Hypertension is exceedingly common among US adults(1) and is a well-known risk factor for CVD(Reference Stamler, Stamler and Neaton2,Reference Vasan, Larson and Leip3) , contributing to nearly half of all cardiovascular deaths in the USA(Reference Ezzati, Lopez and Rodgers4). Primary prevention of hypertension focuses on key lifestyle strategies including weight reduction, Na restriction, increased physical activity, moderation of alcohol intake and other dietary modifications(Reference Whelton, He and Appel5).
The Dietary Approaches to Stop Hypertension (DASH), characterised by higher intakes of fruits and vegetables (FV) and low-fat dairy products, have been shown to have blood pressure (BP)-lowering effects in clinical trial settings.(Reference Appel, Moore and Obarzanek6) Since the initial DASH clinical trial, other studies have confirmed the beneficial effects of DASH on BP and cardiovascular risk in different population groups(Reference Sacks, Svetkey and Vollmer7–Reference Azadbakht, Fard and Karimi9). Recently, a small randomised cross-over study compared a standard DASH pattern with low-fat dairy foods with a full-fat DASH pattern and found comparable reductions in BP(Reference Chiu, Bergeron and Williams10).
The PREMIER clinical trial was designed to test the BP-lowering effects of two interventions among individuals with pre-hypertension or stage 1 hypertension, one of which included standard lifestyle modifications, while the other also added a DASH dietary intervention(Reference Svetkey, Harsha and Vollmer11). Both interventions were designed to promote weight loss, and both led to reductions in BP, compared with an advice-only control group(Reference Appel, Champagne and Harsha12); the intervention including DASH had a slightly greater beneficial effect on weight loss and systolic BP (SBP) and diastolic BP (DBP) reduction. Subsequent secondary analyses suggested that plant rather than animal protein in the PREMIER diets was linked with greater BP-lowering effects(Reference Wang, Yancy and Yu13). In those analyses, the effect of dairy products on BP outcomes was null, while FV intakes were beneficial. These analyses did not account for baseline dietary patterns.
The selection of the population at risk is a crucial element for any clinical trial since beneficial treatment effects are likely to be greater among individuals at higher risk for the outcome. For example, exercise interventions are likely to provide greater benefit for sedentary individuals who are at higher risk for CVD and other disorders than for those who are already exercising regularly. In these analyses, we hypothesised that baseline intakes of dairy products and FV in PREMIER might have impacted the effects of the DASH intervention on weight and BP in that study. Specifically, we explored whether changes in the level of dairy products and FV intake during the initial 6-month exposure period in the PREMIER study were associated with changes in weight and BP during the intervention period.
Methods
Original study
The PREMIER study was a randomised multicentre clinical trial that was designed to compare the BP-lowering effects of two comprehensive non-pharmacological lifestyle interventions (a standard ‘lifestyle’ intervention and the ‘lifestyle plus DASH’ intervention) with an advice-only control group. Details of the study design have been previously published(Reference Svetkey, Harsha and Vollmer11). Begun in 2000, the study recruited 810 men and women (34 % African American) from four study centres (Baltimore, MD; Baton Rouge, LA; Durham, NC and Portland, OR). Eligible subjects were adults at least 25 years of age with a BMI between 18·5 and 45·0 kg/m2 with pre-hypertension (SBP 120–139 mmHg or DBP 80–89 mmHg) or untreated stage 1 hypertension (SBP 140–159 mmHg or DBP 90–95 mmHg) based on mean BP from three screening visits. The primary intervention period was 6 months. Since the present analyses are based on dietary change, those who were missing dietary data at 6 months were excluded (n 119). In addition, nineteen subjects with missing activity data at 6 months and seven subjects with missing weight measures at 6 months were excluded, leaving data for 665 of the original 810 study participants. These data were obtained from the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repository Information Coordinating Center (BIOLINCC).
In the PREMIER study, both intervention arms employed an identical series of group and individual sessions to facilitate adoption of lifestyle changes and skills to achieve the intervention goals, while the advice-only group (control) also targeted weight loss and physical activity but without the behavioural intervention. The lifestyle intervention targeted the following behaviour changes: (1) reduce weight by 7 kg (about 15 lbs) or more if overweight; (2) limit daily Na intake to 100 mmol or less; (3) limit fat intake to 30 % of energy content intake or less; (4) engage in 180 min/week (or its equivalent) of moderate physical activity and (5) limit alcohol intake to no more than one ounce of ethanol per d for men and no more than 0·5 ounce/d for women. Finally, the lifestyle plus DASH intervention added advice on a DASH-style eating pattern (9–12 servings/d of FV, 2–3 servings/d of low-fat dairy products, reduced intakes of saturated and total fat) to the behavioural intervention.
Trained telephone interviewers assessed diet among study participants using unannounced 24-h recalls. Two recalls (one weekday and one weekend) per participant were collected at baseline and 6 months. Intakes of total energy, nutrients and food groups (from both whole foods and composite food ingredients) were calculated at the Diet Assessment Center of Pennsylvania State University using the Nutrition Data System, version NDS-R 1998 (University of Minnesota). In this way, servings/d of dairy products (milk, yogurt and cheese) as well as FV were calculated. Total dairy product intake included both full-fat and reduced-fat dairy products, although study recommendations were for consumption of low-fat dairy products.
BP was measured by trained and certified individuals following a standardised protocol and using a random zero sphygmomanometer and an appropriate-sized cuff. Two separate measures were taken with at least 30 s between measurements. The means of the two SBP and DBP measures were calculated as well as the mean of two measures of weight (to the nearest 0·25 pound (0·11 kg)) using a calibrated digital scale. Weight, SBP and DBP from the baseline and 6-month follow-up visits are used in these analyses.
Two measures of height to the nearest 0·1 cm were taken using a calibrated wall-mounted stadiometer. BMI was calculated as weight, in kg, divided by the square of height, in metres. Physical activity was assessed by means of a standardised 7-d physical activity recall(Reference Sallis, Haskell and Wood14,Reference Blair, Haskell and Ho15) . Each subject’s self-reported daily activities were categorised as light, moderate, hard or very hard, and the number of minutes spent in activities of various intensity levels was used to derive each individual’s estimated average daily energy expenditure for the previous week.
Statistical analysis
We first estimated mean baseline intakes of dairy products and FV in the two original PREMIER intervention groups and the control group as well as mean change in intake from baseline to 6 months (i.e. mean intake at 6 months minus mean baseline intake). Change in weight, SBP and DBP was calculated in the same manner.
Since the recommended intakes of dairy products in the DASH intervention were 2 or more servings/d, we initially classified dairy product intake as follows: <1, 1–<2 and ≥2 servings/d. However, only 22 % of subjects at baseline consumed at least 2·0 servings/d and only 28 % consumed that amount during the intervention. For FV, recommended intakes were 9 or more servings/d. However, since only twenty-five subjects at baseline consumed that amount, we chose to use the following cut-off values: <3, 3–<5 and ≥5 servings/d. Finally, to improve statistical power, we chose to dichotomise dietary intakes based on information from sensitivity analyses as follows: <1·5 v. ≥1·5 servings of dairy products per d and <5 v. ≥5 servings of FV per d.
These dichotomous categorisations were used to classify intake simultaneously during the baseline and intervention periods. For example, four exposure categories for FV intake were defined as follows: (1) low baseline (<5·0 servings)/low intervention (<5·0 servings) intake, (2) low baseline (<5·0 servings)/high intervention (≥5·0 servings), (3) high baseline/low intervention and (4) high baseline/high intervention. Those with low intakes of FV at both baseline and 6 months (low/low) served as the referent group. Those in the low/high groups were those who increased their FV intake during the intervention. The same approach was used to consider baseline/intervention dairy product intakes.
We used ANCOVA to compare the mean weight loss in three groups (low/high, high/low and high/high) compared with the referent groups (low/low). These models were to derive the adjusted mean weight loss controlling for confounding by factors (age, sex, race and physical activity) found to change the mean differences by 5 % of more. Factors that led to no change in the adjusted means were dropped from the final models. This included both non-dietary factors (i.e. height, education, pack-years of smoking) and dietary factors such as intervention intakes of minerals (i.e. Na, K, Ca and Mg), protein foods (i.e. meats, eggs, poultry and fish), sweets and snacks, total grains, and nuts, beans, and soya. Similar analyses were used to evaluate change in SBP and DBP.
Results
Table 1 shows the characteristics of subjects in the PREMIER trial according to baseline intakes of dairy products and FV. Fewer African Americans were in the highest categories of intake for both food groups. Those with the highest baseline intakes of dairy products consumed more saturated fat but not more total fat, while both total fat and saturated fat intakes were inversely associated with FV consumption. Despite randomisation, there were some apparent differences in baseline dairy product intake across the original PREMIER intervention group assignments. In particular, those with the lowest baseline dairy product intakes were least likely to be assigned to the DASH intervention group, a finding that could have led to confounding of the original results if baseline intake is a determinant of response to the intervention.
Table 1. Baseline characteristics of PREMIER subjects according to baseline intake of dairy products and fruits and vegetables
(Mean values and standard deviations; numbers and percentages)
carbs, Carbohydrates; SBP, systolic blood pressure; DBP, diastolic blood pressure.
* To convert kcal to kJ, multiply by 4·184.
† To convert pounds to kg, multiply by 0·453592.
‡ Moderate to vigorous physical activity.
Fig. 1 shows changes in dairy products and FV intakes (from baseline to 6 months) in each assigned PREMIER intervention group, stratifying by baseline intake. For example, those with the lowest baseline dairy product intakes (<1 serving/d) in the ‘advice-only’ (control) group increased their dairy product intakes by 0·46 servings/d, while those with higher baseline intakes (≥2 servings/d) decreased their intakes by 0·85 servings/d. The same pattern was evident in all three intervention groups – those with lower baseline intakes increased consumption and those with higher baseline intakes reduced consumption. The same was not the case for FV intake where those in the DASH intervention group who already were consuming at least 5 servings/d still increased their intakes by an additional 1·44 servings/d.
Fig. 1. Changes in dairy product and fruit and vegetable (FV) intake according to baseline intake in each intervention group. (a) Changes in dairy product intake according to baseline dairy product intake in the three intervention groups. (b) Changes in FV intake according to baseline FV intake in the three intervention groups. Data are mean changes with their standard errors. (a) 
, <1 serving; 
, 1–<2 servings; 
, ≥2 servings; (b) 
, <3 servings; 
, 3–<5 servings; 
, ≥5 servings. DASH, Dietary Approaches to Stop Hypertension.
Fig. 2 shows changes in weight and BP according to intakes of dairy products and FV during the intervention period. For example, panel A shows that those consuming ≥2 servings of dairy products or ≥5 servings of FV during the intervention had statistically significantly greater weight loss than those consuming less (P trend: 0·009 and <0·001, respectively). Higher FV intakes (panels B, C) were also strongly inversely associated with both SBP and DBP (P trend: 0·002 and 0·027, respectively).
Fig. 2. Changes in weight and blood pressure according to intervention intakes. Changes in weight (a), systolic blood pressure (SBP) (b) and diastolic blood pressure (DBP) (c) according to intervention intakes of dairy products or fruits and vegetables (FV). Data are mean differences from baseline with their standard errors. Mean differences are adjusted for age, sex, race and physical activity. * To convert pounds to kg, multiply by 0·453592.
Since higher combined intakes of dairy products and FV are used to reflect the main components of a DASH-style eating pattern, we examined these combined intakes (online Supplementary Table S1) irrespective of the original PREMIER study group assignment. Those with higher intakes of both dairy products and FV had the greatest weight loss (12·5 pounds (5·7 kg) in the high dairy/high FV group v. 5·6 pounds (2·5 kg) lost in the low dairy/low FV group, P < 0·0001) and the greatest reductions in SBP and DBP. Since weight loss is strongly associated with physical activity and level of obesity, we also stratified by these factors. Regardless of activity level, participants with higher intakes of both dairy products and FV had greater reductions in weight than those with lower intakes of dairy products or FV. Similarly, both obese and non-obese subjects with higher intakes of both dairy products and FV had the greatest weight loss and reductions in BP.
Table 2 shows changes in weight and BP associated with the change in dairy products and FV intakes from baseline to the intervention period. In these analyses, those who started with low baseline intakes (<1·5 servings/d of dairy products or <5 servings/d of FV) but who then increased their intakes (to ≥1·5 servings of dairy products and ≥5 servings of FV) during the intervention had the greatest weight loss. Of note, participants with lower dairy product intakes during the intervention period (low/low and high/low groups) lost similar amounts of weight (7·0 and 7·1 lbs (3·18 and 3·22 kg), respectively). Likewise, those with higher intakes of dairy products during the intervention period also lost comparable amounts of weight (10·6 and 9·9 lbs (4·81 and 4·49 kg) in the low/high and high/high groups, respectively). Patterns were similar for FV intake. As a result of the comparable amounts of weight loss in the low/low and high/low groups (and to improve statistical power), we combined these two categories for the analyses in Table 3.
Table 2. Change in intakes of dairy products and fruits and vegetables (FV) and change in weight and blood pressure
(Adjusted mean values with their standard errors)
SBP, systolic blood pressure; DBP, diastolic blood pressure; Ref, referent.
* To convert pounds to kg, multiply by 0·453592.
† Mean differences adjusted for age, sex, race and physical activity using ANCOVA models.
‡ Low/low = low baseline intake/low intervention intake (<1·5 for dairy products, <5 for FV); low/high = low baseline intake/high intervention intake (≥1·5 for dairy products, ≥5 for FV); high/low = high baseline intake/low intervention intake; high/high = high baseline intake/high intervention intake.
Table 3. Changes in weight and blood pressure according to change in intakes of dairy products and fruits and vegetables (FV), stratifying by intervention intake
(Adjusted mean values with their standard errors)
SBP, systolic blood pressure; DBP, diastolic blood pressure; Ref, referent.
* Models adjusted for age, sex, race and physical activity.
† Low v. high intake: <1·5 v. ≥1·5 servings/d for dairy products; <5 v. ≥5 servings/d for FV.
‡ Low intake here combines low/low and high/low categories from Table 2.
To determine whether the effect of change in dairy product intake was modified by the concurrent intake of FV (and vice versa), we carried out stratified analyses in Table 3. As seen in Table 2, participants who increased their intakes of dairy products or FV during the intervention period (low/high) as well as those who had higher intakes throughout (high/high) had greater reductions in weight than those having low intakes during the intervention. This was also the case for SBP and DBP, although most changes did not reach statistical significance. When stratifying by concurrent intake of FV, it is evident here that the beneficial effect of increasing dairy product intake (low/high) is modified by FV intake. Compared with the referent group (low dairy products among those with low FV intakes) with an average weight loss of 5·7 pounds (2·6 kg), those who increased their dairy product intake and who also consumed at least 5 servings of FV/d had an average weight loss of 13·6 pounds (6·2 kg). In contrast, those who increased their dairy product intakes but who had lower FV intakes lost only 8·7 pounds (3·9 kg). The same pattern was evident for BP change. Finally, higher FV intake (high/high) and increasing FV intake (low/high) led to the greatest weight loss and reductions in BP among those with higher dairy product intakes during the intervention.
Discussion
The original PREMIER randomised clinical trial studied the impact of two lifestyle-based interventions, one with a DASH-style dietary pattern and one without, on weight loss and BP reduction. These present secondary analyses examined how baseline intakes of dairy products and FV, the two primary DASH components, may have influenced the effects of the intervention on weight loss and changes in BP.
We found that baseline intake, especially for dairy products, may have influenced the effects of the original DASH intervention in PREMIER since those participants consuming ≥2 servings of dairy products per d at baseline and who were assigned to the DASH intervention group actually reduced their dairy product intake during the intervention period. In contrast, participants in the DASH intervention arm of the trial increased their intakes of FV regardless of baseline intake.
We found that the amount of both dairy products and FV consumed during the intervention was directly associated with weight loss and BP reduction, with higher intakes being associated with greater declines in weight and BP. Further, higher combined intakes of both dairy products and FV (as recommended by DASH) were associated with greater weight loss and greater SBP and DBP reductions than seen in those with lower intakes of both. We found that those who increased their intakes of either dairy products (to ≥1·5 servings/d) or FV (to ≥5 servings/d) had substantially greater weight loss after 6 months than those with lower intervention intakes. In addition, more active participants and those who were obese at baseline lost more weight than their more sedentary or non-obese counterparts (online Supplementary material). Finally, the effects of dairy consumption on mean weight and BP change during the trial were found to be modified by FV intake and vice versa. Increasing dairy consumption led to the greatest weight loss and reductions in SBP and DPB among those with higher FV intakes. Similarly, the greatest FV benefits were seen among those with higher dairy consumption during the intervention. Thus, these analyses support a beneficial effect of the DASH pattern on both weight loss and BP reduction and demonstrate that these benefits were more apparent for dairy product intake in particular among those whose baseline intakes were lower.
The original PREMIER study was designed to evaluate a DASH eating plan as part of a multicomponent intervention that simultaneously targeted weight loss (of 15 lbs (6·8 kg) or more) and BP reduction(Reference Svetkey, Harsha and Vollmer11). While the primary outcome was SBP reduction, it is difficult to separate fully the BP changes from weight reduction in this trial since most subjects lost substantial weight. In the report of the main results, both the lifestyle-only intervention and the lifestyle plus DASH intervention led to statistically significant reductions in weight and BP(Reference Appel, Champagne and Harsha12). However, participants in the lifestyle plus DASH group had only very slightly greater declines in BP than the established lifestyle intervention (i.e. 0·6 mmHg greater decline in SBP, 0·9 mmHg greater decline in DBP). These results contrast with a previous meta-analysis of seventeen controlled trials that found a DASH diet to be associated with an average reduction in SBP of 6·74 mmHg and in DBP of 3·54 mmHg(Reference Saneei, Salehi-Abargouei and Esmaillzadeh16). These results are also at odds with a recent randomised crossover trial showing that a high-dairy product diet alone resulted in greater reductions (P = 0·009) in both SBP (4·6 mmHg) and DBP (3·0 mmHg) than a diet with lower intakes of dairy products(Reference Rietsema, Eelderink and Joustra17). Since the high-dairy diet in that study comprised 5 or 6 servings for men and women, respectively, compared with <1 serving in the low-dairy group, it is possible that the lower intakes of dairy products in the DASH arm of the PREMIER may explain the weaker dairy product-related effects in that study.
The effects observed in the PREMIER study were much weaker than those seen in the earlier DEW-IT trial in which subjects in a lifestyle intervention group (a low-energy, low-Na DASH plus exercise) had substantially greater reductions in BP than those in the advice-only group(Reference Miller, Erlinger and Young18). In an editorial accompanying the publication of the PREMIER study, Pickering explored the possible explanations for this weaker effect of the DASH intervention arm on BP(Reference Pickering19). One potential explanation for these weaker effects was poorer adherence to the DASH plan in the PREMIER study (since foods were purchased and prepared by the subjects rather than provided). Adherence issues are supported by the finding that the lifestyle plus DASH group in the PREMIER study had only a 28·3 % increase in K intake(Reference Elmer, Obarzanek and Vollmer20); in contrast, the K intake in the combined diet group in the original DASH study was 152·0 % higher than that of the control diet(Reference Appel, Moore and Obarzanek6). The weaker observed effects of DASH in the PREMIER study could also have resulted from a greater reduction in BP in the control group than that was observed among controls in previous studies(Reference Appel, Moore and Obarzanek6,Reference Sacks, Svetkey and Vollmer7) . Further, the PREMIER authors speculated that weaker effects on BP in the PREMIER study might have resulted from a subadditivity of intervention effects (i.e. the total effect of the lifestyle intervention plus DASH may have been less than the sum of the two individual interventions), since it may have been difficult for subjects to self-manage two interventions simultaneously(Reference Svetkey, Harsha and Vollmer11).
In the present analyses, we propose an additional explanation for the weaker effects of the DASH intervention on BP in PREMIER. Specifically, we propose that subjects with higher baseline intakes of dairy products should likely have been excluded from the trial. Alternatively, they could have maintained their already-adequate intake of dairy products rather than decreasing intake during the intervention. Our present analyses suggest that subjects who increased their intake (e.g. from low to high) had a somewhat greater benefit (although not statistically significant) than those who had higher intakes throughout. It is possible that those who increased their intakes may have replaced less healthy food choices with healthier foods recommended as part of the DASH pattern. Previous substitution analyses have suggested that replacement of red meat, for example, with low-fat dairy products, fish or other protein sources could lower CVD occurrence and mortality risk(Reference Bernstein, Sun and Hu21,Reference Song, Fung and Hu22) . However, in confounding analyses in the present study, we controlled for changes in snacking behaviours and other changes in diet but we found that none of these altered the results.
In the present study, we focused on dairy products and FV intakes as the primary components of the DASH pattern. There are several mechanisms by which these foods may benefit BP. FV are rich in polyphenols and also raise antioxidant capacity, lowering oxidative stress. These mechanisms have been linked with beneficial effects on BP regulation and downstream effects on cardiovascular health(Reference Lopes, Martin and Nashar23,Reference Medina-Remón, Zamora-Ros and Rotchés-Ribalta24) . Glutamic acid, an amino acid in vegetable protein, has also been inversely associated with BP(Reference Stamler, Brown and Daviglus25). Further, K, Mg and Ca (all of which are found in abundance in fruits, vegetables and dairy products) have been linked with lower BP through regulation of vascular resistance and promotion of vasodilation(Reference Houston and Harper26). A higher K:Na ratio has been linked with decreases in renal Na retention, thereby lowering BP(Reference Gallen, Rosa and Esparaz27). Higher K intakes have also been shown to reduce salt sensitivity, while Mg aids in the regulation of intracellular Ca, K and Na, thereby promoting insulin sensitivity and reducing vascular resistance. The role of dairy products per se in BP regulation has been somewhat controversial. However, at least one meta-analysis of nine randomised controlled trials found that dairy-derived tripeptides lower BP(Reference Xu, Qin and Wang28). Specifically, angiotensin-I-converting enzyme inhibitory peptides that inhibit the renin-angiotensin system have been found in both milk and cheese(Reference Kris-Etherton, Grieger and Hilpert29).
There are some limitations to the present study but the biggest may be the limited power for stratified analyses. We were unable to look at sex- or race-specific differences and some categories in the present analyses had small numbers. In addition, we used lower cut-off values to define ‘higher’ intakes of both dairy products (1·5 servings/d) and FV (5 servings/d) since too few subjects consumed the recommended 2–3 servings of dairy products and 9–12 servings of FV. Another limitation of the present study relates to the self-report of diet (since the PREMIER study was not a feeding trial) which could have introduced bias in the reported intakes of dairy products and FV. In addition, dietary intake in the present study was assessed at each time point using only two 24-h recalls, which may have yielded less precise estimates of intake than would have been obtained if more days of recall had been available(Reference Basiotis, Welsh and Cronin30). However, a previous study from the EFCOSUM group shows that 2 d of dietary records may be used to estimate intakes of frequently-consumed foods in different population groups(Reference Biro, Hulshof and Ovesen31). Nonetheless, the present study is strengthened by the use of carefully controlled measured data for the weight and BP outcomes as well as important potential confounders.
The present study adds important new data supporting the beneficial effects of a DASH intervention on weight loss among overweight individuals and BP reduction in those with borderline high BP or stage 1 hypertension. It also demonstrates that the combined benefits of dairy products and FV are greater than that seen with individual food groups. Finally, the present study demonstrates that the DASH pattern is most beneficial to those with lower baseline intakes of dairy products and FV.
Acknowledgements
These secondary analyses were supported by the National Heart Lung and Blood Institute (NHLBI; grant no. 5T32HL125232) and the National Dairy Council. None of the funders had any role in the analysis, design or writing of this manuscript. This manuscript was prepared using A Trial of Lifestyle Interventions for Blood Pressure Control (PREMIER) Research Materials obtained from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center and does not necessarily reflect the opinions or views of the PREMIER or NHLBI.
Author contributions were as follows: L. L. M. and R. T. P.: study concept and design; L. L. M.: acquisition of data; R. T. P. and M. R. S.: management and analysis of data; R. T. P., L. L. M. and M. L. B.: drafting of manuscript; R. T. P., L. L. M., M. L. B. and M. R. S.: critical revision of manuscript. All authors approved the final manuscript.
None of the authors has conflicts of interest to declare.
Supplementary material
For supplementary material referred to in this article, please visit https://doi.org/10.1017/S0007114520000112
