Beruflich Dokumente
Kultur Dokumente
Received November 21, 2015; first decision December 9, 2015; revision accepted January 27, 2016.
From the Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL (H.C.G.); Department of Medicine, University
of Texas Southwestern Medical Center, Dallas (S.G.R.); Department of Cardiology, Emory University School of Medicine, Atlanta, GA (V.V.); and
Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA (V.V., M.K.A.).
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.
115.06853.
Correspondence to Mohammed K. Ali, Rollins School of Public Health, Emory University, 1518 Clifton Rd NE, Atlanta, GA 30322. Email mkali@emory.edu
© 2016 American Heart Association, Inc.
Hypertension is available at http://hyper.ahajournals.org DOI: 10.1161/HYPERTENSIONAHA.115.06853
standard follow-up; (5) reported net change in systolic BP (SBP) study duration (<12 versus 12–24 versus >24 months), sample size
and diastolic BP (DBP; or the information with which to calculate (<100 versus 100–1000 versus >1000 participants), body mass index
these data) and its associated variance, confidence intervals (CIs), or (BMI) at baseline (<30 versus 30–35 versus >35 kg/m2), presence
P value, and (6) had a duration of at least 6 months. Reasons for ex- or absence of physical activity recommendations, and whether BP
clusion included: (1) trials conducted in populations with secondary reduction was considered the trial’s primary outcome. The pooled ef-
hypertension, that is, chronic kidney disease, renovascular disease, fect size for subgroups was determined using a random effects mod-
or certain endocrine disorders; (2) trials conducted exclusively in el, and ANOVA was used to test for statistical differences between
patients with congestive heart failure (these patients are regularly subgroups. In addition, meta-regression analyses were conducted
prescribed medications that affect BP); (3) trials with overlapping to examine the impact of weight reduction (when reported) on net
participants; (4) trials of nutritional supplementation (such as fish oil BP effect, as weight loss has been independently correlated with BP
or docosahexaenoic acid) as opposed to alteration of dietary pattern change.12
consumption; and (5) lack of a control group. There was no sample Please review the online-only Data Supplement for a full review of
size requirement for inclusion or exclusion.) bias assessment methods.
Please review the online-only Data Supplement for a full review of Comprehensive meta-analysis, version 3 (Biostat, Englewood,
the data extraction methods and dietary categorization criteria. NJ), was used to perform meta-regression and subgroup analy-
ses, and Review Manager, version 5 (The Cochrane Collaboration,
Statistical Analysis Copenhagen, Denmark) was used for all other analyses.
Net BP effect during the duration of follow-up, between intervention
and control groups, was calculated by subtracting the baseline (b) to Results
follow-up (fu) change in the control group from the corresponding
change in the intervention arm: (Ifu−Ib)−(Cfu−Cb). Treatment effects Study Selection and Characteristics
for each trial were weighted by the inverse of the variance (stan- The initial search returned 3201 studies once duplicates were
dard error [SE] for the net effect). If not reported directly, SEs were
derived from the CIs or P values of the net effect, or by calculation
removed. Figure 1 shows the number of studies identified and
from the individual standard deviation (SD) or SE of effects within excluded at various stages of the selection process. Ultimately,
parallel groups (assuming a correlation of 0.5 between variances at 24 individual trials were included in the analysis. From these
baseline and follow-up, as described by Follmann et al).16 The overall trials, there were a total of 39 comparison groups (39 unique
estimated mean net effect size of dietary modifications on BP was intervention groups and 27 control groups) with 23 858 total
pooled across trials using a random effects model to account for the
heterogeneity between studies with regard to the different dietary participants.
interventions and study designs, as well as participant ethnicity, age, Characteristics of the 24 trials with their respective inter-
sex, and health status. Heterogeneity was quantitatively assessed ventions and participants are detailed in Table S1 (online-only
using Q and I2 statistics. Data Supplement). Intervention arms varied in size from 11 to
Several preplanned subgroup analyses were conducted. These a
priori subanalyses involved examining net BP effect by: pre-existing
2570 participants (median: 129). Trial duration ranged from
hypertensive status, antihypertensive use, age (<50 versus ≥50 years), 6 to 48 months of follow-up (median: 12 months). Twenty-
sex (<50% versus ≥50% male), pre-existing diabetes mellitus status, one of the 36 comparison groups that reported sex distribution
Downloaded from http://hyper.ahajournals.org/ at Emory University on February 23, 2016
Gay et al Diet and Blood Pressure 3
Figure 2. Average net effect for (A) systolic blood pressure and (B) diastolic blood pressure, and corresponding 95% confidence intervals
(CIs), among all diets. Average net BP effect is calculated as the net incremental change in the diet group vs the control group. Horizontal
lines indicate 95% CIs of the estimate. All results are reported in mm Hg.
were predominantly male. Ten trials were conducted in the and −4.22 mm Hg [95% CI, −5.87 to −2.57], respectively).
United States, and 14 were based internationally. Race dis- Mediterranean diets (4 trials, 5 comparison groups) had no
tribution was not consistently reported in trials conducted statistically significant net effect on SBP (−1.17 mm Hg [95%
outside of the United States, whereas race was predominately CI, −2.81 to 0.46]), but did for DBP (−1.44 mm Hg [95% CI,
white in US studies. Age ranged from 34 to 67 years (median: −2.11 to −0.76]). Low-sodium diets (6 comparison groups)
45 years). There were 20 trials that reported hypertension led to a pooled net decrease in SBP and DBP of −2.06 mm Hg
status at baseline; among them, 14 comparisons exclusively (95% CI, −3.50 to −0.63) and −1.30 mm Hg (95% CI, −2.37
included hypertensive patients, whereas another 11 com- to −0.23), respectively. Participants adopting combined low-
parisons included a mix of hypertensive and normotensive sodium, high-potassium diets (5 comparison groups) experi-
patients. Ten trials (14 comparison groups) included patients enced net SBP and DBP decreases of −3.14 mm Hg (95% CI,
who were taking BP medication at baseline; in 4 of these, −6.27 to −0.02) and −2.01 mm Hg (95% CI, −3.40 to −0.62),
100% of participants were receiving an antihypertensive med- respectively, whereas those adopting low-sodium, low-calorie
ication. Mean BMI was in the overweight range (25–<30) in diets (5 comparison groups) experienced net SBP and DBP
22 comparison groups and the obese range (≥30) for all others decreases of −2.38 mm Hg (95% CI, −3.79 to −0.98) and
(14 comparison groups). There were 3 comparison groups that −1.33 mm Hg (95% CI, −2.04 to −0.62). Among 13 compari-
exclusively included participants with diabetes mellitus and son groups (11 trials) implementing low-calorie diets (some
14 comparisons composed of people without diabetes mel- with low-fat components), the pooled net SBP effect was
litus. Average baseline SBP and DBP ranged from 123.7 to −3.18 mm Hg (95% CI, −4.24 to −2.11) and −1.28 mm Hg
158.0 mm Hg (mean: 136.2 mm Hg) and 69.9 to 101.0 mm Hg (95% CI, −1.87 to −0.69) for DBP.
(mean: 85.7 mm Hg), respectively.
Subgroup Analysis and Meta-Regression
Change in BP Table summarizes the pooled net BP effects observed across
Forrest plots for net SBP and DBP effect among all diets are different subgroups. All subgroups experienced statistically
presented in Figure 2. Net SBP and DBP effect ranged from significant incremental BP reductions except: (1) people
−12.10 to 7.00 mm Hg and −9.32 to 0.20 mm Hg, respectively. with diabetes mellitus, SBP (95% CI, −5.14 to 0.86); (2)
The overall pooled net effect of diet on SBP was −3.07 mm Hg baseline BMI >35, SBP (95% CI, −5.14 to 0.86); and (3) tri-
(95% CI, −3.85 to −2.30) and for DBP was −1.81 mm Hg als with >1000 participants, SBP (95% CI, −3.77 to 0.22).
(95% CI, −2.24 to −1.38). Larger net SBP (P=0.03) and DBP (P=0.02) reductions
When trials were grouped by dietary patterns adopted were noted among participants with pre-existing hyperten-
(Figure 3; Figure S1 ), the largest net BP effect was seen sion at baseline, compared with normotensives. Similarly,
among Dietary Approaches to Stop Hypertension (DASH) participants not already taking antihypertensive medications
diets (4 trials, 5 comparison groups with pooled net SBP experienced significantly greater net SBP (P=0.01) and DBP
and DBP effects of −7.62 mm Hg [95% CI, −9.95 to −5.29] (P=0.008) declines, compared with their counterparts already
Downloaded from http://hyper.ahajournals.org/ at Emory University on February 23, 2016
4 Hypertension April 2016
Figure 3. Average net effect for (A) systolic blood pressure and (B) diastolic blood pressure, and corresponding 95% confidence intervals
(CIs), summarized by diet type. Average net BP effect is calculated as the net incremental change in the diet group versus the control
group. Horizontal lines indicate 95% CIs of the estimate. All results are reported in mm Hg.
receiving pharmacological therapies. Groups with longer fol- subgroups. Although the overall net effect was modest from
low-up (>24 months) had a significantly smaller effect size an individual perspective, it is important to note that the con-
(P<0.001 for both SBP and DBP) than those with medium trol groups also experienced some benefit and the pooled
(21–24 months) and short (<12 months) follow-ups. Smaller results were incremental BP reductions experienced by those
trials (n<100 participants) exhibited larger net effects (SBP groups adopting dietary interventions. Furthermore, from a
[P<0.001] and DBP [P=0.001]) than medium (n=100–1000) population standpoint, even relatively small reductions in BP
and larger (n>1000) studies. There were significantly smaller can dramatically reduce the incidence of cardiovascular dis-
net effects for both SBP (P=0.05) and DBP (P=0.04) in trials ease and mortality.17 The range of effects also suggests that
where the primary outcome was BP reduction. Net BP change some patients may benefit to a greater degree than others.
was not significantly different across participants of different Our results have important clinical and public health impli-
age, sex, diabetes mellitus status, baseline BMI, or in trials cations, suggesting that dietary modifications are an effective
which also encouraged physical activity. method for controlling BP within the population and that cer-
There were 30 comparison groups reporting net weight tain approaches are better targeted to individuals with specific
change (baseline to follow-up change in intervention minus characteristics.
the corresponding change in control), which ranged from For all dietary interventions, compared with control
−16.0 to +1.4 kg (mean: −4.56 kg). In meta-regression analy- groups, our study identified an incremental BP-lowering
sis (Figure S2), there was a significant relationship noted effect of −3.07 mm Hg and −1.81 mm Hg for SBP and DBP,
between mean incremental weight loss in the intervention respectively. Among the various diet subtypes, the DASH diet
group and net BP effect for both SBP (P<0.001) and DBP was associated with the greatest overall reduction in BP, with
(P=0.01). Specifically, for every additional 1 kg of weight loss a net SBP effect of −7.62 mm Hg and a DBP effect of −4.22
experienced by intervention participants, there was an associ- mm Hg. Importantly, this magnitude is similar to trials exam-
ated 0.36 mm Hg additional reduction in SBP (95% CI, 0.20– ining single drug therapies in mild hypertension, suggesting
0.52) and 0.13 mm Hg additional reduction in DBP (95% CI, the DASH dietary pattern may be an alternative to medica-
0.03–0.24). tion initiation in early stage hypertension.6,18 Significant, but
Please review the online-only Data Supplement for bias smaller BP reductions were also evident among other dietary
assessment results. patterns. Furthermore, a recent meta-analysis found that veg-
etarian diets also led to lower SBP and DBP, though the trials
Discussion included in that study were either too short for our analysis
This meta-analysis of 24 randomized controlled trials, cover- or conducted before 1990.9 Interestingly, although partici-
ing studies during a 25-year period and including over 23 000 pants adopting Mediterranean diets had significantly greater
participants, showed that adopting healthful dietary modifica- net DBP reductions, the effect on SBP was not statistically
tions led to significant incremental reductions in both SBP and greater than that experienced among control groups. This
DBP versus control. These effects were generally consistent is meaningful as a recent, major clinical trial showed that
across different types of diets and among various population Mediterranean diets are associated with a lower incidence of
Downloaded from http://hyper.ahajournals.org/ at Emory University on February 23, 2016
Gay et al Diet and Blood Pressure 5
Table. Average Net Change in Systolic and Diastolic Blood Pressure by Defined Study Design and Participant Characteristic
Subgroups
Systolic Blood Pressure Diastolic Blood Pressure
Stratum Comparison Mean Difference Intracategory; ANOVA; Comparison Mean Difference Intracategory; ANOVA;
Groups* (95% CI) P Value† P Value‡ Groups* (95% CI) P Value† P Value‡
Hypertensive status
Hypertensive 14 −3.31 (−4.86 to −1.77) <0.001 0.03 14 −2.24 (−3.16 to −1.32) <0.001 0.02
Normotensive 11 −1.56 (−2.05 to −1.06) <0.001 11 −0.97 (−1.44 to −0.51) <0.001
Antihypertensive medications
Yes 14 −1.87 (−3.01 to −0.73) 0.001 0.01 14 −1.18 (−1.46 to −0.91) <0.001 0.008
No 23 −3.83 (−4.80 to −2.86) <0.001 23 −2.07 (−2.67 to −1.48) <0.001
Age, y
<50 22 −3.52 (−4.59 to −2.44) <0.001 0.21 22 −1.97 (−2.65 to −1.29) <0.001 0.31
≥50 16 −2.48 (−3.70 to −1.27) <0.001 16 −1.54 (−2.00 to −1.07) <0.001
Sex
≥50% Male 21 −2.84 (−3.80 to −1.88) <0.001 0.88 21 −1.63 (−2.24 to −1.03) <0.001 0.79
<50% Male 15 −2.96 (−4.17 to −1.76) <0.001 15 −1.76 (−2.44 to −1.08) <0.001
Diabetic status
Diabetic 3 −2.14 (−5.14 to 0.86) 0.16 0.69 3 −1.20 (−1.73 to −0.68) <0.001 0.20
Nondiabetic 14 −2.80 (−4.01 to −1.60) <0.001 14 −1.73 (−2.68 to −1.28) <0.001
Study duration
<12 mo 13 −5.25 (−7.16 to −3.35) <0.001 <0.001 13 −2.95 (−4.40 to −1.51) <0.001 <0.001
12–24 mo 13 −3.27 (−4.34 to −2.20) <0.001 13 −1.92 (−2.36 to −1.47) <0.001
>24 mo 13 −1.36 (−1.99 to −0.73) <0.001 13 −0.96 (−1.31 to −0.61) <0.001
Intervention sample size
<100 18 −5.02 (−6.32 to −3.72) <0.001 <0.001 18 −2.76 (−3.55 to −1.97) <0.001 0.001
100–1000 17 −2.05 (−2.65 to −1.45) <0.001 17 −1.17 (−1.55 to −0.80) <0.001
>1000 4 −1.77 (−3.77 to 0.22) 0.08 4 −1.23 (−1.73 to −0.73) <0.001
Baseline BMI
<30 22 −3.63 (−4.75 to −2.51) <0.001 0.30 22 −2.03 (−2.62 to −1.45) <0.001 0.12
30–35 11 −2.58 (−3.45 to −1.70) <0.001 11 −1.63 (−2.34 to −0.91) <0.001
>35 3 −2.14 (−5.14 to 0.86) 0.16 3 −1.20 (−1.73 to −0.68) <0.001
BP primary outcome
Yes 24 −2.47 (−3.46 to −1.49) <0.001 0.05 24 −1.45 (1.93 to −0.97) <0.001 0.04
No 15 −4.15 (−5.51 to −2.79) <0.001 15 −2.44 (−3.25 to −1.64) <0.001
Physical activity encouraged
Yes 21 −2.70 (−3.23 to −2.08) <0.001 0.30 21 −1.68 (−2.09 to −1.26) <0.001 0.74
No 18 −3.54 (−4.99 to −2.09) <0.001 18 −1.83 (−2.60 to −1.06) <0.001
BMI recorded in kg/m . BMI indicates body mass index; BP, blood pressure; and CI, confidence interval.
2
*The number of included comparison groups may not total 39 as a result of missing data.
†P value reflects effect of dietary intervention vs control within each specific subgroup.
‡P value reflects intracategory variance in mean effect as determined through ANOVA methodology.
cardiovascular events and death.19 This finding, along with Subgroup analyses indicated that net BP reduction was
the results of our meta-analysis, suggest that there may be not as great in trials of longer duration. Specifically, trials
alternative cardiovascular advantages, beyond BP control, that lasted longer than 24 months had lower mean incremen-
that contribute to the mortality benefit of the Mediterranean tal BP reductions than trials conducted between 12 to 24
diet, as has been proposed in other studies.20 Further research months, whereas the greatest BP effect was noted in trials
is needed in this area. lasting <12 months. This is a finding that has been described
Downloaded from http://hyper.ahajournals.org/ at Emory University on February 23, 2016
6 Hypertension April 2016
in other analyses of lifestyle interventions and BP, such as that the 2013 American College of Cardiology/American Heart
conducted by Whelton et al.10 A similar effect was also seen Association lifestyle management guidelines.13 In addition, as
within individual trials, such as the PREMIER trial, which explained above, there were greater BP reductions among stud-
showed a declining level of BP reduction when participants ies of shorter duration and smaller size; not all interventions
were examined after 18 months of follow-up compared with were equally distributed across the follow-up or sample size
the original trial of 6-month follow-up.21,22 The most probable spectrum, which allows for the possibility of bias within our
explanation for this finding is that of declining adherence to dietary subanalysis. For example, although DASH diet inter-
the diet over time, and suggests that participant adherence is of ventions exhibited the largest mean net effect size for both SBP
primary importance in any lifestyle intervention program. We and DBP, all the studies included in this category were short-
also found that larger (n>1000) and medium (n=100–1000) term (<12 months) trials, which may partially explain the size
sized trials had less incremental BP effects than smaller tri- of the effect. Given our inclusion criteria for study duration,
als, possibly reflecting that larger trials tended to be of longer important clinical trials such as the original DASH trial and
duration with a wider pool of individual risk and motivation, the OmniHeart trial were excluded from the analysis. However,
whereas smaller studies may include the most motivated par- as noted above, dietary adherence has important implications
ticipants. Trials conducted among individuals with hyperten- on BP effect size, and this exclusion requirement limits any
sion showed significantly greater net BP changes than those over estimation of long-term BP reduction. In addition, the
conducted in normotensive participants, which may reflect subanalyses were limited by what data were available in pub-
healthier baseline diet and lifestyle practices among nonhy- lished studies; as such, subanalyses based on race, which may
pertensive individuals. Other studies have described a ben- interact in important ways with dietary interventions such as
eficial influence of physical activity level on BP reduction, sodium reduction, were not possible.
however, in our analysis; there was no incremental BP benefit
in trials encouraging physical activity versus those that did Perspectives
not.10 There was a larger BP effect experienced by individuals This meta-analysis provides important evidence that dietary
who were not already on antihypertensive medications, which interventions offer clinically significant net BP reductions,
may have ramifications as treated individuals make up a large and that some dietary patterns may be more effective than oth-
portion of the prevalent hypertensive population. It is impor- ers. The public health and clinical implications of this research
tant to note, however, that there was still a significant effect are important, and healthcare providers should consider these
appreciated (for both SBP and DBP) in patients currently when giving dietary recommendations. The DASH interven-
undergoing pharmacological therapy, and it may be true that tion that contains many aspects of other approaches (low
dietary modifications would allow for medication withdrawal sodium recommendation and high-potassium, low-fat food
over time. groups) may be the most appropriate initial recommendation
We detected an independent association of weight loss when BP control is the principle objective, although weight
with BP reduction. This result was observed irrespective of loss and other factors are certainly relevant. This is true for
baseline BMI, which had no correlation with the magnitude patients who are in the prehypertension range as well as those
of net BP effects. Similar findings have been described in already taking antihypertensive medications. Importantly, the
previous analyses and several biologically plausible explana- magnitude of the net effect for the DASH dietary interven-
tions have been put forward.12 Weight loss was not, however, tion was similar to the effect of drug monotherapies for mild
associated with net BP reduction among low-sodium or low- hypertension, providing a possible alternative to medication
sodium, high-potassium diets; a probable reflection of the initiation in early stage hypertension. Adherence to any life-
intricate involvement of these electrolytes in renal BP regula- style modification is of primary importance and should always
tion, which likely has many influences other than weight.23 be addressed and evaluated. Additional studies are needed to
assess the long-term mortality and morbidity effects from BP
Strengths and Limitations reduction through dietary intervention.
Strengths of this study include a rigorous and robust level of
clinical evidence by limiting inclusion to only randomized
Disclosures
controlled trials, thereby restricting potential confounders. None.
Similarly, 3 comprehensive databases were used to complete
literature searches, covering a wide breadth of published stud-
ies and ultimately including 24 trials with >23 000 total par- References
1. Lim SS, Vos T, Flaxman AD, et al. A comparative risk assessment of bur-
ticipants. Furthermore, we limited our inclusion criteria for den of disease and injury attributable to 67 risk factors and risk factor clus-
study duration to trials lasting at least 6 months, which mini- ters in 21 regions, 1990-2010: a systematic analysis for the Global Burden
mizes any over estimation of long-term BP reduction arising of Disease Study 2010. Lancet. 2012;380:2224–2260. doi: 10.1016/
from short-term dietary modifications. S0140-6736(12)61766-8.
2. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional,
There are also some important limitations to point out. We and national age-sex specific all-cause and cause-specific mortality for
chose to limit the review and analysis to clinical trials completed 240 causes of death, 1990–2013: a systematic analysis for the global bur-
within the past 25 years; we acknowledge that trials were being den of disease study 2013. Lancet. 2015;385:117–171.
3. Mozaffarian D, Benjamin EJ, Go AS, et al; American Heart Association
conducted on this issue outside of this time frame, but feel that
Statistics Committee and Stroke Statistics Subcommittee. Heart disease and
25 years encompasses a relevant and robust body of evidence stroke statistics–2015 update: a report from the American Heart Association.
and is similar to date ranges used in the evidence review for Circulation. 2015;131:e29–322. doi: 10.1161/CIR.0000000000000152.
4. Murray CJ, Atkinson C, Bhalla K, et al; U.S. Burden of Disease Association Task Force on Practice Guidelines. Circulation. 2014;129(25
Collaborators. The state of US health, 1990-2010: burden of diseases, suppl 2):S76–S99. doi: 10.1161/01.cir.0000437740.48606.d1.
injuries, and risk factors. JAMA. 2013;310:591–608. doi: 10.1001/ 14. James PA, Oparil S, Carter BL, et al. 2014 evidence-based guideline for
jama.2013.13805. the management of high blood pressure in adults: report from the panel
5. Miller ER III, Erlinger TP, Young DR, Jehn M, Charleston J, Rhodes members appointed to the Eighth Joint National Committee (JNC 8).
D, Wasan SK, Appel LJ. Results of the Diet, Exercise, and Weight Loss JAMA. 2014;311:507–520. doi: 10.1001/jama.2013.284427.
Intervention Trial (DEW-IT). Hypertension. 2002;40:612–618. 15. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred
6. Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, Sacks FM, reporting items for systematic reviews and meta-analyses: the PRISMA
Bray GA, Vogt TM, Cutler JA, Windhauser MM, Lin PH, Karanja N. A statement. Ann Intern Med. 2009;151:264–9, W64.
clinical trial of the effects of dietary patterns on blood pressure. DASH 16. Follmann D, Elliott P, Suh I, Cutler J. Variance imputation for over-
Collaborative Research Group. N Engl J Med. 1997;336:1117–1124. doi: views of clinical trials with continuous response. J Clin Epidemiol.
10.1056/NEJM199704173361601. 1992;45:769–773.
7. Appel LJ, Sacks FM, Carey VJ, Obarzanek E, Swain JF, 17. Cook NR, Cohen J, Hebert PR, Taylor JO, Hennekens CH. Implications of
Miller ER III, Conlin PR, Erlinger TP, Rosner BA, Laranjo NM, small reductions in diastolic blood pressure for primary prevention. Arch
Charleston J, McCarron P, Bishop LM; OmniHeart Collaborative Intern Med. 1995;155:701–709.
Research Group. Effects of protein, monounsaturated fat, and car- 18. The Treatment of Mild Hypertension Research Group. The treatment
bohydrate intake on blood pressure and serum lipids: results of the of mild hypertension study. A randomized, placebo-controlled trial of
OmniHeart randomized trial. JAMA. 2005;294:2455–2464. doi: 10.1001/ a nutritional-hygienic regimen along with various drug monotherapies.
jama.294.19.2455. Arch Intern Med. 1991;151:1413–1423.
8. Sacks FM, Campos H. Dietary therapy in hypertension. N Engl J Med. 19. Estruch R, Ros E, Salas-Salvadó J, et al; PREDIMED Study Investigators.
2010;362:2102–2112. doi: 10.1056/NEJMct0911013. Primary prevention of cardiovascular disease with a Mediterranean diet.
9. Yokoyama Y, Nishimura K, Barnard ND, Takegami M, Watanabe M, N Engl J Med. 2013;368:1279–1290. doi: 10.1056/NEJMoa1200303.
Sekikawa A, Okamura T, Miyamoto Y. Vegetarian diets and blood pres- 20. Rees K, Hartley L, Flowers N, Clarke A, Hooper L, Thorogood M,
sure: a meta-analysis. JAMA Intern Med. 2014;174:577–587. doi: Stranges S. ‘Mediterranean’ dietary pattern for the primary prevention of
10.1001/jamainternmed.2013.14547. cardiovascular disease. Cochrane Database Syst Rev. 2013;8:CD009825.
10. Whelton SP, Chin A, Xin X, He J. Effect of aerobic exercise on blood doi: 10.1002/14651858.CD009825.pub2.
pressure: a meta-analysis of randomized, controlled trials. Ann Intern 21. Appel LJ, Champagne CM, Harsha DW, Cooper LS, Obarzanek E, Elmer
Med. 2002;136:493–503. PJ, Stevens VJ, Vollmer WM, Lin PH, Svetkey LP, Stedman SW, Young
11. Rebholz CM, Friedman EE, Powers LJ, Arroyave WD, He J, Kelly TN. DR; Writing Group of the PREMIER Collaborative Research Group.
Dietary protein intake and blood pressure: a meta-analysis of random- Effects of comprehensive lifestyle modification on blood pressure control:
ized controlled trials. Am J Epidemiol. 2012;176(suppl 7):S27–S43. doi: main results of the PREMIER clinical trial. JAMA. 2003;289:2083–2093.
10.1093/aje/kws245. doi: 10.1001/jama.289.16.2083.
12. Neter JE, Stam BE, Kok FJ, Grobbee DE, Geleijnse JM. Influence of 22. Elmer PJ, Obarzanek E, Vollmer WM, Simons-Morton D, Stevens VJ,
weight reduction on blood pressure: a meta-analysis of randomized Young DR, Lin PH, Champagne C, Harsha DW, Svetkey LP, Ard J, Brantley
controlled trials. Hypertension. 2003;42:878–884. doi: 10.1161/01. PJ, Proschan MA, Erlinger TP, Appel LJ; PREMIER Collaborative
HYP.0000094221.86888.AE. Research Group. Effects of comprehensive lifestyle modification on diet,
13. Eckel RH, Jakicic JM, Ard JD, et al; American College of Cardiology/ weight, physical fitness, and blood pressure control: 18-month results of a
American Heart Association Task Force on Practice Guidelines. 2013 randomized trial. Ann Intern Med. 2006;144:485–495.
AHA/ACC guideline on lifestyle management to reduce cardiovascular 23. Langford HG. Sodium-potassium interaction in hypertension and hyperten-
risk: a report of the American College of Cardiology/American Heart sive cardiovascular disease. Hypertension. 1991;17(suppl 1):I155–I157.
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://hyper.ahajournals.org/content/early/2016/02/22/HYPERTENSIONAHA.115.06853
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Hypertension can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial
Office. Once the online version of the published article for which permission is being requested is located,
click Request Permissions in the middle column of the Web page under Services. Further information about
this process is available in the Permissions and Rights Question and Answer document.
Hawkins C. Gay MD MPH1, Shreya G. Rao MD MPH2, Viola Vaccarino MD PhD3,4, and
1
Northwestern University Feinberg School of Medicine, Chicago, IL
2
University of Texas Southwestern Medical Center, Dallas, TX
3
Emory University School of Medicine, Atlanta, GA
4
Rollins School of Public Health, Emory University, Atlanta, GA
Data Extraction
We retrieved original articles to extract the following study characteristics using a
standardized entry database: primary author, date of publication, title, country of origin, sample
size, participant demographics, participant health status (i.e. diabetes, coronary artery disease,
etc.), use of antihypertensive medication, dietary intervention details, any physical activity
details, setting, duration of follow-up for all reported measurements, method of BP
ascertainment, effect size, weight loss, and bias assessment. If BP measurements were reported
at multiple points in time, data were analyzed for the period of longest follow-up. Some
individual studies examined several unique dietary interventions versus a common control; when
this was the case, each intervention was considered as a separate comparison against the control
group. In cases where incomplete data were presented in the published article, authors were
contacted in an effort to obtain the missing information.
Studies were categorized into different dietary pattern interventions based on the
approach examined in the trial. Dietary Approaches to Stop Hypertension (DASH) diets
followed a protocol involving increased consumption of fruits, vegetables, low-fat dairy, whole
grains, lean meats (fish and poultry), nuts and beans, as well as a reduced intake of red meat and
fat. Mediterranean diets were characterized by reduced saturated fat in favor of monounsaturated
and n-3 fatty acids from nuts, olive oil, and fish, as well as an increased amount of grains,
vegetables, and fruits, with decreased levels of red meat. Low sodium diets entailed consuming
below 2.3g (100 mmol/day) of total daily sodium. High potassium diets had this as a stated
objective of the trial. Low fat diets required that less than 30% of total calories be consumed
from fat. Low calorie diets adjusted the daily caloric intake to reduce weight by at least 4.5 kg or
5% of total body weight. When more than one dietary modification was included in the
intervention, the study was categorized as reporting on a combined intervention, e.g. as a low
sodium, high potassium diet.
Bias Assessment
The presence of publication bias was assessed via creation and visual inspection of funnel
plots, by plotting SEs against the effect size for each study. The Egger test was used to quantify
any asymmetry among the funnel plots, and the trim-and-fill method was utilized to adjust for
any omissions and potential bias. A Cochrane Risk of Bias tool was used to evaluate any
potential methodological concerns amongst included studies. Each trial was assessed for:
selection bias (random sequence allocation and allocation concealment), performance bias
(blinding participants and personnel), detection bias (blinding outcome assessors), attrition bias
(incomplete outcome data), and reporting bias (selective reporting). Each trial was classified as
low, high, or unclear on the risk of bias scale.
RESULTS
Bias Assessment
Funnel plots (Figure S3) of SE versus effect size for each study demonstrate that the
distribution of effect size estimates for individual comparison groups was approximately
symmetrical around the pooled estimate; however, potential missing studies were visually noted
in the region of small sample size and low to negative effect for both SBP and DBP. Further
assessment was completed via the Egger test, which did not identify any asymmetry in either
SBP (P = 0.48) or DBP (P = 0.45) plots. Sensitivity analysis was completed utilizing the trim-
and-fill method which did not identify any missing studies or change the effect size. Taken
together, this suggests there is a low likelihood of publication bias influencing the results
observed. Potential methodologic biases are detailed in Figure S4. Studies mostly generated a
low risk of bias, though in a small number of studies a comprehensive bias assessment revealed a
number of unclear risk components.
References
Table S1. Study design and baseline characteristics of randomized controlled trials of dietary interventions for BP.
Base Line BP
Duration Age Male HTN SBP DBP
†
Study *
Year Country N (mos) (y) (%) BMI Trmt (%)‡ (mmHg) (mmHg) Intervention
DPP 2005 US 1,079 36 51 32.0 33.9 17.0 123.7 78.6 Low Calorie,
Group1 Low Fat
HPT 1990 US 125 36 38 75.2 29.0 0.0 125.3 83.0 Low Calorie
Group2a
HPT 1990 US 196 36 39 63.3 27.0 0.0 124.0 82.6 Low Sodium
Group2b
HPT 1990 US 195 36 38 65.1 26.0 0.0 124.1 82.3 Low Sodium,
Group2c High
Potassium
HPT 1990 US 129 36 39 65.1 29.0 0.0 124.4 82.6 Low Sodium,
Group2d Low Calorie
Look 2007 US 2,570 12 59 40.7 35.9 75.3 128.2 69.9 Low Calorie,
AHEAD Low Fat
Group3
PREMIER 2003 US 268 6 50 35.1 33.0 0.0 135.5 85.0 Low Sodium,
Group4b Low Calorie
TOHP I 1992 US 327 18 43 70.9 27.1 0.0 124.8 83.7 Low Sodium
Group5a
TOHP I 1992 US 308 18 43 72.7 29.5 0.0 124.3 83.7 Low Calorie
Group5b
TOHP II 1997 US 594 36 44 64.8 30.9 0.0 127.7 86.1 Low Sodium
Group6a
TOHP II 1997 US 595 36 43 63.0 31.0 0.0 127.6 86.0 Low Calorie,
Group6b Low Fat
TOHP II 1997 US 597 36 44 68.8 30.9 0.0 127.4 86.0 Low Sodium,
Group6c Low Calorie,
Low Fat
Azadbakht 2005 Iran 11 6 41 100.0 29.9 0.0 143.0 88.0 Low Calorie
et al.8c
Azadbakht 2005 Iran 27 6 41 0.0 29.9 0.0 144.0 85.0 Low Calorie
et al.8d
Coppell et 2010 New 52 6 57 38.0 35.1 84.0 131.9 79.8 Low Calorie,
al.9 Zealand Low Fat
Geleijnse 1994 Netherlands 49 24 66 53.1 27.1 0.0 158.0 89.8 Low Sodium,
et al.12 High
Potassium
Hjelstuen 2007 Norway 51 12 54 100.0 29.5 100.0 137.6 87.7 Low Calorie,
et al.13 Low Fat
Jula et 1990 Finland 32 12 44 100.0 28.5 0.0 147.3 98.2 Low Sodium,
al.15a High
Potassium
Jula et 1990 Finland 15 12 44 0.0 26.2 0.0 154.1 98.3 Low Sodium,
al.15b High
Potassium
Kastarinen 2002 Finland 360 24 54 48.0 28.9 48.0 149.0 91.0 Low Sodium
et al.16
Makela et 2008 Finland 40 6 44 70.0 27.3 0.0 149.9 98.0 Low Sodium
al.18
Mattila et 2003 Finland 368 12 50 46.1 29.4 63.8 139.5 90.5 Low Sodium,
al.20 Low Calorie,
Low Fat
Toledo et 2013 Spain 2,441 48 67 41.7 29.9 68.0 148.0 83.0 Mediterranean
al.21a Dietary
Program
(EVOO)
Toledo et 2013 Spain 2,367 48 67 46.1 29.7 68.4 149.0 83.0 Mediterranean
al.21b Dietary
Program
(Nuts)
Whelton et 1998 US 340 30 67 50.9 27.9 100.0 127.3 71.3 Low Sodium
al.24a
Whelton et 1998 US 147 30 66 49.0 31.0 100.0 128.6 70.7 Low Calorie
al.24b
Whelton et 1998 US 147 30 66 56.0 31.2 100.0 127.6 71.3 Low Sodium,
al.24c Low Calorie
*
a, b, c, and d indicate different comparison groups within the same individual trial.
†
number of participants in each strata (excludes controls, as these were typically the same between strata of the same trial).
‡
% of participants taking an antihypertensive medication at the start of the trial.
BMI recorded in kg/m2.
Supplemental Figures
A.
B.
Figure S1. Average net change in (a) SBP and (b) DBP, and corresponding 95% confidence
intervals by diet type. Average net BP effect is calculated as the net incremental change in the
diet group versus the control group. Horizontal lines indicate 95% CIs of the estimate. SBP
indicates systolic blood pressure and DBP indicates diastolic blood pressure. All results are
reported in mmHg.
A.
Regression of Mean Difference SBP vs Weight Loss
12.50
10.00
7.50
5.00
Difference in Mean SBP
2.50
0.00
-2.50
-5.00
-7.50
-10.00
-12.50
-15.00
-17.50
-22.5 -20.0 -17.5 -15.0 -12.5 -10.0 -7.5 -5.0 -2.5 0.0 2.5 5.0 7.5
Weight Loss
B.
Regression of Mean Difference in DBP vs Weight Loss
4.00
2.00
0.00
Difference in Mean DBP
-2.00
-4.00
-6.00
-8.00
-10.00
-12.00
-22.5 -20.0 -17.5 -15.0 -12.5 -10.0 -7.5 -5.0 -2.5 0.0 2.5 5.0 7.5
Weight Loss
Figure S2. Meta-regression analysis of mean difference in (a.) SBP vs weight loss and (b.) DBP
vs. weight loss, in the 30 trials with data available for weight loss.
A. B.
Figure S3. Funnel plots of (a) systolic blood pressure and (b) diastolic blood pressure in
randomized controlled trials, plotted as net change in effect vs. SE of the net change.
Figure S4. Risk of bias summary; Green = Low Risk, Red = High Risk, Blank = Unclear Risk