Subjects
This retrospective longitudinal study was conducted at the physical examination center of the geriatric department of Tongji Hospital and was approved by the medical ethics committee of Tongji Hospital (TJ-IRB20190410). The study protocol conforms to the Declaration of Helsinki. We collected physical examination records from March 2011 to July 2019. In total, 612 healthy males aged 18 years or older were selected. All of them had measurements of baPWV and a complete lipid panel (including TC, TG, LDL-C, and HDL-C) at baseline and had their second baPWV measurement after a delay of more than three years. The exclusion criteria were: inheritable dyslipidemia, use of lipid-lowering medications, hypertension [defined as systolic blood pressure (SBP) ≥ 140 mm Hg, diastolic blood pressure (DBP) ≥ 90 mm Hg, or use of antihypertensive drugs], diabetes [defined as fasting blood glucose (FBG) of ≥ 7.0 mmol/L, HbA1c ≥ 6.5%, or use of hypoglycemic drugs], coronary disease, stroke, obvious arrhythmia (persistent atrial fibrillation, frequent premature beats, or wearing a pacemaker), cardiomyopathy, valvular heart disease, chronic liver or kidney disease, cancer, ankle-brachial index (ABI) less than 0.9, and missing data.
Clinical characteristics
Every time the patients visited the physical examination center, trained personnel conducted standardized in-person interviews with the patients to collect information regarding age, sex, current cigarette smoking, medical history, and medication use. Anthropometric indexes including height and weight were measured. Body mass index (BMI) was computed as the weight in kilograms divided by the square of the height in meters. Blood pressure and heart rate were measured using an OMRON sphygmomanometer (OMRON Corporation, Japan). The blood pressure and heart rate used in the analysis were calculated as the average of three measured values. Mean arterial pressure (MAP) was calculated from the standard equation MAP = (2/3)DBP + (1/3)SBP (in mm Hg). Fasting venous blood samples were collected and sent to the hospital’s clinical chemistry laboratory. TC, TG, LDL-C, HDL-C, FBG, HbA1c, creatinine, and uric acid were measured using standard certified assays. We also calculated values for non-HDL-C (TC minus HDL-C), TC/HDL-C (TC divided by HDL-C), TG/HDL-C (TG divided by HDL-C), LDL-C/HDL-C (LDL-C divided by HDL-C), and non-HDL-C/HDL-C (non-HDL-C divided by HDL-C).
Arterial stiffness measurements and definition of the outcomes
BaPWV and ABI were measured using the Vascular Profiler BP-203RPEⅢ system (Omron, Kyoto, Japan). Trained technicians placed the pressure cuffs on the subjects, i.e., one on the upper part of each arm and one on each ankle. Then, the subjects were examined after ten minutes of rest in the supine position. The device simultaneously recorded the bilateral pulse waves of the brachial and posterior tibial arteries using an oscillometric method. BaPWV was calculated as the ratio of the traveled distance (which was automatically estimated from the body height) divided by the transit time of the pulse wave between the brachial and posterior tibial arteries. We classified the outcome in 2 different manners. Outcome 1: the baseline and follow-up baPWV were both divided into quartiles, respectively. Then we classified subjects into two subgroups: those who decreased their quartile distribution or persisted within the two lower quartile groups, and those who increased their quartile distribution or persisted within the two higher quartile groups [13]. Outcome 2: a cutoff value of more than 1400 cm/s for baPWV was used to diagnose arterial stiffness [4,14]. Individuals without arterial stiffness at baseline were divided into “non-arterial stiffness” and “incident arterial stiffness” groups based on their follow-up baPWV levels.
Statistical analysis
Data were analyzed using R and RStudio 3.6.2. Continuous variables were presented as the means ± standard deviation or medians (interquartile range), as appropriate for the distribution. Categorical variables were shown as counts and proportions. Paired t-tests were used to determine if the baPWV levels changed over the follow-up period. We compared the baseline variables between groups using unpaired t-test, Mann-Whitney U test, and Chi-squared test accordingly. Crude and multivariable-adjusted binary logistic regression models were developed to estimate odd ratios (ORs) and 95% confidence intervals (CIs) for the two outcomes associated with each serum lipid, respectively. Prior to regression analysis, TG and TG/HDL-C were log-transformed to achieve normality. Age and MAP, the most important determinants for baPWV, were controlled for in the first model. Further adjustments were made for BMI, smoking (current smoker and non-current smoker), heart rate, FBG, HbA1c, creatinine, and uric acid in the second model. The improvement in the ability of each serum lipid to predict outcomes was summarized using area under receiver operating characteristic (ROC) curves (AUC), positive net reclassification improvement (NRI), and integrated discrimination improvement (IDI) [15]. In sensitivity analyses, TC, TG, HDL-C, LDL-C, and non-HDL were divided into three or two levels by clinical cut-points [16]; TC/HDL-C, TG/HDL-C, LDL-C/HDL-C, and non-HDL-C/HDL-C ratios were divided into three levels by tertile cut-points. ORs and 95% CIs for the outcomes across categories of each serum lipid were also calculated, using the most favorable category as the reference. Trends in ORs across categories of each lipid were calculated by modeling the lipid categories as an ordinal variable. Two-tailed p-values < 0.05 were considered significant.