Trends in the Prevalence, Awareness, Treatment, and Control of Hypertension in the Adult US Population

The National Health and Nutrition Examination Survey (NHANES) is conducted by the National Center for Health Statistics, Centers for Disease Control and Prevention, to provide information on the health and nutritional status of the civilian noninstitutionalized population of the United States. Data from four previous surveys and the first phase of the current survey, NHANES III, are used to report trends in blood pressure (BP) distribution and hypertension prevalence from 1960 to 1991. In addition, trends in the percentage of the population with hypertension who were aware of their condition, who were treated with an antihypertensive drug, and whose hypertension was controlled with an antihypertensive drug (subsequently termed rates of awareness, treatment, and control) are presented.

Data from three previous national surveys, spanning the interval from 1960 to 1980,¹ revealed a downward trend in mean systolic BP and hypertension prevalence. Rates of awareness, treatment, and control increased progressively in succeeding surveys.² Provisional data from the first 3 years of NHANES III (1988-1991) confirmed and extended this pattern.³⁴

Hypertension is a major risk factor for coronary heart disease, stroke, congestive heart failure, renal insufficiency, and peripheral vascular disease.⁵ For 23 years the National High Blood Pressure Education Program has attempted to increase public and professional awareness regarding the detection, evaluation, treatment, and control of hypertension.³ The goal of these efforts has been to reduce hypertension-related death and disability. Mortality from coronary heart disease and stroke has decreased dramatically from the 1960s to the present. Quantitation of the status of hypertension prevalence, awareness, treatment, and control is not only essential to monitoring the success of education programs but is also important for predicting and influencing the health of the nation.⁶

Methods

Survey Design

The first National Health Examination Survey (NHES I) and NHANES were designed to produce data regarding the health and nutrition profile of the civilian noninstitutionalized population of the United States. NHES I collected data from 1960 to 1962⁷ ; the subsequent NHANES I, II, and III (phase 1) collected data during 1971-1974, 1976-1980, and 1988-1991, respectively.⁸⁹¹⁰ In addition to the national surveys, the Hispanic Health and Nutrition Examination Survey (HHANES) collected data on Mexican Americans living in five southwestern states (Arizona, California, Colorado, New Mexico, and Texas).¹¹ Although HHANES was not a national survey, the Mexican American sample was a probability sample of the civilian noninstitutionalized Mexican American population living in the Southwest. This universe represented approximately 84% of the 1980 US Mexican American population. Therefore, HHANES should provide a reasonable comparison with the national Mexican American sample in NHANES III. All five surveys used stratified multistage probability sample designs. The design of the surveys mandates the use of sampling weights to derive national estimates. Data were age adjusted using the direct method to the 1990 adult noninstitutionalized population aged 18 to 74 of the United States. Detailed descriptions of the contents and design of each survey have been published.^789101112

Trends in the estimates of BP distribution and hypertension prevalence in the general population, as well as rates of awareness, treatment, and control, are based on comparisons of findings in the five surveys. Among these surveys, there was variation in sample size within each subgroup of the population, in the protocol for BP measurement, and in the potential for measurement error.

Table 1 presents the numbers of people aged 18 to 74 years who were sampled and examined in the four national surveys. Each survey had an insufficient sample size to report on races other than black and white separately, but other races are included in the overall estimates throughout this article. The response rates among adults for the examination component of the four national surveys were 87%, 70%, 69%, and 77% for NHES I, NHANES I, NHANES II, and NHANES III, respectively. The number of Mexican Americans sampled and examined from HHANES is given in Table 4; the response rate among adults was 77%.

BP Measurement

The five surveys differed in the protocols for BP measurement.^{1314151617181920} Table 2 summarizes these differences. BPs were measured solely in mobile examination centers (MECs) until NHANES III. Three BP readings were obtained on each occasion in NHES I, NHANES II, and NHANES III. In NHANES I, one third of the participants had three BP measurements and two thirds had one. In HHANES, two BP measurements were taken. In NHANES I and II, BP was measured twice with subjects in the seated position and once in the supine position in individuals with three measurements. In the other surveys, all BP measurements were taken with subjects in the seated position. In NHANES III, BP measurements were taken on two occasions. The first set of three BPs was measured in the home by a lay interviewer and the second by a physician at the MEC. For the present trend analysis, only BP measurements taken with subjects in the seated position obtained during the MEC examinations have been included. Data for both occasions from NHANES III are presented elsewhere to describe hypertension prevalence in adults aged 18 and older.¹⁹ In summary, NHANES I estimates are based either on a single measurement taken with subjects in the seated position or on the mean of two measurements taken with subjects in the seated position. HHANES and NHANES II estimates are based on the mean of two measurements taken with subjects in the seated position and NHES I and NHANES III estimates on the mean of three measurements with subjects in the seated position.

Training of the BP observers was not documented in NHES I, NHANES I, or NHANES II. Previous publications reported that the observers received no standardized training for the first three surveys.^1314161718 Formal training was instituted in HHANES, with a 1-day training period. Quality-control measures included quarterly site visits, during which the accuracy of BP measurements obtained by field observers was confirmed by an experienced independent observer using a double-headed stethoscope. The training period for NHANES III lasted 1 days. Satisfactory performance using standardized videotapes and simultaneous measurement with an instructor using a double-headed stethoscope were required at the end of the initial training and quarterly thereafter.²⁰ The proportions of zero end digits were an indication of the success of the standardized training efforts. Before HHANES, the proportion of zero end digits ranged from one third in NHES I to more than one half in NHANES II; in HHANES and NHANES III, the proportion ranged from 19% to 27%.

In evaluating trends in BP levels, the lack of a variety of sizes of BP cuffs in the first three surveys is of major concern. A large proportion of the US adult population is obese.²¹²² For example, the prevalence of overweight (defined as a body mass index [BMI] greater than or equal to the 85th percentile for women aged 20 to 29 years) was 43.8% for black women during NHANES II. Use of an inappropriately small BP cuff yields falsely high readings for both systolic and diastolic BP measurements.²³ In NHES I, only an adult-sized BP cuff was available; for NHANES I and II, a child-sized cuff was added. In HHANES and NHANES III, four sizes were available: child, adult, large adult, and thigh.

Table 2 presents the published definitions of diastolic BP recorded in each survey. For NHANES III, phase 1, all observers were trained to record the level of mercury 2 mm below where they heard the last sound. How the published definitions of diastolic BP in the other surveys were interpreted by the examining physicians was not documented. The published definition of the systolic BP level has remained standard over the five surveys as the point at which the first Korotkoff sound was heard. In NHANES III, observers were trained to record the first of two consecutive sounds as the first sound.

The protocol for BP measurement varied in several other ways from survey to survey. The temporal sequence of the BP measurement within the entire examination and within the physician’s component was different in each survey. In NHES I, as many as 39 resident physicians were hired as temporary employees to measure BP. In contrast to the other surveys, NHES I physicians were instructed to use the bell of the stethoscope for auscultation of BP and to record both muffling and disappearance of sound for the diastolic measurements. A nurse placed the cuff on the arm and positioned the sphygmomanometer for the physician.

In NHANES I, 23 physicians measured BP, and approximately one third of the BP measurements were taken with an aneroid sphygmomanometer. In all other surveys, a standard mercury sphygmomanometer was used. In HHANES and NHANES III, the maximum inflation level²⁴ was determined before the first BP was measured. For the Mexican American sample of HHANES, two physicians performed 96% of the BP measurements. In NHANES III, five physicians each took approximately 16% or 33% of the BP measurements.

The defined threshold for hypertension changed during the interval from 1960 to 1991.²⁵²⁶ Therefore, rates of prevalence, awareness, treatment, and control are presented using two definitions: (1) hypertension or high BP is defined as a single BP measurement obtained with subjects in the seated position or the mean of two or more BP measurements obtained with subjects in the seated position with systolic BP >140 mm Hg and/or diastolic BP >90 mm Hg and/or current treatment with an antihypertensive drug; and (2) systolic BP >160 mm Hg and/or diastolic BP >95 mm Hg and/or current antihypertensive drug treatment. Awareness reflects reporting of a prior diagnosis of hypertension or high BP among those people with hypertension according to one of these definitions. Treatment is defined as a history of current antihypertensive drug therapy at the time of the interview. Control is defined as hypertension with systolic BP <140 (160) mm Hg and diastolic BP <90 (95) mm Hg in the setting of concurrent antihypertensive drug therapy. Awareness and treatment are prerequisites to control; therefore, the rate of control in the subgroup of people with treated hypertension is presented.

Results

Hypertension Prevalence

Trends in the age-adjusted prevalence of hypertension varied from one population group to another (Table 3). During the 1970s (NHANES I and II), the age-adjusted prevalence at ≥160/95 mm Hg remained stable. Although black and white women experienced a small decline in prevalence during this interval, the main exception to this relatively stable pattern occurred among black men, who experienced an 8.5% decrease between NHANES I and II. There was a substantial decline in the prevalence of hypertension (≥160/95 mm Hg) between NHANES II and III. This was consistent across the four major race-sex groups (black and white men and women). When 140/90 mm Hg was used as the BP threshold to classify hypertension, a relatively uniform decrease across all population groups was noted between NHANES I and II, with an additional and greater decline between NHANES II and III.

Race-sex-age–specific rates of hypertension for both thresholds were at their lowest at NHANES III in almost all age categories. For the 140/90 mm Hg cut point, most of the race-sex-age–specific subgroups experienced their highest prevalence during NHANES I. Between NHANES II and III, the prevalence of hypertension (140/90 mm Hg) among black men aged 50 to 59 years remained relatively stable at 54.7% and 53.3%, respectively. The prevalence for black men aged 60 to 74 increased from 67.0% to 71.2%. In every other age-specific group, hypertension prevalence between NHANES II and III declined markedly. The proportionate decrease was greatest among 18- to 29-year-olds and least among 60- to 74-year-olds.

During the 7-year interval between HHANES and NHANES III, a small decline was seen in the prevalence of hypertension in Mexican Americans (Table 4). The age-adjusted prevalence for Mexican Americans was similar to that for the white US population in NHANES III (Table 3).

Mean BP Levels

Age-adjusted mean systolic BP increased 2 to 3 mm Hg between NHES I and NHANES I in all population subgroups except black women. Thereafter, a progressive decline occurred from NHANES I to NHANES III (Table 5). For black men and women, the decline in age-specific systolic BP tended to be steepest between NHANES I and II; the decline in white men and women was greatest between NHANES II and III for most age groups. Age-adjusted diastolic BP increased between NHES I and NHANES I and then declined in all subgroups between NHANES I and III. Age-specific diastolic BP also declined markedly and progressively between NHANES I and III for every group.

Hypertension Awareness, Treatment, and Control

During the interval between NHANES II and III, the BP threshold recommended for the definition of hypertension changed from 160/95 to 140/90 mm Hg.²⁵ Consequently, the rates of awareness, treatment, and control at the 140/90 mm Hg threshold are presented only for the NHANES II and III surveys. Rates of awareness, treatment, and control of hypertension defined as ≥160/95 mm Hg or current treatment with antihypertensive medication remained stable between NHES I and NHANES I, with the exception of black men, for whom all three rates increased from their comparatively very low rates in 1960-1962 (Table 6), which were estimated from a small sample size (Table 1). The other three race-sex groups experienced marked increases in hypertension awareness, treatment, and control during the relatively short interval between NHANES I (1971-1974) and NHANES II (1976-1980), although further improvements for black men were smaller and somewhat inconsistent.

During the 12-year interval between NHANES II and III, awareness of hypertension (≥160/95 mm Hg) reached at least 83% for every subgroup; black women had already achieved a rate of 88% at the NHANES II evaluation. Awareness at the 140/90 mm Hg threshold increased from one half (51%) to almost three quarters (73%) of the population with hypertension between NHANES II and III. Increases in awareness were greater for whites than blacks during this interval. Awareness was higher for women than men: 67% of black and white men and 79% and 82% of black and white women, respectively.

Trends in the rate of treatment with antihypertensive drugs in the population with hypertension have mirrored trends in awareness. In NHANES III, the large difference in awareness at the 140/90 mm Hg cut point between men and women was associated with a 19% difference in treatment rates (46% and 65%, respectively). Higher rates of treatment (as well as awareness and control) of hypertension in women compared with men were apparent in both blacks and whites. During NHANES III, black and white women had identical treatment rates (65%) for hypertension defined as ≥140/90 mm Hg; the corresponding rates were 78% and 88% for black and white women, respectively, at the 160/95 mm Hg cut point. For both cut points, increases in treatment rates between NHANES II and III were somewhat greater for whites than for blacks.

Between NHANES I and III, the rates of control increased fourfold for hypertension at the 160/95 mm Hg threshold. It increased nearly threefold at the 140/90 mm Hg cut point in the 12 years between NHANES II and III, remaining, however, only at 29%. All groups experienced a proportionally greater increase in the percent controlled between NHANES II and III, except for black women. During this interval, improvement was greater for whites than blacks, especially among women.

When only people with treated hypertension are considered, the rate of control to <160/95 mm Hg almost doubled between NHANES I and III, from 42% to 82%, with a similar increase in each of the four race-sex subgroups. An increase in control from approximately one third (32%) to more than one half (55%) occurred for the 140/90 mm Hg threshold between NHANES II and III. The control rates for people with treated hypertension were 47% and 49% in black men and women and 50% and 60% in white men and women, respectively.

Treatment Effect in the Populations

Hypertension prevalence was low in the youngest age group for every race-sex subgroup, in contrast to extremely high rates in the oldest age group (Table 3). Therefore, the distribution of systolic BP in these two extreme population subgroups should reflect different influences over time, especially with regard to antihypertensive treatment. Changes in the youngest subgroup should be mostly independent of treatment, whereas changes in BP distribution in the group aged 60 to 74 years should most strongly reflect a treatment effect. Fig 1 compares trends for the smoothed weighted frequency distributions for the population aged 18 to 29 with those for ages 60 to 74. Before NHANES III, the distribution of systolic BP for the population aged 18 to 29 was very similar, with NHES I and NHANES II almost superimposed. In NHANES III, the distribution was shifted downward by an average of about 4 to 5 mm Hg compared with NHES I and NHANES II and by about 8 mm Hg compared with NHANES I. In contrast, the older population has shown a steady continuous downward shift across each successive survey. The difference between the median and 90th percentile of the 60- to 74-year age group narrowed between NHES I and NHANES III from 44 to 28 mm Hg.

Measurement Error

To assess the potential effects of measurement error related to the absence of large BP cuffs before HHANES, we undertook an analysis that controlled partially for the increasing prevalence of obesity. Quintile cutoffs for BMI—defined as weight divided by height squared—were calculated from NHANES III data. Mean systolic and diastolic BP values (Fig 2), prevalence of BP >140/90 mm Hg, treatment rates of people with hypertension, and the total proportion of the adult population treated for high BP (Table 7) were calculated within each BMI quintile. The same BMI cutoffs were applied to NHANES II data. For both systolic and diastolic BP values, the difference in mean BP between NHANES II and III was greater at higher levels of BMI (Fig 2). Although the absolute difference in the age-adjusted prevalence of hypertension between NHANES II and III increased with increasing BMI, the relative decrease within quintiles did not show a trend (Table 7). The percentage of the population reporting treatment with antihypertensive drugs within BMI quintiles remained virtually unchanged between NHANES II and III, indicating that the propensity to treat the obese population for hypertension did not increase or decrease in this interval between 1976-1980 and 1988-1991. Therefore, the decreased overall prevalence is not likely to be due to changes in treatment propensity. The percentage of people with treated hypertension increased from 24% to 27% in every quintile except the second. Each sex-race subgroup in the second quintile (not shown) had an increase in treatment ranging from 3% to 12%. The extremely low proportion treated in quintile 2 is influenced heavily by a treatment rate of only 24% in white men.

Discussion

As in a previous report,¹ our findings document a downward trend in mean systolic and diastolic BP values in three national health examination surveys spanning the interval from 1971 through 1991. The prevalence of hypertension (≥160/95 mm Hg including individuals treated with antihypertensive medication) remained stable in the interval between NHANES I and II but decreased from NHANES II to III. The age-adjusted prevalence of hypertension at the 140/90 mm Hg cut point began declining during the interval between NHANES I and II, with a net absolute decline of 15.9% from 1971-1974 to 1988-1991. The proportions of the population with hypertension who were aware, treated, or controlled increased from 1971-1974 to 1976-1980,²⁷ presumably explaining much of the decline in mean BP levels during that interval.

Data from NHANES III indicate that significant progress in hypertension awareness, treatment, and control continued between 1976-80 and 1988-1991. Hypertension awareness has continued to increase substantially at each of the cut points defining high BP. Two thirds (64%) of all people with hypertension now have their BP controlled if the 160/95 mm Hg threshold is used; however, at 140/90 mm Hg, the percentage is still less than one third (29%). Despite the low rate of control to <140/90 mm Hg, the distribution of systolic BP in the population aged 60 to 74 years has consistently shifted to lower levels between 1960-1962 and 1988-1991. This suggests that the improvements in awareness, treatment, and control account for at least part of the observed decline in mean systolic and diastolic BP values in every age-sex-race subgroup. Possible reasons for the dramatic declines in prevalence rates between NHANES II and III in blacks and whites are discussed below.

In a previous examination of the NHANES III data, more than 7% of the adults aged 18 years and older were classified as normotensive but reported a history of hypertension.¹⁹ About half of this group reported modifying their lifestyle to control their high BP. It is likely that at least some of this group had hypertension and that they should be included in the prevalence estimate for hypertension. To the extent that lifestyle modification was increasingly and successfully attempted in such hypertensive people, this would account for some of the observed decline in prevalence.

The relatively low prevalence of hypertension in Mexican Americans noted in HHANES relative to the earlier NHANES II general population data has been discussed extensively elsewhere.²⁸²⁹³⁰ Seven years later during NHANES III the rates among Mexican Americans were only slightly lower than those noted in HHANES and were similar to those of the white population in NHANES III. In contrast to the four national surveys, the study methods, including the approach to standardization of BP measurement, were similar in HHANES and NHANES III. This suggests that measurement methods may account for some of the decrease in BP from NHANES II to III in the black and white populations. However, it is possible that the first data point of the Mexican American population in 1982-1984 reflects a lower prevalence of hypertension than was present but undetermined in 1976-1980.

As presented in “Methods,” different protocols and levels of standardization were used in BP measurement in the five health examination surveys. In particular, the BP measurements in NHANES III were standardized better than in previous national surveys. A full discussion of all possible biases, comparing NHANES III with those surveys, is beyond the scope of this article. However, the lack of use of large cuffs before NHANES III, the marked preference for digits ending in zero, and variable definitions of diastolic BP may have conferred an upward bias on the prevalence results of prior surveys.

A study of obese subjects found that the difference in BP readings among adult, large, and thigh cuffs was greater with increasing arm circumference. In people with a large arm circumference, the mean difference between BP values obtained with a large cuff and a regular cuff was 5.9 mm Hg systolic and 4.4 mm Hg diastolic. The use of a regular cuff when a large cuff was warranted resulted in a 37% overestimation of hypertension in this population.²³ Based on this experience, formulas were developed to correct the error introduced by incorrect cuff selection and were incorporated in the 1988 American Heart Association’s “Recommendations for Human Blood Pressure Determination by Sphygmomanometer.”³¹ Such corrections, however, are only a partial substitute for use of the correct cuff size, as was done in NHANES III.

A potentially profound effect on hypertension prevalence is possible when end-digit preference exists.³² In a recent study,³³ 78% of the BP readings ended in zero. In this study, change in the systolic BP threshold from ≥140 to >140 mm Hg reduced hypertension prevalence from 25.9% to 13.3%. However, digit preference should have a much smaller effect when surveys are compared with the use of consistent cut points, and the directions of such biases are difficult to predict but may be different for systolic and diastolic BP. Because the trends in the current analyses are generally consistent for systolic and diastolic BP, digit preference probably plays a minor role.

In past surveys the use of the fourth Korotkoff sound (the point at which muffling occurs) as well as the fifth Korotkoff sound (the point of disappearance) as diastolic BP was likely to have resulted in the NHANES III diastolic BP being systematically lower. The American Heart Association guidelines given to physicians to review for NHANES I and II recommended that phase IV Korotkoff sounds were a better index of diastolic BP.³¹ Therefore, it is likely that the physicians who read the guidelines used the fourth sound. Once again, because trends in systolic and diastolic BP are consistent, this would not play a role in systolic BP.

The NHANES III data used to describe trends in this article are subject to regression to the mean; ie, BP values that are extremely high or low are likely to be lower and higher, respectively, at a repeat examination. The effect of regression to the mean on BP distribution has been well established.^34353637 For example, in one study of BP measurement in surveys, the variability of BP measurement indicated that classifying individuals as hypertensive or normotensive based on observation at a single examination would result in more than one third being incorrectly classified as hypertensive and 5% being falsely classified as normotensive. The resultant estimate of hypertension prevalence would have been overestimated by 17%.³⁶ Similar substantial decreases in systolic BP have been noted in other studies of people with hypertension.³⁸³⁹ The use of only one measurement in two thirds of NHANES I participants probably accounts in part for the generally higher mean BP values and prevalence rates compared with the preceding and subsequent surveys, especially in younger people, who have more labile BP.

In summary, the prevalence of hypertension reported from NHANES III may be partially explained by the increased standardization of the BP measurement technique. The use of BP measurements obtained with subjects in the seated position in MECs represented the second set of measurements, and the BP values on this occasion may represent the effect of regression to the mean. In addition to this phenomenon, it is possible that people have become more familiar with having their BP measured and are less likely to experience white coat hypertension. Nevertheless, analyses that use all the measurements, including those taken in the home by nonphysicians, revealed only small differences from those reported in this article.¹⁹

The trend toward decreasing BP levels also has been noted in regional studies of cardiovascular risk factors. The Minnesota Heart Study, a population-based study designed to monitor trends in cardiovascular risk factors including hypertension, noted a decline in hypertension prevalence between 1980-1982 and 1985-1987, although the decrement in high BP was not statistically significant. Age-adjusted systolic and diastolic BP levels declined during the 5-year period of the study from 115.1 and 71.0 mm Hg to 113.6 and 70.4 mm Hg, respectively, in the 25- to 74-year-old population.⁴⁰ In the Canadian Heart Health Survey, conducted from 1986 to 1990, adults aged 18 to 74 years had their BP measured on two occasions with a standardized technique similar to that used in NHANES III. The mean systolic and diastolic BP values for men and women lay approximately halfway between those of NHANES II and III.⁴¹ Prevalence rates (≥140/90 mm Hg) were similar to those described in the present article for NHANES III. This was true for men and women and for both sexes combined. However, the Canadian definition of hypertension included those individuals treated with a low sodium diet or weight control. Therefore, prevalence comparable to the US definition would be lower in Canada.

In the East Boston Senior Health Project, a cohort of the National Institute on Aging’s Established Populations for Epidemiologic Study of the Elderly, systolic BP declined markedly between 1982-1983 and 1988-1989, with most of the decline occurring between 1985-1986 and 1988-1989. Initiation of or changes in antihypertensive drug therapy explained much but not all of the decline.⁴²

Although differences in technique make it impossible to quantify trends in BP and prevalence of high BP precisely, it is clear that hypertension awareness, treatment, and control have improved substantially. Mean systolic and diastolic BP values were lower in NHANES III at all BMI levels, but especially among those subjects with higher BMIs. Although absolute declines in prevalence rates were greater at higher BMIs, relative declines were similar across these strata. Therefore, although much of the observed decline is attributable to methodological differences among the surveys, our tentative conclusion is that much of the decline in prevalence is also real and should encourage further exploration to identify the responsible factors. One powerful determinant—overweight and obesity—has been trending in an adverse direction.²² Some available data suggest that the declining use of alcohol and dietary salt may be playing a role.²⁴³

The prevalence of hypertension is unacceptably high in the US population and affects nearly three quarters of black people and one half of whites aged 60 to 74 years. Further improvements not only in treatment but especially in prevention of hypertension are necessary and possible, as indicated by the low proportion of people with hypertension whose BP is controlled to levels below 140/90 mm Hg; even among patients receiving treatment, only 55% are controlled to nonhypertensive levels. Trends in subsequent national surveys will be helpful in confirming the pattern of BP changes noted in this analysis. If the current pattern continues over time, this will strengthen the possibility that an environmental change in risk factors for high BP is occurring and will document the progressive improvement in the detection, treatment, and control of established hypertension.

Reprint requests to Edward Roccella, PhD, National High Blood Pressure Education Program, National Heart, Lung, and Blood Institute, Building 31, Room 4A-05, Bethesda, MD 20892.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Footnotes