Content uploaded by Francesco Sofi
Author content
All content in this area was uploaded by Francesco Sofi on Jun 13, 2017
Content may be subject to copyright.
Content uploaded by Francesco Sofi
Author content
All content in this area was uploaded by Francesco Sofi on Feb 19, 2016
Content may be subject to copyright.
Vegetarian, vegan diets and multiple health outcomes: A systematic review with
meta-analysis of observational studies
Monica Dinu
a
,
b
, Rosanna Abbate
a
, Gian Franco Gensini
a
,
c
, Alessandro Casini
a
,
b
, and Francesco Sofi
a
,
b
,
c
a
Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy;
b
Unit of Clinical Nutrition, University Hospital of Careggi,
Florence, Italy;
c
Don Carlo Gnocchi Foundation Italy, Onlus IRCCS, Florence, Italy
ABSTRACT
Background: Beneficial effects of vegetarian and vegan diets on health outcomes have been supposed in
previous studies. Objectives: Aim of this study was to clarify the association between vegetarian, vegan
diets, risk factors for chronic diseases, risk of all-cause mortality, incidence, and mortality from cardio-
cerebrovascular diseases, total cancer and specific type of cancer (colorectal, breast, prostate and lung),
through meta-analysis. Methods: A comprehensive search of Medline, EMBASE, Scopus, The Cochrane
Library, and Google Scholar was conducted. Results: Eighty-six cross-sectional and 10 cohort prospective
studies were included. The overall analysis among cross-sectional studies reported significant reduced
levels of body mass index, total cholesterol, LDL-cholesterol, and glucose levels in vegetarians and vegans
versus omnivores. With regard to prospective cohort studies, the analysis showed a significant reduced
risk of incidence and/or mortality from ischemic heart disease (RR 0.75; 95% CI, 0.68 to 0.82) and incidence
of total cancer (RR 0.92; 95% CI 0.87 to 0.98) but not of total cardiovascular and cerebrovascular diseases,
all-cause mortality and mortality from cancer. No significant association was evidenced when specific
types of cancer were analyzed. The analysis conducted among vegans reported significant association
with the risk of incidence from total cancer (RR 0.85; 95% CI, 0.75 to 0.95), despite obtained only in a
limited number of studies. Conclusions: This comprehensive meta-analysis reports a significant protective
effect of a vegetarian diet versus the incidence and/or mortality from ischemic heart disease (¡25%) and
incidence from total cancer (¡8%). Vegan diet conferred a significant reduced risk (¡15%) of incidence
from total cancer.
KEYWORDS
Vegetarian; vegan; diet;
meta-analysis
Introduction
Vegetarian diet, defined as a dietary profile characterized by
abstention from consuming meat and meat products, poultry,
seafood and flesh from any other animal, is experiencing a con-
siderable popularity in the general population (Leitzmann,
2014). The reasons for adoption of this dietary profile are dif-
ferent, ranging from ethical motivations, religious beliefs, envi-
ronmental and cultural issues, to health-related aspects (Craig
and Mangels, 2009; Leitzmann, 2014). Health benefits of vege-
tarian diet have been widely reported by cross-sectional and
prospective cohort studies during the last 50 years, but uncer-
tainties due to the limited sample sizes of some of these studies
and to the fact that some large prospective cohort studies
included particular cohort of subjects still remain (Fraser, 1999,
2009). Indeed, generally speaking, vegetarians tend to be more
conscious for the health aspects, slimmer, and in better health
when compared with omnivores, and specific cohorts have
been demonstrated to be not generalizable to the general popu-
lation for the low prevalence of risk factors (Kwok et al., 2014).
These findings might indicate the presence of flaws in the anal-
ysis of possible health benefits of vegetarian diet. To date, vegan
diet, i.e., the total exclusion of any animal-derived substance is
a pattern that is attracting a relevant interest among the general
population. Few studies reported that vegan diet appears to be
healthful, but no conclusive data have been obtained (Craig,
2009; Le and Sabat!
e, 2014). The aim of this study was to con-
duct a comprehensive systematic review with meta-analysis of
all cross-sectional and cohort studies hitherto published in
order to obtain an estimate of the association between vegetar-
ian, vegan diets, and multiple health outcomes, including risk
factors for chronic diseases, risk of all-cause mortality, inci-
dence and mortality from cardio-cerebrovascular diseases, total
cancer and specific types of cancer.
Methods
Search strategy, inclusion criteria and data extraction
The review question was structured using the following ele-
ments—Population of interest (P); Intervention (I); Compari-
sons (C); Outcome (O); and Time frame (T)—namely, the
PICOT format (Whittemore and Knafl,2005). For this study,
CONTACT Monica Dinu monicarodica.dinu@unifi.it Department of Experimental and Clinical Medicine, University of Florence, Largo Brambilla 3, 50134
Florence, Italy.
Color versions of one or more of the figures in this article can be found online at www.tandfonline.com/bfsn.
© 2017 Taylor & Francis Group, LLC
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION
2017, VOL. 57, NO. 17, 3640–3649
https://doi.org/10.1080/10408398.2016.1138447
Setting (S) was also included. The operationalisation of these
elements is displayed in Table 1.
According to the PRISMA (Preferred Reporting Items for
Systematic Reviews and Meta-Analyses) statement (Moher
et al., 2009) we systematically identified all potentially relevant
articles through a computerized search of main electronic data-
bases: Medline (1950 through April 2015), Embase (1980
through April 2015), Scopus (through April 2015), The
Cochrane Library, and Google Scholar. Additional searches
were conducted by scanning references of the identified articles,
reviews and meta-analyses. Search terms included the following
key words, used in combination as MeSH terms and text words:
“vegetarian,”“vegetarians,”“vegetarianism,”“vegetarian diet,”
“vegetarian diets,”“vegan,”“vegans,”“veganism,”“vegan diet,”
“vegan diets,”and their variants, which were used in combina-
tion with words relating to health status “plasma lipids,”“cho-
lesterol,”“triglycerides,”“glycemia,”“hematic parameters,”
“cancer,”“circulatory diseases,”“cardiovascular disease,”
“ischemic heart disease,”“cerebrovascular disease,”“mortality,”
“health effects,”“health status,”and their variants. The search
was limited to human studies. When multiple articles for a sin-
gle study were present, we used the latest publication and sup-
plemented it, if necessary, with data from the most complete or
updated publication.
Eligible studies included any observational study conducted
in humans (i.e., cross-sectional studies, case-control, nested
case-control, or case-cohort design) that reported a measure of
association (such as hazard ratios or incident rate ratios for
prospective studies) between vegetarian or vegan diet, assessed
by questionnaires, and risk factors for chronic degenerative dis-
eases [body mass index (BMI), total cholesterol, LDL-
cholesterol, HDL-cholesterol, triglycerides, blood glucose], risk
of all-cause mortality, incidence and mortality from cardio-
cerebrovascular diseases, total cancer and specific types of can-
cer, confirmed by medical records or registry linkage.
The decision to include studies was hierarchical and initially
made on the basis of the study title, then of the study abstract,
and finally of the complete study manuscript. Eligible studies
were included if they met the inclusion criteria for study design,
study population (clinically healthy subjects !18 years old),
exposure (vegetarian diet, defined as a diet excluding meat and
meat products, poultry, seafood and flesh from any animal;
vegan diet, defined as a diet that omit all the animal-derived
products), reference group (omnivore diet, defined as a diet
consuming all types of foods including meat and meat-prod-
ucts, poultry, seafood and flesh from any animal), outcome and
statistics (sufficient data to allow calculation of differences
between individuals consuming a vegetarian or a vegan diet
and those consuming an omnivore diet).
Two reviewers (M.D., F.S.) independently extracted data
from all the studies fulfilling the inclusion criteria and any dis-
agreement was resolved by consensus. The following data were
extracted from the original articles by using a standardized data
extraction form: lead author, year of publication, country of the
study population, study design characteristics, characteristics of
different groups, follow-up duration, outcomes, effect size
measurements (i.e., hazard ratio/relative risk, mean difference)
and variables that entered into the multivariable model as
potential confounders.
Assessment of methodological quality
Two reviewers (M.D., F.S.) assessed the methodological quality
independently, and any incongruity was discussed and
resolved. The methodological quality of the trials included was
assessed using elements of the Newcastle-Ottawa Scale (NOS)
for assessing risk of bias in observational studies. A total of 9
points coming from 3 domains were reported for each study:
selection, comparability, and ascertainment of exposure(s) or
outcome(s) (Higgins and Green, xxxx).
Statistical analysis
We used Review Manager (RevMan, version 5.3 for Windows;
The Cochrane Collaboration, Copenhagen, Denmark) to pool
data for each risk factor and outcome of interest. We conducted
pooled analyses using the generic inverse variance method with
random-effects weighting. As for cross-sectional studies, we
calculated the weighted mean differences (WMD) between the
subjects following vegetarian or vegan diet and those following
an omnivore diet with 95% confidence intervals (CIs). With
regard to prospective studies, pooled results were reported as
relative risks (RRs) and presented with 95% CIs. P<0.05 was
considered statistically significant. When available, we used the
results of the original studies from multivariate models with
the most complete adjustment for potential confounders.
Statistical heterogeneity was evaluated by using the I
2
statis-
tic, which assessed the appropriateness of pooling the individ-
ual study results. The I
2
value provided an estimate of the
amount of variance across studies because of heterogeneity
rather than chance. Where I
2
was >50%, the heterogeneity
was considered substantial. Moreover, to further investigate the
heterogeneity across the studies we performed sensitivity analy-
ses by dividing studies into groups according to their main
characteristics. Subgroup analyses were then performed accord-
ing to cohorts (Adventists; Non-Adventists), mean sample size
of the study populations (<9,500; !9,500), country (U.S.; non-
U.S.), mean duration of follow-up (<14 years; !14 years), and
quality of the studies (Moderate Dstudies with scores ranging
from 4 to 6 on the NOS, High Dstudies with scores !7 on the
Table 1. Use of the PICOTS format, as applied to this study.
PICOTS format Description
Population Presumably healthy subjects !18 years old
Intervention Vegetarian diet, defined as a diet excluding meat and
meat products, poultry, seafood and flesh from any
animal
Vegan diet, defined as a diet that omit all the animal-
derived products
Comparisons Omnivore diet, defined as a diet consuming all types of
foods including meat and meat-products, poultry,
seafood and flesh from any animal
Outcome Risk factors for chronic degenerative diseases (body mass
index, total cholesterol, LDL-cholesterol, HDL-
cholesterol, triglycerides, blood glucose); all-cause
mortality; incidence and mortality from cardio-
cerebrovascular diseases, total cancer and specific
type of cancer (colorectal, breast, prostate, lung)
Time Not applicable
Setting Institutional and community setting
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3641
NOS). We removed each single study from the meta-analyses
and recalculated the summary association (the “leave one out”
approach) (Greenhouse and Iyengar, 2009). A study whose
removal either pushed the significance level of the overall asso-
ciation from <0.05 to !0.05 (or vice versa), or altered the nom-
inal effect size by 10% or more, was considered an influential
outlier. If !5 studies were available, we explored the possibility
of publication bias by visual inspection of funnel plot of effect
size against standard error.
Results
Literature search
Our search yielded a total of 10,516 unique citations. After
review and excluding duplicate reports we identified 433 cita-
tions as potentially relevant for the analysis. Of these, 325 were
excluded after full-text reviews for the reasons described in
Figure 1. Overall, a total of 108 articles were finally included in
the meta-analysis (Figure 1).
Selected cross-sectional studies examined the effect of vege-
tarian diet (nD86) and vegan diet (nD24) on the following
risk factors for chronic degenerative diseases: BMI (71 studies
for vegetarian diet; 19 studies for vegan diet), total cholesterol
(64 studies for vegetarians, 19 for vegans), LDL-cholesterol
(46 for vegetarians, 13 for vegans), HDL-cholesterol (51 for
vegetarians, 15 for vegans), triglycerides (55 for vegetarians, 13
for vegans), and blood glucose (27 for vegetarians, 4 for
vegans). Selected cohort prospective studies examined the asso-
ciation between vegetarian diet (nD10), vegan diet (nD2)
and different clinical outcomes: all-cause mortality (5 studies
for vegetarians, 2 studies for vegans), incidence and mortality
from cardiovascular diseases (4 studies, all for vegetarians),
ischemic heart disease (5 studies, all for vegetarians), cerebro-
vascular disease (3 studies, all for vegetarians), incidence of
total cancer (2 studies for vegetarians; 2 studies for vegans) and
mortality from total cancer (3 studies for vegetarians). In
addition, some of these prospective studies reported the associ-
ation with a specific localization of cancer and the vegetarian
diet, such as incidence of breast cancer (2 studies), mortality
from breast cancer (2 studies), prostate cancer (2 studies), colo-
rectal cancer (3 studies), and lung cancer (2 studies).
Cross-sectional studies
Characteristics of the included cross-sectional studies
reporting the effect of vegetarian and vegan diets on differ-
ent risk parameters, including the risk-of-bias assessment,
are reported in Supplementary tables 1–12. The overall
analysis comprised a total number of 56,461 vegetarians
and 8,421 vegans compared with 184,167 omnivorous and
the mean age varied widely, ranging from 18 to 81 years
old. The risk-of-bias assessment for each cross-sectional
study included in the meta-analysis reported a low risk of
bias only in 2 studies, whereas in the others a moderate-to-
high risk was present.
Tables 2 and 3 show the pooled estimates of effect size
and 95% CIs expressed as WMD for the effects of vegetar-
ian and vegan diets vs. omnivore diet on risk factors for
chronic degenerative diseases. At the overall analysis, vege-
tarian diet was significantly associated with lower BMI
(¡1.49), serum total cholesterol (¡28.16 mg/dL), LDL-cho-
lesterol (¡21.27 mg/dL), HDL-cholesterol (¡2.72 mg/dL),
serum triglycerides (¡11.39 mg/dL), and blood glucose lev-
els (¡5.08 mg/dL) with respect to omnivores. Similarly,
vegan diet reported significantly lower BMI (¡1.72), serum
total cholesterol (¡31.02 mg/dL), LDL-cholesterol
(¡22.87 mg/dL), and blood glucose levels (¡6.38 mg/dL),
but nonsignificant lower HDL-cholesterol and triglycerides
with respect to omnivores. Significant heterogeneity
(P<0.001) was present among the studies.
Prospective cohort studies
The characteristics of the included prospective cohort studies,
including risk-of-bias assessment, are presented in Tables 4,5.
The overall analysis for all the different clinical outcomes com-
prised a total number of 72,298 vegetarians followed for a
period ranging from 4.1 to 21 years. One study included only
women and 8 studies included men and women. The risk-
of-bias assessment for the included study reported a low risk of
bias in 4 studies (Key et al., 2009, Cade et al., 2010, Crowe
et al., 2013, Key et al., 2014) and a moderate risk for the
remaining.
The results of the pooled analysis for the all included studies
are depicted in Figure 2. The relation between vegetarian diet
and all-cause mortality was evaluated in 5 studies (Key et al.,
1999,2009; Appleby et al., 2002; Chang-Claude et al., 2005;
Orlich et al., 2013) including 66,018 vegetarians and 8,216
deaths by obtaining a nonsignificant (PD0.24) association
with an RR of 0.94 (95% CI 0.86 to 1.04) and a significant het-
erogeneity (I
2
D83%; P<0.001). After exclusion of studies by
Key et al. (2009), which included the cohorts of the Adventist
Mortality and Health-1 studies the heterogeneity disappeared
(I
2
D21%; PD0.28) and the result did not change (RR 1.01,
95% CI 0.95 to 1.07). For vegans and all-cause mortality the
Figure 1. PRISMA flow diagram for search strategy.
3642 M. DINU ET AL.
risk ratio was 0.88 (RR 0.88, 95% CI 0.75 to 1.02; PD0.42).
Similarly, no significant association (PD0.07) was also found
among vegetarians when incidence and/or mortality from car-
diovascular diseases were taken as a unique outcome (RR 0.93,
95% CI 0.86 to 1.00). However, as incidence and/or mortality
from ischemic heart disease were analyzed separately, vegetar-
ian diet was found to be significantly (p <0.001) associated
with the outcome, with a reduced risk of ¡25% (RR 0.75, 95%
CI 0.68 to 0.82), and nonsignificant heterogeneity (I
2
D35%;
PD0.16), while nonsignificant (PD0.39) association for inci-
dence and/or mortality from cerebrovascular disease (RR 0.93,
95% CI 0.78 to 1.10) was observed.
With regard to incidence of total cancer, meta-analytic pool-
ing under a random-effects model showed significant
(PD0.002) lower risk of cancer among vegetarians (RR 0.92,
95% CI 0.87 to 0.98) and vegans (RR 0.85 95% CI 0.75 to 0.95)
with a nonsignificant heterogeneity among the studies
(I
2
D0%; PD0.35, PD0.71 for vegetarians and vegans, respec-
tively). Finally, by analyzing different localizations of cancer,
nonsignificant reduced risk of incidence of breast cancer (RR
0.94, 95% CI 0.84 to 1.06), as well as mortality from colorectal
(RR 0.90, 95% CI 0.76 to 1.05), breast (RR 0.94, 95% CI 0.56 to
1.58), prostate (RR 0.90, 95% CI 0.63 to 1.29) and lung
(RR 0.86, 95% CI 0.62 to 1.19) cancer was reported when
vegetarians were compared to omnivores.
Sensitivity analysis and publication bias
In order to investigate the possible differences across the studies
we performed some sensitivity analyses by grouping studies
according to some characteristics such as cohorts (Adventists/
Non-Adventists), size of the studies (mean size of the study
samples: 9,500), country (U.S./Non-U.S.), length of follow-up
(mean duration: 14 years), and study quality (Moderate Dstud-
ies with scores ranging from 4 to 6 on the NOS, High Dstudies
with scores !7 on the NOS). As for all-cause mortality and
breast cancer mortality, vegetarian diet demonstrated a
significant association only among studies conducted in the
U.S. Adventist cohorts, with a shorter duration of follow-up
whereas studies conducted among non-Adventists cohorts liv-
ing in European countries did not report any significant associ-
ation with the outcome. With regard to ischemic heart disease,
sensitivity subgroups did not change the significant association
reported in the overall analysis (Table 6). Publication bias was
assessed by both funnel plot and Egger’s linear regression test.
Both methods demonstrated no evidence of publication bias. In
Figure 3, funnel plot for the outcome of ischemic heart disease
is reported.
Discussion
The present is the first systematic review with meta-analysis
that encompasses all the available observational studies esti-
mating the association between vegetarian and vegan diets and
multiple health outcomes including risk factors for chronic
diseases, as well as incidence and mortality from cardio-
cerebrovascular and neoplastic diseases. The overall analysis
comprised a large amount of studies (98 cross-sectional studies
and 10 cohort prospective studies) for a total population of
over than 130,000 vegetarians and 15,000 vegans.
The results of the present meta-analysis report that vegeta-
rians and vegans show significantly lower levels of the most rel-
evant risk factor for chronic disease such as BMI, lipid
variables and fasting glucose, when compared to nonvegetar-
ians and nonvegans. These findings, however, are significantly
affected by the nature of the cross-sectional studies, which are
highly susceptible to biases, as otherwise observed by the mod-
erate-to-high risk of bias assessment in each included study.
Nevertheless, as cohort prospective studies are taken into
account, significant results in terms of reduction for risk of inci-
dence of ischemic heart disease (¡25%) and incidence of total
cancer (¡8%) were observed for vegetarians. Similarly,
although in a very limited number of studies, vegan diet
Table 3. Effects of vegan diet on risk factors for chronic degenerative diseases in case-control studies, expressed as weighted mean difference (WMD).
Outcome N studies Vegans (n) Omnivores (n) WMD 95% CI Pvalue
BMI (kg/m
2
) 19 8 376 123 292 ¡1.72 ¡2.21 to ¡1.22 <0.0001
Total cholesterol (mg/dL) 19 1 272 12 213 ¡31.02 ¡34.82 to ¡27.21 <0.0001
LDL–cholesterol (mg/dL) 13 728 11 670 ¡22.87 ¡29.92 to ¡15.82 <0.0001
HDL-cholesterol (mg/dL) 15 1 175 12 114 ¡1.54 ¡2.96 to ¡0.12 0.61
Triglycerides (mg/dL) 13 483 10 110 ¡9.35 ¡20.28 to 1.57 0.09
Blood glucose (mg/dL) 4 83 125 ¡6.38 ¡12.35 to ¡0.41 0.04
BMI: body mass index; To convert mmol/L cholesterol to mg/dL, we multiplied mmol/L by 38.67. To convert mmol/L triglyceride to mg/dL, we multiplied mmol/L by
88.57. To convert mmol/L blood glucose to mg/dL, we multiplied mmol/L by 18.
Table 2. Effects of vegetarian diet on risk factors for chronic degenerative diseases in case-control studies, expressed as weighted mean difference (WMD).
Outcome Nstudies Vegetarians (n) Omnivores (n) WMD 95% CI Pvalue
BMI (kg/m
2
) 71 57 724 199 230 ¡1.49 ¡1.72 to ¡1.25 <0.0001
Total cholesterol (mg/dL) 64 5 561 23 573 ¡28.16 ¡31.22 to ¡25.10 <0.0001
LDL-cholesterol (mg/dL) 46 5 583 22 934 ¡21.27 ¡24.27 to ¡18.27 <0.0001
HDL–cholesterol (mg/dL) 51 6 194 23 660 ¡2.72 ¡3.40 to ¡2.04 <0.0001
Triglycerides (mg/dL) 55 4 008 22 083 ¡11.39 ¡17.42 to ¡5.37 0.02
Blood glucose (mg/dL) 27 2 256 2 192 ¡5.08 ¡5.98 to ¡4.19 <0.0001
BMI: body mass index; To convert mmol/L cholesterol to mg/dL, we multiplied mmol/L by 38.67. To convert mmol/L triglyceride to mg/dL, we multiplied mmol/L by
88.57. To convert mmol/L blood glucose to mg/dL, we multiplied mmol/L by 18.
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3643
Table 4. Characteristics of prospective cohort studies evaluating vegetarian diet and different clinical outcomes.
Author (year) Country Cohort Sex Age (year) F-up, y n/NOutcome RR (95% IC) Adjustment Risk of bias
Key et al. (1999) U.S. Adventist mortality
study
M/F 52.5 5.6 1635/10258 All-cause mortality 0.83 (0.76 to 0.92) Age, sex, smoking status Moderate
598/10258 IHD 0.74 (0.63 to 0.88)
182/10258 Cerebrovascular disease 0.65 (0.48 to 0.87)
41/10258 Colorectal cancer mortality 1.37 (0.73 to 2.56)
6/10258 Lung cancer mortality 0.59 (0.10 to 3.28)
26/10258 Breast cancer mortality 0.65 (0.28 to 1.52)
15/10258 Prostate cancer mortality 1.41 (0.49 to 4.04)
Key et al. (1999) U.S. Adventist Health
Study-1
M/F 52.5 11.1 3564/8003 All-cause mortality 0.80 (0.74 to 0.87) Age, sex, smoking status Moderate
921/8003 IHD 0.62 (0.53 to 0.73)
317/8003 Cerebrovascular disease 0.93 (0.73 to 1.19)
104/8003 Colorectal cancer mortality 1.01 (0.66 to 1.56)
96/8003 Lung cancer mortality 0.69 (0.37 to 1.27)
64/8003 Breast cancer mortality 0.52 (0.27 to 0.97)
66/8003 Prostate cancer mortality 0.79 (0.44 to 1.41)
Key et al. (1999)GermanyHeidelbergStudyM/F46.59.931/1083Cerebrovasculardisease1.69(0.69to4.15)Age,sex,smokingstatus Moderate
5/1083 Colorectal cancer mortality 0.35 (0.06 to 2.11)
5/1083 Breast cancer mortality 1.09 (0.18 to 6.67)
3/1083 Prostate cancer mortality 1.67 (0.14 to 19.6)
Appleby et al. (2002) U.K. Oxford Vegetarian
Study
M/F 42.3 17.6 532/4674 All-cause mortality 1.01 (0.89 to 1.14) Age, sex, smoking status Moderate
214/4674 Cardiovascular diseases 0.93 (0.77 to 1.12)
109/4674 IHD 0.86 (0.67 to 1.12)
63/4674 Cerebrovascular disease 1.08 (0.75 to 1.54)
156/4674 Cancer mortality 0.89 (0.72 to 1.10)
25/4674 Colorectal cancer mortality 1.20 (0.68 to 2.13)
16/4674 Lung cancer mortality 0.82 (0.44 to 1.56)
22/4674 Breast cancer mortality 1.02 (0.57 to 1.84)
8/4674 Prostate cancer mortality 0.50 (0.22 to 1.17)
Appleby et al. (2002) U.K. Health Food
Shoppers Study
M/F 42.8 18.7 963/4600 All-cause mortality 1.03 (0.95 to 1.13) Age, sex, smoking status Moderate
445/4600 Cardiovascular diseases 0.95 (0.84 to 1.07)
256/4600 IHD 0.85 (0.71 to 1.01)
141/4600 Cerebrovascular disease 0.99 (0.79 to 1.24)
203/4600 Cancer mortality 1.12 (0.95 to 1.32)
33/4600 Colorectal cancer mortality 0.79 (0.51 to 1.22)
24/4600 Lung cancer mortality 1.05 (0.64 to 1.72)
41/4600 Breast cancer mortality 1.73 (1.11 to 2.69)
16/4600 Prostate cancer mortality 1.24 (0.64 to 2.41)
Chang-Claude
et al. (2005)
Germany Heidelberg Study M/F 50 21 322/1225 All-cause mortality 1.10 (0.89 to 1.36) Age, gender, smoking, level of
activity, alcohol
consumption, education
level, BMI.
Moderate
145/1225 Cardiovascular diseases 0.83 (0.62 to 1.12)
41/1225 IHD 0.70 (0.41 to 1.18)
76/1225 Cancer mortality 1.04 (0.86 to 1.34)
Key et al. (2009) U.K. EPIC Oxford Study M/F 37 17 385/16081 All-cause mortality 1.05 (0.93 to 1.19) Age, sex, smoking status, alcohol
consumption
Low
118/16081 Cardiovascular diseases 0.97 (0.78 to 1.21)
46/16081 Cerebrovascular disease 1.10 (0.77 to 1.58)
3644 M. DINU ET AL.
Cade et al. (2010) U.K. United Kingdom
Women’s Cohort
Study
F 49 9 130/6491 Breast cancer incidence 0.88 (0.69 to 1.11) Age, energy intake, menopausal
status, calorie adjusted fat,
BMI, physical activity, OCP
use, HRT use, smoking status,
parity, age at menarche,
ethanol, educational level,
total days breast feeding,
socioeconomic class
Low
Tantamango-Bartley et al.
(2012)
U.S. Adventist Health
Study - 2
M/F 57.5 4.1 878/19735 Cancer incidence 0.95 (0.86 to 1.04) Race, family history of cancer,
BMI, education, smoking,
alcohol, age at menarche,
pregnancies, OCP use,
breastfeeding, menopause
status, HRT.
Moderate
Crowe et al. (2013) U.S. EPIC Oxford study M/F 40.1 11.6 NS/15120 IHD 0.72 (0.61 to 0.85) Age, smoking status, alcohol,
physical activity, educational
level, Townsend Deprivation
Index, OCP use or HRT use
for menopause in woman,
BMI
Low
Orlich et al. (2013) U.S. Adventist Health
Study - 2
M/F 57.5 5.9 815/21177 All-cause mortality 0.91 (0.82 to 1.00) Age, race, smoking status,
exercise, personal income,
educational level, marital
status, alcohol, region, sleep
Moderate
NS/21177 Cardiovascular diseases 0.90 (0.76 to 1.06)
NS/21177 IHD 0.82 (0.62 to 1.06)
NS/21177 Cancer mortality 0.90 (0.75 to 1.09)
Key et al. (2014) U.K. Oxford Vegetarian
Study CEPIC
Oxford Study
M/F 40 14.9 1098/18298 Cancer incidence 0.90 (0.84 to 0.97) BMI, smoking status, alcohol
consumption, physical
activity level, for the women
only-cancers, parity, OCP
Low
325/18298 Breast cancer incidence 0.96 (0.84 to 1.10)
Orlich et al. (2015) U.S. Adventist Health
Study - 2
M/F 58.3 7.3 147/22424 Colorectal cancer mortality 0.83 (0.66 to 1.05) Age, race, sex, BMI, educational
level, alcohol use, moderate
or vigorous exercise, HRT,
history of peptic ulcer, family
history of colorectal cancer,
dietary energy, history of
inflammatory bowel disease,
treatment for diabetes
mellitus within the past year,
used aspirin at least weekly
at least 2 of the past 5 years,
used statins at least 2 of the
past 5 years, supplemental
calcium use, supplemental
vitamin D, fiber intake, prior
colonoscopy or flexible
sigmoidoscopy
Moderate
IHD: Ischemic heart disease; BMI: body mass index; OCP: oral contraceptive pills; HRT: hormone replacement therapy; NS: not specified.
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3645
Table 5. Characteristics of prospective cohort studies evaluating vegan diet and different clinical outcomes.
Author, y Country Cohort Sex Age, y F-up, y n/NOutcomes RR (95% IC) Adjustment Risk of bias
Key et al., (1999) U.S. AMS CAHS-1 CHEIDELB
COXF CHFSS
M/F 52.5 10.6 68/753 All-cause mortality 1.00 (0.70 to 1.44) Age, sex, smoking status Moderate
Tantamango-Bartley
et al., (2012)
U.S. Adventist Health Study - 2 M/F 57.5 4.1 190/4922 Cancer incidence 0.86 (0.73 to 1.00) Race, family history of cancer,
BMI, education, smoking,
alcohol, age at menarche,
pregnancies, breastfeeding,
OCP use, menopause status,
HRT.
Moderate
Orlich et al., (2013) U.S. Adventist Health Study - 2 M/F 57.5 5.9 197/5548 All-cause mortality 0.85 (0.73 to 1.01) Age, sex, race, smoking,
exercise, personal income,
educational level, marital
status, alcohol, region, sleep,
menopause, hormone
therapy.
Moderate
Key et al., (2014) U.K. Oxford Vegetarian Study C
EPIC Oxford Study
M/F 37.8 14.9 105/2246 Cancer incidence 0.82 (0.68 to 1.00) BMI, smoking status, OCP,
alcohol consumption, parity,
physical activity level, for the
women only-cancers.
Low
BMI: body mass index; OCP: oral contraceptive pills; NS: not specified.
AHS-1: Adventist Health Study-1; AMS: Adventist Mortality Study; OXF: Oxford Vegetarian Study; HFSS: Health Food Shoppers Study; EPIC: EPIC Oxford Study.
3646 M. DINU ET AL.
showed a significant association with a reduced risk of total
cancer incidence (¡15%).
In the last years, the number of subjects who began to adopt
a vegetarian and/or vegan dietary pattern has increased with
respect to the past, when the population of vegetarians was lim-
ited only to few and selected cohorts (Craig and Mangels, 2009;
Leitzmann, 2014). Accordingly, the healthy aspect of these die-
tary profiles has gained interest in both medical and lay com-
munities, but some uncertainties in the literature still remain
(Key et al., 2014). Historically, the hypothesis that vegetarian
diet is able to determine a reduced risk of occurrence of disease
and mortality was linked to data whose strength of evidence
was limited, suffering from some drawbacks. In fact, cross-sec-
tional studies suffer from a high risk of bias and scarce quality
and were, in most of the cases, old and conducted in a limited
number of subjects; moreover, the low prevalence of some rele-
vant cardiovascular risk factors (e.g., smoking habit, hyperten-
sion, high body mass index, among all) in certain cohorts raise
some concerns about the generalizability of these results in gen-
eral population (Kwok et al., 2014).
We aimed to conduct the present systematic review with
meta-analysis in order to give an insight into the intricate liter-
ature on this issue. Other systematic reviews with meta-analysis
that analyzed the possible association between vegetarian pat-
tern and clinical outcomes have been published so far (Ashen,
2013; Kwok et al., 2014). This paper, however, is the first that
conducted a comprehensive analysis of the literature in differ-
ent types of analytical studies (cross-sectional and cohort pro-
spective), in different outcomes, and the first that included also
vegan diet.
With regard to the analysis of cross-sectional studies we
found, in a total population of more than 56,000 subjects con-
suming a plant-based dietary pattern, significantly lower levels
of body mass index, total cholesterol, LDL-cholesterol, trigly-
cerides, and blood glucose when vegetarians were compared
with nonvegetarians, and body mass index, total cholesterol
Figure 2. Forest plot summary of all-cause mortality, incidence and mortality from cardio-cerebrovascular diseases, total cancer, and specific type of cancer. Pvalue is for
Z test of no overall association between exposure and outcome; Phet is for test of no differences in association measure among studies; I
2
estimates from heterogeneity
rather than sampling error.
Table 6. Subgroup analyses.
nAll-cause mortality nIschemic heart disease nBreast cancer mortality
Cohort
Adventists 3 0.84 (0.78 to 0.90) 3 0.70 (0.60 to 0.82) 2 0.57 (0.34 to 0.95)
Non-Adventists 4 1.04 (0.98 to 1.10) 4 0.79 (0.71 to 0.88) 3 1.40 (0.98 to 2.01)
Duration of follow-up
<14 years 3 0.84 (0.78 to 0.90) 4 0.84 (0.73 to 0.96) 3 0.59 (0.36 to 0.98)
!14 years 4 1.04 (0.98 to 1.10) 3 0.70 (0.63 to 0.78) 2 1.38 (0.82 to 2.30)
Country
U.S. 3 0.84 (0.78 to 0.90) 4 0.84 (0.73 to 0.96) 2 0.57 (0.34 to 0.95)
Non-U.S. 4 1.04 (0.98 to 1.10) 3 0.70 (0.63 to 0.78) 3 1.40 (0.98 to 2.01)
Sample size
<9,500 subjects 4 0.97 (0.83 to 1.13) 4 0.75 (0.62 to 0.92) 4 1.02 (0.56 to 1.86)
!9,500 subjects 3 0.92 (0.81 to 1.05) 3 0.74 (0.67 to 0.83) 1 0.65 (0.28 to 1.51)
Study quality
Medium (4–6 points, NOS) 6 0.93 (0.84 to 1.03) 6 0.74 (0.68 to 0.81) 5 0.94 (0.56 to 1.58)
High (7–9 points, NOS) 1 1.05 (0.93 to 1.19) 1 0.72 (0.61 to 0.85) 0 —
NOS: Newcastle-Ottawa Scale.
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3647
and LDL-cholesterol when vegans were compared to nonve-
gans. Actually, the reasons for the beneficial effects of vegetari-
anism and veganism on total and LDL-cholesterol are different
and lie mainly on the lower intake of total and saturated fats,
but reasons can also reside on the large consumption of foods
known to decrease these parameters, such as soybean, legumes,
nuts and vegetable oils. Similarly, the lower body mass index
found in the vegetarian/vegan people was not surprising as this
is in total agreement with the literature, being linked to the
lower intake of energy usually reported by these populations.
However, data obtained from cross-sectional studies need to be
interpreted with caution because of the moderate-to-high risk
of bias reported in the vast majority of these studies, and also
because of the high degree of heterogeneity evidenced in our
overall analysis.
The overall analysis among prospective cohort studies docu-
mented a 25%-reduction of incidence and/or mortality from
ischemic heart disease (Ashen, 2013) but not of incidence and/
or mortality from total cardiovascular and cerebrovascular dis-
eases, and an 8%-reduction of incidence of total cancer but not
of mortality from cancer when vegetarians were compared to
nonvegetarians. These results, although partly surprising, could
be explained by the fact that incidence and mortality are 2 very
different outcomes, with cardiovascular and cancer mortality
being greatly influenced by the treatment approaches. More-
over, the overall analysis in the cohort studies reported no sig-
nificant association with specific localizations of cancer disease,
such as incidence and mortality from breast cancer, as well as
incidence of lung, colon-rectum and stomach cancer. This fact
can be explained by the low statistical power, due to a low num-
ber of studies evaluating this aspect and a low sample size.
The present study has some strengths and limitations. This is
the first systematic review and meta-analysis that analyzed all
the available data evaluating both vegetarian and vegan diets
from different types of studies (cross-sectional, cohort) in rela-
tion to different health parameters and outcomes. Such a large
number of studies included allowed us to perform sensitivity
analyses by grouping studies with similar characteristics. By ana-
lyzing studies according to some specificcharacteristics,wehave
noted a difference of association for vegetarians and all-cause
mortality according to cohort (Adventists; Non-Adventists),
duration of follow-up (<14 years; >14 years) and country of
origin of the cohort (U.S.; Non-U.S.). U.S. Adventists reported to
have a greater significant estimate of association versus all-causes
mortality as compared to Europeannon-Adventists.Suchdiffer-
ence has been already partly reported by the other recent meta-
analysis on cardiovascular mortality but not on all-cause mortal-
ity, (Kwok et al., 2014)thusreinforcingthehypothesisthatthe
studies coming from Adventist cohorts present a low degree of
generalizability when compared to other cohorts.
However, our study suffers from some limitations, which are
intrinsic of the studies included in the overall analysis. For
instance, we could not analyze an important datum such as the
duration of adherence to the vegetarian or to the vegan pattern
in the different cohorts. Indeed, only one study explicated this
finding that is extremely relevant for understanding the rela-
tionship with mortality and incidence of disease. In addition,
the definition of the control group, i.e., those following an
omnivorous diet was not really well-defined, including in some
cases subjects consuming a high intake of meat and meat prod-
ucts and in other cases subjects with a reduced consumption of
meat and derivatives. A final potential weakness is the accuracy
of the assessment of vegetarian and vegan status. There are sev-
eral slight differences in the population of vegetarians through-
out the world, and the possibility that some studies could have
included vegetarians and vegan altogether cannot be ruled out.
In conclusion, through using a systematic review and meta-
analytical approach we attempted to give some answers to com-
mon questions such as: are the vegetarian and vegan diets asso-
ciated with a protection versus cardiovascular and cancer
disease? From the analysis of the studies available in the litera-
ture we were able to determine that a significant protection ver-
sus ischemic heart disease and cancer is present in vegetarian
subjects, but that this protection is not significant for overall
mortality, cardio and cerebrovascular diseases. In addition,
vegan diet seems to be associated with a lower rate of cancer
incidence, but this result must be interpreted with caution,
because of the very small sample size and the low number of
studies evaluating this aspect. These findings are extremely
interesting for helping to give correct information to subjects
who want to adopt such dietary patterns.
Authors’contributions
Conception and design: F. Sofi, M. Dinu; Analysis and interpretation of the
data: F. Sofi, M. Dinu; Drafting of the article: M. Dinu, R. Abbate, A.
Casini, F. Sofi; Critical revision of the article for important intellectual con-
tent: R. Abbate, G.F. Gensini, A. Casini; Final approval of the article: R.
Abbate, G.F. Gensini, A. Casini, F. Sofi; and Statistical expertise: F. Sofi, M.
Dinu.
Financial disclosure
Nothing to declare.
Disclosure of interest
All authors declare that there are no conflicts of interest.
Guarantor of the paper
F. Sofi.
Figure 3. Funnel plot for studies investigating the incidence and/or mortality of
ischemic heart disease in vegetarians.
3648 M. DINU ET AL.
ORCID
Francesco Sofihttp://orcid.org/0000-0001-7113-7424
References
Appleby, P. N., Key, T. J., Thorogood, M., Burr, M. L. and Mann, J. (2002).
Mortality in British vegetarians. Public Health Nutr. 5:29–36.
Ashen, D. M. (2013). Vegetarian diets in cardiovascular prevention. Curr
Treat Options Cardiovasc Med. 15:735–745.
Chang-Claude, J., Hermann, S., Eilber, U. and Steindorf, K. (2005). Life-
style determinants and mortality in German vegetarians and health-
conscious persons: Results of a 21-year follow-up. Cancer Epidemiol
Biomarkers Prev 14:963–968.
Craig, W. J., Mangels, A. R. and American Dietetic Association. (2009).
Position of the american dietetic association: Vegetarian diets. J Am
Diet Assoc. 109:1266–1282.
Craig, W. J. (2009). Health effects of vegan diets. Am J Clin Nutr.
89:1627S–1633S.
Cade, J. E., Taylor, E. F., Burley, V. J. and Greenwood, D. C. (2010). Com-
mon dietary patterns and risk of breast cancer: Analysis from the
United Kingdom Women’s Cohort Study. Nutr Cancer 62:300–306.
Crowe, F. L., Appleby, P. N., Travis, R. C. and Key, T. J. (2013). Risk of
hospitalization or death from ischemic heart disease among British
vegetarians and nonvegetarians: Results from the EPIC-Oxford cohort
study. Am J Clin Nutr. 97:597–603.
Fraser, G. E. (1999). Associations between diet and cancer, ischemic heart
disease, and all-cause mortality in non-Hispanic white California Sev-
enth day Adventists. Am J Clin Nutr. 70:532S–538S.
Fraser, G. E. (2009). Vegetarian diets: What do we know of their effects on
common chronic diseases? Am J Clin Nutr. 89:1607S–1612S.
Greenhouse, J. B. and Iyengar, S. (2009). Sensitivity analysis and diagnos-
tics. In: The Handbook of Research Synthesis and Meta-Analysis. 2nd
ed. pp. 423–424. Cooper, H. M., Hedges, L. V. and Valentine, J. C., eds.
Russell Sage Foundation.
Huang, T., Yang, B., Zheng, J., Li, G., Wahlqvist, M. L. and Li, D. (2012).
Cardiovascular disease mortality and cancer incidence in vegetarians:
A meta-analysis and systematic review. Ann Nutr Metab. 60:233–240.
Higgins, J. P. T., Green, S. (eds). (xxxx). Cochrane handbook for systematic
reviews of interventions. Version 5.1.0. Available at: http://www.
cochrane-handbook.org
Key, T. J., Fraser, G. E., Thorogood, M., Appleby, P. N., Beral, V.,
Reeves, G., Burr, M. L., Chang-Claude, J., Frentzel-Beyme, R.,
Kuzma, J. W., Mann, J. and McPherson, K. (1999). Mortality in
vegetarians and nonvegetarians: Detailed findings from a collabora-
tive analysis of 5 prospective studies. Am J Clin Nutr. 70:516S–
524S.
Key, T. J., Appleby, P. N., Spencer, E. A., Travis, R. C., Roddam, A. W. and
Allen, N. E. (2009). Mortality in British vegetarians: Results from the
European Prospective Investigation into Cancer and Nutrition (EPIC-
Oxford). Am J Clin Nutr. 89:1613S–1619S.
Kwok, C. S., Umar, S., Myint, P. K., Mamas, M. A. and Loke, Y, K. (2014).
Vegetarian diet, Seventh Day Adventists and risk of cardiovascular
mortality: A systematic review and meta-analysis. Int J Cardiol.
176:680–686.
Key, T. J., Appleby, P. N., Crowe, F. L., Bradbury, K. E., Schmidt, J. A. and
Travis, R. C. (2014). Cancer in British vegetarians: Updated analyses of
4998 incident cancers in a cohort of 32,491 meat eaters, 8612 fish eat-
ers, 18,298 vegetarians, and 2246 vegans. Am J Clin Nutr. 100:378S–
385S.
Leitzmann, C. (2014). Vegetarian nutrition: Past, present, future. Am J Clin
Nutr. 100:496S–502S.
Le, L. T. and Sabat!
e, J. (2014). Beyond meatless, the health effects of
vegan diets: Findings from the Adventist cohorts. Nutrients
6:2131–2147.
Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G. and PRISMA Group.
(2009). Preferred reporting items for systematic reviews and meta-anal-
yses: The PRISMA statement. BMJ 339:b2535.
Orlich, M. J., Singh, P. N., Sabat!
e, J., Jaceldo-Siegl, K., Fan, J., Knutsen, S.,
Beeson, W. L. and Fraser, G. E. (2013). Vegetarian dietary patterns and
mortality in Adventist Health Study 2. JAMA Intern Med. 173:1230–
1238.
Orlich, M. J., Singh, P. N., Sabat!
e, J., Fan, J., Sveen, L., Bennett, H .,
Knutsen, S. F., Beeson, W. L., Jaceldo-Siegl, K., Butler, T. L., Her-
ring, R. P. and Fraser, G. E. (2015). Vegetarian dietary patterns
and the risk of colorectal cancers. JAMA Intern Med. 175:767–
776.
Tantamango-Bartley, Y., Jaceldo-Siegl, K., Fan, J. and Fraser, G. (2013).
Vegetarian diets and the incidence of cancer in a low-risk population.
Cancer Epidemiol Biomarkers Prev. 22:286–294.
Whittemore, R. and Knafl, K. (2005). The integrative review: Updated
methodology. J Adv Nur 52:546–553.
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3649
