During 2002 to 2006, a population-based sample of 1,022 men aged 40–49 was randomly selected by three centers of the ERA-JUMP Study: 310 White Americans and 107 African Americans from Allegheny County, PA, U.S.; 303 Japanese-Americans from Honolulu, HI, U.S.; 302 Koreans from Ansan, Gyeonggi-do, South Korea [19‐22]. Originally, the ERA-JUMP study consisted of a population sample of the Japanese population whose PWV measures had not been examined. All participants were without clinical cardiovascular disease or other severe diseases [19]. For the purpose of recruiting healthy participants for the present study, we excluded those with peripheral arterial disease, hypertension, diabetes, or hyperlipidemia. Of the original samples, we excluded (1) those having ≤ 0.9 of either left or right ankle-brachial index (n = 7), (2) those taking antihypertensive, diabetes, or lipid-lowering medications (n = 166); and (3) those with values missing for all four regional PWVs (i.e., hfPWV, cfPWV, faPWV, and baPWV) (n = 62). The final sample was 784 (232 White Americans, 68 African Americans, 202 Japanese-Americans, and 282 Koreans).

Written informed consent was obtained from all participants. The study was approved by the Institutional Review Boards of the following institutions: the University of Pittsburgh, Pittsburgh, PA, U.S.; the Kuakini Medical Center, Honolulu, HI, U.S.; and Korea University, Seoul, South Korea.

The PWV was automatically generated using a noninvasive and automated waveform analyzer (Colin-VP2000/1000, Omron, Japan/WaveNexus,TX). This device records the electrocardiogram (ECG), phonocardiogram, and pressure waveforms. ECG electrodes were placed on both wrists. A phonocardiogram was placed on the left edge of the 4^th-rib-case; heart sounds S₁ and S₂ were detected. Pressure waveforms of the brachial and tibial arteries were recorded by a plethysmographic sensor and an oscillometric pressure sensor using the occlusion/sensing cuffs adapted to both arms (i.e., brachial arteries) and both ankles (i.e., tibial arteries). Pressure waveforms of the carotid and femoral arteries were recorded using multiarray tonometry sensors placed at the left carotid and the left femoral arteries. Following 10 minutes of rest in a supine position, pressure waveforms were stored for a sampling time of 10 seconds with automatic gain analysis and quality adjustment.

PWV (cm/second) was calculated as (path length between arterial sites)/(time interval). The path length for cfPWV was obtained by using the actual distance of participants’ body (in cm): ([cm from suprasternal notch to bottom of umbilicus] + [cm from botton of umbilicus to femoral sensor]) – [cm from carotid to suprasternal notch]. Furthermore, the path length for the heart-brachial (Dhb), the heart-femoral (Dhf), the heart-ankle (Dha), the femoral-ankle (Dfa), and the brachial-ankle (Dha – Dhb) segments were obtained using the following formulas: Dhb = 0.2195×height – 2.0734; Dhf =0.5643×height – 18.381; Dha = 0.8129×height + 12.328; Dfa = 0.2486×height – 30.709. The PWV results were obtained with the left sides of cf, ba and fa sites. Reproducibility of the automated PWV measurement was assessed by the University of Pittsburgh Ultrasound Research Laboratory, and used 19 and 12 participants for obtaining intraclass and interclass correlation coefficients, respectively. Intraclass correlation coefficients of 0.86 (hfPWV), 0.76 (cfPWV), 0.96 (faPWV), and 0.97 (baPWV) were obtained within technicians. Interclass correlation coefficients of 0.93 (hfPWV), 0.73 (cfPWV), 0.87 (faPWV), 0.91 (baPWV) were achieved between technicians [23].

We standardized the PWV measurement across the 3 centers. Before the study started, staff at the Ultrasound Research Laboratory, University of Pittsburgh visited Honolulu to train the sonographers in Honolulu and from South Korea in PWV measurements.

All participants underwent a physical examination, and completed a lifestyle questionnaire (e.g., current smokers, alcohol drinking [two times or greater per week], and use of anti-hypertensives, diabetic medications, or lipid-lowering medications) and a laboratory assessment as described previously [20]. Venipuncture was performed early in the clinic visit after a 12-h fast. Samples were storied at -80°C and shipped on dry ice to the University of Pittsburgh [20]. Biochemical measurements of the blood samples were centralized at the Heinz Laboratory of the University of Pittsburgh. Serum lipids were determined with the standardized methods according to the Centers for Disease Control and Prevention, including total cholesterol, low-density-lipoprotein cholesterol (LDL-C), high-density-lipoprotein cholesterol (HDL-C), and triglycerides [24]. Serum fasting glucose was determined by a hexokinase -glucose-6-phosphate-dehydrogenase-enzymatic assay; serum fasting insulin by a radioimmunoassay (Linco Research Inc., Saint Charles, MO). Blood pressure (BP) (mmHg) and heart rate (beats per minute) were measured at right brachial artery in the sitting position after a 5-minute rest with the bladder emptied, by using an automated sphygmomanometer (BP-8800, Colin Medical Technology, Komaki, Japan) at the clinic visit. An average of two measurements was used. All the data collection was standardized across all 3 centers.

We analyzed the pooled data of White Americans (n = 232), African Americans (n = 68), Japanese-Americans (n = 202), and Koreans (n = 282). Sociodemographic and clinical characteristics and regional PWVs were described by using means (standard deviations), medians (interquartile ranges), frequencies, and percentages as appropriately. To obtain correlation coefficients, we performed the partial correlation analysis with adjustment for age and race. To examine race-adjusted associations between PWVs and CV risk factors, we performed multiple regression analysis by placing each CV risk factor as a predictor variable in the model after adjusting for race (placed in dummy variables). Next, to identify CV risk factors associated with each PWV, we performed stepwise regression analysis with the backward selection that specified the significance level (P value > .1) with adjustment for race as a lockterm. The CV risk factors included as predictor variables into the models were age, systolic BP, heart rate, body mass index (BMI), current smoking status, LDL-C, the ratio of total cholesterol vs. HDL-C, triglycerides (log-transformed), fasting glucose, and fasting insulin (log-transformed). Statistical significance was considered to be P < .05. All statistical analyses were performed with STATA 10.0 for Windows (StataCorp LP, College Station, TX).

All the regional PWVs correlated significantly with each other. The correlations of cfPWV and baPWV with either hfPWV or faPWV were not independent (Table 2). cfPWV correlated strongly with hfPWV (r = .81, P < .001), but weakly with faPWV (r = .12, P = .001). baPWV correlated moderately with both hfPWV (r = .47, P < .001) and faPWV (r = .62, P < .001). According to the data from this correlation matrix, cfPWV mostly reflected properties of central arterial stiffness, while baPWV reflected mixed properties of both central and peripheral arterial stiffness. baPWV moderately correlated with cfPWV (r = .42, P < .001). Meanwhile, the correlation between hfPWV and faPWV was also significant, but minimal (r = .08, P = .030).

Age was significantly and positively associated with hfPWV and baPWV (Table 3). Systolic BP and heart rate were significantly and positively associated with all the regional PWVs. Similarly, triglycerides and fasting insulin were significantly and positively associated with all the regional PWVs; however, LDL-C was not significantly associated with any regional PWV.

BMI was significantly and positively associated with hfPWV, cfPWV, and baPWV, but not with faPWV. Current smoking was significantly and positively associated with faPWV and baPWV.

After the stepwise regression analyses with adjustments for race, the CV risk factors of age, systolic BP, BMI, heart rate, triglycerides, and current smoking were included in the regression models as significant correlates for at least one regional PWV (Table 4). Significant correlates for hfPWV were age, systolic BP, and BMI. Significant correlates for cfPWV were systolic BP and BMI. Significant correlates for faPWV were systolic BP, BMI, triglycerides, and current smoking. Significant correlates for baPWV were age, systolic BP, heart rate, triglycerides, and current smoking. As a result, all the regional PWVs had a common significant correlate, i.e., systolic BP. In particular, cfPWV shared common correlates with both hfPWV and faPWV (i.e., systolic BP and BMI). However, BMI was significantly and positively associated with hfPWV and cfPWV, and negatively associated with faPWV. Meanwhile, baPWV shared common correlates with hfPWV (i.e., age and systolic BP) and also with faPWV (i.e., systolic BP, triglycerides, and current smoking).

Finally, based on R² values in the multivariable adjusted models, the CV risk factors explained 21%, 19%, 21%, and 34% of the total variances in hfPWV, cfPWV, faPWV, and baPWV, respectively (Table 4).