Distribution, determinants and normal reference values of aortic arch width: Thoracic aortic geometry in the Framingham Heart Study

Abstract

Study objective

Aortic arch geometry changes with age, including an increase in aortic arch width (AAW). High AAW is a predictor of incident adverse cardiovascular disease (CVD) events, but its distribution and determinants are unknown. We hypothesized that traditional CVD risk factors, in addition to age, are associated with increased AAW in community-dwelling adults.

Study design

Framingham Offspring and Third Generation cohort participants (N = 3026, 52 % men) underwent thoracic multidetector computed tomography (MDCT). A referent group (733M, 738W) free of clinical CVD, hypertension, dyslipidemia, smoking, and diabetes was used to generate sex and 10-year age-group specific upper 90th percentile (P90) cut-points for AAW. AAW was measured as the distance between the cross-sectional centroids of the ascending and descending thoracic aorta. Multivariable logistic regression models were used to identify clinical correlates of high AAW (≥referent P90) in the overall study group.

Results

Among referent participants, AAW increased with greater age-group, p for trend <0.0001 in each sex. Overall and within each age group, AAW was greater in men than women, p < 0.0001 all comparisons. Across all participants, high AAW was associated with greater age (odds ratio, OR = 1.34/10 years; 95 % confidence interval 1.20–1.50), body surface area (OR = 1.97/SD; 1.62–2.40), diastolic blood pressure (OR = 1.59/10 mm Hg; 1.40–1.81), pack-years smoked (OR = 1.07; 1.02–1.13), and prevalent CVD (OR = 1.64; 1.08–2.49).

Conclusion

AAW increases with greater age, body size, diastolic blood pressure and burden of smoking. High AAW (≥referent P90) is also associated with prevalent (clinically apparent) CVD. AAW is often seen on and easily measured from tomographic thoracic images and has prognostic value.

1. Introduction

The thoracic aorta increases in diameter, elongates, and stiffens with advancing age [1], [2], [3]. Age-associated changes in aortic geometry include dilation of the ascending and descending limbs of the thoracic aorta, increase in length of the aortic arch, and an increase in the width and the radius of curvature of the aortic arch [4], [5], [6]. Adverse remodeling of the thoracic aorta may be a precursor of or marker for hypertension and left ventricular concentric remodeling and may be associated with long-term adverse cardiovascular disease (CVD) events [7]. Aortic arch width (AAW), which is one aspect of aortic geometry determined with a single measurement from an axially-oriented image plane at the level of the main pulmonary artery bifurcation, has been shown to be an important prognostic marker. Chuang et al. demonstrated that AAW is a predictor of incident adverse CVD events beyond traditional Framingham risk factors and coronary artery calcium (CAC) in the Framingham Offspring and Third Generation cohorts [8]. However, a comprehensive analysis of the distribution and determinants of AAW in a population of community dwelling adults remains uncharacterized.

Over 3000 Framingham Heart Study (FHS) participants underwent thoracic CT to assess CAC, and these scans were suitable for measurement of AAW as well. Because AAW provides valuable and essentially “free,” prognostic data because it is a single measurement that requires trivial image processing, we sought to characterize the distribution of AAW by age and sex, and to identify clinical correlates of high AAW. We performed parallel analyses for the cross-sectional areas of the ascending (AA) and descending thoracic (DTA) aorta measured from the same imaging plane as for AAW.

2. Materials and methods

2.1. Study population

Framingham Offspring [9] and Third Generation [10] cohorts who underwent cardiac multidetector computed tomography (MDCT) scanning (2002–2005) and had complete risk profiles were included. Eligibility criteria for the MDCT sub-study comprised age ≥35 years (men) or ≥40 years (women). Eligible women were non-pregnant (verified by questionnaire and urine pregnancy test ≤24 h prior to MDCT study). All participants weighed <160 kg. All participants provided written informed consent. The study was approved by the institutional review boards of the Boston University Medical Center and The Massachusetts General Hospital.

Offspring and Third Generation cohorts have been previously described [8], [9]. Briefly, FHS participants undergo periodic “cycle” examinations every several years. Clinical covariates were collected at the Offspring cycle 7 (1998–2001) or Third Generation cycle 1 (2002–2005) examinations. Height and weight were measured with participants in light clothing. Body mass index (BMI) was calculated as weight divided by the square of height (kg/m²). Body surface area (BSA) was calculated using the Mosteller formula [11]. A morning venous blood draw was obtained after 12-hour overnight fast. Interim history, including medications taken and smoking status, was obtained via interview and questionnaires. A physician-performed examination, including duplicate brachial blood pressure measurements using a mercury sphygmomanometer; the average of the two measures was used.

Current cigarette smoking was defined by smoking ≥1 cigarette daily over the past year. Hypertension was defined as a systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or use of antihypertensive medication. Diabetes was defined as a fasting plasma glucose ≥126 mg/dL or treatment with insulin or anti-hyperglycemic medication. Hyperlipidemia was defined as a serum cholesterol ≥240 mg/dL or use of pharmacologic treatment. Cardiovascular disease (CVD) events were adjudicated using standardized criteria [12] by a three-physician panel blinded to risk factor or CT results, after review of all available records.

2.2. Image acquisition and analysis

MDCT scanning was performed on an 8-slice multidetector system (Lightspeed Ultra, General Electric, Milwaukee, WI) during a single, mid-inspiratory breath hold. Image acquisition was prospectively initiated at 50 % of the RR cycle using electrocardiographic triggering. Imaging parameters included 120 kVp with 320-mA tube current (400 mA if body weight was ≥100 kg), 500-ms gantry rotation time and 3:1 table feed. Estimated radiation exposure was 1 mSv (1.25 mSv for 400 mA tube current). Each participant was scanned twice consecutively per the CAC protocol; the first set of scans was used for AAW measurements.

Images were analyzed off-line on a dedicated workstation (Aquarius 3D, TeraRecon Inc., San Mateo, California). A single, trained observer unaware of participant characteristics (R.M.M.) identified the level of the main pulmonary artery bifurcation and measured AAW as the distance between the centroids of the region of interest for the AA and DTA at that level. Using the same software, the cross-sectional areas of the AA and DTA were measured using the elliptical region-of-interest tool. Examples of these measurements are shown in Fig. 1. Observer reproducibility of AAW was assessed on a subset of 100 participants distributed across the Offspring and Third Generation cohorts. Two trained observers unaware of each other's results performed independent measurements of the 100-participant subset in random order at a time-point separate from primary analyses. Intra- and inter-observer intra-class correlation coefficients were 0.988 and 0.985, respectively.

Fig. 1.

Aortic arch width (AAW) and cross-sectional areas of the ascending (AA) and descending thoracic (DTA) aorta at the level of the main pulmonary artery bifurcation.

2.3. Statistical analysis

All analyses were conducted using SAS v9.4 (SAS Institute, Cary, NC). “Normal values” were based on a healthy referent group free of prevalent CVD, hypertension, dyslipidemia, smoking, and diabetes. We compared referent AAW and aortic cross-sectional areas between men and women using 2-sample t-tests. We stratified within each sex by 10-year age-groups, as we hypothesized that AAW and aortic areas would vary with age. For each sex-and-age group, we determined mean values and used linear regression, with age-group as an ordinal variable, to test for within-sex trends across age-groups. We also calculated upper 90th percentile (P90) cut-points for AAW by sex and age group for the referent participants. In secondary analyses, referent participants were stratified by BSA, in addition to sex and age group, as AAW may vary with body size. Similar analyses were performed for the aortic cross-sectional areas (AA, DTA). Using sex-and-age group specific P90 cut-points, we determined prevalence of high (≥P90) AAW and aortic areas in the overall study sample; sex differences in prevalence of enlargement were assessed using the chi square test for homogeneity. We also tested for within-sex trends across age-groups. Sex-specific, age-adjusted Pearson's correlation coefficients were determined to assess linear associations between body size and AAW and aortic cross-sectional areas.

Multivariable logistic regression was used to identify determinants of high AAW in the entire sample. Non-normal variables were log-transformed. Each candidate covariate was entered into individual age-and-sex adjusted logistic regression models. When comparing surviving (p ≤ 0.05) variables likely to be co-linear, e.g. systolic and diastolic blood pressure, the variable with greater odds ratio per comparable increment was selected for entry into the multivariable model. The final model included age, sex, standardized BSA, BMI, diastolic blood pressure, high-density lipoprotein (HDL) cholesterol, log triglycerides, log pack-years smoked, diabetes, treatment for hypertension, and prevalent CVD. Multivariable-adjusted linear regression, using covariates identified from logistic analyses, was performed subsequently. Parallel analyses were conducted for aortic cross-sectional areas.

3. Results

Baseline characteristics of the 3026 participants are shown in Table 1. The overall study sample included 1560 men (51.6 %) with a mean age of 50 ± 11 years and 1466 women aged 52 ± 10 years. The referent cohort comprised 1471 adults (50.2 % men). The mean AAW across all referent men was 74.2 ± 9.2 mm, which was greater than the mean AAW for referent women, 66.5 ± 8.3 mm (sex-difference p < 0.0001). As shown in Table 2, mean AAW increased with greater age-group in each sex (p for age trend <0.0001 for each sex). Similarly, among all study participants, mean AAW was greater in men (79.2 ± 11.8 mm) than women (70.4 ± 10.1 mm) (p < 0.0001), and AAW increased with greater age-group (p < 0.0001) in each sex (Table 3). Overall, the prevalence of high AAW (Table 3) did not differ between men (14.9 %) and women (15.1 %), p = 0.83. However, in the oldest age-group, men (22.8 %) had greater prevalence of high AAW than their female counterparts (16.3 %), p = 0.048. In both sexes, there was an increase in prevalence of high AAW with greater age-group; the trend was significant among men (p = 0.0006) but not women (p = 0.17).

Table 1.

Clinical characteristics of the study population.

	Men	Women
N	1560	1466
Offspring	598 (38.3 %)	682 (46.5 %)
Third generation	962 (61.7 %)	784 (53.5 %)
Age, years	49.9 ± 10.7	52.2 ± 9.9
Height, m	1.77 ± 0.07	1.63 ± 0.06
Weight, kg	89.0 ± 15.2	71.8 ± 15.2
Body surface area, m2	2.08 ± 0.19	1.79 ± 0.21
Body mass index, kg/m2	28.4 ± 4.6	27.1 ± 5.9
Systolic blood pressure, mm Hg	124 ± 15	120 ± 18
Diastolic blood pressure, mm Hg	78 ± 9	74 ± 9
Total cholesterol, mg/dL	194 ± 34	198 ± 36
HDL-C, mg/dL	46 ± 12	61 ± 17
Triglycerides, mg/dL	115 [76, 172]	93 [66, 137]
Pack years smoked	0 [0, 11.3]	0 [0, 8.0]
Hypertension treatment	317 (20 %)	307 (21 %)
Adult onset diabetes	103 (7 %)	70 (5 %)
Prevalent cardiovascular disease	124 (8 %)	70 (5 %)

Data are expressed as mean ± SD, median [quartiles], or as number (percentage).

Table 2.

Mean and standard deviation of AAW and thoracic aortic areas for the healthy referent group by sex and age with corresponding upper 90th percentile (P90) cut-points.

	Men					Women
Age	<45	45–54	55–64	≥65	p for trend	<45	45–54	55–64	65	p for trend
N	386	229	79	44		278	273	114	68
AAWa, mm	70.1 ± 7.0	76.2 ± 8.0	80.8 ± 8.4	88.7 ± 8.6	<0.0001	62.2 ± 5.7	66.1 ± 7.0	71.7 ± 7.7	76.9 ± 9.1	<0.0001
AAb area, mm2	803 ± 153	912 ± 169	947 ± 175	1035 ± 185	<0.0001	712 ± 133	761 ± 160	838 ± 164	862 ± 170	<0.0001
DTAc area, mm2	462 ± 75	509 ± 79	565 ± 95	628 ± 67	<0.0001	379 ± 57	409 ± 67	449 ± 70	475 ± 79	<0.0001
Upper 90th percentile cut-points
AAW, mm	79.2	87.5	94.1	100.6		69.3	75.1	80.5	90.3
AA area, mm2	1004	1132	1206	1302		900	984	1063	1083
DTA area, mm2	560	618	671	704		451	497	531	585

^aAortic arch width.

^bAscending aorta.

^cDescending aorta.

Table 3.

Mean values and prevalence of high AAW (≥healthy referent P90) and aortic cross-sectional areas in the overall study group.

	Men					Women
Age group	<45	45–54	55–64	≥65	p for trend	<45	45–54	55–64	≥65	p for trend
N	555	442	283	280		364	459	343	300
AAWa, mm	71.0 ± 7.1	77.6 ± 8.3	83.5 ± 8.8	93.3 ± 11.3	<0.0001	62.7 ± 5.9	67.5 ± 7.4	73.4 ± 8.0	80.6 ± 10.2	<0.0001
AAb area, mm2	819 ± 162	931 ± 180	1004 ± 195	1088 ± 220	<0.0001	724 ± 138	785 ± 161	859 ± 163	904 ± 173	<0.0001
DTAc area, mm2	468 ± 79	525 ± 86	584 ± 99	683 ± 154	<0.0001	382 ± 57	419 ± 70	468 ± 79	532 ± 133	<0.0001
High AAW, N	71 (13 %)	58 (13 %)	39 (14 %)	64 (23 %)⁎	0.0006	46 (13 %)	71 (15 %)	56 (16 %)	49 (16 %)	0.17
High AA, N	69 (12 %)	60 (14 %)	42 (15 %)	43 (15 %)	0.19	45 (12 %)	56 (12 %)	43 (13 %)	45 (15 %)	0.33
High DTA, N	68 (12 %)	57 (13 %)	46 (16 %)	98 (35 %)⁎	<0.0001	43 (12 %)	62 (14 %)	67 (20 %)	77 (26 %)	<0.0001

^aAortic arch width.

^bAscending aorta.

^cDescending aorta.

^⁎p < 0.05 for men vs. women in corresponding age groups.

Among referent-group, age-adjusted correlations of AAW with measures of body size were significant in both sexes, with p < 0.0001 for each correlation: height (men: r = 0.21, women: r = 0.19), weight (men: r = 0.48, women: r = 0.55), BSA (men: r = 0.48, women: r = 0.46). Across all participants, AAW again was correlated with height (men: r = 0.15, women r = 0.15), weight (men: r = 0.42, women: r = 0.43) and BSA (men: r = 0.44, women: r = 0.44), with p < 0.0001 for each correlation. Supplemental Table 1 shows mean and P90 AAW by sex, age group, and BSA. These are secondary analyses due to the low numbers of participants in each category in the older age groups, but greater BSA appears to be associated with greater AAW.

In multivariable-adjusted logistic regression, greater age, BSA, diastolic blood pressure, pack-years smoked, and prevalent CVD were associated with high AAW (Table 4). The same covariates were entered into a multivariable linear regression model (Table 5). Each additional 10 years of age was associated with 6.3-mm increase in AAW. Other factors in the multivariable linear regression model associated with increasing AAW were BSA, treatment for hypertension, diastolic blood pressure and burden of smoking. Female sex was associated with 9.6-mm lower AAW. Adult-onset diabetes was associated with slightly (1.3 mm) but significantly lower average AAW.

Table 4.

Multivariable-adjusted logistic regression model for high AAW (≥healthy referent P90) with traditional CVD risk factors across all study participants (c = 0.763).

	Odds ratio	95 % CI	p-Value
Age, /10 years	1.34	1.20–1.50	<0.0001
Female sex	1.28	0.99–1.66	0.06
Body surface area, per SD	1.97	1.62–2.40	<0.0001
Body mass index, kg/m2	1.01	0.98–1.05	0.51
Diastolic blood pressure, /10 mm Hg	1.59	1.40–1.81	<0.0001
Hypertension treatment	1.20	0.92–1.57	0.18
HDL-C, /10 mg/dL	0.93	0.85–1.02	0.14
log(Triglycerides)	0.87	0.68–1.11	0.27
Diabetes	0.70	0.45–1.09	0.12
log(Pack years smoked)	1.07	1.02–1.13	0.01
Prevalent CVD	1.64	1.08–2.49	0.02

Table 5.

Multivariable-adjusted linear regression model for AAW across all study participants.

	β	S.E.	p-Value
Age, /10 years	6.28	0.13	<0.0001
Female sex	−9.58	0.31	<0.0001
Body surface area, per SD	2.84	0.23	<0.0001
Body mass index, kg/m2	0.06	0.05	0.20
Diastolic blood pressure, /10 mm Hg	1.66	0.15	<0.0001
Hypertension treatment	1.87	0.35	<0.0001
HDL-C, /10 mg/dL	−0.09	0.10	0.39
log(Triglycerides)	−0.46	0.28	0.10
Adult onset diabetes	−1.31	0.58	0.02
log(Pack years smoked)	0.22	0.06	0.0007
Prevalent CVD	0.97	0.58	0.09

Age-adjusted correlations for AA and DTA cross-sectional areas with AAW were significant for AA area (men: r = 0.55, women: r = 0.52, p < 0.0001 for each sex) as well as DTA area (men: r = 0.44, women: r = 0.45, p < 0.0001 for each sex). Age and sex-adjusted linear regression models showed AAW increased with greater AA area (3.7 mm per standard-deviation increment) as well as DTA area (2.5 mm/SD). In the same model, each decade of age was associated with an additional 3.8 mm of AAW; while female sex was associated with 9.8 mm lower AAW.

Among referent-group participants, AA and DTA areas were significantly correlated with height (r = 0.14 to 0.18) in both sexes, but more strongly correlated with weight and BSA (r = 0.28 to 0.37), p < 0.001 for all. Across the entire study sample, AA and DTA areas correlated with weight and BSA (r = 0.25 to 0.30, p < 0.0001 all), but not with height (p > 0.19). Among referent participants, mean AA and DTA areas were greater in men (866 ± 17 and 498 ± 91 mm² respectively) than in women (764 ± 161 and 410 ± 72 mm²), p < 0.0001 for both comparisons. AA and DTA areas increased monotonically with greater age-group in both sexes (all p < 0.0001) as shown in Table 2.

Across all study participants the age-pooled prevalence of high AA area did not differ between men (13.7 %) and women (12.9 %), p = 0.50, nor did prevalence of high DTA area differ (men: 17.2 %, women: 17.0 %), p = 0.85. The prevalence of high AA area increased minimally with greater age-group in each sex, but trends did not reach significance (Table 3). There was a significant increase in high DTA with increasing age-group in each sex (p < 0.0001 both), but prevalence within each age-group did not differ between sexes, except for the ≥65 years age-group, p < 0.014 (Table 3).

In multivariable-adjusted logistic regression models (Supplemental Table 2), greater age, BSA and blood pressure were associated with high AA and DTA areas, but smoking was associated with high DTA area only. Interestingly, in age and sex-adjusted models, the odds ratio for high DTA area associated with systolic blood pressure was greater than the odds ratio associated with diastolic blood pressure; the opposite was true for high AA area. Accordingly, systolic blood pressure was entered into the multivariable model for high DTA area, and diastolic blood pressure into the high AA area model. Linear regression results (Supplemental Table 3) were consistent with logistic findings, and further indicated that female sex was associated with smaller aortic cross-sectional area.

4. Discussion

In this study of over 3000 community-dwelling adults, we found AAW is greater in men than women and increases with advancing age in both sexes. We determined sex- and age-specific mean and upper 90th percentile limits for AAW in a referent group drawn from the FHS population. The prevalence of high AAW was similar between men and women, and was associated with greater age, body size, diastolic blood pressure, burden of smoking, and prevalent CVD. Cross-sectional areas of the AA and DTA also increased with greater age and were greater in men than women.

4.1. In the context of the current literature

Prior studies have shown there may be an association between changes in thoracic aortic geometry and adverse CVD events. The Rotterdam study demonstrated that thoracic aortic dilation in both men and women was associated with adverse events; however, there was a differential impact of sex on this association. Rueda-Ochoa et al. found that among 2178 participants aged 69 years, women had a 1.37-fold increased risk of stroke per 1 SD increment in the AA diameter. A similar relationship was not seen with men. Increasing AA and DTA diameters were associated all-cause mortality among both men and women. Women were also at increased risk of CVD mortality [7]. In a younger population of Offspring and Third Generation Cohorts from the FHS (N = 3318) who were aged 48.9 ± 10.3 years, we found there was no statistically significant association of high (based on 90th percentile cutpoints constructed for a healthy referent group within the FHS study population) AA and DTA with adverse CVD events above traditional CVD risk factors [13]. These variations in findings may be related to differences in underlying risk factors such as population age. However, aortic diameters alone may not offer a thorough assessment of the geometric changes that occur alongside aortic dilation such as an increase in the length, width, and radius of curvature of the aortic arch. AAW may serve as a more accurate summary measure of these changes in thoracic aortic morphology.

Chuang et al. have specifically evaluated the prognostic value of this summary measure in 2880 FHS participants drawn from the population described in the present study [8]. Over 8.9 years, high AAW augmented prediction of CVD events above Framingham risk factors and CAC in multivariable- adjusted Cox models (HR = 1.55; p = 0.032). AAW improved prediction of CVD risk with a net reclassification index (NRI = 0.31, 95 % CI = 0.15–0.48) indicating appropriate movement of individual risk in the AAW-augmented model. These findings may have significant clinical implications for the general population as thoracic CT scanning is commonly obtained for a variety of indications, and AAW can be easily ascertained from axial CT images. AAW may essentially provide “free” prognostic information related to future CVD risk. Therefore, it is imperative that the distribution and determinants of AAW are characterized in a community dwelling cohort.

However, there have been relatively few reports regarding the distribution and determinants of AAW and aortic geometry in general. Thijssen et al. have shown that among 943 participants from the Rotterdam Study who were aged 62 years, characteristics such as age, male sex, BMI, and diastolic blood pressure were determinants of increasing AA and DTA [14]. In our assessment of AA and DTA areas, we found similarly that diastolic blood pressure, age, and sex, were determinants of increasing areas in addition to risk factors such as smoking, prevalent CVD, and systolic blood pressure. Sugawara et al. also studied 256 healthy adults (51 % men), ranging in age from 19 to 79 years, using magnetic resonance imaging and found an increase in overall length of the thoracic aorta with advancing age [4]. Aortic elongation was principally due to increase in the length of the ascending aorta; lengths of the descending thoracic and abdominal aorta were not associated with age. The authors postulated that age-associated increase in ascending aorta length was due to fatigue, fracture and disarray of elastin fibers in the proximal aorta, leading to remodeling [4]. Sugawara et al. did not address aortic diameter or AAW in their study.

Using custom analysis software, Craiem et al. analyzed the impact of hypertension on thoracic aortic size and geometry among 200 normotensive and 200 hypertensive asymptomatic men with at least one coronary heart disease risk factor who underwent non-contrast CT for risk stratification [5]. Both age and hypertension were associated with increased AAW and radius-of-curvature of the arch. In multivariable-adjusted linear regression models, hypertension was equivalent to 2–7 years of additional aging. Increases in segmental aortic volume were more strongly associated with diastolic, rather than systolic blood pressure, but in contrast to our findings, Craiem et al. found that increasing AAW was more strongly associated with systolic (12 % per 10 mm Hg) rather than diastolic (3 % per 10 mm Hg) blood pressure. Overall, the results of Craiem et al. are largely concordant with our findings, but direct comparison is limited by different study populations. The former study was limited to men referred for coronary heart disease risk assessment, while our study encompassed men and women from a community-dwelling cohort with markedly lower prevalence of hypertension reflective of the general population.

Redheuil et al. used magnetic resonance imaging to determine aortic size and geometry in 100 adults (55 women) free of overt CVD [6]. As with the study of Sugawara et al., the ascending aorta increased in length with advancing age, whereas the descending aorta did not. AAW was greater in men than women and increased significantly with advancing age. Comparing study subjects <30 and >70 years of age, the authors observed a 34 % relative increase (+0.40 mm/year) in AAW. The increase in AAW was principally due to changes in the ascending portion of the aortic arch. Redheuil et al. found no association between AAW, or other markers of arch geometry, and body height, but body weight was correlated with AAW (r = 0.34, p = 0.001) and ascending aortic diameter (r = 0.24, p = 0.01). AAW was correlated with multiple measures of blood pressure, including central and brachial systolic, diastolic, mean and pulse pressures. The present study is largely concordant with the findings of Redheuil et al. We were unable to measure arch length; to minimize radiation exposure, our imaging protocol did not cover the aortic arch since the main purpose of imaging was assessment of CAC. We also found that greater age and hypertension are associated with increasing AAW. We further identified burden of smoking and prevalent CVD as contributors to increased AAW, whereas diabetes was associated with a minimal decrease in AAW. Finally, our study suggests that AA area is more strongly associated with AAW than DTA area in both sexes, like Redheuil et al., who measured AA diameter instead of area. Our findings extend those of Redheuil et al. to a markedly larger, community-dwelling population of free-living adults.

Recently, Lee et al. defined “aortic unfolding” as the “longest distance between the ascending and descending aorta” on an axial slice at the same level as used in the present study [15]. This “unfolding” measure is approximately equivalent to our AAW measure plus the radii of the ascending and descending aortic cross-sections. In 219 adults, self-referred for CAC screening, the unfolding measure was greater in men and was associated with age, body surface area, and hypertension. Unfolding was also associated with CAC score on age-and-sex adjusted analyses. Our findings appear to be broadly concordant with those of Lee et al., but cannot be directly compared since AAW and “unfolding” differ in their definition.

4.2. Potential mechanisms

AAW increases with greater age, greater body surface area, diastolic blood pressure, treatment for hypertension, and burden of smoking, while diabetes was associated with a minimal but statistically significant decrease in quantitative AAW. As noted by Redhuil et al., the mechanisms that underlie age-related changes in aortic geometry, namely transverse dilation, have been well-described and consist of histologic alterations from constant pulsatile stress over time leading to aortic stiffening. However, their findings uniquely showed that a greater degree of age-related aortic arch elongation occurred, which coupled with transverse dilation may augment the proximal aortic volume thus increasing systolic blood volume storage capacity to preserve proximal aortic elasticity [6]. Furthermore, the association of blood pressure with high AAW supports Sugawara et al.'s hypothesis that proximal aortic remodeling is in part due to age and wear-and-tear associated breakdown of elastin in the portion of the aorta that serves as a windkessel, or distensible pressure reservoir [4]. Weight and BSA were more strongly associated with AAW than height. Weight and BSA may better represent circulating blood volume and overall metabolic demand than height [16]. Greater blood volume could account for excess baseline and pulsatile load on thoracic aortic walls, thereby accelerating aortic remodeling. Smoking is associated with acute [17] and long-term deterioration of aortic elasticity [18], so the association of burden of smoking with high AAW is consistent with overall smoking-related aortic remodeling. Smoking has been associated with increased diameter of the descending thoracic and abdominal [19], [20], [21], but not ascending [22], [23] aorta; which is concordant with the findings of the present study. Finally, the finding that diabetes was associated with lower AAW is consistent with prior reports that diabetes is inversely associated with abdominal aortic aneurysm [19] and may be due in part to chronic hyperglycemia associated structural alterations in the aortic wall [24], [25]. However, given the borderline (p = 0.02) association of diabetes with AAW in the linear regression model, and that diabetes was not significant (p = 0.12) with high AAW in the multivariable logistic regression model, it is also possible that our results regarding diabetes and AAW are due to the play of chance.

4.3. Clinical implications and future directions

AAW increases with multiple cardiovascular disease risk factors and with overt CVD itself. AAW may be a simple summary measure of aortic arch geometry. Since the ascending and descending limbs of the thoracic aorta are often seen incidentally on CT scans of the chest, AAW might be an essentially “free” marker of CVD risk with no additional imaging and only minimal analysis (a single linear measurement) overhead. In our study of AAW, we have previously found it has independent prognostic value beyond traditional CVD risk factors and CAC. Therefore, it may offer additional risk stratification, particularly for individuals at intermediate 10-year risk of CVD events who may undergo MDCT scanning for CAC screening. While we have identified cross-sectional correlates of high AAW in the present study, the risk factors associated with progression of AAW remain to be determined.

4.4. Strengths and limitations

The FHS comprises community-dwelling adults who have been followed with meticulous collection of clinical characteristics. AAW from non-contrast CT images is highly reproducible in terms of landmark (pulmonary artery bifurcation) identification for correct slice selection and in terms of the actual AAW length. However, our study is not without limitations. The relationships between AAW and traditional CVD risk factors identified herein are cross-sectional; causality cannot be inferred. The Framingham Offspring and Third Generation cohorts are overwhelmingly of European ethnicity; generalization to other ethnicities may be limited.

5. Conclusion

Aortic arch width, which is a known prognostic marker of adverse CVD events, may be a simple summary measure of aortic-arch geometry, which is known to vary with age and burden of CVD risk factors. We found that AAW is greater in men than women; in both sexes AAW increases with greater age, body size, blood pressure, burden of smoking, and with prevalent CVD. We also presented sex and age-stratified upper limits for AAW based on a healthy referent cohort.

The following are the supplementary data related to this article.

Supplemental Table 1

Mean AAW and 90th percentile cutpoints (P90) by age, sex and BSA among all participants.

Supplemental Table 2

Multivariable-adjusted logistic regression models for AA and DTA cross-sectional areas across all study participants.

Supplemental Table 3

Multivariable-adjusted linear regression models for AA and DTA cross-sectional areas across all study participants.

Funding

Supported by the National Heart, Lung, and Blood Institute's Framingham Heart Study (contract numbers N01-HC-25195 and N01-HC-38038). Dr. O'Donnell, Dr. Fox, and Dr. Gona were supported by the NHLBI Division of Intramural Research.

CRediT authorship contribution statement

Saadia Qazi: Investigation, conceptualization, methodology, writing original draft and review; Philimon N. Gona: Conceptualization, methodology, analysis, software, writing review Rebecca M. Musgrave: investigation, writing review; Caroline S. Fox: Writing review; Joseph M. Massaro: Conceptualization, methodology, analysis, software, writing review; Udo Hoffmann: Investigation, conceptualization, methodology, writing review; Michael L. Chuang: Investigation, conceptualization, methodology, writing review; Christopher J. Odonnell: Investigation, conceptualization, methodology, writing review, supervision.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.