Blood Pressure and Mortality Risk in Older People - Wiley Online Library

Blood Pressure and Mortality Risk in Older People: Comparison Between African Americans and Whites Dan G. Blazer, MD, PhD,* Lawrence R. Landerman, PhD,* Judith C. Hays, RN, PhD,* Tana A. Grady, MD,† Richard Havlik, MD, MPH,‡ and Maria-Chiara Corti, MD, PhD§

OBJECTIVES: To determine the risk from hypertension for all-cause mortality in a racially mixed sample of community-dwelling older adults. DESIGN: Baseline blood pressure was assessed between 1985 and 1986 in a sample of persons 65 years of age and older from five counties of the Piedmont of North Carolina (N ⫽ 4,162). All-cause mortality was monitored annually over the subsequent 6 years as part of the Established Populations for Epidemiologic Studies of the Elderly (EPESE) sponsored by the National Institute on Aging. SETTING: Eighteen percent of all respondents in the sample had a systolic blood pressure of ⬎160 (17% for whites and 18% for African Americans) and 16% had a diastolic blood pressure of ⬎90 (14% for whites and 20% for African Americans). During the 6 years of follow-up, 29% of the sample died (with no difference in mortality rates between whites and African Americans). PARTICIPANTS: 4,000 community-dwelling people age 65 years and older; 1,846 were white and 2,154 were African American. MEASUREMENTS: Systolic and diastolic blood pressure and all-cause mortality. RESULTS: Systolic blood pressure positively related to mortality during the 6 years of follow-up (relative risk ⫽ 1.05). Among whites the relationship of diastolic pressure to mortality was nonlinear, with those at the upper and lower ends of the distribution at increased risk. Among African Americans, diastolic pressure was unrelated to mortality. The analyses were controlled for age; gender; educa-

From the *Duke University Medical Center and the Duke University Center for the Study of Aging and Human Development, Durham, North Carolina; †University of Rochester School of Medicine, Rochester, New York; ‡Epidemiology, Demography and Biometry Program, National Institute on Aging, National Institutes of Health, Bethesda, Maryland; and §The University of Padova, Padova, Italy. This research was performed pursuant to contract N01-AG-4-2110 from the National Institute on Aging in support of the Established Populations for Epidemiologic Studies of the Elderly. Address correspondence to Dan G. Blazer, MD, PhD, J. P. Gibbons Professor of Psychiatry and Behavioral Sciences, Box 3003 DUMC, Durham, NC 27710.

JAGS 49:375–381, 2001 © 2001 by the American Geriatrics Society

tion; body mass index (BMI); smoking history; taking a medication to manage blood pressure; a history of cancer, diabetes mellitus, heart attack, or stroke; poor subjective health; impaired functional status; and cognitive impairment. CONCLUSIONS: The findings confirm that among older adults there is a significant relationship overall between systolic blood pressure and mortality over 6 years of follow-up in both whites and African Americans. Diastolic pressure was a risk factor for whites only. J Am Geriatr Soc 49:375–381, 2001. Key words: aging; blood pressure; hypertension; longitudinal study; mortality; race; African American

T

he association between hypertension and mortality has been well documented in middle-aged populations.1–3 Reports from older populations, however, have exhibited a more complex relationship between blood pressure and mortality, for increased mortality can also be seen among older persons with low blood pressure.4–6 Yet few studies of community populations have included large-enough numbers of whites and African Americans within the same geographic region to compare mortality rates among these two racial groups of elders.7 In a previous study at Duke of the Established Populations for Epidemiologic Studies of the Elderly (EPESE) by our group, hypertension was found to be more prevalent among African Americans than among whites in both crude and adjusted analyses.8 In this study, 44% of white men, 52% of white women, 50% of African-American men, and 66% of African-American women experienced hypertension (systolic blood pressure ⬎160 or diastolic blood pressure ⬎90). Older age, female gender, lower socioeconomic status, obesity, and diabetes mellitus were all associated with hypertension. However, even after adjusting for these covariates, African-American race remained an independent risk factor for high blood pressure, with an adjusted risk ratio of 1.30. In an analysis of the first National Health and Nutrition Examination Survey (NHANES-1), the relative risk of death associated with systolic blood pressure of ⱖ140 mmHg for all age groups

0002-8614/01/$15.00

376

BLAZER ET AL.

was higher for African Americans than for whites (1.7 vs 1.5).9 Therefore, the increased prevalence of hypertension among older African-Americans coupled with findings that African Americans may experience greater risk of mortality from hypertension than whites across the life cycle suggests that the risk of mortality among older African Americans from hypertension might be greater than that for whites. In a sample of 4,000 community-dwelling older people, 1,846 whites and 2,154 African Americans, systolic and diastolic blood pressure were assessed at baseline along with a series of control variables including age, gender, education, income, body mass index (BMI), smoking history, use of medicine for hypertension, a history of cancer, a history of diabetes, and cognitive impairment. These subjects were followed for 6 years. The investigators posed the following two hypotheses: 1. Elevated systolic and diastolic blood pressure will be associated with mortality for both whites and African Americans over 6 years in this sample of older adults. 2. In controlled analyses, African Americans will continue to experience an increased risk of mortality secondary to hypertension than will whites. METHODS The overall EPESE methods have been summarized elsewhere.10 The North Carolina sample consisted of 4,162 community residents 65 years of age and older selected from one urban and four rural counties in the North Carolina Piedmont. In order to optimize both racial differences (African Americans vs whites) and urban/rural differences, the sample was drawn to insure assessment of approximately equal numbers of subjects from the urban and rural counties and equal numbers of African Americans and whites. Forty-eight percent of those sampled were from the urban county, and 52% were from the rural counties. There was a high proportion of African Americans compared with whites in all five counties, ranging, at the time of the 1980 U.S. Census, from one-third in the urban county to nearly two-thirds in one of the rural counties. However, African Americans were oversampled, such that 54% of the sample was African-American, and 46% of the sample was non-African-American (almost exclusively white). The baseline survey, conducted during 1985 and 1986, included data on demographic characteristics, medical history, standard symptom questionnaires, prescription and nonprescription drug use, health habits, and healthcare utilization. Blood pressure and pulse measurements were also obtained. For those participants unable to respond for themselves, proxy informants (n ⫽ 162) were used. Follow-up surveys were conducted yearly for 6 years after the baseline interview. In-person follow-ups were conducted in 1989/1990 and 1992/1993. Telephone follow-ups were conducted in 1987/1988, 1988/1989, 1990/1991, and 1991/ 1992. Our study examines the relationship between baseline systolic blood pressure, diastolic blood pressure, pulse pressure, and all-cause mortality over the first 6 years of follow-up. As discussed below, we go beyond prior studies

APRIL 2001–VOL. 49, NO. 4

JAGS

with respect to how blood pressure is measured and the controls employed. One hundred and sixty-two proxies were eliminated due to missing data on several predictors of interest, leaving an analysis sample of 4,000. This exclusion is consistent with Taylor et al.6 Analyses were performed for all respondents and for white respondents and African-American respondents separately. Three blood pressure measurements were taken using a standard mercury sphygmomanometer with appropriate cuff size after the participant had been seated for at least 5 minutes, according to the standard protocol used. The average of the final two blood pressure measurements were used in this report. Two hundred sixty eight (of 4,000) respondents were missing blood pressure data. Following Cohen and Cohen,11 these missing cases were retained in the analyses, and coded 1 on a dummy variable for missing data. In initial analyses, low, middle, and high blood pressure were defined a priori to provide clinically recognizable groups of adequate size and maintain consistency with previous analyses by Taylor et al.6 For systolic blood pressure, the low, middle, and high strata consisted of those with baseline systolic pressures of ⬍130, 130–159, and ⱖ160 mmHg, respectively. Diastolic blood pressure strata consisted of those with baseline diastolic pressures of ⬍75, 75–89, and ⱖ90 mmHg. The analyses were then repeated treating systolic blood pressure, diastolic blood pressure, and pulse pressure as continuous variables. The rationale for these additional analyses was that the use of conventional cutpoints for low, medium, and high pressure might distort or obscure underlying relationships that were continuous or had different break points. Initial control variables were also consistent with Taylor et al.,6 and included age; gender; education; income; body mass; cigarette pack years; medication; and the presence of cancer, diabetes, or cognitive impairment. Reasoning that unmodeled comorbid conditions might increase mortality among those with low blood pressure and thereby mask the impact of high blood pressure, we added controls for baseline heart attack, stroke, poor subjective health, and cognitive impairment in subsequent models. Demographic information on age and gender as well as history of major chronic conditions diagnosed by a physician (including hypertension and its current treatment status); history of heart attack, stroke, cancer, or diabetes; and self-assessment of health were obtained by self-report. Functional status was assessed using scales frequently used in epidemiological studies and used across all EPESE sites.10 Medications taken during the past week were recorded from the container labels or direct identification and coded uniformly. Height and weight measures were obtained by self-report but not measured. BMI was calculated based on self-report of height and weight. Race/ethnicity was determined using self-report and observation. Cognitive status was determined using the Short Portable Mental Status Questionnaire.12 Vital status of all participants was ascertained by local surveillance mechanisms, specifically checking for local death certificates and obtaining information from the National Death Index and follow-up interviews. Deaths were recorded for sample members during the 6 years following the first interview. Death certificates were obtained for all

JAGS


BLOOD PRESSURE AND MORTALITY RISK

377

Table 1. Descriptive Statistics of the Independent Variables Used in this Study (N ⫽ 4,000 Nonproxy Respondents). Weighted Means (SD*), Proportions, and Unweighted Ns Are Presented All Respondents White African American Male Female (N ⫽ 4,000) (n ⫽ 1,846) (n ⫽ 2,154) P-Value (n ⫽ 1,402) (n ⫽ 2,596) P-Value

Independent Variables Age at baseline (years) Female African American Education (years) Yearly income (thousands) Body mass index Pack/years smoking Takes blood pressure medication Self-reported cancer Self-reported diabetes mellitus Cognitive impairment (SPMSQ) ADL impairment (Katz) ADL impairment (Rosow-Breslau) Subjective health Heart attack Stroke

73.1 (6.4) .62 .35 9.3 (4.1) 12.8 (11.5) 25.5 (4.6) 1.5 (1.7) .40 .12 .15 .11 .20 (.70) .90 (1.10) 2.48 (.90) .12 .07

73.1 (7.5) .62 — 10.2 (4.4) 15.5 (4.8) 24.8 (5.0) 1.6 (2.15) .36 .15 .12 .11 .18 (.46) .84 (1.22) 2.39 (.84) .14 .06

73.1 (5.2) .63

72.4 (6.2) —

7.5 (3.3) 7.7 (5.4) 26.8 (4.1) 1.3 (1.3) .47 .06 .22 .09 .21 (.52) .95 (1.29) 2.54 (.78) .11 .08

.000 .000 .000 .000 .000 .000 .000 .000 .20 .002 .001 .011 .20

73.5 (6.4)

9.0 (4.6) 9.4 (3.7) 15.9 (13.1) 10.9 (10.0) 25.4 (4.1) 25.5 (4.9) 2.5 (1.8) .89 (1.4) .31 .44 .13 .11 .15 .16 .09 .12 .15 (.66) .19 (.69) .63 (1.01) .95 (1.21) 2.42 (.93) 2.44 (.88) .16 .10 .09 .05

.000

.000 .000 .000 .000

.013 0.07 0.000 0.47 .000 .000

* For dichotomous variables, means represent the proportion coded 1 and standard deviations are not given. ADL ⫽ activity of daily living; SPMSQ ⫽ Short Portable Mental Status Questionnaire; SD ⫽ standard deviation.

decedents and coded for underlying cause of death. (Specific cause of death is not analyzed in this study.) RESULTS In Table 1, the independent variables for the sample are presented by gender and race. As can be seen, mean age is virtually identical across gender and race groups. Whites had completed on average 4 more years of education than had African Americans, and whites and men had higher incomes. BMI was somewhat higher among African Americans. Men were more likely to report more pack years of cigarette smoking; African Americans and women were more likely to use medication for blood pressure. Whites were more likely to report cancer and African Americans more likely to report diabetes. The blood pressure measurement by race and gender is presented in Table 2. As reported in previous studies, without adjusting for other factors, diastolic but not systolic blood pressure is significantly higher among African

Americans than whites, and men than women. In Table 3, 6-year and annual mortality rates are presented. As can be seen, 29% of the sample died over 6 years with no difference by race but with men having significantly higher mortality than women. As would be expected, the frequency of death by year increased gradually in waves six and seven. Results from two Cox proportional hazards models are presented for all respondents and separately for whites and African Americans in Table 4. The dependent variable is the number of days between the baseline interview and death. Model coefficients are risk ratios and represent multiplicative effects on the hazard of mortality (the proportional hazards assumption was verified). The covariates in models 1 through 3 are consistent with Taylor et al.6 In models 4 through 6, additional controls on baseline health have been added to the model. As can be seen, in models 4 through 6, higher systolic blood pressure (ⱖ160 compared with ⬍130) becomes a significant risk factor for mortality (relative risk (RR) ⫽ 1.38, P ⱕ .01) when these

Table 2. Blood Pressure by Race and Gender: Weighted Proportions Blood Pressure (mmHg) Systolic ⬍130 Systolic 130–159 Systolic ⱖ160 Missing on systolic Diastolic ⬍75 Diastolic 75–89 Diastolic ⱖ90 Missing on diastolic

All Respondents (N ⫽ 4,000)

White (n ⫽ 1,846)

African American (n ⫽ 2,154)

.23 .53 .18 .06 .34 .43 .16 .07

.23 .53 .17 .06 .37 .43 .14 .07

.23 .52 .18 .07 .28 .44 .21 .07

P-Value

.000 .000

Male (n ⫽ 1,402)

Female (n ⫽ 2,596)

P-Value*

.25 .54 .16 .05 .30 .46 .19 .05

.22 .52 .18 .07 .36 .41 .14 .07

.007 .000 .02 .000 .002

* P-values are reported when statistically significant (t-tested mean differences are present for African Americans vs whites and men vs women, respectively).

378

BLAZER ET AL.


JAGS

Table 3. Proportions of Subjects Who Died by Race and Gender: Weighted Proportions

Mortality Between P1 and P3 (cumulative) P1–T1 T1–T2 T2–P2 P2–T3 T3–T4 T4–P3

All Respondents (N ⫽ 4,000)

White (n ⫽ 1,846)

African American (n ⫽ 2,154)

.29 .04 .04 .05 .04 .05 .06

.29 .04 .04 .05 .04 .05 .06

v .29 .04 .04 .04 .04 .05 .06

P-Value

Male Female (n ⫽ 1,402) (n ⫽ 2,596) .37 .06 .06 .06 .05 .06 .07

.24 .03 .03 .04 .04 .04 .05

P-Value*

.000 .000 .001 .002 .017 .046

* P-values are reported when statistically significant (t-tested mean differences are present for African Americans vs whites and men vs women, respectively). P1 ⫽ 1985/86 baseline interview; P2 ⫽ 1989/90 in-person survey; P3 ⫽ 1992/93 in-person survey; T1 ⫽ 1987/88 telephone survey; T2 ⫽ 1988/89 telephone survey; T3 ⫽ 1990/91 telephone survey; T4 ⫽ 1991/92 telephone survey.

additional controls on baseline health are added to the model. No racial differential is present for this effect. Compared with those with diastolic blood pressure ⬍75, diastolic pressure between 75 and 89 is protective against mortality, an effect that does not change with additional controls on baseline health. This protective effect does not vary with race. While the effect of diastolic pressure ⱖ90

varies with race, it is not significant in either group. Other control variables shown to be risk or protective factors related to mortality were as would be expected. Older age; male gender; lower income; higher body mass; history of increased pack years; taking blood pressure medication; and a history of cancer, diabetes, stroke, or heart attack were risks for higher mortality. Functional and cognitive

Table 4. Adjusted Relative Risk of Mean Systolic Blood Pressure and Diastolic Blood Pressure (Columns 1–3) and Pulse Pressure (Columns 4–6) on 6-Year Mortality (P1–P3)

Systolic BP (Systolic BP)2 Diastolic BP (Diastolic BP)2 Pulse Pressure (Pulse Pressure)2 Missing on BP Age Female African American Education Income BMI BMI2 Packyears Takes BP Medication Cancer Diabetes Cognitive impairment Katz ADL impairment Rosow Breslau impairment Subjective health Heart attack Stroke

1 All Respondents (N ⫽ 4,000)

2 Whites (n ⫽ 1,846)

1.005** NA† .952*** 1.0003***

1.005* NA† .948*** 1.0003**

1.37*** 1.06*** .53*** .94 1.002 .994* .86*** 1.002** 1.13*** 1.13* 1.24** 1.52*** 1.36*** 1.06* 1.37*** 1.12*** 1.29*** 1.62***

1.45** 1.08*** .52*** — 1.000 .995 .86** 1.002 1.20*** 1.22** 1.13 1.42** 1.29** 1.09* 1.36*** 1.14** 1.31** 1.58**

3 African Americans (n ⫽ 2,154)

4 All

5 Whites

6 African Americans

1.005*** NA† 1.35** 1.06*** .53*** .94 1.001 .994* .86** 1.002*** 1.13*** 1.12* 1.25** 1.52*** 1.37*** 1.06 1.37*** 1.12** 1.31*** 1.61***

1.006** NA† 1.44** 1.08*** .52*** — .997 .996 .86** 1.002 1.20*** 1.18** 1.14 1.42** 1.31** 1.09* 1.37*** 1.13** 1.38** 1.52**

1.003 NA† 1.33* 1.04*** .49*** — 1.001 .990 .84*** 1.002*** 1.01 1.01 1.66*** 1.65*** 1.45*** 1.02 1.38*** 1.09 1.26*** 1.62***

1.004 NA† 1.001 1.000

1.37** 1.04*** .50*** — 1.002 .991 .85*** 1.002*** 1.02 .99 1.71*** 1.68*** 1.41*** 1.02 1.37*** 1.09** 1.27** 1.65***

Note: Underlining indicates a significant race-by-blood-pressure interaction. *P ⱕ .10; **P ⱕ .05; ***P ⱕ .01. † Quadratic effect was nonsignificant and dropped from the model. P1 ⫽ 1985/86 baseline interview; P2 ⫽ 1989/90 in-person survey; P3 ⫽ 1992/93 in-person survey; BP ⫽ blood pressure; NA ⫽ not applicable; ADL ⫽ activities of daily living.

JAGS



impairment and poor self-rated health were also predictive of mortality. The additional controls here for baseline health may explain the difference between our findings and those reported by Taylor et al.6 In Table 5, systolic and diastolic pressure and pulse pressure are entered as continuous variables. Columns 1 through 3 report results for systolic and diastolic blood pressure and controls, while columns 4 through 6 report results for parallel models where pulse pressure was substituted for systolic and diastolic pressure. In initial runs, squared as well as linear terms were included for systolic, diastolic, and pulse pressure based on the expectation that each would be nonlinear (J-shaped) in its relationship to mortality. Squared terms for systolic pressure (in models 1–3) and for pulse pressure (models 4–6) were not significant, indicating the absence of a nonlinear effect. We therefore dropped these squared terms and estimated the reduced models reported in Table 5. A significant linear effect (RR ⫽ 1.005, P ⱕ .01) is present for systolic pressure, indicating that a 10-point increase in systolic pressure increases the risk of mortality by 5%. The effect of diastolic pressure is significant and nonlinear (RR: linear ⫽ 0.952, P ⱕ .01; squared term ⫽ 1.003, P ⱕ .01). This quadratic pattern (described in Table 4) implies that an increase in diastolic blood pressure is protective in its lower range and becomes a risk in its upper range. This nonlinear relationship varies with race and

379

is limited to whites (RR: linear ⫽ 0.948, P ⱕ .01; squared term ⫽ 1.003, P ⬍ .05). Among African Americans, diastolic blood pressure is not related to mortality. While pulse pressure is also related to mortality (RR ⫽ 1.005), the fact that pulse pressure combines a linear effect (of systolic pressure) with a nonlinear effect (of diastolic pressure) renders its interpretability (and utility) problematic. Based on the results in Table 5, Table 6 presents the calculated effects of diastolic pressure at 10-point intervals over its range in the sample (minimum ⫽ 21, maximum ⫽ 138). Among African Americans, diastolic pressure is unrelated to mortality, (RR ⫽ 1.001 at all intervals). Among whites, increased diastolic pressure is protective until diastolic pressure exceeds 90 mmHg. Above 90, increases in diastolic pressure are associated with an increased risk of mortality. Above 110, the risk associated with increased diastolic pressure (RR ⫽ 1.01–1.03) is more than twice that associated with increased systolic pressure (RR ⫽ 1.005). Because the effect of diastolic pressure varies significantly with race, its effect among all respondents (an average of within-race effects) is not informative. DISCUSSION A significant and direct relationship between increasing levels of systolic and diastolic blood pressure and risk of all-cause mortality has been consistently reported for middle-aged populations.1,3 However, the relationship be-

Table 5. Adjusted Relative Risk of Baseline Systolic and Diastolic Blood Pressure† on Mortality During 6 Year Follow-Ups (P1–P3)

Systolic 130–150 Systolic ⱖ160 Diastolic 75–89 Diastolic ⱖ90 Missing on blood pressure Age Female Black Education Income BMI BMI2 Packyears Takes BP medication Cancer Diabetes mellitus Cognitive impairment Katz impairment Rosow impairment Subject health Heart attack Stroke

1 All Respondents (N ⫽ 4,000)

2 White (n ⫽ 1,846)

3 African American (n ⫽ 2,154)

1.12 1.20 (P ⫽ .11) 0.81** 0.86 1.25 1.07*** 0.58*** 0.89 0.98* 0.99*** 0.84*** 1.00*** 1.14*** 1.27*** 1.30** 1.70*** 1.66***

1.13 1.18 0.78** 0.69** 1.11 1.09*** 0.57*** — 0.98 0.99** 0.82*** 1.003** 1.21*** 1.38*** 1.17 1.54*** 1.65***

1.13 1.18 0.88 1.21 1.59** 1.06*** 0.55*** — 0.99 0.98* 0.81*** 1.003*** 1.03 1.09 1.82*** 1.93*** 1.63***

4 All

5 White

6 African American

1.19* 1.38*** 0.82*** 0.91 1.38*** 1.06*** 0.55*** 0.99 1.004*** 0.99** 0.86*** 1.002*** 1.11*** 1.07 1.38*** 1.54*** 1.39*** 1.08** 1.32*** 1.12*** 1.28*** 1.59***

1.19 1.36** 0.84* 0.77* 1.46*** 1.08*** 0.55*** — 1.01 0.99* 0.83*** 1.003** 1.21*** 1.22* 1.17 1.41*** 1.28** 1.10* 1.34*** 1.16*** 1.32** 1.53***

1.21* 1.32** 0.81** 1.03 1.44** 1.04*** 0.51*** — 1.00 0.99 0.85*** 1.002*** 1.03 0.98 1.79*** 1.63*** 1.43*** 1.06 1.31*** 1.09* 1.25** 1.58***

Note: Effects are presented for all respondents (columns 1 and 4), and for whites and African Americans separately (columns 2, 3, 5 and 6). Underlining indicates a significant race-by-blood-pressure interaction. †Coefficients for middle and high blood pressure groups represent difference between these groups and those in the low pressure groups (systolic ⬍130; diastolic ⬍75). *P ⱕ .10; **P ⱕ .05; ***P ⱕ .01. BMI ⫽ body mass index; BP ⫽ blood pressure.

380

BLAZER ET AL.


Table 6. The (Nonlinear) Effect of Diastolic Blood Pressure on Mortality at Different Levels of Diastolic Pressure* Diastolic Pressure All Respondents White (mmHg) (RR) (RR) African American 20 30 40 50 60 70 80 90 100 110 120 130 140

0.963 0.969 0.975 0.981 0.987 0.993 0.999 1.005 1.011 1.017 1.023 1.029 1.035

0.958 0.964 0.970 0.976 0.982 0.987 0.993 0.999 1.005 1.011 1.018 1.024 1.030

1.001 (NS) A A A A A A A A A A A A

*Effects are calculated as: Exp[b1 ⫹ 2Db2] where: b1 ⫽ linear effect on the logged hazard of mortality; b2 ⫽ quadratic (squared) effect on the logged hazard of mortality; D ⫽ level of diastolic blood pressure. NS ⫽ not significant; RR ⫽ relative risk; A ⫽ all are the same at all intervals.

tween blood pressure and mortality in later life has been demonstrated to be more complex.4,6 In addition, there have been consistent reports of a relationship between hypertension and race, with higher levels among African Americans than among whites. The relationship has been suggested as a prime cause of the increased mortality rates of African Americans compared with whites in midlife.13,14 Therefore, we hypothesized that systolic and diastolic hypertension would be associated with all-cause mortality for both whites and African Americans over 6 years of follow-up in a community-dwelling sample of older adults. The data presented here support the hypothesis regarding systolic hypertension and are similar to those of Taylor et al.6 They also extend Taylor et al. by showing that the relationship of systolic pressure to mortality is continuous and linear and that this relationship can be obscured if baseline health is not adequately controlled. Thus (as shown in Table 6) trichotomized systolic pressure is not significantly related to mortality in initial models using minimal controls for baseline health. We found a complex relationship between diastolic blood pressure and mortality. Among African Americans, diastolic pressure was unrelated to mortality. Among whites, increased diastolic pressure was protective until diastolic pressure exceeded 90 mmHg. Above this level, diastolic pressure increased the risk of mortality. Above a level of 110, the risk associated with increased diastolic pressure was more than twice that associated with systolic pressure. This nonlinear relationship was obscured and distorted when diastolic pressure was trichotomized using conventional cutpoints (in Table 6). Nonlinear effects have been reported elsewhere. For example, Sutton-Tyrrell et al.15 found that systolic hypertension is strongly correlated with carotid stenosis, yet, among those with isolated systolic hypertension, low diastolic blood pressure is a marker for carotid stenosis. Starr

JAGS

et al.16 found a nonlinear relationship between blood pressure and mortality. Persons with the highest systolic blood pressures (160–169 mmHg) were slightly less likely to die than those with lower pressures (150–159 mmHg). The reason for a nonlinear relationship between diastolic pressure and mortality is undetermined but may be due to vascular stiffness.16–19 Given the higher prevalence of hypertension among African Americans in this sample and previous reports that hypertension may not only be more prevalent among African Americans but also more lethal, we hypothesized that older African Americans would experience an increased risk of mortality compared with whites if they were hypertensive.8,13 The data presented reject this hypothesis. In fact, our data indicate that increases in diastolic pressure (⬎90) are more lethal for whites. Perhaps the virtually identical rates of all-cause mortality for the two racial groups render the possibility of demonstrating a racial difference due to hypertension difficult to establish. Yet the results from such a large number of both whites and African Americans coupled with excellent surveillance over 6 years of follow-up suggest that these findings do reflect the actual relationships in community samples. A number of explanations have been posited for the complex relationship between hypertension and mortality that apply to both whites and African Americans in later life.6 The higher frequency of comorbid illness at baseline in older adults may alter the risk for mortality. For example, someone with comorbid arteriosclerotic coronary heart disease, other comorbid diseases, or frailty may have lower blood pressure yet have an excess risk of death. Unlike in the Taylor et al. study,6 this sample was adjusted for self-reported cancer and diabetes, two of the most frequent chronic illnesses in later life. Nevertheless, selfreport of illness is far from accurate. Another potential factor is that the baseline sample had already been censored for hypertension, for the greatest risk of mortality from hypertension may occur in midlife rather than later in life. This possibility should be seriously considered, given that many of the racial differences observed earlier in life appear to be less frequent when comparing the health status of older whites with older African Americans. For example, in the oldest old, the race differences in mortality apparently shift, with African Americans experiencing a lower death rate than whites in the 85 and older age group.19,20 In summary, we report a systematic longitudinal study of the relationship between blood pressure and mortality in a racially mixed sample of community-dwelling older people at baseline. Systolic pressure was positively related to mortality among both African Americans and whites. Diastolic pressure was nonlinearly related to mortality among whites only. REFERENCES 1. Shurleff D. Section 30. In: Kannel W, Gordon T, eds. Some Characteristics Related to the Incidence of Cardiovascular Disease in Death: Framingham Study 18-Year Follow-Up (DHEW Publication # NIH 74-599). Washington, DC: U.S. Government Printing Office, 1974, pp 74–599. 2. Rabkin S, Mathewson F, Tate R. Predicting risk of ischemic heart disease and cerebral vascular disease from systolic and diastolic blood pressure. Ann Intern Med 1978;88:342–345. 3. Lichtenstein M, Shipley M, Rose G. Systolic and diastolic blood pressure as

JAGS

4. 5. 6. 7. 8.

9. 10.

11. 12. 13.


predictors of coronary heart disease mortality in the White Hall Study. BMJ 1985;291:243–245. Mattila K, Haavisto M, Ragala S. Blood pressure and survival in the very old. BMJ 1988;296:887–889. Moye L, Davis B, Hawkins C et al. Conclusions and implications of the systolic hypertension in the elderly program. Clin Exp Hypertens 1993;15:911–924. Taylor J, Cornoni-Huntley J, Curb J et al. Blood pressure and mortality risk in the elderly. Am J Epidemiol 1991;134:489–501. Corti M, Guralnik J, Ferrucci L et al. Evidence for a black-white crossover in all-cause mortality. Am J Public Health 1999;89:308–314. Svetkey L, George L, Burchette B et al. African-American/white differences in hypertension in the elderly: An epidemiological analysis in central North Carolina. Am J Epidemiol 1993;137:64–73. Cooper R, Liao Y, Rotimi C. Is hypertension more severe among U.S. blacks, or is severe hypertension more common? Ann Epidemiol 1996;6:173–180. Blazer D, Burchette B, Service C et al. The association of age and depression among the elderly: An epidemiologic exploration. J Gerontol A Biol Sci Med Sci 1991;46:M210–M215. Cohen J, Cohen P. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. New York: John Wiley, 1975. Pfeiffer E. A short, portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. J Am Geriatr Soc 1975;23:433–441. Jamerson K. Prevalence of complications and response to different treat-


14.

15.

16. 17.

18.

19.

20.

381

ments of hypertension in African Americans and white Americans in the U.S. Clin Exp Hypertens 1993;15:979–995. Sacco R, Kargman D, Zamanillo M. Race-ethnic differences in stroke risk factors among hospitalized patients with cerebral infarction: The Northern Manhattan Stroke Study. Neurology 1995;45:659–663. Sutton-Tyrrell K, Alcorn H, Wolfson S et al. Predictors of carotid stenosis in older adults with and without isolated systolic hypertension. Stroke 1993;24: 355–361. Starr J, Inch S, Cross S et al. Blood pressure and mortality in healthy old people: The r-shaped curve. BMJ 1996;313:1243–1244. Bots M, Hofman A, de Bruyn A et al. Isolated systolic hypertension and vessel wall thickness of the carotid artery. The Rotterdam Study. Arterioscler Thromb 1994;13:64–69. Psaty B, Furberg C, Kuller L et al. Isolated systolic hypertension and subclinical cardiovascular disease in the elderly: Initial findings from the Cardiovascular Health Study. JAMA 1993;269:214–215. Flynn R, Field T, Rosner B et al. Evidence for a positive linear relation between blood pressure and mortality in elderly people. Lancet 1995;345: 825–829. Manton K, Soldo B. Disability and mortality among the oldest-old: Implications for current and future health and long-term-care service needs. In: Suzman R, Willil D, Manton K, eds. The Oldest Old. New York: Oxford University Press, 1992, pp 17–43.