The relation between diet and cardiovascular disease has been the focus of substantial investigative effort for several decades (1). The observed reduction in cardiovascular death rates in the United States over the past 30 years has been attributed, in part, to changing dietary patterns (1, 2). Nevertheless, hypertensive cardiovascular disease remains the principal cause of morbidity and mortality in America (1). Life-style factors, particularly dietary patterns (3), have been implicated as major contributors to the continued prevalence of high blood pressure in the United States.
The analysis of epidemiologic data presented in this article was prompted by three facts. First, a large data base of the National Center for Health Statistics, Health and Nutrition Examination Survey I (HANES I) (4), has not been analyzed for possible associations between the demographics of hypertension in the United States and exposure to all relevant nutrients in the diet. Second, the hypothesized relation of nutrients such as sodium (5, 6), potassium (7, 8), cholesterol (9, 10), and calcium (11, 12) to the development of hypertensive carbdiovascular disease has yet to be confirmed in the United States. Third, specific nutrients, particularly the mineralions, serve fundamental functions in the regulation of both cardiac output and peripheral resistance, the principal determinants of blood pressure in humans. Utilizing HANES I, we sought to address these issues, recognizing that the results do not prove causality, but do provide valuable insight for future studies of the pathogenesis and treatment of hyperetensive cardiovascular disease. Sample Composition
HANES I collected measures of health and nutrition obtained from interviews and examinations of 20,749 persons (4). The sample was scientifically designed to be representative of the U.S. civilian noninstitutionalized population 1 to 74 years of age.
Data tapes from HANES I were obtained from the National Center for Health Statistics. We used tape 4704 (24-hour food consumption), tape 4701 (version 2, dietary frequency and adequacy), tape 4233 (medical examination), and tape 4111 (anthropometry, goniometry, skeletal age, bone density, and cortical thickness) for the analysis. HANES I data on nutrient intake were analyzed with guidelines provided by Adams (13) and Pennington and Church (14).
Data on blood pressure and nutrient consumption were obtained in 20,749 persons for a 24-hour period. To form the cohort used for our analysis, all individuals under age 18 were excluded, leaving 13,671 persons. Subsequently, the following questions from the medical history questionnaire were used to exclude pregnant women and people with a history of hypertension:
1) “During the past 6 months, have you used any medicine, drugs, or pills for … high blood pressure?”
2) “Are you on a special diet? Low salt?”
3) “Are you pregnant now?”
Individuals answering affirmatively to any of these questions were excluded from the analysis, leaving 10,419 persons. In addition, 47 individuals for which data on one of the 17 nutrients were missing were eliminated, leaving 10,372 subjects for the analysis.
In the course of our analysis, we used three definitions to form hypertension groups: (i) systolic pressure of 140 mmHG or above, (ii) systolic pressure of 160 mmHg or above, and (iii) the upper 10 percent of systolic pressures in age-, sex-, and race-specific subgroups. Twenty-four-hour nutrient consumption was analyzed for calories, protein, fat, carbohydrates, calcium, phosphorus, iron, sodium, potassium, vitamin A, thiamin, riboflavin, preformed niacin, vitamin C, saturated fats, oleic acid, linoleic acid, and cholesterol. Average nutrient intake was calculated for each blood pressure group; for subgroups based on age, race, sex, and body mass index; and for alcohol consumption patterns by using Statistical Programs for the Social Sciences (SPSS). Additional SPSS programs for regression analysis, analysis of variance, and multivariate discriminant analysis were used (15).
The demographic composition of the 10,372 subjects is shown in Table 1. The age, race, and sex distribution was similar to the overall composition of the HANES I population (16). There was a slight overrepresentation of women because there were more women volunteers than men. Nutrient and Caloric Intakes Among Groups: Results
Table 2 shows mean intakes of 11 nutrients and of total calories for several blood pressure groups. Percentage differences in nutrient intake between groups are also listed. Except as noted, intakes were less in hypertensives.
When the definition of hypertension as a systolic blood pressure of 160 mmHg or above was applied (Table 2), 9.2 percent to the population was classified as hypertensive. Reported caloric intake was 15.8 percent lower in individuals who met the criterion for high blood pressure. Only intakes of calcium and linoleic acid were lower (19.6 and 21.2 percent, respectively) in the hypertensives when compared to their 15.8 percent reduction in total caloric intake. A similar pattern was observed when the more liberal definition of hypertension (systolic blood pressure of at least 140 mmHg) was applied. Multivariate discriminant analysis showed calcium to be the nutrient whose intake was most predictive of hypertension (systolic pressure of at least 160 mmHg) after controlling for age, race, and sex. In addition, controlling for alcohol consumption patterns (number of drinks per week) did not change the observed relation between lower calcium intake and higher blood pressures. Carbohydrates, vitamin C, and linoleic acid were the nutrients that entered the multivariate analysis after calcium.
The second comparison of nutrient intake in Table 2 is based on the definition of hypertension as the upper 10 percent of systolic blood pressures for the following age groups: 18 to 24, 25 to 34, 35 to 44, 45 to 54, 55 to 64, and 65 to 74 years. This definition adjusts for the increase in blood pressure with age in our society (17) and for those individuals at greatest risk for cardiovascular complications. Intakes of calcium, potassium, vitamin A, and vitamin C were all reduced (7.6, 7.1, 11.8, and 11.9 percent, respectively) in subjects above the 90th percentile. Linoleic acid, fat, and carbohydrate were eliminated as nutrients that were reduced in the hypertensives.
Because not only age but also sex and race are correlates of blood pressure in the United States (18), we defined hypertension as the upper 10 percent of individuals, based on systolic blood pressure, for their respective age, race, and sex group. This analysis ensures that identified differences in nutrient intake are not simply related to demographic characteristics. After inclusion of age, race, and sex, it was found that potassium, calcium, vitamin A, and vitamin C were the nutrients whose intakes were significantly lower in hypertensive subjects, as determined by partial F tests (P <0.001 for potassium and vitamins A and C and P <0.003 for calcium).
With the epidemiologic technique of direct rate standardization, we standardized means for each nutrient for age, race, and sex, using the entire subgroup as the reference population. Figure 1 depicts the standardized differences on the basis of the 160-mmHg cutoff. By this calculation, calcium, potassium, vitamin A, and vitamin C are the four nutrients anticipated to be reduced the most in individuals with high blood pressure.
Body mass index (kilograms per square meter) is a strong correlate of blood pressure (19). Consequently, this variable was added to age, race, and sex. Intakes of calcium, potassium, vitamin A, and vitamin C were all significantly less (P <0.001, partial F test) in the hypertensive subjects with this final adjustment. Figure 2A depicts total calories for the various blood pressure and body mass index groups. Caloric intake in both lean and obese hypertensive individuals was less than in the corresponding normotensive groups. Figure 2B depicts the values for calcium intake for the normotensive and hypertensive body mass index groups. Two points are evident. First, none of the hypertensive subgroups had a mean intake of calcium equal to the current recommendation of the National Academy of Sciences (800 mg/day) (20). Second, higher intake of calcium was negatively correlated with body mass index (R = -0.588). Calcium intake decreased with increasing body mass index.
Potassium intake (Fig. 2C) showed a pattern similar to that of calcium, but sodium intake (Fig. 2D) did not differ on the basis of blood pressure or body mass index. Neither cholesterol consumption (Fig. 2E) nor phosphorus intake (Fig. 2F) were consistently different among the various blood pressure and body mass index groups.
Figure 3 shows the relation of daily consumption of calcium, potassium, and sodium to systolic blood pressure for the entire cohort. For this analysis, the population was grouped by increments of 75 mg for calcium intake, 150 mg for potassium intake, and 150 mg for sodium intake. Mean systolic blood pressure was then calculated for each group. Standard deviations varied between 16 and 26 mmHg for calcium, 14 and 27 mmHg for potassium, and 11 to 20 mmHg for sodium intake, and increased slightly in states of both extremes of nutrient intake. The number of individuals per stratum was not less than 60 for calcium, 20 for potassium, and 31 for sodium. For each of the three nutrients, increased consumption was negatively correlated with systolic blood pressure (r = -0.604, -0.461, and -0.279 for calcium, potassium, and sodium, respectively).
The proportions of individuals with systolic blood pressures above 160 mmHg for increments of reported calcium (75 mg), potassium (150 mg), and sodium (150 mg) consumption are shown in Fig. 4. The risk of being hypertensive increased with decrements in the ingestion of each of these nutrients, including sodium. Calcium intakes of less than 300 mg/day were associated with a risk of 11 to 14 percent, while intakes greater than 1500 mg carried a risk of 3 to 4 percent. Potassium intakes under 900 mg carried a risk of 12 to 14 percent and intakes above 4200 mg/day a risk of 4 to 5 percent. Sodium intakes below 1600 mg had an associated risk of 9 to 12 percent; intakes above 4800 mg, 2 to 4 percent, the lowest observed.
Figure 5A depicts an individual’s risk of having a systolic blood pressure in the upper 10 percent for his respective age, race, and sex cohort on the basis of calcium intake (75-mg increments). This analysis controls for variability in the prevalence of hypertension in different age, race, and sex groups as well as for variability in the calcium requirements of different subgroups (such as premenopausal and postmenopausal females). For the entire population, the risk of being above the 90th percentile was 3 to 4 percent if reported calcium intake was 1600 mg/day or greater. Below this consumption level, an individual’s risk increased in a roughly linear fashion. The risk increased to 11 to 12 percent at intakes below 300 mg/day. Results for potassium are shown in Fig. 5B. At a potassium consumption level of 4600 mg or greater, the risk was approximately 5 percent that, within a demographic subgroup, an individual would be in the upper 10 percent of the blood pressure profile.
Table 3 shows the results of using multivariate discriminant analysis with U.S. government food groups to predict hypertension (systolic pressure of at least 160 mmHg) in subjects over 34 years of age. Ninety-six percent of all the identified hypertensive individuals were above this age. After controlling for age, race, and sex, differences in dairy product consumption proved to be the best predictor of hypertension. The variables listed correctly identified 79.6 percent of all the individuals with a systolic blood pressure of at least 160 mmHg. Evaluation of HANES I Analysis
HANES I is a scientifically designed, representative sample of the U.S. population and its nutrition habits (4). Calcium, potassium, vitamin C, and vitamin A are the nutrients that distinguish Americans at greatest risk for hypertensive cardiovascular disease from those with lower pressures and thus less risk.
The HANES I data were intended to provide useful epidemiologic insights into relations among various nutrients, demographic correlates, and measures of wellness. The associations and correlations of various nutrient intake patterns with systolic blood pressure that emerged in this analysis do not prove causality (15), but do suggest the relation of a given nutrient to the blood pressure profile of the United States and whether the pattern of its consumption reflects or predicts the demographics of high blood pressure.
The sample size and the individual medical histories allowed us to identify a demographically representative population of adult Americans free of known hypertensive cardiovascular disease and who denied intentional modification of their dietary habits. Combining the medical data with the survey’s comprehensive assessment of each individual’s diet on the previous day provides a unique opportunity to address the relation between diet and blood pressure in healthy Americans. The 24-hour dietary recall design used in HANES I is imprecise for extrapolating to an individual’s lifetime exposure to nutrients but is the best technique for cross-sectional analyses intended to identify differences among specific populations (21).
We chose systolic blood pressure to define our populations because the risk of cardiovascular complications appears to be more closely associated with that measure of arterial pressure than with diastolic blood pressure (22, 23). The three different definitions of hypertension allowed us to test whether the definition of blood pressure utilized influenced the nutrients that would be identified and whether the differences that emerged were predictive of hypertension in those individuals at greatest risk for cardiovascular disease the upper 10 percent of the U.S. population. By adjusting the blood pressure groups so that the hypertensive population was defined as the upper 10 percent of individuals controlled for age, race, sex, and body mass index, we were able to determine which nutritional variables were associatted with hypertension independent of these demographic and anthropometric correlates of blood pressure. In the younger age ranges, the upper 10 percent definition results in some individuals being classified as hypertensive even though they do not meet an arbitrary definition (such as [is not >] 140 mmHg). Nevertheless, these individuals are likely to be in the same high-risk population in the future as the blood pressure profile increases with age (17).
Regardless of the definition of hypertension applied to the population we studied, and, in part, independent of the effects of age, race, sex, body mass index, or alcohol consumption, lower consumption of four nutrients–calcium, potassium, vitamin A, and vitamin C–was statistically associated with hypertension. Ackley et al. (12), in a study of dairy products and blood pressure, also noted that the association was independent of these same variables. Nutrition and Blood Pressure Associations: Implications
Calcium. Calcium was the nutrient for which reduced intake was most consisttent in hypertensive individuals, regardless of how that population was defined. Furthermore, across the entire cohort of 10,372 subjects, increasing systolic blood pressure was correlated most strongly with decrements in daily calcium ingestion. The observed range of the reduction in calcium intake in the hypertensives 17.6 to 19.6 percent) is similar to that in a recently reported diet survey (11). In that survey low intake of calcium was the nutritional factor that distinguished subjects with essential hypertension from those with normal blood pressures. Our findings are also consistent with epidemiologic observations (24-26), spanning several decades, that have suggested an association between adequate exposure to calcium and protection against various cardiovascular disorders. In addition, recent clinical reports have linked disordered calcium metabolism to human hypertension (27-30).
As portrayed in this analysis, the average calcium intake for the U.S. population at greatest rish for hypertensive cardiovascular disease is significantly less than the 800 mg/day recommended by the National Academy of Sciences. In addition, the calcium-depleting effects of the high content of protein, phosphorous, sodium, and alcohol in the U.S. diet may necessitate a calcium intake above 800 m/day in some individuals if the current minimum daily requirement is to be met (31).
Potassium. The relation of potassium consumption to blood pressure was similar to that of calcium. However, the observed differences in potassium intake between hypertensives and normotensives were consistently less than those noted for calcium. Both epidemiologic observations and animal studies have suggested that an increased dietary potassium intake is protective against the development of hypertensive cardiovascular disease (32).
Potassium may influence cell membrane and intracellular mechanisms that contribute to vascular smooth muscle cell regulation as well as humoral and volume factors (33) that are functionally important in cardiovascular regulation. For calcium the putative mechanisms are less certain. While the cation is essential for smooth muscle contraction (33), a membrane stabilization and vascular smooth muscle relaxation effect (34, 35) appears to be an equally important action. The effect of calcium on cell membranes may be related, in part, to inhibitory actions on membrane-associated calcium channels (36, 37). Whatever specific pathway is involved, it is apparent that calcium contributes to both vascular smooth muscle cell relaxation and contraction (33).
The observed relation between calcium and potassium intake and blood pressure provides substantive evidence of nutrient interactions. Dairy products, which account for an individual’s principal exposure to both calcium (60 to 70 percent) and potassium (35 to 45 percent) (38), were the food group whose consumption best predicted whether an individual over the age of 34 was hypertensive. The greater an individual’s consumption of daiy products, the less likely it was that he or she was hypertensive. Increased consumption of dairy products would be associated with a correction in potassium and calcium deficits.
Vitamins A and C. Intake of vitamins A and C differed between the blood pressure groups when the populations were defined as the upper 10 percent and lower 90 percent of systolic blood pressures, as adjusted for demographic variables in the standardized analysis. With the simpler definitions of hypertension ([is greater than or =] 140 or [is greater than or=] 160 mmHg), a consistent difference was less evident. A role of a deficiency of either or both of these vitamins in the pathogenesis of hypertension has not, to our knowledge, been previously postulated. The lower intake of these two vitamins in hypertensives may reflect, in part, their close association in the diel with calcium (vitamin A) and potassium (vitamin C).
Sodium. Regardless of the definition of hypertension and the demographic variable controlled for, hypertensives tended to consume less soddium than normotensives. This conclusion is consistent with a recent report from the National Center for Health Statistics (39) that identified an association of increased salty snack food consumption and frequent salt shaker use with lower blood pressures in the HANES I data. Of all the nutrients measured in the HANES I survey, sodium intake is subject to the greatest underestimation. The dietary recall data do not account for discretionary sodium added during food preparation and at the table. Previous studies have suggested that discretionary sodium accounts for 20 to 40 percent of total sodium consumption in the United States (38). Other diet surveys indicate that an individual’s discretionary sodium use parallel the sodium content of the food consumed (40, 41); still others have demonstrated that dietary recall information on sodium consumption parallel 24-hour urinary sodium excretion (42, 43). Consequently, it seems likely that the lower dietary sodium content of the hypertensives in our study reflects reduced intake of the nutrient among hypertensives in general. Because the individuals that formed our sample denied a history of hypertension and intentional manipulation of their diet, the lower sodium intake reported by individuals at greated risk for hypertension-related cardiovascular disease cannot be ascribed to changes in dietary habits related to concerns about hypertension and cardiovascular disease. Earlier reports, based on the HANES I data, that addressed the relation of sodium intake to measures of wellness suggested a similar inverse relation between dietary sodium and high blood pressure in the United States (39–41).
Previous attempts to establish the putative link between sodium consumption and blood pressure in the United States have failed (6, 44). The HANES I population, however, was specifically identified so that observations based on its analysis could be extrapolated to the entire U.S. population (4). Unlike any previous assessment of sodium’s relation to blood pressure in more limited U.S. surveys, HANES I provides the opportunity to evaluate the association of blood pressure with extremes of sodium intake. With HANES I, sufficiently large numbers of individuals can be identified who spontaneously reported diets either relatively lwo or relatively high in sodium (fig. 4C). While “salt-sensitive” individuals are not specifically identified, it is evident that subjects reporting low-sodium diets are at two to three times greater risk of being hypertensive than those who report a high sodium intake. Such cross-sectional dietary recall data are not necessarily indicative of individual patterns but do indicate the characteristics of the population as a whole.
The relation between greater sodium consumption and lower blood pressure is consistent with both nutritional and physiological interactions of nutrients. First, as mentioned above, dairy products are a significant source of sodium as well as calcium and potassium. Second, the actions of sodium are closely linked to those of calcium and potassium at both the cellular and organ levels (45–47).
Cholesterol. Cholesterol consumption and serum cholesterol concentration have been considered concurrent risk factors with high blood pressure for the development of cardiovascular disease in Western societies (3, 48). In our study cholesterol intake did not differ consistently between hypertensive and normotensive individuals.
Obesity. The known link between obesity and blood pressure has not been explained by past investigations (18, 19, 49). Excessive sodium intake in the obese subject was long though to be a contributing factor (50). However, controlled clinical investigations in the past few years have largely discounted that explanation (51, 52). Our analysis suggests that obesity-related hypertension is not associated with excessive sodium consumption. Furthermore, the total caloric intake of obese subjects was less than that of lean individuals across the HANES I population, a finding noted previously (53). Hypertensive also reported a lower caloric intake than normotensives. This was evident principally in the leanest and most obese subgroups of hypertensives. It is unlikely that the heavier subjects and hypertensive subjects intentionally underreported their caloric intake for several reasons. First, the inverse relation between body mass index and calories was strong and continuous, even within the leaner portion of the population. Second, none of the subjects perceived him or herself as hypertensive. Third, the association between a significant reduction in calories and elevated blood pressure was apparent even in the leanest 30th percentile of the hypertensive population (Fig. 2A).
It appears that nutritional deficiencies and not excesses are what distinguish overweight or hypertensive individuals from normal subjects in the United States. Caloric restriction increases the risk of further reducing an individual’s exposure to nutrients that may be essential for maintaining normal mean arterial pressures. Conclusion
Our analysis of the HANES I survey suggests the following:
1) There are predictable nutritional differences between individuals with high blood pressure and those with normal blood pressure.
2) Deficiencies rather than excesses are the principle nutritional patterns that characterize the hypertensive person in America.
3) Reduced consumption of calcium and potassium is the primary nutritional marker of hypertension, with reductions in vitamins A and C also being noted.
4) Dairy products are the food group for which reduced consumption is most closely related to high blood pressure in the United States.
5) These observations are largely independent of age, race, sex, body mass index, and alcohol consumption.
6) Diets low in sodium are associated with higher blood pressures, while high-sodium diets are associated with the lowest blood pressures.
Implicit in the application of the nutrient and blood pressure interactions we have characterized from the HANES I data is a note of caution. Clinical use of sodium- or cholesterol-restricted diets for patients with high blood pressure of cardiovascular disease must be monitored closely to avoid inadverten, simultaneous reduction in calcium and potassium intakes below current recommended daily allowances (38). In addition, these epidemiologic data raise the important question of whether sodium restriction is routinely advisable in many hypertensives (6, 54). The identified “Salt-sensitive” patient and patients with compromised cardiac or renal function would be the obvious exceptions.
It must be emphasized that these findings do not prove causality. They simply indicate potentially important relations among nutrients and blood pressure regulation in humans. It is possible that low consumption of dairy products (the major source of calcium) serves as a marker of hypertension; however, if the dietary patterns of smokers and individuals with sedentary or stressful life-styles involve low intake of dairy products, one cannot be sure from these findings which is the cause and which the effect. Only future clinical and bases laboratory investigations can ascertain the importance of these correlations in the application of health measures intended to reduce the prevalence of hypertensive cardiovascular disease in adult Americans. If validated, our observations do not indicate that it is routinely necessary to ingest any nutrient, including calcium or potassium, above the current recommended levels. Rather, they suggest the consumption of a diet balanced in all the essential nutrients and appropriate for the individual’s level of physical activity.