This cohort study assesses whether long-term variability and rate of change of blood pressure from young adulthood to midlife are associated with cardiovascular disease and all-cause mortality by middle age.
Key Points
Question
Is the long-term variability of blood pressure across clinical visits from young adulthood to midlife associated with cardiovascular disease and all-cause mortality by middle age?
Findings
In a cohort study of 3394 African American and white individuals, higher long-term visit-to-visit systolic blood pressure variability from young adulthood to midlife was associated with an increased risk for cardiovascular disease and all-cause mortality by middle age.
Meaning
These findings suggest that the assessment of visit-to-visit systolic blood pressure variability may help identify young adults at increased risk for cardiovascular disease and all-cause mortality.
Abstract
Importance
Determining blood pressure (BP) patterns in young adulthood that are associated with cardiovascular disease (CVD) events in later life may help to identify young adults who have an increased risk for CVD.
Objective
To determine whether the long-term variability of BP across clinical visits and the rate of change in BP from young adulthood to midlife are associated with CVD and all-cause mortality by middle age, independently of mean BP during young adulthood and a single BP in midlife.
Design, Setting, and Participants
This prospective cohort study included a community-based sample of 3394 African American and white participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study, enrolled from March 1985 through June 1986. Patterns of systolic BP (SBP) were evaluated with measurements at year 0 (baseline) and 2, 5, 7, and 10 years after baseline. Visit-to-visit SBP variability was estimated as BP variability independent of the mean (VIM). Data were collected from March 1985 through August 2015 and analyzed from June through October 2019.
Main Outcomes and Measures
Cardiovascular disease and all-cause mortality experienced through August 2015 were adjudicated. The associations of each SBP pattern with CVD events and all-cause mortality were determined using Cox proportional hazards regression models.
Results
At year 10, the mean (SD) age of the 3394 participants was 35.1 (3.6) years; 1557 (45.9%) were African American; 1892 (55.7%) were women; and 103 (3.0%) were taking antihypertensive medication. During a median follow-up of 20.0 (interquartile range, 19.4-20.2) years, 162 CVD events and 181 deaths occurred. When all BP pattern measurements were entered into the same model including a single SBP measurement at the year 10 examination, the hazard ratios for CVD events for each 1-SD increase in SBP measures were 1.25 (95% CI, 0.90-1.74) for mean SBP, 1.23 (95% CI, 1.07-1.43) for VIM SBP, and 0.99 (95% CI, 0.81-1.26) for annual change of SBP. The VIM for SBP was the only BP pattern associated with all-cause mortality (hazard ratio, 1.24; 95% CI, 1.09-1.41).
Conclusions and Relevance
The results of this study suggest that the assessment of visit-to-visit SBP variability may help identify young adults at increased risk for CVD and all-cause mortality later in life.
Introduction
The US Preventive Services Task Force and the 2017 American College of Cardiology/American Heart Association blood pressure (BP) guideline recommends using a mean of multiple BP measurements over time to screen for and manage high BP in young adults. Although BP is well known to vary across visits, little is known regarding the clinical relevance of that variability (ie, visit-to-visit BP variability) and the rate of change in BP over time in young adults.1,2
Prior studies have demonstrated that higher visit-to-visit BP variability and greater rate of change in BP over time are each associated with an increased risk for cardiovascular disease (CVD) events, independently of mean BP levels.3,4,5,6 These studies were conducted only in adults 50 years and older. Furthermore, little is known regarding specific BP patterns spanning from young adulthood to midlife that may be associated with CVD events in later life, independently of a single BP measurement in midlife.
We assessed BP patterns in young adulthood and evaluated the associations with CVD events and all-cause mortality by middle age using data from the Coronary Artery Risk Development in Young Adults (CARDIA) Study.7 Patterns of BP assessed included mean BP, cumulative BP exposure, visit-to-visit BP variability, and mean annual change in BP.
Methods
The CARDIA study enrolled 5114 African American and white adults aged 18 to 30 years of age from 4 US field centers (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California) from March 1985 through June 1986.7,8 All participants provided written informed consent at each study visit, and institutional review boards at each field center and the coordinating center approved the study annually. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Data were collected from March 1985 through August 2015. Follow-up examinations were conducted 2, 5, 7, 10, 15, 20, 25, and 30 years after baseline (year 0). Participants were contacted every 6 months by mail, by telephone, or electronically. Overall, 82.6% and 91.0% of the cohort members participated in a clinical examination or semiannual telephone interview in the 2 and 5 years, respectively, preceding the year 30 study examination.
BP and Other Measurements
At the baseline (year 0) and year 2, 5, 7, and 10 examinations, trained research staff measured BP 3 times in each participant’s right-arm brachial artery at 1-minute intervals after the participant had been sitting in a quiet room for 5 minutes, using a random-zero sphygmomanometer (Hawksley & Sons, Ltd). The mean of the second and third measurements was used for the analysis. The standardized BP measurement techniques in this study used validated equipment and met the recommendations in the 2017 American College of Cardiology/American Heart Association guideline for obtaining accurate clinical BP measurements. Other data were collected using standardized protocols and quality control procedures across study centers at each visit, as described previously (eMethods in the Supplement).7,8
BP Pattern Measurements
eFigure 1 in the Supplement shows the timeline for assessment of BP patterns and outcome events. Mean BP levels were calculated across 5 visits (baseline and 2, 5, 7, and 10 years after baseline) for each participant. Hypertension was defined as a mean systolic BP (SBP) of at least 130 mm Hg, mean diastolic BP (DBP) of at least 80 mm Hg, or use of antihypertensive medication.2 For SBP and DBP, visit-to-visit BP variability measurements included BP variability independent of the mean (VIM), SD, coefficient of variation (CV), and mean real variability (MRV). The SD of SBP was calculated as
 |
The SD, CV, and MRV data are correlated with mean BP; thus, distinguishing the effect of BP variability from that of mean BP on outcomes may be difficult. Conversely, VIM has been shown to have no correlation with mean BP levels but a strong correlation with SD, CV, and MRV.3,5 Therefore, in the primary analysis, we used VIM as a measure of visit-to-visit BP variability. In the secondary analyses, we used SD and MRV. Cumulative exposure to BP (millimeters of mercury times years) was defined as the summed mean BP for each pair of consecutive examinations multiplied by the time between these 2 consecutive visits in years.9 Mean annual change in BP from the year 0 through year 10 examinations was calculated using a linear regression model. A single BP measurement in midlife was defined using the BP measurement obtained at the year 10 examination. We illustrated how each BP pattern was calculated in the Figure. We also calculated within-visit BP variability at the year 0 and year 10 examinations using the first, second, and third readings within each examination.
Figure. Illustration of Systolic Blood Pressure (SBP) Patterns.
Example of individual follow-up data of SBP across 5 visits (year 0 to year 10 examinations). The variability independent of the mean (VIM) of SBP was defined as the intraindividual standard deviation (SD) of SBP across examinations (M/x)p where x is individual mean SBP across visits, M is the mean of individual mean 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 mean SBP was calculated as (year 0 SBP + year 2 SBP + year 5 SBP + year 7 SBP + year 10 SBP) / 5. The coefficient of variation (CV) of SBP was calculated as the SD of SBP divided by the mean SBP level. The absolute differences of SBP between successive SBP measurements are shown as Δ1 through Δ4 (blue arrows). For example, Δ1 represents the absolute difference in SBP between year 0 and year 2 measurements. The mean real variability (MRV) of SBP was calculated as (Δ1 + Δ2 + Δ3 + Δ4) / 4. The mean SBP between successive SBP measurements is shown as A1 through A4 (gray squares). Cumulative exposure to SBP was calculated as (A1 × 2 years) + (A2 × 3 years) + (A3 × 2 years) + (A4 × 3 years) and is shown by the gray shaded area, representing millimeters of mercury times years. The mean annual change in SBP from the year 0 to year 10 examinations (orange arrow) was calculated using a linear regression model. A single BP measurement in midlife was defined using the BP measurement obtained at the year 10 examination (oval).
Cardiovascular Outcomes
The protocol to assess outcomes is detailed in the eMethods in the Supplement. The primary outcome was composite CVD events (fatal and nonfatal coronary heart disease, hospitalization for heart failure, stroke, transient ischemic attack, or intervention for peripheral artery disease).10,11,12 All-cause mortality was examined as a secondary outcome. For the current analysis, outcomes were available through August 2015.
Statistical Analyses
Data were analyzed from June through October 2019. Descriptive statistics are presented as means with SDs and proportions as appropriate. We calculated the incidence rate of CVD events and all-cause mortality for participants overall and in quartiles of each BP measurement pattern. Using Cox proportional hazards regression models, we calculated the hazard ratios (HRs) and 95% CIs for CVD events and all-cause mortality associated with each BP pattern measurement. In the Cox model, BP measurements were modeled as continuous (ie, per 1-SD increase). The proportionality assumption for the Cox analyses was confirmed via the inclusion of a time-by-BP interaction. The year 10 examination date was defined as the time origin for time-to-event analysis. Follow-up time was censored on the date the event occurred. Participants who did not have events were censored at the last contact with the participant on or before August 2015. For participants who had CVD events and then died, their CVD events and deaths were each counted as outcomes. We calculated HRs in an unadjusted model (model 1) and after adjustment for age, sex, race, study site, educational level, and clinical and behavioral characteristics at the year 10 examination (body mass index [calculated as weight in kilograms divided by height in meters squared], smoking status, physical activity, total and high-density lipoprotein cholesterol levels, diabetes, and antihypertensive medication use) (model 2). Covariates were selected a priori because they have been shown to be associated with BP and CVD events.2,13 If a participant had missing information for a covariate at the year 10 examination, we adjusted for levels of that covariate at the closest examination before the year 10 examination. We entered each BP pattern measurement and BP at the year 10 examination into the same model (model 3). Furthermore, we included in the same model (1) mean SBP, VIM SBP, annual change in SBP from the year 0 to year 10 examination, and a single SBP measurement at the year 10 examination (model 4) and (2) mean DBP, VIM DBP, annual change in DBP from the year 0 to year 10 examinations, and a single DBP measurement at the year 10 examination (also model 4). The variable inflation factor for mean SBP and cumulative exposure to SBP when both were included in the same model was greater than 9.0.14 Owing to their high degree of collinearity, we did not include mean BP and cumulative exposure to SBP in the same multivariable model. We tested for heterogeneity in the association between each BP pattern measurement and outcomes by sex and race with the inclusion of multiplicative interaction terms. Stratified analyses were considered when a statistically significant interaction was present (P < .05).
We computed the change in the Harrell concordance (C) statistic,15 comparing models including covariates, mean BP and VIM or annual change in SBP from the year 0 to year 10 examinations, and a single BP measurement at the year 10 examination against models including covariates, mean BP from the year 0 to year 10 examinations, and a single BP measurement at the year 10 examination. We conducted 3 sensitivity analyses. First, we excluded participants taking antihypertensive medication at 1 or more examinations from the year 0 to year 10 examinations or with mean SBP of at least 130 mm Hg or mean DBP of at least 80 mm Hg from the year 0 to year 10 examinations. Second, we imputed missing data for BP measurements and covariates from the year 0 to year 10 examinations using multiple imputation with chained equations and 10 iterations as described by Raghunathan et al.16 We did not impute missing BP values in the primary analysis, because the number of missing measurements was substantial (6541 of 25 120 readings [26.0%]; 5 BP measurements per person times the number of people recruited in the current analysis). Third, we fitted Fine-Gray models17 to account for competing risk of CVD events vs all-cause deaths.
All statistical analyses were performed with Stata, version 12.1 (StataCorp LLC). Statistical significance was defined as P < .05 using 2-sided tests.
Results
Of the 5114 participants enrolled in the CARDIA study, we excluded 1 who withdrew study consent, 15 who experienced CVD events before follow-up started (ie, before their year 10 examination), 75 lost to or unavailable for follow-up before the year 10 examination, and 1630 participants with at least 1 visit without BP measurements from the year 0 to year 10 examinations (eFigure 2 in the Supplement), leaving a final analytical sample of 3394 participants (Table 1). Mean (SD) age of participants was 35.1 (3.6) years; 1892 (55.7%) were women, and 1502 (44.3%) were men; 1557 (45.9%) were African American, and 1837 (54.1%) were white.
Table 1. Characteristics of CARDIA Participants at the Year 10 Examinationa.
| Characteristic | Data |
|---|---|
| Year 10 examination | |
| Age, y | 35.1 (3.6) |
| Women, No. (%) | 1892 (55.7) |
| African American, No. (%) | 1557 (45.9) |
| BMI | 27.4 (6.4) |
| Current smokers, No. (%) | 829 (24.4) |
| Educational level, y | 14.8 (2.9) |
| Fasting glucose level, mg/dL | 88.1 (18.6) |
| Total cholesterol level, mg/dL | 178.2 (34.6) |
| HDL cholesterol level, mg/dL | 50.2 (13.9) |
| Physical activity, exercise units | 330.0 (269.7) |
| Antihypertensive medication use, No. (%) | 103 (3.0) |
| SBP, mm Hg | 109.8 (12.7) |
| DBP, mm Hg | 72.3 (10.2) |
| BP pattern using measurements from years 0 to 10 examinations | |
| Duration from the years 0 to 10 examinations, y | 10.0 (0.3) |
| Mean SBP, mm Hg | 108.71 (9.80) |
| Mean DBP, mm Hg | 69.29 (7.86) |
| Cumulative exposure to SBP, mm Hg × y | 1081.98 (101.27) |
| Cumulative exposure to DBP, mm Hg × y | 689.92 (82.15) |
| VIM SBP, U | 6.25 (2.78) |
| VIM DBP, U | 6.25 (2.95) |
| MRV SBP, mm Hg | 7.14 (3.64) |
| MRV DBP, mm Hg | 7.06 (3.49) |
| SD SBP, mm Hg | 6.29 (3.12) |
| SD DBP, mm Hg | 6.24 (2.94) |
| CV SBP, % | 5.75 (2.61) |
| CV DBP, % | 9.09 (4.42) |
| Mean annual change in SBP, mm Hg/y | 0.004 (1.03) |
| Mean annual change in DBP, mm Hg/y | 0.38 (0.95) |
Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); BP, blood pressure; CV, coefficient of variation; DBP, diastolic blood pressure; HDL, high-density lipoprotein; MRV, mean real variability; SBP, systolic blood pressure; VIM, BP variability independent of the mean.
SI conversion factors: To convert total and HDL cholesterol to millimoles per liter, multiply by 0.0259; glucose to millimoles per liter, multiply by 0.0555.
Includes 3394 participants. Unless otherwise indicated, data are expressed as mean (SD). Each BP pattern measurement was calculated using 5 clinical BP measurements from the year 0 to year 10 examinations.
The VIM SBP was correlated with CV SBP and MRV SBP (Pearson r, 0.80-0.99) (eTable 1 in the Supplement) but not with mean SBP levels or cumulative exposure to SBP (Pearson r, 0.01-0.03). Mean SBP was correlated with a single SBP measurement at the year 10 examination (Pearson r, 0.84), whereas the correlation between VIM SBP and a single SBP measurement at the year 10 examination was modest (Pearson r, 0.12) (eTables 2 and 3 in the Supplement). African American race, smoking, and prevalent diabetes at the year 10 examination were associated with higher mean SBP levels, cumulative exposure to SBP, VIM SBP, and mean annual change in SBP (P < .05 for all; eTables 2-5 in the Supplement). The associations between SD SBP from the year 0 to year 10 examinations and within-visit SD SBP at the year 0 or year 10 examination were low (Pearson r, 0.03-0.08).
During a median follow-up of 20.0 (interquartile range, 19.4-20.2) years, 162 CVD events (2.49 events per 1000 person-years) and 181 all-cause deaths (2.74 events per 1000 person-years) occurred. In an unadjusted model (Table 2), each 1-SD increase in mean SBP (per 9.80 mm Hg; HR, 1.97; 95% CI, 1.77-2.20), VIM SBP (per 2.78 U; HR, 1.44; 95% CI, 1.27-1.63), and mean annual change in SBP (per 1.03 mm Hg/y; HR, 1.55; 95% CI,1.39-1.73) was associated with an increased risk for CVD events. When a single SBP measurement at the year 10 examination was entered into the model that included mean SBP, VIM SBP, or annual change in SBP (model 3), the HRs for CVD events for each 1-SD increase in SBP at the year 10 examination (per 12.7 mm Hg) were 1.46 (95% CI, 1.16-1.84) in the mean SBP model, 1.50 (95% CI, 1.31-1.71) in the VIM SBP model, and 1.66 (95% CI, 1.36- 2.02) for annual change in the SBP model. When mean SBP, VIM SBP, and annual change in SBP were entered into the same model including SBP at the year 10 examination (model 4), the HRs for CVD events were 1.25 (95% CI, 0.90-1.74) for mean SBP, 1.23 (95% CI, 1.07-1.43) for VIM SBP, and 0.99 (95% CI, 0.81-1.26) for annual change in SBP. The results for each BP pattern measurement were similar when using DBP instead of SBP (Table 2) and using cumulative exposure to SBP instead of mean SBP levels (eTable 6 in the Supplement). There was no evidence of interaction between each BP pattern measurement and sex or race in associations with CVD events (P > .22 for interaction). The results were similar when using MRV and SD instead of VIM (eTable 7 in the Supplement). When MRV SBP or SD SBP were entered into the same model including mean SBP, annual change in SBP, and SBP at year 10 (model 3), the HRs for CVD events were 1.15 (95% CI, 0.998-1.32) for ARV SBP and 1.18 (95% CI, 1.03-1.35) for SD SBP.
Table 2. Hazard Ratios for CVD Associated With 1-SD Increase in Level of BP Patternsa.
| Variable | Mean BP | VIM | Mean Annual Change |
|---|---|---|---|
| Systolic BP | |||
| 1-SD change | 9.80 mm Hg | 2.78 U | 1.03 mm Hg/y |
| Model, HR (95% CI) | |||
| 1 | 1.97 (1.77-2.20) | 1.44 (1.27-1.63) | 1.55 (1.39-1.73) |
| 2 | 1.68 (1.44-1.96) | 1.37 (1.20-1.56) | 1.37 (1.22-1.54) |
| 3 | 1.15 (0.87-1.53) | 1.20 (1.04-1.38) | 0.97 (0.81-1.15) |
| 4 | 1.25 (0.90-1.74) | 1.23 (1.07-1.43) | 0.99 (0.81-1.26) |
| Diastolic BP | |||
| 1-SD change | 7.86 mm Hg | 2.95 U | 0.95 mm Hg/y |
| Model, HR (95% CI) | |||
| 1 | 2.03 (1.79-2.31) | 1.29 (1.15-1.46) | 1.43 (1.25-1.64) |
| 2 | 1.58 (1.35-1.86) | 1.20 (1.05-1.38) | 1.28 (1.11-1.46) |
| 3 | 1.22 (0.96-1.56) | 1.05 (0.91-1.22) | 0.97 (0.81-1.15) |
| 4 | 1.35 (0.99-1.84) | 1.11 (0.93-1.31) | 1.04 (0.82-1.32) |
Abbreviations: BP, blood pressure; CVD, cardiovascular disease; HR, hazard ratio; VIM, blood pressure variability independent of the mean.
Includes 3394 participants. Model 1 is unadjusted. Model 2 includes adjustment for age at the year 10 examination, race, sex, educational level, study site, and clinical and behavioral characteristics at the year 10 examination or the closest prior examination to the year 10 examination (body mass index, smoking status, physical activity, total and high-density lipoprotein cholesterol levels, diabetes, and antihypertensive medication use). Model 3 includes the variables in model 2 and a single systolic BP (SBP) measurement at the year 10 examination for analysis of SBP or a single diastolic BP (DBP) measurement at the year 10 examination for analysis of DBP. Model 4 includes the variables in model 3 and each SBP pattern measurement (mean SBP, VIM SBP, and annual change in SBP) for analysis of SBP or each DBP pattern measurement (mean DBP, VIM DBP, and annual change in DBP) for analysis of DBP.
In an unadjusted model (Table 3), higher mean SBP, VIM SBP, and mean annual change in SBP were associated with an increased risk for all-cause mortality. When mean SBP, VIM SBP, and annual change in SBP were entered into the same model including SBP at the year 10 examination (model 4), the HRs for all-cause mortality were 0.97 (95% CI, 0.71-1.31) for mean SBP, 1.24 (95% CI, 1.09-1.41) for VIM SBP, and 0.97 (95% CI, 0.80-1.18) for annual change in SBP. The results were similar when using DBP instead of SBP for each BP pattern (Table 3) and using cumulative exposure to SBP instead of mean SBP levels (eTable 8 in the Supplement). There was no evidence of interaction between each BP measurement and sex or race in associations with all-cause mortality (P > .07 for interaction). The results were similar when using MRV and SD instead of VIM (eTable 9 in the Supplement).
Table 3. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase in Level of BP Patternsa.
| Variable | Mean BP | VIM | Mean Annual Change |
|---|---|---|---|
| Systolic BP | |||
| 1-SD | 9.80 mm Hg | 2.78 U | 1.03 mm Hg/y |
| Model, HR (95% CI) | |||
| 1 | 1.51 (1.34-1.72) | 1.40 (1.25-1.58) | 1.29 (1.14-1.47) |
| 2 | 1.15 (0.98-1.35) | 1.27 (1.13-1.44) | 1.14 (1.00-1.29) |
| 3 | 0.93 (0.71-1.21) | 1.24 (1.09-1.41) | 1.03 (0.87-1.22) |
| 4 | 0.97 (0.71-1.31) | 1.24 (1.09-1.41) | 0.97 (0.80-1.18) |
| Diastolic BP | |||
| 1-SD change | 7.86 mm Hg | 2.95 U | 0.95 mm Hg/y |
| Model, HR (95% CI) | |||
| 1 | 1.44 (1.26-1.64) | 1.28 (1.15-1.44) | 1.26 (1.10-1.45) |
| 2 | 1.04 (0.88-1.22) | 1.20 (1.05-1.36) | 1.18 (1.03-1.35) |
| 3 | 0.88 (0.70-1.12) | 1.18 (1.03-1.35) | 1.17 (0.99-1.39) |
| 4 | 1.04 (0.78-1.39) | 1.15 (0.99-1.33) | 1.11 (0.89-1.38) |
Abbreviations: BP, blood pressure; HR, hazard ratio; VIM, blood pressure variability independent of the mean.
Includes 3394 participants. Model 1 is unadjusted. Model 2 includes adjustment for age at the year 10 examination, race, sex, educational level, study site, and clinical and behavioral characteristics at the year 10 examination or the closest prior examination to the year 10 examination (body mass index, smoking status, physical activity, total and high-density lipoprotein cholesterol levels, diabetes, and antihypertensive medication use). Model 3 includes the variables in model 2 and a single systolic BP (SBP) measurement at the year 10 examination for analysis of SBP or a single diastolic BP (DBP) measurement at the year 10 examination for analysis of DBP. Model 4 includes the variables in model 3 and each SBP pattern measurement (mean SBP, VIM SBP, and annual change in SBP) for analysis of SBP or each DBP pattern measurement (mean DBP, VIM DBP, and annual change in DBP) for analysis of DBP.
There were statistically significant increases in C statistics for CVD events (C statistic, 0.781 vs 0.771; mean change, 0.010; 95% CI, 0.001-0.020) when we incorporated VIM SBP from the year 0 to year 10 examinations into base models consisting of covariates, mean BP from the year 0 to year 10 examinations, and a single BP measurement at the year 10 examination (Table 4). Changes in C statistics for all-cause mortality were not statistically significant when we incorporated VIM or annual change in BP into base models.
Table 4. Changes in Discrimination for CVD Events and All-Cause Mortality for BP Patternsa.
| Variable | CVD Events (n = 162) | All-Cause Mortality (n = 181) | ||
|---|---|---|---|---|
| C Statistic (95% CI) | Change in C Statistic From Base Model, Mean (95% CI) | C Statistic (95% CI) | Change in C Statistic From Base Model, Mean (95% CI) | |
| Systolic BP | ||||
| Base model | 0.771 (0.717 to 0.825) | 0 [Reference] | 0.697 (0.641 to 0.753) | 0 [Reference] |
| Base model plus VIM | 0.781 (0.729 to 0.833) | 0.010 (0.001 to 0.020) | 0.709 (0.654 to 0.765) | 0.012 (−0.002 to 0.026) |
| Base model plus annual change | 0.767 (0.713 to 0.821) | −0.004 (−0.010 to 0.002) | 0.694 (0.637 to 0.750) | −0.004 (−0.007 to 0.0005) |
| Diastolic BP | ||||
| Base model | 0.768 (0.717 to 0.819) | 0 [Reference] | 0.694 (0.638 to 0.750) | 0 [Reference] |
| Base model plus VIM | 0.772 (0.722 to 0.823) | 0.005 (−0.001 to 0.010) | 0.702 (0.647 to 0.757) | 0.007 (−0.006 to 0.021) |
| Base model plus annual change | 0.772 (0.722 to 0.822) | 0.001 (−0.004 to 0.001) | 0.701 (0.646 to 0.757) | 0.007 (−0.003 to 0.017) |
Abbreviations: BP, blood pressure; CVD, cardiovascular disease; VIM, BP variability independent of the mean.
Harrell C was used to calculate C statistics for each outcome. The differences in C statistics and 95% CIs for each outcome after each BP measure was added to base models are shown. Base models include adjustment for age at the year 10 examination, race, sex, educational level, study site, and clinical and behavioral characteristics at the year 10 examination or the closest prior examination to the year 10 examination (body mass index, smoking status, physical activity, total and high-density lipoprotein cholesterol levels, diabetes, and antihypertensive medication use), a single systolic BP (SBP) measurement at the year 10 examination for analysis of SBP or a single diastolic BP (DBP) measurement at the year 10 examination for analysis of DBP, and mean SBP from the year 0 to year 10 examinations for analysis of SBP or mean DBP from the year 0 to year 10 examinations for analysis of DBP.
Sensitivity Analyses
Among 3010 participants not taking antihypertensive medication and who had a mean SBP of less than 130 mm Hg and mean DBP of less than 80 mm Hg from the year 0 to year 10 examinations, 110 CVD events (1.89 per 1000 person-years) and 137 all-cause deaths (2.32 per 1000 person-years) occurred. The HRs for CVD events and all-cause mortality for each SBP and DBP pattern measurement were similar to those from the primary analyses (eTables 10 and 11 in the Supplement). We imputed missing BP measurements and covariates for missing data collected from the year 0 to year 10 examinations (eTable 12 in the Supplement) for the 5114 participants. One participant withdrew study consent; this individual’s data were excluded from the analyses. We also excluded 90 participants who experienced CVD events or were lost to follow-up before the year 10 examination, leaving a final analytical sample size of 5024 participants. Mean values of each BP pattern measurement using imputation were similar to those without imputation (eTable 13 in the Supplement). Among the 5024 participants, 242 CVD events and 292 all-cause deaths occurred. Results with and without imputing missing BP were similar in terms of the point estimate for CVD events and all-cause mortality for each BP pattern measurement (eTables 14 and 15 in the Supplement). In Fine-Gray models (eTables 16 and 17 in the Supplement), the results with and without using competing risk models were similar in terms of the point estimate for risks of CVD events and all-cause mortality for each BP pattern measurement.
Discussion
In a community-based sample of African American and white individuals, greater long-term visit-to-visit SBP variability through young adulthood and into midlife was associated with a higher risk for CVD events and all-cause mortality by middle age, independently of mean SBP during young adulthood and a single SBP measure in midlife. A systematic review and meta-analysis including 24 studies (15 intervention trials and 9 observational studies)18 reported higher CVD risk for higher SD SBP and VIM SBP, independently of mean SBP levels. However, all of the studies included in the meta-analysis were conducted among adults 50 years and older or in high-risk populations (eg, participants with a history of stroke, coronary heart disease, or chronic kidney disease). These comorbidities might lead to higher visit-to-visit BP variability and CVD risk. Moreover, prior studies included participants taking antihypertensive medication.18 Antihypertensive drugs can affect visit-to-visit BP variability.19 The present study extends existing knowledge by demonstrating the association of long-term visit-to-visit BP variability among young adults without known CVD with CVD events and all-cause mortality in later life, and the associations were present among individuals not taking antihypertensive medication.
The correlation of mean BP levels with visit-to-visit BP variability measurements was modest. The risk for CVD associated with higher VIM SBP remained statistically significant after we adjusted for a single SBP measure at the year 10 examination. Conversely, when mean SBP or the rate of change in SBP from the year 0 to year 10 examinations and a single SBP at the year 10 examination were entered into the same model, CVD risk associated with higher mean SBP levels or the rate of change in SBP was attenuated, and only a single SBP measurement retained a statistically significant association with CVD events. Only visit-to-visit BP variability was associated with all-cause mortality. From this study, whether higher visit-to-visit BP variability is a causal driver for CVD events or a marker of poor health remains uncertain. Some meta-analyses that included post hoc studies from randomized clinical trials have suggested that changes in visit-to-visit BP variability attributed to intensification of antihypertensive medication were associated with greater reduction in stroke risk, independently of changes in mean BP levels.19,20 Calcium channel blockers have a stronger effect in reducing visit-to-visit BP variability compared with other classes of antihypertensive medication (eg, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and β-blockers).19,20,21 Thus, persons with high visit-to-visit BP variability may benefit more when prescribed calcium channel blockers instead of other classes of antihypertensive medication.
There is no consensus on a criterion standard approach to measure visit-to-visit BP variability.3 Variability independent of the mean is unlikely to be affected by mean BP levels. However, the estimation of VIM is derived from the distribution of BP within each population. Conversely, the assessments of SD and MRV vs VIM are easier measures for health care professionals to perform. Furthermore, the estimation of SD and MRV is not derived from the distribution of BP within each population, and thus the value itself can be used in other populations. For example, each 3.12-mm Hg higher level of SD SBP during young adulthood was associated with an 18% higher risk (95% CI, 3%-35%) for CVD events, independently of mean SBP levels during young adulthood and a single SBP measurement in midlife. 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 and MRV, which could help health care professionals identify individuals at high risk for CVD events. However, discrimination of participants with and without CVD events was modestly improved when assessing visit-to-visit SBP variability in addition to mean SBP during young adulthood and a single SBP measurement in midlife. 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.
Masked hypertension is a phenotype associated with high CVD risk.2 Controversy persists regarding whether CVD risk increases in association with white-coat hypertension compared with sustained normotension (ie, nonhypertensive BP inside and outside of the clinic).2 In general, masked hypertension and white-coat hypertension are diagnosed based on BP obtained during a single clinic visit, and it is unclear whether a person’s BP level would be different with additional clinic visits that included BP measurements. In the CARDIA Study, no participants underwent BP measurements outside of the clinic using 24-hour ambulatory BP monitoring at the year 0, 2, 7, and 10 examinations, and only a sample of 300 participants underwent ambulatory BP monitoring at the year 5 examination. Therefore, we were unable to investigate whether persons with masked or white-coat hypertension had higher office visit-to-visit SBP variability compared with those who had sustained normotension in CARDIA.
Strengths and Limitations
Strengths of this study include the large, well-phenotyped, community-based cohort of white and African American individuals, adjudication of suspected cardiovascular outcomes by a panel of physicians using detailed evaluation criteria, high retention, and the standardized data collection protocols and rigorous quality control. The young age of participants included in the CARDIA study means that all events in these analyses should be considered premature CVD events and deaths.
It remains unclear how BP measurements in this study collected in a highly controlled research setting correspond to BP measurements commonly obtained in a typical clinic setting. Further, the method applied to estimate visit-to-visit BP variability was based on BP measurements collected over 5 visits at intervals of 2 to 3 years. This approach is clinically relevant, because the US Preventive Services Task Force recommends screening for high BP in adults aged 18 to 39 years every 3 to 5 years.1 The interval between BP measurements and the number of readings not only influences visit-to-visit BP variability but could also affect the strength of its association with health outcomes.4,5 Therefore, visit-to-visit BP variability based on measurements taken at intervals of 2 to 3 years may have a different association with outcomes when compared with visit-to-visit BP variability based on measurements taken at intervals of 1 to 12 months. Determining how best to assess visit-to-visit BP variability to better identify young adults at high risk of adverse outcomes merits further investigation. Our results may not be generalizable to other racial and ethnic groups (eg, Asian and Hispanic).
Conclusions
In this population-based cohort study of African American and white individuals, greater visit-to-visit SBP variability at younger than 40 years was associated with a higher risk for CVD events and all-cause mortality over the ensuing 20 years. These outcomes occurred independently of mean SBP during young adulthood and a single SBP measure in midlife.
eMethods. CARDIA Study
eFigure 1. Study Design
eFigure 2. Flow Chart: Sample for the Analyses, CARDIA
eTable 1. Correlations Among Blood Pressure Pattern Measurements (n = 3394)
eTable 2. Associations of Each Systolic Blood Pressure Pattern Measurement With Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 3. Differences in Each Systolic Blood Pressure Pattern Measurement by Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 4. Associations of Each Diastolic Blood Pressure Pattern Measurement With Characteristics at the Year 0 or Year 10 Examinations (n = 3394)
eTable 5. Differences in Each Diastolic Blood Pressure Pattern Measurement by Demographic Variables and Clinical Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 6. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 7. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 8. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 9. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 10. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns Among Participants not Taking Antihypertensive Medication and With Mean SBP<130 mm Hg and DBP<80 mm Hg From the Year 0 to Year 10 Examinations (n = 3010)
eTable 11. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns Among Participants not Taking Antihypertensive Medication and With Mean SBP<130 mm Hg and DBP<80 mm Hg From the Year 0 to Year 10 Examinations (n = 3010)
eTable 12. Observation Number of Imputed Blood Pressure and Covariates: a Sensitivity Analysis With Imputation of Missing Blood Pressure and Covariate Measurements
eTable 13. Mean (SD) of Each Blood Pressure Pattern Measurement in a Multiple Imputation Sample (n = 5024)
eTable 14. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase in Blood Pressure Patterns in a Multiple Imputation Sample (n = 5024)
eTable 15. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns in a Multiple Imputation Sample (n = 5024)
eTable 16. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns: Fine-Gray Model
eTable 17. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns: Fine-Gray Model
eReferences.
References
- 1.Siu AL; US Preventive Services Task Force . Screening for high blood pressure in adults: US Preventive Services Task Force recommendation statement. Ann Intern Med. 2015;163(10):778-786. doi: 10.7326/M15-2223 [DOI] [PubMed] [Google Scholar]
- 2.Whelton PK, Carey RM, Aronow WS, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Hypertension. 2018;71(6):e13-e115. doi: 10.1161/HYP.0000000000000065 [DOI] [PubMed] [Google Scholar]
- 3.Yano Y. Visit-to-visit blood pressure variability—what is the current challenge? Am J Hypertens. 2017;30(2):112-114. doi: 10.1093/ajh/hpw124 [DOI] [PubMed] [Google Scholar]
- 4.Shimbo D, Newman JD, Aragaki AK, et al. Association between annual visit-to-visit blood pressure variability and stroke in postmenopausal women: data from the Women’s Health Initiative. Hypertension. 2012;60(3):625-630. doi: 10.1161/HYPERTENSIONAHA.112.193094 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Rothwell PM, Howard SC, Dolan E, et al. Prognostic significance of visit-to-visit variability, maximum systolic blood pressure, and episodic hypertension. Lancet. 2010;375(9718):895-905. doi: 10.1016/S0140-6736(10)60308-X [DOI] [PubMed] [Google Scholar]
- 6.Muntner P, Whittle J, Lynch AI, et al. Visit-to-visit variability of blood pressure and coronary heart disease, stroke, heart failure, and mortality: a cohort study. Ann Intern Med. 2015;163(5):329-338. doi: 10.7326/M14-2803 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Friedman GD, Cutter GR, Donahue RP, et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol. 1988;41(11):1105-1116. doi: 10.1016/0895-4356(88)90080-7 [DOI] [PubMed] [Google Scholar]
- 8.National Heart, Lung, and Blood Institute Coronary Artery Risk Development in Young Adults (CARDIA) Study. https://www.nhlbi.nih.gov/research/resources/obesity/population/cardia.htm. Accessed February 1, 2018.
- 9.Liu K, Colangelo LA, Daviglus ML, et al. Can antihypertensive treatment restore the risk of cardiovascular disease to ideal levels? the Coronary Artery Risk Development in Young Adults (CARDIA) Study and the Multi-Ethnic Study of Atherosclerosis (MESA). J Am Heart Assoc. 2015;4(9):e002275. doi: 10.1161/JAHA.115.002275 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Rosamond WD, Chang PP, Baggett C, et al. Classification of heart failure in the atherosclerosis risk in communities (ARIC) study: a comparison of diagnostic criteria. Circ Heart Fail. 2012;5(2):152-159. doi: 10.1161/CIRCHEARTFAILURE.111.963199 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Luepker RV, Apple FS, Christenson RH, et al. ; AHA Council on Epidemiology and Prevention; AHA Statistics Committee; World Heart Federation Council on Epidemiology and Prevention; European Society of Cardiology Working Group on Epidemiology and Prevention; Centers for Disease Control and Prevention; National Heart, Lung, and Blood Institute . Case definitions for acute coronary heart disease in epidemiology and clinical research studies: a statement from the AHA Council on Epidemiology and Prevention; AHA Statistics Committee; World Heart Federation Council on Epidemiology and Prevention; the European Society of Cardiology Working Group on Epidemiology and Prevention; Centers for Disease Control and Prevention; and the National Heart, Lung, and Blood Institute. Circulation. 2003;108(20):2543-2549. doi: 10.1161/01.CIR.0000100560.46946.EA [DOI] [PubMed] [Google Scholar]
- 12.Adams HP Jr, Bendixen BH, Kappelle LJ, et al. Classification of subtype of acute ischemic stroke: definitions for use in a multicenter clinical trial: TOAST: Trial of Org 10172 in Acute Stroke Treatment. Stroke. 1993;24(1):35-41. doi: 10.1161/01.STR.24.1.35 [DOI] [PubMed] [Google Scholar]
- 13.D’Agostino RB Sr, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117(6):743-753. doi: 10.1161/CIRCULATIONAHA.107.699579 [DOI] [PubMed] [Google Scholar]
- 14.Hair JF, Anderson R, Tatham RL, Black WC. Multivariate Data Analysis. Upper Saddle River, NJ: Prentice Hall; 2006. [Google Scholar]
- 15.Newson RB. Comparing the predictive powers of survival models using Harrell’s C or Somers’ D. Stata J. 2010;10:339-358. doi: 10.1177/1536867X1001000303 [DOI] [Google Scholar]
- 16.Raghunathan TE, Lepkowski JM, van Hoewyk J, Solenberger P. A multivariate technique for multiply imputing missing values using a sequence of regression models. Surv Methodol. 2001;27:85-95. [Google Scholar]
- 17.Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999;94:496-509. doi: 10.1080/01621459.1999.10474144 [DOI] [Google Scholar]
- 18.Stevens SL, Wood S, Koshiaris C, et al. Blood pressure variability and cardiovascular disease: systematic review and meta-analysis. BMJ. 2016;354:i4098. doi: 10.1136/bmj.i4098 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Webb AJ, Fischer U, Mehta Z, Rothwell PM. Effects of antihypertensive-drug class on interindividual variation in blood pressure and risk of stroke: a systematic review and meta-analysis. Lancet. 2010;375(9718):906-915. doi: 10.1016/S0140-6736(10)60235-8 [DOI] [PubMed] [Google Scholar]
- 20.An S, Bao M, Wang Y, et al. Relationship between cardiovascular health score and year-to-year blood pressure variability in China: a prospective cohort study. BMJ Open. 2015;5(10):e008730. doi: 10.1136/bmjopen-2015-008730 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kollias A, Stergiou GS, Kyriakoulis KG, Bilo G, Parati G. Treating visit-to-visit blood pressure variability to improve prognosis: is amlodipine the drug of choice? Hypertension. 2017;70(5):862-866. doi: 10.1161/HYPERTENSIONAHA.117.10087 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. CARDIA Study
eFigure 1. Study Design
eFigure 2. Flow Chart: Sample for the Analyses, CARDIA
eTable 1. Correlations Among Blood Pressure Pattern Measurements (n = 3394)
eTable 2. Associations of Each Systolic Blood Pressure Pattern Measurement With Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 3. Differences in Each Systolic Blood Pressure Pattern Measurement by Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 4. Associations of Each Diastolic Blood Pressure Pattern Measurement With Characteristics at the Year 0 or Year 10 Examinations (n = 3394)
eTable 5. Differences in Each Diastolic Blood Pressure Pattern Measurement by Demographic Variables and Clinical Characteristics at the Year 0 or 10 Examinations (n = 3394)
eTable 6. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 7. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 8. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 9. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns (n = 3394)
eTable 10. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns Among Participants not Taking Antihypertensive Medication and With Mean SBP<130 mm Hg and DBP<80 mm Hg From the Year 0 to Year 10 Examinations (n = 3010)
eTable 11. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns Among Participants not Taking Antihypertensive Medication and With Mean SBP<130 mm Hg and DBP<80 mm Hg From the Year 0 to Year 10 Examinations (n = 3010)
eTable 12. Observation Number of Imputed Blood Pressure and Covariates: a Sensitivity Analysis With Imputation of Missing Blood Pressure and Covariate Measurements
eTable 13. Mean (SD) of Each Blood Pressure Pattern Measurement in a Multiple Imputation Sample (n = 5024)
eTable 14. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase in Blood Pressure Patterns in a Multiple Imputation Sample (n = 5024)
eTable 15. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns in a Multiple Imputation Sample (n = 5024)
eTable 16. Hazard Ratios for Cardiovascular Disease Associated With 1-SD Increase of Blood Pressure Patterns: Fine-Gray Model
eTable 17. Hazard Ratios for All-Cause Mortality Associated With 1-SD Increase of Blood Pressure Patterns: Fine-Gray Model
eReferences.