Selection of clinical trials and observational studies for meta-analysis of association between vegetarian diets and blood pressure.
Effects on systolic BP (A) and on diastolic BP (B) are depicted as squares; error bars indicate 95% CIs. Meta-analysis yielded pooled estimates of −4.8 mm Hg (95% CI, −6.6 to −3.1) for systolic BP and −2.2 mm Hg (−3.5 to −1.0) for diastolic BP, which are depicted as blue diamonds. Vegan diets were defined as omitting all animal products; vegetarian diets may include some animal products as indicated by the terms lacto (dairy products) and ovo (eggs).
Effects on systolic BP are depicted as squares; error bars indicate 95% CIs. Meta-analysis yielded a pooled estimate of −6.9 mm Hg (95% CI, −9.1 to −4.7) for systolic BP, which is depicted as a blue diamond. Arrows indicate that the 95% CI exceeds the left line. Vegan diets were defined as omitting all animal products; vegetarian diets may include some animal products as indicated by the terms lacto (dairy products) and ovo (eggs).
Effects on diastolic BP are depicted as squares; error bars indicate 95% CIs. Meta-analysis yielded a pooled estimate of −4.7 mm Hg (95% CI, −6.3 to −3.1) for diastolic BP, which is depicted as a blue diamond. Arrows indicate that the 95% CI exceeds the left line. Vegan diets were defined as omitting all animal products; vegetarian diets may include some animal products as indicated by the terms lacto (dairy products) and ovo (eggs).
Funnel plot of study weights against change in systolic blood pressure (BP) in clinical trials (A) and observational studies (B). Blood pressure results in individual studies are depicted as circles scattered around the pooled BP estimate. A trim-and-fill method indicated that 3 clinical trials and no observational studies might have been missing owing to publication bias. After adjustment for putative missing data, the overall differences for systolic BP increased to −5.2 mm Hg (95% CI, −6.9 to −3.5) in clinical trials.
eTable 1. Electronic database search strategy
eTable 2. Designs and characteristics of populations of studies of vegetarian diets and blood pressure
eTable 3. Blood pressure responses of subgroups to vegetarian diets (clinical trials)
eTable 4. Blood pressure responses of subgroups to vegetarian diets (observational studies)
Yokoyama Y, Nishimura K, Barnard ND, Takegami M, Watanabe M, Sekikawa A, Okamura T, Miyamoto Y. Vegetarian Diets and Blood PressureA Meta-analysis. JAMA Intern Med. 2014;174(4):577-587. doi:10.1001/jamainternmed.2013.14547
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Previous studies have suggested an association between vegetarian diets and lower blood pressure (BP), but the relationship is not well established.
To conduct a systematic review and meta-analysis of controlled clinical trials and observational studies that have examined the association between vegetarian diets and BP.
MEDLINE and Web of Science were searched for articles published in English from 1946 to October 2013 and from 1900 to November 2013, respectively.
All studies met the inclusion criteria of the use of (1) participants older than 20 years, (2) vegetarian diets as an exposure or intervention, (3) mean difference in BP as an outcome, and (4) a controlled trial or observational study design. In addition, none met the exclusion criteria of (1) use of twin participants, (2) use of multiple interventions, (3) reporting only categorical BP data, or (4) reliance on case series or case reports.
Data Extraction and Synthesis
Data collected included study design, baseline characteristics of the study population, dietary data, and outcomes. The data were pooled using a random-effects model.
Main Outcomes and Measures
Net differences in systolic and diastolic BP associated with the consumption of vegetarian diets were assessed.
Of the 258 studies identified, 7 clinical trials and 32 observational studies met the inclusion criteria. In the 7 controlled trials (a total of 311 participants; mean age, 44.5 years), consumption of vegetarian diets was associated with a reduction in mean systolic BP (−4.8 mm Hg; 95% CI, −6.6 to −3.1; P < .001; I2 = 0; P = .45 for heterogeneity) and diastolic BP (−2.2 mm Hg; 95% CI, −3.5 to −1.0; P < .001; I2 = 0; P = .43 for heterogeneity) compared with the consumption of omnivorous diets. In the 32 observational studies (a total of 21 604 participants; mean age, 46.6 years), consumption of vegetarian diets was associated with lower mean systolic BP (−6.9 mm Hg; 95% CI, −9.1 to −4.7; P < .001; I2 = 91.4; P < .001 for heterogeneity) and diastolic BP (−4.7 mm Hg; 95% CI, −6.3 to −3.1; P < .001; I2 = 92.6; P < .001 for heterogeneity) compared with the consumption of omnivorous diets.
Conclusions and Relevance
Consumption of vegetarian diets is associated with lower BP. Such diets could be a useful nonpharmacologic means for reducing BP.
The relationship between blood pressure (BP) and cardiovascular disease risk is continuous, consistent, and independent of other risk factors.1 According to Lewington et al,2 in individuals aged 40 to 70 years, each increment of 20 mm Hg in systolic BP or 10 mm Hg in diastolic BP is associated with more than twice the risk of cardiovascular disease across the BP range from 115/75 to 185/115 mm Hg.
A substantial body of evidence supports the role of modifiable factors, including diet, body weight, physical activity, and alcohol intake, in the risk of developing hypertension.3 Dietary modifications have been shown3 to be particularly effective in preventing and managing hypertension.
Vegetarian diets are defined as dietary patterns that exclude or rarely include meats; some vegetarian diets include dairy products, eggs, and fish. All vegetarian diets emphasize foods of plant origin, particularly vegetables, grains, legumes, and fruits. In observational studies,4,5 consumption of vegetarian diets is associated with a lower prevalence of hypertension. Although some randomized clinical trials6,7 have found that adoption of a vegetarian diet reduces BP, others8,9 have not yielded similar results. To our knowledge, the available evidence regarding the association between vegetarian diets and BP has not been subjected to meta-analysis. To clarify the nature of this association and provide a valid estimate of the effect size regarding the effects of consumption of vegetarian diets on BP, both of which could prove useful in formulating dietary guidance, we performed a meta-analysis of studies that had examined associations between vegetarian diets and BP.
The electronic search strategy is shown in the Supplement (eTable 1). MEDLINE and Web of Science were searched for articles published in English from January 1, 1946, to November 7, 2013, and from January 1, 1900, to November 7, 2013, respectively, containing 1 or more of the keywords or phrases for vegetarian diets (plant-based diet or diet, vegetarian or vegetarian diets or vegetarianism or diets, vegan or vegan diets) and for blood pressure (blood pressure or hypertension). The reference lists of the retrieved articles were subsequently reviewed for identification of additional articles. If necessary, the relevant authors were contacted by the investigators to acquire missing information (Figure 1).
Two reviewers (Y.Y. and M.T.) independently scanned the retrieved abstracts to identify studies that met the following inclusion criteria: (1) use of a sample of participants older than 20 years; (2) an intervention or exposure consisting of a vegetarian diet, defined as a diet generally excluding or rarely including meats; these may include semivegetarian diets, defined as rarely including meat; vegan diets, defined as omitting all animal products; or vegetarian diets that include some animal products as indicated by the terms lacto (dairy products), ovo (eggs), or pesco (fish); (3) collection of sufficient data to allow calculation of mean differences in systolic/diastolic BP between individuals consuming a vegetarian diet and those consuming a referent or control diet; and (4) use of a controlled trial or observational study design. The exclusion criteria were (1) use of a sample consisting of twins; (2) use of multiple interventions (ie, use of lifestyle interventions in addition to dietary interventions); (3) reporting only categorical BP data; or (4) reliance on case series or case reports.
For each study, data regarding systolic and diastolic BP and variance measures; study methodology and sample size; baseline characteristics of the study population, including mean age, sex (proportion of men), BP, antihypertensive medication use, body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), alcohol intake, and dietary data (including type of diets examined and duration of their consumption); and outcomes, including adjustment factors used for each analytic model, BP measurements, and dietary measurements, were extracted. Mean values for baseline age, proportion of men, systolic and diastolic BP, BMI, and alcohol intake were calculated.
Mean differences in systolic and diastolic BP between groups consuming vegetarian or comparison diets were calculated. The pooled SE for the net difference in BP associated with the consumption of a vegetarian diet was obtained or, when not given, estimated using the method of Follmann et al,10 assuming a correlation of 0.50 between the baseline and final BP values (parallel design) or between the BP values during the intervention and control periods (crossover design). For studies comparing more than 1 exposure group or treatment arm, such as those comparing vegan and lacto-ovo-vegetarian groups, a pooled effect was calculated for each study using a random-effects model and then used to conduct the overall calculation.
Estimates of net change in BP associated with the consumption of vegetarian diets were combined using a random-effects model, which assigns a weight to each study on the basis of an individual study’s inverse variance. Overall estimates were derived for controlled trials and observational studies separately, using the study as the unit of analysis. Estimates of BP differences were reported within 95% CIs. Differences were considered significant at 2-sided P < .05.
Stratified analyses by mean age, sex, BMI, diet type, sample size, duration of vegetarian diet consumption, antihypertensive medication use, baseline hypertensive status, and location (country) were performed separately for controlled trials and observational studies. As a sensitivity analysis, we conducted a one-study-removed analysis to assess the effect of each study on the combined effect. Calculation of I2 and meta-regression was performed with subgroups, using the study as the unit of analysis to assess heterogeneity among studies.11
Funnel plots were developed and examined to identify publication bias, and the Egger test was performed to assess the relationship between sample size and effect size. The trim-and-fill method was used to adjust for publication bias. The trim-and-fill method determines where missing studies are likely to fall, adds them to the analysis, and then recomputes the combined effect. These analyses were conducted separately for controlled trials and observational studies. All analyses were performed using Comprehensive Meta-analysis, version 2, software (Biostat).
The search of the MEDLINE and Web of Science databases led to the retrieval of 258 studies. Of these, 7 clinical trials and 32 observational studies met the inclusion criteria (Figure 1).
Seven clinical trials were identified (Table 1).6- 8,12- 15 The 7 trials included a total of 311 participants (median sample size, 38; range, 11-113), with a mean age of 44.5 years (range, 38.0-54.3 years). All were open (nonmasked) controlled trials conducted for 6 or more weeks (mean, 15.7 weeks). Of these, 6 were randomized clinical trials.6- 8,13- 15 As shown in Table 1, several participants in 1 clinical trial13 used antihypertensive medication. All except 1 study15 provided foods to the participants. Vegan diets were examined in 2 trials,12,13 a lacto-vegetarian diet in 1,15 and lacto-ovo-vegetarian diets in 4.6- 8,14 Four studies12- 15 used parallel designs, and 3 trials6- 8 used crossover designs. All studies6- 8,12- 15 reported repeated BP measurements. Adjustments for potential confounders for each trial are shown in the Supplement (eTable 2).
Thirty-two observational studies were identified (Table 2).16- 47 These studies included 21 604 participants (median sample size, 152; range, 20-9242) with a mean age of 46.6 years (range, 28.8-68.4 years). Each of the 32 observational studies used cross-sectional designs. As shown in Table 2, several participants in 5 observational studies17,28,32,38,40 used antihypertensive medication. Because pooled effects were not reported, male and female subgroups (10 studies)24,25,32,33,35,36,41,44- 46 and racial subgroups (1 study)40 were included in the subgroup analyses (Table 2). In 22 of these studies, participants had been following vegetarian diets for more than 1 year.16- 20,22- 24,26- 31,33,37,39- 43,45 Five studies focused on vegan diets,23,27,31,39,41 2 on lacto-vegetarian diets,18,24 10 on lacto-ovo-vegetarian diets,16,19,20,22,26,29,37,43,44,47 and 15 on mixed diet types (vegan, lacto, lacto-ovo, pesco, and/or semivegetarian).17,21,25,28,30,32- 36,38,40,42,45,46 In 20 studies, diets were assessed by using questionnaires, typically food frequency questionnaires17,29,32,34,35,38,40 or 24-hour diet recalls.16,20,22,25,27,31,33,37,39,41,42,44,45 Interviews or self-report were used in 7 studies,18,19,24,28,36,43,47 and weighing methods were used for 1 study21; the means of dietary assessment were not reported in 4 studies.23,26,30,46 Of the 32 observational studies, 12 conducted repeated measurements of BP.16- 18,24,27,28,31,34,35,38,40,44 The adjusted factors in each study are shown in the Supplement (eTable 2).
In the clinical trials, consumption of vegetarian diets was associated with a mean reduction in systolic BP (−4.8 mm Hg; 95% CI, −6.6 to −3.1; P < .001; I2 = 0; P = .45 for heterogeneity) and diastolic BP (−2.2 mm Hg; 95% CI, −3.5 to −1.0; P < .001; I2 = 0; P = .43 for heterogeneity) compared with the consumption of omnivorous diets (Figure 2).
In the observational studies, consumption of vegetarian diets was associated with lower mean systolic BP (−6.9 mm Hg; 95% CI, −9.1 to −4.7; P < .001; I2 = 91.4; P < .001 for heterogeneity) and diastolic BP (−4.7 mm Hg; 95% CI, −6.3 to −3.1; P < .001; I2 = 92.6; P < .001 for heterogeneity) compared with consumption of omnivorous diets (Figure 3 and Figure 4).
In the meta-regression investigating the sources of heterogeneity in the observational trials, the potential sources were sex (proportion of men) (β coefficient, −0.03; P < .001), baseline systolic BP (−0.13; P = .003), baseline diastolic BP (−0.30; P < .001), sample size (0.001; P < .001), and BMI (−0.46; P = .02). These factors were not significant in the meta-regression of clinical trials (data not shown). These results suggest that the association between vegetarian diets and lower BP in adults is stronger among men and those with higher baseline BP and BMI. The association is also stronger in studies with smaller sample sizes.
Pooled changes in BP associated with consumption of vegetarian diets in planned strata are summarized in the Supplement (eTables 3 and 4). In the clinical trials, no heterogeneity was found in any subgroup and the estimated effect sizes were very similar.
For observational studies, subgrouping reduced heterogeneity in most cases, and vegetarian diets were associated with lower BP regardless of subgroup, although effect sizes were attenuated in some groups. Lower systolic BP values were reported in the predominantly (50%-99%) male subgroups compared with the 100% female subgroups. Lower systolic and diastolic BP were found in both BMI subgroups (<25 and ≥25), sample size subgroups (<100 and ≥100), and baseline BP subgroups (normal, prehypertension, and stage 1 hypertension).
In the one-study-removed analysis, results were largely unchanged, with BP differences between the vegetarian and comparison groups ranging from −5.3 to −3.5 mm Hg for systolic BP and −2.9 to −1.8 mm Hg for diastolic BP in clinical trials (all results, P < .05) and from −7.2 to −6.3 mm Hg for systolic BP and −5.0 to −4.3 mm Hg for diastolic BP for observational studies (all results, P < .001).
For clinical trials, visual examination of the funnel plot revealed that smaller trials that reported small reductions in systolic BP were possibly overrepresented (Figure 5A). In the absence of publication bias, study results would be symmetrically represented about the mean effect size; our findings suggest that a few studies were missing in the bottom left side. This visual impression was confirmed by the Egger test (P = .04). The results of use of the trim-and-fill method suggest that 3 trials might have been missing such that their addition would change the overall effect on systolic BP to −5.2 mm Hg (95% CI, −6.9 to −3.5).
For observational studies, visual examination of the funnel plot revealed that larger trials that reported generous reductions in systolic BP were possibly overrepresented. Our findings suggest that a few studies were missing in the middle right side (Figure 5B). This visual impression was confirmed by the Egger test (P < .001). The results of the trim-and-fill method suggest that no study was missing.
This meta-analysis of 7 controlled trials and 32 observational studies indicates that consumption of vegetarian diets is associated with lower BP compared with consumption of omnivorous diets. The meta-analysis indicates an overall difference in systolic BP of −4.8 mm Hg in controlled trials and −6.9 mm Hg in observational studies. For diastolic BP, the differences were −2.2 mm Hg in controlled trials and −4.7 mm Hg in observational studies. These effect sizes are similar to those observed with commonly recommended lifestyle modifications, such as adoption of a low-sodium diet48 or a weight reduction of 5 kg,49 and are approximately half the magnitude of those observed with pharmaceutical therapy, such as administration of angiotensin-converting enzyme inhibitors to individuals with hypertension.50 According to Whelton et al,51 a reduction in systolic BP of 5 mm Hg would be expected to result in a 7%, 9%, and 14% overall reduction in mortality due to all causes, coronary heart disease, and stroke, respectively.
The findings of the present study are consistent with those of a previous review of observational studies.5 They also accord with those of the Dietary Approaches to Stop Hypertension study,52,53 which was based on the observation that consumption of vegetarian diets was associated with a reduced risk of hypertension and found that a diet rich in vegetables and fruits, along with other dietary changes, reduced systolic BP and diastolic BP.
Specific diet and lifestyle factors are known to influence BP. Obesity, excessive sodium intake, and excessive alcohol use are associated with increased BP and risk of hypertension; potassium intake and physical activity are associated with lower BP.54,55 In addition, intake of unsaturated fat, protein, magnesium, and dietary fiber may be associated with differences in BP.5 The details provided in the studies included in the present review were insufficient to justify subgroup analyses that might have investigated the influence of these factors on the observed BP differences. Nonetheless, the following factors merit consideration as possible explanations for the observed associations. First, compared with omnivores, vegetarians typically have lower BMIs and a lower risk of obesity, which is mainly attributable to the lower energy density of the diet that results from higher fiber content and lower fat content.56 Weight differences do not fully explain the observed BP differences, however, because studies controlling for body weight have demonstrated a BP-lowering effect of vegetarian diets.6 Second, potassium is abundant in vegetarian diets.57 Meta-analyses58,59 of randomized clinical trials have reported that potassium supplementation decreases BP. It is hypothesized that a high potassium intake increases vasodilation and glomerular filtration rate while decreasing renin level, renal sodium reabsorption, reactive oxygen species production, and platelet aggregation.60 Third, some reports61 have suggested that vegetarian diets may be lower in sodium; however, others57 have shown no clear differences in sodium intake between nonvegetarians and vegetarians. Fourth, some studies32,36,37,41 have reported that alcohol consumption is lower in vegetarian populations compared with the general population. However, of the 7 clinical trials included in our study, 5 were limited to participants with no more than modest alcohol consumption; their results are unlikely to be substantially affected by alcohol intake. Vegetarian diets are often proportionately lower in saturated fatty acids and richer in polyunsaturated fatty acids compared with omnivorous diets; both of these dietary characteristics are associated with lower BP.5,62,63 Consumption of vegetarian diets has also been associated with reduced blood viscosity, which may affect BP.64 The consumption of vegetable protein has been shown to be inversely associated to BP.65
The present meta-analysis has several strengths. First, the available clinical trials and observational studies provided a reasonably large overall sample size that fosters confidence in the findings as well as permitting subgroup analyses in specific population groups. Second, its focus on dietary patterns rather than on the use of dietary supplements or artificial dietary manipulations makes the findings easily applicable to general or clinical populations.
This review also has several limitations. First, although no heterogeneity existed among the controlled trials, heterogeneity was high among the observational studies. Meta-regression and subgroup analyses showed that sex, baseline BP, sample size, and BMI may be key reasons for this heterogeneity. Nonetheless, lower BP was evident in all subgroups, although the differences were not significant for some subgroups. Second, this meta-analysis carried forward design limitations of the included studies. Most notable in this regard are small sample sizes and the fact that all observational studies used cross-sectional rather than prospective designs; however, the latter limitation is partially compensated for by the inclusion of several randomized clinical trials. Third, some of the observational studies did not adjust for lifestyle factors, such as alcohol intake or physical activity level. Finally, foods that make up vegetarian diets and the nutrient composition of the diets differ from person to person and from country to country. Further studies are needed to explore the relationships between specific foods and nutrients and BP. Nevertheless, the results of the meta-analysis of the controlled trials suggest a robust relationship between consumption of vegetarian diets and lower BP.
Consumption of vegetarian diets is associated with lower BP. Further studies are required to clarify which types of vegetarian diets are most strongly associated with lower BP. Research into the implementation of such diets, either as public health initiatives aiming at prevention of hypertension or in clinical settings, would also be of great potential value.
Accepted for Publication: December 8, 2013.
Corresponding Author: Yoko Yokoyama, PhD, MPH, Department of Preventive Medicine and Epidemiologic Informatics, National Cerebral and Cardiovascular Center, 5-7-1 Fujishirodai, Suita-city, Osaka 565-8565, Japan (firstname.lastname@example.org).
Published Online: February 24, 2014. doi:10.1001/jamainternmed.2013.14547.
Author Contributions: Dr Yokoyama had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Yokoyama, Nishimura, Barnard, Okamura, Miyamoto.
Acquisition of data: Yokoyama.
Analysis and interpretation of data: Yokoyama, Nishimura, Barnard, Takegami, Watanabe, Sekikawa.
Drafting of the manuscript: Yokoyama, Nishimura, Barnard, Miyamoto.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Yokoyama, Nishimura.
Obtained funding: Yokoyama.
Administrative, technical, or material support: Yokoyama, Takegami, Miyamoto.
Study supervision: All authors.
Conflict of Interest Disclosures: None reported.
Funding/Support: Financial support for this study was provided by a grant-in-aid for Japan Society for the Promotion of Science Fellows grant 23-10883.
Role of the Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Richard Holubkov, PhD, Division of Critical Care, Department of Pediatrics, University of Utah School of Medicine, provided statistical advice, and T. Colin Campbell, PhD, Nutritional Biochemistry, Division of Nutritional Sciences, Cornell University, provided conceptual advice. Dr Holubkov received compensation for his assistance.