Association between age and systolic and diastolic blood pressure and pulse pressure in the overall sample.
Association between systolic blood pressure and age by daily calcium intake.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hajjar IM, Grim CE, George V, Kotchen TA. Impact of Diet on Blood Pressure and Age-Related Changes in Blood Pressure in the US PopulationAnalysis of NHANES III. Arch Intern Med. 2001;161(4):589–593. doi:10.1001/archinte.161.4.589
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
The impact of diet on blood pressure and the age-related changes in blood pressure have been difficult to detect within one population. We designed this analysis to study the association of major dietary factors with blood pressure and with age-related changes in blood pressure in a representative sample of the US population.
Data were obtained on all individuals 20 years or older (n = 17 030) surveyed in the Third National Health and Nutrition Examination Survey (NHANES III), including demographic data, anthropometric data, dietary intake (sodium, potassium, calcium, magnesium, protein, alcohol, and total energy) based on 24-hour recall, and blood pressure. Multivariate models relating diet to blood pressure were constructed using stepwise regression, best subset regression, and multiple regression.
Systolic blood pressure was positively associated with higher sodium, alcohol, and protein intakes (P<.05) and negatively associated with potassium intake (P = .003). Diastolic blood pressure was negatively associated with potassium and alcohol intakes (P<.001). Pulse pressure was positively associated with sodium, protein, and alcohol intakes (P<.001). A higher intake of calcium (P = .01) was associated with a lower rate of rise in systolic blood pressure with age.
A diet low in sodium, alcohol, and protein is associated with lower systolic blood and pulse pressure. Potassium intake was associated with lower systolic and diastolic blood pressure, whereas alcohol intake was associated with lower diastolic blood pressure. In addition, the age-related changes in systolic blood pressure were attenuated by higher calcium and protein intakes. Magnesium was not associated with any changes in blood pressure.
RESULTS OF the recently reported Dietary Approaches to Stop Hypertension (DASH) trial document the effects of overall dietary patterns on blood pressure levels.1 Although many specific dietary factors have been linked with changes in blood pressure, relatively strong evidence exists for only a few of them. Higher blood pressure levels have been associated with high intakes of alcohol, sodium, and protein and low intakes of potassium, calcium, and magnesium.2-11 The effects of these nutrients on blood pressure are more apparent in studies among several populations, as it is difficult to consistently detect the impact of specific dietary factors on blood pressure or age-related changes in blood pressure within one population. For example, although a relationship between sodium consumption and blood pressure has been demonstrated in across-population studies, at the time this article was written, this was not the case in within-population studies. An analysis of the first National Health and Nutrition Examination Survey (NHANES I) data concluded that a high sodium intake was associated with lower blood pressure.4 The difficulty of accurately measuring nutrient intakes and their limited range within a single population contribute to the difficulty of assessing the relationship between specific dietary factors and blood pressure within a population11,12
Moreover, few studies have included methodological approaches to adjust for the effect of the high correlation and interaction among different dietary factors.13 These studies have usually assessed the association of isolated dietary factors with blood pressure rather than assessing multiple factors taken together.14 We designed this analysis of NHANES III to study the associations between the major dietary factors and systolic (SBP) and diastolic (DBP) blood pressure and pulse pressure (PP), as well as age-related changes in blood pressure.
The NHANES III is a stratified multistage probability sample of the US population. The sampling methods and the survey protocols have been previously described.15 Dietary information was obtained via a 24-hour recall questionnaire, and standardized blood pressure and anthropometric measurements were taken by a trained observer. Of the 39 695 participants in NHANES III, our analysis included only those participants 20 years of age or older (n = 17 030).
Our data set included demographic information (age, sex, and ethnicity), anthropometric information (weight, height, and body mass index [BMI], as reported in the survey), and blood pressure (the mean of 3 readings of SBP and DBP). We selected dietary factors commonly thought to be associated with blood pressure, including sodium, potassium, calcium, magnesium, protein, alcohol, and total energy consumed per day. We excluded the other macronutrients (fat and carbohydrates) because they lacked evidence suggesting an association. Pulse pressure was calculated by subtracting the DBP from the SBP for each participant.
Intakes of the different dietary factors evaluated were highly correlated with each other. A consequence of this intercorrelation (or multicollinearity) among dietary factors was the inability to attribute an independent effect of any single nutrient on blood pressure.16,17 Standardization of individual dietary factors using density measures (either as an energy ratio obtained by dividing the amount consumed by total energy per day or a mass ratio obtained by dividing the amount consumed by the BMI of each participant) decreased this degree of intercorrelation. We elected to use the density measure for our statistical analysis.
Univariate and multivariate analyses relating consumption of dietary factors to blood pressure were performed. For the univariate analysis, weighted linear and polynomial regression models were used.18 Adjustments were made for demographic variables and BMI. To develop the multivariate model, a random subsample (n = 8529) was selected from the total sample and used to choose the best set of variables to be included in the final models. These variables included the demographic variables, BMI, dietary factors using density measures, and dietary factor interaction variables. Standard stepwise regression18-20 and best subset regression19,21 models were used to select these variables. A principal component analysis was also performed.16,17 These methods minimize the effect of intercorrelation on the model developed.17 The model was then tested on the remainder subsample (n = 8501) using a multiple-weighted regression model.18 The variable inflation factor (VIF) was also calculated for each variable to measure the effect of intercorrelation in the final model.22,23 A VIF of 10 or less indicates a low degree of intercorrelation and, hence, more reliable results.17 The final model developed was applied to the overall sample (n = 17 030). Minitab Release 12.22 software (Minitab Inc, State College, Pa) was used for statistical analysis.
The overall sample that satisfied our selection criteria (age ≥20 years) included 17 030 participants. The mean ± SD age of the overall sample was 48.8 ± 0.2 years; the population was 47% male. The ethnic breakdown was 42% white, 28% African American, 26% Hispanic, and 4% other. The mean ± SD BMI (calculated as weight in kilograms divided by the square of height in meters) of the sample was 27.1 ± 0.2. The dietary intake of sodium in the overall sample was high, whereas the daily potassium, calcium, and magnesium intakes were low in relation to the recommended dietary guidelines.24,25 African Americans consumed more sodium than the other ethnic groups, but less potassium, calcium, and magnesium (P<.001 for each dietary factor). Consumption of all dietary factors was higher in men than in women (P<.001) (Table 1).
There was a linear association between age and SBP and a curvilinear association between age and both DBP and PP (P<.001 for all 3 associations) (Figure 1). In the univariate analysis, after adjustments for age, sex, ethnicity, and BMI, SBP was positively associated with the sodium-potassium ratio (P<.001), alcohol consumption (P<.001), and total energy consumption (P = .007) and negatively associated with potassium intake (P<.001). Systolic blood pressure was not associated with the sodium, magnesium, calcium, or protein intake. Diastolic blood pressure was negatively associated with calcium intake (P<.001) and positively associated with alcohol consumption (P<.001). No other associations were detected, including total energy consumption. Pulse pressure was negatively associated with magnesium intake (P<.001) and positively associated with alcohol consumption (P<.001); no other associations were found.
When the standardized daily intake per kilocalorie was used to measure the dietary factors, different associations were found. Lower SBP and DBP were associated with higher daily intakes of calcium, magnesium, and potassium per kilocalorie, whereas higher SBP and DBP were associated with a higher intake of alcohol per kilocalorie (P<.001 for each association) (Table 2).
In the multivariate analysis, higher SBP was associated with a higher daily intake of sodium, protein, and alcohol (P = .048 for sodium; P < .001 for protein and alcohol) (Table 2). The associations of alcohol and protein intake with SBP were less prominent in women than in men (P = .048; VIF, 1.5, for men; P < .001, VIF, 9.6, for women). Calcium potentiated the effect of protein on SBP (P = .03; VIF, 9.4), while potassium blunted that effect (P < .001; VIF, 9.4) (Table 3). Lower SBP was associated with a higher potassium intake (P = .003). Systolic blood pressure was not associated with any other dietary factors. In the final multivariate model for DBP, lower DBP was only associated with the higher consumption of potassium and alcohol (P < .001 for both). In the final multivariate model for PP, higher PP was associated with higher sodium, protein, and alcohol intake (P < .001 for all 3) but not potassium, magnesium, or calcium intake (Table 2). The association between alcohol and PP was lower in women (P < .001; VIF, 1.6) (Table 3). Calcium intake was associated with an attenuated rate of rise in SBP with age. The mean (SD) rate of increase of SBP with age was 0.65 (0.01) mm Hg per year for the lower tertile of calcium consumption, 0.58 (0.01) mm Hg for the middle tertile, and 0.53 (0.02) mm Hg for the upper tertile (P = .02; VIF, 4.4) (Figure 2).
Overall, the final models explained 40.9% of the SBP variability, 16.7% of the DBP variability, and 40.9% of the PP variability in the sample. Age alone contributed about 35.3% and diet about 2% of the SBP variability. About 6% of the DBP variability was explained by age and 3% by diet, whereas about 39% of the PP variability was explained by age and 2% by diet.
The statistical approach of this analysis, specifically, the multistep multivariate analysis and the standardization to the individual's total energy intake, led to more valid results than those previously reported, and the models in this analysis were more stable. Consistent with previous results in other populations,13 we found a high degree of multicollinearity among the different dietary factors evaluated in NHANES III. Our approach was able to detect and correct for this prevalent but unnoticed phenomenon of multicollinearity. By including multiple factors together, evaluating their interactions, and adjusting for their high degree of interdependence, we were able to develop models that allowed us to detect specific associations between dietary factors and blood pressure in addition to identifying patterns of dietary associations with blood pressure. In particular, we were able to study the impact of diet on age-related changes in blood pressure by including age and diet interaction measures in our models.
Our analysis of NHANES III documents that higher sodium, protein, and alcohol intakes are associated with higher SBP and PP. These associations with sodium and protein intake were apparent only in the multivariate model, whereas the associations of alcohol intake with SBP and PP were observed in both the univariate and multivariate models. Our finding of an association between sodium intake and SBP is in accordance with animal models, across-population surveys, and intervention trials.6,10,26-29 Earlier studies provided conflicting results about the association between protein intake and SBP,30-32 while our analysis demonstrates a positive association. This association is blunted by a higher potassium intake and magnified by a higher calcium intake. Our analysis also adds evidence for a positive association between alcohol intake and SBP,6,8,17,19,33-35 although the converse was true in the univariate analysis, highlighting the importance of adjusting for covariates. The associations of higher PP and lower DBP with higher alcohol intake have not been described previously. Furthermore, our analysis suggests that the associations between alcohol intake and SBP and PP are more prominent in men than in women. Similarly, the Belgian Interuniversity study showed that alcohol consumption is associated with higher SBP only in men.36 In accordance with prior evidence,4,7,37-40 we observed that higher potassium consumption was associated with lower SBP and DBP. Our study shows the importance of potassium intake for blood pressure levels in the population.
Until the present study, no relationship between dietary factors and PP has been reported. Recent evidence suggests that PP is more predictive of cardiovascular events, especially coronary events, than either SBP or DBP or even mean arterial pressure.41 The relationship of PP to cardiovascular events is present even in the absence of an elevated SBP, and the incidence of all these events is decreased by measures that reduce PP.41 Our observations suggest that this risk factor may be favorably modified by reducing dietary intakes of sodium, protein, and alcohol.
Our results suggest that the age-related changes in blood pressure within the US population can be modified by dietary manipulations. Higher calcium and protein intakes attenuated the age-related rise in SBP, whereas higher sodium, alcohol, and protein intakes increased the rate of the age-related change in DBP. These observations remain to be confirmed using longitudinal data.
Consistent with previous studies,11,37 blood pressure and dietary factors in our sample were associated less strongly than blood pressure and demographic factors or BMI. Despite this smaller effect, the impact of diet in the general population is significant, since small changes in blood pressure in a large population would have a significant effect. In addition, this magnitude of the effect of specific nutrients on blood pressure within the general population is not dissimilar from results of clinical trials.2
In summary, this is one of the largest analyses to investigate the association of diet and blood pressure within a population, especially within older age groups. The results have important clinical and public health implications. Since dietary habits are potentially modifiable, the manipulation of diet could have a significant impact not only on blood pressure levels but also on the rise in blood pressure with age. More specifically, our results suggest that a diet low in sodium, protein, and alcohol and rich in potassium and calcium would affect the blood pressure levels in the general population, including PP. Furthermore, our study shows that, since the impact of an individual nutrient on blood pressure is modified by the intake of other nutrients, it is important to assess the overall diet rather than any single nutrient in isolation when measuring the impact of diet on blood pressure.
Accepted for publication September 14, 2000.
Corresponding author: Ihab M. Hajjar, MD, MS, University of South Carolina/Palmetto Health Alliance, 9 Medical Park Dr, #230, Columbia, SC 29203 (e-mail: email@example.com).