Original Article
Is Pulse Pressure an Independent Risk Factor for Incident Stroke, REasons for Geographic And Racial Differences in Stroke Stephen P. Glasser,1 Daniel L. Halberg,2 Charles D. Sands,3 Aleena Mosher,4 Paul M. Muntner,5 and George Howard4 BACKGROUND Pulse pressure (PP) is a potential risk factor of stroke. The relationship of incident stroke with systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial pressure (MAP), and PP was examined. METHODS Data were from the REasons for Geographic And Racial Differences in Stroke national cohort study of 30,239 black and white participants aged ≥45 years, enrolled between 2003 and 2007. PP (SBP−DBP) and MAP (MAP = DBP + 1/3*PP) were calculated. Telephone followup occurred every six months for self or proxy-reported suspected stroke events, confirmed using expert adjudication. Cox-proportional hazards models examined the association of incident stroke for the different BP measurements with multivariable adjustment for sociodemographic and clinical risk factors including gender and race.
maximum 10 years), 916 strokes occurred. Unadjusted PP (hazard ratio [HR] = 1.30; 95% confidence interval [CI] 1.24–1.35), SBP (HR = 1.22; 95% CI 1.18–1.32), MAP (HR = 1.24; 95% CI 1.16–1.32), and DBP (HR = 1.09; 95% CI 1.01–1.17) were associated with stroke risk; however, after adjustment for SBP and other risk factors, the association with PP was attenuated (HR = 0.98; 95% CI 0.90–1.07), whereas SBP persisted as a predictor (HR = 1.14; 95% CI 1.06–1.23). These associations were consistent across age (younger vs. older >70 years) and race (black vs. white).
CONCLUSIONS PP is positively associated with incident stroke, but not independently from SBP; and, there were no significant gender, racial, or regional differences in that association.
RESULTS Men and women without prevalent stroke at baseline were analyzed (n = 25,462). During follow-up (mean 6.3 ± 2.3 years,
Keywords: blood pressure; hypertension; pulse pressure, incident stroke.
Current definitions of hypertension are primarily based on systolic blood pressure (SBP) and diastolic blood pressure (DBP) but not on pulse pressure (PP). More recently, increased attention has been given to PP as a predictor of stroke risk, with PP hypothesized to be an indicator of the stiffness of large arteries, especially the aorta.1–4 SBP and DBP increase with age in a parallel manner until about age 60, after which SBP continues to rise and DBP begins to decrease. This age-related phenomenon results in the large increase in PP after age 60 and an increase in the prevalence of isolated systolic hypertension.5 As a result, PP may be a key blood pressure (BP) measure in older individuals and may be important as a risk factor for cardiovascular disease, including stroke, myocardial infarction (MI), and death.3
Some evidence suggests PP may be an independent predictor of stroke, even independent of SBP.6Although studies have explored a possible relationship between PP and occurrence of stroke events, most of these studies have included predominantly white populations.2,5 Few data on this subject are available in African-Americans (AA), and it is unclear if there is a correlation between PP and the incidence of stroke in this population. Differences in the association of PP and stroke between women and men have also been reported, but not extensively.1 The primary purpose of this study was to further investigate if PP is a useful predictor for stroke risk. We postulated that differences in PP, independent of SBP may represent one possible contributor to the racial differences in stroke risk.
Correspondence: Stephen P. Glasser ([email protected]).
1Department of Medicine, Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA; 2Department of Pharmaceutical Sciences, McWhorter School of Pharmacy, Samford University, Birmingham, Alabama, USA; 3Department of Epidemiology, Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA; 4Department of Biostatistics, University of Alabama at Birmingham, Birmingham, Alabama, USA; 5Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama, USA.
Initially submitted September 5, 2014; date of first revision October 7, 2014; accepted for publication November 30, 2014; online publication January 14, 2015.
doi:10.1093/ajh/hpu265
© American Journal of Hypertension, Ltd 2015. All rights reserved. For Permissions, please email: [email protected]
American Journal of Hypertension 28(8) August 2015 987
Glasser et al.
As such, The REasons for Geographic And Racial Differences in Stroke (REGARDS) could provide valuable information as to whether racial differences exist relative to PP, and its effect on incident stroke. We also explored whether AAs have a greater stroke risk at same PP compared with whites (that is a differential susceptibility compared with whites), and if regional differences in that association exist. METHODS Study population
REGARDS is a national, population-based, biracial, longitudinal cohort study designed to examine underlying causes for racial and regional differences in stroke and CHD mortality.7 The study oversampled AAs and persons living in the Stroke Belt region of the United States, an area that has stroke mortality rates higher than the rest of the country. Between January 2003 and October 2007, 30,239 individuals were enrolled, including 42% AA, 58% white, 45% men, and 55% women. The sample enrolled 21% of participants from the Stroke Belt Buckle (coastal plain region of North Carolina, South Carolina, and Georgia), 35% from the Stroke Belt states (remainder of North Carolina, South Carolina, and Georgia, plus Alabama, Mississippi, Tennessee, Arkansas, and Louisiana), and the remaining 44% from the other 40 contiguous states (referred to as non-Belt). REGARDS participants were selected from commercially available lists (Genesys). A letter and brochure informed participants of the study and an upcoming phone call. During that call, verbal consent was obtained and a 45-minute questionnaire was administered. Including an estimate of eligibility among participants not reached, the telephone participation rate was 33%; the cooperation rate among those with confirmed eligibility was 49% (similar to the Multi-Ethnic Study of Atherosclerosis, which had a 39.8% cooperation rate among those contacted and to whom the study was explained).8 A participant was considered enrolled in the study if they completed the 45-minute telephone questionnaire and the in-person physical examination at baseline. Using a computer-assisted telephone interview, demographic information and medical history were obtained by trained interviewers. Consent was obtained verbally by telephone and subsequently in writing during the in-person physical examination. The physical examination, 3–4 weeks after the telephone interview, included anthropometric and BP measurements, collection of a blood sample, and an electrocardiogram. Height, weight, and BP measurements were obtained from the in-person component. SBP and DBP were defined as the average of 2 measurements taken by a trained technician using a standard protocol and regularly tested aneroid sphygmomanometer, measured after the participant was seated for five minutes.9 BP quality control was monitored by central examination of digit preference and retraining of technicians took place as necessary. A medication inventory was conducted via pill bottle review during the examination. PP was then calculated as the difference in SBP and DBP, whereas mean arterial pressure (MAP) was calculated as DBP plus one-third of PP (MAP = DBP + .33*PP). Selfadministered questionnaires including the Block98 Food 988 American Journal of Hypertension 28(8) August 2015
Frequency Questionnaire (NutritionQuestTM, Berkeley, CA) were left with the participant to gather information. Participants were followed by telephone at six-month intervals for surveillance of stroke events. Report of a potential event triggered medical retrieval and reports of death triggered interviews with the next-of-kin or other proxies provided by the participant in addition to retrieval of any hospital records that corresponded to a hospitalization near the time of death. The National Death Index was also queried for the cause of death. Stroke events were defined as either meeting the World Health Organization definition (“an acute neurologic dysfunction of vascular origin with sudden (within seconds) or at least rapid (within hours) occurrence of symptoms and signs corresponding to the involvement of focal areas in the brain” or as having symptoms lasting >24 hours and neuroimaging consistent with acute ischemia or hemorrhage.10,11 Study methods were reviewed and approved by all involved institutional review boards. Additional methodological details are provided elsewhere.7 Risk factors considered in this analysis included diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia, and anti-hypertension medication use. Diabetes was defined as fasting glucose level ≥ 126 (or non-fasting ≥ 200) or self-report of glucose control medication. History of heart disease was defined as self-report of cardiovascular events or revascularization procedures, or evidence of MI obtained from electrocardiogram. Self-reported levels of smoking were never, past, and current. Atrial fibrillation was defined from self-report or electrocardiogram evidence. Dyslipidemia was defined as selfreported use of lipid-lowering medication. Antihypertensive medication was defined by self-reported current use.
Figure 1. Exclusionary cascade. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure.
Pulse Pressure and Stroke Table 1. Demographic and clinical characteristics of subjects at baseline examination according to SBP and DBP categories DBP < 80
80 ≤ DBP < 90
90 ≤ DBP
SBP < 120 Participants/events Age (years)
7,037/174
887/13
13/0
62.8 ± 9.4
61.1 ± 9.1
61.1 ± 11.1
Black
2,125 (30%)
339 (38%)
3 (23%)
Male
2,756 (39%)
395 (45%)
4 (31%)
Heart disease
1,016 (14%)
83 (9%)
1 (8%)
Dyslipidemia
2,154 (31%)
251 (28%)
3 (23%)
570 (8%)
69 (8%)
1 (8%)
Atrial fibrillation LVH
386 (5%)
61 (7%)
1 (8%)
Smoke (current)
1,046 (15%)
106 (12%)
9 (69%)
Diabetes
1,056 (15%)
113 (13%)
0 (0%)
120 ≤ SBP < 140 Participants/events
6,967/262
4,872/169
484/13
Age (years)
66.6 ± 9.2
63.1 ± 8.7
60.4 ± 8.6
Black
2,599 (37%)
2,183 (45%)
262 (54%)
Male
3,153 (45%)
2,388 (49%)
255 (53%)
Heart disease
1,273 (18%)
699 (14%)
56 (12%)
Dyslipidemia
2,501 (36%)
1,507 (31%)
130 (27%)
Atrial fibrillation
549 (8%)
362 (7%)
42 (9%)
LVH
647 (9%)
507 (10%)
60 (12%)
853 (12%)
654 (13%)
86 (18%)
1,627 (23%)
909 (19%)
78 (16%)
Smoke (current) Diabetes 140 ≤ SBP Participants/events
1,440/93
2,167/114
1,595/78
Age (years)
70.8 ± 8.5
Black
616 (43%)
66.8 ± 8.6
63.4 ± 8.8
1,069 (49%)
982 (62%)
Male
675(47%)
1,055 (49%)
797 (50%)
Heart disease
365 (25%)
430 (20%)
252 (16%)
Dyslipidemia
543 (38%)
749 (35%)
438 (27%)
Atrial fibrillation
165 (11%)
187 (8%)
126 (8%)
LVH
194 (13%)
305 (14%)
300 (19%)
Smoke (current)
190 (13%)
353 (16%)
298 (19%)
Diabetes
464 (32%)
602 (28%)
411 (26%)
Abbreviations: DBP = diastolic blood pressure; SBP = systolic blood pressure; Values = mean ± SD or percentage; LVH = left ventricular hypertension.
Statistical analysis
Baseline demographic and clinical characteristics of eligible participants were summarized within strata of SBP and DBP. Continuous variables were summarized by means and standard deviation, whereas categorical variables were summarized by proportions. Cox-proportional hazards models were used to examine the association of incident stroke with the different BP measurements (SBP, DBP, PP, and MAP) in a series of incremental models. Model 1 is the unadjusted hazard ratios (HRs),
whereas model 2 adjusts for age, race, age by race interaction (previously shown to be significant for stroke risk), gender, income, education, and geographic region. Model 3 adds adjustments for stroke risk factors including diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia. Model 4 adds adjustment for anti-hypertension medication use. Using HRs, we assessed if these relations were consistent within age strata, where the population was dichotomized at age 70 (age selected to have approximately the same number of stroke events in each strata), among black and white participants, and for ischemic and hemorrhagic stroke outcomes. Unadjusted American Journal of Hypertension 28(8) August 2015 989
Glasser et al.
HRs were also ascertained for categorical SBP (SBP < 120, 120 < SBP < 140, SBP > 140) and DBP (DBP < 80, 80 < DBP < 90, DBP > 90) classifications for all participants, older and younger participants, and black and white participants. Proportional hazards assumptions were tested by including the interaction of PP with log-transformed follow-up time in the multivariable Cox models. Multiple imputation was used with stroke outcomes to avoid potential biases introduced from differences in the characteristics of those where medical records could not be retrieved for adjudication, or where records were still in the adjudication process. As in other longitudinal cohort studies, the process of endpoint determination requires retrieval and physician adjudication of medical records. A corollary of this approach is a suspected event cannot be confirmed without successful medical record retrieval. However, there are factors of interest that affect the likelihood of being successful in retrieving records, for example, medical records are more difficult to retrieve following hospitalization of AA. We employed multiple imputation in order to remove the potential bias introduced by factors potentially confounded with record retrieval.12 Briefly, based on records where we were successful in retrieval, we developed a logistic regression model predicting the probability that a suspected event would be confirmed in adjudication. Clearly, the most powerful predictor in this model was the reason why the record was being sought, for example with a much higher likelihood of there truly being a stroke if the participant reported they were hospitalized for a stroke than when they simply reported having stroke symptoms that were not judged to be a stroke by a clinician. We then applied this prediction model to all suspected events where medical records were not successfully retrieved, and generated 10 data sets where events were generated using a binomial distribution with the modeled probabilities. These data sets were then analyzed using the approach of Rubin.13 Additional details of
the imputation process are available elsewhere.12 Cox regression models and descriptive statistics were calculated with SAS version 9.3 (SAS Institute, Cary, NC). RESULTS
Of the total REGARDS sample (n = 30,239), 1,930 reported a history of stroke and were excluded from the primary analysis. After further exclusion of informed consent errors (n = 56), missing follow-up data (n = 469), missing any covariate data (n = 2,258), and missing SBP and/or DBP (n = 64), the analytic sample was 25,462 men and women ages 45–98 (mean age 64.6 ± 9.4 years, 40.0% black, 45.1% male, Figure 1). The characteristics of the study sample are shown in Table 1, according to joint SBP and DBP classifications. As expected, age at the time of stroke was greater in each race and sex category (age at time of stroke was 71.6 ± 8.9 vs. 67.7 ± 8.6 years in white vs. black men; and, 72.2 ± 8.2 vs. 68.9 ± 9.6 years in white vs. black women). Current cigarette smoking was reported in 30% vs. 19% of black men when compared with without incident stroke, and this held for all race/sex comparisons although with lower percentages. In addition, only 13 participants had SBP < 120 and DBP > 90 mm Hg. Generally, participants with higher SBP were older and sicker, whereas participants with higher DBP were younger and healthier. Table 2 presents the HRs for stroke associated with levels of SBP, DBP, MAP, and PP (per 10 mm Hg). A total of 916 strokes over 160,211 person-years of follow-up revealed unadjusted associations between each BP measure (SBP, DBP, PP, and MAP) and incident stroke. The HR for incident stroke associated with PP held for all models. Further, PP remained significant after adjustment for MAP, but was substantially attenuated and was not statistically significant after adjustment for SBP, risk factors, and comorbid conditions (HR 0.98,
Table 2. Overall HRs of the various BP (per 10 mm Hg) parameters and incident stroke Model 1
Model 2
Model 3
Model 4
PP
1.30 (1.24–1.35)
1.14 (1.09–1.20)
1.12 (1.07–1.18)
1.11 (1.06–1.16)
MAP
1.24 (1.16–1.32)
1.19 (1.12–1.27)
1.19 (1.12–1.27)
1.17 (1.10–1.25)
SBP
1.22 (1.18–1.26)
1.14 (1.09–1.18)
1.13 (1.08–1.17)
1.11 (1.07–1.16)
DBP
1.09 (1.01–1.17)
1.14 (1.05–1.22)
1.15 (1.07–1.23)
1.13 (1.05–1.22)
Single pressure models
Two (pairwise) pressure models PP
1.27 (1.20–1.33)
1.10 (1.04–1.16)
1.07 (1.02–1.13)
1.07 (1.01–1.12)
MAP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
PP
1.21 (1.11–1.32)
1.01 (0.93–1.10)
0.98 (0.90–1.07)
0.98 (0.90–1.07)
SBP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
PP
1.29 (1.24–1.35)
1.14 (1.09–1.20)
1.12 (1.07–1.17)
1.11 (1.06–1.16)
DBP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
Abbreviations: BP = blood pressure; DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. SBP and DBP in mm Hg. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
990 American Journal of Hypertension 28(8) August 2015
Pulse Pressure and Stroke Table 3. HRs for incident stroke associated with PP, MAP, SBP, and DBP (per 10 mm Hg) in older vs. younger participants Model 1
Model 2
Model 3
Model 4
Older participants Single pressure models PP
1.12 (1.05–1.19)
1.10 (1.03–1.17)
1.10 (1.03–1.17)
1.09 (1.03–1.17)
MAP
1.18 (1.09–1.29)
1.18 (1.08–1.29)
1.18 (1.08–1.29)
1.17 (1.07–1.28)
SBP
1.12 (1.07–1.18)
1.12 (1.06–1.17)
1.11 (1.06–1.17)
1.11 (1.05–1.17)
DBP
1.13 (1.02–1.25)
1.14 (1.03–1.26)
1.14 (1.03–1.26)
1.13 (1.02–1.25)
1.06 (0.98–1.13)
1.06 (0.98–1.13)
1.05 (0.98–1.13)
Two (pairwise) pressure models PP
1.07 (1.07–1.15)
MAP
1.13 (1.13–1.25)
1.13 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
PP
0.99 (0.99–1.11)
0.97 (0.86–1.09)
0.97 (0.86–1.09)
0.97 (0.86–1.09)
SBP
1.13 (1.13–1.25)
1.14 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
PP
1.12 (1.12–1.19)
1.10 (1.04–1.18)
1.10 (1.04–1.17)
1.10 (1.03–1.17)
DBP
1.13 (1.13–1.25)
1.14 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
1.20 (1.11–1.29)
1.14 (1.06–1.23)
1.12 (1.04–1.21)
Younger participants Single pressure models PP
1.32 (1.24–1.42)
MAP
1.30 (1.19–1.42)
1.20 (1.10–1.32)
1.19 (1.09–1.30)
1.16 (1.06–1.27)
SBP
1.24 (1.18–1.31)
1.16 (1.09–1.23)
1.13 (1.07–1.20)
1.11 (1.05–1.18)
DBP
1.19 (1.07–1.32)
1.14 (1.02–1.26)
1.15 (1.04–1.28)
1.13 (1.02–1.25)
Two (pairwise) pressure models PP
1.25 (1.15–1.36)
1.15 (1.05–1.25)
1.09 (1.00–1.18)
1.08 (0.99–1.17)
MAP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
PP
1.15 (1.01–1.32)
1.07 (0.94–1.22)
1.00 (0.88–1.14)
1.00 (0.88–1.14)
SBP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
PP
1.31 (1.22–1.40)
1.19 (1.10–1.28)
1.13 (1.05–1.22)
1.12 (1.03–1.20)
DBP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
Abbreviations: DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
0.90–1.07), whereas SBP persisted as a significant predictor (HR 1.14, 1.06–1.23) in the presence of PP. There was no significant influence of self-reported antihypertensive drug use on these associations. In the overall analysis, the HRs for the various measures in older vs. younger and in blacks vs. whites was not appreciably different (Tables 3 and 4). We also assessed, in a series of incremental models, the results restricted to only normotensive participants, with PP as the exposure variable; and, the results were similar to the models with all participants. Also, we compared the imputed to the non-imputed data, and the most the HR changed was 0.05 (data not shown). DISCUSSION
Our analysis of the REGARDS data suggests that PP, SBP, DBP, and MAP are positively associated with stroke, but
counter to our hypothesis, after multivariable adjustment for SBP, the association of PP is attenuated and becomes statistically insignificant. Cardiovascular mortality and its various clinical manifestations such as angina, MI, stroke, and cardiac failure are increased in the presence of elevated BP. Traditionally, this has been measured by sphygmomanometry, which yields 2 measures: SBP and DBP. It is increasingly argued that neither SBP nor DBP may be as strong a predictor of a stroke as PP. Thus, PP may be a key BP measure, particularly in older individuals, and may increase in importance as a risk factor for cardiovascular disease, including stroke, MI, and death. However, arguments persist as to whether PP is an independent measure (particularly from SBP) of stroke risk.1 The Framingham study reported on its longitudinal follow-up of persons over 50 years of age, finding that cardiovascular (and in particular coronary) American Journal of Hypertension 28(8) August 2015 991
Glasser et al. Table 4. HRs of the association of incident stroke and PP (per 10 mm Hg) in black vs. white participants Model 1
Model 2
Model 3
Model 4
PP
1.26 (1.17–1.34)
1.15 (1.07–1.24)
1.13 (1.05–1.22)
1.12 (1.03–1.20)
MAP
1.24 (1.14–1.35)
1.23 (1.13–1.34)
1.22 (1.12-1.33)
1.20 (1.10–1.31)
SBP
1.20 (1.14–1.26)
1.15 (1.09–1.22)
1.14 (1.08–1.20)
1.13 (1.06–1.19)
DBP
1.14 (1.02–1.26)
1.19 (1.08–1.32)
1.19 (1.08–1.32)
1.18 (1.06–1.30)
1.21 (1.11–1.31)
1.08 (0.99–1.18)
1.06 (0.97–1.15)
1.05 (0.97–1.15)
MAP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
PP
1.13 (0.99–1.28)
0.97 (0.85–1.11)
0.95 (0.83–1.08)
0.95 (0.84–1.08)
SBP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
PP
1.25 (1.17–1.34)
1.14 (1.06–1.23)
1.12 (1.04–1.20)
1.11 (1.03–1.19)
DBP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
Black participants Single pressure models
Two (pairwise) pressure models PP
White participants Single pressure models PP
1.32 (1.25–1.40)
1.14 (1.07–1.21)
1.12 (1.05–1.20)
1.11 (1.04–1.18)
MAP
1.21 (1.11–1.33)
1.15 (1.05–1.26)
1.16 (1.06–1.27)
1.14 (1.04–1.25)
SBP
1.23 (1.17–1.30)
1.12 (1.06–1.18)
1.12 (1.06–1.18)
1.10 (1.04–1.16)
DBP
1.02 (0.92–1.13)
1.08 (0.97–1.20)
1.10 (0.99–1.22)
1.08 (0.98–1.20)
Two (pairwise) pressure models PP
1.31 (1.23–1.40)
1.11 (1.03–1.19)
1.08 (1.01–1.17)
1.08 (1.00–1.16)
MAP
1.03 (0.93–1.13)
1.09 (0.93–1.20)
1.10 (1.00–1.22)
1.09 (0.98–1.21)
PP
1.29 (1.15–1.45)
1.05 (0.93–1.18)
1.02 (0.90–1.15)
1.02 (0.90–1.15)
SBP
1.03 (0.93–1.13)
1.09 (0.98–1.20)
1.10 (1.00–1.22)
1.09 (0.98–1.21)
PP
1.32 (1.25–1.40)
1.14 (1.07–1.21)
1.12 (1.05–1.20)
1.11 (1.04–1.18)
DBP
1.03 (0.93–1.13)
1.09 (0.98–1.20)
1.10 (1.06–1.22)
1.09 (0.98–1.21)
Abbreviations: DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
mortality is associated with increased PP.5 In that study, mortality was related independently with initial systolic and diastolic pressure, but the strongest association was with PP, and when systolic pressure was initially considered, there was a negative association with diastolic pressure. In other words, for a given systolic pressure, lower diastolic pressure was associated with greater mortality. We have reported the association of PP and incident CHD from the same REGARDS population and found that although PP was positively associated with CHD, this association was attenuated after multivariable adjustment that included SBP.14 As Millar and Lever15 pointed out, the superiority of PP as a cardiac risk predictor in hypertension is supported by “3 strands of evidence”: PP is a risk factor for coronary events (MI, angina, heart failure, and cardiac death); PP is strongly associated with, and a potential determinant of 992 American Journal of Hypertension 28(8) August 2015
several surrogate markers of cardiac risk such as echocardiographically determined left atrial size and ventricular mass,16 electrocardiographic indices of ischemia and cardiac size,17 and carotid wall thickness;18,19 and, there is a physiologically plausible mechanism linking increased conduit artery stiffness to cardiac risk via raised PP. That is, increased arterial wall stiffness (itself a well-recognized index of cardiovascular mortality)20 causes increased systolic BP during systole, increased pulse wave velocity, and early return of the reflected pressure wave during late systole rather than during diastole. These effects increase left ventricular work and oxygen requirements simultaneously and tend to diminish coronary perfusion.21,22 On the other hand, cerebral blood flow occurs throughout the cardiac cycle and the relationship between PP and stroke is weak.14 Others have reported that high PP reflecting large artery stiffness is a significant independent risk factor for
Pulse Pressure and Stroke
cardiovascular, especially coronary, mortality in different populations. Madhavan et al. 23 reported that untreated hypertensive subjects with a PP of 63 mm Hg had an increased risk of cardiovascular complications. In addition, they found that these subjects were at greater risk of MI when there was too great a fall in DBP after treatment.5 Franklin et al. also reported that in a larger population of treated and untreated hypertensive subjects, PP was the only BP measurement independently related to the treatment incidence of MI.20 The link between PP and cardiovascular complications has also been shown in subjects who had MI with left ventricular dysfunction.1 Our findings did not confirm those previous studies that support our hypothesis of an association between PP and stroke risk that was independent of SBP. In our study, the correlation of PP and SBP was 0.82, making it mathematically challenging to establish an independent effect. Our data does support the previous studies that did not demonstrate an association of PP (independent of SBP) and incident stroke.1,24,25 We postulate that the main reason for the difference in the association of stroke with CHD is the strong association of stroke with SBP, mitigating the effects of other BP variables. In addition, the pathophysiology of stroke and CHD is not identical. Our study has several limitations worth noting. Some non-laboratory risk factors were based on self-report (although this is common to many epidemiologic studies), and individuals without telephones were necessarily excluded from selection into the study population. These excluded individuals who may be of lower socioeconomic status, and therefore, may have different risk factor profiles than those included in this analysis. In conclusion, in the REGARDS population, the role of SBP as the dominant index of BP as a predictor of stroke risk is supported. Although PP does predict stroke risk, its association is not as strong and is not independent of the effect of SBP. These data potentially suggest that efforts to evaluate and manage stroke risk through BP control should focus predominantly on SBP. This is in contrast to our prior report that did demonstrate and association of PP and CHD.6
ACKNOWLEDGMENTS
This research project is supported by a cooperative agreement from the National Institute of Neurological Disorders and Stroke (U01 NS041588), National Institutes of Health, and Department of Health and Human Service. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health. Representatives of the funding agency have been involved in the review of the manuscript but not directly involved in the collection, management, analysis, or interpretation of the data. We thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at http://www.regardsstudy.org.
DISCLOSURE
The authors declared no conflict of interest.
REFERENCES 1. Benetos A, Safar M, Rudnichi A, Smulyan H, Richard JL, Ducimetieère P, Guize L. Pulse pressure: a predictor of long-term cardiovascular mortality in a French male population. Hypertension 1997; 30:1410–1415. 2. Brown DW, Giles WH, Greenlund KJ. Blood pressure parameters and risk of fatal stroke, NHANES II mortality study. Am J Hypertens 2007; 20:338–341. 3. Geeganage C, Sare G, Bath PM. Pulse pressure as a predictor of stroke. Expert Rev Neurother 2008; 8:165–167. 4. Verdecchia P, Schillaci G, Reboldi G, Franklin SS, Porcellati C. Different prognostic impact of 24-hour mean blood pressure and pulse pressure on stroke and coronary artery disease in essential hypertension. Circulation 2001; 103:2579–2584. 5. Franklin SS, Khan SA, Wong ND, Larson MG, Levy D. Is pulse pressure useful in predicting risk for coronary heart Disease? The Framingham heart study. Circulation 1999; 100:354–360. 6. Okada K, Iso H, Cui R, Inoue M, Tsugane S. Pulse pressure is an independent risk factor for stroke among middle-aged Japanese with normal systolic blood pressure: the JPHC study. J Hypertens 2011; 29:319–321. 7. Howard VJ, Cushman M, Pulley L, Gomez CR, Go RC, Prineas RJ, Graham A, Moy CS, Howard G. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology 2005; 25:135–143. 8. MESA Coordinating Center, University of Washington. MESA Exam 1 Participation Rate. http://www.mesa-nhlbi.org/participation.aspx 2006. Accessed 25 June 2013. 9. U.S. Department of Health and Human Services, NIoH. National Heart Lung and Blood Institute. http://www.nhlbi.nih.gov/index.htm Accessed 5 June 2014. 10. Anonymous. Stroke—1989. Recommendations on stroke prevention, diagnosis, and therapy. Report of the WHO Task Force on Stroke and other Cerebrovascular Disorders. Stroke 1989; 20:1407–1431. 11. Luepker RV, Apple FS, Christenson RH, Crow RS, Fortmann SP, Goff D, Goldberg RJ, Hand MM, Jaffe AS, Julian DG, Levy D, Manolio T, Mendis S, Mensah G, Pajak A, Prineas RJ, Reddy KS, Roger VL, Rosamond WD, Shahar E, Sharrett AR, Sorlie P, Tunstall-Pedoe H; 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; 108:2543–2549. 12. Howard G, McClure LA, Moy CS, Safford MM, Cushman M, Judd SE, et al. Imputation of incident events in longitudinal cohort studies. Am J Epidemiol 2011; 174:718–726. 13. Rubin DB. Multiple Imputation for Nonresponse in Surveys. J. Wiley & Sons: New York, 1987. 14. Halberg DL, Sands C, Le A, Howard VJ, Safford M GS, Muntner P. Pulse and mean arterial pressure as predictors of stroke in the REGARDS study [abstract]. Stroke 2012; 43:A2593. 15. Millar JA, Lever AF. Implications of pulse pressure as a predictor of cardiac risk in patients with hypertension. Hypertension 2000; 36:907–911. 16. Gardin JM, Arnold A, Gottdiener JS, Wong ND, Fried LP, Klopfenstein HS, O’Leary DH, Tracy R, Kronmal R. Left ventricular mass in the elderly. The Cardiovascular Health Study. Hypertension 1997; 29:1095–1103. 17. Darne B, Girerd X, Safar M, Cambien F, Guize L. Pulsatile versus steady component of blood pressure: a cross-sectional analysis and a prospective analysis on cardiovascular mortality. Hypertension 1989; 13:392–400.
American Journal of Hypertension 28(8) August 2015 993
Glasser et al. 18. Mitchell GF. Pulse pressure, arterial compliance and cardiovascular morbidity and mortality. Curr Opin Nephrol Hypertens 1999; 8:335–342. 19. Zanchetti A, Bond MG, Hennig M, Neiss A, Mancia G, Dal Palù C, Hansson L, Magnani B, Rahn KH, Reid J, Rodicio J, Safar M, Eckes L, Ravinetto R. Risk factors associated with alterations in carotid intimamedia thickness in hypertension: baseline data from the European Lacidipine Study on Atherosclerosis. J Hypertens 1998; 16:949–961. 20. Blacher J, Pannier B, Guerin AP, Marchais SJ, Safar ME, London GM. Carotid arterial stiffness as a predictor of cardiovascular and all-cause mortality in end-stage renal disease. Hypertension 1998; 32:570–574. 21. Franklin SS. Cardiovascular risks related to increased diastolic, systolic and pulse pressure. An epidemiologist’s point of view. Pathol Biol (Paris) 1999; 47:594–603.
994 American Journal of Hypertension 28(8) August 2015
22. Safar ME, Siche JP, Mallion JM, London GM. Arterial mechan ics predict cardiovascular risk in hypertension. J Hypertens 1997; 15:1605–1611. 23. Madhavan S, Ooi WL, Cohen H, Alderman MH. Relation of pulse pressure and blood pressure reduction to the incidence of myocardial infarction. Hypertension 1994; 23:395–401. 24. Domanski M, Davis B, Pfeffer M, Kastantin M, Mitchell GF. Isolated systolic hypertension: prognostic information provided by pulse pressure. Hypertension 1999; 34:375–380. 25. Domanski M, Norman J, Wolz M, Mitchell G, Pfeffer M. Cardiovascular risk assessment using pulse pressure in the first national health and nutrition examination survey (NHANES I). Hypertension 2001; 38:793–797.
Is Pulse Pressure an Independent Risk Factor for Incident Stroke, REasons for Geographic And Racial Differences in Stroke Stephen P. Glasser,1 Daniel L. Halberg,2 Charles D. Sands,3 Aleena Mosher,4 Paul M. Muntner,5 and George Howard4 BACKGROUND Pulse pressure (PP) is a potential risk factor of stroke. The relationship of incident stroke with systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial pressure (MAP), and PP was examined. METHODS Data were from the REasons for Geographic And Racial Differences in Stroke national cohort study of 30,239 black and white participants aged ≥45 years, enrolled between 2003 and 2007. PP (SBP−DBP) and MAP (MAP = DBP + 1/3*PP) were calculated. Telephone followup occurred every six months for self or proxy-reported suspected stroke events, confirmed using expert adjudication. Cox-proportional hazards models examined the association of incident stroke for the different BP measurements with multivariable adjustment for sociodemographic and clinical risk factors including gender and race.
maximum 10 years), 916 strokes occurred. Unadjusted PP (hazard ratio [HR] = 1.30; 95% confidence interval [CI] 1.24–1.35), SBP (HR = 1.22; 95% CI 1.18–1.32), MAP (HR = 1.24; 95% CI 1.16–1.32), and DBP (HR = 1.09; 95% CI 1.01–1.17) were associated with stroke risk; however, after adjustment for SBP and other risk factors, the association with PP was attenuated (HR = 0.98; 95% CI 0.90–1.07), whereas SBP persisted as a predictor (HR = 1.14; 95% CI 1.06–1.23). These associations were consistent across age (younger vs. older >70 years) and race (black vs. white).
CONCLUSIONS PP is positively associated with incident stroke, but not independently from SBP; and, there were no significant gender, racial, or regional differences in that association.
RESULTS Men and women without prevalent stroke at baseline were analyzed (n = 25,462). During follow-up (mean 6.3 ± 2.3 years,
Keywords: blood pressure; hypertension; pulse pressure, incident stroke.
Current definitions of hypertension are primarily based on systolic blood pressure (SBP) and diastolic blood pressure (DBP) but not on pulse pressure (PP). More recently, increased attention has been given to PP as a predictor of stroke risk, with PP hypothesized to be an indicator of the stiffness of large arteries, especially the aorta.1–4 SBP and DBP increase with age in a parallel manner until about age 60, after which SBP continues to rise and DBP begins to decrease. This age-related phenomenon results in the large increase in PP after age 60 and an increase in the prevalence of isolated systolic hypertension.5 As a result, PP may be a key blood pressure (BP) measure in older individuals and may be important as a risk factor for cardiovascular disease, including stroke, myocardial infarction (MI), and death.3
Some evidence suggests PP may be an independent predictor of stroke, even independent of SBP.6Although studies have explored a possible relationship between PP and occurrence of stroke events, most of these studies have included predominantly white populations.2,5 Few data on this subject are available in African-Americans (AA), and it is unclear if there is a correlation between PP and the incidence of stroke in this population. Differences in the association of PP and stroke between women and men have also been reported, but not extensively.1 The primary purpose of this study was to further investigate if PP is a useful predictor for stroke risk. We postulated that differences in PP, independent of SBP may represent one possible contributor to the racial differences in stroke risk.
Correspondence: Stephen P. Glasser ([email protected]).
1Department of Medicine, Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA; 2Department of Pharmaceutical Sciences, McWhorter School of Pharmacy, Samford University, Birmingham, Alabama, USA; 3Department of Epidemiology, Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, Alabama, USA; 4Department of Biostatistics, University of Alabama at Birmingham, Birmingham, Alabama, USA; 5Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama, USA.
Initially submitted September 5, 2014; date of first revision October 7, 2014; accepted for publication November 30, 2014; online publication January 14, 2015.
doi:10.1093/ajh/hpu265
© American Journal of Hypertension, Ltd 2015. All rights reserved. For Permissions, please email: [email protected]
American Journal of Hypertension 28(8) August 2015 987
Glasser et al.
As such, The REasons for Geographic And Racial Differences in Stroke (REGARDS) could provide valuable information as to whether racial differences exist relative to PP, and its effect on incident stroke. We also explored whether AAs have a greater stroke risk at same PP compared with whites (that is a differential susceptibility compared with whites), and if regional differences in that association exist. METHODS Study population
REGARDS is a national, population-based, biracial, longitudinal cohort study designed to examine underlying causes for racial and regional differences in stroke and CHD mortality.7 The study oversampled AAs and persons living in the Stroke Belt region of the United States, an area that has stroke mortality rates higher than the rest of the country. Between January 2003 and October 2007, 30,239 individuals were enrolled, including 42% AA, 58% white, 45% men, and 55% women. The sample enrolled 21% of participants from the Stroke Belt Buckle (coastal plain region of North Carolina, South Carolina, and Georgia), 35% from the Stroke Belt states (remainder of North Carolina, South Carolina, and Georgia, plus Alabama, Mississippi, Tennessee, Arkansas, and Louisiana), and the remaining 44% from the other 40 contiguous states (referred to as non-Belt). REGARDS participants were selected from commercially available lists (Genesys). A letter and brochure informed participants of the study and an upcoming phone call. During that call, verbal consent was obtained and a 45-minute questionnaire was administered. Including an estimate of eligibility among participants not reached, the telephone participation rate was 33%; the cooperation rate among those with confirmed eligibility was 49% (similar to the Multi-Ethnic Study of Atherosclerosis, which had a 39.8% cooperation rate among those contacted and to whom the study was explained).8 A participant was considered enrolled in the study if they completed the 45-minute telephone questionnaire and the in-person physical examination at baseline. Using a computer-assisted telephone interview, demographic information and medical history were obtained by trained interviewers. Consent was obtained verbally by telephone and subsequently in writing during the in-person physical examination. The physical examination, 3–4 weeks after the telephone interview, included anthropometric and BP measurements, collection of a blood sample, and an electrocardiogram. Height, weight, and BP measurements were obtained from the in-person component. SBP and DBP were defined as the average of 2 measurements taken by a trained technician using a standard protocol and regularly tested aneroid sphygmomanometer, measured after the participant was seated for five minutes.9 BP quality control was monitored by central examination of digit preference and retraining of technicians took place as necessary. A medication inventory was conducted via pill bottle review during the examination. PP was then calculated as the difference in SBP and DBP, whereas mean arterial pressure (MAP) was calculated as DBP plus one-third of PP (MAP = DBP + .33*PP). Selfadministered questionnaires including the Block98 Food 988 American Journal of Hypertension 28(8) August 2015
Frequency Questionnaire (NutritionQuestTM, Berkeley, CA) were left with the participant to gather information. Participants were followed by telephone at six-month intervals for surveillance of stroke events. Report of a potential event triggered medical retrieval and reports of death triggered interviews with the next-of-kin or other proxies provided by the participant in addition to retrieval of any hospital records that corresponded to a hospitalization near the time of death. The National Death Index was also queried for the cause of death. Stroke events were defined as either meeting the World Health Organization definition (“an acute neurologic dysfunction of vascular origin with sudden (within seconds) or at least rapid (within hours) occurrence of symptoms and signs corresponding to the involvement of focal areas in the brain” or as having symptoms lasting >24 hours and neuroimaging consistent with acute ischemia or hemorrhage.10,11 Study methods were reviewed and approved by all involved institutional review boards. Additional methodological details are provided elsewhere.7 Risk factors considered in this analysis included diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia, and anti-hypertension medication use. Diabetes was defined as fasting glucose level ≥ 126 (or non-fasting ≥ 200) or self-report of glucose control medication. History of heart disease was defined as self-report of cardiovascular events or revascularization procedures, or evidence of MI obtained from electrocardiogram. Self-reported levels of smoking were never, past, and current. Atrial fibrillation was defined from self-report or electrocardiogram evidence. Dyslipidemia was defined as selfreported use of lipid-lowering medication. Antihypertensive medication was defined by self-reported current use.
Figure 1. Exclusionary cascade. Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure.
Pulse Pressure and Stroke Table 1. Demographic and clinical characteristics of subjects at baseline examination according to SBP and DBP categories DBP < 80
80 ≤ DBP < 90
90 ≤ DBP
SBP < 120 Participants/events Age (years)
7,037/174
887/13
13/0
62.8 ± 9.4
61.1 ± 9.1
61.1 ± 11.1
Black
2,125 (30%)
339 (38%)
3 (23%)
Male
2,756 (39%)
395 (45%)
4 (31%)
Heart disease
1,016 (14%)
83 (9%)
1 (8%)
Dyslipidemia
2,154 (31%)
251 (28%)
3 (23%)
570 (8%)
69 (8%)
1 (8%)
Atrial fibrillation LVH
386 (5%)
61 (7%)
1 (8%)
Smoke (current)
1,046 (15%)
106 (12%)
9 (69%)
Diabetes
1,056 (15%)
113 (13%)
0 (0%)
120 ≤ SBP < 140 Participants/events
6,967/262
4,872/169
484/13
Age (years)
66.6 ± 9.2
63.1 ± 8.7
60.4 ± 8.6
Black
2,599 (37%)
2,183 (45%)
262 (54%)
Male
3,153 (45%)
2,388 (49%)
255 (53%)
Heart disease
1,273 (18%)
699 (14%)
56 (12%)
Dyslipidemia
2,501 (36%)
1,507 (31%)
130 (27%)
Atrial fibrillation
549 (8%)
362 (7%)
42 (9%)
LVH
647 (9%)
507 (10%)
60 (12%)
853 (12%)
654 (13%)
86 (18%)
1,627 (23%)
909 (19%)
78 (16%)
Smoke (current) Diabetes 140 ≤ SBP Participants/events
1,440/93
2,167/114
1,595/78
Age (years)
70.8 ± 8.5
Black
616 (43%)
66.8 ± 8.6
63.4 ± 8.8
1,069 (49%)
982 (62%)
Male
675(47%)
1,055 (49%)
797 (50%)
Heart disease
365 (25%)
430 (20%)
252 (16%)
Dyslipidemia
543 (38%)
749 (35%)
438 (27%)
Atrial fibrillation
165 (11%)
187 (8%)
126 (8%)
LVH
194 (13%)
305 (14%)
300 (19%)
Smoke (current)
190 (13%)
353 (16%)
298 (19%)
Diabetes
464 (32%)
602 (28%)
411 (26%)
Abbreviations: DBP = diastolic blood pressure; SBP = systolic blood pressure; Values = mean ± SD or percentage; LVH = left ventricular hypertension.
Statistical analysis
Baseline demographic and clinical characteristics of eligible participants were summarized within strata of SBP and DBP. Continuous variables were summarized by means and standard deviation, whereas categorical variables were summarized by proportions. Cox-proportional hazards models were used to examine the association of incident stroke with the different BP measurements (SBP, DBP, PP, and MAP) in a series of incremental models. Model 1 is the unadjusted hazard ratios (HRs),
whereas model 2 adjusts for age, race, age by race interaction (previously shown to be significant for stroke risk), gender, income, education, and geographic region. Model 3 adds adjustments for stroke risk factors including diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia. Model 4 adds adjustment for anti-hypertension medication use. Using HRs, we assessed if these relations were consistent within age strata, where the population was dichotomized at age 70 (age selected to have approximately the same number of stroke events in each strata), among black and white participants, and for ischemic and hemorrhagic stroke outcomes. Unadjusted American Journal of Hypertension 28(8) August 2015 989
Glasser et al.
HRs were also ascertained for categorical SBP (SBP < 120, 120 < SBP < 140, SBP > 140) and DBP (DBP < 80, 80 < DBP < 90, DBP > 90) classifications for all participants, older and younger participants, and black and white participants. Proportional hazards assumptions were tested by including the interaction of PP with log-transformed follow-up time in the multivariable Cox models. Multiple imputation was used with stroke outcomes to avoid potential biases introduced from differences in the characteristics of those where medical records could not be retrieved for adjudication, or where records were still in the adjudication process. As in other longitudinal cohort studies, the process of endpoint determination requires retrieval and physician adjudication of medical records. A corollary of this approach is a suspected event cannot be confirmed without successful medical record retrieval. However, there are factors of interest that affect the likelihood of being successful in retrieving records, for example, medical records are more difficult to retrieve following hospitalization of AA. We employed multiple imputation in order to remove the potential bias introduced by factors potentially confounded with record retrieval.12 Briefly, based on records where we were successful in retrieval, we developed a logistic regression model predicting the probability that a suspected event would be confirmed in adjudication. Clearly, the most powerful predictor in this model was the reason why the record was being sought, for example with a much higher likelihood of there truly being a stroke if the participant reported they were hospitalized for a stroke than when they simply reported having stroke symptoms that were not judged to be a stroke by a clinician. We then applied this prediction model to all suspected events where medical records were not successfully retrieved, and generated 10 data sets where events were generated using a binomial distribution with the modeled probabilities. These data sets were then analyzed using the approach of Rubin.13 Additional details of
the imputation process are available elsewhere.12 Cox regression models and descriptive statistics were calculated with SAS version 9.3 (SAS Institute, Cary, NC). RESULTS
Of the total REGARDS sample (n = 30,239), 1,930 reported a history of stroke and were excluded from the primary analysis. After further exclusion of informed consent errors (n = 56), missing follow-up data (n = 469), missing any covariate data (n = 2,258), and missing SBP and/or DBP (n = 64), the analytic sample was 25,462 men and women ages 45–98 (mean age 64.6 ± 9.4 years, 40.0% black, 45.1% male, Figure 1). The characteristics of the study sample are shown in Table 1, according to joint SBP and DBP classifications. As expected, age at the time of stroke was greater in each race and sex category (age at time of stroke was 71.6 ± 8.9 vs. 67.7 ± 8.6 years in white vs. black men; and, 72.2 ± 8.2 vs. 68.9 ± 9.6 years in white vs. black women). Current cigarette smoking was reported in 30% vs. 19% of black men when compared with without incident stroke, and this held for all race/sex comparisons although with lower percentages. In addition, only 13 participants had SBP < 120 and DBP > 90 mm Hg. Generally, participants with higher SBP were older and sicker, whereas participants with higher DBP were younger and healthier. Table 2 presents the HRs for stroke associated with levels of SBP, DBP, MAP, and PP (per 10 mm Hg). A total of 916 strokes over 160,211 person-years of follow-up revealed unadjusted associations between each BP measure (SBP, DBP, PP, and MAP) and incident stroke. The HR for incident stroke associated with PP held for all models. Further, PP remained significant after adjustment for MAP, but was substantially attenuated and was not statistically significant after adjustment for SBP, risk factors, and comorbid conditions (HR 0.98,
Table 2. Overall HRs of the various BP (per 10 mm Hg) parameters and incident stroke Model 1
Model 2
Model 3
Model 4
PP
1.30 (1.24–1.35)
1.14 (1.09–1.20)
1.12 (1.07–1.18)
1.11 (1.06–1.16)
MAP
1.24 (1.16–1.32)
1.19 (1.12–1.27)
1.19 (1.12–1.27)
1.17 (1.10–1.25)
SBP
1.22 (1.18–1.26)
1.14 (1.09–1.18)
1.13 (1.08–1.17)
1.11 (1.07–1.16)
DBP
1.09 (1.01–1.17)
1.14 (1.05–1.22)
1.15 (1.07–1.23)
1.13 (1.05–1.22)
Single pressure models
Two (pairwise) pressure models PP
1.27 (1.20–1.33)
1.10 (1.04–1.16)
1.07 (1.02–1.13)
1.07 (1.01–1.12)
MAP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
PP
1.21 (1.11–1.32)
1.01 (0.93–1.10)
0.98 (0.90–1.07)
0.98 (0.90–1.07)
SBP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
PP
1.29 (1.24–1.35)
1.14 (1.09–1.20)
1.12 (1.07–1.17)
1.11 (1.06–1.16)
DBP
1.07 (1.00–1.15)
1.13 (1.05–1.21)
1.14 (1.06–1.23)
1.13 (1.05–1.21)
Abbreviations: BP = blood pressure; DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. SBP and DBP in mm Hg. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
990 American Journal of Hypertension 28(8) August 2015
Pulse Pressure and Stroke Table 3. HRs for incident stroke associated with PP, MAP, SBP, and DBP (per 10 mm Hg) in older vs. younger participants Model 1
Model 2
Model 3
Model 4
Older participants Single pressure models PP
1.12 (1.05–1.19)
1.10 (1.03–1.17)
1.10 (1.03–1.17)
1.09 (1.03–1.17)
MAP
1.18 (1.09–1.29)
1.18 (1.08–1.29)
1.18 (1.08–1.29)
1.17 (1.07–1.28)
SBP
1.12 (1.07–1.18)
1.12 (1.06–1.17)
1.11 (1.06–1.17)
1.11 (1.05–1.17)
DBP
1.13 (1.02–1.25)
1.14 (1.03–1.26)
1.14 (1.03–1.26)
1.13 (1.02–1.25)
1.06 (0.98–1.13)
1.06 (0.98–1.13)
1.05 (0.98–1.13)
Two (pairwise) pressure models PP
1.07 (1.07–1.15)
MAP
1.13 (1.13–1.25)
1.13 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
PP
0.99 (0.99–1.11)
0.97 (0.86–1.09)
0.97 (0.86–1.09)
0.97 (0.86–1.09)
SBP
1.13 (1.13–1.25)
1.14 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
PP
1.12 (1.12–1.19)
1.10 (1.04–1.18)
1.10 (1.04–1.17)
1.10 (1.03–1.17)
DBP
1.13 (1.13–1.25)
1.14 (1.04–1.26)
1.14 (1.03–1.26)
1.13 (1.03–1.25)
1.20 (1.11–1.29)
1.14 (1.06–1.23)
1.12 (1.04–1.21)
Younger participants Single pressure models PP
1.32 (1.24–1.42)
MAP
1.30 (1.19–1.42)
1.20 (1.10–1.32)
1.19 (1.09–1.30)
1.16 (1.06–1.27)
SBP
1.24 (1.18–1.31)
1.16 (1.09–1.23)
1.13 (1.07–1.20)
1.11 (1.05–1.18)
DBP
1.19 (1.07–1.32)
1.14 (1.02–1.26)
1.15 (1.04–1.28)
1.13 (1.02–1.25)
Two (pairwise) pressure models PP
1.25 (1.15–1.36)
1.15 (1.05–1.25)
1.09 (1.00–1.18)
1.08 (0.99–1.17)
MAP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
PP
1.15 (1.01–1.32)
1.07 (0.94–1.22)
1.00 (0.88–1.14)
1.00 (0.88–1.14)
SBP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
PP
1.31 (1.22–1.40)
1.19 (1.10–1.28)
1.13 (1.05–1.22)
1.12 (1.03–1.20)
DBP
1.13 (1.02–1.26)
1.11 (1.00–1.23)
1.14 (1.03–1.26)
1.11 (1.00–1.23)
Abbreviations: DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
0.90–1.07), whereas SBP persisted as a significant predictor (HR 1.14, 1.06–1.23) in the presence of PP. There was no significant influence of self-reported antihypertensive drug use on these associations. In the overall analysis, the HRs for the various measures in older vs. younger and in blacks vs. whites was not appreciably different (Tables 3 and 4). We also assessed, in a series of incremental models, the results restricted to only normotensive participants, with PP as the exposure variable; and, the results were similar to the models with all participants. Also, we compared the imputed to the non-imputed data, and the most the HR changed was 0.05 (data not shown). DISCUSSION
Our analysis of the REGARDS data suggests that PP, SBP, DBP, and MAP are positively associated with stroke, but
counter to our hypothesis, after multivariable adjustment for SBP, the association of PP is attenuated and becomes statistically insignificant. Cardiovascular mortality and its various clinical manifestations such as angina, MI, stroke, and cardiac failure are increased in the presence of elevated BP. Traditionally, this has been measured by sphygmomanometry, which yields 2 measures: SBP and DBP. It is increasingly argued that neither SBP nor DBP may be as strong a predictor of a stroke as PP. Thus, PP may be a key BP measure, particularly in older individuals, and may increase in importance as a risk factor for cardiovascular disease, including stroke, MI, and death. However, arguments persist as to whether PP is an independent measure (particularly from SBP) of stroke risk.1 The Framingham study reported on its longitudinal follow-up of persons over 50 years of age, finding that cardiovascular (and in particular coronary) American Journal of Hypertension 28(8) August 2015 991
Glasser et al. Table 4. HRs of the association of incident stroke and PP (per 10 mm Hg) in black vs. white participants Model 1
Model 2
Model 3
Model 4
PP
1.26 (1.17–1.34)
1.15 (1.07–1.24)
1.13 (1.05–1.22)
1.12 (1.03–1.20)
MAP
1.24 (1.14–1.35)
1.23 (1.13–1.34)
1.22 (1.12-1.33)
1.20 (1.10–1.31)
SBP
1.20 (1.14–1.26)
1.15 (1.09–1.22)
1.14 (1.08–1.20)
1.13 (1.06–1.19)
DBP
1.14 (1.02–1.26)
1.19 (1.08–1.32)
1.19 (1.08–1.32)
1.18 (1.06–1.30)
1.21 (1.11–1.31)
1.08 (0.99–1.18)
1.06 (0.97–1.15)
1.05 (0.97–1.15)
MAP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
PP
1.13 (0.99–1.28)
0.97 (0.85–1.11)
0.95 (0.83–1.08)
0.95 (0.84–1.08)
SBP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
PP
1.25 (1.17–1.34)
1.14 (1.06–1.23)
1.12 (1.04–1.20)
1.11 (1.03–1.19)
DBP
1.11 (1.00–1.22)
1.17 (1.06–1.29)
1.18 (1.07–1.30)
1.16 (1.05–1.28)
Black participants Single pressure models
Two (pairwise) pressure models PP
White participants Single pressure models PP
1.32 (1.25–1.40)
1.14 (1.07–1.21)
1.12 (1.05–1.20)
1.11 (1.04–1.18)
MAP
1.21 (1.11–1.33)
1.15 (1.05–1.26)
1.16 (1.06–1.27)
1.14 (1.04–1.25)
SBP
1.23 (1.17–1.30)
1.12 (1.06–1.18)
1.12 (1.06–1.18)
1.10 (1.04–1.16)
DBP
1.02 (0.92–1.13)
1.08 (0.97–1.20)
1.10 (0.99–1.22)
1.08 (0.98–1.20)
Two (pairwise) pressure models PP
1.31 (1.23–1.40)
1.11 (1.03–1.19)
1.08 (1.01–1.17)
1.08 (1.00–1.16)
MAP
1.03 (0.93–1.13)
1.09 (0.93–1.20)
1.10 (1.00–1.22)
1.09 (0.98–1.21)
PP
1.29 (1.15–1.45)
1.05 (0.93–1.18)
1.02 (0.90–1.15)
1.02 (0.90–1.15)
SBP
1.03 (0.93–1.13)
1.09 (0.98–1.20)
1.10 (1.00–1.22)
1.09 (0.98–1.21)
PP
1.32 (1.25–1.40)
1.14 (1.07–1.21)
1.12 (1.05–1.20)
1.11 (1.04–1.18)
DBP
1.03 (0.93–1.13)
1.09 (0.98–1.20)
1.10 (1.06–1.22)
1.09 (0.98–1.21)
Abbreviations: DBP = diastolic blood pressure; HR = hazard ratio; MAP = mean arterial pressure; PP = pulse pressure; SBP = systolic blood pressure. Model 1: Crude model. Model 2: Adjusting for age, race, age*race, gender, region, income category, and education level. Model 3: Model 2 and additional adjustment for diabetes, heart disease, smoking, atrial fibrillation, and dyslipidemia.
mortality is associated with increased PP.5 In that study, mortality was related independently with initial systolic and diastolic pressure, but the strongest association was with PP, and when systolic pressure was initially considered, there was a negative association with diastolic pressure. In other words, for a given systolic pressure, lower diastolic pressure was associated with greater mortality. We have reported the association of PP and incident CHD from the same REGARDS population and found that although PP was positively associated with CHD, this association was attenuated after multivariable adjustment that included SBP.14 As Millar and Lever15 pointed out, the superiority of PP as a cardiac risk predictor in hypertension is supported by “3 strands of evidence”: PP is a risk factor for coronary events (MI, angina, heart failure, and cardiac death); PP is strongly associated with, and a potential determinant of 992 American Journal of Hypertension 28(8) August 2015
several surrogate markers of cardiac risk such as echocardiographically determined left atrial size and ventricular mass,16 electrocardiographic indices of ischemia and cardiac size,17 and carotid wall thickness;18,19 and, there is a physiologically plausible mechanism linking increased conduit artery stiffness to cardiac risk via raised PP. That is, increased arterial wall stiffness (itself a well-recognized index of cardiovascular mortality)20 causes increased systolic BP during systole, increased pulse wave velocity, and early return of the reflected pressure wave during late systole rather than during diastole. These effects increase left ventricular work and oxygen requirements simultaneously and tend to diminish coronary perfusion.21,22 On the other hand, cerebral blood flow occurs throughout the cardiac cycle and the relationship between PP and stroke is weak.14 Others have reported that high PP reflecting large artery stiffness is a significant independent risk factor for
Pulse Pressure and Stroke
cardiovascular, especially coronary, mortality in different populations. Madhavan et al. 23 reported that untreated hypertensive subjects with a PP of 63 mm Hg had an increased risk of cardiovascular complications. In addition, they found that these subjects were at greater risk of MI when there was too great a fall in DBP after treatment.5 Franklin et al. also reported that in a larger population of treated and untreated hypertensive subjects, PP was the only BP measurement independently related to the treatment incidence of MI.20 The link between PP and cardiovascular complications has also been shown in subjects who had MI with left ventricular dysfunction.1 Our findings did not confirm those previous studies that support our hypothesis of an association between PP and stroke risk that was independent of SBP. In our study, the correlation of PP and SBP was 0.82, making it mathematically challenging to establish an independent effect. Our data does support the previous studies that did not demonstrate an association of PP (independent of SBP) and incident stroke.1,24,25 We postulate that the main reason for the difference in the association of stroke with CHD is the strong association of stroke with SBP, mitigating the effects of other BP variables. In addition, the pathophysiology of stroke and CHD is not identical. Our study has several limitations worth noting. Some non-laboratory risk factors were based on self-report (although this is common to many epidemiologic studies), and individuals without telephones were necessarily excluded from selection into the study population. These excluded individuals who may be of lower socioeconomic status, and therefore, may have different risk factor profiles than those included in this analysis. In conclusion, in the REGARDS population, the role of SBP as the dominant index of BP as a predictor of stroke risk is supported. Although PP does predict stroke risk, its association is not as strong and is not independent of the effect of SBP. These data potentially suggest that efforts to evaluate and manage stroke risk through BP control should focus predominantly on SBP. This is in contrast to our prior report that did demonstrate and association of PP and CHD.6
ACKNOWLEDGMENTS
This research project is supported by a cooperative agreement from the National Institute of Neurological Disorders and Stroke (U01 NS041588), National Institutes of Health, and Department of Health and Human Service. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health. Representatives of the funding agency have been involved in the review of the manuscript but not directly involved in the collection, management, analysis, or interpretation of the data. We thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at http://www.regardsstudy.org.
DISCLOSURE
The authors declared no conflict of interest.
REFERENCES 1. Benetos A, Safar M, Rudnichi A, Smulyan H, Richard JL, Ducimetieère P, Guize L. Pulse pressure: a predictor of long-term cardiovascular mortality in a French male population. Hypertension 1997; 30:1410–1415. 2. Brown DW, Giles WH, Greenlund KJ. Blood pressure parameters and risk of fatal stroke, NHANES II mortality study. Am J Hypertens 2007; 20:338–341. 3. Geeganage C, Sare G, Bath PM. Pulse pressure as a predictor of stroke. Expert Rev Neurother 2008; 8:165–167. 4. Verdecchia P, Schillaci G, Reboldi G, Franklin SS, Porcellati C. Different prognostic impact of 24-hour mean blood pressure and pulse pressure on stroke and coronary artery disease in essential hypertension. Circulation 2001; 103:2579–2584. 5. Franklin SS, Khan SA, Wong ND, Larson MG, Levy D. Is pulse pressure useful in predicting risk for coronary heart Disease? The Framingham heart study. Circulation 1999; 100:354–360. 6. Okada K, Iso H, Cui R, Inoue M, Tsugane S. Pulse pressure is an independent risk factor for stroke among middle-aged Japanese with normal systolic blood pressure: the JPHC study. J Hypertens 2011; 29:319–321. 7. Howard VJ, Cushman M, Pulley L, Gomez CR, Go RC, Prineas RJ, Graham A, Moy CS, Howard G. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology 2005; 25:135–143. 8. MESA Coordinating Center, University of Washington. MESA Exam 1 Participation Rate. http://www.mesa-nhlbi.org/participation.aspx 2006. Accessed 25 June 2013. 9. U.S. Department of Health and Human Services, NIoH. National Heart Lung and Blood Institute. http://www.nhlbi.nih.gov/index.htm Accessed 5 June 2014. 10. Anonymous. Stroke—1989. Recommendations on stroke prevention, diagnosis, and therapy. Report of the WHO Task Force on Stroke and other Cerebrovascular Disorders. Stroke 1989; 20:1407–1431. 11. Luepker RV, Apple FS, Christenson RH, Crow RS, Fortmann SP, Goff D, Goldberg RJ, Hand MM, Jaffe AS, Julian DG, Levy D, Manolio T, Mendis S, Mensah G, Pajak A, Prineas RJ, Reddy KS, Roger VL, Rosamond WD, Shahar E, Sharrett AR, Sorlie P, Tunstall-Pedoe H; 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; 108:2543–2549. 12. Howard G, McClure LA, Moy CS, Safford MM, Cushman M, Judd SE, et al. Imputation of incident events in longitudinal cohort studies. Am J Epidemiol 2011; 174:718–726. 13. Rubin DB. Multiple Imputation for Nonresponse in Surveys. J. Wiley & Sons: New York, 1987. 14. Halberg DL, Sands C, Le A, Howard VJ, Safford M GS, Muntner P. Pulse and mean arterial pressure as predictors of stroke in the REGARDS study [abstract]. Stroke 2012; 43:A2593. 15. Millar JA, Lever AF. Implications of pulse pressure as a predictor of cardiac risk in patients with hypertension. Hypertension 2000; 36:907–911. 16. Gardin JM, Arnold A, Gottdiener JS, Wong ND, Fried LP, Klopfenstein HS, O’Leary DH, Tracy R, Kronmal R. Left ventricular mass in the elderly. The Cardiovascular Health Study. Hypertension 1997; 29:1095–1103. 17. Darne B, Girerd X, Safar M, Cambien F, Guize L. Pulsatile versus steady component of blood pressure: a cross-sectional analysis and a prospective analysis on cardiovascular mortality. Hypertension 1989; 13:392–400.
American Journal of Hypertension 28(8) August 2015 993
Glasser et al. 18. Mitchell GF. Pulse pressure, arterial compliance and cardiovascular morbidity and mortality. Curr Opin Nephrol Hypertens 1999; 8:335–342. 19. Zanchetti A, Bond MG, Hennig M, Neiss A, Mancia G, Dal Palù C, Hansson L, Magnani B, Rahn KH, Reid J, Rodicio J, Safar M, Eckes L, Ravinetto R. Risk factors associated with alterations in carotid intimamedia thickness in hypertension: baseline data from the European Lacidipine Study on Atherosclerosis. J Hypertens 1998; 16:949–961. 20. Blacher J, Pannier B, Guerin AP, Marchais SJ, Safar ME, London GM. Carotid arterial stiffness as a predictor of cardiovascular and all-cause mortality in end-stage renal disease. Hypertension 1998; 32:570–574. 21. Franklin SS. Cardiovascular risks related to increased diastolic, systolic and pulse pressure. An epidemiologist’s point of view. Pathol Biol (Paris) 1999; 47:594–603.
994 American Journal of Hypertension 28(8) August 2015
22. Safar ME, Siche JP, Mallion JM, London GM. Arterial mechan ics predict cardiovascular risk in hypertension. J Hypertens 1997; 15:1605–1611. 23. Madhavan S, Ooi WL, Cohen H, Alderman MH. Relation of pulse pressure and blood pressure reduction to the incidence of myocardial infarction. Hypertension 1994; 23:395–401. 24. Domanski M, Davis B, Pfeffer M, Kastantin M, Mitchell GF. Isolated systolic hypertension: prognostic information provided by pulse pressure. Hypertension 1999; 34:375–380. 25. Domanski M, Norman J, Wolz M, Mitchell G, Pfeffer M. Cardiovascular risk assessment using pulse pressure in the first national health and nutrition examination survey (NHANES I). Hypertension 2001; 38:793–797.
