Relationship of Blood Pressure to Cardiovascular Death: The Effects of Pulse Pressure in the Elderly

Transcription

1 Relationship of Blood Pressure to Cardiovascular Death: The Effects of Pulse Pressure in the Elderly MEI-LING T. LEE, PhD, BERNARD A. ROSNER, PhD, AND SCOTT T. WEISS, MD PURPOSE: To investigate the relationship of systolic and diastolic blood pressure to fatal myocardial infarction, fatal stroke and other death related to cardiovascular diseases (CVD). METHODS: The study was based on a prospective longitudinal study conducted by the Veterans Administration at the Boston Outpatient Clinic. Participants are male volunteers from the greater Boston area. Main outcome measures are fatal myocardial infarction, fatal stroke and other deaths related to cardiovascular diseases. The method of pooled logistic regression was used for statistical analysis. RESULTS: For younger men (age 21 59), after adjusting for effects of other risk factors, when systolic and diastolic blood pressure were considered separately, SBP was predictive of cardiovascular death (SBP: RR 1.23; 95% CI (1.05, 1.45) per 10 mmhg of increase), and DBP showed a nonsignificant positive trend in relation to cardiovascular death (DBP: RR 1.27; 95% CI (0.95, 1.69) per 10 mmhg of increase). For older men (age 60 85), when SBP and DBP were considered separately, SBP (RR 1.26; 95% CI (1.02, 1.55) per 15 mmhg of increase) was directly related, but DBP (RR 1.05; 95% CI (0.83, 1.32) per 8 mmhg of increase) was not related to cardiovascular death. However, for the elderly group, when SBP and DBP were considered jointly in the regression model, then the regression coefficient of DBP ( 0.018, p 0.30) was of approximately the same absolute magnitude as that of SBP ( 0.021, p 0.02) but opposite in sign. For younger men, when SBP and DBP were considered jointly, SBP ( 0.021, p 0.049) but not DBP ( 0.001, p 0.953) was positively related to cardiovascular death. CONCLUSIONS: We found that, for the elderly, pulse pressure (SBP-DBP) may be a more accurate predictor of cardiovascular death than either SBP or DBP alone. The relative risk per 35 mmhg of increase of pulse pressure, which equals the approximate interval from the 10th to the 90th percentile in the elderly group, is 2.1 with 95% CI (1.1, 3.8). In younger subjects, SBP, but not DBP, is an independent predictor of fatal CVD. Ann Epidemiol 1999;9: Elsevier Science Inc. All rights reserved. KEY WORDS: Logistic Regression, Longitudinal Data, Myocardial Infarction, Stroke. decade of age. This phenomenon of decreasing DBP for the elderly also has been reported by the Manitoba follow-up study (10) and the Framingham study (11). Therefore, pulse pressure, defined as SBP minus DBP, increases slowly up to age 60 but then rises rapidly with increasing age. In this paper, we discuss the use of pulse pressure, in comparison to SBP or DBP alone, in predicting fatal myocar- dial infarction, fatal stroke and other deaths related to CVD in the elderly. Other risk factors considered include age, body mass index (BMI Weight/Height 2 ), cholesterol level (mg/dl), diabetes, prior MI, prior stroke, and smoking status. The study cohort of the NAS consists of initially healthy men with age range initially between 21 to 80 in Risk factors were recorded at each visit from 1963 to INTRODUCTION Abnormal blood pressure has long been recognized as a major risk factor for the development of coronary artery disease, stroke, and other cardiovascular disease (CVD) outcomes in adults (1, 2). Isolated systolic hypertension is recognized as an important predictor of adverse cardiovascular events in the elderly (3 6). Isolated systolic hypertension is known to predict myocardial infarction (MI), stroke, and death from congestive heart failure (7, 8). In a previous analysis (9) of the Normative Aging Study (NAS) data, we noticed that while the mean systolic blood pressure (SBP) increased with age for all age groups, the mean diastolic blood pressure (DBP) increased with age only prior to age 57.5; after age 57.5 the DBP decreased by 1 2 mmhg per From the Channing Laboratory, Brigham and Women s Hospital and METHODS Harvard Medical School, Boston, Massachusetts, USA. Address reprint requests to: Dr. Mei-Ling T. Lee, Channing Laboratory/ Study Population BWH, Harvard Medical School, 181 Longwood Avenue, Boston, MA Received December 18, 1997; revised April 14, 1998; accepted May The NAS (2) is a longitudinal study of health and aging 19, established by the Veterans Administration (VA) at the 1999 Elsevier Science Inc. All rights reserved /99/$ see front matter 655 Avenue of the Americas, New York, NY PII S (98)

2 102 Lee et al. AEP Vol. 9, No. 2 RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD February 1999: Selected Abbreviations and Acronyms the hospital chart and the death certificate together. If BMI body mass index necessary the NAS record is reviewed and/or the primary CI confidence interval physician is contacted to resolve any problem with missing CVD cardiovascular diseases or incomplete data. Codings of these fatal endpoints are DBP diastolic blood pressure based on the International Classification of Diseases (9th ICD9CM International Classification of Diseases (9th revision) Revision) Clinical Modification (ICD9CM) such that Clinical Modification MI myocardial infarction cause-of-death (either primary, secondary, or other causes) NAS Normative Aging Study with code ranging from 410 to denotes acute myocar- RR relative risk dial infarction, denotes stroke, , 411 SBP systolic blood pressure 429.9, , and 795 denotes other cardiovascular re- VA Veterans Administration lated deaths. Risk factors considered initially include body mass index (BMI), SBP and DBP, age at visit, cholesterol level, smoking status, prior MI, prior stroke, and current diagnosis of diabetes Boston Outpatient Clinic in Community residents mellitus. Because BMI, SBP, DBP, smoking status, and from around the greater Boston area were contacted and cholesterol levels were used as covariates in the logistic volunteered to be screened according to specific health criteria, regression models, observations with missing information and those with chronic health conditions including on any of these covariates were excluded from the analysis asthma, chronic bronchitis, obesity, diabetes mellitus, and data set. Accordingly, from 1963 to 1992, a total of 4842 hypertension were excluded. The study cohort initially in- visits from 1828 subjects with visit age less than 60, and a cluded 2280 community-dwelling male volunteers from the total of 1772 visits from 897 subjects with visit age older greater Boston area who were years of age at en- than 60 were included in the observational study. Due to rollment. the low incidence of diabetes in this population (less than 1.5%), diabetes was excluded as a risk factor in the regression Eligibility models. Only initially healthy individuals who met screening criteria based on current medical status and medical history to investigate Statistical Methods nonpathologic aging were enrolled. Blood pressure Since the data set contains repeated measurements such as was measured by an examining physician using a standard BMI and cholesterol level, we consider the method of pooled mercury sphygmomanometer while the subject was seated. logistic regression (12) for our analysis. Treating each exami- Systolic blood pressure and (SBP) and fifth-phase diastolic nation interval as a mini follow-up study, this method pools blood pressure (DBP) were measured to the nearest 2 mmhg, repeated observations of all time intervals. and the average of right and left arm values was used in The pooled person-visit data from the NAS were assemthese analyses. All physicians are trained annually in the bled as follows. Assume that subject i has v recorded visits. blood pressure measurement protocol. It is the same protocol Let Y i,j denote the indicator variable for fatal CVD event that has been used for the 30 years of the study. These subjects for subject i at visit j. Hence, Y i, j 0 for j 1,...,v. If were examined every 5 years before age 52, and every 3 years either fatal MI, fatal stroke or any other cardiovascular after age 52. The examination of subjects continues at present. related death is recorded after the last visit, then we define The longitudinal data were not equally spaced due to design; Y i,v 1 1; otherwise, Y i,v 1 0. Corresponding to the outalso subjects did not always return at the date of their come variable Y i,j is the covariate vector X ij (x ij1,...,x ijp ) t, scheduled visits according to the protocol. Thus the data denoting p risk factors observed at the j th visit. Given risk were not equally spaced even within the 52 and 52 factors X ij at the j th visit, j 1,...,v, let p i,j 1 Prob(Y i,j 1 age range, respectively. To eliminate possible confounding 1 X ij ) denote the conditional probability that subject i has effects of medication use, we restrict our attention to the a fatal MI, fatal stroke or other cardiovascular deaths after subcohort of those subjects who have never been treated with visit j. anti-hypertensive medication during the course of the study. For the logistic regression model, define, for j 1,...,v, Measurements The endpoints of interest are fatal MI, fatal stroke, and other deaths related to CVD. All death certificates and hospital charts are reviewed by a specially trained research nurse supervised by a physician to accurately determine the cause of death. NAS coded cause of death is not simply based on the death certificate but on a complete review of logit p i,j 1 log p i,j 1 1 p i,j 1 0 r k 1 k x ijk To investigate the relationship of blood pressure measurements with fatal cardiovascular outcomes, we ran a set of four logistic regression models. In these four regression models we included (a) SBP alone, (b) DBP alone, (c) both SBP and DBP, and (d) pulse pressure as a representation of blood

3 AEP Vol. 9, No. 2 February 1999: Lee et al. RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD 103 TABLE 1. Summary of frequencies and percentages of former and current smoking status by age Summary statistics of risk factors mean (standard deviation) at current visit a Summary of smoking status observed frequencies [row percent] Current Current Pulse Former smoker smoker Age at visit N b SBP DBP pressure BMI Cholesterol smoker 1 24/day 25 /day (10.22) (7.41) (9.18) (2.98) (43.07) [18.3%] [38.2%] [22.1%] (11.22) (8.07) (8.53) (2.88) (46.03) [31.9%] [27.2%] [19.9%] (11.87) (7.77) (8.65) (2.99) (46.48) [41.0%] [19.5%] [13.8%] (14.49) (8.57) (10.70) (3.20) (45.08) [48.3%] [12.1%] [10.8%] (16.03) (8.60) (12.47) (3.28) (43.54) [57.1%] [7.5%] [4.8%] (17.99) (8.83) (15.24) (3.16) (40.32) [57.3%] [5.2%] [1.8%] (19.02) (10.36) (15.89) (3.29) (39.47) [42.4%] [1.5%] [0.0%] a Excluding subjects treated with anti-hypertensive medication. b Total number of person visits: 1828 subjects have visits at age 60; 897 subjects have visits at age 60. pressure status, respectively. For each model, we included of cardiovascular related diseases at age less than 60 (36 the following covariates: BMI, total cholesterol level, prior died of MI, 3 of stroke, 18 of asymptomatic ischemic heart MI, prior stroke, a dichotomous variable for former smoking, disease with hypertensive disease, and 14 of other cardiovastwo dichotomous variables indicating current smokers num- cular disease). ber of cigarettes per day (1 24 cigarettes per day, 25 Table 1 gives summary statistics of the means and stancigarettes per day, respectively), and indicator variables of dard deviations for SBP and DBP, pulse pressure, BMI, and age at visit in 5-year groups. cholesterol level by age at current visit. For this table, the Since it was apparent in a preliminary analysis that the same subject may be included in different age intervals. It relationship of blood pressure to CVD death was different can be seen that the relationship between both mean SBP for subjects with age less than 60 years old than for subjects and DBP versus age is not linear. The standard deviation who are 60 years and older at a given visit, to control for of both SBP and DBP tend to rise with age. A summary of the effect of aging, separate analyses were presented for frequencies and percentages of former and current smoking these two age groups. status by age is also shown in Table 1. Over 50% of the men had smoked at some time in their lives, but only 20% of year-old men were current smokers with the percent RESULTS of current smoking declining sharply with age. Demographic Data For the younger people, over 10% of them smoked 25 or more cigarettes per day. Less than 5% of the elderly (age Of the selected study participants, the age distribution at 60 and up) smoked 25 or more cigarettes per day. their baseline visits has a range from 20 to 80. The mean Percentile distributions of SBP, DBP, and pulse pressure actual age at each subject s initial visit is 43 years old with by age group are presented in Table 2. The standard deviaa standard deviation of When the study began, 77% tion of SBP, DBP, and pulse pressure are approximately 15 of subjects were years old, 17% were years old, mmhg, 8 mmhg, and 10 mmhg, respectively. Furthermore, and 6% were Except for one person who had eight the approximate range from the 10th to the 90th percentile visits, subjects have up to seven visits. Over 80% of subjects of pulse pressure in the elderly group is 35 mmhg. have two or more visits. At the baseline visit, 33% of the subjects were former smokers and 40% of the subjects were current smokers. 64 of the 897 (7.13%) subjects with age Regression Results 60 and up at visits died of cardiovascular related diseases Tables 3 and 4 present pooled logistic regression results for (29 died of MI, 5 of stroke, and 30 of other cardiovascular subjects with age less than 60 and for subjects with age 60 disease including vascular heart disease, pericardial disease, and older, respectively. cardiomyopathy and pulmonary heart disease). For subjects Risk factors including BMI, total cholesterol, age group, younger than 60 at any visit, 71 of the 1828 (3.88%) died prior MI, and smoking status, are included in each of Models

4 104 Lee et al. AEP Vol. 9, No. 2 RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD February 1999: TABLE 2. Percentiles of SBP, DBP, and pulse pressure; Normative Aging Study: ; excluding subjects on anti-hypertensive medication SBP SBP SBP SBP SBP SBP SBP SBP SBP Age N Q1 Q5 Q10 Q25 Q50 Q75 Q90 Q95 Q DBP DBP DBP DBP DBP DBP DBP DBP DBP Q1 Q5 Q10 Q25 Q50 Q75 Q90 Q95 Q PulseP PulseP PulseP PulseP PulseP PulseP PulseP PulseP PulseP Q1 Q5 Q10 Q25 Q50 Q75 Q90 Q95 Q In addition to these risk factors, Model 1 includes SBP, the same absolute magnitude with SBP being the dominant Model 2 includes DBP, Model 3 includes both SBP and DBP, predictor of CVD death. On the other hand, unlike the elderly and Model 4 includes the pulse pressure as representation of group, total cholesterol and indicator variables for current blood pressure status, respectively. For the elderly, prior smoking status are significant for predicting fatal CVD for stroke is also included in Models 1 4 as a risk factor. subjects with visit age less than 60 years old. For the elderly, when SBP and DBP are considered separately, the regression coefficients are (p 0.033) for SBP in Model 1, and (p 0.689) DISCUSSION for DBP in Model 2, respectively. When SBP and DBP were both included in Model 3, the regression coefficients are We utilized the method of pooled logistic regression to (p 0.019) for SBP, and analyze longitudinal data concerning the relationship of (p 0.303) for DBP. Notice that, in this case, the regression SBP and DBP and other cardiovascular risk factors to fatal coefficients for SBP and DBP are of approximately the same CVD in the NAS. We found that there was a significant absolute magnitude but opposite in sign. When pulse pressure effect of pulse pressure in predicting fatal cardiovascular (SBP-DBP) was used in Model 4, it was significant with regrespersons outcomes in the elderly (age 60 and up). Conversely, for sion coefficient (p 0.017). Table 4 shows under age 60, only systolic blood pressure was a that, for the elderly, replacing the two variables SBP and DBP significant predictor of fatal CVD. (Model 3) by a single pulse pressure variable (Model 4) not In a previous analysis (9) of the NAS data set we noticed only provided about the same log likelihood, but reduced the that the relationship between both mean SBP and DBP, degrees of freedom by one. Hence we conclude that, for when plotted against age, was not linear. Specifically, mean subjects age 60 and older, pulse pressure has significant SBP increased by 0 1 mmhg per decade prior to age 47.5, effects in predicting fatal CVD, after adjusting for other and 2 7 mmhg per decade after 47.5; mean DBP increased possible confounders. In addition, prior MI and prior stroke by 1 2 mmhg per decade prior to age Subjects entering are significant predictors of fatal CVD for the elderly. the study at age showed slight increases in mean Table 3 presents regression results based on subjects less DBP after age 57.5, while subjects entering the study after than 60 years old. Although pulse pressure is significant in age 52.5 showed decreases of 1 2 mmhg per decade. Similar Model 4, when SBP and DBP were considered jointly in Model results on the decrease of DBP in elderly men have been 3, the regression coefficient of DBP and SBP were not of reported in the Manitoba Study (10) for 3983 Canadian

5 AEP Vol. 9, No. 2 February 1999: Lee et al. RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD 105 TABLE 3. Logistic regression analysis: Fatal CVD on selected cardiovascular risk factors; from 1828 subjects less than 60 years old Coefficient Model 1A: Model 2A: Model 3A: with Model 4A: pulse (p-value) with SBP only with DBP only both SBP & DBP pressure (BP-DBP) Intercept (0.000) (0.000) (0.000) (0.000) SBP (0.010) (0.049) DBP (0.109) (0.953) SBP-DBP (0.032) BMI (0.945) (0.994) (0.953) (0.808) Total cholesterol (0.009) (0.009) (0.009) (0.007) Current smoker /day (0.015) (0.014) (0.016) (0.023) Current smoker /day (0.001) (0.000) (0.001) (0.001) Former smoker (0.788) (0.770) (0.790) (0.828) Previous nonfatal MI (0.056) (0.053) (0.056) (0.060) Age (0.737) (0.755) (0.736) (0.716) Age (0.594) (0.641) (0.592) (0.560) Age (0.169) (0.182) (0.169) (0.154) Age (0.044) (0.046) (0.044) (0.038) Age (0.187) (0.157) (0.187) (0.162) 2logL d.f Reference age group: age 30 34; reference smoking group: non-smoking. Subjects never on anti-hypertensive medication; 71 events out of 4842 person visits. the analyses. (For the elderly, the relative risk per 35 mmhg increase in pulse pressure was 2.0 with 95% CI (1.2, 3.3)). The reasons for a reduction in DBP after age 60 have been attributed to a variety of different causes. One hypothesis is that an age-related decrease in cardiac output is the cause for the fall in DBP of age greater than 60, but this is inconsis- tent with the rise of SBP with age greater than 60. Franklin and colleagues (14) suggest that, after the sixth decade of life, increasing pulse pressure and decreasing DBP are surrogate measurements for large artery stiffness. The most likely explanation for the fall of DBP after age 60 is decreased compliance of large arteries (15 17). It is worth noting the mechanism by which stiffening of the proximal aorta may contribute to cardiac disease. Such stiffening will result in an increase in systolic blood pressure which in turn will result in increased left ventricular afterload and resultant left ventricular hypertrophy. In addition, the stiffening will lower diastolic blood pressure, re- men over a 40-year period between 1948 and 1988, and in the Framingham Heart Study (3) for 2036 subjects (890 men and 1146 women) in a population-based cohort. For subjects in the Manitoba study, DBP fell after age 55. In the Framingham study, a similar fall of DBP was noted after age 60. Thus for the elderly group the pulse pressure tended to increase with age. Our findings that pulse pressure, rather than systolic or diastolic pressure alone, was the best predictor of adverse cardiovascular events in the elderly is consistent with a known correlation of SBP with DBP (13) and, also, with the concept that large artery elasticity is the primary manifestation of a generalized arthrosclerotic process. To avoid the possible confounding effects of medication use, we eliminated all subjects who ever were on antihypertensive medications. However, we repeated the analysis by including subjects who received anti-hypertensive medication during the course of the study and obtained similar results for all

6 106 Lee et al. AEP Vol. 9, No. 2 RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD February 1999: TABLE 4. Logistic regression analysis: Fatal CVD on selected cardiovascular risk factors; from 897 subjects 60 years of age Model 3: Model 4: Coefficient Model 1: Model 2: with both pulse pressure (p-value) with SBP only with DBP only SBP & DBP (SBP-DBP) Intercept (0.001) (0.016) (0.005) (0.001) SBP (0.033) (0.019) DBP (0.689) (0.303) SBP-DBP (0.017) BMI (0.178) (0.209) (0.242) (0.247) Total cholesterol (0.113) (0.091) (0.110) (0.108) Current smoker /day (0.684) (0.723) (0.747) (0.757) Current smoker /day (0.224) (0.230) (0.260) (0.264) Former smoker (0.828) (0.722) (0.869) (0.871) Previous nonfatal MI (0.024) (0.025) (0.025) (0.026) Previous nonfatal stroke (0.030) (0.046) (0.029) (0.030) Age (0.464) (0.560) (0.433) (0.432) Age (0.599) (0.415) (0.694) (0.699) Age (0.248) (0.101) (0.347) (0.351) Age 80 and up (0.009) (0.004) (0.020) (0.020) 2log likelihood d.f Reference age group: age 60 64; reference smoking group: non-smoking. Subjects never on anti-hypertensive medication; 64 events out of 1772 person visits. sulting in decreased coronary artery driving pressure and decreased perfusion. A widening pulse pressure may be more predictive of cardiac risk than either SBP or DBP alone as both will increase cardiac work and the risk of heart failure and myocardial infarction. Benetos and colleagues (18) have confirmed a relationship of widening pulse pressure to increased cardiovascular mortality in 19,083 men years old followed for an average of 20 years. Similar results have been reported previously by Madhavan and colleagues (19), and by the SAVE investigators (20). One of the strengths of our investigation is that it is longitudinal and considers some of the relevant covariates influencing cardiovascular risk. An additional strength is the length of follow-up and the ability to track age-related blood pressure trends over 30 years of time. There are two potential limitations to the current investigation. First, our population is predominantly Caucasian and thus, we are unable to make any comments concerning African-American or other ethnic groups. The second limitation is that the NAS is solely male and, hence, we have no information on female trends in blood pressure and their influence on cardiovascular risk. These data do support the conclusion that, in elderly subjects age 60 or above, pulse pressure is a more accurate predictor than either systolic or diastolic pressure alone in predicting cardiovascular deaths. However, in younger subjects with age less than 60, SBP was a more important predictor of fatal CVD then DBP or pulse pressure. The research is supported in part by grant HL of the National Institutes of Health. The Normative Aging Study is sponsored by the Health Services Research and Development Service of the Department of Veterans Affairs. We thank Dr. Stanley Franklin for helpful comments and Soma Datta for programming assistance.

7 AEP Vol. 9, No. 2 February 1999: Lee et al. RELATIONSHIP OF PULSE PRESSURE TO FATAL CVD Franklin SS, Gustin W, Wong ND, Larson MG, Weber MA, Kannel REFERENCES WB, et al. Hemodynamic patterns of age-related changes in blood Joint National Committee. The 1988 Report of the Joint Na- pressure: The Framingham Heart Study. Circulation. 1997;96: tional Committee on detection, evaluation and treatment of high blood pressure. Arch Intern Med. 1988;148: D Agostino RB, Lee M-L, Belanger AJ, Cupples LA, Anderson K, 2. Cassano P, Segal M, Vokonas P, Scott W. Body fat distribution, blood Kannel WB. Relation of pooled logistic regression to time dependent pressure, and hypertension: A prospective cohort study of men in the Cox regression analysis: The Framingham Heart Study. Statist Med. normative aging study. Ann Epidemiol. 1990;1: ;(9): Kannel WB, Gordon P, Schwartz MJ. Systolic versus diastolic blood 13. Kannel WB, Wolfe PA, McGee DL, Dawber TR, McNamara P, Castelli pressure in risk of coronary heart disease. The Framingham Study. WP. Systolic blood pressure, arterial rigidity, and risk of stroke. The Amer J Cardial. 1971;27: Framingham Study. JAMA. 1981;245: Gifford RW Jr. Are systolic/diastolic and isolated systolic hypertension 14. Franklin SS, Sutton-Tyrrel K, Belle SH, Weber MA, Kuller LH. The risk factors for the elderly? Geriatrics. 1989;44: importance of pulsatile components of hypertension in predicting ca- rotid stenosis in older adults. J Hypert. 1997;15: Bassadi BM, Furberg CD, Kuller LH, Borhani NO, Rautaharju PM, 15. Safar ME. Pulse pressure and essential hypertension: Clinical and O Leary DH, et al. Isolated systolic hypertension and subclinical cardiotherapeutic implications. J Hypert. 1989;7: vascular disease in the elderly. Initial findings from the Cardiovascular Health Study. JAMA. 1992;268: Franklin SS, Weber MA. Measuring hypertensive cardiovascular risk: The cardiovascular overload concept. Am Heart J. 1994;128: Kannel WB. Blood pressure as a cardiovascular risk factor: Prevention and Treatment. JAMA. 1996;275: Franklin SS. The concept of vascular overload in hypertension. Cardiol Clin. 1995;13: Perry HM Jr, Smith WM, McDonald RH, Black B, Cutler JA, Furgberg CD, et al. Morbidity and mortality in the systolic hypertension in the 18. Benetos A, Safar M, Rudnichi A, Smulyan H, Richard J-L, Ducimetiere elderly program (SHEP) pilot study. Stroke. 1989;20:4 13. P, et al. Pulse pressure: A predictor of long-term cardiovascular mortal- ity in a French male population. Hypertension. 1997;30: Kannel WB, Wolfe PA, Verter J, McNamara PM. Epidemiologic assess- 19. Madhavan S, Ooi WL, Cohen H, Alderman MH. Relation of pulse ment of the role of blood pressure in stroke. JAMA. 1996;276: pressure and blood pressure reduction to the incidence of myocardial infarction. Hypertension. 1994;23: Lee M-LT, Rosner BA, Vokonas, PS, Weiss, ST. Longitudinal analysis 20. Mitchell GF, Moye LA, Braunwald E, Rouleau JL, Bernstein V, Geltof adult male blood pressure: The Normative Aging Study, man EM, et al. Sphygmomanometrically determined pulse pressure is J Epidemiol Biostat. 1996;1: a powerful independent predictor of recurrent events after myocardial 10. Tate RB, Manfreda J, Krahn AD, Cuddy TE. Tracking of blood pressure infarction in patients with impaired left ventricular function. SAVE over a 40-year period in the University of Manitoba follow-up study. Investigators. Survival and Ventricular Enlargement. Circulation. 1997; Am J Epidemiol. 1995;142: :