Diet plays an important role in the modulation of blood pressure in hypertensive and normotensive adults [1,2,3,4,5,6,7,8,9,10,11,12]. Short-term randomized clinical trials, including the DASH (Dietary Approaches to Stop Hypertension) and OmniHeart, demonstrated that a combination diet rich in fish, lowfat dairy products, nuts and seeds, fruit, vegetables, legumes, and/or whole grains as well as lower in red and processed meat substantially lowered systolic and diastolic blood pressure in pre-hypertensive to moderately hypertensive white and black men and women [1,10,13]. In addition, a meta-analysis of 14 clinical trials conducted in the U.S., Europe, and Australia showed a significant blood pressure lowering effect of the Mediterranean Diet (MedDiet) pattern (a diet rich in olive oil, fish, legumes, nuts, vegetables, fruit, and moderate alcohol but low in meat and dairy products) [14]. These diet patterns have in common greater intakes of fruit, vegetables, legumes, nuts, and fish and low intake of meat; however, notable differences between these diet patterns are the inclusion of more lowfat dairy products in the DASH diet [1], whereas, a MedDiet includes more olive oil but fewer dairy products [14].

In observational studies, dietary pattern analysis has been useful in predicting diet-disease associations because a diet pattern captures total dietary intake instead of individual nutrient or food intakes [15]. However, study findings have been inconsistent possibly due to different study designs and study populations, method of diet assessment, or method of diet pattern derivation [11,16,17,18,19]. “A posteriori” derived dietary patterns, including the Western and Prudent patterns, are data driven and created through factor or cluster analysis [20]. “A priori” diet patterns or scores, such as the Healthy Eating Index (HEI) and Diet Diversity Score (DDS) [21,22], have been created based on dietary recommendations or other evidence. Food scores based on epidemiologic evidence, including a healthy food score has been inversely related to elevated blood pressure and incident myocardial infarction [19,23], while other diet and food scores have been unrelated to blood pressure [24]. Given the differential study results reported in dietary pattern analysis, there is also a need to elucidate the contribution of individual food groups in the diet pattern to lower or higher risk of developing high blood pressure.

We propose to examine the associations of an “a priori” healthy diet pattern [19] that is similar to a DASH-type diet pattern: a diet rich in dairy products, nuts, fruit, vegetables, and whole grain, but low in sugar-sweetened beverages and red/processed meat with the incidence of high normal blood pressure (HNBP) and hypertension in middle-aged African American and white men and women. We hypothesize that this healthy diet pattern will be inversely related to the development of HNBP and hypertension. In addition, we will determine whether certain food groups and usual number of servings of these foods explain a major proportion of the relation.

In this ARIC sample free of baseline high normal blood pressure (n = 5260) and hypertension (n = 9913), the mean ± SD age at baseline was 54.0 ± 5.7 years in men and 53.0 ± 5.6 years in women. The mean ± SD systolic and diastolic blood pressures at baseline for the total population were 114 ± 12 mmHg and 70 ± 9 mmHg, respectively (Table 1); by Exam 4 the mean systolic blood pressure increased to 123 ± 17 mmHg, while the mean diastolic blood pressure remained 70 ± 10 mmHg. Among the four sex-ethnic groups, white women had the lowest mean systolic and diastolic blood pressures at baseline. Over nine years of follow-up, 1663 participants developed high normal blood pressure and 2853 participants developed hypertension. The incident proportions of high normal blood pressure did not vary much across gender-ethnic groups. However, blacks, particularly women, had the highest incidence of hypertension compared with whites.

Participants who were categorized with a lower HFscore (Quintile 1) were more likely to be younger, African American, current smokers, physically inactive, and have less education than those in Quintile 5 (Table 2). Intakes of fiber, calcium, vitamin D, alcohol, and servings of dairy, fish and seafood, fruit and vegetables, fruit juice, legumes, whole grain, and refined grain were higher, while % kcal from fat and servings of meat were lower across successive quintiles of HFscore intake.

There was no relation for the risk of developing high normal blood pressure across quintiles of HFscore (p_trend = 0.63) (Table 3, Model 2), while the HFscore was inversely associated with development of hypertension (Table 4, Full HFscore models). Participants categorized in the highest HFscore quintile were at 14% lower risk of incident hypertension (Model 2; p_trend = 0.01). After adjusting this model for BMI and diabetes at baseline and change in BMI, the relation was slightly attenuated (Model 3; p_trend = 0.06).

To determine which food groups in the HFscore that contributed most to the risk of developing hypertension, we examined individual food groups relative to incident hypertension (Supplemental Table 1). Greater intake of dairy products and nuts and lower intake of meat were significantly and inversely related to hypertension; however, no other food group was related. Three partial HFscore models were created by removing the “dairy”, “nuts” and/or “meat” groups from the HFscore variable to examine the independent relations of dairy, nuts, meat, and the partial HFscores on the risk of developing hypertension (Table 4; Partial HFscore models). We found that the partial HFscores did not remain significant when each of “meat”, “dairy”, or “nuts” was omitted from the HFscore (Partial HFscore models a, b, and c), but these partial HFscores were still inversely related to hypertension incidence. When the three food groups “meat’, “dairy”, and “nuts” were all omitted from the HFscore (Partial HFscore Model d), the combination of the remaining food groups was not related to hypertension (p_trend = 0.90); however, compared to Quintile 1 intake, consuming one or less daily serving of meat, more than two daily servings of dairy products, and about one quarter serving of nuts per day (Quintile 5 intake) was independently related to lower risk of developing hypertension. Adjusting the model for calcium and vitamin D enhanced the inverse association for dairy (data not shown). Race or gender did not modify the relations between the HFscore and incident high normal blood pressure or hypertension.

The HFscore was inversely related to the risk of developing hypertension, but not HNBP, in middle-aged African American and white men and women. Moreover, in studying the components of the diet pattern, the inverse relation with hypertension was largely attributable to greater intakes of dairy products and nuts, and lower intake of meat. Specifically, study participants with the lowest risk of developing hypertension consumed more than two daily servings of dairy products, at least one serving of nuts per week, and one or fewer daily servings of meat, in addition to the combination of greater intakes of fruit, vegetables, whole grains, legumes, and fish and lower intake of sugar sweetened beverages.

Our findings were consistent with those observed in the CARDIA study of young African American and white men and women, the study from which the HFscore used in ARIC was adapted [19]. Even though different dietary assessment instruments were used—the 66-item FFQ [28] in ARIC and the Diet History questionnaire [34] in CARDIA—similar food groups and the same scoring scheme were used to create the food score in each study. One diet quality score, the Recommended Foods Score, which includes only foods recommended for health but not other foods consumed in the rest of the diet, has been related to lower blood pressure [24]. Other diet quality scores, such as the Dietary Diversity Score or the Healthy Eating Index (HEI), are also based on dietary food recommendations, but have had limited success in explaining disease risk [24,35]. The MedDiet pattern, based on the traditional diet of the Mediterranean area and which features olive oil consumption, has consistently been inversely related to a variety of chronic diseases [36], including high blood pressure [37], despite the use of a variety of study designs, study populations, and diet assessment instruments. Although the DASH diet successfully reduced blood pressure in short-term clinical trials [1], not all observational studies have shown a long-term beneficial relation between a DASH-like diet and development of hypertension in longitudinal studies [38]. Finally, numerous “data driven” dietary patterns derived by factor or cluster analysis have been inconsistently related to change in blood pressure or incident hypertension. The inconsistent findings may potentially be explained by differences in the method of diet assessment, the study population and the geographic location of each study (e.g., food availability) [18,39,40,41].

Our findings demonstrate a lower risk of hypertension with higher intake of dairy products and nuts, but lower intake of meat, after adjusting for the rest of the diet pattern and other confounding factors. Inconsistent results have been reported for studies about dairy products, whether total dairy, high-fat or lowfat dairy, relative to elevated blood pressure (defined as SBP ≥130 mmHg and/or DBP ≥85 mmHg) [19] and high blood pressure [11,42,43,44]. Dietary patterns with higher intake of dairy products, may lower blood pressure by modulating insulin secretion [45], although the evidence is inconsistent [46]. Another potential blood pressure lowering mechanism may be the proteins in milk that include both casein and whey which contain blood pressure lowering properties [47]. In addition, dairy products contain vitamin D, calcium magnesium, and individual fatty acids 15:0 and 17:0 which correlate inversely with cardiovascular risk factors [48,49]. Alternatively, dietary calcium may suppress 1,25-dihydroxyvitamin D concentrations, thereby decreasing vascular smooth muscle cell intracellular calcium and also reducing peripheral vascular resistance and blood pressure [50]. Thus, we speculate that a beneficial effect of dairy products on blood pressure may be due to the myriad of nutrients and dietary compounds in dairy products.

Generally, greater meat consumption has been associated with higher risk of developing elevated or high blood pressure among a variety of populations [11,19,51,52]. In contrast to dairy products, meat contains more saturated fatty acid and cholesterol, which may increase risk of coronary heart disease [53]. In a meta-analysis of total protein intake relative to blood pressure, eight cross-sectional studies showed an inverse relation, while four prospective studies reported no relation between total protein intake and blood pressure [54]. Further, both plant protein and protein from meat sources have been inconsistently related to blood pressure [54,55]. In addition, red meat contains high levels of heme iron, which has an unfavorable effect on blood pressure levels [56]. Moreover, processed meat contains considerable salt/sodium, a known risk factor for hypertension. Otherwise, higher consumption of meat may be a factor contributing to weight gain, which is positively associated with elevated blood pressure [57]. Our results showed that adjustment for BMI attenuated the HFscore- hypertension relation, which further supports the mediating effect of BMI on blood pressure.

Finally, our study adds to the accumulating evidence that nut consumption has a blood pressure lowering effect [58]. Nuts are comprised of several components that have been related to lower blood pressure, including plant protein, fiber, antioxidants, monounsaturated and polyunsaturated fatty acids, polyphenols, and the minerals magnesium and calcium.

There are a few limitations to our study. Because blood pressure fluctuates during the day and over time, we captured blood pressure at only one sitting during each exam; however, participants had the opportunity for four blood pressure measurements between 1987–1989 and 1996–1998. Second, even though the FFQ used to assess dietary intake included only 66-items, which may misclassify food intake somewhat, study participants had the opportunity to report other foods not queried, which were coded and included in analysis. Nevertheless, daily nutrient intake may be underestimated; however, food intake is most likely correctly ranked. While acknowledging these limitations, the FFQ has been previously tested for validity and reliability of diet patterns yielding correlations ranging from 0.45 to 0.74 between food records and two FFQs [28]. Finally, statistical models were adjusted for numerous potential factors which may confound the relation between dietary intake and blood pressure category; however, residual confounding may still exist.

The strengths of this study include its prospective study design, high participation rate, and the standardized procedure for measuring blood pressure at four clinic exams. Further, dietary intake was measured on two occasions [33]. In addition, few, if any, cohort studies have previously examined the relation between the HFscore and high normal blood pressure in Caucasian and African American populations.

These findings are consistent with the 2010 U.S. Dietary Guidelines for Americans [59], and suggest that a healthy diet pattern is related to a lower risk of developing hypertension, especially a diet rich in dairy and nut products, but low in meat.