Mendelian randomization study supports the causal effects of air pollution on longevity via multiple age-related diseases

Air pollution, particularly fine particulate matter (PM2.5), is a major global health concern and has been associated with increased risk of chronic diseases and premature mortality. Observational studies suggest improved life expectancy with reductions in PM2.5, but causal inference is complicated by numerous confounders such as socioeconomic status, education, income, occupation, noise pollution, and access to green space. Genetic factors influence both susceptibility to air pollution and longevity, and the heritability of lifespan is substantial. In the absence of randomized controlled trials, the study employs Mendelian randomization (MR) to test the causal effect of genetically predicted PM2.5 concentration on longevity and to identify mediating pathways through age-related diseases and risk factors. The primary research questions are: (1) Does genetically predicted PM2.5 causally affect longevity? (2) Do cardiometabolic, cardiovascular, respiratory, autoimmune, or neurodegenerative traits mediate the effect of PM2.5 on longevity? The study also adjusts for potential confounding by other air pollutants and smoking using multivariable MR.

Design: Two-sample Mendelian randomization (MR) framework using large-scale GWAS summary statistics to estimate causal effects of genetically predicted PM2.5 on longevity and potential mediators, with univariable MR, multivariable MR (MVMR), and two-step MR mediation analysis. Data sources: PM2.5 exposure instruments from UK Biobank within ESCAPE, modeled by land use regression for 2010 (N≈423,796; mean 9.99 μg/m3, SD 1.06). Longevity outcomes from a GWAS meta-analysis across 20 cohorts of European, East Asian, and African American ancestry, defining cases as survival above the 90th (11,262 cases) or 99th percentile (3,484 cases) and controls as ≤60th percentile (25,483 controls). Potential confounders included nitrogen dioxide and nitrogen oxides air pollution (UK Biobank) and cigarettes per day (GSCAN; N=337,334). Potential mediators spanned cardiometabolic risk factors, cardiovascular/cerebrovascular diseases, lung function, autoimmune diseases, and neurodegenerative diseases, with GWAS datasets from GIANT, ICBP, DIAMANTE, CMDKP, HERMES, AFGen/Broad AF, UK Biobank, SpiroMeta, ILCCO, IIBDGC, IMSGC, IPDGC, IGAP, and others. All included participants for mediator traits were of predominantly European ancestry. Instrument selection: Independent genome-wide significant SNPs (P<5×10^-8) associated with exposures were selected, clumped at r2<0.001 using 1000 Genomes European reference. Palindromic/incompatible alleles were removed during harmonization. F-statistics were calculated and instruments with F<10 were considered weak and excluded. Univariable MR: Primary analysis by inverse variance weighted (IVW) method combining SNP-specific Wald ratios. Weighted median estimators used as sensitivity. Sensitivity analyses: MR-Egger regression to assess directional pleiotropy; MR-PRESSO to detect and remove outlier instruments and re-estimate IVW effects. Conservative analyses performed to evaluate robustness. Multivariable MR: MVMR IVW modeled PM2.5 jointly with nitrogen dioxide, nitrogen oxides, and smoking to estimate direct effects on mediators and longevity, allowing SNPs to be associated with multiple exposures. Mediation analysis: Two-step MR. Step 1 estimated causal effects of PM2.5 on each potential mediator using PM2.5 instruments. Step 2 estimated effects of each mediator on longevity using mediator-specific instruments. Indirect effects computed via product-of-coefficients; standard errors via delta method. Mediation proportion calculated as indirect/total effect. Multiple testing correction for 36 mediators used P<0.0014 as significance threshold; general statistical significance also reported at P<0.05. MR implemented with TwoSampleMR (v0.5.6) and MRPRESSO (v1.0).

- Univariable MR showed no significant direct causal effect of genetically predicted PM2.5 on longevity: 90th survival percentile OR=0.56 (95% CI 0.12–2.63; P=0.47); 99th percentile OR=0.32 (95% CI 0.03–3.61; P=0.36). Sensitivity analyses were consistent. - Genetically predicted PM2.5 was causally associated with several potential mediators: - Hypercholesterolaemia: OR=1.07 per 1.06 μg/m3 increase (95% CI 1.01–1.15; P=0.03). - Hypertension: OR=1.08 (95% CI 1.02–1.16; P=0.014). - Angina pectoris: OR=1.05 (95% CI 1.01–1.08; P=0.0084). - Asthma: OR=1.01 (95% CI 1.00–1.02; P=0.011). - Hypothyroidism: OR=1.10 (95% CI 1.06–1.13; P=2.93×10^-8). - Alzheimer’s disease (AD): OR=1.62 (95% CI 1.01–2.60; P=0.044). - Diastolic blood pressure (DBP) was reduced: OR=0.07 (95% CI 0.02–0.27; P=9.24×10^-5) in the reported analysis. - MR-PRESSO sensitivity (after removing pleiotropic SNPs) suggested additional associations: increased risk of type 2 diabetes (OR=1.86, 95% CI 1.20–2.87; P=0.039) and atrial fibrillation (OR=1.06, 95% CI 1.01–1.10; P=0.046). - Multivariable MR (adjusting for nitrogen dioxide, nitrogen oxides, and smoking) identified stronger direct effects of PM2.5 on: - Angina pectoris: OR=1.14 (95% CI 1.05–1.23; P=0.0012). - Hypercholesterolaemia: OR=1.43 (95% CI 1.23–1.67; P=5.09×10^-6). - Hypothyroidism: OR=1.12 (95% CI 1.03–1.23; P=0.012). No direct causal effect of air pollution on longevity was observed in MVMR. - Mediation (two-step MR) indicated indirect effects of PM2.5 on longevity via specific mediators: - DBP: mediated proportion ≈31.5% (90th percentile) and 28.4% (99th percentile); mediation effect β=0.18 (P=0.00029) and 0.21 (P=0.0006), respectively. - Hypertension: mediated proportion 70.9% (90th; P=0.018) and 40.9% (99th; P=0.023). - Angina pectoris: small but significant mediation; mediated proportion 2.5% (90th; P=0.020) and 1.5% (99th; P=0.026). - Hypercholesterolaemia: significant mediation in the 90th percentile subgroup; mediated proportion reported up to 100% (P=0.043) for 90th, suggestive in 99th (P≈0.051). - Alzheimer’s disease: mediation significant in 90th percentile (P≈0.05) with mediated proportion ≈24.7%; suggestive in 99th percentile (P≈0.053). Overall, while PM2.5 did not show a direct causal effect on longevity, it influenced longevity indirectly through multiple age-related diseases and risk factors.

The study addressed whether PM2.5 causally affects longevity and through which pathways. Using MR to mitigate confounding and reverse causality, no direct effect of genetically predicted PM2.5 on longevity was detected; however, robust associations were found between PM2.5 and several cardiometabolic, cardiovascular, endocrine, and neurodegenerative traits. Multivariable MR confirmed independent PM2.5 effects on angina pectoris, hypercholesterolaemia, and hypothyroidism after adjusting for other air pollutants and smoking, reinforcing specificity of PM2.5’s role. Two-step MR mediation demonstrated that PM2.5 may reduce the probability of exceptional longevity indirectly via elevated blood pressure and hypertension, coronary syndromes (angina pectoris), dyslipidemia (hypercholesterolaemia), and Alzheimer’s disease. These findings align with prior observational evidence linking PM2.5 with cardiometabolic disease and dementia and support biological mechanisms involving oxidative stress, inflammation, vascular dysfunction, and potential neurotoxicity via particle translocation into circulation. The results collectively suggest that mitigating PM2.5 exposure could improve life expectancy by preventing key intermediary diseases rather than exerting a direct effect on lifespan.

This MR study indicates that PM2.5 exposure does not have a detectable direct causal effect on longevity but exerts indirect detrimental effects via multiple mediators, notably DBP/hypertension, angina pectoris, hypercholesterolaemia, hypothyroidism, and Alzheimer’s disease. Public health measures and policies aimed at reducing ambient PM2.5 concentrations are likely to confer longevity benefits by lowering the incidence of these mediating conditions. Future research should expand sample sizes, include diverse ancestries, improve instrument strength for additional pollutants (e.g., PM10, PM2.5–10), refine mediation MR methodologies, and explore time-varying exposures and gene–environment interactions to better quantify intervention impacts on lifespan.

- Potential sample overlap between UK Biobank-derived mediators and the air pollution cohort could bias estimates; alternative data sources were limited. - Lack of genome-wide significant instruments for PM10 and PM2.5–10 precluded MR analyses for these exposures. - Differences between MR and observational/RCT estimates mean MR results should not be directly interpreted as intervention effect sizes; MR estimates may be larger than observational ones. - MR-PRESSO removal of pleiotropic instruments can reduce power or produce overly precise estimates; associations for T2D and AF are considered suggestive. - Power limitations common to gene–environment interaction research may have hindered detection of small direct effects; the genetically instrumented PM2.5 variation likely represents only part of true exposure variability. - Predominantly European ancestry samples may limit generalizability to other populations.