Circulating lipids, lipoproteins, and the risk of idiopathic pulmonary fibrosis: a cohort study and genetic analysis

Abstract

Circulating lipids and lipoproteins have been implicated in lung health and disease, yet their associations with the risk of developing idiopathic pulmonary fibrosis (IPF) remain poorly characterized. We integrated a prospective cohort study of 479,738 United Kingdom Biobank participants (1,236 incident IPF cases during 10 years of follow-up) with two-sample Mendelian randomization (MR) using the largest available genome-wide association data (5,159 IPF cases and 27,459 controls). Circulating lipids and lipoproteins were quantified via NMR spectroscopy (138 biomarkers) and clinical biochemistry (6 biomarkers). Multivariable-adjusted Cox proportional hazards models estimated associations with incident IPF and MR using genetic variants as instruments was performed to assess potential causal relationships. After multiple-testing correction, 83 lipid biomarkers were significantly associated with incident IPF in the prospective analysis. Apolipoprotein B (ApoB)-containing lipoproteins, including LDL and VLDL, showed broadly inverse associations across lipid components and particle sizes. A size-dependent pattern was observed for HDL particles, with small HDL subfractions inversely associated with IPF risk, whereas very large HDL subfractions were positively associated. Inverse associations were also observed for apolipoprotein B, apolipoprotein A1 (ApoA-1), and broader lipid classes, including sphingomyelins. Triangulating evidence across approaches, MR analyses supported inverse associations of genetically predicted total cholesteryl esters and sphingomyelins with IPF risk. This study identifies circulating lipids and lipoproteins associated with IPF risk and highlights cholesterol and sphingolipid metabolism as potential contributors to IPF susceptibility. Further mechanistic studies and validation in independent cohorts are warranted.

Idiopathic pulmonary fibrosis (IPF) is a progressive and ultimately fatal interstitial lung disease characterized by irreversible scarring of the pulmonary parenchyma and a sustained decline in respiratory function (1). Driven by global population aging and escalating burden of environmental stressors such as air pollution, the incidence and prevalence of IPF have intensified worldwide (2, 3). Although the introduction of antifibrotic therapies has improved clinical management, these agents primarily slow disease progression rather than halting or reversing the fibrotic remodeling. This underscores a clinical imperative to further elucidate the molecular drivers of IPF and to identify robust biomarkers that could enable early detection and targeted intervention.

Emerging evidence suggests that lipids and lipoproteins play a critical role in the pathogenesis of several respiratory diseases, including pulmonary fibrosis (4, 5). Beyond their canonical function in surfactant production and the maintenance of alveolar stability, lipids are increasingly recognized as active regulators of fibroblast biology, immune responses of macrophages, and oxidative stress (6). Clinically, hyperlipidemia and cardiovascular comorbidities are more prevalent among patients with IPF than in the general population (7). Several small-scale observational studies have identified distinct circulating lipid signatures, such as alterations in triglycerides, cholesterol, and sphingolipids, that distinguish patients with IPF from healthy individuals (8, 9, 10). Cross-sectional analyses have further linked HDL cholesterol to interstitial lung abnormalities, which are considered potential precursors of IPF (11, 12). However, it remains unclear whether these lipid alterations represent causal drivers of IPF development, or instead reflect systemic consequences of established disease, or confounding from medications and lifestyle factors. Large-scale population-based studies evaluating the longitudinal associations and potential causal links between circulating lipids, lipoproteins, and IPF risk are currently lacking.

To address these knowledge gaps, we systematically evaluated the associations of circulating lipids and lipoproteins, measured by standard biochemical assay and high-throughput NMR spectroscopy, with future risk of IPF in a population-based prospective cohort. To further triangulate evidence for causality, we integrated a two-sample Mendelian randomization (MR) framework utilizing genome-wide association study (GWAS) data derived from clinically characterized IPF studies. The study design is outlined in Figure 1.

Fig. 1.

Study design overview. This study integrates a prospective cohort analysis with over 10 years of follow-up and a two-sample MR framework to investigate the associations between circulating lipids, lipoproteins, and the risk of IPF. The validity of the MR approach relies on three key assumptions for instrumental variables: (1) genetic variants are robustly associated with the lipid exposure (relevance assumption); (2) genetic variants are independent of confounders (independence assumption); and (3) genetic variants influence the IPF outcome exclusively through the lipid exposure of interest (exclusion restriction assumption; also known as absence of horizontal pleiotropy). Created in BioRender. Li, Y. (2026) https://BioRender.com/il9y8ne. IPF, idiopathic pulmonary fibrosis; MR, Mendelian randomization.

Materials and methods

Prospective cohort

Study population

The United Kingdom Biobank (UK Biobank) is a prospective cohort comprising over 500,000 participants aged 40–69 years who were recruited between 2006 and 2010 from 22 assessment centers across England, Wales, and Scotland (13). At baseline, participants completed extensive questionnaires, underwent physical assessments, and provided biological samples. These data are linked to national health registries. The participant selection process for the present study is illustrated in supplemental Fig. S1. Briefly, we excluded participants who withdrew consent, those with prevalent IPF at baseline, and individuals with missing data on lipid measurements or key covariates. The UK Biobank study received ethical approval from the North West Multi-centre Research Ethics Committee (11/NW/0382), and all participants provided written informed consent. This study was conducted in accordance with the principles of the Declaration of Helsinki.

Assessment of circulating lipid levels

Metabolic biomarkers were quantified from EDTA plasma samples collected at baseline from approximately 491,000 participants using a high-throughput proton NMR spectroscopy platform (Nightingale Health Ltd., Finland). Of the 168 biomarkers quantified as absolute concentrations, 138 correspond to lipid and lipoprotein traits. These include triglyceride, free cholesterol, cholesteryl ester, and phospholipid content of various lipoprotein subclasses, including HDL, LDL, IDL, and VLDL, as well as lipoprotein particle concentrations, mean particle diameters, apolipoprotein A1 (ApoA-1), and apolipoprotein B (ApoB).

Biomarkers assessment, quality control, and preprocessing procedures of the NMR spectroscopy data have been described previously (14). We further addressed unwanted technical variation using the R package "ukbnmr" (v3.3.0) (15). To facilitate comparison, we also examined 6 lipid biomarkers measured via standard biochemistry assays on serum samples, including total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, ApoA-1, and ApoB (16). All lipid measurements were standardized using rank-based inverse normal transformation prior to association analyses.

Assessment of incident IPF

Incident IPF cases were identified through record linkage to hospital inpatient data and national death registries using the International Classification of Diseases, 10th Revision code J84.1. To enhance the specificity of the IPF definition, we excluded cases of secondary pulmonary fibrosis based on self-reported medical history or relevant diagnostic codes (supplemental Table S1), consistent with previous studies (17, 18, 19).

Statistical analysis

Associations of individual lipid and lipoprotein biomarkers with the risk of incident IPF were evaluated using Cox proportional hazards models, with follow-up time as the underlying timescale. Person-years were calculated from the date of baseline recruitment to the date of first IPF diagnosis, death, loss to follow-up, or the study end date (December 31, 2019), whichever occurred first. The proportional hazards assumption was assessed using Schoenfeld residual tests and graphical diagnostics. Hazard ratios (HRs) and 95% CIs were estimated per SD increase in lipid biomarkers under two models. Model 1 was adjusted for age (years), sex (male or female), and ethnicity (White or non-White). Model 2 was additionally adjusted for Townsend Deprivation Index (a proxy for socioeconomic status), BMI (kg/m²), accumulative smoking exposure (pack-years), alcohol consumption frequency (never, occasional, or regular), fasting time prior to blood sampling (hours), and baseline use of lipid-lowering medications (yes or no). As missing data for covariates were minimal (1.44% of participants), complete-case analysis was performed without imputation.

To account for multiple testing in the hypothesis-free assessment of highly correlated lipid biomarkers, we applied a principal component approach using the R package “poolr” (v1.2-0), as previously done (20, 21, 22). This analysis identified 14 principal components explaining > 99.5% of the total variance in lipid biomarkers. Accordingly, statistical significance was defined using a Bonferroni-corrected threshold of P < 0.0036 (0.05/14).

Several sensitivity analyses were conducted to ensure the robustness of our findings. First, we excluded participants who developed IPF within the first two years of follow-up to minimize potential reverse causation. Second, we applied the Fine-Gray subdistribution hazard model to account for the competing risk of death. Third, analyses were repeated using a broader definition of IPF that included International Classification of Diseases, 10th Revision codes J84.0, J84.1, J84.8, and J84.9 (23). In addition, subgroup analyses stratified by sex were performed, and multiplicative interactions were assessed by including cross-product terms in the models.

Two-sample MR

Genetic data sources

To assess the potential causal effects of circulating lipids and lipoproteins on IPF risk, we conducted a two-sample MR analysis. Summary statistics for the 138 NMR-derived lipid biomarkers were obtained from a large-scale GWAS meta-analysis encompassing the UK Biobank and Estonian Biobank (n = 599,249; of European ancestry) (24). These lipid measurements underwent quality control and normalization procedures consistent with those applied in our prospective cohort analysis.

Summary statistics for IPF were sourced from a recent GWAS meta-analysis by the International IPF Genetics Consortium, comprising seven clinically characterized cohorts (5,159 cases and 27,459 controls of European ancestry) (25). All IPF cases were diagnosed in accordance with the American Thoracic Society/European Respiratory Society guidelines (26). In both the exposure and outcome GWAS, genetic association estimates were adjusted for age, sex, and genetic principal components.

A potential for sample overlap exists because a subset of control participants (n = 10,000) in the IPF GWAS was drawn from the UK Biobank, which also contributed to the lipid GWAS. However, this overlap represents a small proportion of the total study population, and the use of strong genetic instruments is expected to minimize potential bias in the two-sample MR estimates (27).

Selection of instruments and data harmonization

To identify genetic instruments, SNPs associated with each lipid biomarker at genome-wide significance (P < 5 × 10⁻⁸) were selected. Linkage disequilibrium clumping was then performed to ensure the independence of the selected instruments (r² < 0.001 within a 10,000 kb window) using the European reference panel from the 1000 Genomes Project.

Data harmonization was conducted to align the effect alleles between the exposure and outcome datasets. Palindromic SNPs with intermediate allele frequencies were excluded to avoid strand ambiguity. The strength of the instrumental variables was quantified using the F-statistic, with an F-statistic >10 considered indicative of sufficiently strong instruments (28).

Statistical analysis

The primary causal effect was estimated using the inverse-variance weighted (IVW) method with multiplicative random effects. This method combines SNP-specific Wald ratio estimates and provides the most precise causal estimates under the assumption that all instrumental variables are valid (29). MR effect estimates were reported as odds ratios (ORs) with 95% CIs per SD increase in lipid biomarkers. Cochran’s Q statistic was used to assess heterogeneity across SNP-specific estimates. To prioritize causal lipid biomarkers from high-dimensional, correlated exposures, multivariable MR approach based on Bayesian model averaging was employed (30). The method ranks candidates based on their marginal inclusion probability, effectively selecting the most robust causal factors by averaging over all potential subset models.

To evaluate the robustness of the findings and assess potential violations of MR assumptions, several complementary sensitivity analyses were performed. First, the weighted median and MR-Egger regression methods were applied, which can provide consistent causal estimates even when some instruments are invalid (31, 32). Second, the MR-Egger intercept test was used to assess directional horizontal pleiotropy, and MR pleiotropy residual sum and outlier was applied to detect and correct for potential pleiotropic outliers (33). Third, multivariable MR analyses using standard IVW method were conducted to account for potential confounding due to established IPF risk factors, including smoking exposure and BMI. Finally, Steiger filtering was applied to verify the direction of causality by confirming that the selected SNPs explained more variance in the lipid exposure than in the IPF outcome (34).

All MR analyses were conducted using the R packages “TwoSampleMR” (v0.7.0), “MendelianRandomization” (v0.10.0), and “mrbma” (v0.1.0) (30, 35). Statistical power was estimated using the R package “mRnd” (36). Evidence for potential causality was defined as associations that met the Bonferroni-corrected significance threshold (P < 0.0036), showed consistent effect directions across prospective cohort and MR analyses, and demonstrated no evidence of violations of key MR assumptions.

Results

Prospective cohort

The prospective analysis included 479,738 participants from the UK Biobank (54.3% female; mean age 56.52 years). During a median follow-up of 10.91 years (totaling 5,186,488 person-years), 1,236 incident cases of IPF were identified. Baseline characteristics of the study population are presented in Table 1. Compared with participants who remained free of IPF, individuals who subsequently developed the disease were older at baseline, more likely to be male, had higher levels of socioeconomic deprivation, higher BMI, greater cumulative smoking exposure, and a higher prevalence of lipid-lowering medication use.

Table 1.

Baseline characteristics of the study participants in the UK Biobank

Characteristic	Total participants (n = 479,738)	Incident IPF (n = 1,236)	Non-IPF (n = 478,502)
Age, years, mean (SD)	56.52 (8.09)	62.93 (5.80)	56.50 (8.09)
Sex, n (%)
Female	260,700 (54.3)	419 (33.9)	260,281 (54.4)
Male	219,038 (45.7)	817 (66.1)	218,221 (45.6)
Ethnicity, n (%)
White	454,741 (94.8)	1,189 (96.2)	453,552 (94.8)
Non-White	24,997 (5.2)	47 (3.8)	24,950 (5.2)
Townsend deprivation index, median (IQR)	−2.16 (−3.66, 0.50)	−1.66 (−3.32, 1.79)	−2.17 (−3.66, 0.49)
Body mass index, kg/m2, mean (SD)	27.41 (4.78)	28.67 (4.86)	27.40 (4.78)
Pack-years of smoking, years, median (IQR)	0.00 (0.00, 7.20)	7.50 (0.00, 33.00)	0.00 (0.00, 7.10)
Alcohol intake frequency, n (%)
Never	38,061 (7.9)	115 (9.3)	37,946 (7.9)
Occasional	108,427 (22.6)	293 (23.7)	108,134 (22.6)
Regular	333,250 (69.5)	828 (67.0)	332,422 (69.5)
Lipid-lowering medications, n (%)
No	393,668 (82.1)	713 (57.7)	392,955 (82.1)
Yes	86,070 (17.9)	523 (42.3)	85,547 (17.9)
Fasting time, hours, median (IQR)	3.00 (2.00, 4.00)	3.00 (2.00, 4.00)	3.00 (2.00, 4.00)

IQR, interquartile range; PF, idiopathic pulmonary fibrosis; SD, standard deviation.

In Cox proportional hazards analyses of 138 NMR-derived and 6 biochemistry-measured lipid biomarkers, Schoenfeld residual tests indicated nominal deviations from the proportional hazards assumption (P < 0.05) for 22 biomarkers (supplemental Table S2). However, visual inspection of the residual plots revealed no clear evidence of systematic time-dependent trends (supplemental Fig. S2), suggesting that the estimation of average HRs remained appropriate in this large cohort. After correction for multiple testing (P < 0.0036), 105 lipid and lipoprotein biomarkers were significantly associated with incident IPF in the minimally adjusted model, of which 83 remained significant after full covariate adjustment (supplemental Table S2).

As shown in Figure 2, a consistent pattern of inverse associations was observed for ApoB-containing lipoproteins, including LDL and VLDL subfractions across nearly all lipid components (cholesterol, free cholesterol, cholesteryl esters, and phospholipids) as well as across particle sizes (small, medium, and large). HRs per SD increase ranged from 0.83 (95% CI: 0.78-0.89) for cholesterol in medium LDL to 0.91 (95% CI: 0.86-0.97) for total lipids in VLDL. Associations for HDL subfractions varied according to particle size. Higher concentrations of small HDL particles were consistently associated with lower risk of IPF, with the strongest inverse association observed for total small HDL particle concentrations (HR per SD increase: 0.82; 95% CI: 0.78-0.87), and similar estimates across their lipid components. In contrast, very large HDL was the only subfraction associated with an increased risk of IPF, with HRs per SD increase ranging from 1.11 (95% CI: 1.04-1.18) for cholesterol to 1.17 (95% CI: 1.09-1.24) for free cholesterol component. Significant inverse associations were also observed for ApoA-1, ApoB, and broader lipid classes, including phosphatidylcholines, cholines, phosphoglycerides, and sphingomyelins.

Fig. 2.

Associations of individual lipid and lipoprotein biomarkers with risk of incident IPF. HRs and 95% CIs were estimated per SD increase in lipid and lipoprotein levels using Cox proportional hazards models adjusted for age, sex, ethnicity, Townsend Deprivation Index, body mass index, pack-years of smoking, alcohol consumption frequency, fasting time prior to blood sampling, and baseline use of lipid-lowering medications. Asterisks (∗) indicate associations that remained statistically significant after correction for multiple testing (P < 0.0036). Lipid biomarkers are grouped by lipid composition and lipoprotein subfractions. ApoA1, apolipoprotein A1; ApoB, apolipoprotein B; -CE, cholesteryl ester; CI, confidence interval; -FC, free cholesterol; HDL, high-density lipoprotein; HR, hazard ratio; IDL, intermediate-density lipoprotein; IPF, idiopathic pulmonary fibrosis; L-, large; LDL, low-density lipoprotein; M-, medium; -P, particle concentration; phosphatidylc, phosphatidylcholines; phosphoglyc, phosphoglycerides; PL, phospholipids; S-, small; SD, standard deviation; TG, triglycerides; VLDL, very-low density lipoprotein; XL-, very-large; XS-, very-small; XXL-, extremely-large.

Results for 6 lipid biomarkers measured using both conventional biochemistry and the NMR platform were broadly consistent (supplemental Fig. S3). Sensitivity analyses yielded similar findings (supplemental Table S3), and sex-stratified analyses showed no evidence of effect modification by sex (all interaction P > 0.05) (supplemental Table S4).

Two-sample MR

In the two-sample MR analysis exploring potential causal effect of lipids and lipoproteins on IPF risk, the number of independent SNPs selected as genetic instruments for each biomarker ranged from 84 to 275, collectively explaining 2.45%–12.97% of the phenotypic variance (supplemental Table S5). All individual F-statistics exceeded 30, indicating that the instruments had good strength. Statistical power calculations demonstrated that while the mean power to detect a modest effect (OR = 1.10 per SD increase) across 138 biomarkers was 46.78%, the study was sufficiently powered (>80%) to detect causal effects of OR > 1.2.

Following correction for multiple comparisons (P < 0.0036), the primary IVW analysis identified three lipid biomarkers significantly associated with IPF risk, two of which showed effect directions consistent with the prospective cohort findings (Fig. 3 and supplemental Table S6). Specifically, genetically predicted higher levels of total cholesteryl esters (instrumented by 166 SNPs; OR per SD increase: 0.80; 95% CI: 0.69-0.92) and sphingomyelins (169 SNPs; OR per SD increase: 0.80; 95% CI: 0.69-0.93) were associated with a lower risk of IPF. There was no evidence of substantial heterogeneity or horizontal pleiotropy for these associations (supplemental Table S6), and no influential outliers or indications of reverse causality were detected (all P > 0.05). Sensitivity analyses using alternative univariable MR methods, as well as multivariable MR adjusting for smoking and BMI, yielded consistent estimates (supplemental Table S7). In the MR approach based on Bayesian model averaging analysis, total cholesteryl esters and sphingomyelins yielded comparable levels of evidence, with marginal inclusion probabilities of 0.635 and 0.560, respectively, both surpassing the 0.5 prioritization threshold.

Fig. 3.

Lipid and lipoprotein biomarkers significantly associated with IPF identified by MR study and comparison with prospective cohort estimates. Effect estimates from MR are presented as ORs with 95% CIs per SD increase in lipid and lipoprotein levels, derived using the inverse-variance weighted method. Corresponding estimates from the prospective cohort are presented as HRs with 95% CIs per SD increase, obtained from Cox proportional hazards models adjusted for age, sex, ethnicity, Townsend Deprivation Index, body mass index, pack-years of smoking, alcohol consumption frequency, fasting time prior to blood sampling, and baseline use of lipid-lowering medications. CI, confidence interval; HR, hazard ratio; IPF, idiopathic pulmonary fibrosis; MR, Mendelian randomization; OR, odds ratio; SD, standard deviation; Total_CE, total cholesteryl esters; XL_HDL_P, concentration of very large high-density lipoprotein particles.

The IVW analysis also suggested an inverse association between genetically predicted levels of very large HDL particles and IPF risk (227 SNPs; OR per SD increase: 0.78; 95% CI: 0.68-0.91) (Fig. 3). However, this association showed evidence of heterogeneity across SNP-specific estimates (Cochran’s Q P < 0.05) and potential pleiotropic outliers (MR pleiotropy residual sum and outlier global test P < 0.05), and the direction of effect was notably inconsistent with that observed in the prospective cohort analysis (HR per SD increase: 1.09; 95% CI: 1.02-1.16).

Discussion

In this large-scale prospective cohort study integrating NMR-based lipid profiling, we systematically evaluated the associations between circulating lipids, lipoproteins, and the subsequent risk of IPF. We identified a broad pattern of inverse associations between multiple lipid biomarkers and incident IPF, particularly among ApoB-containing lipoprotein subclasses and several major lipid classes. By integrating prospective epidemiological evidence with genetic analyses, total cholesteryl esters and sphingomyelins were prioritized as lipid traits potentially contributing to lower IPF susceptibility. These findings provide population-level evidence supporting a potential contribution of circulating lipid metabolism to IPF development and highlight the importance of investigating lipid-related pathways in disease susceptibility and future biomarker discovery.

Previous observational studies have primarily investigated circulating lipid profiles in patients with IPF, focusing on their associations with disease severity and clinical outcomes. For example, circulating HDL characteristics have been inversely associated with clinical outcomes (e.g., mortality and lung transplantation) in IPF as well as subclinical interstitial lung abnormalities and circulating biomarkers of pulmonary injury and extracellular matrix remodeling (e.g., SP-A and MMP-7) (11, 37, 38). These findings, consistent with our observations, suggest that specific lipid and lipoprotein profiles may be associated with more favorable pulmonary phenotypes. However, previous studies were largely conducted after disease establishment, making it difficult to determine whether lipid alterations represent early biological determinants of IPF development or secondary consequences of progressive fibrosis. By leveraging a prospective cohort design, our study extends previous observations by suggesting that circulating lipid profiles may be involved in the biological processes underlying IPF susceptibility. Notably, we observed heterogeneous associations across HDL subclasses. Small HDL particles showed inverse associations with IPF risk, whereas very large HDL particles showed positive associations with IPF risk. These findings highlight that HDL particles are biologically heterogeneous and that their associations with IPF may depend on particle size and composition rather than HDL concentration alone (39).

Accumulating evidence indicates that circulating lipids and lipoproteins may influence pulmonary lipid homeostasis and regulate cellular processes involved in IPF pathogenesis. Alveolar type II (AT2) cells represent a major interface between systemic lipid metabolism and the lung, as circulating lipoproteins can be taken up by AT2 cells and provide lipid substrates for pulmonary surfactant synthesis and secretion (40). Disruption of AT2 lipid metabolic pathways and surfactant homeostasis has been implicated in epithelial dysfunction, impaired repair capacity, and progressive fibrotic remodeling in IPF (6, 41). In addition to epithelial cells, lipid metabolism is critical for maintaining macrophage-mediated alveolar homeostasis. Altered cholesterol handling and impaired lipid efflux in macrophages may promote lipid accumulation, inflammatory activation, and macrophage states associated with fibrotic remodeling within the lung microenvironment (42, 43). Furthermore, fibroblasts undergo substantial metabolic reprogramming during fibrosis, including alterations in lipid synthesis, storage, and utilization, which contribute to myofibroblast activation and extracellular matrix deposition (44, 45). Together, these findings provide a mechanistic framework supporting the hypothesis that circulating lipid profiles may influence IPF susceptibility through coordinated regulation of epithelial integrity, immune responses, and fibroblast activation.

Among the lipid traits prioritized by our genetic analyses, total cholesteryl esters are of particular interest given the central role of cholesterol homeostasis in pulmonary biology (5). The lung has a unique dependence on circulating lipoproteins for cholesterol supply while requiring tightly regulated cholesterol trafficking and recycling to maintain surfactant function and alveolar homeostasis. Dysregulated cholesterol handling has been implicated in lipid accumulation in macrophages, inflammatory activation, and fibrotic remodeling. Importantly, the biological effects of cholesteryl esters may vary according to their fatty acid composition, with saturated cholesteryl esters reported to enhance fibrotic marker expression in lung fibroblasts, whereas unsaturated cholesteryl esters may exert opposing effects (46). The inverse association between genetically predicted total cholesteryl ester levels and IPF risk observed in our study may reflect a systemic lipid profile associated with preserved cholesterol homeostasis. However, future lipid species-level analyses are needed to determine which cholesteryl ester subclasses contribute to IPF susceptibility.

Sphingomyelins represent another lipid class prioritized by our genetic analyses. As major structural components of cellular membranes and lipid rafts, sphingomyelins contribute to membrane organization and regulate signaling pathways involved in inflammation, stress responses, and cellular survival. Alterations in sphingolipid metabolism, including pathways involving the bioactive mediator sphingosine-1-phosphate, have been implicated in pulmonary fibrosis through regulation of epithelial cell apoptosis, myofibroblast differentiation, and TGF-β signaling (47). Consistent with the pathogenic relevance of this pathway, pharmacological modulation of sphingolipid metabolism has shown beneficial effects in experimental models of lung injury and fibrosis (48). The strong genetic correlation (r = 0.94) between sphingomyelins and cholesteryl esters suggests shared genetic regulation of these lipid traits (24), although the specific mechanisms linking circulating sphingomyelin levels to IPF susceptibility require further investigation.

From a clinical perspective, circulating lipid profiles may contribute to future IPF risk stratification approaches. Current strategies for IPF diagnosis largely rely on clinical symptoms and radiological evidence, often after irreversible fibrotic remodeling has occurred. Circulating metabolic signatures associated with future IPF risk may provide complementary information for identifying individuals with increased susceptibility, particularly when integrated with established genetic risk factors (e.g., MUC5B variants) and environmental exposures (e.g., smoking). In addition, our findings highlight lipid metabolism as a potential biological axis in IPF pathogenesis and suggest that cholesterol and sphingolipid pathways warrant further mechanistic investigation. However, whether lipid profiles improve clinical prediction beyond existing risk factors or whether modulation of lipid metabolism can influence IPF development requires validation in independent cohorts and prospective interventional studies.

To the best of our knowledge, this is the first study to investigate the association between NMR-derived lipid measurements and incident IPF risk in a large population-based prospective cohort. Several limitations should be considered. First, residual confounding cannot be completely excluded despite extensive adjustment for demographic characteristics, lifestyle factors, and medication use. Second, incident IPF cases were identified through linked electronic health records rather than prospective clinical adjudication, which may introduce some degree of outcome misclassification. Third, although MR analyses using genetic variants from well-defined IPF cohorts provide evidence supporting potential causal relationships, genetic instruments explain only a proportion of the variance in lipid traits and may be affected by horizontal pleiotropy. The consistency between prospective and MR analyses nevertheless strengthens the interpretation of our findings. Fourth, our study focused on 138 NMR-derived lipid traits and did not capture the full spectrum of lipid species. Finally, because the study populations were predominantly of European ancestry, replication in independent IPF cohorts with diverse ancestries and clinically adjudicated diagnoses is required to assess the generalizability of our findings.

In conclusion, this study provides evidence that circulating lipids and lipoproteins are associated with future IPF risk. By integrating prospective epidemiological analyses with genetic evidence, our findings support a potential role for systemic lipid metabolism in IPF development and highlight cholesterol and sphingolipid pathways as important areas for further investigation. Future studies incorporating independent cohorts and mechanistic experimental validation are warranted to clarify the biological mechanisms underlying these associations and their potential clinical relevance.

Data availability

This research was conducted using the UK Biobank resource under application number 95180. Access to UK Biobank data is available upon application (https://www.ukbiobank.ac.uk/enable-your-research/apply-for-access/). Genetic association summary statistics for IPF were obtained through application to the International IPF Genetics Consortium (https://github.com/genomicsITER/PFgenetics). Summary statistics for NMR metabolites are publicly available in the GWAS Catalog (https://www.ebi.ac.uk/gwas/; accession numbers: GCST90451106-GCST90451603).

Supplemental data

This article contains supplemental data.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.