risk prediction across ancestries. Nature. 2022;611(7934):115–23.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74.

    Article  PubMed  Google Scholar 

  44. Papadimitriou N, Dimou N, Tsilidis KK, Banbury B, Martin RM, Lewis SJ, et al. Physical activity and risks of breast and colorectal cancer: a Mendelian randomisation analysis. Nat Commun. 2020;11(1):597.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Verbanck M. MRPRESSO: Performs the Mendelian randomization pleiotropy RESidual sum and outlier (MR-PRESSO) test. 2017.

  46. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, Lewis, et al. The MR-Base platform supports systematic causal inference across the human phenome. Elife. 2018. https://doi.org/10.7554/eLife.34408.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Broadbent JR, Foley CN, Grant AJ, Mason AM, Staley JR, Burgess S. MendelianRandomization v0.5.0: updates to an R package for performing Mendelian randomization analyses using summarized data. Well Open Res. 2020;5:252. https://doi.org/10.12688/wellcomeopenres.16374.2.

    Article  Google Scholar 

  48. Sanderson E, Spiller W, Bowden J. Testing and correcting for weak and pleiotropic instruments in two-sample multivariable Mendelian randomization. Stat Med. 2021;40(25):5434–52.

    Article  PubMed  PubMed Central  Google Scholar 

  49. Mounier N, Kutalik Z. Bias correction for inverse variance weighting Mendelian randomization. Genet Epidemiol. 2023;47(4):314–31.

    Article  CAS  PubMed  Google Scholar 

  50. Yuan S, Larsson SC. An atlas on risk factors for type 2 diabetes: a wide-angled Mendelian randomisation study. Diabetologia. 2020;63(11):2359–71.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Wang K, Shi X, Zhu Z, Hao X, Chen L, Cheng S, et al. Mendelian randomization analysis of 37 clinical factors and coronary artery disease in East Asian and European populations. Genome Med. 2022;14(1):63.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Li HQ, Feng YW, Yang YX, Leng XY, Zhang PC, Chen SD, et al. Causal relations between exposome and stroke: a Mendelian randomization study. J Stroke. 2022;24(2):236–44.

    Article  PubMed  PubMed Central  Google Scholar 

  53. Keenan T, Zhao W, Rasheed A, Ho WK, Malik R, Felix JF, et al. Causal assessment of serum urate levels in cardiometabolic diseases through a mendelian randomization study. J Am Coll Cardiol. 2016;67(4):407–16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Gill D, Cameron AC, Burgess S, Li X, Doherty DJ, Karhunen V, et al. Urate, blood pressure, and cardiovascular disease: evidence from Mendelian randomization and meta-analysis of clinical trials. Hypertension. 2021;77(2):383–92.

    Article  CAS  PubMed  Google Scholar 

  55. Au Yeung SL, Borges MC, Lawlor DA, Schooling CM. Impact of lung function on cardiovascular diseases and cardiovascular risk factors: a two sample bidirectional Mendelian randomisation study. Thorax. 2022;77(2):164–71.

    Article  PubMed  Google Scholar 

  56. Zhu J, Zhao H, Chen D, Tse LA, Kinra S, Li Y. Genetic correlation and bidirectional causal association between type 2 diabetes and pulmonary function. Front Endocrinol. 2021;12: 777487.

    Article  Google Scholar 

  57. Martin-Campos JM, Carcel-Marquez J, Llucia-Carol L, Lledos M, Cullell N, Muino E, et al. Causal role of lipid metabolome on the risk of ischemic stroke, its etiological subtypes, and long-term outcome: a Mendelian randomization study. Atherosclerosis. 2023;386: 117382.

    Article  CAS  PubMed  Google Scholar 

  58. Li Y, Li D, Lin J, Zhou L, Yang W, Yin X, et al. Proteomic signatures of type 2 diabetes predict the incidence of coronary heart disease. Cardiovasc Diabetol. 2025;24(1):120.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Lawrence EM. Why do college graduates behave more healthfully than those who are less educated? J Health Soc Behav. 2017;58(3):291–306.

    Article  PubMed  PubMed Central  Google Scholar 

  60. Zajacova A, Lawrence EM. The relationship between education and health: reducing disparities through a contextual approach. Annu Rev Public Health. 2018;39:273–89.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Baccarelli AA, Ordovas J. Epigenetics of early cardiometabolic disease: mechanisms and precision medicine. Circ Res. 2023;132(12):1648–62.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Darmon N, Drewnowski A. Does social class predict diet quality? Am J Clin Nutr. 2008;87(5):1107–17.

    Article  CAS  PubMed  Google Scholar 

  63. Tang WH, Wang Z, Levison BS, Koeth RA, Britt EB, Fu X, et al. Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med. 2013;368(17):1575–84.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  64. Munzel T, Hahad O, Kuntic M, Keaney JF, Deanfield JE, Daiber A. Effects of tobacco cigarettes, e-cigarettes, and waterpipe smoking on endothelial function and clinical outcomes. Eur Heart J. 2020;41(41):4057–70.

    Article  PubMed  PubMed Central  Google Scholar 

  65. Messner B, Bernhard D. Smoking and cardiovascular disease: mechanisms of endothelial dysfunction and early atherogenesis. Arterioscler Thromb Vasc Biol. 2014;34(3):509–15.

    Article  CAS  PubMed  Google Scholar 

  66. Lavie CJ, Ozemek C, Carbone S, Katzmarzyk PT, Blair SN. Sedentary behavior, exercise, and cardiovascular health. Circ Res. 2019;124(5):799–815.

    Article  CAS  PubMed  Google Scholar 

  67. Borghi C, Agabiti-Rosei E, Johnson RJ, Kielstein JT, Lurbe E, Mancia G, et al. Hyperuricaemia and gout in cardiovascular, metabolic and kidney disease. Eur J Intern Med. 2020;80:1–11.

    Article  CAS  PubMed  Google Scholar 

  68. Albano GD, Gagliardo RP, Montalbano AM, Profita M. Overview of the mechanisms of oxidative stress: impact in inflammation of the airway diseases. Antioxidants. 2022;11(11):2237.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  69. Corbi G, Bianco A, Turchiarelli V, Cellurale M, Fatica F, Daniele A, et al. Potential mechanisms linking atherosclerosis and increased cardiovascular risk in COPD: focus on Sirtuins. Int J Mol Sci. 2013;14(6):12696–713.

    Article  PubMed  PubMed Central  Google Scholar 

  70. Duan M, Zhao X, Li S, Miao G, Bai L, Zhang Q, et al. Metabolic score for insulin resistance (METS-IR) predicts all-cause and cardiovascular mortality in the general population: evidence from NHANES 2001–2018. Cardiovasc Diabetol. 2024;23(1):243.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  71. Dang K, Wang X, Hu J, Zhang Y, Cheng L, Qi X, et al. The association between triglyceride-glucose index and its combination with obesity indicators and cardiovascular disease: NHANES 2003–2018. Cardiovasc Diabetol. 2024;23(1):8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We gratefully acknowledge the participants and investigators involved in the original genome-wide association studies and the developers of the GWAS Catalog (https://www.ebi.ac.uk/gwas/home).

Funding

This study was supported by the National Science and Technology Innovation 2030, Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2024ZD0524300, 2024ZD0524301), and National Natural Science Foundation of China (No. 72342017). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author information

Authors and Affiliations

  1. School of Public Health, Tianjin Medical University, Tianjin, 300070, China

    Dun Li, Lihui Zhou, Yujian Li, Xinyu Zhang, Weili Xu & Yaogang Wang

  2. School of Nursing, Tianjin Medical University, Tianjin, 300070, China

    Jing Lin & Li Sun

  3. School of Basic Medical Sciences, Tianjin Medical University, Tianjin, 300070, China

    Hongxi Yang

  4. School of Integrative Medicine, Public Health Science and Engineering College, Tianjin University of Traditional Chinese Medicine, Tianjin, 301617, China

    Zhe Xu & Yaogang Wang

  5. Aging Research Center, Department of Neurobiology, Health Care Sciences and Society Karolinska Institutet, Stockholm University, 17165, Stockholm, Sweden

    Weili Xu

  6. National Institute of Health Data Science at Peking University, Peking University, Beijing, 100191, China

    Yaogang Wang

Authors
  1. Dun Li
    View author publications

    Search author on:PubMed Google Scholar

  2. Jing Lin
    View author publications

    Search author on:PubMed Google Scholar

  3. Hongxi Yang
    View author publications

    Search author on:PubMed Google Scholar

  4. Lihui Zhou
    View author publications

    Search author on:PubMed Google Scholar

  5. Yujian Li
    View author publications

    Search author on:PubMed Google Scholar

  6. Zhe Xu
    View author publications

    Search author on:PubMed Google Scholar

  7. Li Sun
    View author publications

    Search author on:PubMed Google Scholar

  8. Xinyu Zhang
    View author publications

    Search author on:PubMed Google Scholar

  9. Weili Xu
    View author publications

    Search author on:PubMed Google Scholar

  10. Yaogang Wang
    View author publications

    Search author on:PubMed Google Scholar

Contributions

Y.W. conceptualized and designed the study, D.L., J.L., H.Y., L.Z., Y.L., and Z.X. performed the data collection and analysis, and D.L. drafted the manuscript. D.L., J.L., H.Y., L.Z., L.S. and X.Z. contributed to the analysis and interpretation of data. Y.W., W.X., L.S., and X.Z. contributed to the revision of the manuscript and approved the final draft. Y.W. obtained funding for the study. Y.W. and W.X. were involved in study supervision. All authors contributed to the intellectual content and critical revisions to the drafts of the paper and approved the final version.

Corresponding author

Correspondence to Yaogang Wang.

Ethics declarations

Ethics approval and consent to participate

All GWAS data involved are de-identified and with public availability, and the ethical approval and consent to participate were obtained by each GWAS, which were described in the original publications. The STROBE-MR (Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization) checklist was provided as STROBE-MR-checklist.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Li, D., Lin, J., Yang, H. et al. Causal association of modifiable factors with cardiometabolic multimorbidity: an exposome-wide Mendelian randomization investigation. Cardiovasc Diabetol 24, 241 (2025). https://doi.org/10.1186/s12933-025-02790-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12933-025-02790-w

Keywords

Cardiovascular Diabetology

ISSN: 1475-2840