Potential Causal Association Between Plasma Lipidomes and Keloid: Analysis from a Two-Sample Mendelian Randomization

Abstract

Background

Keloid formation is a multifaceted biological response to skin injury characterized by fibrous connective tissue, and functional and aesthetic impairments. Recent research highlights the role of plasma lipidomes in various diseases, yet their specific relationship with keloid formation has not been extensively studied.

Objective

This study aims to elucidate the causal relationship between plasma lipidomes and keloid formation using a two-sample Mendelian randomization (MR) approach.

Methods

Firstly, we employed a two-sample MR framework, utilizing genetic instrumental variables (IVs) derived from large-scale genome-wide association studies (GWAS). The exposure data comprised 179 plasma lipid species measured in 7174 individuals of European ancestry, while outcome data on keloids were sourced from a GWAS involving 436,200 individuals. IVs were selected using a significance threshold of P < 1×10⁻⁵ and linkage disequilibrium criteria, yielding 4507 independent single-nucleotide polymorphisms (SNPs). Causal effect estimates were calculated using multiple MR methods, and sensitivity analyses were conducted to evaluate the robustness of the findings.

Results

The analysis suggested a significant association between elevated levels of Phosphatidylcholine (O-16:0_16:0) and increased risk of keloid formation (OR 1.399, 95% CI 1.166–1.678, p-value < 0.001) by Inverse-variance weighted (IVW). Sensitivity analyses confirmed the absence of heterogeneity and significant horizontal pleiotropy, reinforcing the validity of the causal inference.

Conclusion

Our findings indicate that specific lipid species, particularly Phosphatidylcholine, may be associated with an increased risk of keloid formation, highlighting a potential role of lipid metabolism in keloid pathophysiology.

Introduction

Keloid formation is a complex biological process that occurs after tissue injury, leading to the replacement of normal tissue with fibrous connective tissue and resulting in pathological changes in skin structure and appearance due to dysregulation of collagen synthesis and degradation.¹ Keloid typically presents with alterations in pigmentation and texture, and may be associated with sensations such as itching or pain, which can adversely affect both physical function and psychological well-being.² Keloids could lead to both physical limitations and psychological distress, accompanied by functional and aesthetic impairments, profoundly affecting the overall well-being of patients.³ The underlying mechanisms of keloid formation involve numerous biological pathways, including inflammation, exosome interactions, epigenetics, fibroblast activation, and extracellular matrix remodeling. Despite advancements in understanding these mechanisms, effective therapeutic strategies to mitigate excessive keloid formation remain limited.^4,5

It is well known that plasma lipidomics plays a crucial role in various physiological and pathological conditions.⁶ Plasma lipidomics refers to the comprehensive analysis of different lipid species present in plasma, which are essential for cellular signaling, membrane structure, and inflammatory responses.⁷ Alterations in lipid metabolism are associated with a range of diseases, including cardiovascular diseases, tumors, and metabolic syndrome, suggesting that lipid profiles may serve as biomarkers of disease risk and even as potential therapeutic targets.⁸ For instance, the study found that phosphatidylinositol (18:1_20:4) is associated with the risk of IgA nephropathy (IgAN) and may serve as a biomarker for early detection and treatment of IgAN.⁹ However, the specific relationship between plasma lipidomes and keloid formation has not been thoroughly investigated.

Systemically, plasma lipids are not isolated metabolic markers, and also participate in lipid transport and exchange among peripheral tissues, including the skin. The alterations in circulating lipid composition may therefore be reflected in, or contribute to, changes in the cutaneous lipid microenvironment.¹⁰ At the tissue level, lipids are essential components of the epidermal barrier and also act as bioactive mediators that regulate inflammatory responses, oxidative stress, keratinocyte function, and wound healing. Disruption of these lipid-dependent processes may prolong inflammation or impair the resolution phase of repair, thereby creating conditions favorable for excessive extracellular matrix deposition.¹¹ At the cellular level, lipid composition can influence dermal fibroblast phenotype and function. Changes in membrane lipids and lipid-derived signaling molecules may affect fibroblast activation and collagen synthesis, which are central events in keloid pathogenesis. Certain lipid classes may promote a shift from regenerative repair toward a profibrotic program by modulating inflammatory and fibrogenic pathways, including those related to transforming growth factor-β signaling and extracellular matrix remodeling.^12,13 Thus, plasma lipidomic alterations may influence keloid risk by affecting lipid transport, skin barrier homeostasis, inflammation, and fibroblast-mediated fibrosis. Based on these reasons, it is possible to utilize genetically predicted plasma lipidomes to explore potential causal relationships with keloid formation.

Epidemiological researches explore the distribution and determinants of health-related states and events in populations, providing critical insights that guide public health policies and disease prevention strategies.¹⁴ Within this framework, Mendelian randomization (MR) capitalizes on genetic variants as instrument variables (IVs) to infer causal relationships between modifiable exposures and health outcomes, thereby addressing confounding issues often present in observational studies.¹⁵ The innovative approach of MR analysis not only enhances the robustness of causal inference but also facilitates the identification of potential therapeutic targets in various complex diseases.¹⁶ Nowadays, MR analysis has been applied across a spectrum of diseases, including cardiovascular disorders, skin diseases, diabetes, and various types of cancer, demonstrating its versatility in elucidating causal relationships and informing therapeutic interventions.^17–19

MR analysis has been utilized to investigate the causal relationships between genetic factors, such as lipid levels and immune responses, and the risk of keloid formation, shedding light on potential underlying mechanisms. For example, a study demonstrates that MR analysis identifies neurotrimin (NTM) and 6 other proteins as potential therapeutic targets for keloids, with NTM significantly upregulated in keloid tissues compared to normal skin, thus highlighting its relevance in keloid treatment strategies.²⁰ Additionally, Zou et al employ MR to reveal a bidirectional causal relationship, indicating that CD66b++ myeloid cell abundance serves as a protective factor against keloid formation.²¹ These studies indicate that MR analysis, either independently or in conjunction with other omics approaches, can provide causal evidence and novel targets for refractory conditions such as keloids, warranting further investigation.

Nonetheless, there is currently a lack of evidence-based reports on the association between plasma lipids and keloids. To address this issue, we intend to clarify the relationship using MR analysis. Methodologically, this study systematically employs a two-sample MR framework to investigate the causal relationship between plasma lipidomes and keloid formation by utilizing genetic instrumental variables (IVs) derived from large-scale genome-wide association studies. Then, we selected independent single-nucleotide polymorphisms (SNPs) associated with specific lipid species, carefully ensuring compliance with the core MR assumptions to eliminate confounding biases. Finally, we applied multiple causal effect estimation methods and sensitivity analyses to validate our findings, thereby elucidating the potential impact of lipid metabolism on keloid development.

Methods

Study Design and MR Framework

This study employs a two-sample MR analysis strategy to assess the potential causal association between plasma lipidomes as exposure and keloid formation as an outcome. The analysis rigorously follows the three fundamental assumptions of MR: (1) Correlation assumption: The selected genetic IVs must have a significant association with the target exposure, namely specific plasma lipidomes, at the genome-wide level; (2) Independence assumption: IVs should be independent of any known or unknown confounders that could bias the exposure-outcome relationship; (3) Exclusivity assumption: The effect of IVs on the outcome, which is keloid, must be entirely mediated through the exposure, which is plasma lipidomes, with no direct effects or alternative biological pathways, thereby excluding horizontal pleiotropy (Figure 1).

Figure 1.

Technical route. The technical route of the study, highlighting the screening process for 179 SNPs associated with plasma lipid traits. SNPs significantly associated with the exposure factor (plasma lipids) were selected filtering criteria. The lines marked with numbers in the figure represent the three core assumptions of Mendelian randomization: 1) strong correlation between the IV and the exposure (plasma lipids); 2) Independence of the IV from confounding factors; 3) the IV affects the outcome (keloid) only through the exposure.

Data Sources and Sample Characteristics

The exposure data on plasma lipidomes were extracted from the genome-wide association studies (GWAS) Catalog database (GCST90277238 - GCST90277416).⁶ This dataset originates from a genome-wide association study (GWAS) involving 7174 individuals of European ancestry. The study quantitatively measured the concentrations of 179 plasma lipidomes using a mass spectrometry-based shotgun lipidomics approach. Genome-wide SNP genotyping was conducted using the HumanCoreExome array platform. Additionally, the outcome data on keloids were also sourced from the GWAS Catalog database (GCST90483368).²² This dataset integrates information from 436,200 individuals of European ancestry, with 1779 cases explicitly diagnosed with keloids and 434,421 healthy control samples.

This study utilized publicly available GWAS data, which were previously collected with ethical approval and informed consent from participants. As this study did not involve new data collection or direct interaction with participants, additional ethical approval was not required.

Selection and Quality Control of IVs

The selection of IVs followed a rigorous protocol to ensure compliance with the core assumptions of MR. Initially, for each target 179 plasma lipidome, all SNPs that reached a genome-wide significance threshold (P < 1×10⁻⁵) were identified. This relatively relaxed threshold was adopted because the number of SNPs associated with individual lipid species at the conventional genome-wide significance level was limited. Therefore, using P < 1×10⁻⁵ allowed us to retain a sufficient number of candidate IVs and improve statistical power for MR analyses. To eliminate linkage disequilibrium (LD) among SNPs and ensure the independence of IVs, a threshold of r² < 0.001 was applied, along with a physical distance window of 10,000 kb for clustering, retaining only the SNP with the lowest P-value from each LD block. Following these screening and LD quality control measures, a total of 4507 independent SNPs were identified and utilized as genetic IVs for subsequent MR analyses. Additionally, SNPs exhibiting palindromic characteristics were directionally corrected or excluded based on allele frequencies from both exposure and outcome GWAS.

Causal Effect Estimation and Sensitivity Analysis

Causal effect estimates were derived using 6 complementary MR methods: Inverse-variance weighted (IVW), MR-Egger regression, Maximum likelihood, Weighted median, Simple mode, and Weighted mode. Evidence of a significant causal effect was considered present when the OR estimates and corresponding 95% CIs obtained from at least three of the six MR methods showed consistent directions, and the IVW method yielded a P value < 0.05. Effect estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs) per genetically predicted one-standard-deviation increase in the exposure. Specifically, lipidomes with OR > 1 and P < 0.05 were considered risk factors for keloid, while those with OR < 1 and P < 0.05 were regarded as protective factors.

To assess the robustness of the results, comprehensive sensitivity analyses were conducted. First, heterogeneity was evaluated using Cochran’s Q statistic based on the MR method, with a P-value > 0.05 indicating no evidence of heterogeneity, suggesting homogeneity among IV estimates. Second, horizontal pleiotropy was examined via the intercept term of MR-Egger regression, with an intercept P-value > 0.05 indicating no significant directional horizontal pleiotropy, supporting the exclusion assumption. Third, leave-one-out analysis was performed, where each IV was sequentially removed, and the combined causal effect estimates, primarily based on IVW or Weighted median, were recalculated. If the combined effect estimates (OR point estimates and confidence intervals) of the remaining IVs did not significantly deviate from the results of the full dataset upon removal of any single IV, this indicated that any single influential SNP did not drive the findings and thus demonstrated robustness. Finally, visual diagnostic tools were employed to intuitively present the stability of the results. Scatter plots depicted the effect sizes and confidence intervals for each instrumental variable, while funnel plots assessed the symmetry of effect sizes against the inverse of standard errors. Additionally, leave-one-out forest plots provided further insights into the robustness of the findings.

Statistical Analysis

All statistical analyses were performed using R software. MR analyses were conducted primarily using the TwoSampleMR package. Data harmonization, LD clumping, causal effect estimation, heterogeneity testing, MR-Egger intercept analysis, and leave-one-out analyses were performed using functions implemented in TwoSampleMR.

Results

SNP Selection and Instrument Variable Determination

After quality control and LD clumping, a total of 4507 independent SNPs were retained as instrumental variables for the 179 plasma lipid species included in the MR analyses (Figure 1). The number of SNPs available for each lipid species varied across exposures, reflecting differences in the genetic architecture and variance explained by individual lipid traits. Detailed information on the SNP instruments and F-statistics for each lipid species is provided in Table S1.

Among the 179 plasma lipid species examined, Phosphatidylcholine (O-16:0_16:0) showed evidence of a positive association with keloid risk in the primary MR analysis. Using the IVW method, genetically predicted higher levels of Phosphatidylcholine (O-16:0_16:0) were associated with an increased risk of keloid formation (OR 1.399, 95% CI 1.166–1.678, p-value < 0.001). IVW results were consistent with MR Egger, Maximum likelihood, and Weighted median. The direction of effect was generally consistent across complementary MR methods, supporting the potential association between this lipid species and keloid formation (Figure 2 and Table 1).

Figure 2.

Mendelian randomization (MR) results table with a forest plot of MR effect estimates. Forest plot presents the results of evaluating the causal association between plasma lipids and keloids using the weighted median method. It displays the names of various lipid exposure factors, the analytical methods used, the number of IVs, as well as the OR and its 95% confidence interval. “exposure” refers to the exposure factor, “Method” indicates the algorithm used to estimate causal effects, “nsnp” denotes the number of IVs for that exposure factor, “pval” represents the corresponding p-value for the algorithm, and “OR (95% CI)” indicates the OR along with 95% confidence interval. The vertical line represents the null line (OR = 1), used to assess the statistical significance of the associations. Bold P-values indicate statistical significance at P < 0.05.

Abbreviations: CI, confidence interval; MR, Mendelian randomization; OR, odds ratio; SNP, single-nucleotide polymorphism.

Table 1.

MR Analysis for the Association Between Phosphatidylcholine (O-160_160) and Keloid Risk

id.exposure	id.outcome	Method	nsnp	b	se	pval	lo_ci	up_ci	or	or_lci95	or_uci95
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	Inverse variance weighted	17	0.335581160407614	0.0928853193285286	0.000302844448669678	0.153525934523698	0.51763638629153	1.39875304894768	1.16593802477734	1.67805668085494
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	MR Egger	17	0.397945130650489	0.182435726331375	0.0454892786079746	0.0403711070409949	0.755519154259984	1.48876234024498	1.04119709807836	2.12871636870901
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	Maximum likelihood	17	0.342880817089425	0.0953085630418312	0.000321183472034173	0.156076033527436	0.529685600651414	1.40900082309983	1.16891507644768	1.69839824936577
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	Weighted median	17	0.384610059174884	0.124764809317985	0.00205145197858398	0.140071032911633	0.629149085438134	1.46904137053577	1.15035550905636	1.87601357263539
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	Simple mode	17	0.435568614885716	0.218048235180256	0.0630579066583083	0.00819407393241522	0.862943155839018	1.54584179775285	1.00822773723994	2.37012608899397
Phosphatidylcholine (O-16:0_16:0) levels	Keloid	Weighted mode	17	0.454064292799945	0.213890826639349	0.049715895255084	0.0348382725868218	0.873290313013069	1.57469923604349	1.03545223424606	2.39477747208827

Sensitivity Analysis and Heterogeneity Testing

To validate the robustness of our findings, we conducted heterogeneity testing on the lipid species identified as exposure factors. For phosphatidylcholine (O-16:0_16:0), Cochran’s Q test detected no evidence of substantial heterogeneity among the instrumental variables, either under the MR-Egger method (Q = 6.5773, P = 0.9683) or the IVW method (Q = 6.7351, P = 0.9780) (Table S2 Sensitivity CochranQ result, and Table S3 Sensitivity Horizontal Pleiotropy result). These nonsignificant results indicate that no evidence of substantial heterogeneity was detected.

Outlier Analysis and Robustness Assessment

To exclude potential outliers among the IVs, we performed causal estimation analyses for individual SNPs. Observations show that as the level of exposure increases, the risk of the outcome also increases (Figure 3A). The consistent trend across six algorithms demonstrates the robustness of the causal relationship between SNPs, exposure, and outcome factors. The IVs showed a symmetrical funnel shape with no significant outliers of low-precision SNPs (Figure 3B). Figure 3A and B illustrate the effect sizes of each IV on exposure and outcome, showing a symmetrical distribution around the regression fit line without any apparent outliers, further confirming the stability of our results.

Figure 3.

Visualization methods for SNP stability analysis. (A) Scatter Plots illustrate the effect sizes of each SNP on both exposure and outcome, with regression fit lines. The x-axis shows the SNP effect on the exposure, and the y-axis represents the effect size of each SNP on the outcome, measured in units that indicate the risk level of the outcome. Each black point represents an SNP, with horizontal lines indicating confidence intervals for the exposure and vertical lines for the outcome. Different colored lines represent regression lines fitted by various algorithms estimating the causal effect. (B) Funnel Plot assesses the symmetry of IVs. The x-axis represents the effect estimate (β) for each IV, while the y-axis shows the inverse of the standard deviation. The light blue and dark blue segments represent fitted curves from the IVW and MR-Egger algorithms, respectively. The symmetric distribution of SNPs around the fitted curves indicates no significant outliers. (C) Leave-One-Out analysis forest plot displays changes in effect size after sequentially excluding SNPs. The y-axis lists the excluded SNP names, and the x-axis shows effect estimates post-exclusion of each SNP, with the red color indicating the original combined effect estimate.

Finally, we employed a leave-one-out analysis, systematically removing each SNP to re-estimate the causal effect, comparing it with the complete effect size. Figure 3C showed that each SNP was excluded one by one, and the causal effect of exposure on the outcome was recalculated using the remaining SNPs to determine if the MR result was driven by an influential outlier. The results indicated that all the effect size estimates stayed on the same side of the null line and were consistent with the estimates derived from the full dataset, eliminating the possibility of any SNP being a strongly influential factor. These analyses detected no obvious asymmetry or evidence that the observed association was driven by a single influential SNP.

Discussion

Keloids result from the complex wound healing process, characterized by the replacement of damaged skin with fibrous tissue, which can lead to various forms such as hypertrophic keloids and keloids, significantly impacting aesthetic and functional outcomes.^23,24 Understanding the etiological factors contributing to keloid formation is crucial for developing effective prevention and treatment strategies. Thus, in this two-sample Mendelian randomization study, we evaluated the potential relationship between 179 plasma lipid species and keloid formation. Our results suggested that genetically predicted higher levels of Phosphatidylcholine (O-16:0_16:0), a specific ether-linked phosphatidylcholine species, were associated with an increased risk of keloid formation in the IVW, MR Egger, Maximum likelihood, and Weighted median analyses. However, the Weighted mode and Simple-mode estimates did not reach nominal statistical significance. Therefore, this association should be interpreted as suggestive rather than definitive causal evidence.

From a biological perspective, several hypothetical mechanisms may explain this observed association. Phosphatidylcholine species are important components of cell membranes and may influence membrane structure, lipid signaling, and inflammatory responses.^25–27 In keloid biology, altered lipid composition could theoretically affect fibroblast activation, TGF-β-related signaling, oxidative stress, and extracellular matrix deposition, including collagen synthesis. However, these mechanisms remain speculative for Phosphatidylcholine (O-16:0_16:0), because most existing evidence is based on total phosphatidylcholine, phosphatidylcholine-containing membranes, or phosphatidylcholine-related enzymes rather than this specific ether-linked molecular species. Therefore, these interpretations should be extrapolated with caution, and further lipid-species-specific experimental studies are needed to determine whether and how Phosphatidylcholine (O-16:0_16:0) contributes to keloid formation.

Furthermore, we also found some similar studies confirming the lipids on keloid formation. This study employed two-sample MR, and identified a significant causal relationship between elevated triglyceride-rich very large VLDL levels and increased keloid formation.²⁸ Additionally, it also revealed that higher cholesterol-rich remnant lipoproteins may exert a protective effect against keloid development, further showing the complex interplay of lipid metabolism in keloid pathology. Furthermore, genetically predicted plasma lipid levels assessed through MR reveal significant positive associations between Sterol ester (27:1/18:2) and Diacylglycerol (18:1_18:3) with keloid formation, while Triacylglycerol (54:7) shows a negative correlation, underscoring distinct lipidomic profiles as potential therapeutic targets for keloid prevention and treatment.²⁹

Consistent with these findings, one notable aspect of our results is the identification of a specific lipid species potentially involved in keloid formation. Phosphatidylcholines are major glycerophospholipids that constitute a substantial component of eukaryotic cell membranes and play important roles in membrane architecture, signaling, and cellular metabolism.³⁰ Previous studies have shown that phosphatidylcholines can modulate inflammatory responses and intracellular signaling pathways, thereby influencing collagen synthesis and degradation.³¹ Based on these observations, we speculate that Phosphatidylcholine (O-16:0_16:0) may contribute to fibrosis and abnormal collagen deposition by altering membrane composition and fluidity, which in turn may affect fibroblast activation, proliferation, and migration. In addition, its downstream metabolites may regulate inflammatory responses and oxidative stress, potentially promoting profibrotic cytokines such as TGF-β and exacerbating fibrotic remodeling. Phosphatidylcholine may also influence extracellular matrix homeostasis by affecting the balance between collagen synthesis and degradation. Together, these mechanisms suggest that phosphatidylcholine-related signaling pathways may represent a novel avenue for understanding keloid pathogenesis and identifying potential therapeutic targets.

The association of Phosphatidylcholine with increased keloid risk suggests that lipid metabolism may influence fibroblast activity, inflammation, and extracellular matrix remodeling. This connection raises the possibility that targeting lipid metabolic pathways could serve as a novel therapeutic strategy to mitigate excessive scarring. Emerging evidence suggests that lipid metabolic reprogramming plays a role in the pathological processes of keloid formation, and targeting specific lipid molecules represents a promising intervention strategy. Lipidomic interventions have shown potential in regulating fibrosis and wound healing.³² For instance, lipid-based nanosystems have been found to improve wound healing by modulating inflammation and the proliferative phase of the healing process, and they represent effective strategies for promoting wound repair.³³ In burn wound treatment, acellular fish skin grafts have been shown to alter lipid mediator profiles, such as EPA and DHA monohydroxylated lipid mediators, which influence the early stages of wound healing and potentially affect scar formation.³⁴ These findings suggest that interventions targeting lipid metabolism and inflammatory pathways may provide novel therapeutic strategies for the prevention and treatment of keloids.

Another important discussion point is the implications of our findings in the context of existing literature on keloid formation. While previous MR analyses have focused on genetic factors related to immune responses and fibroblast activation, our study highlights the need to consider metabolic factors, particularly lipid profiles, in understanding keloid pathogenesis. Abnormalities in arachidonic acid synthesis and interactions between butyric acid, prostaglandin E2, and sphingosine 1-phosphate significantly contribute to keloid hyperfibrosis, indicating that targeting these lipid pathways may enhance therapeutic strategies for keloid treatment.³² The interplay between lipid metabolism and immune cell activity, such as that of macrophages and mast cells, could provide deeper insights into the multifactorial nature of keloid formation. For instance, Weighted gene co-expression network analysis and machine learning revealed legumain (LGMN) as a novel lipid metabolism-related biomarker for keloid, with high expression linked to M2 macrophage infiltration and immune response, suggesting its potential role in fibroblast proliferation and apoptosis inhibition as therapeutic targets for keloid treatment.³⁵ It is necessary to elucidate the mechanistic pathways linking lipid metabolism to immune modulation in keloid tissue development.

However, it is essential to recognize the limitations inherent in the two-sample MR design. Although the overall outcome sample size was large, the number of keloid cases was relatively limited, which may have reduced statistical power for lipid traits with low variance explained by their instruments. Therefore, the null associations should be interpreted cautiously, as they may reflect limited power rather than the absence of a causal relationship. The reliance on genetic variants as IVs may not fully capture the complexity of lipid metabolism and its interaction with environmental factors. For example, there is a report that identified DUSP1 and HOXA5 as central hub genes in keloid pathogenesis, with MR analysis revealing their opposing roles, while highlighting the critical involvement of mast cells and macrophages, and the IL17 signaling pathway as promising therapeutic targets.³⁶ The contrasting findings from MR analysis and the study sample outcomes suggest a complex interplay between genetic factors and environmental influences, indicating the need for further investigation into their differential roles in keloid formation. And, our study mainly focuses on individuals of European ancestry may also limit the broader applicability of the findings to varied populations. It should also be noted that, after adjustment for multiple comparisons, only one lipid species in our analysis retained statistical significance, which raises the possibility of inflation due to multiple testing or a type I error. Moreover, although the MR-Egger intercept test did not indicate evidence for directional horizontal pleiotropy, the potential presence of residual pleiotropy or reverse causation cannot be entirely ruled out. These issues show the necessity of validating the findings in independent cohorts with diverse ancestral backgrounds, while integrating longitudinal lipidomic profiling, clinical outcome data, and mechanistic investigations in experimental models to substantiate the associations and elucidate the underlying biological pathways. Future research should aim to replicate these findings in broader, more heterogeneous cohorts and explore the role of environmental and lifestyle factors in lipid metabolism and keloid formation.

Conclusion

Finally, our two-sample MR study suggested that genetically predicted higher levels of Phosphatidylcholine (O-16:0_16:0) may be associated with an increased risk of keloid formation. Although sensitivity analyses did not indicate substantial heterogeneity or directional horizontal pleiotropy, the findings should be interpreted cautiously given the mixed MR estimates and multiple-testing considerations. These results provide preliminary genetic evidence supporting a potential role of lipid metabolism in keloid pathophysiology and highlight the need for further validation and mechanistic studies. Interventions targeting lipid metabolic pathways would provide valuable approaches for inhibiting keloids.

Funding Statement

Public Hygiene and Health Commission of Shenzhen Municipality of Integration of Medical and Preventive Services with Traditional Chinese Medicine Group Project ([2019]No.25) and Natural Science Foundation of Hubei Province (Grant No. 2026AFB244).

Abbreviations

GWAS, genome-wide association studies; IgAN, IgA nephropathy; IVs, instrumental variables; IVW, inverse-variance weighted; LD, linkage disequilibrium; MR, Mendelian randomization; NTM, neurotrimin; OR, odds ratio; SNPs, single-nucleotide polymorphisms.

Data Sharing Statement

The GWAS summary statistics for plasma lipid species are publicly available from the GWAS Catalog under accession numbers GCST90277238–GCST90277416, and the keloid GWAS summary statistics are publicly available under accession number GCST90483368. All analytical methods involved in this study were implemented through custom-written R scripts. To ensure the reproducibility of the results, the complete code and datasets have been made publicly available in the following code repository: https://github.com/gongsongtongji/lipisome_mr_analysis/tree/main

Ethical Statement

This study was based solely on publicly available, de-identified GWAS summary statistics from the GWAS Catalog. The analysis did not involve individual-level data, identifiable private information, biological samples, direct participant contact, or any intervention. According to Article 32, items 1 and 2, of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects issued in China on February 18, 2023, research using legally obtained publicly available data or anonymized information that does not involve personal privacy or identifiable human information may be exempt from additional ethics review. Therefore, the present secondary analysis of publicly available, de-identified summary-level GWAS data was exempt from additional institutional ethical approval.

Disclosure

The authors declare no conflicts of interest.