Introduction

The American Heart Association (AHA) Strategic Impact 2020 Goal provides guidelines and metrics to measure both the current and longitudinal status of cardiovascular health (CVH) with the goal of improving CVH for the US population.¹ In contemporary community-based cohorts in the US population, there is a low prevalence (≤5%) of ideal CVH.² Further, poor CVH is associated with significantly elevated morbidity and mortality rates from cardiovascular diseases (CVD).^3,4 Studies demonstrating an association between maintenance of ideal CVH and the prevention of CVD from community-based population, while limited, had been reported.⁵ However, it remains unknown whether a change in CVH profile or maintenance of ideal CVH provides incremental benefit by preventing progression of subclinical atherosclerosis. We sought to characterize these potential associations that would facilitate the development of novel approaches for precision medicine to primary prevention of CVD in lower risk adults.

See Editorial by Yeboah

See Clinical Perspective

Coronary artery calcium (CAC) is a well-recognized biomarker of the presence and burden of subclinical coronary atherosclerosis and of the progression of coronary atherosclerosis.⁶ Population-based studies provide compelling evidence for significant associations between an elevated CAC score and the risk of future CVD, independent of traditional CVD risk factors.^7,8 Conversely, persons free of CAC carry a substantially lower risk of future CVD.⁹ In asymptomatic middle-aged adults with zero CAC score, however, CHD risk varies considerably across diverse populations.^10,11 A significant association between CAC progression and an unfavorable CVH profile has been reported for patients with type I diabetes mellitus such that patients with an unfavorable CVH profile have a 1.5-fold higher odds of CAC progression.¹² In participants of the FHS (Framingham Heart Study) at low baseline CVD risk, we tested whether a decline in CVH confers significant, additional value for the prediction of CAC progression.

Methods

All data, methods used in the analysis, and materials will not be made directly available by the authors for purposes of reproducing the results. However, researchers are encouraged to request all data and materials through the Framingham Heart Study data request process, for which detailed information is available at www.framinghamheartstudy.org.

Study Population

The study population comprised FHS Offspring and Third-Generation participants. The FHS is a community-based prospective study designed to investigate the incidence of CVD and related risk factors.¹³ The study design and participant characteristics of the offspring and Third-Generation cohorts have been published.^14,15 Participants who attended the Third-Generation examination 1 (from 2002 to 2005) and Offspring examination 7 (1998–2001) and who also participated in the MDCT (Multi-Detector Computed Tomography) substudy I (2002–2005) comprised the baseline study population. For the MDCT studies, we excluded participants with the following conditions: weighed ≥160 kg, pregnant, men <35 years of age, and women <40 years of age.⁸ For the current study, participants of the MDCT I study who were unable to attend the MDCT II study (second MDCT, 2008–2011) were also excluded. After these exclusions, the final study population consisted of 1119 participants with a zero Agatston CAC score at baseline. The study protocol was approved by the Institutional Review Board of the Boston University Medical Center, and all participants gave written informed consent.

Determination of CVH

To characterize ideal CVH status, we applied 6 AHA guideline metrics (total cholesterol, fasting glucose, blood pressure (BP), body mass index, cigarette smoking, and exercise) and the cut point definition for ideal, intermediate, or poor status for each CVH metric.¹ Information on these 6 metrics was obtained through routine FHS examination and laboratory measurements. At each examination cycle, a physician conducted an interview and physical examination of each participant and collected information on medical history and medication use. Blood samples were collected at each examination cycle for serum cholesterol and fasting blood glucose. The ideal threshold for total cholesterol was <200 mg/dL without cholesterol-lowering medication use. Cut point levels for intermediate and poor CVH for total cholesterol were 200 to 240 and 240, respectively. Cholesterol-lowering medication user those had total cholesterol <240 mg/dL were grouped into the group of intermediate CVH. For fasting blood glucose, the ideal threshold was <100 mg/dL without diabetic medication use. Intermediate CVH group included medication user for diabetes mellitus or participants those had fasting blood glucose of 100 to 126 mg/dL. Participants with fasting blood sugar of 126 mg/dL or higher had poor CVH. Diabetes mellitus was defined by fasting blood glucose >126 mg/dL or use of diabetic medication. Systolic and diastolic BPs were generated using mean value of the physician measurements. Ideal BP was defined as systolic BP <120 and diastolic BP <80 mm Hg without antihypertension medication use. Cut points for poor BP metrics were 140 mm Hg for systolic BP or 90 mm Hg for diastolic BP. Body mass index was calculated as weight (kg) divided by the square of height (m²), and an ideal body mass index was <25 kg/m². Cut points for intermediate and poor CVH metric were 25 and 30, respectively. Current cigarette smoking was defined as smoking ≥1 cigarettes per day in the year preceding the examination. We defined former smokers as participants who smoked at any previous examination cycle but not the current examination cycle. An ideal CVH status for cigarette smoking was defined as never having smoked. For the current study, we defined intermediate CVH status for former smoker that ideal CVH status included never smokers. Information on physical activity was collected at the examination cycle on the basis of typical hours per day for 5 activity categories, sleep, sedentary, slight activity, moderate activity, and heavy activity.¹⁶ A physical activity score was generated as the weighted sum of hours of the 5 activities with a weight factor of 1, 1.1, 1.5, 2.4, and 5, respectively, that was correlated with oxygen consumption of the activity. For the current study, we defined those with poor physical activity as those with an activity index score of ≤27.9. An intermediate status for physical activity was defined by a physical activity score between 28.0 and 36.9. The ideal CVH category for physical activity was defined by a physical activity score of ≥37.

Information on coronary heart disease (myocardial infarction and CHD death), with follow-up up to December 31, 2013, was abstracted to identify its prevalence and examine the potential associations of atherosclerotic CVD (asCVD) with CAC progression. Using prespecified criteria, all available clinical information on each CHD diagnosis, including medical records and participant-reported information, was reviewed by the 3-physician end point panel to verify a CHD diagnosis.¹⁷ Information on cholesterol-lowering medication use was collected through on-site interview followed with prescription/label verification and standardized coding using the World Health Organization Anatomic Therapeutic Chemical classifications.

For each CVH metric, we defined ideal, intermediate, and poor categories on the basis of the guideline recommendations for the AHA 2020 Strategic Goals. We stratified participants into 3 groups according to the number of ideal metrics: favorable (5+ ideal metrics), intermediate (3 or 4 metrics), and unfavorable (≤2 metrics). We conducted analyses of the impact of unfavorable CVH on CAC progression using either a categorical CVH profile or a semiquantitative continuous numeric CVH status based on the number of ideal CVH metrics. For the categorical CVH profile, we generated binary variables referring to unfavorable and intermediate CVH profiles. For changes in CVH profiles, 1 binary variable was defined by failure to maintain a favorable status when the starting point was a baseline favorable CVH profile. A second binary variable was defined, for participants with either an unfavorable or intermediate profile at baseline, by failure to gain ≥2 ideal CVH metrics. For the semiquantitative numeric CVH profile, we calculated the change in number of ideal CVH metrics by subtracting the number of ideal CVH metrics in follow-up compared with the number of ideal CVH metrics at baseline. For the current study, our primary analysis tested for significant association between CAC progression and numeric CVH status. Our secondary analysis tested for associations between CAC progression and categorical CVH (results are shown in the Tables in the Data Supplement).

Measurement of Coronary Artery Calcium

The extent of coronary artery calcium (CAC) was quantified in both the baseline and follow-up FHS MDCT studies. The comprehensive study protocol and reproducibility for FHS MDCT I have been previously reported.⁸ In brief, scans were conducted using an 8-slice MDCT scanner (LightSpeed Ultra; General Electric, Milwaukee, WI) with prospective ECG triggering during a single breath hold in midinspiration (≈18 s) using sequential data acquisition. A calcified lesion was defined as an area of ≥3 connected pixels with a computed tomographic (CT) attenuation >130 Hounsfield units. A modified Agatston score, calculated as the sum of lesions from each individual cross section, was generated.

Cut points for CAC progression were defined by estimating interindividual difference with multiple scoring results for CAC. The estimates of agreement yielded a proxy for measurement noise.¹⁸ At the baseline CT scan, a second CAC scan was obtained after brief repositioning of each participant in the scanner, providing duplicate CAC Agatston scoring for 3390 participants. A total of 2803 participants had one or both CAC Agatston scores of zero. We evaluated statistical distributions for participants with at least 1 zero CAC score and observed an SD of 3.4 for those with a zero CAC score. We applied this estimate of variation to define a threshold for CAC progression for the 1119 participants who attended both the MDCT I and MDCT II examinations. In a secondary analysis, we applied a more conservative threshold Agatston score of 11.6 to define CAC progression, to account for the magnitude of change not explained by measurement alone.¹⁹

Statistical Methods

Demographic characteristics and CVD-related risk factors were provided as mean±SD for continuous variables and number and percent of participants for dichotomous or categorical variables. To account for within nuclear family correlation structure, we used multivariate generalized estimating equations (GEE) logistic regression to test for significant associations between CAC progression (outcome) and unfavorable CVH profile (independent variable of interest), assuming an exchangeable (compound symmetry) correlation structure. We generated 3 sequential multivariate models to characterize the significant impact of CVH profile on prediction of CAC progression. In model I, the covariates were age, sex, HDL (high-density lipoprotein) cholesterol, and triglycerides. In model II, we included the covariates in model I plus the number of ideal CVH metrics at baseline. In model III, we included covariates in model II, and we further adjusted for decreasing number of ideal CVH metrics. We also performed an additional adjustment for cholesterol-lowering medication statin use to further verify the significant association between nonideal CVH and CAC progression (model IIb and model IIIb).²⁰ In addition to individual parameter estimation, we evaluated information of model fit to characterize potential predictors of CAC progression. Quasi-likelihood criteria for the GEE model were applied to identify the best-fitting model, and the rule of model selection was similar to that for the Akaike information criteria.²¹ We generated the predictive probability of CAC progression using the best-fitting GEE equation and calculated CAC progression risk by number of nonideal CVD metrics.

We calculated participants’ asCVD risk using the Pooled Cohort Equation, and we characterized significant associations between change in asCVD risk and CAC progression using the GEE model.²² A 2-sided P value of <0.05 was considered statistically significant. All analyses were performed using SAS version 9.3.

Results

The mean age of the study population at baseline was 48.4 years, and the population was 58% female. At baseline, 278 (25%) had a favorable CVH profile (5+ ideal metrics), 576 (51%) had an intermediate profile, and 265 (24%) participants had an unfavorable profile (<3 ideal CVH metrics). The mean number of ideal CVH metrics was 3.5 and was similar in both men and women. Table 1 shows distributions of baseline characteristics for the 1119 FHS participants free of baseline CAC. After an average of 6.1 years of follow-up, we observed CAC progression defined by the occurrence of newly detected CAC of ≥3.4 Agatston score in 191 participants (17.1%). Among these 191 participants, the mean Agatston score change values, grouped by tertiles of Agatston score change, were 7, 21, and 63, respectively. Characteristics of ideal CVH among the 3 subgroups by increase of Agatston score were not significantly different. Table I in the Data Supplement shows the distributions of Agatston scores for the 191 participants with CAC progression.

In contrast to the 928 participants free of progression, participants with CAC progression possess a higher CVD risk at baseline, with a higher frequency of male sex, higher body mass index, greater level of triglycerides, lower levels of HDL cholesterol, and higher frequencies of cigarette smoking, diabetes mellitus, antihypertensive medication use, and lipid-lowering medication use. In the GEE model adjusting for age, sex, HDL cholesterol, triglycerides, and familial correlation, there were a 2.43-fold (95% confidence interval, 1.40–4.23; P=0.0017) higher odds of CAC progression in participants with an unfavorable CVH profile (Table II in the Data Supplement, model II). In regression analysis testing for significant association between number of nonideal baseline CVH metrics and CAC progression, there were a 1.24-fold (95% confidence interval, 1.08–1.42; P=0.0018) higher odds of CAC progression per increase of 1 nonideal metric (Table 2, model II). Additional adjustment for statin use did not change the significant association (Table 2, model IIb).

Failure to gain ideal CVH metrics did not influence CAC progression after adjustment for baseline CVH profile. At follow-up, the prevalence rates for favorable, intermediate, and unfavorable CVH profiles were 15%, 45%, and 40%, respectively. In multivariable-adjusted models, there was a nonsignificant association of failure to maintain favorable CVH profile with incidence of CAC progression (Table 2, model III). We estimated the probability of developing CAC progression by number of nonideal metrics at baseline (model II). There was a marked increase in the proportion with CAC progression as the number of nonideal CVH metrics increased (Figure 1). Participants with 6 nonideal CVH metrics at baseline had a 41% risk of CAC progression; in contrast, there was a 5% risk of progression for participants with 6 ideal metrics at baseline.

• Download figure
• Download PowerPoint

In secondary analysis to examine the occurrence of clinical CVD events, there is a very low event risk in our study population. The mortality rate was zero over the follow-up period after MDCT II. Four participants developed myocardial infarction, including 2 participants who had evidence of CAC progression and 2 who were free of CAC progression.

In sensitivity analyses, using a more conservative threshold for CAC progression (Agatston score ≥11.6; n=125; 11.1% CAC progression), we observed a significant association only for a higher number of nonideal CVH metrics. The association between baseline unfavorable CVH profile and CAC progression was not significant. For each unit increase in nonideal CVH metric, there were a 1.17-fold higher odds of CAC progression (Table 3, model II). In the overall model fit, however, an increase in nonideal CVH metrics over time did not confer additional predictive potential for CAC progression.

Applying the American College of Cardiology/AHA recommended pooled cohort equation for status of CVH, we calculated the 10-year risk of the first asCVD and observed similar asCVD risk for the 1119 participants at the baseline and follow-up CT scans.²² For both time periods, the 191 participants with CAC progression had a significantly higher risk of asCVD (mean±SD, 10.4% ± 12.48) in contrast to that for the 928 participants free of progression (mean±SD, 4.8±6.83). Results of the GEE models adjusted for familial correlation showed significant association between 10-year asCVD risks and CAC progression (Table 4). When we grouped participants into high versus low asCVD risk groups using a cut point of 7.5% at baseline, we observed a nonsignificant association between high asCVD risk and CAC progression. Change in 10-year asCVD risk was not associated with CAC progression (Table III in the Data Supplement). Results of the separate GEE models suggested that higher asCVD risk score or number of nonideal CVH metrics predicted CAC progression (Table 4). Further, there was a stronger association between number of nonideal CVH metrics and CAC progression in contrast to the association between baseline 10-year risk of asCVD and CAC progression (Figure 2).

• Download figure
• Download PowerPoint

Discussion

In this community-based study of middle-aged men and women at low cardiovascular risk by virtue of having a CAC Agatston score of zero at baseline, a quarter of the population had an unfavorable CVH profile. In the follow-up period for this population, we found evidence for significant progression to CAC in 17% and a substantial increase in both intermediate and unfavorable CVH profile over 6.1 years, correlating with a concomitant decrease in favorable CVH profile. There was a significant increase in progression to CAC as a function of increases in nonideal CVH at baseline. Results of the analysis for both categorical and numeric indices of CVH did not demonstrate an incremental influence of change in CVH status over time on prediction of CAC progression. By providing quantitative estimates of the risks for CAC progression conferred by increasing numbers of nonideal CVH factors, the current study extends and adds to prior evidence that the maintenance of ideal CVH may be beneficial for primordial prevention.²³

Most persons are free from CAC in early life, but the factors predicting continued freedom from CAC have not been comprehensively characterized. Patsalas et al²⁴ reported a high percentage of hemodialysis patients (13 of 17 [76%]) maintaining zero CAC over 30 months of follow-up. A similar rate of maintenance of zero CAC was reported for a clinical study of 422 patients referred for CAC scanning.²⁵ A significantly lower frequency for maintaining zero CAC (33%) was reported by Alagoz et al²⁶ in 75 kidney donors. In the current study in a community-based population, we observed that 80% of participants maintained a zero Agatston score (901/1119), not markedly different from prior findings in disease-based studies. Results of our study are consistent with other studies that describe variations in CVD and related health well-being varies within an otherwise low-risk middle-aged population free of CAC. As previously reported, 1 common predictor for both the presence of CAC at baseline and the occurrence of new CAC over time (progression) is older age.²⁷ Further studies in larger cohorts may identify additional specific health factors at baseline and changes in health factors over time that prevent progression to subclinical coronary atherosclerosis.

There are challenges to interpretation of CAC progression studies using repeated CAC measurements. Prior reports have noted the crucial role of interscan variability and the time-dependent changes of CAC²⁸ and the lack of universal agreement for defining CAC progression.²⁹ Results of our more stringent sensitivity tests confirmed our significant findings that variation in a numeric CVH profile significantly predicts CAC progression, in both the overall study population and the subset of nonstatin users. In the relatively short follow-up time after MDCT II, we noted very few CVD events, however, we observed significant variation in baseline CVH associated with CAC progression. Recent attention to precision medicine has highlighted a potential opportunity to focus on precision prevention and primordial prevention in younger persons free of clinical or subclinical disease.^30,31 Our data help lay the groundwork for future studies to identify more precise CVH factors, including biomarkers, genomic markers, and imaging measures for CVD prevention.

Strengths of the current study include our detailed characterization of CVH profile in a well-described community-based study population with repeated clinical examinations over time. We focused on participants with zero CAC Agatston score, yielding a study population with a relatively homogeneous CVD and minimal medication use. In addition, information on changes in CVD risk factors and CVD disease status was available for all participants that showed a very low CVD risk and zero CVD-related mortality for this group of participants with minimum influence by comorbidity or medication treatment.

Nevertheless, this study had several limitations. Although we had robust data for 6 ideal CVH metrics, we lacked detailed information on dietary metrics, so we were unable to calculate the AHA ideal diet metric. However, in contrast to the other CVH metrics, the ideal diet metric status has a very low prevalence rate in the general population, so it is likely that the large majority of our participants fell in the nonideal diet category.³² The study population was primarily of European ancestry, so additional studies in multiethnic cohorts would be needed to assess the generalizability of these associations and the rate of CAC progression in other racial/ethnic populations. We also acknowledge the possibility that results may differ using other definitions of CAC progression. Multiple CT scanning and scorings for duplicate CAC were available at the baseline CT scan only. The lack of multiple scoring at the follow-up CT scan hampered the evaluation of interstudy variation. The limitation to maximize homogeneity of the study population by recruiting participants those with zero CAC includes low statistical power. Cut point for defining CAC progression was derived from a single study center that should had been verified in replication cohorts. In addition, we applied 1 method of defining progression that study findings may alter if other methods were applied. In addition, the length of follow-up (mean, 6.1 years) hamper association test for maintaining ideal CVH and incidence of CAC progression.

In summary, the additive loss of ideal CVH factors is associated with CAC progression in low-risk middle-aged men and women with zero CAC at baseline. Significant associations between an unfavorable CVH profile and CAC progression provide supportive evidence for public health measures to prevent CVD by promoting favorable CVH profiles in persons free of both clinical and subclinical CVD.

Footnotes