Mosaic Loss of Chromosome Y, Smoking, and Risk of Symptomatic Peripheral Artery Disease: Insights From a Longitudinal Cohort Study

Abstract

Background

Mosaic loss of chromosome Y (mLOY) has been linked to several cardiovascular diseases, but its association with peripheral artery disease (PAD) remains unclear. We examined the relationship between mLOY and symptomatic PAD.

Methods

This longitudinal study included 217 047 men without prevalent PAD. mLOY was identified using the Mosaic Chromosomal Alterations pipeline. Incident symptomatic PAD was ascertained using International Classification of Diseases, Tenth Revision (ICD‐10), and Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 codes. Cox regression was used to evaluate the association between mLOY and PAD. We also performed exploratory mediation analyses of immune‐inflammatory biomarkers, examined the joint association of mLOY and smoking, and assessed the contribution of mLOY to sex‐related PAD disparities.

Results

During a median follow‐up of 13.6 years, 4910 men developed symptomatic PAD. mLOY was associated with a higher risk of symptomatic PAD (hazard ratio [HR], 1.25 [95% CI, 1.17–1.33]; P<0.001), with the strongest association observed for expanded mLOY (mosaic cell fraction ≥10%) (HR, 1.35 [95% CI, 1.24–1.47]; P<0.001). Exploratory analyses suggested that several immune‐inflammatory biomarkers were statistically related to the mLOY‐PAD association. Additive‐scale interactions were observed between mLOY and former or current smoking. Compared with nonsmokers without mLOY, current smokers carrying mLOY had the highest PAD risk (HR, 4.69). Compared with matched women, men with mLOY had an ≈1.74‐fold higher PAD risk.

Conclusions

mLOY was associated with symptomatic PAD, independently of other risk factors. These findings provide biologically informative evidence linking mLOY to symptomatic PAD and highlight smoking cessation as an important public‐health priority in men.

Nonstandard Abbreviations and Acronyms

mCF

mosaic cell fraction

mLOY

mosaic loss of chromosome Y

Clinical Perspective.

What Is New?

• This large‐scale longitudinal study suggests that mosaic loss of chromosome Y is independently associated with an increased risk of incident symptomatic peripheral artery disease, particularly among men with an expanded mosaic cell fraction.

• We observed an additive‐scale interaction between mosaic loss of chromosome Y and smoking status, with current smokers carrying mosaic loss of chromosome Y showing the highest risk of developing symptomatic peripheral artery disease.

What Are the Clinical Implications?

• These findings are biologically informative for understanding the contribution of acquired somatic mosaicism to peripheral artery disease risk in men and highlight the public health importance of smoking cessation in men.

Peripheral artery disease (PAD) is an atherosclerotic condition affecting >200 million people worldwide, with major adverse consequences, including amputation, functional impairment, poor quality of life, and increased risks of cardiovascular events and mortality. ¹ , ² , ³ Notably, PAD is an age‐related disease, with its prevalence increasing sharply from <5% before the age of 50 years to nearly 20% by the age of 80 years. ² Considering the effect of population aging and the severe consequences of PAD, identifying new risk factors for PAD and implementing effective interventions for high‐risk individuals are critical. ²

Mosaic loss of chromosome Y (mLOY) is a common somatic mutation in hematopoietic cells among male individuals. As male individuals age, a portion of their cells lose the Y chromosome, with >40% of men aged ≥70 years exhibiting detectable mLOY in hematopoietic cells. ⁴ Accumulating evidence suggests that mLOY is associated with decreased longevity and an increased risk of age‐related diseases, ⁵ including lung diseases, ⁶ Alzheimer disease, ⁷ and cardiovascular diseases, such as coronary artery disease, ⁸ heart failure, ⁹ aortic aneurysm, ¹⁰ valvular heart disease, ¹¹ and atrial fibrillation. ¹² , ¹³ An experimental study in mice has further revealed a causal link between mLOY and age‐related profibrotic pathologies. ⁹ These findings underscore the contribution of mLOY to age‐related diseases. However, the underlying association of mLOY with the risk of PAD has not been well studied. mLOY was associated with immune function alterations and blood cell differentiation, ¹⁴ , ¹⁵ , ¹⁶ but whether these immune‐inflammatory factors are statistically related to the association between mLOY and PAD remains unclear. In addition, smoking is a well‐established risk factor for both PAD and mLOY, ³ , ¹⁷ yet whether smoking and mLOY are jointly associated with excess PAD risk remains largely unknown.

Investigating the potential association between mLOY and PAD could provide insight into the contribution of acquired somatic mosaicism to PAD risk, as well as sex‐specific disparities in disease susceptibility. In this study, we used genotyping data from the UK Biobank to estimate the associations of mLOY with the risk of PAD and further explored the potential role of immune‐inflammatory markers. We also evaluated the joint effects of mLOY and smoking on PAD risk and assessed the potential modifying influence of smoking behavior. Finally, an exploratory analysis was conducted to investigate the contribution of mLOY to sex‐related disparities in PAD.

METHODS

UK Biobank resource data under application number 55177 were used for this research. The UK Biobank data are available on application (www.ukbiobank.ac.uk/).

Study Population

The UK Biobank is a large prospective cohort that recruited ≈500 000 adults from 22 assessment centers in England, Scotland, and Wales between March 2006 and October 2010. The study was approved by the North West Multicentre Research Ethics Committee, Manchester, United Kingdom (21/NW/0157), and all participants provided informed consent.

Genome‐wide array genotyping data were obtained from peripheral blood samples of 488 377 UK Biobank participants at the initial assessment visit. We excluded individuals who lacked data for mLOY calling, had incomplete quality control information, had mismatched genetic sex and self‐reported gender, had sex chromosome aneuploidy, had >10 third‐degree relatives identified or were excluded from the kinship inference process, were outliers for heterozygosity or missingness rates, or had PAD at baseline. In addition, we restricted the analysis to subjects with a sample call rate ≥0.97 and an autosomal B‐allele frequency deviation ≤0.03 to ensure high‐quality samples, leaving 217 047 participants (Figure S1).

Detection of mLOY Events

As previously described, ⁴ , ¹⁸ , ¹⁹ we identified mLOY carriers using the Mosaic Chromosomal Alterations caller. Briefly, the Mosaic Chromosomal Alterations caller applies a Viterbi hidden Markov model to integrate log2 R ratios and B‐allele frequencies derived from raw genotyping array intensity data, leveraging haplotype information to detect allelic imbalance within cell populations. The fraction of cells carrying mLOY (“mosaic cell fraction” [mCF]) was estimated using B‐allele frequency values in the pseudoautosomal region 1.

Ascertainment of Outcome

Incident clinically recognized PAD was defined by International Classification of Diseases, Tenth Revision (ICD‐10), codes and the Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (ICD‐10 codes: I700, I7000, I7001, I702, I7020, I7021, I708, I7080, I709, I7090, I738, and I739; Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4, codes: X093, X094, X095, L216, L513, L516, L518, L521, L522, L541, L544, L548, L591, L592, L593, L594, L595, L596, L597, L598, L601, L602, L631, L635, L639, and L667). In this study, hospital inpatient admission data were updated through October 31, 2022, for England, August 31, 2022, for Scotland, and May 31, 2022, for Wales. For each UK Biobank participant, person‐years of follow‐up were calculated from the date of their initial visit to the assessment center until the earliest occurrence of the outcome diagnosis, death, or the end of the follow‐up period. In cases where participants had multiple PAD‐related codes, the date of the first recorded code was assigned as the incident date.

Genetic Risk Score for PAD

The weighted genetic risk score was constructed based on 19 single‐nucleotide polymorphisms significantly associated with PAD at genome‐wide significance from a prior genome‐wide association study ²⁰ , ²¹ (Table S1). These genome‐wide association study findings were subsequently replicated and validated in the UK Biobank cohort. ²⁰ The genetic risk score exhibited a normal distribution, with higher values reflecting increased genetic susceptibility to PAD (Figure S2). According to the distribution of the genetic risk score, participants were categorized into tertiles of genetic risk (low, intermediate, and high risk).

Immune‐Inflammatory Biomarkers

Baseline peripheral blood counts of white blood cells (leukocytes), neutrophils, monocytes, lymphocytes, and platelets were included in this study. From these counts, we derived combined inflammation biomarkers, including the neutrophil/lymphocyte ratio (neutrophils/lymphocytes), lymphocyte/monocyte ratio (lymphocytes/monocytes), platelet/lymphocyte ratio (platelets/lymphocytes), systemic inflammation response index (monocyte×neutrophil/lymphocyte), and the systemic immune‐inflammation index (neutrophils×platelets/lymphocytes). Serum CRP (C‐reactive protein) at baseline was also measured.

Assessment of Covariates

Several variables were considered potential confounders, including age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/General Certificate of Secondary Education [GCSE] or equivalent, others), employment status (in paid employment or self‐employed or others), smoking status (never, previous, or current), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), International Physical Activity Questionnaire activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), high‐density lipoprotein cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. For diet, the healthy diet score ranged from 0 to 5. One point was assigned for each of the following 5 dietary components: daily vegetable intake of at least 4 tablespoons, daily fruit intake of at least 3 pieces, fish intake at least twice weekly, unprocessed red meat intake no more than twice weekly, and processed meat intake no more than twice weekly. A higher score reflected a healthier dietary pattern. The number and proportion of missing values for each covariate are reported in Table S2. Missing covariate data were multiply imputed using the “mice” R package. Missing covariates were imputed using stochastic regression models that included covariates associated with missingness, age, and self‐reported ethnicity, but not the exposure variable or the outcome variable, generating 5 imputed data sets with a maximum of 5 iterations per chain.

Statistical Analysis

Baseline characteristics are presented as mean±SD for continuous variables and percentages for categorical variables. Comparisons of baseline characteristics across the categories of mLOY were made using the Student t test for continuous variables or the Pearson χ² test for categorical variables.

Primary analyses examined the association of mLOY with PAD using Cox proportional hazards models. Hazard ratios (HRs) and 95% CIs were calculated, and the proportional hazards assumption was assessed by plotting Schoenfeld residuals over time. Models were adjusted for age, self‐reported ethnicity, assessment center, Townsend deprivation index, educational attainment, employment status, smoking status, alcohol intake frequency, International Physical Activity Questionnaire activity group, diet, sleep duration, body mass index, use of lipid‐lowering medication, high‐density lipoprotein cholesterol, cholesterol, diabetes status, hypertension status, genotype array, and genetic principal components 1 to 10. Given that mCF reflects the proportion of blood cells carrying mLOY, participants with mLOY were further classified as nonexpanded (mCF <10%) or expanded (mCF ≥10%) according to prior studies. ⁶ , ¹³ We also performed a restricted mean survival time analysis. Restricted mean survival time represents the average survival time within a specific (restricted) time frame and can quantify the absolute difference in restricted mean survival time between 2 groups without imposing any model assumptions. In addition, the interaction between the effects of mLOY and PAD genetic susceptibility was evaluated using the multiplicative interaction term.

To explore whether immune‐inflammatory biomarkers were statistically related to the association between mLOY and PAD, exploratory mediation analyses were conducted using the R mediation package. We selected the immune‐inflammatory biomarkers for the mediation analyses based on prior knowledge of potential causal pathways. ¹⁶ , ²² , ²³ Although we assumed causal relationships and temporal precedence among exposure, mediator, and outcome variables, these assumptions cannot be confirmed in this observational study. Therefore, the estimated indirect effects and proportions were interpreted as exploratory statistical estimates only. The mediation analysis is used only for exploratory purposes. The false discovery rate correction was applied to adjust for multiple testing.

To investigate the joint association between mLOY and smoking status on PAD, participants were divided into 6 groups, with nonsmokers without mLOY as the reference group. Multiplicative interaction between mLOY and smoking status in relation to PAD was assessed using an interaction term, whereas additive interaction was evaluated by calculating the relative excess risk attributable to interaction and the attributable proportion. We reported 97.5% CIs for relative excess risk attributable to interaction and attributable proportion to account for multiple testing.

In addition, we conducted an exploratory analysis to assess sex differences in PAD incidence and to evaluate whether mLOY statistically explained part of the male–female difference in PAD risk. Propensity score matching was applied to select female participants without PAD at baseline who had baseline characteristics similar to male participants (Data S1). In the propensity score–matched cohort, Cox proportional hazards models were first used to compare PAD incidence among men with mLOY, men without mLOY, and matched women. We then conducted a coefficient attenuation analysis in the matched cohort using 2 sequential Cox proportional hazards models. Model 1 included sex, comparing men with matched women, together with covariates. Model 2 additionally included an mLOY indicator that contrasted men carrying mLOY with the combined group of men without mLOY and matched women. The proportional attenuation in the coefficient for sex after adding the mLOY indicator was calculated. Bootstrap resampling with 1000 iterations was used to estimate CIs.

Several sensitivity analyses were conducted to verify the reliability of the results. First, Fine and Gray models were constructed to account for competing risks by all‐cause mortality. Second, we excluded participants who developed PAD within the first 3 years of follow‐up to minimize the potential influence of reverse causality. Third, to address potential residual confounding by age, Cox proportional hazards models were refitted using age as the time scale, with participants entering at baseline age and exiting at age at event or censoring. In addition, age was modeled using restricted cubic splines to flexibly account for potential nonlinear associations. Fourth, to assess the robustness of our findings to missing‐data handling, we excluded individuals with imputed covariates. We additionally performed a sensitivity analysis in which mLOY status, PAD event status, and follow‐up time were included as predictors in the imputation model for missing covariates. Fifth, to assess the robustness of PAD case ascertainment, we conducted several analyses using alternative outcome definitions, including PAD cases defined by Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4, codes, atherosclerosis‐specific PAD ICD‐10 codes, and exclusion of nonspecific codes. The detailed definition codes are provided in Table S3. We further conducted subgroup analyses stratified by age and several lifestyle factors (alcohol consumption, sleep duration, and physical activity).

All statistical analyses were conducted using R 4.1.0 software.

RESULTS

Baseline Characteristics

A total of 217 047 participants were included in the analysis, of whom 35 204 (16.2%) had detectable mLOY, consistent with the age‐related accumulation of this somatic alteration. Among mLOY carriers, the distribution of mCF is as shown in Figure S3, with most individuals showing relatively low levels of mosaicism. Specifically, 12 401 carriers (35.2%) were classified as having expanded mLOY. The mean±SD age of the study participants was 56.7±8.2 years (Table 1). Participants with mLOY were older (62.0±5.9 versus 55.7±8.2 years; P<0.001) and had a higher prevalence of diabetes and hypertension, along with greater use of lipid‐lowering therapies. Additionally, mLOY carriers were more likely to be current or former smokers compared with noncarriers (61.2% versus 48.4%; P<0.001).

Table 1.

Baseline Characteristics of the Study Population According to mLOY Status

Characteristic	Overall (N=217 047)	No mLOY (N=181 843)	mLOY (N=35 204)	P value*
Age, y	56.7±8.2	55.7±8.2	62.0±5.9	<0.001
White race	198 155 (91.3)	164 511 (90.5)	33 644 (95.6)	<0.001
Missing†	1225 (0.6)	1038 (0.6)	187 (0.5)
Center				0.013
England	192 855 (88.9)	161 417 (88.8)	31 438 (89.3)
Scotland	15 135 (7.0)	12 790 (7.0)	2345 (6.7)
Wales	9057 (4.2)	7636 (4.2)	1421 (4.0)
TDI	−1.3±3.1	−1.2±3.2	−1.4±3.1	<0.001
Missing	279	246	33
Education				<0.001
College or university degree	78 029 (36.0)	63 048 (34.7)	14 981 (42.6)
A levels/AS levels or equivalent/O levels/GCSEs or equivalent	62 505 (28.8)	53 067 (29.2)	9438 (26.8)
Other	73 593 (33.9)	63 325 (34.8)	10 268 (29.2)
Missing	2920 (1.3)	2403 (1.3)	517 (1.5)
Employment status				<0.001
In paid employment or self‐employed, others	131 448 (60.6)	117 081 (64.4)	14 367 (40.8)
Other	84 433 (38.9)	63 720 (35.0)	20 713 (58.8)
Missing	1166 (0.5)	1042 (0.6)	124 (0.4)
BMI, kg/m2	27.8±4.2	27.9±4.3	27.6±4.0	<0.001
Missing	962	822	140
PA				<0.001
Low	33 319 (15.4)	28 405 (15.6)	4914 (14.0)
Moderate	67 725 (31.2)	56 506 (31.1)	11 219 (31.9)
High	75 281 (34.7)	63 182 (34.7)	12 099 (34.4)
Missing	40 722 (18.8)	33 750 (18.6)	6972 (19.8)
Smoking status				<0.001
Never	106 248 (49.0)	92 790 (51.0)	13 458 (38.2)
Previous	83 003 (38.2)	67 383 (37.1)	15 620 (44.4)
Current	26 622 (12.3)	20 683 (11.4)	5939 (16.9)
Missing	1174 (0.5)	987 (0.5)	187 (0.5)
Alcohol intake frequency				<0.001
>3 or 4 times a week	104 757 (48.3)	88 880 (48.9)	15 877 (45.1)
Other	111 773 (51.5)	92 507 (50.9)	19 266 (54.7)
Missing	517 (0.2)	456 (0.3)	61 (0.2)
Healthy diet score	3.1±1.2	3.1±1.2	3.2±1.2	<0.001
Missing	3827	3309	518
Sleep duration, h/d				<0.001
7–8	145 667 (67.1)	121 950 (67.1)	23 717 (67.4)
>8	15 736 (7.3)	12 393 (6.8)	3343 (9.5)
<7	54 294 (25.0)	46 346 (25.5)	7948 (22.6)
Missing	1350 (0.6)	1154 (0.6)	196 (0.6)
Use of lipid‐lowering medication	49 051 (22.6)	38 358 (21.1)	10 693 (30.4)	<0.001
Missing	2251 (1.0)	1928 (1.1)	323 (0.9)
HDL	1.3±0.3	1.3±0.3	1.3±0.3	<0.001
Missing	25 906	21 619	4287
Cholesterol	5.5±1.1	5.5±1.1	5.4±1.1	<0.001
Missing	10 073	8281	1792
Diabetes status	14 828 (6.8)	12 253 (6.7)	2575 (7.3)	<0.001
Hypertension status	67 520 (31.1)	54 587 (30.0)	12 933 (36.7)	<0.001

Data are presented as mean±SD or number (percentage). BMI indicates body mass index; HDL, high‐density lipoprotein; mLOY, mosaic loss of chromosome Y; PA, physical activity; and TDI, Townsend deprivation index.

^* P values for differences in baseline characteristics between the mLOY and no mLOY groups were estimated using Student t test for continuous variables and Pearson χ² test for categorical variables.

^† Baseline characteristics shown in Table 1 are based on the original data before multiple imputation, whereas the main analyses were conducted using the multiply imputed data sets. The number and proportion of missing values for each covariate are reported in Table S2.

Cumulative Incidence and Association of mLOY With PAD

During a median follow‐up of 13.6 (interquartile range, 12.8–14.28) years, 4910 participants (2.3%) developed symptomatic PAD, with an incidence rate of 1.7 (95% CI, 1.6–1.8) events per 1000 person‐years. Among these incident cases, 4326 cases were ascertained via ICD‐10 diagnosis codes and 584 via Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4, procedure codes. The incidence of PAD was 1.9% in noncarriers, 4.0% in overall mLOY carriers, and highest (5.2%) in the expanded mLOY subgroup.

As shown in Figure 1A, compared with participants without mLOY, the HR for PAD was 1.25 (95% CI, 1.17–1.33; P<0.001) for participants with mLOY, after adjusting for confounding variables. When classifying mLOY by cell fraction, a dose–response relationship was observed between mLOY‐derived mCF levels and the risk of PAD (P for trend <0.001) (Figure 1B). Compared with noncarriers, nonexpanded mLOY carriers had an 18% higher risk of PAD (HR, 1.18 [95% CI, 1.09–1.28]; P<0.001), whereas expanded mLOY carriers exhibited a more pronounced 35% increased risk (HR, 1.35 [95% CI, 1.24–1.47]; P<0.001). Restricted mean survival time analysis also suggested that mLOY carriers have a shorter PAD‐free survival time, which decreases as mCF increases (Table S4).

Figure 1. Association between mLOY and PAD.

Cumulative incidence plots were constructed using the Kaplan–Meier method, stratified by mLOY status (A) or mCF levels (B). Forest plot of multivariable Cox proportional hazards regression analysis for the association of mLOY status (A) or mCF levels (B) with the PAD. mLOY was classified as nonexpanded (mCF <10%) or expanded (mCF ≥10%). HRs and 95% CIs are shown. Model was adjusting for age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/GCSEs or equivalent, or others), employment status (in paid employment or self‐employed or others), smoking status (never, previous, or current), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), IPAQ activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), HDL cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. HDL indicates high‐density lipoprotein; HR, hazard ratio; IPAQ, International Physical Activity Questionnaire; mCF, mosaic cell fraction; mLOY, mosaic loss of chromosome Y; PAD, peripheral artery disease; and Ref, reference.

In genetic analyses, we evaluated whether PAD genetic risk modified the association between mLOY and PAD, but no significant interaction was detected (P for interaction=0.697). We further performed joint analyses, and the cumulative incidence of PAD stratified by mLOY status and genetic risk is presented in Figure 2A. Compared with individuals with lower genetic risk and no mLOY, those with both higher genetic risk and mLOY had a markedly elevated risk of PAD (HR, 1.80 [95% CI, 1.62–2.00]; P<0.001) (Figure 2B).

Figure 2. Joint association of combined genetic risk and mLOY with incident PAD.

A, Cumulative incidence of PAD stratified by combined categories of polygenic risk score (low, intermediate, or high) and mLOY status. Curves were generated using the Kaplan–Meier method. B, Forest plot from a Cox proportional hazards model showing the joint associations of GRS and mLOY with PAD incidence. HRs and 95% CIs are shown. Model was adjusting for age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/GCSEs or equivalent, or others), employment status (in paid employment or self‐employed or others), smoking status (never, previous, or current), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), IPAQ activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), HDL cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. GRS indicates genetic risk score; HDL, high‐density lipoprotein; HR, hazard ratio; IPAQ, International Physical Activity Questionnaire; mLOY, mosaic loss of chromosome Y; PAD, peripheral artery disease; and Ref, reference.

Exploratory Mediation Analyses of mLOY on PAD

Individuals with mLOY displayed marked alterations in peripheral blood cell profiles compared with those without mLOY, including higher white blood cell, neutrophil, monocyte, and platelet counts, higher CRP, neutrophil/lymphocyte ratio, platelet/lymphocyte ratio, systemic immune‐inflammation index, and systemic inflammation response index, and lower lymphocyte count and lymphocyte/monocyte ratio (Figure 3A). After false discovery rate correction, regression models revealed that mLOY was associated with all immune‐inflammatory biomarkers, and exploratory mediation analyses suggested that several of these biomarkers were statistically related to the association between mLOY and PAD (Figure 3B and Table S5).

Figure 3. Mediation analysis of the association between mLOY and PAD.

A, Violin plots show the distribution of immune‐inflammatory biomarkers in individuals with and without mLOY. The plots represent values corresponding to the 95th percentile of all measurements for each biomarker. The Wilcoxon rank‐sum test was used to assess differences between groups. ***P<0.001; **P<0.01; *P<0.05. B, The associations between mLOY and immune‐inflammation indicators, immune‐inflammation indicators and PAD, as well as the proportion of mediation. Model was adjusting for age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/GCSEs or equivalent or others), employment status (in paid employment or self‐employed or others), smoking status (never, previous, or current), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), IPAQ activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), HDL cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. CRP indicates C‐reactive protein; HDL, high‐density lipoprotein; HR, hazard ratio; IPAQ, International Physical Activity Questionnaire; LMR, lymphocyte/monocyte ratio; mLOY, mosaic loss of chromosome Y; NLR, neutrophil/lymphocyte ratio; PAD, peripheral artery disease; PLR, platelet/lymphocyte ratio; SII, systemic immune‐inflammation index; and SIRI, systemic inflammation response index.

Joint Association of mLOY and Smoking With the Risk of PAD

The joint effect of mLOY and smoking status on the risk of PAD was assessed. As shown in Table 2, compared with nonsmokers without mLOY, we observed that current smokers carrying mLOY had the highest risk of PAD (HR, 4.69 [95% CI, 4.18–5.27]; P<0.001; 10‐year cumulative incidence, 5.36%). For the interactive effects, no evidence of multiplicative interaction between mLOY and smoking was observed (P for interaction=0.245). However, positive additive interactions were found between mLOY, former smoking (relative excess risk attributable to interaction, 0.54 [97.5% CI, 0.28–0.81]; attributable proportion, 0.24 [97.5% CI, 0.14–0.35]; synergy index, 1.80 [97.5% CI, 1.27–2.53]), and current smoking (relative excess risk attributable to interaction, 0.69 [97.5% CI, 0.08–1.30]; attributable proportion, 0.15 [97.5% CI, 0.03–0.27]; synergy index, 1.23 [97.5% CI, 1.03–1.48]) on the risk of PAD.

Table 2.

Joint Association and Interaction of mLOY and Smoking With the Risk of PAD (N=215 873)

Subgroup	Cases/Total, N	Incident cases per 1000 person‐years	10‐y Cumulative incidence, %	HR (95% CI)	P value	P interaction	RERI (97.5% CI)	AP (97.5% CI)	SI (97.5% CI)
Nonsmoker
No mLOY	938/92790	0.76	0.56	Reference
mLOY	233/13458	1.27	0.96	1.06 (0.91–1.23)	0.443
Former smoker							0.54 (0.28–0.81)	0.24 (0.14–0.35)	1.80 (1.27–2.53)
No mLOY	1652/67383	1.89	1.59	1.66 (1.53–1.80)	P<0.001	0.245
mLOY	721/15620	3.69	3.10	2.27 (2.05–2.51)	P<0.001
Current smoker							0.69 (0.08–1.30)	0.15 (0.03–0.27)	1.23 (1.03–1.48)
No mLOY	892/20683	3.39	2.89	3.96 (3.60–4.35)	P<0.001
mLOY	448/5939	6.29	5.36	4.69 (4.18–5.27)	P<0.001

A total of 1174 participants whose smoking status was unavailable were excluded. Model was adjusting for age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/GCSEs or equivalent, or others), employment status (in paid employment or self‐employed or others), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), International Physical Activity Questionnaire activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), high‐density lipoprotein cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. AP indicates attributable proportion; HR, hazard ratio; mLOY, mosaic loss of chromosome Y; PAD, peripheral artery disease; RERI, relative excess risk attributable to interaction; and SI, synergy index.

mLOY and Sex Differences in PAD

We also conducted an exploratory analysis to evaluate the role of mLOY in sex‐related disparities in PAD. A total of 124 418 male participants were matched 1:1 with female participants having similar baseline characteristics. Among them, 19 655 male participants had detectable mLOY. The distribution and standardized mean differences for baseline characteristics between male and female participants before and after matching were presented in Tables S6 and S7. The cumulative incidence of PAD was markedly elevated in men with mLOY compared with both men without mLOY and women (Figure 4). Compared with matched women, men without mLOY had a 1.42‐fold increased risk of PAD (HR, 1.42 [95% CI, 1.33–1.53]; P<0.001), whereas men with mLOY showed a further increase in PAD incidence (HR, 1.74 [95% CI, 1.58–1.91]; P<0.001) (Figure 4). In addition, the coefficient attenuation analysis suggested that mLOY statistically explained ≈12.91% (95% CI, 6.11%–20.51%; P<0.001) of the observed male–female difference in PAD incidence, indicating that most of the sex difference remained unexplained by mLOY.

Figure 4. Risk of PAD incidence during follow‐up among male and matched female participants.

Cumulative incidence of PAD was estimated using the Kaplan–Meier method and compared among 3 groups: (1) men with mLOY, (2) men without mLOY, and (3) a matched cohort of women. Forest plot of multivariable Cox proportional hazards models showing the HRs for PAD. HRs and 95% CIs are shown. Model was adjusting for age (continuous, years), self‐reported ethnicity (White or other ethnic backgrounds [including the following UK Biobank self‐reported groups: Mixed, Asian or Asian British, Black or Black British, Chinese, and Other ethnic group]), assessment center (England, Scotland, or Wales), Townsend deprivation index (continuous), educational attainment (college or university degree, A levels/AS levels or equivalent/O levels/GCSEs or equivalent or others), employment status (in paid employment or self‐employed or others), smoking status (never, previous, or current), alcohol intake frequency (daily or almost daily/3 or 4 times a week, once or twice a week/1 to 3 times a month/special occasions only/never), IPAQ activity group (low, moderate, or high), diet (continuous, 0–5 points), sleep duration (<7, 7–8, or >8 hours), body mass index (continuous, kg/m²), use of lipid‐lowering medication (yes or no), HDL cholesterol (continuous, mmol/L), cholesterol (continuous, mmol/L), diabetes status (yes or no), hypertension status (yes or no), genotype array (UK BiLEVE array or UK Biobank Axiom array), and genetic principal components 1 to 10. HDL indicates high‐density lipoprotein; HR, hazard ratio; IPAQ, International Physical Activity Questionnaire; mLOY, mosaic loss of chromosome Y; and PAD, peripheral artery disease.

Sensitivity and Subgroup Analyses

To verify the robustness of our results, sensitivity analyses were performed. The association between mLOY and PAD remained consistent after accounting for competing risks by all‐cause mortality (Table S8), excluding participants who developed PAD within the first 3 years of follow‐up (Table S9), further refitting the models using age as the time scale and modeling age with restricted cubic splines (Table S10), and applying different missing‐data approaches, including complete‐case analysis and an imputation model that additionally included mLOY status, PAD event status, and PAD follow‐up time as predictors for missing covariates (Table S11). The findings were robust and remained largely consistent across alternative definitions of PAD (Table S12). Moreover, the effect of mLOY on PAD was similar in different age groups (Table S13), physical activity groups (Table S14), and sleep duration groups (Table S15). An interaction between mLOY and alcohol consumption was observed for PAD (P‐interaction=0.019), with a stronger association observed in frequent drinkers (Table S16).

DISCUSSION

Results from this large longitudinal analysis of >217 000 men suggest that mLOY was associated with an increased risk of symptomatic PAD, with the risk increasing in proportion to higher levels of mCF. Exploratory mediation analyses suggested that several immune‐inflammatory biomarkers were statistically related to the association between mLOY and symptomatic PAD. We also observed positive additive interactions between mLOY and smoking on PAD risk. In addition, our findings suggest that mLOY partly contributes to sex‐related differences in PAD incidence.

mLOY is the most frequent somatic mutation observed in men and represents a form of clonal hematopoiesis, with prevalence increasing markedly with age. For many years, it was thought to represent a harmless feature of aging. However, more recent studies have revealed its links to reduced longevity and age‐related disorders, including Alzheimer disease, ⁷ age‐related pulmonary diseases, ⁶ and solid tumor. ²⁴ In the UK Biobank, higher levels of mLOY have also been linked to cardiovascular conditions, such as valvular heart disease, ¹¹ heart failure, ⁹ and atrial fibrillation. ¹² It has been shown that mLOY in mice led to lower left ventricular function, accumulation of fibroblasts in the myocardium, and increased diffuse cardiac fibrosis. ⁹ Furthermore, mLOY is independently associated with secondary major cardiovascular events in individuals with severe atherosclerosis, ²⁵ and has been linked to poor prognosis in vascular diseases, such as coronary artery disease and ischemic stroke. ²⁶ , ²⁷ Our study extends these investigations to symptomatic PAD, suggesting that mLOY may represent an acquired somatic marker associated with PAD risk. The absence of a significant interaction between mLOY and the PAD genetic risk score suggests that mLOY may contribute to PAD risk largely independently of inherited PAD susceptibility. This finding further suggests that acquired somatic mosaicism may represent an additional layer of vascular risk beyond germline genetic predisposition.

The mechanisms underlying the association between mLOY and PAD have been underexplored. It has been reported that leukocytes with mLOY are associated with the dysregulation of many autosomal genes, primarily involved in immune functions. ¹⁴ Proteomics research has revealed that the mLOY‐associated proteins are significantly overexpressed in pathways related to inflammation and immune regulation. ²⁸ In the present study, exploratory mediation analyses suggested that several immune‐inflammatory biomarkers, including neutrophil count and the systemic inflammation response index, were statistically related to the mLOY‐PAD association. Neutrophils play an important role in all stages of cardiovascular inflammation. ²³ Activated neutrophils release reactive oxygen species as well as various proinflammatory alarmins and proteases, leading to endothelial dysfunction. ²⁹ , ³⁰ Neutrophils also secrete neutrophil extracellular traps, which can thin the fibrous cap and promote the formation of rupture‐prone plaques. ²³ Furthermore, neutrophil extracellular traps are closely linked to thrombosis, suggesting a potential role in the pathogenesis of PAD. ³¹ , ³² In addition, the heritability‐enriched regions of LOY showed substantial overlap with (Friend leukemia integration 1) FLI1 transcription factor binding sites, which are fate‐determining factors that promote hematopoietic stem cell differentiation toward platelets rather than red blood cells. ¹⁵ Consistent with previous studies, ¹⁵ we observed a positive association between mLOY and platelet count. Platelet activation is elevated in patients with PAD, and its degree is positively associated with PAD severity. ³ , ³³ Interestingly, although mLOY was associated with elevated platelet count, consistent with FLI1‐related hematopoietic differentiation mechanisms, our mediation analysis did not support platelet count as a statistical mediator of the mLOY‐PAD association. This discrepancy warrants further investigation and may reflect limitations of baseline cross‐sectional biomarker analyses, the complexity of platelet activation beyond platelet count, or distinct roles of platelet biology at different stages of PAD development.

Notably, we observed an additive interaction between smoking and mLOY in relation to incident symptomatic PAD, indicating that individuals jointly exposed to smoking and mLOY constituted a high‐risk group from a public health perspective. The joint exposure to current smoking and mLOY drove the 10‐year cumulative incidence of PAD to its peak (5.36%), an almost 10‐fold increase compared with nonsmokers without mLOY (0.56%). Furthermore, former smokers exhibited a lower excess risk than current smokers, a pattern consistent with the public health importance of smoking cessation among men. From a preventive perspective, based on the observed absolute risk differences, a hypothetical smoking cessation scenario would correspond to ≈13 fewer incident symptomatic PAD cases per 1000 noncarriers and ≈23 fewer cases per 1000 mLOY carriers over 10 years. However, these estimates should be interpreted cautiously because our observational design cannot establish the causal effect of smoking cessation interventions. We also observed a nominal interaction between mLOY and alcohol consumption, with a stronger association between mLOY and PAD among frequent drinkers. Given prior evidence linking alcohol consumption to PAD risk and alcohol‐related oxidative stress, ³⁴ , ³⁵ this finding may have potential biological plausibility, although it should be interpreted cautiously because residual confounding cannot be excluded.

In addition, this study offers a biological perspective on sex differences in PAD. Most evidence based on intermittent claudication suggests PAD is more common in men, ³⁶ , ³⁷ whereas findings based on the ankle‐brachial index (ABI) are inconsistent, ³⁸ , ³⁹ , ⁴⁰ because of the more frequent presentation of atypical symptoms and a modestly lower intrinsic ABI in women. ⁴¹ , ⁴² Overall, more severe or symptomatic PAD tends to be more prevalent in men. In this context, men with mLOY had a higher incidence of clinically recognized PAD than both men without mLOY and matched women. The coefficient attenuation analysis suggested that mLOY statistically explained ≈13% of the male–female difference in PAD risk, indicating that mLOY may contribute modestly to sex disparities in PAD. However, most of the sex difference remained unexplained, suggesting that other factors, such as hormonal factors, lifestyle exposures, health care–seeking behavior, clinical recognition patterns, and other genetic or biological mechanisms, may also contribute to the observed sex disparity.

The advantages of this study lie in the large sample size and the longitudinal analysis of the relationship between mLOY and PAD, which fills the knowledge gap on mLOY in peripheral artery disease. In addition, our findings have public health relevance by reinforcing the importance of smoking cessation among men, particularly given the high PAD risk observed among smokers carrying mLOY. However, some limitations must be recognized. The major limitations of this study are related to baseline‐only assessment of mLOY, clinically recognized PAD ascertainment, and the exploratory nature of the mediation analyses. First, given mLOY was assessed from genetic data derived exclusively from baseline peripheral blood samples, we could not evaluate longitudinal changes within individuals. The proportion of mLOY increases with age, meaning some participants classified as noncarriers at baseline likely developed mLOY during follow‐up. This introduces nondifferential exposure misclassification, which typically biases HRs toward the null, suggesting our observed estimates are likely conservative. Therefore, the longitudinal changes of mLOY and their impact on PAD warrant further investigation. Second, PAD was identified using ICD‐10 diagnosis and procedure codes, which may have led to misclassification. Although this approach is susceptible to potential misclassification, we combined multiple data sources, including inpatient records and surgical procedures data, to capture a broad spectrum of PAD manifestations. Furthermore, any such misclassification is likely nondifferential relative to mLOY status, which would typically bias our HR estimates toward the null. We also conducted sensitivity analyses using alternative case definitions for PAD. In addition, ABI measurements were unavailable in this study. Therefore, our outcome likely reflects clinically recognized PAD and may preferentially capture symptomatic or more severe disease. Asymptomatic PAD may have been missed, leading to underascertainment and potentially lower incidence estimates. This limitation may also influence the interpretation of sex differences, because women may have more atypical symptoms or lower clinical recognition of PAD. Future research should explore the association between mLOY and PAD defined by ABI. Third, our mediation analyses should be interpreted cautiously. Although mLOY was assessed at baseline and PAD occurred during follow‐up, the immune‐inflammatory biomarkers were measured at the same baseline time point as mLOY. Therefore, temporal ordering between mLOY and these biomarkers could not be established. The estimated indirect effects reflect exploratory statistical associations and should not be interpreted as evidence of biological mediation. Future studies with repeated measurements of mLOY and inflammatory biomarkers are needed to clarify whether dynamic immune‐inflammatory changes lie on the pathway between mLOY and PAD development. Several additional considerations should also be noted. Because most participants included in this study were of White European ancestry, whether our findings apply to other ethnic groups requires further investigation. In addition, the UK Biobank is subject to healthy volunteer bias, as participants tend to be healthier, more educated, and more health conscious than the general population. This may limit the generalizability of our incidence estimates and may have influenced the magnitude of the observed associations. Finally, although we adjusted for a broad range of demographic, lifestyle, and clinical covariates, residual confounding cannot be fully excluded. For example, detailed information on smoking intensity, pack‐years, smoking duration, and time since smoking cessation was not fully captured, which may have influenced the observed joint association and additive interaction estimates for smoking and mLOY.

Future studies should validate these findings in more diverse populations and in cohorts with systematic ABI assessment to capture both symptomatic and asymptomatic PAD. Repeated measurements of mLOY are also needed to assess its temporal dynamics and determine whether changes in mLOY burden over time are associated with PAD risk. In addition, mechanistic studies are warranted to clarify whether immune‐inflammatory alterations, oxidative stress, thrombosis, or other biological processes link mLOY to PAD development. Future work integrating mLOY with other clonal hematopoiesis markers may provide a more comprehensive understanding of acquired somatic mosaicism in vascular disease. Furthermore, standardized and clinically feasible methods for detecting and quantifying mLOY are needed before its potential translational application can be considered. Finally, predictive modeling studies are needed to determine whether mLOY provides incremental value beyond established PAD risk factors before its potential clinical utility can be evaluated.

In summary, mLOY was associated with an increased risk of symptomatic PAD, with the risk increasing in proportion to higher levels of mCF, providing new insight into the pathogenesis of PAD.

Sources of Funding

This research was supported by the National Natural Science Foundation of China (X.T., 82570128), the Development Project of Zhejiang Province (X.T., 2025C01119), and the Rut and Arvid Wolff Memorial Foundation (X.T., 2023‐02467). The funders had no role in the conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Disclosures

There are no conflicts of interest.

Supporting information

Supplemental Methods S1.

Tables S1–S16.Figures S1–S3.

STROBE Statement—Checklist S2.