This observational study determines whether long-term cumulative systolic blood pressure in the risk equations improves the predictive ability compared with single systolic blood pressure measurements.
Key Points
Question
Does inclusion of long-term cumulative systolic blood pressure (SBP) in the American College of Cardiology/American Heart Association 10-year atherosclerotic cardiovascular disease risk equation improve the predictive ability compared with using single SBP measurements?
Findings
This cohort study included 11 767 adults from the Lifetime Risk Pooling Project, and both 5-year and 10-year cumulative SBP improved model performance as measured by the net reclassification index at event rate and the integrated discrimination index.
Meaning
Using long-term measures of cumulative SBP instead of single measurements can improve the ability of cardiovascular disease risk prediction models to correctly classify individuals in terms of their risk for cardiovascular disease.
Abstract
Importance
Long-term cumulative systolic blood pressure (SBP) is significantly associated with increased rates of atherosclerotic cardiovascular disease (ASCVD) development independent of single SBP levels. However, published ASCVD risk prediction algorithms only include currently measured SBP.
Objective
To determine whether including long-term (5- and 10-year) cumulative SBP in risk equations improves the predictive ability compared with single SBP measurements.
Design, Setting, and Participants
Adults aged 45 to 65 years at the time of risk estimation with at least 20 years of follow-up (5 and 10 years prior to risk estimation and 10 years of event follow-up). The Lifetime Risk Pooling Project included data from the following cohorts: Coronary Artery Risk Development in Young Adults Study, Atherosclerosis Risk in Communities Study, and Framingham Heart Study (both the original and offspring).
Exposures
Ten-year ASCVD risk, calculated using the approach of the 2013 American College of Cardiology/American Heart Association 10-year ASCVD risk equations, first with current SBP and then substituting 5- and 10-year cumulative SBP levels.
Main Outcomes and Measures
Incident ASCVD events that occurred over 10 years of follow-up, compared with the predicted risks, using the C statistic, net reclassification index at event rate, and the integrated discrimination index.
Results
This study included 11 767 participants with a mean (SD) age of 59.1 (4.7) years at risk estimation. A total of 6873 participants (58%) were women, and 1499 (13%) were African American. In the 10 years of follow-up from risk estimation, 1887 participants (16%) had an ASCVD event. There were no significant improvements in the C statistic when including 5- or 10-year cumulative SBP. However, the addition of cumulative SBP resulted in significant improvements in the net reclassification index at event rate (10-year net reclassification index for men, 0.04 [95% CI, 0.02-0.06]; 10-year net reclassification index for women, 0.03 [95% CI, 0.01-0.06]) and the relative integrated discrimination index (10-year relative integrated discrimination index for men, 0.12; 10-year relative integrated discrimination index for women, 0.10).
Conclusions and Relevance
Using long-term measures of cumulative blood pressure, instead of single measurements, can modestly improve the ability of cardiovascular disease risk prediction models to correctly classify individuals in terms of their risk for cardiovascular disease.
Introduction
Longitudinal systolic blood pressure (SBP) levels are significantly associated with increased incidence of atherosclerotic cardiovascular disease (ASCVD) independent of baseline or proximate SBP levels, suggesting that cumulative SBP may have additional predictive value over single SBP measures.1,2,3,4,5 The 2013 American College of Cardiology/American Heart Association pooled cohort 10-year ASCVD risk equations include SBP measured at a single point in time for quantitative risk assessment, ignoring previous blood pressure history that may be informative to ASCVD risk.6 However, the developers of these equations specifically called for evaluation of whether long-term risk factor levels impact the model performance.6 In this study, we tested whether using long-term (5- and 10-year) cumulative SBP measures in the 2013 pooled cohort equations could improve the model performance compared with single SBP measurements.
Methods
Study Population
This study used data from the Lifetime Risk Pooling Project cohorts that had adequate blood pressure follow-up during middle age as well as adjudicated ASCVD events: Coronary Artery Risk Development in Young Adults Study, Atherosclerosis Risk in Communities Study, and Framingham Heart Study (both the original and offspring cohorts).7 We included participants aged 45 to 65 years who were free of ASCVD with at least 20 years of observation time, including up to 10 years prior to risk estimation at the index age and for 10 years or longer of event follow-up after the index age. Included cohorts were approved by institutional review boards at their coordinating institutions, and all participants provided informed consent at each examination.
Blood Pressure Measurement
Blood pressure was reported as the mean of 2 or 3 seated measurements taken by trained personnel using standardized methods. Cumulative SBP was calculated by taking the means of SBP levels from each pair of consecutive examinations, multiplying each mean value by the years between the consecutive visits, and summing all available multiplied means (prior to the index age) to yield a 5-year or 10-year cumulative SBP exposure, expressed as mm Hg–years (similar to pack-years of cigarette exposure).
ASCVD Risk and Event Ascertainment
Using the 2013 American College of Cardiology/American Heart Association pooled cohort equations,6 three 10-year CVD risks were calculated for each participant, first using current SBP at the index age and then using 5- and 10-year cumulative SBP prior to the index age to represent SBP. These models also included age, sex, race/ethnicity, total cholesterol, high-density lipoprotein cholesterol, antihypertensive medication use, current smoking status, and diabetes. Event ascertainment occurred after the risk estimation and included incident ASCVD, defined as coronary heart disease death, nonfatal myocardial infarction, and fatal or nonfatal stroke. These data were available up to the most recent time of event adjudication by the included cohorts (December 2014 to December 2016).
Statistical Analysis
First, 10-year ASCVD risk equations were fitted using the approach of the 2013 American College of Cardiology/American Heart Association pooled cohort equations but with local β coefficient estimates corresponding the data of the included cohorts. Three sets of equations were fitted using each of the blood pressure measures of interest (current SBP, 5-year cumulative SBP, and 10-year cumulative SBP) in identical samples. Discrimination of the risk equations fitted with each of these 3 blood pressure variables was compared with observed ASCVD events by computing the C statistic.8 We also illustrated the predictive ability using calibration plots that compared the expected vs observed risk by deciles of predicted risk. Separately, the predictive ability of the 5-year and 10-year cumulative blood pressure was compared against current SBP using the net reclassification at event rate index and the integrated discrimination index (IDI). Standardized net benefit was plotted against the classification threshold for each blood measure, separately by sex (eFigure in the Supplement). Further description and interpretation of model performance measures are provided in eTable 1 in the Supplement. Analyses were stratified by sex for the main tables and are additionally stratified by race/ethnicity, age, and antihypertensive medication use in eTables 2 to 4 in the Supplement.
Results
This study included 11 767 participants; the mean (SD) index age was 59.1 (4.7) years at risk estimation, 6873 participants (58%) were women, and 1499 (13%) were African American. Those in the highest tertile of 10-year cumulative SBP were older, more likely to be men, and had a higher burden of other CVD risk factors (Table 1). In follow-up from risk estimation, 1887 participants (16%) had an ASCVD event. The ASCVD incidence rate was higher with increasing tertiles of 10-year cumulative SBP from 5.9 events per 1000 person-years in the lowest tertile to 9.9 and 18.4 in the second and third tertiles, respectively.
Table 1. Demographic Characteristics at the Time of Risk Prediction by Tertiles of 10-Year Cumulative Systolic Blood Pressure.
| Variable, Mean (SD) | Tertile of 10-y Cumulative SBP | Test Statistic P Valuea | ||
|---|---|---|---|---|
| Low (n = 3921) | Intermediate (n = 3925) | High (n = 3921) | ||
| 10-y cumulative SBP, mm Hg | 1066 (63.0) | 1216 (37.9) | 1411 (112) | <.001 |
| Age, y | 58.1 (4.8) | 59.0 (4.7) | 60.1 (4.3) | <.001 |
| Men, No. (%) | 1425 (36.3) | 1777 (45.3) | 1692 (43.2) | <.001 |
| African American, No. (%) | 390 (10.0) | 529 (13.5) | 580 (14.8) | <.001 |
| SBP, mm Hg | 111 (10.3) | 126 (11.3) | 144 (16.7) | <.001 |
| DBP, mm Hg | 67.4 (7.9) | 74.1 (8.3) | 80.7 (10.7) | <.001 |
| Body mass indexb | 26.9 (4.8) | 28.6 (5.4) | 29.2 (5.7) | <.001 |
| Total cholesterol, mg/dL | 206 (38.3) | 209 (40.9) | 219 (43.0) | <.001 |
| HDL cholesterol, mg/dL | 52.9 (16.9) | 49.9 (16.2) | 49.8 (16.0) | <.001 |
| Current smoker, No. (%) | 802 (20.5) | 737 (18.8) | 761 (19.4) | .17 |
| Antihypertensive medication, No. (%) | 577 (14.7) | 1174 (29.9) | 1949 (49.7) | <.001 |
| Lipid-lowering medication, No. (%) | 360 (9.2) | 400 (10.2) | 370 (9.5) | .29 |
| Diabetes, No. (%) | 249 (6.4) | 478 (12.2) | 635 (16.2) | <.001 |
| ASCVD event, No. (%)c | 307 (7.8) | 546 (13.9) | 1034 (26.4) | <.001 |
| ASCVD event rate per 1000 person-years | 5.9 | 9.9 | 18.4 | <.001 |
Abbreviations: ASCVD, atherosclerotic cardiovascular disease; DBP, diastolic blood pressure; HDL, high-density lipoprotein; SBP, systolic blood pressure.
SI conversion factor: To convert total and HDL cholesterol to millimoles per liter, multiply by 0.0259.
One-way analysis of variance F test used for continuous variables; χ2 test used for categorical variables.
Calculated as weight in kilograms divided by height in meters squared.
Atherosclerotic cardiovascular disease defined as coronary heart disease death, nonfatal myocardial infarction, and fatal or nonfatal stroke.
Current SBP and 5-year cumulative SBP were highly correlated (R = 0.86), as were current SBP and 10-year cumulative SBP (R = 0.80). Discrimination was similar for the 3 models across SBP measures for men and women. Likewise, calibration plots visually demonstrated high correlation between predicted and observed events for all 3 blood pressure measures (Figure). However, the regression coefficients were greater in magnitude for the 5-year and 10-year cumulative SBP compared with single SBP (hazard ratios per multiplicative increase in SBP of 2.69 and 2.74 for 5- and 10-year, respectively, compared with 2.12 for single SBP).
Figure. Calibration Plots of 10-Year ASCVD Risk by Blood Pressure Measures and Sex.
Data points indicate expected vs observed 10-year risk of ASCVD events by deciles of predicted risk. Bold line indicates perfect fit between predicted risk and observed events. In a situation of perfect calibration, the plotted points would lie on this line. Points above the line indicate underprediction; points below the line indicate overprediction. ASCVD indicates atherosclerotic cardiovascular disease; SBP, systolic blood pressure.
There were significant improvements, although modest, in model performance for both measures of cumulative SBP among both sexes (Table 2). For men, the net reclassification index was 0.04 (95% CI, 0.02-0.06) comparing 10-year cumulative SBP and current SBP and 0.03 (95% CI, 0.01-0.05) comparing 5-year cumulative SBP and current SBP. For women, the net reclassification index was 0.03 (95% CI, 0.01-0.06) comparing 10-year cumulative SBP and current SBP and 0.04 (95% CI, 0.02-0.06) comparing 5-year cumulative SBP and current SBP. For men, the absolute IDI was 0.006 (95% CI, 0.004-0.008) comparing 10-year cumulative SBP and current SBP and 0.006 (95% CI, 0.004-0.006) comparing 5-year cumulative SBP and current SBP. For women, the absolute IDI was 0.002 (95% CI, 0.001-0.003) comparing 10-year cumulative SBP and current SBP and 0.002 (95% CI, 0.001-0.003) comparing 5-year cumulative SBP and current SBP. When looking at the relative IDI among men, the 5-year and the 10-year SBP relative IDI was 0.12. Among women, the relative IDI was 0.09 for 5-year cumulative SBP and 0.10 for 10-year cumulative SBP.
Table 2. Predictive Ability of Blood Pressure Measures for 10-Year Atherosclerotic Cardiovascular Disease Risk by Sexa.
| Model | Current SBP | 5-y Cumulative SBP | 10-y Cumulative SBP |
|---|---|---|---|
| Men (n = 4894) | |||
| C statistic (95% CI) | 0.67 (0.65-0.69) | 0.68 (0.66-0.71) | 0.68 (0.66-0.70) |
| Discrimination slope | 0.033 | 0.038 | 0.038 |
| Category-free NRI (95% CI) | 1 [Reference] | 0.30 (0.23-0.38) | 0.27 (0.20-0.34) |
| NRI(p) (95% CI) | 1 [Reference] | 0.03 (0.01-0.05) | 0.04 (0.02-0.06) |
| Absolute IDI (95% CI) | 1 [Reference] | 0.006 (0.004-0.007) | 0.006 (0.004-0.008) |
| Relative IDI | [Reference] | 0.12 | 0.12 |
| Women (n = 6873) | |||
| C statistic (95% CI) | 0.68 (0.65-0.71) | 0.68 (0.65-0.71) | 0.68 (0.66-0.71) |
| Discrimination slope | 0.028 | 0.031 | 0.032 |
| Category-free NRI (95% CI) | 1 [Reference] | 0.17 (0.08-0.22) | 0.16 (0. 06-0.20) |
| NRI(p) (95% CI) | 1 [Reference] | 0.04 (0.02-0.06) | 0.03 (0.01-0.06) |
| Absolute IDI (95% CI) | 1 [Reference] | 0.002 (0.001-0.003) | 0.002 (0.001-0.003) |
| Relative IDI | [Reference] | 0.09 | 0.10 |
Abbreviations: IDI, integrated discrimination improvement; NRI(p), net reclassification index; SBP, systolic blood pressure.
Atherosclerotic cardiovascular disease defined as coronary heart disease death, nonfatal myocardial infarction, and fatal or nonfatal stroke. Risk prediction model also includes study-specific coefficients for age, race/ethnicity, total cholesterol, high-density lipoprotein cholesterol, current smoker, antihypertensive medication, and diabetes.
Discussion
Despite relatively high correlation between current SBP and long-term cumulative SBP, substituting either 5- or 10-year cumulative SBP in the ASCVD risk prediction equations significantly improved their ability to correctly classify individuals in terms of their risk for ASCVD, although improvements were small in magnitude. These improvements are comparatively modest when considering coronary artery calcium score, another novel cardiovascular risk predictor,9 but blood pressure is frequently measured and widely available in longitudinal electronic health record data. Implementation of use of cumulative SBP for risk prediction could feasibly occur without change to standard clinical practice, although it might be preferentially implemented in large health systems with robust electronic data capabilities.
Other studies have shown the importance of cumulative blood pressure on ASCVD risk2,4,5; in a 2015 study, cumulative SBP was a significant predictor of ASCVD events over a 20-year follow-up period, independent of current SBP.5 Our results extend these findings to test the impact of including cumulative SBP in an ASCVD risk estimator commonly used to support clinical decision making. However, future studies are needed to determine the most appropriate value of cumulative SBP, including time length, whether all measurements in the time period should be given equal weight, and whether functional forms such as trajectories may provide better characterization of cumulative blood pressure than numerical values.2,10
Limitations
This study has several limitations. The generalizability of these findings may be limited to populations similar to the included cohorts. In particular, the sample was predominantly white, and thus findings may not be consistent in other racial/ethnic groups. Further, the sample was limited to those who were CVD-free at the time of risk estimation (mean [SD] age, 59.1 [4.7] years), and the predictive ability of cumulative SBP may not be true for individuals who experience CVD events in early middle age. Lastly, blood pressure measurement is conducted in the included cohorts via detailed research protocols, which only moderately correlate to clinical blood pressure values.11 Likewise, these measures may differ from ambulatory blood pressure values. Additional validation in cohorts that use real-world blood pressure measures is warranted.12
Conclusions
Using long-term measures of cumulative blood pressure instead of single measurements can modestly improve the ability of CVD risk prediction models to correctly classify individuals in terms of their risk for CVD.
eFigure. Standardized Net Benefit versus Classification Threshold for Each Blood Pressure Measure by Sex
Table 1. Description and Interpretation of Performance Measures for Risk Prediction Models
eTable 2. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Race
eTable 3. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Age
eTable 4. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Antihypertensive Medication Use
eReferences
References
- 1.Lauer MS, Anderson KM, Levy D. Influence of contemporary versus 30-year blood pressure levels on left ventricular mass and geometry: the Framingham Heart Study. J Am Coll Cardiol. 1991;18(5):1287-1294. doi: 10.1016/0735-1097(91)90549-O [DOI] [PubMed] [Google Scholar]
- 2.Vasan RS, Massaro JM, Wilson PW, et al. ; Framingham Heart Study . Antecedent blood pressure and risk of cardiovascular disease: the Framingham Heart Study. Circulation. 2002;105(1):48-53. doi: 10.1161/hc0102.101774 [DOI] [PubMed] [Google Scholar]
- 3.Lee DS, Massaro JM, Wang TJ, et al. Antecedent blood pressure, body mass index, and the risk of incident heart failure in later life. Hypertension. 2007;50(5):869-876. doi: 10.1161/HYPERTENSIONAHA.107.095380 [DOI] [PubMed] [Google Scholar]
- 4.Allen N, Berry JD, Ning H, Van Horn L, Dyer A, Lloyd-Jones DM. Impact of blood pressure and blood pressure change during middle age on the remaining lifetime risk for cardiovascular disease: the cardiovascular lifetime risk pooling project. Circulation. 2012;125(1):37-44. doi: 10.1161/CIRCULATIONAHA.110.002774 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bonifonte A, Ayer T, Veledar E, Clark A, Wilson PW. Antecedent blood pressure as a predictor of cardiovascular disease. J Am Soc Hypertens. 2015;9(9):690-696. doi: 10.1016/j.jash.2015.06.013 [DOI] [PubMed] [Google Scholar]
- 6.Goff DC Jr, Lloyd-Jones DM, Bennett G, et al. ; American College of Cardiology/American Heart Association Task Force on Practice Guidelines . 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. 2014;63(25 Pt B):2935-2959. doi: 10.1016/j.jacc.2013.11.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wilkins JT, Karmali KN, Huffman MD, et al. Data resource profile: the cardiovascular disease lifetime risk pooling project. Int J Epidemiol. 2015;44(5):1557-1564. doi: 10.1093/ije/dyv150 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Pencina MJ, D’Agostino RB. Overall C as a measure of discrimination in survival analysis: model specific population value and confidence interval estimation. Stat Med. 2004;23(13):2109-2123. doi: 10.1002/sim.1802 [DOI] [PubMed] [Google Scholar]
- 9.Yeboah J, McClelland RL, Polonsky TS, et al. Comparison of novel risk markers for improvement in cardiovascular risk assessment in intermediate-risk individuals. JAMA. 2012;308(8):788-795. doi: 10.1001/jama.2012.9624 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Allen NB, Siddique J, Wilkins JT, et al. Blood pressure trajectories in early adulthood and subclinical atherosclerosis in middle age. JAMA. 2014;311(5):490-497. doi: 10.1001/jama.2013.285122 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ahmad FS, Chan C, Rosenman MB, et al. Validity of cardiovascular data from electronic sources: the multi-ethnic study of atherosclerosis and HealthLNK. Circulation. 2017;136(13):1207-1216. doi: 10.1161/CIRCULATIONAHA.117.027436 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.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]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eFigure. Standardized Net Benefit versus Classification Threshold for Each Blood Pressure Measure by Sex
Table 1. Description and Interpretation of Performance Measures for Risk Prediction Models
eTable 2. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Race
eTable 3. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Age
eTable 4. Predictive Ability of Blood Pressure Measures for 10-year ASCVD Risk, by Antihypertensive Medication Use
eReferences
