598

Risk Factors for Hypertension in

a National Cohort Study
Earl S. Ford and Richard S. Cooper

Hypertension continues to be a major public health problem in the United States. We used data
from the National Health and Nutrition Examination Survey Epidemiologic Followup Study
(1971-1984) to examine predictors of hypertension for the 7,073 participants free from
hypertension at the baseline examination. The follow-up period averaged 10 years. Body mass
index was positively related to the probability of hypertension developing among white men
(n=2,370), white women (n=3,949), black men (n=231), and black women (n=523). Education
was inversely associated with the probability of hypertension developing among white women
and was of borderline significance among white men and black women. In a subanalysis of
white men (n = l,790) and white women (n=3,063) who completed the 24-hour recall dietary
questionnaire, dietary consumption of sodium, calcium, and potassium did not predict the
development of hypertension. The failure of our study to support findings relating intake of
dietary cations to the development of hypertension may be attributable to imprecision in the
measurement of dietary data and misclassification of hypertension status. These data reinforce
the importance of weight control in the primary prevention of hypertension. (Hypertension
1991;18:598-606)

H ypertension is a major public health problem

in the United States, with 57,100,000
Americans estimated to have the disease.1
Despite extensive research into the etiology and
contributing causes of essential hypertension, the
pathogenesis of the condition is still not explained.
Nonpharmacological approaches to the treatment of
hypertension have included a reduction of salt intake
and increases in dietary potassium and calcium con-
sumption, weight loss, and increased physical activity.
Although the small reductions in blood pressure
from these nonpharmacological measures could the-
oretically have important beneficial public health
consequences,2 it has not been shown that these
approaches can serve as primary measures to retard
or prevent the onset of hypertension in the general
population.
Most of the available data have come from cross-
sectional and experimental studies. However, determi-
nants of blood pressure have been examined in few
prospective studies. We studied the relations of several
sociodemographic, behavioral, and biochemical param-
eters on the incidence of hypertension among a group
From the Division of Diabetes Translation, Center for Chronic
Disease Prevention and Health Promotion, Centers for Disease
Control, Atlanta, Ga., and the Department of Preventive Medicine
and Epidemiology, Loyola University Medical Center, Maywood, 111.
Address for correspondence: Earl Ford, MD, MPH, Division of
Diabetes Translation, Centers for Disease Control, 1600 Clifton
Road, NE, Mailstop K10, Atlanta, GA 30333.
Received May 10, 1990; accepted in revised form June 27, 1991.
of participants followed from the first National Health
and Nutrition Examination Survey (NHANES I),
which was conducted from 1971 through 1975.
Methods
Between 1971 and 1975, 20,729 Americans aged
25-74 years participated in the NHANES I study.
The plans and operations for the NHANES I and
NHANES I Epidemiologic Followup Study have
been published previously.3-6 Of the 14,407 partici-
pants who were 25-74 years old at baseline and
underwent a medical examination, 13,380 (92.9%)
were successfully traced between 1982 and 1984.7
Among the participants, a more detailed examination
was performed on 6,913 persons. These included
3,059 participants in a 1975 augmentation survey plus
a random sample of participants examined from 1971
through 1974. Survivors and proxy respondents for
deceased participants and participants incapable of
responding were interviewed, and blood pressure
measurements were obtained from survivors. Mean
follow-up time was 10 years for traced participants
still alive, with a range of 5-12 years. The study and
the confidentiality of the data from the study were
thoroughly discussed with the participants. By agree-
ing to participate in the study, participants gave
implied consent.
By design, various subgroups of the final sample
completed different study components and question-
naires. A detailed medical examination was per-

formed on 6,913 persons. This included the 3,059
persons from the augmentation survey that was con-
ducted from July 1974 through September 1975 as
well as 3,854 randomly sampled persons from partic-
ipants between 1971 and 1974.8 Consequently, not all
information is available for each respondent. For
example, dietary data were available only for 11,348
participants, serum calcium data for 6,913 partici-
pants, and serum potassium and sodium data for
3,059 participants, whereas serum magnesium data
were available for all participants.
At the baseline examination, a single blood pres-
sure measurement was obtained in accordance with
American Heart Association criteria.9 We defined
hypertension at baseline as a systolic blood pressure
(SBP) of greater than or equal to 160 mm Hg, a
diastolic blood pressure (DBP) of greater than or
equal to 95 mm Hg, or using antihypertensive medi-
cation. The latter factor was determined at the
baseline examination, when all participants between
the ages of 12 and 74 years were asked, "During the
past 6 months, have you used any medicine, drugs, or
pills for high blood pressure?" Participants who
responded that they regularly or occasionally used
antihypertensive medications were eliminated from
the analysis, as were participants with missing infor-
mation that precluded the establishment of hyperten-
sive status. For participants receiving the more de-
tailed examination, the wording of the follow-up
questionnaire differed from the original question-
naire. First, respondents were asked if they had ever
been told by a doctor that they had hypertension or
high blood pressure. Those who responded affirma-
tively were then asked, "Has a doctor ever prescribed
medicine for your (high blood pressure/hyperten-
sion)?" Those who responded affirmatively to this
question and replied to a subsequent question that
they used antihypertensive medication all the time,
often, or once in a while, were classified as hyperten-
sive and excluded from the analysis.
At the follow-up examination between 1980 and
1982, three blood pressure measurements were ob-
tained from each living participant where this was
possible. The average SBP and DBP data from the
last two readings were used to determine hyperten-
sive status. The same definition (SBP ^160 mm Hg,
DBP s95 mm Hg, or the use of antihypertensive
medications) classified participants as hypertensive.
Independent variables were age, income, educa-
tion, leisure-time physical activity, non-leisure-time
physical activity, body mass index, blood cholesterol
level, resting pulse rate, alcohol use, and serum
magnesium level. For the dietary subanalysis, dietary
intakes of sodium, potassium, and calcium and total
number of calories were examined. The continuous
variables of age, income, education, body mass index,
and alcohol use were divided into discrete categories,
as shown in Table 1. Body mass index was divided
into three categories. Normal weight was defined as
less than 27.8 kg/m2 for men and less than 27.3 kg/m2
for women. Obesity was defined as 27.8 to less than
Ford and Cooper Risk Factors for Hypertension 599
Scxnm
Total Diet catfom
Prior to deletion 14.407 11.348 3,059
Delete raca other than white or black
14.235 11.214 3.021
Delete km to foDowup
13,243 10/J37 2J06
Delete p e n o n wtm hypcrteuioo
att
5.653 1,767
Ddete reconb with r'f'Hiig data
Number of reconb available for anatytb 7.073 5/411 1.548
FIGURE 1. Sample sizes available for analysis after various
deletions, National Health and Nutrition Examination Survey
(NHANES I) Epidemiologic Followup Study, 1971-1984.
31.1 kg/m2 for men and 27.3 to less than 32.3 kg/m2
for women. Severe obesity was defined as 31.1 kg/m2
or greater for men and 32.3 kg/m2 or greater for
women.10 Alcohol use was calculated as the number
of drinks of alcohol consumed per day. This variable
was categorized as less than 1 drink/day, 1-2 drinks/
day, and 3 or more drinks/day. Dietary calcium,
potassium, and sodium data were divided into quar-
tiles based on the distribution of the entire sample,
and the quartile with the lowest consumption was
used as the reference category. Similarly, serum
levels of calcium, potassium, and sodium were di-
vided into quartiles based on the distribution of the
entire sample, with the lowest serum values used as
the reference category.
To describe differences in categorical baseline
variables among the four race and sex groups, we
used log-linear modeling. To test for differences in
continuous variables among the four race and sex
groups, we used analysis of variance. The relation-
ship of the study variables to the development of
hypertension was examined in several ways. For each
race and sex group, we determined the incidence of
hypertension for each level of each study variable.
Duferences in hypertension incidence were exam-
ined with x1 tests. Furthermore, we used logistic
regression analysis to examine the relation between
each individual study variable and hypertension sta-
tus. To examine the independent association be-
tween the study variables and hypertension status,
multiple logistic regression analysis was used.
The NHANES I study was designed as a multistage,
stratified probability survey. Oversampling of certain
groups (women of childbearing age, the elderly, and
persons living in low-income areas) was performed.
Rather than producing national prevalence estimates,
we wished to examine relations between risk factors
and mortality in a large community sample and, there-
fore, we did not use sample weights.
600 Hypertension Vol 18, No 5 November 1991

TABLE 1. Baseline Distribution of Study Variables in the NHANES I Study, 1971-1975

Black White Black White
men men women women
(=231) (n=2370) (n=523) (n=3,949) p Value
Variable n % n % n % n % Race Sex Interaction
Age(yr) 03147 0.0001 0.001
25-44 103 44.6 1,173 49.5 383 73.2 2,510 63.6
45-64 84 36.4 808 34.1 96 18.4 997 25.3
65 44 19.1 389 16.4 44 8.4 442 11.2
Income 0.0001 0.0016
<S10,000 157 68.0 1,021 43.1 411 78.6 1,859 47.1
$10,000-14,999 43 18.6 675 28.5 64 12.2 1,059 26.8
$15,000-19,999 16 6.9 339 14.3 33 6.3 552 14.0
$20,000-24,999 10 4.3 173 7.3 9 1.7 257 6.5
&$25,000 5 2.2 162 6.8 6 1.2 222 5.6
Education (yr) 0.0001 0.0001
<12 133 57.6 803 33.9 270 51.6 1,198 30.3
12 57 24.7 795 33.5 173 33.1 1,764 44.7
2:12 41 17.8 772 32.6 80 15.3 987 25.0
Nonleisure-timc physical activity 0.2222 0.0021 0.0044
Very active 131 56.7 1,185 50.0 243 46.5 1,844 46.7
Moderately active 91 39.4 988 41.7 226 43.2 1,830 46.3
Inactive 9 3.9 197 8.3 54 10.3 275 7.0
Leisure-time physical activity 0.0001 0.0001
Much exercise 50 21.7 668 28.2 49 9.4 659 16.7
Moderate exercise 78 33.8 987 41.7 142 27.2 1,647 41.7
Little or no exercise 103 44.6 715 30.2 332 63.5 1,643 41.6
Weight 0.0001 0.0020 0.001
Normal weight 174 75.3 1,873 79.0 318 60.8 3,138 79.5
Obese 35 15.2 366 15.4 128 24.5 577 14.6
Severe obesity 22 9.5 131 5.5 77 14.7 234 5.9
Alcohol use (drinks/day) 0.0417 0.0001 0.0079
<1 187 81.0 1,731 73.0 475 90.8 3,588 90.9
1-2 26 11J 372 15.7 28 5.4 282 7.1
2=3 18 7.8 267 11.3 20 3.8 79 2.0

Results
Out of the 14,407 cohort members of this study,
7,073 participants (231 black men, 2,370 white men,
523 black women, and 3,949 white women) were
available for analysis not involving serum or dietary
data after exclusions (Figure 1). Because of the
relatively small number of black participants in-
volved, some of the results, especially for black men,
are based on small numbers; therefore, the findings
should be interpreted with caution.
The distribution of categorical variables and means
with standard errors of continuous variables at base-
line for the four race and sex groups are presented in
Tables 1 and 2, respectively. The four groups are
characterized by differences in race or sex or both
race and sex distributions of all study variables.
The 10-year incidence of hypertension increased
with age and body mass index in all four groups and
decreased with increasing education in all groups
except black men (Table 3). Except among white
women, the incidence of hypertension was higher
among the least active group during leisure time than
among more active participants. The incidence of
hypertension varied little across serum magnesium
and alcohol consumption levels.
Mean systolic blood pressure was 123.0 mm Hg at
baseline and 125.1 mm Hg at follow-up. The change
in systolic blood pressure ranged from a decrease of
65 mm Hg to an increase of 91 mm Hg. Mean dia-
stolic blood pressure was 78.5 mm Hg at baseline and
76.0 mm Hg at follow-up. The change in diastolic
blood pressure ranged from a decrease of 61 mm Hg
to an increase of 68 mm Hg.
The univariate relations were examined simultane-
ously by multiple logistic regression analysis (Table
4). Increasing body mass index was the only variable
that was consistently associated with the develop-
ment of hypertension in all groups. In models where
we categorized body mass index, the odds ratios for
men with a body mass index of 27.3 or greater and
women with a body mass index of 27.8 kg/nr or
greater compared with those who had a smaller body
mass index were 2.78 for black men (p=0.0032), 1.64
 Ford and Cooper Risk Factors for Hypertension 601

TABLE 2. Means and Standard Deviations of Continuous Variables at Baseline: NHANES I-Epldemiologic FoUownp Study, 1971-1984

Black men White men Black women White women P Value

Variable (=231) (n=2,370) (n=523) (n=3,949) Race Sex
Cholesterol level (mg/dl) 215.849.0 215.443.2 209.1 44.9 213.747.7 0.0730 0.0578
Systolic blood pressure (mm Hg) 126.913.1 125.813.0 122.214.6 121.214.9 0.1599 0.0001
Diastolic blood pressure (mm Hg) 81.48.3 80.6+8.0 79.28.8 77.09.0 0.0001 0.0001
Resting heart rate (beats/min) 74.512.1 77.011.9 79.611.4 80.912.2 0.0030 0.0001
Dietary intake (#i=119) (n = 1349) (n=285) (n=2,418)
Calcium (mg/day) 704.8482.2 938.7+660.1 464.5382.2 657.2463.1 0.0001 <0.0001
Potassium (mg/day) 2,074.9l,281.7 2,674.6 1,267.4 l,439.1851.0 l,924.7902.5 0.0001 0.0381
Sodium (mg/day) 2,424.7 1,504.8 2,770.8 1,637.6 1,570.1 1,033.5 1,835.7 1,088.0 0.0001 <0.0001
Calories (NoVday) 2,227.8931.9 2,426.6 998.6 1,523.1 780.6 1,582.5637.4 0.0001 <0.0001
Serum cation levels (meq/1)
Calcium 9.80.5 9.70.4 9.61.0 9.60.4 0.4791 0.0001
Potassium 4.10.3 4.10.3 4.1 0.3 4.0 03 0.6581 0.0001
Sodium 142.42.8 142.73.4 141.43.6 142.23.1 0.2580 0.0029
Magnesium 1.60.2 1.70.1 1.60.2 1.70.1 0.0001 0.0001

for white men (p<0.0001), 1.84 for black women Discussion

0=0.0020), and 2.16 for white women (/><0.0001)
(data not shown). Education was inversely related to
the probability of hypertension developing among
white women and was of borderline significance
among white men and black women. Serum magne-
sium levels were inversely related to the incidence of
hypertension. White men who consumed 3 or more
drinks/day were at increased risk for hypertension
developing compared with white men who drank 1-2
drinks/day (odds ratio 1.55, 95% confidence limits
1.04-2.32). The regression coefficients for alcohol
use among black men suggested a similar association
but were not statistically significant. This is in con-
trast to white women, who had a lower risk of
hypertension if they consumed less than 1 drink/day.
Physical activity did not predict the development of
hypertension.
For the dietary analysis, 4,772 white subjects free
from hypertension at baseline and with complete
data for the dependent and independent variables
were available for analysis (Table 5). Because of
small numbers, the analyses were restricted to white
participants. Among white men (n = 1,709), no signif-
icantfindingswere observed in statistical models that
adjusted for age or age and caloric intake. Among
white women (n=3,063), calcium, potassium, and
sodium intake were all inversely related to hyperten-
sion in statistical models that adjusted for age only.
However, in models that adjusted for age and caloric
intake, no significant results were found. Body mass
index appears to confound the relation between
calcium and potassium intake and hypertension but
not that of sodium intake.
There was no evidence that baseline serum levels
of calcium, potassium, or sodium were predictive of
the future risk of hypertension in either univariate or
multivariate analysis (Table 6).
Relatively few epidemiological investigations have
prospectively examined potential causes of hyperten-
sion. We examined the relation of a number of
variables to hypertension development after a mean
follow-up period of 10 years. Education and body
mass index were the two variables that most consis-
tently predicted the incidence of hypertension across
all four race and sex groups.
A hypertension threshold of 160 mm Hg or more
for SBP or 95 mm Hg or more for DBP was chosen
for two reasons. First, during the period of this study,
the prevailing definition of hypertension was based
on this threshold. Only in 1984 was the definition
changed to a SBP of 140 mm Hg or more or a DBP of
90 mm Hg or more.1 Second, only one blood pressure
measurement was obtained at baseline. Conse-
quently, the possibility of misclassifying a participant
as hypertensive is greater at the lower threshold of
hypertension.
Various anthropometric measures have been asso-
ciated with blood pressure.11 Our study demonstrates
that, independent of age, body mass index was a risk
factor for hypertension. Data from the NHANES II
study showed that the prevalence of obesity (com-
bined overweight and severe overweight categories)
was 32.2% among white men, 35.9% among black
men, 34.9% among white women, and 63.1% among
black women.10 If the odds ratios for obesity previ-
ously calculated are assumed to be approximations of
the relative risk, then the population attributable risk
is 12.6% for white men, 23.0% for black men, 18.7%
for white women, and 28.8% for black women. These
data underscore the importance of weight control in
the prevention of hypertension, especially among
black men and women.
Several studies have shown that socioeconomic
status is inversely related to blood pressure.12-13 Our
results also show that participants with more ad-
602 Hypertension Vol 18, No 5 November 1991

TABLE 3. Ten-Year Incidence of Hypertension,,NHANES I-Epldemlologlc Followup Study, 1971-15*84

Black men White men Black women White women
(n = 231) (n=2,370) (n=523) (n=3,949)
Variable n % p Value n % p Value n % p Value n % p Value
Age (yr) 0.034 <0.001 0.011 <0.001
25-44 27 26.2 156 13.3 130 33.9 337 13.4
45-64 37 44.1 216 26.7 36 37.5 297 29.8
65 17 38.7 106 27.3 25 56.8 180 40.7
Income 0.021 0.415 0.368 <0.001
<$ 10,000 58 36.9 215 21.1 157 38.2 468 25.2
$10,000-14,999 7 16.3 138 20.4 21 32.8 178 16.8
$15,000-19,999 7 43.8 62 18.3 8 24.2 88 15.9
$20,000-24,999 6 60.0 27 15.6 2 22.2 44 17.1
2:$25,000 3 60.0 36 22.2 3 50.0 36 16.2
Education (yr) 0.116 <0.001 0.042 < 0.001
<12 54 40.6 195 24.3 113 41.9 354 29.6
12 15 26.3 159 20.0 62 35.8 317 18.0
212 12 293 124 16.1 16 20.0 143 14.5
Nonleisure-time physical activity 0.952 0.878 0.042 0.115
Very active 45 34.4 240 20.3 82 33.7 355 19.3
Moderately active 33 36.4 201 20.3 81 35.8 395 21.6
Inactive 3 33.3 37 18.8 28 51.9 64 23.3
Leisure-time physical activity 0.006 0.046 0.023 0.128
Much exercise 16 32.0 122 18.3 17 34.7 130 19.7
Moderate exercise 18 23.1 190 19.3 39 27.5 320 19.4
Little or no exercise 47 45.6 166 23.2 135 40.7 364 22.2
Weight 0.006 < 0.001 <0.001 <0.001
Normal weight 51 29.3 340 18.2 95 29.9 532 17.0
Obese 18 51.4 97 26.5 59 46.1 191 33.1
Severe obesity 12 54.6 41 31.3 37 48.1 91 38.9
Alcohol use (drinks/day) 0.868 0.111 0.944 0.470
<1 67 35.8 352 20.3 173 36.4 731 20.4
1-2 8 30.8 63 16.9 10 35.7 66 23.4
3 6 33.3 63 23.6 8 40.0 17 21.5
Magnesium level (meq/1) 0.486 0.836 0.451 0.426
sl.59 29 32.2 86 21.2 70 32.7 193 19.3
> 1.59-1.68 16 37.2 120 20.4 53 41.1 213 19.9
> 1.68-1.77 22 43.1 135 20.5 40 37.4 214 21.6
>1.77 14 29.8 137 19.1 28 38.4 194 21.8

vanced education were less likely to become hyper-
tensive than were participants with fewer years of
education. Income was not a particularly good pre-
dictor of hypertension.
Two studies have demonstrated an inverse associ-
ation between the incidence of hypertension and
physical activity orfitness.14-15Among 6,039 men and
women aged 20-65 years, those with low levels of
fitness had a relative risk of 1.52 for developing
hypertension (defined either as ever being told by a
physician that they had hypertension or as the use of
antihypertensive medications).14 A lack of strenuous
exercise was a risk factor for physician-diagnosed
hypertension developing among 14,998 male Harvard
alumni.15 Our present findings could not confirm the
results of these studies.
The relation between sodium use and blood pressure
has been much studied and continues to remain con-
troversial.16.17 The INTERSALT study, a large cross-
sectional study, has shown that a reduction in sodium
intake of 100 mmol/day would lead to a mean reduction
of 2.2 mm Hg in SBP and 0.1 mm Hg in DBP.18 Our
analysis of data from the NHANES I Epidemiologic
Followup Study shows no evidence that sodium con-
sumption was related to the incidence of hypertension
in a large sample of the US population. In this regard,
our study joins other within-population studies that
have failed to confirm a positive relation between
sodium consumption and blood pressure.
Also, the data from this study did not show a
reduction in the risk of the incidence of hypertension
with increased use of dietary potassium at baseline.
 Ford and Cooper Risk Factors for Hypertension 603

TABLE 3. Continued
Black men White men Black women White women
(n=231) (n =2370) (*=523) (n=3,949)
Variable n % p Value n % p Value n % pi Value n % p Value
Calories (No./day) 0.456 0.139 0.091 <0.001
s i , 230.85 8 33.3 33 25.4 83 43.2 255 25.3
> 1,230.85-1,671.09 12 41.4 50 18.7 46 37.4 212 22.7
>l,671.09-2,252.66 21 41.2 109 23.9 25 28.4 122 16.1
>2,252.66 23 29.1 168 19.7 17 32.1 56 153
Dietary calcium intake
(mg/day) 0.497 0.563 0.076 0.034
343.75 20 41.7 47 21.9 94 41.8 186 213
>343.75-591.11 17 37.8 78 20.4 35 29.2 183 22.7
>591.11-953.77 18 32.1 104 23.2 31 41.9 172 22.2
> 953.77 9 26.5 131 19.8 11 29.7 104 16.8
Dietary sodium intake (mg/day) 0.539 0.878 0.388 <0.001
1,161.8 13 35.1 41 21.4 78 40.0 232 25.0
>1,161.8-1,782.12 16 43.2 68 22.4 53 39.9 197 22.4
> 1,782.12-2,671.93 20 35.7 % 20.0 25 32.1 139 18.8
>2,671.93 15 28.3 155 21.2 15 30.0 77 14.9
Dietary potassium intake
(mg/day) 0.517 0.009 0.624 0.061
1,336.34 27 39.1 28 16.5 102 39.7 202 23.6
> 1336.34-1,919.28 12 36.4 93 26.9 34 33.3 189 21.7
> 1,919.28-2,633.59 14 36.8 102 21.8 18 33.3 157 19.8
> 2,633.59 11 25.6 137 18.9 17 39.5 97 18.0
Serum calcium level (mg/dl) 0.164 0.198 0.420 0.658
9.4 7 25.9 78 24.1 19 35.2 129 22.3
>9.4-9.7 10 52.6 83 21.0 10 26.3 101 20.6
>9.7-9.9 7 31.8 66 18.5 17 44.7 76 19.4
>9.9 8 24.2 44 17.7 11 35.5 38 19.2
Serum potassium level (meq/l) 0.234 0.694 0.184 0.744
3.9 5 333 31 17.0 9 36.0 61 17.5
>3.9-4.1 3 333 31 18.2 2 11.8 40 19.6
>4.1-43 2 18.2 32 20.5 2 22.2 34 19.3
>4.3 7 58.3 23 19.7 7 43.8 33 21.6
Serum sodium level (meq/l) 0.934 0.121 0.227 0.475
141 5 38.5 31 16.9 4 16.7 56 19.2
>141-143 5 29.4 41 21.7 11 40.7 61 20.5
>143-144 4 40.0 24 20.9 2 22.2 19 143
>144 3 37.5 22 12.7 3 42.9 33 203

Although we did note an inverse relation between
the two variables among white women in univariate
analysis, this relation no longer persisted in multi-
variate analysis. Previous NHANES I study analyses
demonstrated a cross-sectional association between
dietary potassium intake and blood pressure.19'20
Other cross-sectional and experimental studies have
also shown an inverse relation.18-21
The association between dietary calcium intake
and blood pressure and the potential benefit of
treating hypertension with increased dietary intake
of calcium remain controversial.162223 Previous
analyses of the NHANES I and II data have yielded
conflicting findings regarding the influence of vari-
ous dietary variables on blood pressure, particularly
calcium.192024-25 Other cross-sectional studies have
shown relations between dietary calcium intake and
blood pressure.2026"35 Data from the Western Elec-
tric Heart Study36 showed that calcium intake was
inversely related to the incidence of elevated DBP
(95 mm Hg or greater) but not of elevated SBP (160
mm Hg or greater). A recent report of a large
cohort study of women showed that dietary calcium
intake was inversely related to hypertension among
women.37
Although alcohol use was shown to be related to
blood pressure in previous cross-sectional analyses of
NHANES I data,20-24 we found conflicting evidence
regarding the role of alcohol consumption in the inci-
dence of hypertension. Among white men, and perhaps
604 Hypertension Vol 18, No 5 November 1991

TABLE 4. Multiple Logistic Regression Analysis Results: NHANES I-Epidemiologk Follownp Study, 1971-1984
White men (n=2,370) Black men (n=231) White women (/i =3,949) Black women (*=523)
Variable P SE p Value P SE p Value P SE p Value P SE p Value
Age(yr) 0.0256 0.0041 <0.0001 0.0165 0.0118 0.15% 0.0405 0.0035 < 0.0001 0.0112 0.0095 0.2384
Household income ($1,000s) 0.0116 0.0077 0.1326 0.0510 0.0269 0.0577 -0.0006 0.0064 0.9210 0.0122 0.0194 0.5312
Education (yr) -0.0348 0.0181 0.0551 -0.0583 0.0456 0.2010 -0.0704 0.0162 <0.0001 -0.0686 0.0389 0.0775
Resting heart rate (beats/min) 0.0100 0.0044 0.0243 0.0211 0.0124 0.0896 0.0046 0.0034 0.1843 0.0025 0.0085 0.7635
Cholesterol level (mg/dl) 0.0023 0.0012 0.0653 -0.0008 0.0032 0.8055 0.0005 0.0010 0.5788 0.0030 0.0024 0.2082
Body mass index (kg/m 2 ) 0.0639 0.0142 <0.0001 0.0746 0.0367 0.0420 0.0768 0.0088 <0.0001 0.0588 0.0174 0.0007
Alcohol use (drinks/day)
(Reference=l-2)
<1 0.1617 0.1554 0.2983 0.2478 0.4824 0.6075 -0.3530 0.1566 0.0242 -0.0478 0.4258 0.9106
2:3 0.4375 0.2054 0.0332 0.5036 0.7136 0.4803 -0.2739 0.3237 0.3975 0.3861 0.6366 05442
Leisure-time physical activity
(Reference=most active)
Moderate -0.0123 0.1385 0.9292 -0.6408 0.4460 0.1508 -0.1117 0.1270 03793 -0.4490 0.3812 0.2390
Sedentary 0.1564 0.1436 0.2761 0.3958 0.3909 0.3113 -0.1249 0.1285 03313 -0.0088 0.3449 0.9797
Nonleisure-time physical activity
(Reference=most active)
Moderate -0.0758 0.1168 0.5166 0.0916 0.3221 0.7761 0.0167 0.0908 0.8542 0.0812 0.2073 0.6952
Sedentary -0.1647 0.2100 0.4327 -0.4324 0.7999 0.5888 -0.0638 0.1702 0.7076 05142 0.3262 0.1149
Magnesium level (meq/1) -0.7485 0.4129 0.0698 0.6969 0.9402 0.4585 -0.2684 0.3192 0.4005 0.4121 0.6462 0.5236
SE, standard error.

among black men as well, the risk of hypertension
developing was highest among those who drank an
average of 3 or more drinks/day compared with those
who drank 1-2 drinks each day. However, among white
women, a lower risk of hypertension was found among
those who did not drink at all (odds ratio 0.70, 95%
confidence limits 0.52-0.95) or who drank less than 1
drink/day compared to women who had 1-2 drinks/day.
This agrees with findings from a large cohort study of
women.38 However, unlike that study, we could not
demonstrate a relation between an increased risk of
hypertension and increased alcohol consumption.
Several study limitations must be considered in the
interpretation of our findings. First, blood pressure was
determined by a single measurement taken at the
baseline examination. In most studies, three measure-

TABLE 5. Logistic Regression Analysis* Results Relating Dietary Cations to Hypertension: NHANES I-Epldemiologic Followup Study,
1971-1984
White men (n = 1,709) Black men (n = 183) White women (n =3,063) Black women (r1=456)
Variable P SE p Value P SE p Value P SE p Value P SE p Value
Continuous levels
Calcium (per 100 meq/1) 0.0076 0.0109 0.4859 -0.0681 0.0463 0.1415 -0.0019 0.0128 0.8834 -0.0098 0.0359 0.7842
Potassium (per 100 meq/1) 0.0035 0.0066 05892 -0.0302 0.0186 0.1053 -0.0029 0.0069 0.6740 0.0100 0.0155 0.5181
Sodium (per 100 meq/1) 0.0057 0.0046 0.2176 -0.0271 0.0170 0.1122 -0.0062 0.0058 0.2838 -0.0030 0.0134 0.8258
Quartilesf:
Calcium intake
Quartile 2 -0.0937 0.2120 0.6587 -0.3787 0.4494 0.3994 0.1085 0.1259 0.3886 -0.5001 0.2585 0.0530
Quartile 3 0.0845 0.2057 0.6811 -0.6153 0.4605 0.1815 0.1528 0.1326 0.2491 0.1082 0.3006 0.7189
Quartile 4 -0.0398 0.2113 0.8505 -1.0549 0.6020 0.0797 0.0085 0.1590 0.9573 -0.2608 0.4365 05501
Potassium intake
Quartile 2 0.6272 0.2438 0.0101 - 0 3 4 2 8 0.4728 0.4684 -0.1387 0.1243 0.2645 -0.1441 0.2654 05873
Quartile 3 03858 0.2428 0.1121 -0.2391 0.4691 0.6104 -0.1052 0.1362 0.4399 -0.1029 03463 0.7663
Quartile 4 0.2758 0.2549 0.2792 -0.7632 05341 0.1530 -0.0676 0.1702 0.6912 0.4961 0.4151 0.2320
Sodium intake
Quartile 2 0.0584 0.2273 0.7972 0.2508 05084 0.6218 -0.1074 0.1201 0.3712 0.1037 0.2443 0.6712
Quartile 3 0.0153 0.2176 0.9439 0.0860 0.4829 0.8587 -0.1793 0.1369 0.1903 -0.1934 03109 05339
Quartile 4 0.1589 0.2225 0.4750 -0.2872 05824 0.6219 -0.3188 0.1741 0.0671 0.0424 0.4077 0.9172
SE, standard error.
Covariates in each model include age and caloric intake.
tFor quartile cutpoints, see Table 3.
 Ford and Cooper Risk Factors for Hypertension 605

TABLE 6. Multiple Logistic Regression Analysis Results Relating Serum Cation Levels to Hypertension : NHANES I-Epidemiologic
Followup Study, 1971-1984*
White men (n=610) White women (n=828)
Variable P SE p Value P SE p Value
Model 1: Continuous
Calcium (mg/dl) -0.0607 0.2494 0.8079 -0.2585 0.2284 0.2577
Potassium (meq/1) -0.1863 03739 0.6182 -0.1033 0.3134 0.7418
Sodium (meq/1) -0.0026 0.0352 0.9405 0.0023 0.0320 0.9436
Model 2: Quartiles
Calcium
Quartile 2 0.1319 0.3010 0.6613 0.0460 0.2365 0.8459
Quartile 3 -0.0773 0.3121 0.8043 0.0732 0.2607 0.7790
Quartile 4 -0.1886 0.3690 0.6092 -0.4775 0.3619 0.1871
Potassium
Quartile 2 -0.0277 03063 0.9280 0.0418 0.2521 0.8684
Quartile 3 0.0176 03146 0.9554 -0.2726 0.2695 0.3118
Quartile 4 -0.3296 0.3375 03288 -0.1390 0.2835 0.6238
Sodium
Quartile 2 0.4299 0.2885 0.1362 0.1474 0.2336 0.5279
Quartile 3 03115 0.3500 0.3734 -0.5311 0.3237 0.1009
Quartile 4 -0.3355 0.3456 0.3317 -0.2025 0.2898 0.4847
SE, standard error.
Adjusted for age, income, education, heart rate, serum cholesterol level, body mass index, alcohol use, physical activity, serum
magnesium and other cation levels, and caloric intake.

ments are taken on one or more occasions, with the
average of the last two measurements used to provide
greater stability of the data. Also, the questions asked
of the augmented sample used to classify participants
as hypertensive differed somewhat. This difference may
have resulted in some people (but probably not very
many) being misclassified as normotensive.
The measurement of several of the independent
variables poses greater concerns. Dietary data were
collected at a single 24-hour recall. With this method,
the intraindividual variability exceeds the interindi-
vidual variability and therefore, it may not be suffi-
ciently accurate to classify persons correctly. The
impact of measurement error using this technique on
potential associations has been discussed previous-
ly.3940 We tried to minimize misclassification by
performing a quartile analysis, but the results from
this analysis did not give strong support for any effect
of intake of dietary cations (Table 5).
The classification of physical activity levels in this
data set is unsophisticated and probably resulted in
significant misclassification. Participants were simply
asked to classify themselves, without a frame of
reference, in one of three activity categories. There-
fore, each participant's belief of what constituted
being active or sedentary dictated an answer. The
reliability of these answers has not been tested, and
the indirect validity of these data in the NHANES I
study is limited.41 The usefulness and limitations of
simplified physical activity questionnaires have been
discussed elsewhere.42-43
Finally, the relatively small numbers of black men
and women reduced the power of the study to detect
associations in these groups.
In conclusion, body mass index and education were
generally the most consistent predictors for the de-
velopment of hypertension after an average of 10
years of follow-up among the four race and sex
groups. These data confirm the importance of weight
control in the prevention of hypertension and pro-
vide a clear target for intervention by both the
physician in his or her daily practice and by persons
in the public health sector. No strong evidence for an
effect of dietary sodium, potassium, or calcium intake
on the incidence of hypertension was observed. How-
ever, the failure of our study to support findings
relating intake of dietary cations to the development
of hypertension may be attributable to imprecision in
the measurement of dietary data and misclassifica-
tion of hypertension status. Future longitudinal stud-
ies using better quantification of dietary variables
should address the importance of dietary intake of
cations on hypertension in the general population.
References
1. Hypertension prevalence and the status of awareness, treat-
ment, and control in the United States: Final report of the
subcommittee on definition and prevalence of the 1984 joint
national committee. Hypertension 1985;7:457-468
2. Rose G: Strategy of prevention: Lessons from cardiovascular
disease. BrMedJ 1981;282:1847-1851
3. National Center for Health Statistics. Plan and operation of
the Health and Nutrition Examination Survey, United States
1971-1973. DHEW publication No. (PHS) 79-1310. (Vital and
Health Statistics, series 1, number 10a). Washington, DC,
National Center for Health Statistics, 1973
4. National Center for Health Statistics. Plan and operation of the
Health and Nutrition Examination Survey, United States
1971-1973. DHEW publication No. (HRA) 77-1310. (Vital and
Health Statistics, series 1, number 10b). Washington, DC,
National Center for Health Statistics, 1977
5. National Center for Health Statistics. Plan and operation of
the HANES I Augmentation Survey of adults 25-74 years,
606 Hypertension Vol 18, No 5 November 1991
United States, 1974-1975. DHEW publication No. (PHS)
78-1314. (Vital and Health Statistics, series 1, number 14).
Washington, DC, National Center for Health Statistics, 1978
6. National Center for Health Statistics. Plan and operation of
the NHANES I Epidemiologic Followup Study 1982-84.
DHHS publication No. (PHS) 87-1324. (Vital and Health
Statistics, series 1, number 22). Washington, DC, National
Center for Health Statistics, 1987
7. Madans JH, Kleinman JC, Cox CS, Barbano HE, Feldman JJ,
Cohen B, Finucane FF, Cornoni-Huntley J: 10 Years after
NHANES I: Report of initial followup, 1982-84. Public Health
Rep 1986;101:465-473
8. Cornoni-Huntley J, Barbano HE, Brody JA, Cohen B, Feld-
man JJ, Kleinman JC, Madans J: National Health and Nutri-
tion Examination I Epidemiologic Followup Survey. Public
Health Rep 1983;98:245-251
9. Bordley J, Connor C, Hamilton WF, Kerr WJ, Wiggers CJ:
Recommendations for human blood pressure determination
by sphygmomanometers. Circulation 1951;4:503-509
10. National Center for Health Statistics. Anthropometric refer-
ence data and prevalence of overweight, United States,
1976-80. DHHS publication No. (PHS) 87-1688. (Vital and
Health Statistics, series 11, no. 238). Washington, DC,
National Center for Health Statistics, 1987
11. Fraser GE: Preventive cardiology. New York, Oxford Univer-
sity Press, 1986, pp 127-147
12. Kraus JF, Borhani ND, Franti CE: Socioeconomic status,
ethnicity and risk of coronary heart disease. Am J Epidemiol
1980;lll:407-414
13. Rose GA, Marmot MG: Social class and coronary heart
disease. Br Heart J 1981;45:13-25
14. Blair SN, Goodyear NN, Gibbons LW, Cooper KH: Physical
fitness and incidence of hypertension in healthy normotensive
men and women. JAMA 1984;252:487-490
15. Paffenbarger RS Jr, Wing AL, Hyde RT, Jung DL: Physical
activity and incidence of hypertension in college alumni. Am J
Epidemiol 1983;117:245-257
16. Kesteloot H: Kationen en bloeddruk. Ned Tijdschr Geneeskd
1985;129:630-634
17. Swales JD: Salt saga continued: Salt has only small importance
in hypertension. Br Med J 1988;297:307-308
18. Intersalt Cooperative Research Group: INTERSALT: An
international study of electrolyte excretion and blood pres-
sure: Results for 24 hour urinary sodium and potassium
excretion. Br Med J 1988;297:319-328
19. McCarron DA, Morris C, Henry H, Stanton JL: Blood pres-
sure and nutrient intake in the United States. Science 1984;
224:1392-1398
20. Gruchow HW, Sobocinski KA, Barboriak JJ: Alcohol, nutrient
intake, and hypertension in US adults. JAMA 1985;253:
1567-1570
21. Smith WCS, Crombie IK, Tavendale RT, Gulland SK, Tun-
stall-Pedoe HD: Urinary electrolyte excretion, alcohol con-
sumption, and blood pressure in the Scottish heart health
study. BrMedJ 1988;297:329-330
22. Maxwell MH, Waks AU: Cations and hypertension: Sodium,
potassium, calcium, and magnesium. Med Clin North Am
1987;71:859-875
23. Langford HG: Nonpharmacologic therapy of hypertension:
Commentary on diet and blood pressure. Hypertension 1989;
13(suppl I):I-98-M02
24. Harlan WR, Hull AL, Schmouder RL, Landis JR, Thompson
FE, Larkin FA: Blood pressure and nutrition in adults. Am J
Epidemiol 1984;120:17-28
25. Sempos C, Cooper R, Kovar MG, et al: Dietary calcium and
blood pressure in National Health and Nutrition Examination
Surveys I and II. Hypertension 1986;8:1067-1074
26. Kesteloot H, Geboers J: Calcium and blood pressure. Lancet
1982;1:813-815
27. Stanton JL, Braitman LE, Riley AM Jr, Khoo CS, Smith JL:
Demographic, dietary, life style, and anthropometric corre-
lates of blood pressure. Hypertension 1982;4(suppl III):III-
135-III-142
28. Ackley S, Barrett-Connor E, Suarez L: Dairy products, cal-
cium, and blood pressure. Am J Clin Nutr 198338:457-461
29. Garcia-Palmieri MR, Costas R Jr, Cruz-Vidal M, Sorlie PD,
Tillotson J, Havlik RJ: Milk consumption, calcium intake, and
decreased hypertension in Puerto Rico: Puerto Rico Heart
Health Program Study. Hypertension 1984;6:322-328
30. Nichaman M, Shekelle R, Paul O: Diet, alcohol, and blood
pressure in the Western Electric Study (abstract). Am J
Epidemiol 1984;120:469-470
31. Reed D, McGee D, Yano K, Hankin J: Diet, blood pressure,
and multicollinearity. Hypertension 1985;7:405-410
32. Kromhout D, Bosschieter EB, Coulander C: Potassium, cal-
cium, alcohol intake and blood pressure: The Zutphen Study.
Am J Clin Nutr 1985;41:1299-1304
33. Kok FJ, Vandenbrouke JP, van der Heide-Wessel C, van der
Heide R: Dietary sodium, calcium, potassium, and blood
pressure. Am J Epidemiol 1986;123:1043-1048
34. Liebman M, Chopin LF, Carter E, Clark AJ, Disney GW,
Hegsted M, Kenney MA, Kirmani ZA, Koonce KL, Korslund
MK, Moak SW, McCoy H, Stallings SF, Wakefield T: Factors
related to blood pressure in a biracial adolescent female
population. Hypertension 1986;8:843-850
35. Joffres MR, Reed DM, Yano K: Relationship between mag-
nesium intake and other dietary factors to blood pressure: The
Honolulu Heart Study. Am J Clin Nutr 1987;45:469-475
36. Zhang HY, Liu K, Shekelle R, Dyer A, Stamler J: The impact
of calcium intake on incidence of elevated blood pressure: The
Western Electric Study (abstract). Am J Epidemiol 1988;128:
916-917
37. Witteman JC, Willett WC, Stampfer MJ, Colditz GA, Sacks
FM, Speizer FE, Rosner B, Hennekens CH: A prospective
study of nutritional factors and hypertension among US
women. Circulation 1989;80:1320-1327
38. Witteman JC, Willett WC, Stampfer MJ, Colditz GA, Kok FJ,
Sacks FM, Speizer FE, Rosner B, Hennekens CH: Relation of
moderate alcohol consumption and risk of systemic hyperten-
sion in women. Am J Cardiol 1990;65:633-637
39. Liu K, Stamler J, Dyer A, McKeever J, McKeever P: Statistical
methods to assess and minimize the role of intra-individual
variability in obscuring the relationship between dietary lipids
and serum cholesterol. / Chronic Dis 1978^1:399-418
40. Sempos CT, Johnson NE, Smith EL, Gilligan C: Effects of
intraindividual and interindividual variation in repeated
dietary records. Am J Epidemiol 1985;121:120-130
41. Farmer ME, Locke BZ, Moscicki EK, Dannenberg AL, Lar-
son DB, RadlofT LS: Physical activity and depressive symp-
toms: The NHANES I Epidemiologic Follow-up Study. Am J
Epidemiol 1989;128:1340-1351
42. Slater CH, Green LW, Vernon SW, Keith VM: Problems in
estimating the prevalence of physical activity from national
surveys. Prev Med 1987;16:107-118
43. Washburn RA, Adams LL, Haile GT: Physical activity assess-
ment for epidemiologic research: The utility of two simplified
approaches. Prev Med 1987;16:636-646
KEY WORDS body mass index calcium diet education
risk factors magnesium potassium population
study sodium