PROGRESS IN CARDIOLOGY
Is obesity-related hypertension cardiovascular risk? The Framingham Study William
B. Kannel,
Boston,
Mass.,
MD:
Tingjie
and Bethesda,
Zhang, MD, and Robert J. Garrison,
METHODOLOGY
The Framingham Study has observed a cohort of 5209 men and women initially aged 30 to 62 years in From sthe Section of Preventive Medicine and Epidemiology, Boston University School of Medicine, Evans Memorial Department of Clinical Research, University Hospital at Boston University Medical Center; and bField Studies and Biometry Branch, Epidemiology and Biometry Program, Division of Epidemiology and Clinical Applications, National Heart, Lung and Blood Institute. Supported by National Heart, Lung, and Blood Institute grant numbers NIH-NOl-HV-92922, NIH-NOl-HV-52971 and by contract number NOlHC-38038, by The Charles A. Dana Foundation; and by Merck Sharp and Dohme, West Point, Pa. for publication
April
2, 1990;
accepted
Reprint requests: William B. Kannel, MD, 720 Harrison Ave., Boston, MA 02118.
4/l/23607
MS.b
Md.
Hypertension and obesity have been frequently noted to coexist.‘-7 In fact, obesity and weight gain have been found prospectively to predispose to the development of elevated blood pressure.3-5 Both obesity and hypertension predispose to the occurrence of cardiovascular disease, including coronary disease, stroke, and cardiac failure.4s7-14 Since obesity and weight gain have also been shown to promote other risk factors such as glucose intolerance, hypercholesterolemia, hyperuricemia, reduced high-density lipoprotein (HDL)-cholesterol, and hyperinsulinemia, it would seem that hypertension accompanied by obesity should be more dangerous than hypertension that is not accompanied by obesity. However, controversy exists as to whether hypertension accompanied by obesity is more15-18 or less1g-26 hazardous. Thus it is not clear whether hypertension has a different prognostic importance in the obese than in the lean. The purpose of this report is to examine this issue in a prospective study that utilizes 34 years of cardiovascular disease information on 5209 subjects biennially characterized by their blood pressure and weight status.
Received
less of a
Doctors
May
25, 1990.
Office
Bldg.,
Suite
1105,
1948 to 1952 for a period of 34 years for the development of cardiovascular disease by means of biennial clinic examinations, surveillance of hospital admissions, death certification, and medical examiner reports. Sampling, response rates, clinical criteria for cardiovascular end points, and laboratory methods employed have been reported in detail elsewhere.27T 28 Blood pressures were measured to the nearest 2 mm Hg, using a mercury sphygmomanometer with cuffs of variable widths, depending on the arm circumference, in the left arm with the patient seated. A participant was designated hypertensive if either of the two systolic blood pressure measurements were >160 mm Hg and either of two diastolic pressure measurements was >95 mm Hg, or if the examining physician determined that the subject was taking an antihypertensive agent. Mild hypertension was defined as pressure in the range of 140 to 159 mm Hg systolic and/or diastolic pressures between 90 and 94 mm Hg. Height and weight were determined on each biennial examination throughout the follow-up. Subjects were weighed on a clinical beam scale, wearing an examination gown. The obesity index chosen for this analysis was the body mass index (BMI) calculated from the ratio of weight (kilograms) to height (meters) squared. The 34-year follow-up period was divided into four segments of 8 years to permit reclassification of individuals at each of four baseline examinations. In this report, the final 2 years of follow-up information is ignored. Biennial examination number 1,5,9, and 13 were used as the baseline examinations at which an individual’s risk factor values and his or her age were “updated.” Unless otherwise specified, only study participants who were free of cardiovascular disease (CVD) as defined below were entered at the baseline examinations, and the time of occurrence of death or the first occurrence of CVD over the subsequent 8 years was noted. All 8-year segments were pooled for analyses, 1195
1 I 96
Kannel,
Zhang, and Garrison
American
November 1990 Heart Journal
I. Eight-year cardiovascular diseaseincidence rates per 100 participants by sex, age, and baselineexamination number: Framingham Study, 34-year follow-up Table
Men
Women
Rasrline exam No. * Age (.vri 30-39 40-49 50-59 60-69
1
5
3.0 (823j 8.9 (766) 18.2 (600) 20.0
(60) 70+
-
9
Baseline exam No. * 1 :I
Total rL~rnt.s/No.
18.8 (309) 22.9 (407) 32.9 (243)
(911) 160 (I 743) 385 (2093) 273 (1193) 13’1 (398) ”
2.3
2:
(88) 8.6 (752) 19.4 (583) 19.8 (329) -
12.0 (22.5) 17.5 (601) 26.0 (396) 34.0 (153)
Number of subjects is shown in parentheses. *Subjects with prevalent cardiovascular disease
at
each baseline are
II. Prevalence of hypertension by relative weight* excluding subjects with prevalent cardiovascular disease
Tertile of BMI Lowest Middle Highest *All trends statistically gression analysis.
Women
Age (yr)
Age fyr)
S-64
659d
X5-64
65-94
11 17
19 32
9 15
44 39
27
30
30
53
significant
(p < 0.001 I by age-adjusted
1.0 (1040) 2.9 (944) 10.2 (773) 16.4
(611
,5 0.0 (111) 2.8 (960) 9.7 1785) 15.9 (484)
9
13 -~
Total events/No. 1 (1151)
x.2 (292) 10.5
(846) 17.5 (560) 28.7 (265)
(2lt) 11.5 (399) 14.1 (661 t 25.2 (445)
290 (2803) 278 (1766) 189 (714)
excluded
Table
Men
I
logistic re-
which included multivariate logistic regression analysis2g for testing the relationship between hypertension prevalence and other attributes. The Cox proportional hazards mode13’ was used to model the time to CVD as a function of hypertension, BMI, and the other CVD risk factors. The coefficients from the model provide point estimates of relative risks of CVD for contrasting levels of the independent variables in the regression analysis. The likelihood ratio test was used to test the significance of the regression coefficients. The appropriateness of the proportional hazards model-i.e., the constancy of the ratio of hazards for various risk factor levels over time-was tested by evaluating the association between model residuals and the rank order of the time to CVD.31 Cardiovascular events included in the analysis were coronary heart disease, stroke, cardiac failure, and occlusive peripheral arterial disease, which were defined using previously published criteria.27* 28,32 When study participants who had prevalent CVD
at the first examination are excluded, there were 978 men and 845 women who suffered an incident CVD event. The age-, sex-, and baseline-specific CVD incidence rates are displayed in Table I. The remarkable uniformity of the age-specific rates across the four baseline examinations justifies pooling of 8-year observations from the 32-year period. OBSERVATIONS
Under age 65 years, the prevalence of hypertension in both sexes increased three-fold from the lowest to the highest tertile of BMI (Table II). Over age 65 years, the upward trend is less striking in men but is statistically significant. The risk of cardiovascular events is increased in both borderline and definite hypertensive individuals at all ages and in both men and women (Table III). The risk of cardiovascular events also increased progressively with relative weight in men (Table IV). This relationship is statistically significant at all ages. On adjustment for other risk factors promoted by obesity, including hypertension, the association remains statistically significant in men. In women, a significant relationship, age- and risk factor-adjusted, is noted under age 65 with borderline statistical significance noted above age 65. The $-year relative risk of an initial cardiovascular event associated with hypertension (compared to non-hypertensive individuals) was estimated from the multivariate proportional hazards regression coefficient. In addition to hypertension status, the model included age in years and cigarette smoking, since the relation of cardiovascular disease to weight may be confounded by the tendency of smokers to be leaner. There is no hint of a diminishing impact of
Volume Number
120 5
hypertension with increasing BMI in either sex at any age (Table V). For cardiovascular mortality, there is also no indication of a diminishing impact of hypertension with increasing BMI, and the lean subjects under age 65 appear to have the lowest relative risk (Table VI). There is a suggestion of a lesser impact in those overweight individuals who are over age 65 that is not statistically significant. For coronary events in particular, there is if anything an increased impact of hypertension in the obese (Table VII). However, a formal statistical test of this hypothesis failed to reject the null hypothesis of no interaction. For coronary heart disease mortality, the trend is not seen and at no age in either sex does the impact of hypertension wane with increasing weight (Table VIII). A graphic display of the data cross-classifying ageadjusted incidence rates of cardiovascular disease by tertiles of BMI and systolic blood pressure indicates no diminishing impact of hypertension with increasing relative weight (Fig. 1). In fact, the risk at any level of blood pressure seems to increase with BMI. Extending the analysis to include the important cardiovascular end points of myocardial infarction and stroke adjusted for age and cigarette smoking, no interaction of BMI and hypertension in any age-sex subgroup as indicated. The net effect of hypertension on cardiovascular disease incidence shows no tendency to diminish with increasing BMI in either sex (Table IX). COMMENTS
Obesity and hypertension have evoked widespread concern because of their high prevalence and influence on public health. In the United States it has been recently noted that 26% of the population between 1976 and 1980 were overweight, and no less than 12.4 million of these 34 million adult Americans were severely overweight. g* 33 Childhood and adolescent obesity has increased dramatically over the past two decades. It is estimated that about one fourth of all American children are overweight, and these children tend to become obese adults?, 34 This situation is not unique to the United States, as it is also a problem in Europe.l’t 23,35 Hypertension is also a ubiquitous major problem in North America, in Europe, and in Asia.36* 37 In the Framingham Study, it has been recently noted that 60% of all 70-year-old individuals are now receiving hypertensive treatment. 35 The Tecumseh Study found that 27 % of men and 37 % of women aged 40 to 59 years were hypertensive.4 A recent epidemiologic study in China (1979 to 1980), sampling more
Obesity-related hypertension
1197
III. Risk of cardiovascular events by hypertensive status; &year age-adjustedrate per 100study participants: Framingham Study, 34-year follow-up Table
Hypertensive status Normal Mild Definite All trends significant
Men
Women
Age (yr)
Age (.vr)
35-64
65-94
35-64
65-94
9.0 15.2 21.7
17.4 30.4 43.0
4.5 7.1 13.3
13.5 21.3 28.2
at p < 0.001.
IV. Risk of cardiovascular events by relative weight; 8-year age-adjusted rates per 100 study participants: Framingham Study, 34-year follow-up Table
BMI
quintile
First Second Third Fourth Fifth Significant
Men
Women
Age lyr)
Age (yr)
35-64
65-94
35-64
65-94
11.1 11.5 13.6 14.9 15.5
20.9 30.7 26.2 31.4 34.0
6.5 6.4 5.7 7.6 9.8
21.7 20.0 17.6 26.0 26.0
pValue
trend
Age-adjusted Risk factor-adjusted
Is obesity-related hypertension cardiovascular risk? The Framingham Study William
B. Kannel,
Boston,
Mass.,
MD:
Tingjie
and Bethesda,
Zhang, MD, and Robert J. Garrison,
METHODOLOGY
The Framingham Study has observed a cohort of 5209 men and women initially aged 30 to 62 years in From sthe Section of Preventive Medicine and Epidemiology, Boston University School of Medicine, Evans Memorial Department of Clinical Research, University Hospital at Boston University Medical Center; and bField Studies and Biometry Branch, Epidemiology and Biometry Program, Division of Epidemiology and Clinical Applications, National Heart, Lung and Blood Institute. Supported by National Heart, Lung, and Blood Institute grant numbers NIH-NOl-HV-92922, NIH-NOl-HV-52971 and by contract number NOlHC-38038, by The Charles A. Dana Foundation; and by Merck Sharp and Dohme, West Point, Pa. for publication
April
2, 1990;
accepted
Reprint requests: William B. Kannel, MD, 720 Harrison Ave., Boston, MA 02118.
4/l/23607
MS.b
Md.
Hypertension and obesity have been frequently noted to coexist.‘-7 In fact, obesity and weight gain have been found prospectively to predispose to the development of elevated blood pressure.3-5 Both obesity and hypertension predispose to the occurrence of cardiovascular disease, including coronary disease, stroke, and cardiac failure.4s7-14 Since obesity and weight gain have also been shown to promote other risk factors such as glucose intolerance, hypercholesterolemia, hyperuricemia, reduced high-density lipoprotein (HDL)-cholesterol, and hyperinsulinemia, it would seem that hypertension accompanied by obesity should be more dangerous than hypertension that is not accompanied by obesity. However, controversy exists as to whether hypertension accompanied by obesity is more15-18 or less1g-26 hazardous. Thus it is not clear whether hypertension has a different prognostic importance in the obese than in the lean. The purpose of this report is to examine this issue in a prospective study that utilizes 34 years of cardiovascular disease information on 5209 subjects biennially characterized by their blood pressure and weight status.
Received
less of a
Doctors
May
25, 1990.
Office
Bldg.,
Suite
1105,
1948 to 1952 for a period of 34 years for the development of cardiovascular disease by means of biennial clinic examinations, surveillance of hospital admissions, death certification, and medical examiner reports. Sampling, response rates, clinical criteria for cardiovascular end points, and laboratory methods employed have been reported in detail elsewhere.27T 28 Blood pressures were measured to the nearest 2 mm Hg, using a mercury sphygmomanometer with cuffs of variable widths, depending on the arm circumference, in the left arm with the patient seated. A participant was designated hypertensive if either of the two systolic blood pressure measurements were >160 mm Hg and either of two diastolic pressure measurements was >95 mm Hg, or if the examining physician determined that the subject was taking an antihypertensive agent. Mild hypertension was defined as pressure in the range of 140 to 159 mm Hg systolic and/or diastolic pressures between 90 and 94 mm Hg. Height and weight were determined on each biennial examination throughout the follow-up. Subjects were weighed on a clinical beam scale, wearing an examination gown. The obesity index chosen for this analysis was the body mass index (BMI) calculated from the ratio of weight (kilograms) to height (meters) squared. The 34-year follow-up period was divided into four segments of 8 years to permit reclassification of individuals at each of four baseline examinations. In this report, the final 2 years of follow-up information is ignored. Biennial examination number 1,5,9, and 13 were used as the baseline examinations at which an individual’s risk factor values and his or her age were “updated.” Unless otherwise specified, only study participants who were free of cardiovascular disease (CVD) as defined below were entered at the baseline examinations, and the time of occurrence of death or the first occurrence of CVD over the subsequent 8 years was noted. All 8-year segments were pooled for analyses, 1195
1 I 96
Kannel,
Zhang, and Garrison
American
November 1990 Heart Journal
I. Eight-year cardiovascular diseaseincidence rates per 100 participants by sex, age, and baselineexamination number: Framingham Study, 34-year follow-up Table
Men
Women
Rasrline exam No. * Age (.vri 30-39 40-49 50-59 60-69
1
5
3.0 (823j 8.9 (766) 18.2 (600) 20.0
(60) 70+
-
9
Baseline exam No. * 1 :I
Total rL~rnt.s/No.
18.8 (309) 22.9 (407) 32.9 (243)
(911) 160 (I 743) 385 (2093) 273 (1193) 13’1 (398) ”
2.3
2:
(88) 8.6 (752) 19.4 (583) 19.8 (329) -
12.0 (22.5) 17.5 (601) 26.0 (396) 34.0 (153)
Number of subjects is shown in parentheses. *Subjects with prevalent cardiovascular disease
at
each baseline are
II. Prevalence of hypertension by relative weight* excluding subjects with prevalent cardiovascular disease
Tertile of BMI Lowest Middle Highest *All trends statistically gression analysis.
Women
Age (yr)
Age fyr)
S-64
659d
X5-64
65-94
11 17
19 32
9 15
44 39
27
30
30
53
significant
(p < 0.001 I by age-adjusted
1.0 (1040) 2.9 (944) 10.2 (773) 16.4
(611
,5 0.0 (111) 2.8 (960) 9.7 1785) 15.9 (484)
9
13 -~
Total events/No. 1 (1151)
x.2 (292) 10.5
(846) 17.5 (560) 28.7 (265)
(2lt) 11.5 (399) 14.1 (661 t 25.2 (445)
290 (2803) 278 (1766) 189 (714)
excluded
Table
Men
I
logistic re-
which included multivariate logistic regression analysis2g for testing the relationship between hypertension prevalence and other attributes. The Cox proportional hazards mode13’ was used to model the time to CVD as a function of hypertension, BMI, and the other CVD risk factors. The coefficients from the model provide point estimates of relative risks of CVD for contrasting levels of the independent variables in the regression analysis. The likelihood ratio test was used to test the significance of the regression coefficients. The appropriateness of the proportional hazards model-i.e., the constancy of the ratio of hazards for various risk factor levels over time-was tested by evaluating the association between model residuals and the rank order of the time to CVD.31 Cardiovascular events included in the analysis were coronary heart disease, stroke, cardiac failure, and occlusive peripheral arterial disease, which were defined using previously published criteria.27* 28,32 When study participants who had prevalent CVD
at the first examination are excluded, there were 978 men and 845 women who suffered an incident CVD event. The age-, sex-, and baseline-specific CVD incidence rates are displayed in Table I. The remarkable uniformity of the age-specific rates across the four baseline examinations justifies pooling of 8-year observations from the 32-year period. OBSERVATIONS
Under age 65 years, the prevalence of hypertension in both sexes increased three-fold from the lowest to the highest tertile of BMI (Table II). Over age 65 years, the upward trend is less striking in men but is statistically significant. The risk of cardiovascular events is increased in both borderline and definite hypertensive individuals at all ages and in both men and women (Table III). The risk of cardiovascular events also increased progressively with relative weight in men (Table IV). This relationship is statistically significant at all ages. On adjustment for other risk factors promoted by obesity, including hypertension, the association remains statistically significant in men. In women, a significant relationship, age- and risk factor-adjusted, is noted under age 65 with borderline statistical significance noted above age 65. The $-year relative risk of an initial cardiovascular event associated with hypertension (compared to non-hypertensive individuals) was estimated from the multivariate proportional hazards regression coefficient. In addition to hypertension status, the model included age in years and cigarette smoking, since the relation of cardiovascular disease to weight may be confounded by the tendency of smokers to be leaner. There is no hint of a diminishing impact of
Volume Number
120 5
hypertension with increasing BMI in either sex at any age (Table V). For cardiovascular mortality, there is also no indication of a diminishing impact of hypertension with increasing BMI, and the lean subjects under age 65 appear to have the lowest relative risk (Table VI). There is a suggestion of a lesser impact in those overweight individuals who are over age 65 that is not statistically significant. For coronary events in particular, there is if anything an increased impact of hypertension in the obese (Table VII). However, a formal statistical test of this hypothesis failed to reject the null hypothesis of no interaction. For coronary heart disease mortality, the trend is not seen and at no age in either sex does the impact of hypertension wane with increasing weight (Table VIII). A graphic display of the data cross-classifying ageadjusted incidence rates of cardiovascular disease by tertiles of BMI and systolic blood pressure indicates no diminishing impact of hypertension with increasing relative weight (Fig. 1). In fact, the risk at any level of blood pressure seems to increase with BMI. Extending the analysis to include the important cardiovascular end points of myocardial infarction and stroke adjusted for age and cigarette smoking, no interaction of BMI and hypertension in any age-sex subgroup as indicated. The net effect of hypertension on cardiovascular disease incidence shows no tendency to diminish with increasing BMI in either sex (Table IX). COMMENTS
Obesity and hypertension have evoked widespread concern because of their high prevalence and influence on public health. In the United States it has been recently noted that 26% of the population between 1976 and 1980 were overweight, and no less than 12.4 million of these 34 million adult Americans were severely overweight. g* 33 Childhood and adolescent obesity has increased dramatically over the past two decades. It is estimated that about one fourth of all American children are overweight, and these children tend to become obese adults?, 34 This situation is not unique to the United States, as it is also a problem in Europe.l’t 23,35 Hypertension is also a ubiquitous major problem in North America, in Europe, and in Asia.36* 37 In the Framingham Study, it has been recently noted that 60% of all 70-year-old individuals are now receiving hypertensive treatment. 35 The Tecumseh Study found that 27 % of men and 37 % of women aged 40 to 59 years were hypertensive.4 A recent epidemiologic study in China (1979 to 1980), sampling more
Obesity-related hypertension
1197
III. Risk of cardiovascular events by hypertensive status; &year age-adjustedrate per 100study participants: Framingham Study, 34-year follow-up Table
Hypertensive status Normal Mild Definite All trends significant
Men
Women
Age (yr)
Age (.vr)
35-64
65-94
35-64
65-94
9.0 15.2 21.7
17.4 30.4 43.0
4.5 7.1 13.3
13.5 21.3 28.2
at p < 0.001.
IV. Risk of cardiovascular events by relative weight; 8-year age-adjusted rates per 100 study participants: Framingham Study, 34-year follow-up Table
BMI
quintile
First Second Third Fourth Fifth Significant
Men
Women
Age lyr)
Age (yr)
35-64
65-94
35-64
65-94
11.1 11.5 13.6 14.9 15.5
20.9 30.7 26.2 31.4 34.0
6.5 6.4 5.7 7.6 9.8
21.7 20.0 17.6 26.0 26.0
pValue
trend
Age-adjusted Risk factor-adjusted
