Long-term Blood Pressure Variability in Young Adulthood and Coronary Artery Calcium and Carotid Intima-Media Thickness in Mid-Life: The CARDIA Study

Abstract

Recent evidence links long-term (visit-to-visit) blood pressure (BP) variability to the risk of cardiovascular disease, independent mean BP levels. Potential associations between long-term BP variability and cardiovascular disease risk may be reflected in early life course alterations in coronary artery calcium (CAC) and carotid intima-media thickness.

We evaluated 2482 Coronary Artery Risk Development in Young Adults study participants (mean [SD] age at the year-20 exam [2005–2006] was 45.4 [3.6] years, 43.2% men, and 41.3% African-American). We included participants with BP assessments across 20-years (year 0, 2, 5, 7, 10, 15, 20 exams) and carotid intima-media thickness and CAC data at the year-20 exam. BP variability was assessed using variability independent of the mean and SD). Adjusted multivariable linear or logistic regression models (as appropriate) were used to assess associations between long-term BP variability measures and carotid intima-media thickness. and CAC (ln [CAC+1] and prevalent CAC). Long-term systolic BP variability independent of the mean (per 1-SD) was positively associated with carotid intima-media thickness.

(β=10 μm, standard error (SE)=3, p=0.002. Similarly, long-term diastolic BP variability independent of the mean was associated with carotid intima-media thickness (β=10μm, SE (3), p=0.001). Long-term BP variability was not associated with either ln [CAC+1] or prevalent CAC.

Long-term variability independent of the mean systolic and diastolic across early adulthood was positively associated with modest adverse mid-life alterations in carotid intima-media thickness but not to CAC. Our findings provide further insights into pathophysiologic mechanisms that link long-term BP variability to cardiovascular disease.

Graphical Abstract

INTRODUCTION

High blood pressure (BP) is a leading cause of cardiovascular disease globally and in the United States.¹ Given the high burden of hypertension, additional efforts to understand the pathobiological mechanisms that link BP to cardiovascular disease remain important.

Current guidelines support the use of the mean of multiple BP recordings across separate visits to identify and treat hypertension. One aspect that has garnered interest in recent times is the tendency of BP to vary across visits and whether this visit-to-visit BP variability is independently associated with unfavorable cardiovascular outcomes. Growing evidence now suggest that long-term (visit-to-visit) BP variability is associated with the risk of cardiovascular disease, independent of mean BP levels.^2–4 The mechanisms that underlie a relationship between BP variability and cardiovascular disease are however incompletely studied. More so in early adulthood, where early life cardiac risk factors can increase the risk of cardiovascular disease, independent of the risk profile in late adulthood.^{5, 6} Oscillations in BP may influence cardiovascular disease early in its life course and may therefore be reflected in alterations in subclinical markers of atherosclerosis such as coronary artery calcium and carotid intima-media thickness, that are established as strong predictors of cardiovascular disease.

Few studies carried out in predominantly elderly populations have examined the association between BP variability and markers of atherosclerosis and their findings have been inconsistent.^7–10 Knowledge of how long-term BP variability from young adulthood is related to coronary artery calcium or carotid intima-media thickness by middle age, may provide further insights into underlying pathophysiologic pathways between BP variability and the risk of cardiovascular disease. We therefore investigated the association between long-term BP variability in young adults and coronary artery calcium and carotid intima-media thickness in mid-life.

METHODS

Study Design and Participants

All data used for this study have been made publicly available and can be requested at https://biolincc.nhlbi.nih.gov. All participants provided written informed consent at each examination, and institutional review board of each field center and the coordinating center approved the study annually.

The Coronary Artery Risk Development in Young Adults (CARDIA) study design and selection criteria have been described in prior reports.¹¹ Briefly, the CARDIA study is a multi-center community-based study that recruited 5115 men and women Caucasian and African-American participants (Birmingham, Alabama; Oakland, California; Chicago, Illinois; and Minneapolis, Minnesota) who were aged 18–30 years at baseline. Follow-up examinations were conducted at the following exams: year-2, 5, 7, 10, 15, 20, 25, and 30 with response rates of 91%, 86%, 81%, 79%, 74%, 72%, 72% and 71%. Out of the 3549 CARDIA participants who attended the year-20 exam, our analytic dataset included persons with BP assessments across 20-years (CARDIA year 0, 2, 5, 7, 10, 15, 20 exams) and carotid ultrasound and computed tomography data at the year-20 exam. This resulted in an analytic sample of 2482 participants. In addition, we carried out sensitivity analysis in a subset of participants (n=1851) with low global burden of atherosclerosis defined as a coronary calcium score of zero.

Long-term Blood Pressure Variability

BP was ascertained using standardized protocols across all study examinations and field centers. Prior to each assessment, participants were asked to fast and refrain from heavy physical activity or smoking activity. After 5 minutes of rest, trained staff obtained 3 readings from the brachial artery. The average of the 2^nd and 3^rd BP readings were used. In the year-0 to year-15 exam, a random-zero sphygmomanometer was used. However, in the year-20 exam because of concern about mercury contained in the BP apparatus, automated oscillometric BP monitors were used. A BP calibration study was carried out to account for differences between random-zero sphygmomanometer and automated oscillometric BP monitors.¹² Detailed description of the CARDIA BP measurement protocol and calibration study are included in supplement (eMethods). Long-term BP variability was calculated across 7 visits (baseline and, 2, 5, 7, 10, 15, and 20 years following baseline) for each participant. For SBP and DBP, visit-to-visit BP variability measurements included BP variability independent of the mean and standard deviation (SD). Variability independent of the mean was defined as the intra-individual standard deviation of BP over visits times ${(\frac{M}{x})}^p$ where x is individual average BP over visits, M is the average of individual average SBP in the overall population, and p is the regression coefficient on the basis of regressing the natural logarithm of SD on the natural logarithm of the multiplication of x and M. The SD of SBP was calculated as:

$$Standard deviation of SBP=\sqrt{\frac{{\sum }_{i=1}^7{\left(SB{P}_i-\overline{SBP}\right)}^2}{6}}$$

SD is correlated with average BP; thus, distinguishing the effect of BP variability from that of BP level change on outcomes may be difficult. Conversely, variability independent of the mean has been shown to have no correlation with average BP levels but a strong correlation with SD.^{13, 14} Therefore, in the current study, we used both variability independent of the mean and SD as a measure of visit-to-visit BP variability.

Covariates

Covariates were obtained with a combination of standardized questionnaires, clinical, and laboratory data. A modified version of the Minnesota leisure time physical activity questionnaire was used to quantify physical activity in exercise units. Body mass index was calculated from weight and height. Fasting samples were collected using standard protocols.¹⁵ Cumulative risk exposure from year-0 to year-20 for continuous variables was calculated by calculating the means of values from each pair of consecutive examinations and multiplying each mean value by the number of years between the consecutive visits. Time-varying anti-hypertensive and cholesterol lowering medication use was obtained from year-0 to year-20 and smoking status (current smoker, former smoker, and never smoker) was self-reported.

Cardiac Computed Tomography

Electron beam cardiac computed tomography (CT) was performed at the year-20 exam using standardized protocols to assess coronary calcium.¹⁶ Total coronary artery calcium was obtained by summing all lesions within a given artery and across all arteries (left main, left anterior descending, left circumflex, and right coronary artery). CT images were transmitted to a central reading center where coronary artery calcium was quantified in 2.5 to 3.0-mm-thick images by experienced technicians and a cardiovascular radiologist. Because coronary artery calcium score distribution was positively skewed, we log-transformed the coronary artery calcium score. The inter observer and intra-observer κ statistic was 0.89 and 0.95, respectively.¹⁶

Carotid Ultrasonography

Details of the carotid ultrasound scanning protocol at the CARDIA year-20 exam have been described previously.¹⁷ Carotid ultrasound images were recorded on videotape and transmitted to an Ultrasound Reading Center (Tufts University) for intima-media thickness scoring. Images of the common carotid artery intima-media thickness, bulb- intima-media thickness, and proximal internal carotid artery intima-media thickness were obtained using high-resolution B-mode ultrasonography (Logiq 700; General Electric Co., UK).¹⁷ Intima-media thickness was measured from the lumen-intima to the media-adventia interfaces. Carotid intima-media thickness was calculated from the average of the mean intima-media thickness for the internal, bulb, and common carotid near and far walls of the right and left sides. Correlation coefficients were 0.86 for common carotid artery intima-media thickness, 0.72 for the bulb intima-media thickness, and 0.88 for the internal carotid artery intima-media thickness.¹⁷

Statistical Analysis

Continuous variables were presented as mean (SD). Categorical variables were reported as numbers and percentages. We used multivariable linear or logistic regression models as appropriate to assess the association between long-term systolic and diastolic BP variability measures (SD and variability independent of the mean) and carotid intima-media thickness and coronary artery calcium as dependent variables. Log transformed (coronary artery calcium) + 1 was assessed in linear regression models with beta coefficients whereas logistic regression models were used to calculate odds ratios for prevalent coronary artery calcium as it was a binary variable (Yes/No CAC). Models were adjusted for age, sex, race, cumulative exposure to risk factors from year-0 to year-20 for continuous covariates (BP, body mass index, ratio of total cholesterol to high density lipoprotein cholesterol, and physical activity), proportion of study participants exposed for categorical variables (smoking and time-varying antihypertensive and cholesterol medication use (year-0 to year-20). In a sensitivity analysis, we examined associations for carotid intima-media thickness in a subset of participants who had a coronary artery calcium score of zero. Linear regression assumptions were not violated and a variance inflation factor of <2.5 was used to test for collinearity. All analysis was performed using Stata version 15.1 and a two-tailed p-value of <0.05 was considered statistically significant.

RESULTS

Study Characteristics

Baseline characteristics for participants in the year-20 exam are shown in Table 1. Mean (SD) age at the year-20 exam was 45.4 (3.6) years, 43.2% were men, and 41.3% were African-American. Mean body mass index and systolic BP was 29.4 (7.2) kg/m² and 115.1 (14.3) mmHg, respectively. The proportion of current smokers was 16.5%. The average SD of long-term systolic BP and diastolic BP was 7.69 mmHg and 6.92 mmHg, respectively. Average variability independent of the mean for systolic and diastolic BP was 7.59 U and 6.92 U, respectively.

Table 1.

Characteristics at the CARDIA Year-20 exam

Characteristics	CARDIA Year-20 Exam (n=2482)
Age, years	45.4 ± 3.6
Men, %	1071 (43.2)
African-Americans, %	1025 (41.3)
Body mass index, Kg/m2	29.4 (7.22)
Anti-hypertensive medications, %	402 (16.2)
Current smoking, %	406 (16.5)
Physical Activity, eu	343.1 (270.9)
Fasting glucose, mg/dL	97.7 (24.7)
Total cholesterol, mg/dL	186.3 ± 35.0
Blood Pressure Parameters
Systolic BP, mmHg	115.1 ± 14.3
Diastolic BP, mmHg	71.7 ± 10.9
Long-term systolic BP SD, mmHg	7.69 ± 3.8
Long-term diastolic BP SD, mmHg	6.92 ± 2.9
Long-term systolic BP VIM, U	7.59 ± 3.2
Long-term diastolic BP VIM, U	6.92 ± 2.8
Outcomes
Prevalent CAC	408 (18.1)
Carotid intima-media thickness, μm	890 ± 170

Data are expressed as the means ± standard deviations or frequency and percentages.; BP means blood pressure; Eu, exercise units; SD, standard deviation; VIM, variance independent of the mean.

Association between Long-term Blood Pressure Variability and coronary artery calcium and carotid intima-media thickness

As shown Table 2, greater long-term systolic BP variability indicators (variability independent of the mean and SD) across early adulthood was positively associated with carotid intima-media thickness but not with either log (coronary artery calcium) +1 or prevalent coronary artery calcium. Similarly, greater long-term diastolic BP variability indicators (variability independent of the mean and SD) across early adulthood was positively associated with carotid intima-media thickness but not with either log (coronary artery calcium) +1 or prevalent coronary artery calcium. As shown in Table 3, among participants with (coronary artery calcium)=0 (n=1851), associations between long-term systolic and diastolic BP variability and carotid intima-media thickness remained significant. There were no race or gender differences in the association between BP variability and coronary artery calcium or carotid intima-media thickness (p>0.1 for interaction).

Table 2.

Association between Long-term Systolic Blood Pressure Variability and Coronary Artery Calcium and Carotid Intima-Media Thickness

Outcomes	Systolic Blood Pressure Variability Indicators				Diastolic Blood Pressure Variability Indicators
	VIM	VIM	SD	SD	VIM	VIM	SD	SD
	Unadjusted	Adjusted	Unadjusted	Adjusted	Unadjusted	Adjusted	Unadjusted	Adjusted
Ln (CAC +1)	0.03 (0.08) p=0.67	−0.003 (0.09) p=0.97	0.17 (0.07) p=0.02	0.01 (0.09) p=0.95	0.09 (0.09) p=0.25	0.09 (0.09) p=0.29	0.19 (0.08) p=0.02	0.10 (0.09) p=0.27
Prevalent CAC	0.98 (0.88–1.09)	0.93 (0.81–1.06)	1.19 (1.08–1.31)	0.92 (0.79–1.06)	1.04 (0.94–1.16)	1.02 (0.89–1.17)	1.15 (1.04–1.27)	1.02 (0.89–1.17)
CIMT, μm	10 (3) p<0.001	10 (3) p=0.002	30 (3) p<0.001	10 (3) p=0.001	10 (3) p<0.001	10 (3) p=0.001	20 (3) p<0.001	10 (3) p<0.001

Multivariable linear or logistic regression models showing the association between long-term systolic and diastolic blood pressure variability and coronary artery calcium and carotid intima-media thickness. Coefficients and standard errors are presented for continuous variables and odds ratios with 95% confidence interval are presented for categorical variables. Models were adjusted for age, sex, race, cumulative exposure to risk factors (year-0 to year-20) for continuous covariates (BP, body mass index, ratio of total cholesterol to high density lipoprotein cholesterol, physical activity), proportion of study participants exposed for categorical variables (smoking (never, former, or current), time varying anti-hypertensive and cholesterol medication use (year-0 to year-20). VIM, means variability independent of the mean; SD, standard deviation; CAC, coronary artery calcium; CIMT, carotid intima-media thickness. CIMT was calculated from the average of the mean IMT of the internal, bulb and common carotid near and far walls of the right and left sides. Because CAC score distribution is positively skewed, we used log-transformed CAC+1. Beta coefficients are per 1 SD increments. For systolic BP SD and VIM, per SD is 4.26 mmHg and 3.46 U, respectively. For diastolic BP SD and VIM, per SD is 2.97 mmHg and 2.83 U, respectively. i.e. beta coefficients for SD of SBP represent each 4.26 mmHg higher level of SD of systolic BP during young adulthood was associated with an 10 μm (standard error 3 μm) increment of common carotid intima-media thickness in midlife.

Table 3.

Association between Long-term Systolic Blood Pressure Variability and Coronary Artery Calcium and Carotid Intima-Media Thickness among 1851 participants with a coronary calcium score of zero

Outcomes	SBP Variability Indicators				DBP Variability Indicators
	Unadjusted	Adjusted	Unadjusted	Adjusted	Unadjusted	Adjusted	Unadjusted	Adjusted
	VIM	VIM	SD	SD	VIM	VIM	SD	SD
CIMT, μm	10 (3) p<0.001	10 (3) p=0.02	10 (3) p<0.001	10 (4) p=0.01	10 (3) p<0.001	10 (3) p=0.004	10 (3) p<0.001	10 (3) p=0.006

Multivariable linear regression models showing the association between long-term systolic and diastolic blood pressure variability and carotid intima-media thickness in participants with CAC=0. Coefficients and standard errors are presented. Models were adjusted for age, sex, race, cumulative exposure to risk factors (year-0 to year-20) for continuous covariates (BP, body mass index, ratio of total cholesterol to high density lipoprotein cholesterol, physical activity), proportion of study participants exposed for categorical variables (smoking (never, former, or current), time varying anti-hypertensive and cholesterol medication use (year-0 to year-20). VIM, means variability independent of the mean; SD, standard deviation; CIMT, carotid intima-media thickness. CIMT was calculated from the average of the mean IMT of the internal, bulb and common carotid near and far walls of the right and left sides.

Beta coefficients are per 1 SD increments. For systolic BP SD and VIM, per SD is 3.74 mmHg and 3.15 U, respectively. For diastolic BP SD and VIM, per SD is 2.97 mmHg and 2.85 U, respectively. i.e. beta coefficients for SD of SBP represent each 4.26 mm Hg higher level of SD of systolic BP during young adulthood was associated with an 10 μm (standard error 4) increment of common carotid intima-media thickness in midlife.

DISCUSSION

Our central finding was that long-term systolic and diastolic BP variability in early adulthood was positively associated with mid-life alterations in carotid intima-media thickness but not to coronary artery calcium. Our study observations of positive associations between long-term systolic and diastolic BP variability and carotid intima-media thickness remained significant, even among a subset of participants with a negligible burden of subclinical atherosclerosis (coronary artery calcium=0). Our study results suggest that the association between long-term BP variability and cardiovascular disease may be, at least in part, influenced by differential associations with markers of subclinical atherosclerosis.

Differential associations between BP variability, coronary artery calcium, and carotid intima-media thickness have been noted in prior studies carried out in predominantly middle age-to-elderly populations. In one previous study of 322 participants, Debarmore et al showed that there were no differences in BP variability among individuals in different categories of coronary artery calcium.¹⁰ Along the same line, in a study 1033 participants from Japan,⁹ there was a lack of association between BP variability and coronary artery calcium. In this same Japanese study, positive associations between BP variability and carotid intima-media thickness were also observed, despite a lack thereof with coronary artery calcium.⁹ In a separate study, there was no association between BP variability and atherosclerotic plaque identified by carotid ultrasound.¹⁸ Collectively, these findings from smaller studies are in agreement with our observations of a lack of association with coronary artery calcium, but our results which additionally account for early life cumulative risk exposure, importantly extend them to young adulthood.

In contrast to coronary artery calcium, and also consistent with our findings, studies in predominantly elderly populations have provided strong evidence of positive associations between BP variability and carotid intima-media thickness.^{8, 19–22}

Taken together, it is not entirely clear why there are differential associations between BP variability and carotid intima-media thickness and coronary artery calcium. However, potential explanations may partly lie in the subtle differences in the information conveyed by carotid intima-media thickness and coronary artery calcium. Despite a significant correlation between coronary artery calcium and carotid intima-media thickness, carotid intima-media thickness is a less sensitive test than coronary artery calcium for the detection of subclinical atherosclerosis.²³ Carotid ultrasound was initially focused on visually identifying atherosclerotic plaque, stenosis, or flow-limiting occlusions^{24, 25} Changes in carotid intima-media thickness was however subsequently noted to be a strong predictor of atherosclerotic cardiovascular disease.^{23, 25} Carotid intima-media thickness may represent atherosclerotic plaque lesions or arterial wall intima or media hyperplasia. Current B-mode carotid ultrasound techniques are however not sufficiently sensitive to delineate between alterations that occur in the arterial wall intima from those that manifest in the media layer.²⁵ Arterial wall thickening predominantly driven by hyperplasia in the arterial wall media layer may also be observed in the process of arterial remodeling with advancing age and the aging-related influence of risk factors.²⁶ Increases in the carotid intima-media thickness may therefore occur with or without atherosclerosis.²⁵ In contrast, coronary artery calcium represents summative scoring of calcium detected in the coronary circulation. Consistent with this notion, differential relationships between other risk factors and carotid intima-media thickness but not with coronary artery calcium have been previously documented.²⁷ In the present study, we further tested the robustness of our results by carrying out a sensitivity analysis in a subset of participants with no coronary calcium (coronary artery calcium =0). In this subgroup, positive associations between long-term BP variability and carotid intima-media thickness remained significant.

Even with some evidence in favor of the supposition that changes in carotid intima-media thickness that may occur with BP variability reflect arterial wall thickening driven by media hyperplasia, rather than atherosclerosis per se, it is unclear why given its strong association with cardiovascular disease, BP variability would be related to arteriosclerosis (arterial wall thickening), but not to coronary artery calcium. One potential contributing factor is that the proximal large elastic arteries, such as the carotid artery, act as a primary buffer to modulate pressure pulsality. Proximal large arteries are thus inherently more exposed to long-term variability of BP compared to more distal medium-sized arteries such as those in the coronary circulation. Oscillating BP levels that have the greatest impact on the proximal large arteries, may induce shear stress on the arterial scaffold resulting in a cascade involving fragmentation of elastin fibers, endothelial dysfunction, inflammation and oxidative stress, and consequent arterial remodelling.^28–30 This premise is supported by experimental data that BP variability may lead to increased oscillatory shear stress to the vascular endothelium contributory to reduced nitric oxide bioavailability and increased expression of adhesion molecules.^{31, 32} This is additionally corroborated by observational data that show consistent positive associations between BP variability and large artery remodeling including arterial stiffnes,^{33, 34} carotid pulse wave velocity,³⁵ and aortic distensibility.³⁶ There is no consensus on a gold-standard approach to measure visit-to-visit BP variability.³⁷Variability independent of the mean is unlikely to be affected by average BP levels. However, the estimation of variability independent of the mean is derived from the distribution of BP within each population. Conversely, the assessments of SD versus variability independent of the mean are easier measures for clinicians to perform. Furthermore, the estimation of SD is not derived from the distribution of BP within each population, and thus the value itself can be used in other populations. For example, the current study suggests that each 4.26 mm Hg higher level of SD of systolic BP during young adulthood was associated with a 10 μm (standard error 3 μm) increment of common carotid intima-media thickness in midlife, independently of cumulative systolic BP levels during young adulthood. In the clinical setting, electronic health record systems could be programmed to calculate a person’s visit-to-visit BP variability over time using SD, which could help clinicians identify individuals at high risk for CVD events. Our results require further testing in an independent cohort to determine whether the assessment of visit-to-visit SBP variability in clinical practice improves detection and subsequent medical management of young adults at higher risk for CVD events.

Strengths and limitations

The core strengths of our study include the use of a contemporary community-based sample of young adults longitudinally followed as they transitioned from early adulthood to middle age, ability to account for long-term cumulative risk factor exposure, and the use of multi-modality imaging data with excellent reproducibility metrics. Several caveats however warrant further consideration. Our study design is observational and causal inferences are not intended. Excluded CARDIA study participants were more likely black, men, and had 1 mmHg higher systolic BP, but no statistical differences in diastolic BP, body mass index, physical activity, or total and high density lipoprotein cholesterol. Our participant sample still represented a diverse spectrum of cardiovascular risk factors, race, and sex. To the extent that less healthy participants are excluded, it would constitute a conservative bias that would result in underestimation of associations. Our results using clinic BP variability may not be generalizable to ambulatory settings. Different BP equipment was used in year-20, however a calibration study was performed to account for systematic differences. We did not assess other measures of arterial remodeling such as arterial stiffness or aortic distensibility. Different BP equipment was used in the year-20 exam, but a calibration study was performed to account for systematic differences in BP recordings. Our analysis did not examine change in coronary artery calcium or carotid intima-media thickness over time, so reverse causation cannot be excluded.

The optimal number of visits and interval time between visits to assess long-term BP variability is uncertain, but one prior report noted that 6 visits may be sufficient to estimate cardiovascular disease risk,³⁸ which is consistent with our current approach. In conclusion, our study observations highlight that long-term systolic and diastolic BP variability in early adulthood was positively associated with modest alterations in carotid intima-media thickness but not with coronary artery calcium in middle age.

PERSPECTIVES

In the setting of the increased utilization of electronic medical records in clinical practice, algorithms can be created to measure visit-to-visit variability in systolic and diastolic blood pressure. Our results suggest that in early life, apart from mean blood pressure values, variability of blood pressure may be associated with subclinical vascular changes that have been reported to antedate cardiovascular disease.

NOVELTY AND SIGNIFICANCE.

What Is New?

• Long-term systolic and diastolic blood pressure variability in early adulthood was positively associated with modest alterations in carotid intima-media thickness but not with coronary artery calcium in middle age.

What Is Relevant?

• The association between blood pressure variability and adverse cardiovascular disease outcomes may be, at least in part, influenced by differential associations with markers of subclinical atherosclerosis.

• Assessing visit-to-visit blood pressure variability may useful in identifying younger adults who may be susceptible to cardiovascular disease in later life.

Summary

• Long-term blood pressure variability beginning in young adulthood was associated with unfavorable changes in a surrogate of subclinical atherosclerosis, independent of mean blood pressure levels.