Inflammation-Mediated Causal Effects of MRI-Defined Body-Composition Obesity Axes on Chronic Obstructive Pulmonary Disease, Asthma, and Allergic Rhinitis: A Mendelian Randomization Study

Abstract

Background

Obesity is heterogeneous; conventional anthropometric measures combine fat distribution, ectopic fat deposition, and muscle volume, limiting refined causal inference for chronic airway diseases. We used four magnetic resonance imaging (MRI)-defined obesity axes to assess their causal effects on chronic obstructive pulmonary disease (COPD), asthma, and allergic rhinitis (AR), and to evaluate mediation by circulating inflammatory proteins.

Methods

Genetic instruments were selected for four MRI-defined obesity axes. Two-sample Mendelian randomization (MR) used inverse-variance weighting as the primary method, with complementary sensitivity analyses and false discovery rate correction. Non-overlapping FinnGen outcome data were used for discovery, with replication in the Global Biobank Meta-analysis Initiative for COPD and asthma and the Million Veteran Program for AR, followed by meta-analysis. Two-step MR assessed inflammatory protein mediation.

Results

Genetically predicted general obesity axis was associated with increased risks of COPD (OR = 1.125, 95% CI: 1.012–1.250), asthma (OR = 1.533, 95% CI: 1.426–1.647), and AR (OR = 1.624, 95% CI: 1.328–1.987). The lower-body fat axis was associated with higher risks of COPD (OR = 1.192, 95% CI: 1.088–1.306) and asthma (OR = 1.135, 95% CI: 1.052–1.224). No causal evidence was observed for the other obesity axes. Replication and meta-analysis largely supported these findings, except for general obesity and AR, which showed borderline pooled evidence (P = 0.055). Mediation analysis identified 17 candidate inflammatory pathways, of which 10 were significant. FLT3L mediated 18.17% of the effect of general obesity on COPD, and CXCL5 mediated 3.78% of its effect on asthma. CD40L receptor and LIFR mediated 5.59% and 12.65% of the lower-body fat–asthma effect, respectively, whereas CD40L receptor, CXCL10, IL-12B, OPG, and TNFRSF9 showed suppressive mediation in selected pathways.

Conclusion

MRI-defined obesity axes provided genetic evidence supporting potential causal associations with airway diseases through inflammatory mediation pathways.

Introduction

Chronic obstructive pulmonary disease (COPD), asthma, and allergic rhinitis (AR) are among the most common chronic airway diseases and are associated with substantial disease burden and frequent comorbidity. Within the classic “one airway, one disease” framework, AR and asthma are considered distinct clinical manifestations of a continuous inflammatory process involving the upper and lower airways, with marked clinical overlap.¹ Approximately 30–50% of individuals with AR subsequently develop asthma, whereas 70–90% of individuals with asthma have concomitant AR.^2,3 In contrast to the allergic inflammatory features of AR and asthma, COPD is commonly linked to long-term inhalation of harmful particles or gases, with smoking, air pollution, and occupational exposure recognized as important modifiable risk factors.⁴ However, COPD and upper airway disease are not completely separate conditions. Approximately 35% of individuals with COPD have comorbid AR, and the prevalence of chronic nasal symptoms in COPD populations may reach 88%. These symptoms are associated with impaired quality of life, increased risks of acute exacerbation and hospitalization, and reduced lung function.^5,6 Therefore, AR, asthma, and COPD may not represent three isolated diseases, but rather distinct phenotypes within an allergic–obstructive airway continuum. Nevertheless, the shared upstream drivers and biological mechanisms that connect this continuum remain insufficiently understood. Identifying modifiable causal exposures and potential mediating pathways is therefore important for risk stratification, early prevention, and precision intervention.

Obesity is a potential upstream driver that warrants particular attention. According to the World Health Organization, one in eight people worldwide were living with obesity in 2022, and the global prevalence of adult obesity has more than doubled since 1990.⁷ Obesity may contribute to the onset and progression of airway diseases through multiple pathways, including mechanical loading, metabolic dysregulation, and chronic low-grade inflammation.^8,9 Adipose tissue also functions as an active endocrine and immune organ, and excess adiposity can alter the secretion of adipokines and pro-inflammatory mediators, including leptin, adiponectin, interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α). These obesity-related inflammatory signals may promote systemic and airway inflammation and contribute to altered airway immune responses and hyperresponsiveness, providing biological plausibility for inflammation as an intermediate pathway linking obesity to chronic airway disease.^10,11 However, obesity is not a single biological entity. Traditional anthropometric measures, such as body mass index, waist circumference, and waist-to-hip ratio, are convenient for clinical and epidemiological use, but they cannot adequately distinguish key differences in body composition, including visceral fat, subcutaneous fat, ectopic fat, and muscle volume.^12,13 As a result, these conventional measures may not fully capture the internal heterogeneity of obesity and may be insufficient for etiological stratification and risk identification in precision medicine. In addition, conventional observational studies are vulnerable to confounding, reverse causation, and measurement error, leaving uncertainty regarding the true causal relationships between obesity and respiratory diseases and the inflammatory mechanisms that may underlie these associations.

Mendelian randomization (MR) uses genetic variants that are robustly associated with an exposure as instrumental variables (IVs). When the core assumptions are satisfied, MR can reduce bias from confounding and reverse causation to some extent, thereby providing stronger evidence for causal inference in exposure–outcome relationships.¹⁴ Previous MR studies based on traditional obesity phenotypes, including body mass index, waist circumference, body fat distribution, and visceral adiposity, have suggested that obesity-related traits may causally influence the risks of asthma, impaired lung function, COPD, AR, and other allergic diseases.^15–19 However, most of these studies relied on single conventional obesity indicators and therefore could not fully characterize the biological heterogeneity of obesity. Recently, a study integrating UK Biobank imaging and multi-omics data derived four independent obesity axes from 24 MRI-derived phenotypes of fat distribution and muscle volume: a general obesity axis, reflecting widespread accumulation of visceral, subcutaneous, and ectopic fat; a muscle-dominant axis, reflecting greater muscle volume; a peripheral fat axis, reflecting relatively higher abdominal and thigh subcutaneous fat; and a lower-body fat axis, characterized by greater lower-body subcutaneous fat and lower ectopic fat.²⁰ Compared with traditional obesity measures, these four axes provide a more refined and biologically interpretable framework for characterizing heterogeneity in body composition, distinguishing differences in overall adiposity, muscle predominance, peripheral subcutaneous fat storage, and lower-body versus ectopic fat distribution that may be obscured by conventional anthropometric measures. Importantly, these axes represent continuous and potentially overlapping dimensions of body composition rather than mutually exclusive clinical obesity categories.²⁰ On this basis, the present study applied two-sample MR to systematically evaluate the causal effects of the four obesity axes on COPD, asthma, and AR. We further conducted two-step mediation MR to investigate the biological pathway linking obesity axes, inflammation, and respiratory diseases, thereby providing new genetic evidence for inflammatory mechanisms underlying obesity-related susceptibility to airway disease and potentially informing phenotype-based risk stratification and future precision prevention or targeted intervention strategies.

Methods

Study Design

This study was reported in accordance with the STROBE-MR guidelines to ensure transparency and reproducibility.²¹ Genetic association data were obtained from relevant genome-wide association studies (GWASs), and the overall study workflow is shown in Figure 1. First, we used the four obesity axes as exposures and selected associated genetic variants as IVs to assess the causal effects of genetically predicted obesity-axis traits on COPD, asthma, and AR. The causal findings identified in the discovery analysis were subsequently evaluated in independent replication datasets, followed by targeted meta-analyses combining the discovery and replication estimates to assess the robustness of the results. Associations that showed evidence of causality were subsequently incorporated into a two-step mediation MR framework to investigate the obesity axis–inflammation–respiratory disease pathway. All data used in this study were derived from publicly available summary-level GWAS datasets. This study was based exclusively on publicly available and anonymized summary-level GWAS data, with no participant contact or intervention. Therefore, additional ethical approval and clinical trial registration were not required, in accordance with Article 32(1)–(2) of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects, effective in China from February 18, 2023.

Figure 1.

Study workflow. Overview of the two-sample Mendelian randomization (MR), replication, meta-analysis, and two-step mediation MR evaluating four MRI-derived obesity axes, respiratory outcomes, and circulating inflammatory proteins.

Abbreviations: AR, allergic rhinitis; COPD, chronic obstructive pulmonary disease; MRI, magnetic resonance imaging.

Assumptions of Mendelian Randomization

Genetic variants selected as IVs must satisfy the three core MR assumptions—relevance, independence, and exclusion restriction—to support valid causal inference. The relevance assumption requires that genetic variants are robustly associated with the exposure. The independence assumption requires that genetic variants are not associated with confounders of the exposure–outcome relationship. The exclusion restriction assumption requires that genetic variants influence the outcome only through the exposure and not through alternative pathways. The relevance assumption can be assessed directly by applying a genome-wide significance threshold and calculating F statistics. Although the independence and exclusion restriction assumptions cannot be tested directly, their plausibility can be evaluated through complementary quality-control and sensitivity analyses, including Steiger testing, linkage disequilibrium clumping, and pleiotropy assessment. In this study, the selection of genetic instruments strictly followed these assumptions, and the detailed procedures are described below.

Data Source and Genetic Instrument Selection for Obesity Axes

Genetic association data for the four obesity axes were obtained from the GWAS conducted by Odoemelam et al.²⁰ In that study, 24 MRI-derived measures of fat distribution and muscle volume were extracted from the UK Biobank imaging cohort (n = 33,122). Sex-stratified principal component analysis was then used to identify four mutually independent obesity axes that did not show clear sex-specific effects. GWAS analyses were subsequently performed in 25,637 White British participants who passed imaging and genotyping quality control, using the BOLT-LMM model. Additional details are provided in the supplementary material.

Using sex-combined summary-level GWAS data, we selected genetic instruments under genome-wide significance and independence criteria. Specifically, single nucleotide polymorphism (SNP) were selected at P < 5×10⁻⁸ and were clumped to ensure independence at r² < 0.001 within a 10-Mb window. Linkage disequilibrium (LD) clumping was performed using the European ancestry reference panel from Phase 3 of the 1000 Genomes Project. The selected SNPs were then matched with the outcome GWAS datasets, and effect alleles were harmonized across exposure and outcome datasets. Ambiguous palindromic SNPs with intermediate allele frequencies were removed during harmonization. To avoid potential uncertainty introduced by proxy matching, proxy SNPs were not used to replace unavailable variants. The overall instrument-selection procedure was consistent with previous MR studies, and additional quality-control procedures are described in the sensitivity analysis section. Because the original GWAS applied inverse-normal transformation, MR estimates represent the effect per 1-standard deviation (SD) increase in the Z-score scale of each obesity axis. For binary outcomes, causal estimates are reported as odds ratios (ORs).

Data Sources for Respiratory Diseases

To minimize potential sample overlap with the UK Biobank exposure GWAS, we used FinnGen R12 as the primary outcome dataset for COPD, asthma, and AR in the discovery stage.²⁰ FinnGen links genotype data from approximately 500,000 Finnish biobank participants with nationwide health registry data. Disease endpoints were defined using ICD diagnosis codes, and GWAS analyses were adjusted for age, sex, the first 10 genetic principal components, genotyping chip, and batch effects.²² In the discovery stage, COPD, asthma, and AR included 24,138 cases and 409,070 controls, 59,100 cases and 259,839 controls, and 15,569 cases and 474,650 controls, respectively. In the replication stage, the replication datasets for COPD and asthma were obtained from the Global Biobank Meta-analysis Initiative (GBMI), which currently provides the largest available GWAS sample sizes for these outcomes, including 58,559 cases and 937,358 controls for COPD and 121,940 cases and 1,254,131 controls for asthma.²³ For AR, the replication dataset was obtained from the Million Veteran Program (MVP), including 92,310 cases and 311,377 controls. Detailed descriptions of all outcome datasets are provided in the supplementary material.

Data Sources for Inflammatory Mediators

GWAS summary statistics for protein quantitative trait loci (pQTL) of 91 circulating inflammatory proteins were obtained from Zhao et al²⁴ This study integrated data from 11 cohorts of European ancestry, comprising 14,824 participants. Plasma inflammatory proteins were measured using the Olink Target 96 Inflammation panel. Each cohort performed association analyses using linear regression models, with inverse-rank normalization applied to protein levels and adjustment for age, sex, genetic principal components, and cohort-specific covariates. Fixed-effect inverse-variance weighted meta-analysis was then conducted using METAL. Quality control procedures accounted for minor allele frequency, Hardy–Weinberg equilibrium, imputation quality, cross-cohort coverage, and heterogeneity. Additional details are provided in the Supplementary Methods.

Statistical Analysis

All analyses were performed using R version 4.2.3. The main R packages included metafor, TwoSampleMR, MVMR, MRcML, and BWMR. The primary MR methods were the Wald ratio and inverse-variance weighted (IVW) method with the default random-effects model.²⁵ The Wald ratio was used when only one valid instrumental SNP was available, providing a causal estimate based on the ratio of the SNP–outcome association to the SNP–exposure association. When more than one independent SNP was available, IVW was used to combine SNP-specific Wald ratio estimates through inverse-variance weighted regression. To evaluate the robustness of the primary findings against potential bias from heterogeneity, horizontal pleiotropy, and weak instruments, we additionally applied MR-Egger, weighted median, penalized weighted median, contamination mixture, robust adjusted profile score (RAPS), debiased IVW (dIVW), constrained maximum likelihood (cML), and Bayesian weighted MR (BWMR). Multiple testing was controlled using the false discovery rate (FDR). Robust evidence of causality was defined as an FDR-adjusted P value below 0.05 for the primary MR estimate, consistent effect directions in more than half of the complementary MR methods, concordant results in sensitivity analyses, and no evidence of substantial heterogeneity or horizontal pleiotropy. Associations with nominal statistical significance but not surviving FDR correction were considered suggestive evidence when P < 0.05 and P_FDR > 0.05.

For genetic instrument quality control, we calculated the variance explained by each SNP (R²) and performed Steiger filtering. This procedure systematically removed SNPs that explained more variance in the outcome than in the exposure, based on R² estimates and the corresponding sample sizes of the exposure and outcome datasets, thereby reducing the likelihood of reverse causation and violations of the MR assumptions. We also calculated the F statistic for each SNP and removed weak instruments with F < 10. The formulas for R² and F statistics have been reported previously.^26,27 In addition, MR-PRESSO was used to identify and remove outlier SNPs to reduce potential bias from horizontal pleiotropy. Horizontal pleiotropy was assessed using the MR-Egger intercept test²⁸ and the MR-PRESSO global test.²⁹ Heterogeneity was evaluated using the Cochran Q statistic.³⁰ Statistical power was assessed using the online mRnd tool (https://shiny.cnsgenomics.com/mRnd/).

In the two-step mediation MR analysis, we first estimated the causal effects of genetically predicted obesity axes on candidate inflammatory proteins and then assessed the causal effects of these proteins on the corresponding respiratory disease outcomes. The indirect effect was calculated as the product of the exposure–mediator effect and the mediator–outcome effect. The total effect was represented by the MR estimate for the association between each obesity axis and each respiratory disease outcome. The direct effect was calculated by subtracting the indirect effect from the total effect. The mediation proportion was defined as the ratio of the indirect effect to the total effect, and the 95% confidence interval was estimated using the delta method.³¹

Results

Genetic Instrument Selection and Evaluation

We first identified 41 genome-wide significant genetic loci for the four obesity axes, including 2 loci for the general obesity axis, 9 loci for the muscle-dominant axis, 15 loci for the peripheral fat axis, and 15 loci for the lower-body fat axis (Table S1). After harmonization with the outcome datasets and subsequent quality-control procedures, the number of genetic instruments included in each MR analysis ranged from 2 to 12. The F statistic for each IV was greater than 10, supporting adequate instrument strength. All retained instruments passed Steiger filtering, with directionality tests indicating TRUE. MR-PRESSO was applied systematically to detect and remove outlier variants when present. Overall, these procedures supported the validity of the genetic instruments with respect to the three core MR assumptions, and the discovery-stage design minimized potential bias from sample overlap. Detailed information for each IV, including alleles, SNP–exposure and SNP–outcome associations, F statistics, and Steiger filtering results, is provided in Table S2.

Causal Associations Between Obesity Axes and Respiratory Diseases

Figure 2a presents the causal estimates, and Table S3 provides detailed results from all MR methods. In the discovery-stage MR analysis, we identified five associations with evidence of causality, including four FDR-significant associations and one suggestive association. Specifically, each 1-SD increase in the genetically predicted general obesity axis was associated with an increased risk of COPD, with suggestive evidence only (OR = 1.125, 95% CI: 1.012–1.250, P = 0.029, P_FDR = 0.070). The general obesity axis was also causally associated with higher risks of AR (OR = 1.624, 95% CI: 1.328–1.987, P = 2.30×10⁻⁶, P_FDR = 1.38×10⁻⁵) and asthma (OR = 1.533, 95% CI: 1.426–1.647, P = 2.53×10⁻³¹, P_FDR = 3.04×10⁻³⁰). For these FDR-significant associations, all complementary MR methods produced effect estimates in the same direction as the IVW estimates (5/5).

Figure 2.

Mendelian randomization estimates of obesity axes and respiratory outcomes. (a) Forest plots of associations between four obesity axes and AR, asthma, and COPD. Points and horizontal lines represent odds ratios (ORs) and 95% confidence intervals (CIs), respectively. (b–f) Scatter plots for associations supported in the primary analysis; points represent genetic instruments and lines represent MR estimates. (g) Pooled estimates from the discovery and replication analyses. Filled squares indicate P < 0.05, open squares indicate P ≥ 0.05, and daggers indicate 0.05 ≤ P < 0.10. The dashed line denotes OR = 1.

Abbreviations: AR, allergic rhinitis; COPD, chronic obstructive pulmonary disease; MR, Mendelian randomization.

In addition, each 1-SD increase in the genetically predicted lower-body fat axis was significantly associated with higher risks of COPD (OR = 1.192, 95% CI: 1.088–1.306, P = 1.60×10⁻⁴, P_FDR = 6.40×10⁻⁴) and asthma (OR = 1.135, 95% CI: 1.052–1.224, P = 1.12×10⁻³, P_FDR = 3.36×10⁻³). Except for MR-Egger, all complementary MR methods yielded directionally consistent evidence with the IVW estimates (7/8). Scatter plots further illustrated the consistency of the causal effect directions across MR methods (Figure 2b–f), with full scatter plots provided in Supplementary Figure S1. Statistical power was 100% for all associations with positive evidence, except for the association between the general obesity axis and COPD, for which the estimated power was 70%. No evidence of causal associations was observed for the remaining exposure–outcome pairs (P > 0.05).

In sensitivity analyses, the MR-Egger regression detected evidence of horizontal pleiotropy only for the association between the peripheral fat axis and COPD (Table S4; MR-Egger intercept = −0.041, P = 0.033). The MR-PRESSO global test did not identify evidence of horizontal pleiotropy in any analysis (Table S5), and the Cochran Q test did not detect evidence of heterogeneity (Table S6). In the independent replication datasets, three associations were replicated with consistent effect directions and statistical significance (Table S7). In the subsequent combined meta-analysis, four associations remained statistically significant (Figure 2g and Table 1). Notably, the association between the general obesity axis and AR did not reach conventional statistical significance in the meta-analysis but showed a consistent direction of effect and borderline evidence of association (P = 0.055). Therefore, this association was interpreted as suggestive evidence.

Table 1.

Discovery, Replication, and Pooled Meta-Analysis Results for Obesity Axes and Respiratory Outcomes

Exposure	Outcome	Discovery Dataset	Replication Dataset	Meta-Analysis
		OR (95% CI)	OR (95% CI)	I2(%)	Q_pval	Model	OR (95% CI)	P-value
General obesity	AR	1.624 (1.328, 1.987)	1.163 (0.941, 1.437)	80.09	0.025	Random	1.376 (0.992, 1.909)	0.055
General obesity	Asthma	1.533 (1.426, 1.647)	1.276 (1.087, 1.497)	76.33	0.040	Random	1.419 (1.187, 1.695)	1.17E-04
General obesity	COPD	1.125 (1.012, 1.250)	1.161 (1.111, 1.214)	0.00	0.589	Fixed	1.156 (1.109, 1.204)	4.51E-12
Lower-body fat	AR	0.955 (0.832, 1.096)	1.006 (0.954, 1.061)	0.00	0.491	Fixed	0.999 (0.951, 1.050)	0.973
Lower-body fat	Asthma	1.135 (1.052, 1.224)	1.057 (1.015, 1.101)	61.84	0.105	Random	1.087 (1.016, 1.164)	0.016
Lower-body fat	COPD	1.192 (1.088, 1.306)	1.075 (0.978, 1.183)	57.15	0.127	Random	1.133 (1.025, 1.254)	0.015
Muscle dominant	AR	1.069 (0.861, 1.328)	1.065 (0.973, 1.165)	0.00	0.973	Fixed	1.065 (0.981, 1.157)	0.135
Muscle dominant	Asthma	1.022 (0.906, 1.152)	1.114 (0.983, 1.263)	0.00	0.328	Fixed	1.065 (0.977, 1.162)	0.153
Muscle dominant	COPD	0.935 (0.845, 1.035)	1.085 (0.988, 1.191)	77.65	0.034	Random	1.009 (0.872, 1.167)	0.909
Peripheral fat	AR	1.023 (0.888, 1.178)	0.997 (0.933, 1.066)	0.00	0.748	Fixed	1.002 (0.943, 1.064)	0.954
Peripheral fat	Asthma	1.043 (0.966, 1.126)	1.010 (0.922, 1.107)	0.00	0.600	Fixed	1.029 (0.971, 1.092)	0.334
Peripheral fat	COPD	1.019 (0.913, 1.138)	0.969 (0.921, 1.020)	0.00	0.420	Fixed	0.978 (0.934, 1.025)	0.350

Two-Step Mediation Analysis

In the two-step mediation MR analysis, we used an intersection-based screening strategy to identify potential inflammatory mediators. First, for obesity axes that showed positive causal evidence in the primary analysis, we evaluated their genetic causal effects on 91 circulating inflammatory proteins. A total of 40 significant associations were observed, including 7 associations for the general obesity axis and 33 associations for the lower-body fat axis (Table S8). Next, using inflammatory proteins as exposures, we applied the same instrumental-variable selection and quality-control procedures as those used in the primary analysis to further assess their causal effects on COPD, AR, and asthma. Because some inflammatory proteins lacked valid instrumental variables that met the selection criteria, 73 inflammatory proteins were ultimately included in the inflammatory protein–disease MR analysis. This analysis identified 42 significant associations, including 8 for COPD, 12 for AR, and 22 for asthma (Table S9). An inflammatory protein was included in the formal mediation analysis for a given obesity axis–disease pair only when all three paths reached statistical significance: the obesity axis–inflammatory protein path, the inflammatory protein–disease path, and the obesity axis–disease path. Based on this intersection-based strategy, 17 candidate mediation paths were identified for further analysis (Figure 3 and Table 2).

Figure 3.

Intersection-based screening of inflammatory mediators. Screening of 40 significant obesity axis–protein and 42 protein–outcome associations identified 17 candidate mediation paths, of which 10 showed significant mediation. Colors indicate the corresponding obesity axis; darker, thicker lines denote significant mediation paths (P < 0.05), whereas lighter lines denote candidate intersections that were not significant in the final mediation analysis.

Abbreviation: COPD, chronic obstructive pulmonary disease.

Table 2.

Summary of Two-Step Mediation Mendelian Randomization Analysis Results

Exposure Axis	Mediator Phenotype	Outcome	Confidence Interval of Total Effect (95% CI)	Confidence Interval of Direct Effect (95% CI)	Confidence Interval of Mediation Effect (95% CI)	Confidence Interval of the Proportion of Mediation Effect	Pval
General obesity	CD40L receptor	COPD	0.118 (0.012, 0.223)	0.140 (0.033, 0.247)	−0.023 (−0.040, −0.006)	−19.23% (−33.75%, −4.71%)	0.009
General obesity	FLT3L	COPD	0.118 (0.012, 0.223)	0.096 (0.011, 0.204)	0.021 (0.003, 0.040)	18.17% (2.18%, 34.16%)	0.025
Lower-body fat	CD40L receptor	COPD	0.176 (0.085, 0.267)	0.169 (0.078, 0.261)	0.006 (−0.002, 0.014)	3.63% (−0.91%, 8.17%)	0.102
Lower-body fat	FLT3L	COPD	0.176 (0.085, 0.267)	0.184 (0.092, 0.276)	−0.008 (−0.018, 0.002)	−4.61% (−10.40%, 1.18%)	0.103
Lower-body fat	TGF-α	COPD	0.176 (0.085, 0.267)	0.202 (0.105, 0.298)	−0.026 (−0.058, 0.006)	−14.78% (−32.72%, 3.15%)	0.092
Lower-body fat	uPA	COPD	0.176 (0.085, 0.267)	0.161 (0.067, 0.254)	0.015 (−0.001, 0.031)	8.61% (−0.35%, 17.57%)	0.053
General obesity	CD40L receptor	Asthma	0.427 (0.355, 0.499)	0.452 (0.379, 0.525)	−0.025 (−0.036, −0.014)	−5.85% (−8.54%, −3.17%)	1.91E-05
General obesity	CXCL5	Asthma	0.427 (0.355, 0.499)	0.411 (0.338, 0.484)	0.016 (0.001, 0.031)	3.78% (0.34%, 7.22%)	0.030
Lower-body fat	CD40L receptor	Asthma	0.126 (0.050, 0.202)	0.119 (0.043, 0.196)	0.007 (0.000, 0.014)	5.59% (0.18%, 10.99%)	0.041
Lower-body fat	CD5	Asthma	0.126 (0.050, 0.202)	0.134 (0.058, 0.211)	−0.008 (−0.017, 0.001)	−6.38% (−13.43%, 0.67%)	0.068
Lower-body fat	CXCL10	Asthma	0.126 (0.050, 0.202)	0.147 (0.069, 0.224)	−0.020 (−0.036, −0.004)	−16.13% (−28.86%, −3.39%)	0.011
Lower-body fat	DNER	Asthma	0.126 (0.050, 0.202)	0.136 (0.059, 0.213)	−0.010 (−0.020, 0.001)	−7.53% (−16.08%, 1.03%)	0.074
Lower-body fat	FGF21	Asthma	0.126 (0.050, 0.202)	0.137 (0.060, 0.214)	−0.011 (−0.023, 0.001)	−8.72% (−18.14%, 0.69%)	0.061
Lower-body fat	IL-12B	Asthma	0.126 (0.050, 0.202)	0.131 (0.055, 0.208)	−0.005 (−0.010, 0.000)	−4.08% (−8.21%, 0.00%)	0.046
Lower-body fat	LIFR	Asthma	0.126 (0.050, 0.202)	0.110 (0.033, 0.188)	0.016 (0.000, 0.032)	12.65% (0.14%, 25.17%)	0.044
Lower-body fat	OPG	Asthma	0.126 (0.050, 0.202)	0.142 (0.064, 0.219)	−0.015 (−0.030, −0.001)	−12.17% (−23.77%, −0.58%)	0.036
Lower-body fat	TNFRSF9	Asthma	0.126 (0.050, 0.202)	0.151 (0.071, 0.231)	−0.025 (−0.050, 0.000)	−19.73% (−39.81%, 0.00%)	0.047

The two-step mediation MR analysis showed several potential inflammatory mediation paths (Table 2). For COPD-related pathways, circulating levels of Fms-related tyrosine kinase 3 ligand (FLT3L) mediated 18.17% of the effect of the general obesity axis on COPD risk (95% CI: 2.18–34.16%). In contrast, CD40L receptor levels showed a significant suppression effect in the association between the general obesity axis and COPD, with a mediation proportion of −19.23% (95% CI: −33.75% to −4.71%). For asthma-related pathways, circulating levels of C-X-C motif chemokine 5 (CXCL5) mediated 3.78% of the effect of the general obesity axis on asthma risk (95% CI: 0.34–7.22%), whereas CD40L receptor levels showed a suppression effect in this pathway, with a mediation proportion of −5.85% (95% CI: −8.54% to −3.17%).

For the association between the lower-body fat axis and asthma, CD40L receptor levels and leukemia inhibitory factor receptor (LIFR) levels mediated 5.59% (95% CI: 0.18–10.99%) and 12.65% (95% CI: 0.14–25.17%) of the total effect, respectively. By contrast, C-X-C motif chemokine 10 (CXCL10), interleukin-12 subunit beta (IL-12B), osteoprotegerin (OPG), and tumor necrosis factor receptor superfamily member 9 (TNFRSF9) showed significant suppression effects, with mediation proportions ranging from −19.73% to −4.08%. The remaining candidate pathways, including inflammatory protein paths related to the lower-body fat axis–COPD association and several lower-body fat axis–asthma paths, did not reach the conventional threshold for statistical significance and therefore indicated potential trends only.

Discussion

In this study, we systematically evaluated the causal relationships between four MRI-derived body-composition-based obesity axes and COPD, asthma, and AR, and further explored inflammatory protein pathways using two-step mediation MR. The results showed that the general obesity axis had the most consistent pattern of association with increased risks of asthma, COPD, and AR, whereas the lower-body fat axis was mainly associated with increased risks of COPD and asthma. In contrast, no stable evidence was observed for the muscle-dominant axis or the peripheral fat axis. These findings support the concept that obesity is not a single exposure with uniform effects on airway diseases. Instead, different body composition patterns may shape susceptibility to respiratory diseases through immune-inflammatory pathways that differ in magnitude and, in some cases, direction.

The general obesity axis reflects an overall increase in visceral fat, subcutaneous fat, and ectopic fat. The causal associations of this axis with increased risks of COPD, AR, and asthma are broadly consistent with previous observational and MR studies,^15–19 suggesting that systemic obesity burden may influence both upper and lower airway disease phenotypes. Compared with a single BMI-based measure, this axis more closely represents a composite phenotype characterized by increased whole-body adiposity, ectopic fat deposition, and metabolic inflammation. Therefore, its effect on airway diseases may not be limited to body weight gain alone. For COPD, increased systemic fat burden may aggravate thoracoabdominal mechanical restriction, reduce pulmonary reserve, and further increase respiratory burden in the presence of small airway disease or airflow limitation.^8,9,32 For asthma, obesity-related adipose tissue dysfunction may promote immunometabolic dysregulation and shift airway inflammation beyond classic allergic inflammation toward a more complex metabolic inflammatory phenotype, characterized by neutrophil recruitment, adipokine imbalance, and reduced glucocorticoid responsiveness.^33–36 For AR, the suggestive association with the general obesity axis may reflect the effects of obesity-related systemic inflammation and adipokine imbalance on upper airway mucosal immunity, including enhanced nasal mucosal inflammation, impaired epithelial barrier function, and increased susceptibility to allergic inflammation. This interpretation is consistent with the “united airway” concept shared by AR and asthma.¹ In the present study, FLT3L and CXCL5 partially mediated the associations of the general obesity axis with COPD and asthma, respectively. These findings further suggest that the general obesity axis may influence respiratory disease risk not only through mechanical loading, but also through dendritic cell-related immune regulation, chemokine-mediated inflammatory cell trafficking, and broader systemic inflammatory networks. Notably, the association between the general obesity axis and AR reached only borderline evidence in the meta-analysis and did not form a stable inflammatory mediation pathway. Therefore, this mechanism requires further validation in larger datasets with harmonized phenotype definitions and data on local inflammation in the nasal mucosa.

The association of the lower-body fat axis with increased risks of COPD and asthma is one of the most notable findings of this study. The lower-body fat axis does not simply represent greater leg fat mass; rather, it reflects greater lower-body subcutaneous fat and lower ectopic fat burden.²⁰ In metabolic disease research, lower-body subcutaneous fat is often regarded as a relatively favorable fat storage depot because it may buffer excess free fatty acid spillover and reduce fat deposition in the liver and skeletal muscle, thereby lowering the risks of insulin resistance and cardiometabolic disease.^20,37–39 However, the present findings suggest that this apparent metabolic benefit should not be directly extrapolated to respiratory diseases. In other words, the lower-body fat axis may represent a fat distribution pattern that is relatively favorable for metabolic health but not necessarily protective for respiratory health. The potential adverse effects of this phenotype on COPD and asthma may arise from total fat burden, lower-limb adiposity, reduced mobility, fatty infiltration of skeletal muscle, altered inflammatory signaling, or pulmonary mechanical load, rather than from ectopic lipotoxicity in the conventional metabolic sense. This finding highlights the organ-specific nature of the disease effects of fat distribution: a fat storage pattern that may reduce metabolic risk does not necessarily reduce the risk of airway disease.

Compared with the general obesity axis and the lower-body fat axis, the muscle-dominant axis and the peripheral fat axis did not show stable causal effects. The muscle-dominant axis primarily captures muscle volume or larger body size, rather than muscle function, muscle quality, or exercise capacity.²⁰ Therefore, the absence of a significant association should not be interpreted as evidence that skeletal muscle health is irrelevant to respiratory outcomes. Instead, it suggests that muscle volume alone may be insufficient to capture the complex relationships among muscle function, sarcopenia, exercise tolerance, and COPD prognosis. The peripheral fat axis mainly reflects abdominal and thigh subcutaneous fat storage. Unlike the lower-body fat axis, it does not simultaneously encode reduced ectopic fat burden. The null findings for the peripheral fat axis therefore support the interpretation that not all subcutaneous fat distribution patterns have consistent effects on airway diseases. Obesity-related respiratory risk may depend more on specific combinations of fat deposition patterns and inflammatory networks than on an isolated increase in fat in a single anatomical region.

The two-step mediation MR analysis further suggested that inflammatory proteins may participate in some pathways linking obesity axes to respiratory diseases. Mechanistically, obesity-related adipose tissue dysfunction may reshape the circulating inflammatory milieu through altered cytokine and adipokine secretion and persistent low-grade systemic inflammation.^8,9 These circulating signals may interact with airway epithelial and immune cells, promoting processes such as dendritic-cell activation, neutrophil recruitment, T-cell co-stimulation, and tissue remodeling, thereby providing a plausible biological bridge between abnormal body composition and chronic airway disease.¹⁰ In this context, the mediators identified in our study may represent different components of an interconnected immune-inflammatory network rather than isolated linear pathways. FLT3L partially mediated the effect of the general obesity axis on COPD. FLT3L is an important regulator of dendritic cell development and homeostasis and may influence the differentiation of conventional and plasmacytoid dendritic cells, thereby shaping antigen presentation, antiviral defense, and the initiation of adaptive immunity.⁴⁰ The development and progression of COPD involve not only airway structural damage and tobacco exposure, but also innate immunity, antigen presentation, infection susceptibility, and persistent activation of chronic inflammation.⁴¹ Therefore, the mediating role of FLT3L suggests that general obesity may increase susceptibility to COPD by altering dendritic cell-related immune regulation. CXCL5 partially mediated the association between the general obesity axis and asthma. CXCL5 is a CXC chemokine involved in neutrophil recruitment and inflammatory cell migration. Obesity-related asthma is frequently accompanied by non-eosinophilic inflammation and enhanced neutrophil-related pathways.^34,35 Therefore, CXCL5 as a mediator is biologically consistent with the non-classical inflammatory features of obesity-related asthma.

In the pathway from the lower-body fat axis to asthma, CD40L receptor and LIFR showed positive mediation effects, whereas CXCL10, IL-12B, OPG, and TNFRSF9 showed suppression effects. The CD40/CD40L axis is an important node in co-stimulatory responses involving antigen-presenting cells, B cells, and T cells, and participates in immunoglobulin class switching, T-cell activation, and amplification of inflammation.⁴² LIFR is a receptor related to the IL-6 cytokine family and may participate in epithelial repair, tissue remodeling, and inflammatory regulation. The positive mediation effects of CD40L receptor and LIFR suggest that the respiratory risk related to lower-body fat may not be explained solely by mechanical burden, but may also involve immune co-stimulation, cytokine signaling, and airway remodeling. By contrast, the negative mediation proportions for CXCL10, IL-12B, OPG, and TNFRSF9 indicate that their indirect effects were opposite in direction to the total effect. CXCL10 and IL-12B are more closely related to Th1/interferon-associated immune pathways, TNFRSF9 is a T-cell co-stimulatory receptor, and OPG is a member of the TNF receptor superfamily with links to osteoimmunology and vascular inflammation.^33,35,36 These pathways may not act in a uniformly pathogenic direction in asthma, but may reflect counter-regulation of Th2 inflammation, allergic immunity, or tissue injury. Thus, the effect of obesity axes on asthma is unlikely to follow a simple model of “increased inflammation leading to increased risk.” Instead, it may arise from multiple immune pathways, including both risk-promoting mediators and counteracting mediators.

The suppression effects observed in this study deserve particular attention from both statistical and biological perspectives. CD40L receptor showed negative mediation proportions in the associations of the general obesity axis with COPD and asthma, indicating that the indirect effect through this mediator was opposite in direction to the total effect. These findings should not be interpreted as a “negative proportion of risk explained.” Rather, they are more appropriately understood as inconsistent mediation or suppression effects, in which a specific mediating pathway counteracts the overall harmful effect.^43,44 From a biological perspective, circulating CD40L receptor levels do not necessarily reflect enhanced local CD40/CD40L signaling in the airways. They may also reflect soluble receptor activity, immune feedback regulation, or changes in antigen-presenting cell homeostasis.⁴² Therefore, the suppression effect of CD40L receptor does not weaken the total effect of obesity axes on respiratory diseases. Instead, it suggests that obesity-related immune alterations may include regulatory modules with opposing directions of effect.

This study has several important strengths. First, we extended the assessment of obesity beyond traditional BMI or waist circumference measures by using MRI-derived obesity axes, thereby avoiding the reduction of whole-body fat burden, muscle volume, peripheral subcutaneous fat, and lower-body fat storage into a single exposure. Second, this study included COPD, asthma, and AR simultaneously, reflecting the clinical concept of an allergic–obstructive airway continuum and showing that different obesity axes may have distinct effects on different airway phenotypes. Third, by integrating discovery analysis, replication analysis, meta-analysis, and two-step mediation MR, this study strengthened causal inference and incorporated circulating inflammatory proteins into mechanism-oriented screening. Several limitations should also be acknowledged. First, the data were derived mainly from populations of European ancestry. Therefore, the generalizability of the findings to Asian, African, or admixed populations requires caution. Second, large meta-analysis resources such as GBMI may include data from multiple biobanks, and differences in phenotype definitions and medical record coding across cohorts may affect replication consistency. This issue was particularly evident for AR. Third, the inflammatory protein mediation analysis was based on circulating protein quantitative trait loci, whereas circulating protein levels may not necessarily reflect local inflammatory states in the nasal mucosa, bronchi, or lung tissue. In addition, although multiple sensitivity analyses were performed to assess horizontal pleiotropy, residual pleiotropic effects or correlated inflammatory pathways cannot be completely excluded. Therefore, the identified mediation signals should be interpreted as mechanistic clues rather than definitive evidence of a single biological pathway. Future studies should further validate these pathways in multi-ancestry cohorts and with tissue-specific protein quantitative trait loci, single-cell transcriptomic data, proteomic data, and prospective imaging cohorts.

The potential value of this study lies in its implication that risk stratification for obesity-related airway diseases should not rely solely on BMI, but should also consider fat distribution and body composition patterns. For individuals at high risk of asthma or COPD, the integration of imaging-based body composition measures and inflammatory protein profiles may help distinguish metabolic-inflammatory, mechanical-load, and immune-regulatory risk subgroups more accurately. Mediating proteins such as FLT3L, CXCL5, LIFR, and CD40L receptor may also serve as candidate targets for future mechanistic research. However, they should not be directly interpreted as therapeutic targets at this stage. MR reflects the effects of long-term genetically predicted exposures, whereas pharmacological interventions are influenced by treatment timing, dose, tissue specificity, and feedback regulation. Therefore, these inflammatory proteins should be regarded as biological clues rather than immediate clinical intervention targets.

Conclusion

To encapsulate, different obesity axes showed markedly heterogeneous associations with COPD, asthma, and AR. The general obesity axis and the lower-body fat axis showed the most prominent associations with increased risks of lower airway diseases, while the general obesity axis showed a suggestive association with AR, although the pooled evidence was borderline in the meta-analysis. Inflammatory protein mediation analysis further indicated that both risk-promoting and suppressive mediation pathways may link obesity axes to airway diseases. These findings provide genetic evidence supporting potential causal associations between distinct obesity patterns and respiratory diseases and suggest that immune-inflammatory pathways may partly contribute to these relationships. They may also inform future body-composition-based risk stratification and the prioritization of inflammatory pathways for targeted prevention or intervention research.

Funding Statement

This work was supported by the project “Research on the Coordinated Development Pathway of Homogeneous Medical Quality Management and Medical Insurance Fund Security under the Framework of Urban Medical Groups” under grant number JZCXZK202624.

Data Sharing Statement

All datasets used in this study are publicly available summary-level GWAS data and were freely downloadable without restrictions for academic research purposes.

Ethics Approval and Informed Consent

The study was a secondary analysis of publicly available data and therefore did not require ethical approval and clinical registration. According to Items 1 and 2 of Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects (effective February 18, 2023, China), research projects that meet either of the following conditions may be exempt from ethical review: (1) the research uses legally obtained public data or data generated through observation without interfering with public behavior; or (2) the research uses anonymized information or data. As this study complies with both conditions, ethical review was not required.

Disclosure

There are no competing interests in this study.