Abdominal aortic calcification on vertebral morphometry images predicts incident myocardial infarction

Authors


Abstract

Abdominal aortic calcification (AAC) measured on spine X-rays is an established risk factor for cardiovascular disease. We investigated whether AAC assessed using vertebral morphometry and a recently developed scoring system (AAC-8) is reliable and associated with cardiovascular risk factors or events. A total of 1471 healthy postmenopausal women and 323 healthy middle-aged and older men participated in 5 and 2 year trials of calcium supplements, respectively. AAC-8 was assessed on vertebral morphometry images at baseline and follow-up. In addition, 163 men also had coronary artery calcification measured using computed tomography. Cardiovascular events during the trials were independently adjudicated. We found strong inter- and intrameasurer agreement for AAC-8 (κ > 0.87). The prevalence of AAC increased with age (p < .01) in women and in men. AAC was associated with many established cardiovascular risk factors, with serum calcium in women (p = .002) and with higher coronary calcium scores in men (p = .03). Estimated 5 year cardiovascular risk increased with increasing AAC-8 score (p < .001) in women and in men. The presence of AAC independently predicted myocardial infarction (MI) in women [hazards ratio (HR) = 2.30, p = .007] and men (HR = 5.32, p = .04), even after adjustment for estimated cardiovascular risk in women. In women, AAC independently predicted cardiovascular events (MI, stroke, or sudden death) (HR = 1.74, p = .007), and changes in AAC-8 score over time were associated with MI and cardiovascular events, even after adjustment for estimated cardiovascular risk. In summary, scoring AAC on vertebral morphometric scans is a reproducible method of assessing cardiovascular risk that independently predicts incident MI and cardiovascular events, even after taking into account traditional cardiovascular risk factors. © 2010 American Society for Bone and Mineral Research

Introduction

Abdominal aortic calcification (AAC) is an established risk factor for cardiovascular disease. Previous studies have shown that AAC seen on plain spine X-rays is associated with subsequent death from cardiovascular disease1–3 and incidence of coronary heart disease and cardiovascular disease,2 myocardial infarction (MI),4 and stroke.5, 6 These relationships persist even after traditional risk factors for cardiovascular disease are taken into account. Recent studies have shown that it is possible to assess AAC on vertebral morphometric images obtained by dual-energy X-ray absorptiometry (DXA).7–9 In a case-control study of older women, a newly developed score, the AAC-8, was shown to independently predict MI and stroke after controlling for traditional cardiovascular risk factors.9

Recently, we have completed follow-up of a cohort of older women and a cohort of middle-aged and older men who took part in clinical trials of calcium supplementation. All participants had vertebral morphometric images performed at baseline and at study completion and independent adjudication of vascular events during follow-up. We set out to determine whether measurement of AAC, as scored using AAC-8, was reliable and reproducible and whether AAC assessed by DXA was associated with cardiovascular risk factors or incident cardiovascular events.

Methods

Participants

A total of 1471 healthy postmenopausal women took part in a 5 year randomized, controlled trial of calcium supplements, and 323 healthy middle-aged and older men took part in a 2 year randomized, controlled trial of calcium supplements. The study design and results of both trials have been previously published.10–12 Briefly, women were aged greater than 55 years, free from major medical conditions, and not taking agents for osteoporosis, including hormone-replacement therapy or vitamin D supplements in doses greater than 1000 IU/day. Men were aged greater than 40 years, free from major medical conditions, and not taking agents that might impact on calcium metabolism. Moreover, 244 men continued in an open-ended observational extension study during which no study medication was given. Vitamin D–deficient participants (serum 25-hydroxyvitamin D < 25 nmol/L) were excluded from both studies.

Abdominal aortic calcification scoring

Vertebral morphometry was performed at baseline and at 60 months in the women using a Lunar Expert densitometer and at baseline and at 2 years in the men using a GE Prodigy densitometer. The AAC-8 score was calculated from vertebral morphometric images as the sum of the total length of calcification of the anterior and posterior aortic walls in front of the L1–4 vertebral segments.8 For each wall, the aggregate length of calcification is scored between 0 and 4 relative to vertebral body height. Thus absent calcification is scored as 0, and if the aggregate length of calcification is 1 or less vertebral body height, the score is 1. If the aggregate length of calcification is greater than 1 and less than or equal to 2 vertebral heights, the score is 2 and so on to a maximum score of 4. The scores of each wall are summed to give an AAC-8 score that ranges between 0 and 8.8

Coronary calcium scores

A total of 163 men (80 placebo group, 83 calcium 1200 mg/day group) had coronary calcium scores determined by computed tomography (CT) an average of 3.5 years after study entry. Scoring was performed using a 64 slice CT scanner (LightSpeed VCT, GE) and analyzed using the Agatston system13 with manufacturer-supplied software (Smartscore, GE). The coronary calcium score (Agatston score) and the calcium score for the proximal ascending aorta including the aortic valve (aortic calcium score) were determined for each subject.

Quality assurance

Interscore reliability for the AAC-8 was established from comparison of scores made by two independent measurers (MB and TW) on 30 randomly selected images. After confirmation of acceptable intermeasurer agreement, one measurer (TW) scored 100 randomly chosen images on two occasions not less than 1 day apart to assess intrameasurer consistency. The quality-assurance process was performed for images obtained from both the Expert and Prodigy densitometers. Inter- and intrameasurer agreement was assessed using the kappa coefficient (κ), and κ > 0.80 was considered to suggest strong agreement.

Cardiovascular event assessment

Assessment of cardiovascular events in these cohorts has been described previously in detail.12 Briefly, participants were seen every 6 months during the trials, and adverse events were inquired after and recorded, but questions about specific symptoms were not asked. Myocardial infarction (MI) was defined in accordance with the Joint European Society of Cardiology/American College of Cardiology Committee criteria for acute, evolving, or recent MI.14 Thus an event was classified as an MI if there was the typical rise and fall of a biochemical marker of myocardial necrosis with at least one of the following: ischemic symptoms, development of pathologic Q waves on the electrocardiogram, electrocardiographic changes indicative of ischemia (ST-segment elevation or depression), or coronary artery intervention. Stroke was defined as a sudden focal neurologic deficit of presumed vascular origin that persisted for more than 24 hours or that led to death, and transient ischemic attack was defined as a sudden focal neurologic deficit of presumed vascular origin that persisted for less than 24 hours. Sudden death was defined in accordance with ICD-9 (798.2) as death occurring in less than 24 hours from onset of symptoms not otherwise explained. All self-reported events were independently verified using medical records and adjudicated by a cardiologist or neurologist. Additionally, causes of death were obtained from hospital records or death certificates, and a search of the national hospital admissions database for unreported events was undertaken in the women's study.

Statistical analysis

Because the women's study was much larger, with longer follow-up, and included participants at higher risk of vascular events, the primary analyses were carried out in the women's study and then repeated in the men's study to see whether similar findings were obtained.

The cohorts were divided by the presence (AAC-8 > 0) or absence (AAC-8 = 0) of baseline AAC, and baseline characteristics were compared between the groups using t tests for continuous variables and Fisher's exact test for categorical variables. Multivariate stepwise logistic regression analysis for the presence of AAC was carried out, entering variables into the model where they were plausibly biologically related to vascular calcification or had p < .15 in the univariate analysis. Estimated 5 year cardiovascular risk was calculated as described by Gaziano and colleagues.15 For women, some of the required laboratory data were not available, so the non-laboratory-based model was employed (which uses age, systolic blood pressure, smoking status, history of diabetes, history of blood pressure treatment, and body mass index), whereas for men the laboratory model was employed (which uses the same variables as the nonlaboratory model but uses total cholesterol instead of body mass index). Estimated 5 year cardiovascular risk was compared with AAC-8 score by analysis of variance (ANOVA) and trend testing with orthogonal contrasts. Linear regression models were used to estimate the amount of variance in estimated 5 year cardiovascular risk explained by AAC and coronary calcium scores. The cohort of men was divided by the presence or absence of AAC at 2 years, and coronary calcium scores were compared between the groups using Wilcoxon tests. Categorical modeling was used to assess changes in the presence or absence of AAC with time. Survival analyses assessing the contributions of treatment allocation, AAC, and estimated cardiovascular risk on incidence of cardiovascular events were performed using Kaplan-Meier analysis and Cox proportional hazards modeling. Finally, the area under the curves (AUCs) obtained from receiver operator characteristics (ROC) analyses were calculated and used to compare the accuracy of AAC and estimated 5 year cardiovascular risk in predicting cardiovascular events. All analyses were performed using the SAS software package (Version 9.1, SAS Institute, Cary, NC, USA) or Prism (Version 5, GraphPad Software, San Diego, CA, USA). All tests were two-tailed and p < .05 was considered significant.

Results

Quality assurance

The κ for intermeasurer agreement for AAC-8 score for the initial 30 images was 0.87 [95% confidence (CI) 0.78–0.95]. The κ values for intrameasurer agreement for AAC-8 score were 0.89 (95% CI 0.81–0.96) for images from the Expert densitometer and 0.91 (95% CI 0.81–1.00) for images from the Prodigy densitometer.

Subject disposition

A total of 1471 women were randomized to placebo (n = 739) or calcium 1 g/day (n = 732). Then 1424 baseline images and 1014 five year images were scored for AAC-8. Missing scores were due to scans not being undertaken (26 baseline, 444 follow-up) or scan quality (21 baseline, 13 follow-up). A total of 323 men were randomized to placebo (n = 107), calcium 600 mg/day (n = 108), or calcium 1200 mg/day (n = 108). Then 320 baseline images and 298 two year images were scored for AAC-8. Missing scores were due to scans not being undertaken (0 baseline, 18 follow-up) or scan quality (3 baseline, 7 follow-up). The baseline characteristics of the cohorts are shown in Table 1. The mean duration of follow-up of the women was 4.4 years and of the men was 2.0 years in the trial, 1.7 years for the 244 men in the observational extension, and 3.3 years for the combined total follow-up.

Table 1. Baseline Characteristics of the Cohorts Divided by the Presence or Absence of Abdominal Aortic Calcification (AAC)
 WomenMen
 Baseline AAC absent (n = 717)Baseline AAC present (n = 707)PBaseline AAC absent (n = 263)Baseline AAC present (n = 57)p
  • Data are mean (SD) or percentage except where stated.

  • a

    Cholesterol was measured in all the men and a subgroup of 223 women.

  • b

    Estimated glomerular filtration rate (eGFR) as recommended by Mathew TH. The Australasian Creatinine Consensus Working Group. Chronic kidney disease and automatic reporting of estimated glomerular filtration rate: a position statement. Med J Aust. 2005;183:138-141.

  • c

    Any use of vitamin D supplements throughout study.

  • d

    Data are median (range).

  • e

    Bone density T-score ≤ −2.5 at proximal femur or spine.

Age (years)73.5 (4.1)75.0 (4.2)<.00155.5 (9.7)61.1 (11.3)<.001
Weight (kg)67.7 (12.3)66.0 (9.9).00583.3 (12.4)81.0 (10.3).19
Blood pressure (mm Hg)
 Systolic133.5 (21.6)139.8 (22.9)<.001130.7 (13)132.4 (15.7).37
 Diastolic70.2 (9.9)71.5 (10.7).0378.8 (8.0)77.5 (7.5).29
Cholesterol (mmol/L)a
 Total6.5 (1.0)6.7 (1.2).085.7 (0.9)5.4 (0.8).08
 High-density lipoprotein1.7 (0.4)1.6 (0.5).281.5 (0.4)1.5 (0.4).37
 Low-density lipoprotein4.1 (1.0)4.4 (1.1).033.6 (0.8)3.4 (0.7).08
 Triglycerides1.5 (0.9)1.6 (0.9).561.4 (0.7)1.2 (0.8).09
Glucose (mmol/L)5.1 (0.63)5.1 (0.77).515.1 (0.5)5.0 (0.4).30
Serum creatinine (umol/L)88 (13)87 (15).0694 (11)94 (11).78
eGFR (mL/min/1.73m2)b61.5 (10.2)60.4 (11.0).0578.3 (11.4)76.5 (11.4).28
Serum total calcium (mmol/l)2.31 (0.09)2.33 (0.09)<.0012.35 (0.09)2.34 (0.08).42
Serum 25-hydroxyvitamin D (nmol/L)55 (19)53 (17).0592 (33)96 (34).43
 <50 nmol/L (%)4245.31651.0
Vitamin D supplement use (%)c2324.572733.41
 Dose (IU/day)d400 (200-2700)400 (400-3300).91400 (200-2200)400 (400-1600).48
Total-body lean mass (kg)36.4 (4.2)35.9 (3.9).0459.5 (7.1)59.0 (5.5).51
Total-body fat mass (kg)27.5 (10.5)26.5 (8.4).0719.8 (7.7)18.2 (6.9).17
Osteoporosis (%)e23.424.6.620.01.8.18
European ethnicity99.099.7.1892.494.7.57
History of
 Ischemic heart disease5.310.3<.0010.40.01.0
 Transient ischemic attack0.10.4.370.00.01.0
 Stroke0.40.7.500.40.01.0
 Hypertension25.532.5.0046.812.3.18
 Diabetes2.13.3.190.40.01.0
 Dyslipidemia6.610.5.014.91.8.48
Current hypertension treatment19.127.9<.00120.531.6.08
Smoking status
 Current smoker2.13.8.063.41.8.88
 Former smoker3541.9.00943.738.6.56

Women

Figure 1 shows the distribution of AAC-8 scores at baseline. The distribution was highly skewed, with 50.4% of the cohort having no detectable AAC. Table 1 shows the baseline characteristics of the cohort divided by the presence or absence of AAC. The presence of AAC was associated with a number of established risk factors, including older age, increased blood pressure, increased low-density lipoprotein (LDL) cholesterol, and history of ischemic heart disease, hypertension, dyslipidemia, and smoking. The proportion of women with AAC increased linearly with age (<70 years, prevalence 31.8%; 70 to 75 years, 46.0%; 75 to 80 years, 59.8%; 80 years or older, 60.8%; p < .001; test for linear trend p < .001). In a multivariate logistic regression analysis, seven variables independently predicted the presence of AAC (Table 2).

Figure 1.

The distribution of AAC-8 scores in the cohorts of women and men at baseline and the distribution of the change in AAC-8 scores at 5 years in women and 2 years in men.

Table 2. Predictors of Baseline Abdominal Aortic Calcification in the Women's Cohort From a Multivariate Logistic Regression Analysis
VariableaOdds ratio (95% confidence interval)p
  • a

    Reference units: age, 1 year; weight, 1 kg; blood pressure, 1 mm Hg; serum calcium, 0.1 mmol/L

Age1.07 (1.04–1.10)<.001
Weight0.99 (0.98–1.00).02
Systolic blood pressure1.01 (1.01–1.02)<.001
Total serum calcium1.23 (1.08–1.40).002
Previous smoker1.46 (1.16–1.83).001
History of ischemic heart disease1.97 (1.28–3.02).002
Current hypertension treatment1.34 (1.03–1.75).03

The mean (SD) estimated 5 year cardiovascular risk was 6.4% (3.2%). In women with baseline AAC, the estimated risk was 7.0% (3.4%) compared with 5.8% (2.9%) in women without AAC (p < .001). Figure 2 shows that when the cohort was divided in six groups by amount of AAC, there was a linear increase in the estimated cardiovascular risk with increasing AAC-8 score (p < .001), and the differences between groups were statistically significant (p < .001). Findings were similar when the cohort was divided into three groups (data not shown).

Figure 2.

Box and whiskers plots showing the estimated 5 year cardiovascular risk for the cohort of women divided into six groups by baseline AAC-8 score. The box shows the median with the interquartile range, and the whiskers show the range. The differences between groups were statistically significant (p < .001), and the estimated cardiovascular risk increased linearly with increasing AAC-8 score (p < .001).

Thirty-four women with baseline AAC had a subsequent MI compared with 15 women without AAC. After adjusting for treatment allocation (to calcium or placebo), the hazard ratio (HR) was 2.30 (95% CI 1.25–4.22, p = .007; Fig. 3). Thirty-five women with baseline AAC had a subsequent stroke compared with 22 women without AAC (HR = 1.48, 95% CI 0.86–2.55, p = .16). Sixty-seven women with baseline AAC had a subsequent cardiovascular event (MI, stroke, or sudden death) compared with 38 women without AAC (HR = 1.74, 95% CI 1.16–2.00, p = .007; see Figure 3). Adjusting for treatment compliance and a time-by-treatment interaction did not alter the relationship between baseline AAC and incident MI, stroke, or cardiovascular event in these models. Estimated 5 year cardiovascular risk also predicted incident MI, stroke, and cardiovascular events (HR = 1.14–1.15 , p < .001 for all three models). After adjusting for estimated 5 year cardiovascular risk, the presence of baseline AAC remained an independent predictor of subsequent MI (HR = 2.05, 95% CI 1.09–3.86, p = .03) but not stroke (p = .16) or cardiovascular event (p = .08). The sensitivity, specificity, and positive and negative predictive values of the presence of baseline AAC for subsequent MI were 69.4%, 51.1%, 4.8%, and 97.9% respectively. Using ROC curve analysis, estimated 5 year cardiovascular risk was more predictive of subsequent MI and cardiovascular event than AAC (MI: AUC for estimated 5 year cardiovascular risk: 0.69, 95% CI 0.62–0.76 versus AUC for AAC: 0.60, 95% CI 0.54–0.67, p for difference in AUC = .038; cardiovascular event: AUC for estimated 5 year cardiovascular risk: 0.69, 95% CI 0.64–0.74 versus AUC for AAC: 0.58, 95% CI 0.53–0.62, p for difference in AUC < .001). Adding AAC to the estimated 5 year cardiovascular risk did not cause a statistically significant increase in the AUC for either MI or cardiovascular events.

Figure 3.

The cumulative incidence of myocardial infarction (upper panel) and cardiovascular events (any of myocardial infarction, stroke, or sudden death) (lower panel) in the cohort of women by the presence or absence of abdominal aortic calcification at baseline.

The distribution of AAC-8 scores at 5 years remained highly skewed and similar to that seen at baseline. Figure 1 shows the distribution of the change in AAC-8 scores over 5 years. Seventy-one percent of women had no change in AAC-8 score, but the proportion of women with AAC at 5 years was 56.5%, a significant increase from baseline (p < .001).

Finally, we assessed whether changes in AAC-8 score over time were associated with cardiovascular events. A 1 unit increase in AAC-8 score from baseline was associated with an increased incidence of MI (HR = 2.14, 95% CI 1.41–3.26, p < .001) and cardiovascular events (HR = 1.62, 95% CI 1.12–2.36, p = .011) but not stroke (HR = 1.28, 95% CI 0.71–2.34, p = .41). After adjusting for estimated 5 year cardiovascular risk, change in AAC-8 score remained independently associated with MI (HR = 2.09, 95% CI 1.36–3.22, p < .001) and cardiovascular events (HR = 1.54, 95% CI 1.05–2.27, p = .03) but not stroke (p = .56).

Men

Figure 1 shows the distribution of AAC-8 scores at baseline. Also, 82.2% of the cohort had no detectable AAC, and only 1.3% had an AAC-8 score greater than 1. Table 1 shows the baseline characteristics of the cohort divided by the presence or absence of AAC. In univariate analyses, age was the only factor associated with AAC. The proportion of men with AAC increased with age (<50 years, prevalence 13.3%; 50 to 60 years, 12.5%; 60 to 70 years, 23.1%, ≥70 years, 38.9%; p = .001). In a multivariate logistic regression, age was the only factor that independently predicted the presence of AAC [odds ratio (OR) = 1.06, 95% CI 1.03–1.09, p < .001].

The mean (SD) estimated 5 year cardiovascular risk was 1.8% (1.3%). In men with baseline AAC, the estimated risk was 2.5% (1.9%) compared with 1.7% (1.1%) in men without AAC (p < .001).

In addition,158 men had both an assessment of coronary calcification and an assessment of AAC at 2 years. Figure 4 shows that coronary calcium scores were higher in men with AAC than in men without AAC (p = .03). Men with AAC also had higher aortic calcium scores (p = .003) than men without AAC. In linear regression models, AAC accounted for 8.5% of the variance in estimated 5 year cardiovascular risk, whereas the coronary calcium score accounted for 19.9% of the variance.

Figure 4.

Scatter plot showing the relationship between abdominal aortic calcification and coronary calcium scores in the cohort of men. The bars show the median ± interquartile range. Data on 44 men with a coronary calcium score of 0 are not shown for clarity (abdominal aortic calcification was present in 10 of these men and absent in 34).

Three men with baseline AAC had a subsequent MI compared with three men without AAC. After adjusting for treatment allocation (to calcium or placebo), the HR was 5.32 (95% CI 1.07–26.6, p = .04). There were no strokes during follow-up, and two sudden deaths occurred in individuals who previously had an MI during the follow-up period. Estimated 5 year cardiovascular risk also predicted incident MI (HR = 1.85, p = .01). After adjusting for the estimated 5 year cardiovascular risk, the presence of AAC was not associated with subsequent MI (HR = 1.50, 95% CI 0.2–11.0, p = .69).

The distribution of AAC-8 scores at 2 years remained highly skewed and similar to that seen at baseline. Figure 1 shows the distribution in the change in AAC-8 scores over 2 years. Eighty-nine percent of men had no change in the AAC-8 score, but the proportion of men with AAC at 2 years was 26%, a significant increase from baseline (p < .001). The study was too small to assess whether changes in AAC-8 score were associated with cardiovascular events.

Discussion

We found that assessment of AAC using the AAC-8 score was reproducible and that the presence of AAC was associated with many established risk factors for cardiovascular disease and was an independent predictor of subsequent cardiovascular events in women. The AAC-8 score has high inter- and intraobserver agreement, making it a suitable for use in clinical practice, as has also been reported by others.8, 9 The AAC-8 score was reproducible and did not change in most of the individuals over time, whereas at a cohort level the proportion of individuals with AAC increased over time. In older women, the presence of AAC was associated with established cardiovascular risk factors, including age, blood pressure, LDL cholesterol, and history of ischemic heart disease, hypertension, dyslipidemia, and smoking. The AAC-8 score also was positively linearly related with estimated 5 year cardiovascular risk. Additionally, in men, the presence of AAC was associated with increased coronary calcium scores, another established cardiovascular risk factor.16 Finally, the presence of AAC independently predicted future MI and cardiovascular events and remained an independent predictor of subsequent MI, with a hazard ratio of 2.05, even after adjusting for estimated 5 year cardiovascular risk. Changes in the AAC-8 score over time also were associated with MI and cardiovascular events, even after adjusting for estimated 5 year cardiovascular risk. Taken together, the results strongly suggest that the presence of AAC provides useful information in assessing cardiovascular risk beyond that provided by traditional risk factors. AAC also was positively related to baseline serum calcium level. We are not aware that relationships between serum calcium and AAC have been described previously outside the setting of chronic renal failure, where vascular calcification occurs commonly and has been inconsistently related to serum calcium levels.17

Calcification of the aorta has long been recognized as a risk factor for cardiovascular events and mortality. Witteman and colleagues reported that AAC was associated with an up to sixfold increased risk of cardiovascular death in men that was independent of major cardiovascular risk factors.1 Wilson and colleagues developed a 24-point scoring method2, 18 and applied this to the Framingham Heart Study to show that increasing AAC was positively associated with incidence of coronary heart disease and cardiovascular disease, as well as cardiovascular mortality, after adjusting for traditional cardiovascular risk factors.2 Rodondi and colleagues also reported that AAC was associated with increased all-cause and cardiovascular mortality in the Study of Osteoporotic Fractures.3 Others have reported associations between AAC and MI,4 stroke,5, 6 congestive heart failure,19 and peripheral vascular disease.6 AAC is also associated with risk factors for cardiovascular disease, including increased age, male sex, smoking, and the presence of diabetes mellitus, hypertension, and renal failure.20 In all these studies, however, AAC was measured by plain X-rays of the spine1–6, 19, 20 or computed tomography.20 It remains unclear whether vascular calcification such as AAC directly causes cardiovascular events or is merely a marker of elevated cardiovascular risk owing to atherosclerosis and other cardiovascular risk factors.

Schousboe and colleagues have led developments in the assessment of AAC by DXA. First, in a pilot study they showed that AAC assessed by DXA has a very high level of agreement with AAC assessed by plain X-rays using 24-point scoring systems.7 Next, they developed a simplified 8-point scoring system (AAC-8) and validated both 8- and 24-point scores obtained by DXA against comparable scores obtained from plain X-rays.8 Finally, they performed a case-control study of 408 women participating in a clinical trial who sustained a MI or stroke. Both the 8- and 24-point scores of AAC obtained by DXA were independently associated with MI and stroke, even after adjustment for traditional cardiovascular risk factors.9 Our results have confirmed and extended these findings, showing that in women at low risk of cardiovascular events the presence of AAC or increases in AAC over time predict cardiovascular events, even after traditional risk factors are taken into account.

The use of AAC-8 score as a screening tool is not appropriate because it has low sensitivity and specificity for MI and cardiovascular events and has less predictive value than traditional cardiovascular risk factors integrated into the estimated 5 year cardiovascular risk score described by Gaziano and colleagues.15 However, at an individual level, the presence of AAC approximately doubled the future risk of MI, whereas the absence of AAC had a very high negative predictive value. Thus it seems reasonable to look for AAC when assessing vertebral morphometric images obtained from DXA because the AAC-8 score is quick, easy to perform, and reliable. Bone densitometry as part of an assessment for osteoporosis is indicated in all women over 65 years of age.21 Vertebral morphometry is commonly performed as part of this assessment because it is quick, has a low radiation dose, and is able to detect previously unrecognized vertebral fractures in men and women.22–24 Importantly, such vertebral fractures are associated with increased risk of future clinical fractures.25 In addition, radiographic vertebral fractures, the majority of which are unrecognized, are associated with increased mortality.26 As a result of these findings, the National Osteoporosis Foundation recommends that anyone with a radiographic vertebral fracture be treated for osteoporosis.27 Thus it is likely that vertebral morphometric assessment using DXA will become more widespread.

Our study has a number of limitations. Both trials were single-center studies, with people of European origin comprising more than 90% of the cohorts, and were restricted to older postmenopausal women and middle-aged and older men. In addition, the studies recruited healthy volunteers with low cardiovascular risk; thus the findings may not be generalizable to other ethnic or age groups or to those with poorer health status or higher cardiovascular risk. Furthermore, the men's trial was small, with few cardiovascular events during follow-up, which limits the ability to draw conclusions. Reproducing our findings in larger cohorts with more cardiovascular events therefore is important. Thirty percent of the women and 7% of the men did not have a follow-up assessment of AAC, which is a further limitation of the study.

In conclusion, this study has shown that assessment of AAC using the AAC-8 score is a reproducible and reliable technique for assessing cardiovascular risk that independently predicts incidence of MI and cardiovascular events. AAC should be looked for on vertebral morphometric images and the presence or absence of AAC conveyed to the clinician requesting the test. Where AAC is present, it should be indicated that this is associated with an increased risk of future cardiovascular events, even after taking into account traditional cardiovascular risk factors.

Acknowledgments

This study was funded by grants from the Health Research Council of New Zealand and from Osteoporosis New Zealand. TW received a scholarship from the National Heart Foundation of New Zealand for this work.

Disclosures

The authors state that they have no conflicts of interest.

Ancillary