Abdominal aortic calcification identified on lateral spine images from bone densitometers are a marker of generalized atherosclerosis in elderly women

Abstract

Objective

Dual-energy X-ray bone absorptiometry (DXA) is a low-cost, minimal-radiation technique used to improve fracture prediction. DXA machines can also capture single-energy lateral spine images and abdominal aortic calcification (AAC) is commonly seen on these images.

Approach and Results

We investigated whether DXA-derived measures of AAC were related to an established test of generalized atherosclerosis in 892 elderly Caucasian women aged over 70 years with images captured during bone density testing in 1998/1999 and B-mode carotid ultrasound in 2001. AAC scores were calculated using a validated 24 point scale into low (AAC24 score, 0 or 1), moderate (AAC24 scores 2-5) and severe AAC (AAC24 scores >5) seen in 45%, 36% and 19% respectively. AAC24 scores were correlated with mean and maximum common carotid artery intimal medial thickness (r=0.12, P<0.001 and r=0.14, P<0.001). Compared to individuals with low AAC, those with moderate or severe calcification were more likely to have atherosclerotic plaque (adjusted prevalence ratio (PR) 1.35, 95%CI; 1.14-1.61, P<0.001 and PR 1.94 95%CI; 1.65-2.32, P<0.001 respectively) and moderate carotid stenosis (adjusted PR 2.22, 95%CI; 1.39-3.54, P=0.001 and 4.82 95%CI; 3.09-7.050, P<0.001). The addition of AAC24 scores to traditional risk factors improved identification of women with carotid atherosclerosis as quantified by C-statistic (+0.075, P<0.001), net reclassification (0.249, P<0.001) and integrated discrimination (0.065, P<0.001).

Conclusions

Abdominal aortic calcification identified on images from a DXA machine were strongly related to carotid ultrasound measures of atherosclerosis. This low-cost, minimal radiation technique used widely for osteoporosis screening is a promising marker of generalized extracoronary atherosclerosis.

Introduction

The pathogenesis of arterial disease commences with the accumulation of lipid-laden cells within the sub-endothelial layers leading to occlusive intimal plaques in the medium to large sized arteries such as the coronary, carotid, vertebral, aortic and iliac arteries¹. These lesions accumulate cellular infiltrate within the intimal layer of the arterial walls ² and express osteochondrogenic transcription factors regulating the deposition of extracellular calcium ³. These lesions appear as raised white areas in the arterial wall ⁴ and are easily visualised using non-invasive imaging modalities. B-mode carotid ultrasound is a validated, non-invasive, sensitive and highly reproducible method of directly visualising and assessing subclinical carotid atherosclerosis not dependent on calcification^5-8. The technique involves measuring the carotid wall lumen-intima to the media-adventitia interface the carotid intimal medial thickness (CCA-IMT) and is considered an indication of the degree of generalized atherosclerosis^9-11. In addition to CCA-IMT the presence of any focal plaque throughout the carotid artery and the degree of arterial narrowing by plaque is measured.

Non-invasive imaging of vascular structure to detect subclinical atherosclerosis within the apparently healthy elderly population with the aim of identifying individuals that are the most and the least likely to benefit from simple and inexpensive lifestyle modifications is an active area of clinical research. A current example is the study of the clinical utility of CT measured coronary artery calcium to predict future clinical events¹². However concerns over the high cost and radiation exposure of CT have limited its use in low to intermediate risk individuals¹³. Low-cost, minimal radiation exposure techniques are therefore needed. One such technique is single-energy lateral spine images from a dual energy X-ray absorptiometry (DXA) machine which are used to diagnose osteoporosis and predict fracture risk. During a DXA examination lateral spine single-energy images can also be captured and are currently used to identify vertebral fractures. These images simultaneously capture the presence of abdominal aortic calcification (AAC) that can be assessed by an expert reader to produce an AAC24 score. AAC from these images has been shown to be associated with cardiovascular disease events in older individuals^14-18 and specific patient populations such as those with rheumatoid arthritis¹⁹. Previous studies using AAC assessed by high radiation exposure CT have found AAC is a marker of generalized coronary and extracoronary atherosclerosis burden^{20, 21}, however no studies have reported the relationship using images captured from a DXA machine.

DXA is recommended for osteoporosis screening and is routinely undertaken in the elderly who have a high prevalence of generalized atherosclerosis. While carotid ultrasound is not routinely recommended as a screening tool for clinical cardiovascular disease risk it is an established tool for assessing the burden of generalized atherosclerosis^9-11. As such we sought to investigate whether this low-cost, minimal radiation technique for assessing abdominal aortic calcification is related to an established test for generalized atherosclerosis presence and burden.

Materials and Methods

Materials and Methods are available in the online-only Data Supplement.

Results

Baseline characteristics (Table 1)

Table 1.

Characteristics of the study population stratified by abdominal aortic calcification scores^*

	Low AAC (AAC24 score 0 or 1)	Moderate AAC (AAC24 score 2-5)	Severe AAC (AAC24 score >5)	P value
Number	399 (45)	325 (36)	168 (19)
Age, years	74.8 ± 2.6	74.8 ± 2.4	75.4 ± 2.7	0.031
Body mass index, kg/m2	27.4 ± 4.5	27.0 ± 4.3	26.3 ± 3.7	0.021
Smoking history
Previous, yes (%)	107 (27.0)	128 (39.5)	69 (41.1)
Current, yes (%)	0 (0)	0 (0)	2 (1.2)	<0.001
Diabetes, yes (%)	17 (4.3)	17 (5.2)	7 (4.2)	0.790
Atherosclerotic vascular disease, yes (%)	25 (6.3)	37 (11.4)	27 (16.1)	0.001
Medication history
Antiplatelet medication, yes (%)	93 (23.3)	92 (28.3)	62 (36.9)	0.004
Antihypertensive medication, yes (%)	153 (38.3)	148 (45.5)	79 (47.0)	0.066
Statins medication, yes (%)	51 (12.8)	60 (18.5)	53 (31.5)	<0.001
Blood Pressure
Systolic blood pressure, mmHg	137 ± 17	137 ± 18	140 ± 19	0.093
Diastolic blood pressure, mmHg	74 ± 10	74 ± 11	72 ± 10	0.139
Mean arterial pressure, mmHg	95 ± 11	95 ± 12	95 ± 12	0.995
Lipid profiles†
Total cholesterol, mg/dL	229 ± 40	224 ± 43	226 ± 37	0.294
Low density lipoprotein cholesterol, mg/dL	144 ± 36	141 ± 40	141 ± 34	0.545
High density lipoprotein cholesterol, mg/dL	58 ± 15	55 ± 14	56 ± 15	0.074
Triglycerides, mg/dL	135 ± 58	138 ± 64	144 ± 70	0.296
Carotid ultrasound
Mean CCA-IMT, mm	0.761 ± 0.112	0.782 ± 0.126	0.802 ± 0.122	0.001
Maximum CCA-IMT, mm	0.902 ± 0.132	0.924 ± 0.150	0.959 ± 0.147	<0.001
Presence of any carotid plaque, yes (%)	143 (35.8)	159 (48.9)	120 (71.4)	<0.001
Carotid stenosis ≥ 25%, yes (%)	25 (6.3)	48 (14.8)	54 (32.1)	<0.001

Data expressed as mean ± SD or number and (%). Abbreviations: mmHg, millimetres mercury; CCA-IMT, common carotid artery intimal medial thickness.

^*Lateral spine images measured in 1998 or 1999, B-mode carotid ultrasound measured in 2001.

^†measured in 852 participants.

Participants that did not have a readable lateral spine image (n=417), or did not participatet in the ancillary study of carotid ultrasound (n=154), or were missing traditional cardiovascular risk factors (n=39) were on average older (P<0.001), more likely to have a history of smoking (P=0.005), diabetes (P=0.001) and use low dose aspirin (P=0.002). An overview of the study is presented in supplementary Figure 1 with characteristics of participants included/not included provided in supplementary Table 1. Characteristics for participants included in the study by AAC severity grouping is presented in Table 1. AAC scores ranged from 0 to 17 with a median AAC24 score of 2 and IQR of 0 to 4. Participants who had previously smoked and those who used CVD medication were more likely to have higher abdominal aortic calcification scores (Table 1).

AAC and common carotid artery intimal medial thickness (CCA-IMT)

Abdominal aortic calcification 24 scores were positively correlated with both mean CCA-IMT (r_s=0.124, P<0.001) and maximum CCA-IMT (r_s=0.139, P<0.001) using Spearman's rank correlation. Mean and maximum CCA-IMT was significantly higher in those with moderate (AAC24 scores 2-5) and severe aortic calcification (AAC24 scores >5) compared to those with low AAC (AAC24 score 0 or 1) before (Table 1) and after multivariable-adjustment for baseline risk factors (Figure 1). In multivariable-adjusted linear regression the addition of AAC24 scores modestly improved the variation explained for mean CCA-IMT (traditional risk factor model r²=0.026, traditional risk factor model with AAC24 score r²=0.037) and maximum CCA-IMT (traditional risk factor model r²=0.028, traditional risk factor model with AAC24 score r²=0.040).

Figure 1.

Multivariable-adjusted mean and SEM of the mean CCA-IMT and maximum CCA-IMT categorized by the severity of abdominal aortic calcification (low AAC24 scores 0 or 1; white column, moderate AAC24 scores 2-5; grey column and severe AAC24 scores 6+; black column). * Significantly different (P<0.05) from low AAC (AAC24 score 0 or 1) and ^† significantly different from low (AAC24 scores 0 or 1) and moderate AAC (AAC24 score 2-5). Multivariable-adjustment includes age, body mass index, current smoking, systolic blood pressure, antihypertensive medication, antiplatelet medication, history of atherosclerotic vascular disease hospitalization, year of scan, prevalent diabetes and treatment code (calcium or placebo).

Diagnostic performance of AAC for focal carotid plaque (Table 2)

Table 2.

Relationship between abdominal aortic calcification (AAC) and measures of carotid atherosclerosis.

	Any atheromatous plaque (n = 422)			Moderate stenosis† (n = 127)
	No. (%)	Prevalence Ratio	P value	No. (%)	Prevalence Ratio	P value
Unadjusted
AAC24 score 0-1, (n = 399)	143 (35.8)	(referent)		25 (6.3)	(referent)
AAC24 score 2-5, (n = 325)	159 (48.9)	1.37 (1.15-1.62)	<0.001	48 (14.8)	2.36 (1.48-3.74)	<0.001
AAC24 score 6+, (n = 168)	120 (71.4)	2.01 (1.15-2.36)	<0.001	54 (32.1)	5.16 (3.33-8.01)	<0.001
AAC, no (n = 242)	73 (30.2)	(referent)		10 (4.1)	(referent)
AAC, yes (n = 650)	349 (53.7)	1.78 (1.46-2.19)	<0.001	117 (18.0)	4.37 (2.32-8.21)	<0.001
Multivariable-adjusted
AAC24 score 0-1, (n = 399)	143 (35.8)	(referent)		25 (6.3)	(referent)
AAC24 score 2-5, (n = 325)	159 (48.9)	1.35 (1.14-1.61)	<0.001	48 (14.8)	2.22 (1.39-3.54)	0.001
AAC24 score 6+, (n = 168)	120 (71.4)	1.94. (1.65-2.32)	<0.001	54 (32.1)	4.82 (3.09-7.05)	<0.001
AAC, no (n = 242)	73 (30.2)	(referent)		10 (4.1)	(referent)
AAC, yes (n = 650)	349 (53.7)	1.74 (1.42-2.14)	<0.001	117 (18.0)	4.01 (2.14-7.53)	<0.001

Data expressed as number and (%). P value represents overall P value by Poisson regression with robust variance. Abbreviations: AAC, abdominal aortic calcification. Models were adjusted for traditional CVD risk factors including baseline age, body mass index, current smoking, systolic blood pressure, antihypertensive medication, antiplatelet medication, history of atherosclerotic vascular disease hospitalization, year of scan, prevalent diabetes and treatment code (calcium or placebo).

* Lateral spine images were captured in 1998-1999 while B-mode carotid ultrasound was assessed in 2001.

^†Carotid stenosis ≥25%.

The C-statistic for AAC24 scores alone for carotid plaque and moderate carotid stenosis were 0.66 ± 0.018 and 0.72 ± 0.024. A maximum Youden's index of 0.24 for carotid plaque and 0.33 for moderate stenosis was achieved with an AAC24 score ≥3 and provided a sensitivity of 55% and specificity of 69% for carotid plaque and sensitivity of 71% and specificity of 62% for moderate carotid stenosis (Supplementary Figure 2). Severe AAC (AAC24 score ≥6) had a specificity of 90% for carotid atherosclerosis and 85% for moderate carotid stenosis while low AAC (AAC24 0 or 1) had a sensitivity of 66% for carotid atherosclerosis and 80% for moderate carotid stenosis. The severity of AAC was strongly associated with the presence of carotid atheroma assessed by presence or absence of plaque and degree of focal stenosis in both unadjusted (Table 2) and multivariable-adjusted models (Table 3). When participants were dichotomized by the presence of any AAC (AAC24 score 0 vs. AAC24 score 1-17) those with any AAC were more likely to have focal carotid atherosclerotic plaques and stenosis (Table 3). Similarly participants with higher degrees of AAC were more likely to have carotid plaque and moderate stenosis compared to those with low AAC24 score 0 or 1 before and after adjusting for baseline risk factors participants (Table 3). In the 852 individuals with lipid profiles available the inclusion of total cholesterol and high density lipoprotein cholesterol in the multivariable-adjusted models did not substantially alter the main findings (Supplementary Table 2 and Figure 3). While the use of cardiovascular medications such as statins are currently not included in traditional risk factor assessment, in further analysis we added these commonly used medications to the traditional models. The addition of these medications did not substantially alter the main findings of the study (difference in C-statistic for plaque +0.07, P<0.001, moderate carotid stenosis +0.10, P<0.001).

Table 3.

Reclassification of athermanous plaque risk individuals with the addition of AAC24 scores to traditional cardiovascular risk factors.

	Traditional cardiovascular risk factors + AAC24 scores
Traditional cardiovascular risk factors	< 40%	40-50%	≥ 50%	Reclassified higher	Reclassified lower	Correctly reclassified
Participants with atheromatous plaque (n = 420)
< 40%	33	19	16
40-50%	70	59	67	102 (24.3%)	109 (26.0%)	−7 (−1.7%)
≥ 50%	6	33	117
Participants without atheromatous plaque (n = 470)
< 40%	84	14	8
40-50%	141	58	54	76 (16.2%)	201 (42.8%)	125 (26.6%)
≥ 50%	18	42	51

Traditional cardiovascular risk model includes age, body mass index, current smoking, systolic blood pressure, antihypertensive medication, antiplatelet medication, history of atherosclerotic vascular disease hospitalization, year of scan, prevalent diabetes and treatment code (calcium or placebo). Overall net reclassification improvement (NRI) = 0.249, P<0.001, with events −1.7%, P=0.630 and without events 26.6%, P<0.001. IDI = 0.065, P<0.001.

Addition of AAC24 scores to traditional risk factors

The addition of the presence of AAC24 scores to traditional cardiovascular risk factors markedly improved the C-statistic AUC from poor (0.60) to fair (0.68) for any plaque and from fair (0.62) to good (0.73) for moderate carotid stenosis (Figure 2). Data on the category-based net reclassification improvements are shown in Table 3 and 4 with overall improvements being 0.249 for plaque and 0.343 for moderate carotid stenosis. Reclassification plots with superimposed cut-points are shown in Supplementary Figure 4. Integrated discrimination was improved after the addition of AAC24 scores to traditional risk factors (0.064, P<0.001 for carotid plaque and 0.084, P<0.001 for moderate carotid stenosis). Model calibration was not significantly different (P<0.05) as assessed by the Hosmer-Lemenshow test for either plaque or moderate stenosis.

Figure 2.

Improvement to the C-statistic with the addition of AAC divided into three groups (low, moderate and high; solid grey lines) and AAC24 scores (continuous 0-24; solid black lines) to traditional cardiovascular risk factors (dashed black line) for the identification of individuals with any carotid plaque (A) and moderate carotid stenosis (B). All improvements to the C-statistic with the addition of AAC measures, P<0.001.

Table 4.

Reclassification of moderate carotid stenosis risk with the addition of AAC24 scores to traditional cardiovascular risk factors.

	Traditional cardiovascular risk factors + AAC24 scores
Traditional cardiovascular risk factors	< 10%	10-15%	≥ 15%	Reclassified higher	Reclassified lower	Correctly reclassified
Participants with moderate carotid stenosis ≥ 25% (n = 127)
< 10%	8	4	1
10-15%	19	17	24	29 (22.8%)	27 (21.3%)	2 (1.6%)
≥ 15%	3	5	46
Participants without moderate carotid stenosis ≥ 25% (n = 765)
< 10%	108	20	17
10-15%	287	83	81	118 (15.4%)	368 (48.1%)	250 (32.7%)
≥ 15%	35	46	88

Traditional cardiovascular risk model includes age, body mass index, current smoking, systolic blood pressure, antihypertensive medication, antiplatelet medication, history of atherosclerotic vascular disease hospitalization, year of scan, prevalent diabetes and treatment code (calcium or placebo). Overall net reclassification improvement (NRI) = 0.343, P<0.001, with events 1.6%, P=0.789 and without events 36.7%, P<0.001. IDI = 0.084, P<0.001.

Similarly the inclusion of the AAC24 scores the inclusion of the graded severity of AAC in 3 groups (low, moderate and high) led to improvements to the C-statistic for plaque and moderate stenosis (+0.07 and +0.10 respectively, P<0.001), net reclassification (plaque; events reclassification improvement 2.6%, P=0.471 and non-events 16.6%, P<0.001, overall NRI 0.192, P<0.001 and moderate stenosis; events reclassification improvement 8.7%, P=0.131 and non-events 21.7%, P<0.001, overall NRI 0.304, P<0.001). Integrated discrimination was also improved for plaque (0.059, P<0.001) and stenosis (0.070, P<0.001).

Further analyses

Interaction tests between baseline cardiovascular risk factors and AAC24 scores for any focal carotid plaque and moderate carotid stenosis identified significant interactions (P<0.1) between AAC24 scores and systolic blood pressure for any focal carotid plaque and body mass index and antiplatelet use for moderate carotid stenosis (Supplementary Table 3). Further sensitivity analyses excluding older women with a BMI greater than 30 kg/m² (n=190), those with an eGFR of < 60 ml/min/1.73m² (n=252) or those with a history of atherosclerotic vascular disease (n=89) did not substantially change the overall results of the study (data not shown).

Discussion

Our findings demonstrate that vascular calcification identified from images obtained on a DXA machine are related to both CCA-IMT and focal carotid plaque before and after adjusting for traditional cardiovascular risk factors. These findings using a low-cost, minimal radiation technique support previous findings by others using higher radiation modalities that abdominal aortic calcification are a marker of generalized atherosclerosis at other sites ^{20, 21}. However these findings need to be confirmed in further studies using DXA-derived l images with other established measures of generalized atherosclerosis.

Another measure of vascular calcification; coronary artery calcification (CAC) score assessed by CT has also been shown to be associated with increased CCA-IMT^{22, 23} and carotid plaque ²⁴ further supporting the concept that vascular calcification is related to both of these measures of subclinical atherosclerosis. Interestingly focal plaque and intimal medial thickness may represent different stages of the pathological processes and provide different information regarding future cardiovascular risk^25-27, suggesting abdominal aortic calcification may capture aspects of both the early and the late pathological processes. As such, assessment of AAC may be useful as a surrogate endpoint for interventions targeting both early and late stage atherosclerosis.

The use of CT to calculate an Agatston score to quantify CAC ²⁸, has been recommended to inform treatment decision on cardiovascular risk where traditional risk-based treatment decisions are uncertain ²⁹. Criqui and colleagues ³⁰ recently found abdominal aortic calcification assessed by CT was a better predictor of CVD and all-cause mortality than CAC with independent and additive prognostic value for coronary heart disease and hard CVD events to CAC alone suggesting that AAC may be a better marker of extracoronary atherosclerosis than CAC. However concerns regarding the cost and radiation exposure of CT have prevented the implementation of widespread screening using this imaging modality ²⁹. Therefore other modalities such as lateral spine images from DXA machines are needed. Lateral spine DXA imaging provide a substantially lower radiation exposure (0.013 mSv) compared to abdominal CT (8 mSv) ³¹, coronary artery CT (2.3 mSv)³² and newer low radiation multi-slice abdominal CT (1.5 mSv)³³ and therefore may be considered as a potential candidate for routine cardiovascular risk assessment in low to intermediate cardiovascular risk patients where there is currently no routine screening recommended.

Regarding agreement with other imaging modalities for aortic calcification, DXA-based AAC24 scores have good-very good agreement with radiography ^{34, 35} and moderate-good agreement with CT ^{36, 37}. The two studies comparing DXA-based AAC24 scores and standard radiographs were large studies with experienced investigators and reported ICC of between 0.80-0.82 ^{34, 35}, however the studies comparing DXA-based AAC24 scores with CT were considerable smaller (n=40-100), and only one study reported an ICC of 0.93³⁷ while the other study ³⁶ used Spearman rank correlation (r_s=0.58, P<0.0001) which does not measure agreement between tests³⁸. Therefore the evidence for the agreement between AAC by DXA-based techniques with CT remains limited.

Despite this, our results suggest that when clinicians detect moderate to severe calcification of the abdominal aorta on lateral spine images for osteoporosis screening they should be aware that these individuals are likely to have generalized atherosclerosis at other vascular beds and consider appropriate cardiovascular disease risk assessment and management. Conversely, if there is no or low calcification then these individuals are less likely to have atherosclerosis at other vascular beds allowing improved monitoring of patients and targeting of scarce healthcare resources.

Interestingly we observed interactions between AAC24 scores and blood pressure for carotid atherosclerosis in elderly women with a stronger relationship in those with systolic blood pressure below 140 mmHg. Hypertension and vascular calcification represent two closely related processes with medial calcification thought to promote more widespread arterial rigidity and hypertension than intimal calcification³⁹. Given this the observed relationship between AAC and carotid atherosclerosis in elderly women with systolic blood pressure below 140 mmHg may be due to a greater proportion of elderly women with intimal calcification. This is consistent with the current view that intimal calcifications are more strongly related to atherosclerosis than medial calcification which are thought to be due to arteriosclerosis⁴⁰. Similarly obesity and abnormal lipid profiles are recognized risk factors for intimal calcification³, while antiplatelet medications are used for atherothrombotic vascular disease and as such the observed interactions are likely to reflect the different aetiologies of the vascular calcifications. However as these are hypothesis generating analyses the findings need to be confirmed in other studies.

The addition of abdominal aortic calcification measures to traditional cardiovascular risk factors substantially improved all indices of classification of subclinical atherosclerosis using poisson regression, ROC analysis as well as the newer net reclassification and integrated discrimination, demonstrating that AAC is likely to be a surrogate marker of generalized atherosclerosis in elderly women independent of traditional risk factors. Indeed the use of AAC24 scores alone yielded substantially better discrimination of individuals with carotid atherosclerosis than traditional cardiovascular risk factors alone, while the addition of AAC24 scores to traditional risk factors led to one in three participants being correctly reclassified at lower risk. The reclassification of low-risk individuals is particularly important as it identifies a population that is unlikely to receive any net benefit from interventions and provides a rationale for less intensive clinical management. However despite correct reclassification in low risk individuals without carotid atherosclerosis, there is insufficient evidence to determine whether AAC24 scores from DXA machines can improve current cardiovascular disease risk prediction.

Regarding strengths and limitations of the study, the strengths include the representative nature of the cohort of elderly women who had assessment of traditional cardiovascular risk factors, B-mode carotid ultrasound and abdominal aortic calcification assessed by a single highly experienced investigator (JTS) blinded to the B-mode carotid ultrasound measures of atherosclerosis. B-mode carotid ultrasound is a highly sensitive measure of both early and late subclinical atherosclerosis ⁴¹ and is considered to indicate the degree of generalized atherosclerosis⁹, as well as providing insights into the development of vascular pathology. Next lateral spine imaging using DXA machines are cost-effective for fracture prediction⁴² and have been approved for the assessment of some patients referred for osteoporosis diagnosis⁴³. Indeed the age group of women who were participants in this population-based cohort commonly have DXA assessments performed to assess fracture risk. Evaluation of these images not only for fracture but also abdominal aortic calcification burden may further improve the cost/benefit ratio of this procedure leading to improved targeting of limited healthcare resources.

However the following limitations must also be mentioned. Firstly the study was observational and as such we cannot rule out bias, especially given the carotid ultrasound measurements were performed after the lateral spine imaging and the women that did not attend the year 3 carotid ultrasound measurements had higher baseline cardiovascular risk. Secondly the AAC24 point scale scoring system is semi-quantitative and operator dependant, however for this study a single highly experienced investigator (JTS) read all images. Indeed a previous study demonstrated very good inter-operator agreement between JTS and another experienced investigator with intra-class correlation coefficients of 0.89 (95% CI: 0.80-0.94) for AAC24 scores ⁴⁴. Additionally this study did not assess atherosclerosis and calcification at other vascular beds using established methodology such as coronary artery or thoracic aortic calcification by CT and as such further studies are needed to validate these findings. Finally this cohort consisted of elderly Caucasian women with an average age of 78 at carotid ultrasound measurement and these findings need to be confirmed in younger cohorts, different ethnicities and elderly men.

In conclusion, we demonstrated abdominal aortic calcification scores identified on images from a DXA machine were associated with established measures of generalized extracoronary atherosclerosis. Abdominal aortic calcification scores were independently and additively predictive for carotid atherosclerosis compared to traditional cardiovascular risk factors including age, smoking, body mass index, systolic blood pressure, antiplatelet medications, atherosclerotic vascular disease, diabetes and prescription of antihypertensive medication.

Significance.

These findings identify this low-cost, minimal radiation technique from bone densitometers already widely used for osteoporosis screening as a potential test for routine assessment of generalized atherosclerosis burden in elderly women.

Abbreviations and Acronyms

AAC

Abdominal aortic calcification

AAC24

Abdominal aortic calcification 24 point scores

ASVD

Atherosclerotic vascular disease

CAC

Coronary artery calcification

CCA-IMT

Common carotid artery intimal medial thickness

CT

Computed tomography

DXA

Dual energy X-ray absorptiometry