Between 1997 and 2003, previously unscreened 3,170 men aged 40-79 were recruited for a population-based prostate cancer screening study on the Caribbean island of Tobago, Trinidad and Tobago [15]. To be eligible, men had to be ambulatory, noninstitutionalized, and not terminally ill. Approximately 60% of all age-eligible men on the island participated (n=3375), and participation was similar across the island parishes. Between 2014 and 2016, 856 men were recruited for a follow-up clinical examination that included measures of subclinical CVD and chest and abdominal CT in a subset of individuals. The current analysis included 408 men from the follow-up visit who had aortic area measured, as well as data for all other subclinical CVD measures. The Institutional Review Boards of the University of Pittsburgh and the Tobago Ministry of Health and Social Services approved this study. All participants provided a written informed consent.

Thoracic and abdominal aortic areas were measured from the chest and abdominal CT scans collected using dual slice, high-speed NX/I computed tomography (GE Medical Systems, Waukesha, WI, USA). Digital imaging and communications in medicine (DICOM) images were securely sent electronically from Tobago to the Vanderbilt University Medical Center for archiving and analysis. Experienced analysts (JGT and SN) trained by a board-certified radiologist (JJC) identified and measured aortic area at all three locations using an Apple workstation running customized OsiriX MD software (https://www.osirix-viewer.com). The ascending (ASC) and descending (DSC) thoracic aortic areas were measured on the same transverse plane at the location of the pulmonary artery. While the abdominal (ABD) aortic area was measured above the bifurcation of the abdominal aorta proximal to where it splits into the iliac arteries. In each case, the vessel was visualized in OsiriX MD 3D multiplanar reconstruction mode to determine the centerline of the artery. The outer wall of the aorta was circumscribed and measured orthogonally to the vessel centerline. The mean of three contiguous 3 mm thick CT slices was used for each aortic area measure. These measures had an intra-class correlation coefficient of 0.92 in a 5% re-read of the study participants.

The use of non-contrast CT for measurement of aortic dimensions was validated in cynomolgus monkeys by comparing CT-derived dimensions with those measured in situ in exsanguinated monkeys sacrificed as part of another experiment. Whole-body CT imaging was performed on anesthetized monkeys using a Toshiba 32 slice scanner (Tokyo, Japan) at 120 kVp and 300 mA and a helical pitch of 21, pitch factor=0.625 with 0.5 sec rotation speed. The thoracic aorta was contoured on contiguous 0.5 mm axial CT slices by an experienced analyst providing excellent agreement with in situ measures in the ascending (r=0.98) and good agreement in the descending aorta (r=0.83). All animal manipulations and procedures were performed according to the guidelines of state and federal laws, the US Department of Health and Human Services, and the Animal Care and Use Committee of Wake Forest University School of Medicine.

Arterial calcification was measured using the same chest and abdomen CT images used for aortic area measurement. Coronary artery calcification (CAC) was measured from chest CT from the carina through the entire inferior aspect (diaphragmatic) of the heart. Whereas abdominal aortic calcification (AAC) was measured from L3 to S1 from the abdominal CT images. These images were scored for calcification presence (yes/no), as well as for severity using the Agatston method [16] by the Carr laboratory at Vanderbilt University Medical Center. Common carotid artery images were obtained using B-mode ultrasonography from a portable ultrasound machine (Cypress, Siemens/Acuson, Munich, Germany). Two images, each of the left and right, near and far walls of the distal common carotid artery (1 cm proximal to the carotid bulb), were collected and sent to the University of Pittsburgh Ultrasound Research Laboratory (URL), where they were read using a semi-automated, edge-detection reading system (Dr. Gustavsson, Sweden). These images produced measures of carotid intima-media thickness (cIMT; mm), carotid inter-adventitial diameter (IAD; mm), and carotid plaque presence (yes/no). Using an automated waveform analyzer (Colin VP2000, Komaki, Japan), brachial-ankle pulse-wave velocity (PWV; cm/s) was measured by recording pulse volume waveforms via cuffs equipped with plethysmographic sensors placed over the brachial and tibial arterial sites. baPWV is the distance in centimeters between the brachial and ankle arterial recording sites divided by the time delay in seconds between the foot of the respective waveforms. The distance or path length for brachial/ankle arterial sites was calculated based on a height-based algorithm [17].

Blood pressure readings were taken three times using an automated blood pressure cuff (Omron Healthcare, Inc, Vernon Hills, IL, USA) while the participant was sitting and with 5 min of rest in between the measurements. The average of the 2^nd and 3^rd readings was used as the mean systolic and diastolic blood pressure (SBP and DBP, respectively). Hypertension was defined using the American College of Cardiology (ACC) 2017 guidelines [18] as SBP ≥130 mmHg, DBP ≥80 mmHg, or current use of antihypertension medication. Standing height was measured twice using a wall-mounted stadiometer and averaged to the nearest 0.1 cm. Weight was measured to the nearest 0.1 kg using a balance beam scale. Body mass index (BMI) was calculated as the ratio of weight to height squared (kg/m²). Diabetes was defined using fasting serum glucose levels (≥126 mg/dl) and/or diabetic medication use. Dyslipidemia was defined by statin or other lipid-lowering medication use. The history of cardiovascular disease was self-reported and included myocardial infarction, stroke, congestive heart failure, angina, rheumatic heart disease, or intermittent claudication.

Standardized interviewer-administered questionnaires were used to collect demographic, health history, family history, and lifestyle information. Smoking was defined as current, former, or never smokers. Former smokers were defined as men who had previously smoked >100 cigarettes, but were not currently smoking. Regular alcohol consumption was defined for this study as drinking ≥4 drinks/week over the previous 12 months because alcohol intake is generally very low in the THS. As activity was difficult to quantify in this cohort using standard questionnaires developed in the US, we have assessed a number of metrics and determined sedentary behavior to be the strongest correlate of cardiometabolic health [19]. Therefore, sedentary behavior was dichotomized at the median into greater vs. lesser using reported TV watched per week (median=14 h).

We examined the distribution of each aortic area and subclinical CVD variable for normality and outliers, and transformed as necessary. In order to account for the largest contributors to arterial size, all analyses were adjusted for age and height. Spearman correlation was used to determine the correlation between aortic area measures. We used age- and height-adjusted linear regression to determine the effect of a 1SD greater or presence of each CVD risk factor on each aortic area. Then, stepwise linear regression was used to determine the CVD risk factors (independent variables) with independent associations with each aortic area (dependent variables). Stepwise regressions forced the age and height to be included in each model, and required a p-value of 0.2 to enter into the model and a p-value of 0.05 to remain in the model. We used model R-square and assessed variation inflation factors to consider related CVD risk factors, including weight vs. BMI, and SBP, DBP and hypertension treatment vs. hypertension status.

To assess the predictive associations of aortic areas (independent variables) with subclinical CVD measures (dependent variables), we used multiple linear or logistic regression, as appropriate. These models were first adjusted for only the CVD risk factors determined to have an independent association with aortic areas in the stepwise modeling procedures (multivariable model), and then additionally adjusted for all measured CVD risk factors (full model). Sensitivity analyses for these results were also conducted in the subset of individuals who did not report a history of CVD (n=385).

Lastly, in order to determine the potential utility of aortic area measures taken only from chest CT, we identified individuals in the top quartile of each of the ascending and descending aortic areas and categorized them into 4 groups: 1) individuals not in the either top quartile (reference group for all analyses, n=262); 2) individuals in the top quartile of the ascending aortic area only (n=45); 3) individuals in the top quartile of the descending aortic area only (n=43); and 4) individuals in the top quartile of both the ascending and descending aortic areas (n=59). Then we used nominal logistic regression to determine the association of being in any of these large aorta groups with the odds of having each subclinical CVD adjusting for all measured CVD risk factors (full model). Having subclinical CVD was defined by either the presence of plaques (for CAC, AAC, and carotid plaque) or by being in the top quartile of continuous subclinical CVD measures (for CAC and AAC severity, cIMT and IAD, and PWV). All analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC, USA).

Men were aged 64 years on average, with a range of 50-89 years. They were slightly overweight, and 80% were hypertensive. The prevalence of other cardiometabolic diseases was 25% for diabetes and 15% for dyslipidemia, but only 6% of men reported having a history of any CVD event. Few men were current smokers (7%) or more than occasional alcohol drinkers (14%), though over half of men reported watching more than 14 h of TV per week. Atherosclerotic plaque was present in 60% (AAC), 30% (CAC) and 15% (carotid plaque) of men. Assuming an Agatston score of 300 as significant, arterial calcifications were generally not severe (median scores: AAC 106 and CAC 0). PWV was 1655 cm/s on average. Mean unadjusted aortic areas were 10.1 cm², 6.2 cm², and 3.3 cm², for the ASC, DSC, and ABD aortic locations, respectively. All aortic areas were significantly correlated with each other (all p<0.001), with Spearman correlation estimates adjusted for age and height being highest for DSC and ABD (0.61) and ASC and DSC (0.60), and lower for ASC and ABD (0.34; data not shown) (Table 1).

Greater age, taller height, greater weight or BMI, greater DBP, hypertension, and hypertension treatment were all associated with larger areas at each measured aortic location (all p<0.05). Additionally, greater SBP was individually associated with DSC (p-value <0.001). However, no other traditional CVD risk factors listed in Table 2 were associated with any aortic area at p<0.05 after adjustment for age and height. Independent correlates for ASC included age, height, BMI, SBP, DBP, and hypertension treatment. For DSC, independent correlates included age, height, BMI, and DBP. While only age, height, and BMI were independent correlates of ABD.

Fig. (1). Odds Ratios for Prevalent Subclinical CVD by Composite Assessment of Ascending and Descending Aortic Areas Men were categorized into four groups based on the chest CT aortic areas as being in the top quartile of the ASC only (n=45), the DSC only (n=43), both aortic areas (n=59), or neither (n=262; referent group). Odds ratio estimates are displayed (diamond marker) and 95% CI (error bars) is shown for subclinical CVD in each panel (a-h) by each large aortic area group compared to the group with neither aortic area in the top quartile. Referent group estimates (all OR = 1.0) are not shown, though the OR=1.0 threshold is highlighted via a dashed line.

After adjustment for independent correlates of aortic areas from Table 2, each of the larger aortic areas at all three locations was associated with greater carotid IAD (all p<0.001) and greater PWV (all p<0.02). When entering all three measures into a single model, no individual aortic area was an independent correlate of IAD; however, DSC was the only independent correlate of PWV (p=0.006; data not shown). In addition, greater ASC was associated with greater cIMT (p=0.011) and nominally associated with carotid plaque presence (p=0.06). Greater ABD was associated with both the presence and severity of AAC (both p<0.001). However, no aortic area was associated with CAC presence or severity. Results were largely similar after adjustment for all measured CVD risk factors (Table 3) and in the subset of men without a history of clinical CVD (n=385; Supplemental Table).

After categorizing men into large vs. not large chest aortic areas (ASC and DSC) based on being in the top quartile for either or both, we determined the association of being in these categories with subclinical CVD throughout the body adjusting for all measured CVD risk factors (Fig. 1). While all groups had greater odds of large carotid IAD, only men with large ASC had significantly greater odds of carotid plaque (OR: 2.78, 95% CI: 1.16-6.66), and only men with both large ASC and DSC had greater odds of a thick cIMT (OR: 2.82, 95% CI: 1.36-5.88). Odds of a high PWV were highest in men with only a large DSC (OR: 4.11, 95% CI: 1.58-10.70); men with both large ASC and DSC also had significantly greater odds of high PWV (OR: 3.54, 95% CI: 1.48-8.47). Interestingly, while no chest aortic area was individually associated with CAC or AAC, men with both large ASC and DSC had significantly higher odds of any AAC and of severe CAC and AAC (all p<0.04). Results were similar, and interpretations were unchanged after including adjustment for CAC in all models (data not shown), suggesting information gained from aortic size on chest CT is independent of coronary calcification.