Study population
The ARIC Study is a prospective cohort study that enrolled 15792 participants aged 45 to 64 years from 4 US communities (Forsyth County, North Carolina; Jackson, Mississippi; eight northern suburbs of Minneapolis, Minnesota; and Washington County, Maryland), aimed at investigating the natural history, etiology, and clinical manifestations of atherosclerotic disease in black and white men and women. The participants were recruited between 1987 to 1989 (visit 1). Four subsequent visits were conducted: visit 2 (1990-1992), visit 3 (1993-1995), visit 4 (1996-1998), and visit 5 (2011-2013). The details about the study design have been previously described [16]. Written informed consent was obtained from all ARIC participants, and the ARIC study was approved by the institutional review boards at each site.
We excluded participants who had PAD diagnosis at baseline (n=613); those who had missing data regarding PAD (n=555); and those who had missing data regarding other covariates of interest (n=1785). We also excluded participants who had no follow-up information on PAD (n=266). This resulted in a final sample of 12573 participants for the analysis of association between baseline TyG index and incident PAD. We further excluded those participants with fewer than three valid TyG index during follow-up visits; the remaining 9296 participants were included in the analysis of association between TyG index group-based trajectory and incident PAD (Figure 1).
Data collection at baseline
Trained interviewers collected information using standardized questionnaires on demographic, lifestyle, and detailed medical information at visit 1. Age, sex, race, educational level, physical activity, alcohol consumption, and smoking status were self-reported. Educational attainment was categorized as basic (less than high school), intermediate (high school graduate or vocational school), and advanced (college, graduate school, or professional school). Smoking and alcohol drinking status were classified as current, former, or never. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Seated blood pressure represented the mean of the last two of three measurements using a random-zero sphygmomanometer after a 5-minute rest. Hypertension was defined as systolic blood pressure readings ≥140 mmHg or diastolic blood pressure readings ≥90 mmHg, or use of antihypertensive drugs in the previous two weeks. Diabetes was defined as fasting glucose level ≥126 mg/dL (≥7.0 mmol/L), non-fasting glucose level ≥200 mg/dL (≥11.1 mmol/L), self-reported physician diagnosis of diabetes, or use of antidiabetic drugs. Prevalent cardiovascular diseases such as coronary heart disease and stroke were determined according to both participants’ self-report and measurements at visit 1. Blood samples were obtained from participants who were asked to fast for 12 hours and stored at -70°C according to standardized protocols until laboratory analysis [17]. Plasma total cholesterol, high-density lipoprotein (HDL) cholesterol, and triglycerides were measured using automated enzymatic procedures, and low-density lipoprotein (LDL) cholesterol was calculated using the Friedewald equation when the concentration of triglycerides is < 400 mg/dL [18]. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration creatinine equation [19]. Medications were determined through self-reported usage in the previous two weeks and inspection of medication containers that participants brought to the visit.
The triglyceride-glucose (TyG) index was calculated as ln(fasting triglycerides [mg/dL]×fasting glucose [mg/dL]/2). Group-based trajectory analysis is designed to identify clusters of individuals with similar patterns of change over time. Trajectory groups were identified and then qualitatively examined and named to describe a visual pattern of change.
Ascertainment of incident PAD
Based on previous literature [8, 20, 21], PAD-related hospitalizations were identified by the following International Classification of Diseases Ninth Revision (ICD-9) discharge codes: peripheral vascular disease, unspecified (443.9); atherosclerosis of native arteries of the extremities, unspecified (440.20); atherosclerosis of native arteries of the extremities with intermittent claudication (440.21); atherosclerosis of native arteries of the extremities with rest pain (440.22); atherosclerosis of native arteries of the extremities with ulceration (440.23); atherosclerosis of native arteries of the extremities with gangrene (440.24); other atherosclerosis of native arteries of the extremities (440.29); atherosclerosis of bypass graft of the extremities (440.3); chronic total occlusion artery extremities (440.4); atherosclerosis of other specified arteries (440.8); coexisting leg amputation (84.11, 84.12, 84.15, 84.17); leg artery revascularization (38.18, 39.25, 39.29, 39.50); lower extremity ulcer and gangrene (707.1x). Incident PAD was defined as if a documented PAD-related hospitalization or a measured ankle-brachial index <0.90 during follow-up visits in patients not diagnosed with PAD at visit 1 [3]. Critical limb ischemia (CLI), the severe form of PAD, was based on the discharge codes (84.11, 84.12, 84.15, 84.17, 707.1x).
Statistical Analysis
Normally distributed continuous data were expressed as mean±SD, and the non-normally distributed continuous data, otherwise, were expressed as the median (interquartile range). Categorical data were expressed as numbers (percentage). Differences among groups were evaluated using analysis of variance (ANOVA) or Kruskal-Wallis h-test when appropriate for the continuous variables, and the χ2 test for the categorical variables. The follow-up period was set as the time from visit 1 (baseline) to the incident of PAD, or loss to follow-up, or September 30, 2015, whichever came first. Kaplan-Meier estimates were used to compute cumulative incidence of incident PAD by TyG index quartiles and the differences in estimates were compared using the log-rank procedure. Cox proportional hazards regression model was used to calculate hazard ratios and 95% CIs between TyG index and time to incident PAD. Three multivariate models with progressive degrees of adjustment were used to adjust for potential confounders. Model 1 adjusted for age, sex, and race. Model 2 further adjusted for antihypertensive drugs, body mass index, coronary heart disease, cholesterol-lowering drugs, diastolic blood pressure, diabetes, drinking status, education level, hypertension, physical activity during leisure time, systolic blood pressure, smoking status, sport during leisure time, stroke. Model 3 was additionally further adjusted for eGFR, insulin, LDL cholesterol and total cholesterol. We further used a restricted cubic spline regression model with 3 knots to assess the nonlinear dose-response association between TyG index and incident PAD. Subgroup analyses were performed stratifying by age, sex, race, smoking status, body mass index, hypertension, and diabetes at baseline, respectively. Moreover, we used latent class models to identify different patterns of longitudinal change in TyG index during follow-up. Models were fit using R 3.6.1 based on R package tidyLPA. Group-based trajectory analysis was designed to identify clusters of individuals with similar patterns of change over time [22]. The optimal number of trajectory groups was determined using a combination of Bayesian information criterion and number of observations in each group. Participants were assigned to the trajectory group for which they had the greatest posterior predictive probability. To estimate the association of trajectory groups with incident PAD, trajectory groups was included as an independent variable in a logistic regression model examining predictors of follow-up incident PAD.
All analyses were conducted in SPSS version 23 (SPSS, Inc, Chicago, Illinois). A two-sided P value of < 0.05 was considered statistically significant.