In this retrospective cross-sectional study, we examined data from 2,827 participants in the National Health and Nutrition Examination Survey (NHANES) conducted in the United States from 1999 to 2004. NHANES collects a variety of health-related information, including demographic details, physical examination results, and laboratory findings (23). The data collection was approved by the ethics review board of the National Center for Health Statistics, with all participants providing written informed consent before their inclusion (24). The NHANES data can be accessed publicly through their website (http://www.cdc.gov/nchs/nhanes.htm).

The NHANES database is highly suitable for investigating PAD, as it provides standardized ankle-brachial index (ABI) measurements—a validated and widely accepted diagnostic criterion for PAD in epidemiological studies—alongside a comprehensive set of demographic, examination, and laboratory variables, enabling robust adjustment for potential confounders.

Of the initial 31,126 participants, we first excluded those younger than 40 years (n = 21,156) because ABI measurements were not performed in this age group. Further exclusions included participants with missing bilateral ABI data (n = 2,408), ABI values >1.4 (n = 65) (25, 26), or missing monocyte/lymphocyte counts (n = 248). After also removing individuals with incomplete data on covariates (education, marital status, income, physical activity, smoking, BMI, total cholesterol, CRP, and HbA1c), 6,118 participants were retained for the primary analysis. To improve national representativeness, we further excluded those with missing fasting weight data (n = 3,291), resulting in a final weighted sample of 2,827 participants.

ABI measurements were conducted on participants aged 40 years and older, following a specific methodology established in previous studies. After a brief rest period, systolic blood pressure readings were obtained from the right arm and both ankles. In cases where right arm measurements were unavailable, the left arm was utilized instead. Participants aged 40–59 had their systolic blood pressure measured twice, whereas those aged 60 and above were assessed once. The ABI was calculated by dividing the average systolic blood pressure of the ankle by that of the arm.

PAD was defined as an ABI of less than 0.9 in either leg, consistent with established epidemiological criteria (23, 27). The NHANES examination did not include assessments for upper limb artery disease. To account for medial arterial calcification—a common source of artifact in individuals with diabetes that causes incompressible arteries and falsely elevated ABI readings—we excluded participants with an ABI value greater than 1.4 in either leg (23). This standard practice enhances the diagnostic specificity of a low ABI for obstructive PAD within the study sample.

MLR is calculated by dividing the monocyte count by the lymphocyte count, with both values readily available from laboratory data files.

We employed standardized surveys to collect socio-demographic and lifestyle data. The Mobile Examination Center provided examination outcomes, which included measurements such as body mass index (BMI), blood pressure, and various biochemical markers. Based on existing literature (27, 28–36), we examined several potential factors, including age, gender, ethnicity, education level, marital status, family income, smoking habits, physical activity levels, BMI, and laboratory test results (such as total cholesterol, C-reactive protein, and glycated hemoglobin A1c). Additionally, we considered pre-existing medical conditions, including diabetes, hypertension, and cardiovascular disease (CVD).

For more detailed information on these variables, including the measurement methods used in the questionnaires and a complete list of variables, please refer to the official NHANES website (https://www.cdc.gov/nchs/nhanes/). Family income was categorized into three groups based on the poverty income ratio (PIR): low (PIR ≤ 1.3), medium (PIR > 1.3–3.5), and high (PIR > 3.5). Smoking status was categorized according to definitions from previous research: never smoked, current smoker, and former smoker. To assess physical activity levels, participants reported on vigorous exercises that significantly increased breathing and heart rate (e.g., swimming or fast cycling) and moderate activities that caused mild to moderate physiological responses (e.g., golfing or leisurely biking). Each activity was required to have lasted at least 10 min in the past month. Physical activity levels were classified into three categories: below moderate (no moderate or vigorous activities), moderate (no vigorous exercise but at least one moderate activity), and high intensity (at least one instance of vigorous exercise).

Hypertension was defined as having an average systolic blood pressure (ASBP) or average diastolic blood pressure (ADBP) of 140/90 mmHg or higher, currently using antihypertensive medication, receiving a diagnosis from a healthcare provider, or meeting any combination of these criteria. Diabetes was characterized by a fasting glucose level exceeding 7 mmol/L, a random glucose level of 11.1 mmol/L or more, a glycated hemoglobin A1c of 6.5% or higher, the use of glucose-lowering medications, or a documented history of diabetes. The presence of cardiovascular diseases was assessed based on self-reported experiences with conditions such as congestive heart failure, coronary heart disease, angina pectoris, myocardial infarction, and stroke. BMI was calculated using the standard formula that involves an individual's weight and height.

In alignment with NHANES analytical recommendations, this secondary analysis accounted for the complex survey design by integrating sampling weights, stratification, and clustering variables to generate nationally representative estimates and minimize Type I error. Data from household interviews and Mobile Examination Center (MEC) examinations were combined, with fasting subsample weights (WTSAF2YR) applied to adjust for differential probabilities of selection and survey nonresponse. The survey design was implemented using the svydesign function in R, incorporating stratification (SDMVSTRA), primary sampling units (SDMVPSU), and appropriate weight adjustments to reflect the U.S. non-institutionalized population aged 40 years and older.

Categorical variables are summarized as unweighted counts with weighted proportions, and continuous variables as means with standard errors. Between-group comparisons employed weighted Student's t-tests for normally distributed continuous data, Mann–Whitney U tests for non-normally distributed variables, and chi-square tests for categorical variables. Covariates were selected based on clinical plausibility, univariate significance, and a pre-specified >10% change-in-estimate criterion for the MLR-PAD association. Variables meeting any of these criteria were included in final models. Supplementary Table S2 provides a complete assessment of each covariate's confounding influence and multicollinearity (VIF < 2.0), validating the approach.The relationship between MLR and PAD was evaluated through weighted multivariable logistic regression using a hierarchical adjustment strategy across five sequential models: Model 1 (unadjusted); Model 2 (adjusted for gender, race, education, marital status, poverty-income ratio, and physical activity); Model 3 (Model 2 + BMI, total cholesterol, CRP, HbA1c, and smoking status); Model 4 (Model 3 + history of cardiovascular disease, hypertension, and diabetes); Model 5 (Model 4 + age).

To examine potential nonlinearity, weighted restricted cubic spline (RCS) regression with four knots was fitted. If the nonlinearity test was significant (P < 0.05), a threshold-effect analysis was conducted using a two-piecewise logistic model, with the inflection point determined by maximum likelihood estimation. The log-likelihood ratio test compared the nonlinear model with a linear specification.

Subgroup analyses were performed to assess effect modification by age (<65 vs. ≥65 years), sex, physical activity level, family income, education, and comorbidities (cardiovascular disease, hypertension, diabetes). Heterogeneity across subgroups was evaluated with multivariate logistic regression, and interaction effects were tested via likelihood ratio tests.

Statistical analyses were performed using R Statistical Software (Version 4.2.2, http://www.R-project.org, The R Foundation) and the Free Statistics Analysis Platform (Version 2.1, Beijing, China, http://www.clinicalscientists.cn/freestatistics). A detailed explanation of their functionalities ensured transparency and reproducibility of the analytical methods used in this study. Overall, the statistical analysis appears to be comprehensive, well executed, and supported by appropriate software tools, contributing to the strength of the study's findings.

In total, 31,126 individuals participated in the in-home interviews. Finally, 2,827 individuals were included in the study (Figure 1). The study included eligible individuals aged 40 years or older and had a prevalence rate of PAD at 3.96%. Table 1 summarizes the weighted baseline characteristics of 2,827 participants stratified by tertiles of the MLR. Participants in the highest MLR tertile (T3) were significantly older, predominantly male, and exhibited elevated cardiovascular risk profiles compared to those in T1 (all p < 0.001). Notably, the prevalence of PAD increased across MLR tertiles, with T3 showing the highest rate (5.87% vs. 2.21% in T1, p < 0.001). Significant differences were also observed in race, inflammatory markers, smoking status, and comorbidities, whereas distributions of education level, marital status, income, and physical activity were comparable across groups. These findings suggest that higher MLR levels are associated with older age, male sex, and a more adverse cardiometabolic phenotype.

The univariate analysis revealed significant associations between various factors, including age, marital status, smoking habits, family income, physical activity level, education level, total cholesterol levels, CRP levels, monocyte count, HbA1c levels, CVD presence, hypertension, and diabetes with the occurrence of PAD (Table 2). Incorporating NHANES sampling weights, we evaluated the association between MLR and PAD across five adjustment models. In the weighted analysis, MLR as a continuous variable showed a strong association with PAD in unadjusted and partially adjusted models (e.g., Model 4: OR = 4.00, p = 0.006), but this relationship was attenuated and became non-significant after full adjustment including age (Model 5: OR = 1.99, p = 0.143). When analyzed categorically, participants in the highest MLR tertile (T3) maintained a significantly elevated odds of PAD even in the fully adjusted model (Model 5: OR = 2.25, p = 0.008), with a significant trend across tertiles (p = 0.008). These weighted results indicate that while elevated MLR is associated with PAD, its effect is substantially confounded by age and other clinical factors (Table 3).

Figure 2 presents the weighted restricted cubic spline analysis evaluating the dose-response relationship between the MLR and PAD prevalence. After adjusting for demographic, lifestyle, anthropometric, biochemical, and clinical covariates, the analysis revealed a nonlinear association (P for non-linearity = 0.025). The odds ratio (OR) for PAD showed a gradual increase with rising MLR values, particularly above the reference point of 0.28. However, the overall association did not reach statistical significance (P for overall = 0.054). These results suggest a complex, non-linear relationship between MLR and PAD risk in this nationally representative sample, with the inflammatory marker exhibiting a threshold effect rather than a simple linear trend. It is important to note that the confidence intervals widen substantially at both the lower and upper extremes of the MLR distribution, reflecting fewer observations and greater uncertainty in these ranges. Consequently, the precise shape of the curve at the tails should be interpreted with caution.

Table 4 presents the threshold analysis of MLR in relation to PAD based on weighted restricted cubic spline modeling. Using an MLR cut-point of 0.343 derived from the RCS curve, the analysis demonstrated a significant association with PAD at lower MLR levels (MLR < 0.343: OR = 1.62, 95% CI 1.24–2.11, p < 0.001). In contrast, no statistically significant association was observed at higher MLR values (MLR ≥ 0.343: OR = 0.94, 95% CI 0.64–1.38, p = 0.738). The log-likelihood ratio test confirmed the presence of a significant threshold effect (p = 0.009). These findings suggest that the association between MLR and PAD follows a nonlinear pattern, with increased risk primarily observed below the threshold of 0.343.

Figure 3 presents weighted subgroup analyses evaluating the MLR–PAD association across demographic and clinical variables. The fully adjusted OR for the overall cohort was 1.99 (95% CI: 0.78–5.07). Subgroup-specific estimates remained generally consistent, with no statistically significant variation observed by age, sex, education, physical activity, income, hypertension, cardiovascular disease, or diabetes status (all P-interaction > 0.05). These findings suggest that the relationship between MLR and PAD—while attenuated after full adjustment—does not differ substantially across key population subgroups in this nationally representative sample.