Effects of Dietary Patterns on Blood Pressure:  Subgroup Analysis of the Dietary Approaches to Stop Hypertension (DASH) Randomized Clinical Trial FREE

HIGH BLOOD pressure is a major cardiovascular risk factor affecting nearly 50 million adults in the United States and significantly increasing their risk of heart failure, heart attack, stroke, and kidney failure.¹ Diet is an important determinant of blood pressure. Current national guidelines recommend 3 nutritional approaches to prevent and treat hypertension: reduced sodium intake, weight reduction in the overweight, and moderation of alcohol intake.^1- 2 Recently, Appel et al³ reported that a "combination" dietary pattern that is rich in fruits, vegetables, and low-fat dairy products and reduced in saturated fat, total fat, and cholesterol, and modestly increased in protein significantly lowered blood pressure in the absence of reduced sodium intake or weight loss (the Dietary Approaches to Stop Hypertension [DASH] study).

Epidemiological and clinical trial data suggest that the effects of diet on blood pressure may be influenced by demographic characteristics such as race, age, and sex and by clinical characteristics such as body habitus and blood pressure.⁴ We report prespecified subgroup analyses to determine the extent to which demographic and clinical characteristics influenced response to the DASH combination diet, and to determine what segments of the population may benefit from this diet.

The trial's design and methods are reported in detail elsewhere⁵ and are summarized here. The DASH trial was a multicenter, randomized, controlled feeding study designed to compare the effects of 3 dietary patterns on blood pressure in persons with high-normal blood pressure and stage 1 hypertension. The 3 patterns consisted of (1) a control diet with a nutrient composition typical of that consumed by Americans⁶; (2) a "fruits-and-vegetables" diet rich in fruits and vegetables but otherwise similar to the control diet; (3) and a "combination" diet rich in fruits, vegetables, and low-fat dairy foods; reduced in saturated fat, total fat, and cholesterol; and modestly increased in protein. These dietary patterns were designed to reflect choices one might make in whole foods or food groups, and differed in the content of selected minerals (including potassium, calcium, and magnesium), macronutrients (including total fat, saturated fat, and protein), and fiber, all of which have been associated with changes in blood pressure in epidemiological studies.⁵ Since sodium reduction^7- 8 and weight loss in the overweight⁹ are known to lower blood pressure,¹ the dietary patterns we tested did not include these interventions. All 3 diets contained similar amounts of sodium (approximately 3000 mg/d), which is somewhat below average US intake (3286 mg/d in whites; 3265 mg/d in blacks).⁶ Energy (calorie) intake was adjusted throughout feeding to maintain constant body weight. The nutrient goals and food group distribution in the control and intervention diets are summarized in Table 1. Chemical analysis of duplicate meals confirmed that the nutrient content of the menus approximated the nutrient goals.^3,5

The protocol consisted of screening for eligibility, a 3-week run-in feeding period, randomization to 1 of the 3 dietary patterns, and an 8-week intervention feeding period. The clinical sites provided participants with all of their food during the run-in and intervention feeding periods. The primary end point was the change in diastolic blood pressure from the end of run-in to the end of the 8-week intervention period. Change in systolic blood pressure was a prespecified secondary end point.

The trial protocol was approved by the institutional review board of each participating center and by an external protocol review committee. Each participant provided written informed consent.

Inclusion criteria were age older than 22 years and average untreated seated systolic blood pressure of less than 160 mm Hg and diastolic blood pressure of 80 to 95 mm Hg, determined as the cumulative average of 6 random-zero measurements taken during 3 screening visits. Major exclusion criteria were poorly controlled diabetes mellitus, hyperlipidemia, cardiovascular event within the previous 6 months, renal insufficiency, chronic disease that might interfere with trial participation, pregnancy or lactation, body mass index (BMI) greater than 35 kg/m², medications that affect blood pressure, unwillingness to stop taking all vitamin and mineral supplements, unwillingness to discontinue use of antacids containing magnesium or calcium, and alcoholic beverage intake of more than 14 drinks per week. Because of the disproportionate burden of hypertension and its complications in minority populations, particularly among African Americans, the study was designed to include two thirds minority participants.¹⁰ Participants were recruited primarily through mass mailings, advertisements, and public service announcements. Participants were paid $150 to $600 for completion of the study. A total of 459 individuals were randomized into the trial.

The trial was conducted in 3 phases: screening, run-in, and intervention. During the screening phase, prospective participants attended 3 visits. Medication-treated hypertensive subjects attended additional visits for supervised medication withdrawal prior to screening. All participants were free of antihypertensive medication for at least 2 weeks prior to the first screening visit. At each screening visit, blood pressure was measured and continued eligibility and interest were ascertained.

The run-in phase of the trial consisted of a 3-week period of controlled feeding during which prospective participants ate the control diet. If participants were sufficiently adherent to the protocol during run-in, they were then randomly assigned to 1 of the 3 treatment diets (control, fruits and vegetables, or combination). Randomization was performed by the coordinating center using computer-generated random allocation assignments, stratified by clinical site and cohort.

For convenience, participants were enrolled and fed in 4 to 5 separate, consecutive cohorts at each of the 4 clinical centers. Within each cohort, participants were randomly assigned to 1 of the 3 experimental diets. Participants ate their assigned treatment diet for the 8-week intervention phase. On each weekday, participants ate 1 meal—lunch or dinner—at the clinical center. After completing the on-site meal, participants "carried out" all foods to be consumed off-site during the subsequent 24-hour period. On Fridays, they also received their weekend food to be consumed off-site. Participants were instructed to drink no more than 3 caffeinated beverages per day (≤400 mg of caffeine) and no more than 2 beverages per day that contained alcohol (≤30 g of alcohol). With the exception of a few discretionary items (eg, some brands of diet soda) that did not affect the nutrient or energy content of their diet, participants were required to eat all study foods and to eat only study foods. For each day of controlled feeding, participants completed a daily diary in which they recorded any nonstudy foods they ate and whether they failed to eat any required study foods. Clinic staff used this information, along with direct observation during on-site meals, to assess adherence.

Clinical center staff adjusted the energy intake of each participant's diet to maintain stable weight during the 11-week feeding period (run-in plus intervention). Initially, energy needs were estimated based on age, sex, weight, and physical activity. Study menus were prepared at 4 energy levels (6694, 8786, 10,878, and 12,970 kJ/d [1600, 2100, 2600, and 3100 kcal/d]), with intermediate energy levels achieved by the addition of the appropriate number of "unit foods" (418-kJ [100-kcal] muffins that met the nutrient specifications of the assigned diets). Body weight was measured at each on-site meal. If weight rose or fell more than 2% from baseline, the number of unit foods and/or the energy level was adjusted to maintain baseline weight.

Trained and certified staff measured blood pressure using random-zero sphygmomanometers and following a common protocol.^11- 12 For each assessment, 2 blood pressure measurements were obtained after participants rested quietly for 5 minutes in the seated position. These 2 measurements were averaged to provide the blood pressure for that occasion. During run-in and intervention, measurements for a given individual were obtained at approximately the same time of day. Measurements were taken by staff who were unaware of the participants' diet assignment.

Baseline blood pressure was defined as the average of the 3 screening and 4 run-in pairs of measurements. Follow-up blood pressure was defined as the average of 4 or 5 pairs of measurements during weeks 7 and 8 of the intervention feeding. Consistent with our intention-to-treat analysis, follow-up blood pressure for 7 participants without measurements during the last 2 weeks was considered to be the average of earlier intervention measurements. For an additional 6 participants without any measurements during the intervention phase, follow-up blood pressure was considered to be the mean blood pressure during screening. For 2 participants prescribed blood pressure medication, the last blood pressure prior to medication was used.

Height and weight were measured in light clothes without shoes, and BMI was calculated as the Quetelet index (weight in kilograms divided by the square of height in meters). Baseline physical activity was measured at the final screening visit with a 7-day physical activity recall interview.¹³ Years of education, income, alcohol consumption, and family history data were obtained by questionnaire.

The primary outcome was change in diastolic blood pressure, defined as the difference between follow-up and baseline measurements. Change in systolic blood pressure was a similarly defined secondary outcome, and all subgroup analyses presented here were prespecified. The target sample size of 456 randomized participants was estimated to provide 85% power to detect a mean between-diet difference of 2.0 mm Hg in diastolic blood pressure. Sample size was chosen to provide adequate power for the primary outcome; power was therefore lower for subgroup analyses.

Prespecified subgroups were defined as follows. Race was self-defined and categorized as non-Hispanic white, African American, or other minority. Participants were considered hypertensive if baseline systolic blood pressure was 140 mm Hg or higher or if baseline diastolic blood pressure was 90 mm Hg or higher. Women were considered obese if BMI was 27.3 or greater; men were considered obese if BMI was 27.8 or greater.^14- 15 Age was dichotomized at the median for categorical analyses and treated as a continuous variable for multivariate analyses. Years of education, physical activity, and family income were dichotomized at the approximate median. Alcohol intake was defined as some vs no intake (finer gradations were not made since 61% of participants reported no alcohol intake and participants who reported more than 2 drinks per day were excluded).

Data analysis was performed on an intention-to-treat basis. All blood pressure changes reported in the accompanying figures and tables are net of the observed change in blood pressure among participants randomized to the control diet. That is, the blood pressure reductions for each diet represent the change in blood pressure from baseline to end-of-feeding for that diet less the corresponding change in the control diet. These estimates were obtained using the LSMEANS statement in PROC GLM of the SAS software package (SAS Institute, Cary, NC) and are adjusted for site and feeding cohort. This procedure adjusts for "regression to the mean" and secular trends, which might otherwise lead to biased estimates of treatment effects.¹⁶ Results were considered statistically significant at P<.05.

We conducted separate but parallel analyses for both systolic and diastolic blood pressure, and all analyses were performed in 2 stages. Initially we fit separate models for each subgroup variable (eg, race, sex, obesity), testing both for the significance of treatment effects within the subgroups defined by that variable (eg, in males and in females) and whether these effects differed between the subgroups (eg, male vs female) by including the appropriate interaction terms. Unlike previous analyses of these data,³ which assessed differences across the 3 diets simultaneously, this analysis independently compared pairwise differences between diets, ie, the combination diet vs the control diet and the fruits-and-vegetables diet vs the control diet, which allowed us to examine the separate effects of each diet. In this setting of hypothesis generating related to subgroup strata, adjustments for multiple comparisons were not made.

The second stage of the analysis incorporated subgroup variables and interactions that were significant in the previous analyses into a multivariate model to assess the joint effects of these variables. Likelihood ratio goodness-of-fit tests were used to develop a best-fitting model from among this collection of variables. Independent variables not significant at the P<.05 level were dropped from the models. All analyses were conducted using standard general linear model techniques (SAS) and included adjustment for site and cohort effects.

Figure 1 illustrates the progress of participants through the trial. Table 2 provides baseline characteristics of the 459 randomized participants by race and sex. Thirty-four percent of participants identified themselves as non-Hispanic white, 60% as African American, and 6% as other minorities. The "other minorities" comprised a small, heterogeneous group and are not included further in any analyses. Forty-nine percent of the study participants were women. Women were relatively underrepresented among white participants (33%) but not among African Americans (59%). The mean baseline blood pressure was similar across race and sex groups, averaging 131.3/84.7 mm Hg. Twenty-nine percent of participants were classified as hypertensive based on their baseline blood pressure (systolic blood pressure of ≥140 mm Hg or diastolic blood pressure of ≥90 mm Hg), with a somewhat higher percentage among African Americans (32%) than among whites (26%). More than half of the study participants were classified as obese, with somewhat higher rates among women and African Americans than men and whites. The physical activity mean is comparable to definitions of "sedentary."¹⁷

Twenty-seven percent of participants were current smokers and 40% reported smoking at some time in their life. Baseline intake of potassium, calcium, and sodium, based on food frequency questionnaires, was 2940, 749, and 3512 mg/d, respectively, in African American participants, and 3396, 998, and 4104 mg/d in whites. Fat comprised 39% of energy in African Americans and 38% in whites. African Americans ate an average of 5.4 servings of fruits and vegetables daily, and whites ate 6.1 servings.

Adherence to the dietary intervention was greater than 95% in all treatment groups based on attendance at the on-site meal, and greater than 93% in all groups based on self-reported consumption of all study foods and no nonstudy foods. Urinary potassium and sodium excretion confirmed high levels of adherence.³ Weight remained stable, decreasing by 0.04 kg or less in each treatment group.³

Table 3 demonstrates the blood pressure change observed in participants eating the combination diet within various demographic and other strata. The blood pressure–reducing effects of the combination diet were present in virtually every subgroup studied. Reduction in blood pressure was statistically significant in all but 2 strata: diastolic blood pressure in participants with lower educational level and in those reporting some alcohol intake. Among all other groups (African Americans and whites, men and women, older and younger, obese and lean, higher income and lower income, sedentary and active, hypertensive and normotensive, with and without familial hypertension), the DASH combination diet significantly lowered both systolic and diastolic blood pressure.

The blood pressure effect was greater in those who were hypertensive at entry to the study. Blood pressure decreased 3.5/2.2 mm Hg (P<.001) in the normotensive participants assigned to the combination diet and 11.6/5.3 mm Hg (P<.001) in the hypertensive participants assigned to this diet, with the decrease in hypertensives significantly greater (P value for interaction ≤.008 for both systolic and diastolic blood pressure).

Figure 2 shows the net change in blood pressure plotted against baseline blood pressure. For participants eating the combination diet, the amount of blood pressure reduction increased significantly as baseline blood pressure increased (P<.001 for linear trend for both systolic and diastolic blood pressure). There was no evidence of an association between change in blood pressure and baseline blood pressure among participants eating the control diet (data not shown).

The effects of the combination diet also differed significantly by race. Among African Americans eating the DASH combination diet, blood pressure decreased by 6.9/3.7 mm Hg (P<.001). Among whites, blood pressure decreased by 3.3/2.4 mm Hg (P<.01). The reduction in systolic blood pressure was significantly greater in the African Americans (P value for interaction =.02).

Figure 3 displays the combined influence of both race and hypertension status. Among African Americans with hypertension, the DASH combination diet reduced blood pressure by 13.2/6.1 mm Hg (95% confidence interval, −18.2 to −8.1/−9.1 to −3.1). Among normotensive African Americans, the combination diet reduced blood pressure by 4.3/2.6 mm Hg (95% confidence interval, −6.5 to −2.2/−4.4 to −0.8). Among whites, blood pressure decreased 6.3/4.4 mm Hg (95% confidence interval, −12.9 to 0.4/−9.4 to 0.6) in hypertensive participants and 2.0/1.2 mm Hg (95% confidence interval, −4.3 to 0.3/−3.2 to 0.8) in normotensive participants.

Table 3 also demonstrates that the combination diet lowered diastolic blood pressure more in participants consuming no alcohol (3.9 mm Hg) than in those consuming some alcohol (1.2 mm Hg) (P value for interaction =.03). Sixty-one percent of participants reported no alcohol consumption at all.

Multivariate regression models confirmed the significant treatment × race and treatment × hypertension status interactions for systolic blood pressure, and the significant treatment × hypertension status and treatment × alcohol intake interactions for diastolic blood pressure. The net effects were comparable to the unadjusted estimates in Figure 3.

The fruits-and-vegetables diet had effects that were generally intermediate between the control and combination diets, though the blood pressure reductions were not always statistically significant when compared with the changes in the control group (Table 4). There was more of a tendency for changes in systolic blood pressure to be statistically significant than corresponding changes in diastolic blood pressure. For most of the variables considered, the effect of the fruits-and-vegetables diet was statistically significant in 1 subgroup but not the other. For example, blood pressure reduction was statistically significant among African Americans but not among whites. Nonetheless, there was no statistically significant interaction between race and the blood pressure effect of the fruits-and-vegetables diet. The only variable that significantly affected the blood pressure response to the fruits-and-vegetables diet was hypertension status. Blood pressure declined 7.1/2.8 mm Hg (P<.001 for systolic blood pressure, P<.01 for diastolic blood pressure) among hypertensive participants assigned to the fruits-and-vegetables diet, and by 0.9/0.4 mm Hg (P>.20 for systolic and diastolic blood pressure) in the normotensive participants assigned to this diet (P value for interaction =.001 for systolic blood pressure, P=.07 for diastolic blood pressure). The fruits-and-vegetables diet lowered blood pressure by 8.0/3.4 mm Hg among hypertensive African Americans, 1.3/0.3 mm Hg in normotensive African Americans, and 5.9/3.1 mm Hg among hypertensive whites, and increased blood pressure by a nonsignificant 0.8/0.4 mm Hg among normotensive whites.

These analyses demonstrate that the DASH combination diet lowers blood pressure in virtually all subgroups and is, therefore, an intervention that has the potential to be broadly applicable and effective. In a number of characteristics, the study population was comparable to the general US population. Specifically, census data indicate that 81.2% of the US population are high school graduates (vs 82% of DASH participants), and that the median family income is $34,076 (vs approximately $35,000 for DASH participants).^18- 19 The US population is 48.8% male (vs 49% of DASH participants) and 12.5% African American (vs 60%, intentionally overrepresented).¹⁹ The mean score on the 7-day Physical Activity Recall survey of 158 kJ/kg per day (37.7 kcal/kg per day) is consistent with means found in population samples in which this instrument has been used.¹³

The DASH combination diet was particularly effective in African Americans and in individuals with hypertension. The joint effect in hypertensive African Americans was a 13.2/6.1 mm Hg decrease in blood pressure, an effect with important clinical and public health relevance. Increased efficacy of the DASH combination diet among African Americans supports other data suggesting racial differences in blood pressure response to diet. For example, in some reports, African Americans tend to consume less dietary potassium²⁰ and may have a greater blood pressure–lowering effect of increased potassium intake.^21- 22 The DASH combination diet includes high levels of potassium, which may explain some of the increased efficacy in this segment of the population. However, the effect cannot be attributed solely to increased "potassium sensitivity" among African Americans because the combination diet reduced blood pressure more than the fruits-and-vegetables diet even though both of these diets provided similar amounts of potassium. Furthermore, there were other nutrient differences in the combination diet compared with the fruits-and-vegetables and the control diets.

The increased efficacy in African Americans cannot be attributed to uneven distribution of female sex and obesity, since neither of these characteristics significantly affected the blood pressure response to the diet. Furthermore, small differences between blacks and whites in dietary intake of potassium (2940 vs 3396 mg/d), calcium (749 vs 998 mg/d), and sodium (3512 vs 4104 mg/d) at entry into the study could not account for the racial difference in blood pressure effect of the DASH combination diet because changes in blood pressure were calculated comparing end-of-intervention measurements with those obtained at the end of the run-in period when all participants had identical intake. Thus, the method by which the DASH combination diet lowers blood pressure is unknown. But regardless of the mechanism of action, a dietary intervention that provides distinctive benefit to African Americans is particularly welcome, since this group suffers a disproportionate burden of morbidity and mortality from hypertension.¹

The finding that the systolic blood pressure effects of the DASH combination diet vary by race is different from that of previous analyses of these data, which reported no diet × race interaction.³ The previous analysis looked at the 3 diets jointly, and demonstrated that they were not identical with regard to blood pressure effects. This analysis refines that observation by separately comparing each intervention diet vs the control diet, and when looked at in this way, there are clear racial differences in response to the diet.

The DASH combination diet had similar effects in subgroups known to respond differently to other blood pressure–lowering interventions, ie, subgroups defined by sex, age, BMI, and socioeconomic status. The similar effect of the DASH combination diet in all subgroup strata may be a function of low power to detect interactions. Nonetheless, significant blood pressure–lowering effects in all these strata suggest the broad applicability of this intervention. The apparent increased effect in nondrinkers compared with those who habitually consume some alcohol is intriguing and warrants further investigation.

The DASH combination diet was also more effective in participants with stage 1 hypertension (diastolic blood pressure of 90-95 mm Hg) compared with those with high-normal blood pressure, suggesting a role for this intervention in the treatment of established hypertension. We do not know the efficacy of this dietary pattern in individuals with more severe levels of hypertension. Typically, nonpharmacologic and drug therapies are more effective as the severity of hypertension increases.²³ In fact, within the range of blood pressure studied, the DASH combination diet lowered blood pressure more as baseline blood pressure increased. Therefore, it is conceivable that the DASH combination diet will offer an effective strategy, presumably in combination with pharmacotherapy, for hypertension of stage 2 and higher, but this hypothesis must be tested directly. Even if we restrict our focus to individuals with stage 1 hypertension, this intervention has a potentially large clinical impact. The greatest number of hypertensive individuals have stage 1 hypertension, and, therefore, the greatest burden of hypertension-related morbidity and mortality occurs in the population with this degree of hypertension.¹ The effect of the DASH combination diet on the blood pressure of individuals with clinical hypertension, an effect comparable to that of pharmacologic monotherapy, suggests that this intervention, if adhered to, has the potential to obviate or reduce the need for antihypertensive medication in a substantial proportion of the hypertensive population.

Despite the greater effect in hypertensive persons, the DASH combination diet was also effective in those with high-normal blood pressure, suggesting a role for this dietary intervention in primary prevention both in clinical populations and in the general population. Individuals with high-normal blood pressure are at increased risk of blood pressure–related morbidity and mortality even though they do not have clinical hypertension.²⁴ In addition, because blood pressure increases with age, it is the population with high-normal blood pressure now that is most likely to develop hypertension in the future. A strategy for lowering blood pressure in this population should reduce blood pressure–related events and should prevent clinical hypertension. Furthermore, if applied broadly to the public, the DASH combination diet could lead to a small decrease in the population distribution of blood pressure that would result in a large decrease in the number of cardiovascular events.²⁵

The ideal preventive strategy is reasonably priced, low risk, and easily implemented. The DASH combination diet, which included readily available foods and which was well accepted by study participants in a controlled feeding study, may meet these criteria. The cost of the DASH combination diet bought off the shelf at a grocery store is approximately $130 per week for a family of 4. This estimate places the DASH combination diet between "low cost" and "moderate," based on US Department of Agriculture estimates of food bills for a typical American family of 4 in January 1997.²⁶ Additional research will be needed to determine the extent to which free-living individuals can implement this dietary pattern.

The DASH combination diet lowered blood pressure while weight and sodium intake were held constant. This is an extremely important aspect of our study results. Both sodium reduction and weight loss are effective nonpharmacologic strategies for the prevention and treatment of hypertension,¹ and our data do not in any way argue against or preclude these established recommendations. The effect of established strategies applied in the context of the DASH combination diet is currently unknown, but the effect of the DASH combination diet with reduced sodium intake is under investigation. In addition, the DASH combination diet is consistent with all other public health recommendations currently advocated for reduction of heart disease, cancer, and osteoporosis.

The observed blood pressure effects were extremely robust. Not only was blood pressure reduced in virtually all subgroups examined but also this pattern was consistently observed across all cohorts and across all 4 sites (data not shown). The robustness of our findings suggests that they are not the result of chance or an unusual study population or locale. Our data suggest that the DASH combination diet is effective regardless of the underlying demographic and biologic characteristics that we investigated. However, there are several other factors that may influence the efficacy of this dietary pattern. For example, baseline dietary pattern, renin activity, insulin resistance, genetic influences, and gene-environment interactions may be important modulators of this response and are currently under investigation. Additional study is needed to determine the extent to which the DASH combination diet lowers blood pressure in the setting of other nonpharmacologic interventions, in patients taking antihypertensive medication, in those with stage 2 or 3 hypertension, and in individuals implementing the DASH combination diet on their own.

In conclusion, the DASH combination diet, without sodium reduction or weight loss, had significant blood pressure–lowering effects in virtually all subgroups. These effects were particularly striking in African Americans and in those with stage 1 hypertension. This intervention adds to our current nonpharmacologic approaches to treatment and prevention of high blood pressure. The DASH combination diet may be an effective strategy for preventing and treating hypertension in a broad cross section of the population, including segments of the population at high risk for hypertension and its complications.

Accepted for publication April 20, 1998.

This work was supported by grants HL50981, HL50968, HL50972, HL50977, HL50982, HL02642, RR02635, and RR00722 from the National Heart, Lung, and Blood Institute, the Office of Research on Minority Health, and the National Center for Research Resources of the National Institutes of Health, Bethesda, Md.

Presented at Biomedicine '97 (Scientific Session of the American Federation of Medical Research), Washington, DC, April 27, 1997.

We are indebted to the trial participants for their sustained commitment to DASH.

Information about DASH is available at http://www.dash.bwh.harvard.edu.

Reprints: Laura P. Svetkey, MD, MHS, Duke Hypertension Center, 3020 Pickett Rd, Suite 401, Durham, NC 27705.

Division of Epidemiology and Clinical Applications, National Heart, Lung, and Blood Institute, Bethesda, Md (sponsor): E. Obarzanek, PhD (project officer), J. A. Cutler, MD, M. A. Proschan, PhD; Kaiser Permanente Center for Health Research, Portland, Ore (coordinating center): T. M. Vogt, MD (principal investigator), N. Karanja, PhD, P. LaChance, R. Laws, C. Eddy, J. Rice, K. Linton, L. Haworth, N. Adams, K. Pearson, L. Diller, and J. Taylor; Brigham and Women's Hospital and Harvard Medical School, Boston, Mass (clinical center): T. J. Moore, MD (principal investigator), F. M. Sacks, MD, L. Jaffe, J. McKnight, MD, M. MacDonald, K. Nauth, and Y. Courtney; Duke University Medical Center, Durham, NC (clinical center): M. Drezner, MD, P.-H. Lin, PhD, C. Bales, PhD, L. Carter-Edwards, PhD, C. Plaisted, K. Hoben, S. Norris, P. Reams, K. Aicher, and R. Fike; Pennington Biomedical Research Center, Louisiana State University, Baton Rouge (clinical center): G. Bray, MD (principal investigator), D. Harsha, PhD, M. M. Windhauser, PhD, C. M. Champagne, PhD, P. J. Wozniak, PhD, B. McGee, and S. Crawford; Johns Hopkins University, Baltimore, Md (clinical center): B. Caballero, MD, PhD, S. Kumanyika, PhD, E. R. Miller, MD, S. Jee, J. Charleston, P. McCarron, S. Cappelli, B. Harnish, and P. Coleman; Virginia Polytechnic Institute, Blacksburg (food analysis coordinating center): K. K. Stewart, PhD, and K. Phillips; Oregon Health Sciences University, Portland (central laboratory): D. McCarron, MD, J.-B. Roulet, and R. Illingworth, PhD; Beltsville Human Nutrition Research Center, Department of Agriculture, Beltsville, Md (research kitchen for Johns Hopkins Clinical Center): S. Burns, E. Lashley, and J. T. Spence.