Content uploaded by Alicja Wolk
Author content
All content in this area was uploaded by Alicja Wolk
Content may be subject to copyright.
Original Research Communications
Risk of overweight and obesity among semivegetarian,
lactovegetarian, and vegan women
1– 4
PK Newby, Katherine L Tucker, and Alicja Wolk
ABSTRACT
Background: Observational studies suggest that a plant-based diet
is inversely related to body mass index (BMI), overweight, and
obesity.
Objective: Our objective was to examine the BMI (kg/m
2
) and risk
of overweight and obesity of self-defined semivegetarian, lactoveg-
etarian, and vegan women.
Design: Data analyzed in this cross-sectional study were from
55 459 healthy women participating in the Swedish Mammography
Cohort. Women were asked whether they considered themselves to
be omnivores (n ҃ 54 257), semivegetarians (n ҃ 960), lactoveg-
etarians (n ҃ 159), or vegans (n ҃ 83), and this question was the main
exposure variable in this study. In secondary analyses, we reclassi-
fied women as lactovegetarians on the basis of food intakes reported
on the food-frequency questionnaire.
Results: The prevalence of overweight or obesity (BMI 욷 25) was
40% among omnivores, 29% among both semivegetarians and veg-
ans, and 25% among lactovegetarians. In multivariate, adjusted lo-
gistic regression analyses, self-identified vegans had a significantly
lower risk of overweight or obesity [odds ratio (OR) ҃ 0.35; 95% CI:
0.18, 0.69] than did omnivores, as did lactovegetarians (OR ҃ 0.54;
95% CI: 0.35, 0.85) and semivegetarians (OR ҃ 0.52; 95% CI: 0.43,
0.62). Risk of overweight or obesity remained significantly lower
among lactovegetarians classified on the basis of the food-frequency
questionnaire (OR ҃ 0.48; 95% CI: 0.30, 0.78).
Conclusions: Even if vegetarians consume some animal products,
our results suggest that self-identified semivegetarian, lactovegetar-
ian, and vegan women have a lower risk of overweight and obesity
than do omnivorous women. The advice to consume more plant
foods and less animal products may help individuals control their
weight. Am J Clin Nutr 2005;81:1267–74.
KEY WORDS Overweight, obesity, BMI, vegetarian, lacto-
vegetarian, vegan
INTRODUCTION
There is a critical need to control the obesity epidemic in light
of the increasing prevalence of overweight and obesity around
the world (1). Two extensive reviews of observational studies
that used eating pattern methods suggest that a plant-based diet
high in fiber-rich foods, such as vegetables, fruits, cereals, whole
grains, and legumes, is inversely related to body mass index
(BMI), overweight, and obesity (2, 3). Protective eating patterns
may also include lean protein foods such as chicken, fish, soy
products, and reduced-fat dairy products (2). The studies in these
reviews, however, varied in their design and adjustment for po-
tential confounders in the analyses. Additional, recent evidence
supports a protective effect of dietary fiber and whole grains
(4 –7), as well as dairy products and calcium (8, 9), although more
studies are needed.
Therefore, the scientific literature suggests that it may be use-
ful to further examine whether plant-based diets are inversely
related to obesity. Moreover, the question remains as to whether
animal products such as lean protein and dairy foods are helpful
for controlling weight. Studies examining different types of veg-
etarian eating patterns seem well poised to assess this relation.
Although some studies have shown associations between vege-
tarianism and blood pressure (10, 11), blood lipids (12–17), can-
cer (18), heart disease (18), and all-cause mortality (18 –20),
surprisingly few studies have rigorously examined the relation
between vegetarian eating patterns and obesity (21–25). Al-
though 3 reviews suggest that vegetarians have a lower BMI on
average (26 –28), much of the data on vegetarian diets and BMI
are baseline associations from studies whose primary outcome
was not obesity (10, 15, 18, 29) or from analyses that were not
adjusted for potential confounders (25). Although a few well-
designed studies have directly examined the relation between
vegetarian eating patterns and BMI (22, 24), additional studies
are needed to confirm these findings. Furthermore, no studies
that we are aware of have examined vegetarian eating patterns
among older Swedish women. The aim of the present study was
to examine the BMI and risk of overweight and obesity among
self-defined semivegetarian, lactovegetarian, and vegan women.
1
From the Jean Mayer US Department of Agriculture Human Nutrition
Research Center on Aging at Tufts University, Boston, MA (PKN and KLT),
and the Division of Nutritional Epidemiology, Department of Environmental
Medicine, Karolinska Institute, Stockholm, Sweden (AW).
2
Any opinions, findings, conclusions, or recommendations expressed in
this publication are those of the authors and do not necessarily reflect the view
of the US Department of Agriculture.
3
Supported by the Swedish Research Council/Section 4-Longitudinal
Studies, the Swedish Cancer Society, the Swedish Foundation for Interna-
tional Cooperation in Research and Higher Learning, and the US Department
of Agriculture (under agreement no. 58-1950-4-401).
4
Reprints not available. Address correspondence to PK Newby, Jean
Mayer USDA Human Nutrition Research Center on Aging at Tufts Univer-
sity, 711 Washington Street, 9th Floor, Boston, MA 02111. E-mail:
pknewby@post.harvard.edu.
Received October 11, 2004.
Accepted for publication January 12, 2005.
1267Am J Clin Nutr 2005;81:1267–74. Printed in USA. © 2005 American Society for Clinical Nutrition
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
SUBJECTS AND METHODS
Study design and participants
Data analyzed in this cross-sectional study were from the first
wave of the Swedish Mammography Cohort (SMC) in 1987–
1990, a population-based mammography screening program in-
troduced in Vastmanland and Uppsala counties in central Swe-
den. All women born between 1914 and 1948 living in these 2
counties were invited to the screening (n ҃ 90 903) and received
a 6-page questionnaire that included items about anthropometric,
reproductive, sociodemographic, and dietary factors. Among
those contacted, 66 651 (74%) completed the questionnaire and
agreed to participate; these methods are further described else-
where (30). In brief, questionnaires with missing or incorrect ID
numbers, questionnaires lacking a date, women who moved out
of the study area, and women who died during the study period
but were missing a date of death were excluded from the study.
Women with a previous cancer diagnosis as reported in the
Swedish Cancer Registry were also excluded. After these exclu-
sions, 61 433 participants in the SMC study remained, from
which we created our study sample.
To obtain a healthy cohort, we first excluded women with a
cardiovascular condition of angina, diabetes, coronary disease,
or stroke (n ҃ 1834); cardiovascular conditions were based on
hospital diagnoses. We next excluded subjects whose energy
intake appeared implausible (쏝2510 or 쏜16 736 kJ; n ҃ 747).
Of the remaining study participants, subjects with missing height
or weight data or with implausible values for weight (쏝40 or
쏜225 kg), height (쏜2.25 or 쏝1.4 m), or BMI (in kg/m
2
: 쏝14 or
쏜50) were excluded (n ҃ 2197). Last, women were excluded if
they did not answer the question describing their overall dietary
pattern, which was our main exposure variable (n ҃ 1196), as
described below. After all exclusions, 55 459 women were avail-
able for the analysis.
Dietary assessment
Dietary intakes were assessed with a food-frequency question-
naire (FFQ). On the FFQ, the women were asked whether they
considered themselves to be omnivorous (consume all foods),
semivegetarian (mostly lactovegetarian, sometime consume fish
or eggs), lactovegetarian (consume no meat, poultry, fish, or
eggs), or vegan (consume no meat, poultry, fish, eggs, or dairy
products). From this question, the women were grouped into 4
mutually exclusive categories of omnivore, semivegetarian, lac-
tovegetarian, and vegan (the last 3 patterns are referred to as
vegetarian eating patterns herein).
The FFQ contained 67 food questions, of which 60 were semi-
quantitative food items in which the subjects were asked how
often, on average, they had consumed the foods in the past 6 mo,
ranging from “never/seldom” to “4 or more times/d.” (The remain-
ing items on the questionnaire queried eating behaviors and were not
used in the present study.) Standard portion sizes were used on the
FFQ (eg, 1 slice of bread ҃ 1 serving). Frequency intakes were
converted to daily intakes (eg, 1–3 times/mo ҃ 0.07 servings/d on
the basis of the midpoint of 2 times/mo). Nutrients were derived
from reported food intakes by using the database from the Swedish
Food Administration. Foods were also categorized into several ma-
jor food groups (ie, fruit, vegetables, potatoes, legumes, cereals,
whole grains, refined grains, dairy products, meat, poultry, and fish)
to examine food group intakes across the different eating patterns.
Anthropometric assessment
Subjects self-reported their weight (kg) and height (m) on the
questionnaire; high validity has been observed for self-reported
height (r ҃ 1.0) and weight (r ҃ 0.9) compared with actual
measurements among Swedish women (31). From height and
weight, BMI (kg/m
2
) was calculated. On the basis of their BMI,
the women were classified as overweight (BMI 25–29.99) or
obese (BMI 욷 30) according to international cutoffs (32).
Covariate assessment
On the 1987 questionnaire, the women reported their age,
education, and marital status. Education was originally measured
in 6 categories (compulsory school, vocational or girls’ school,
junior secondary school, secondary school, university or college,
and other training) and was collapsed into 4 mutually exclusive
categories of less than high school, high school, university, and
other training. Marital status was reported as single, cohabitating
but not married, married, divorced, or widowed.
Because only a limited set of covariates was assessed in 1987–
1990, we used data from the second wave of the SMC study in
1997 to further adjust our regression analyses, including only
those variables that could not have changed between 1987 and
1997. Specifically, the questionnaire in 1997 included questions
on the age at birth of the first child, parity, and body shape at 10 y
of age. Subjects were asked to recall their childhood body shape
(“How was your figure at 10 y of age?”) by choosing among 7
figures of increasing body size, ranging from smallest (figure 1)
to largest (figure 7). The women were also asked to recall their
age at the birth of their first child and to report how many children
they had. Smoking habit (current, previous, or never) was also
assessed. We re-categorized smokers as never smokers and ever
smokers, because a never smoker in 1997 would also be a never
smoker in 1987 but it could not be determined whether a current
or previous smoker in 1997 would have had the same smoking
status in 1987.
Statistical analyses
Sample characteristics were described by using means and
SDs for continuous variables and frequencies (number and per-
cent) for categorical variables. We used Tukey’s honestly sig-
nificant differences test for continuous covariates and the chi-
square test for categorical covariates to examine differences in
sample characteristics across groups. Mean (앐SE) intakes of
macronutrients and food groups among each eating pattern group
were calculated by using a generalized linear model and were
adjusted for multiple comparisons by using Tukey’s honestly
significant differences test.
Linear and logistic regression analyses were performed, and
each analysis included an indicator variable for each vegetarian
eating pattern (eg, semivegetarian, lactovegetarian, or vegan) in
the same model; omnivores were treated as the reference group.
Two sets of linear regression analyses were performed, and alpha
was set at 0.05. In the first analysis, the outcome was BMI, and
in the second, the outcome was weight. For each outcome, the
first model was adjusted for age and energy intake and the second
model was multivariate adjusted for age, energy intake, alcohol
intake, education, marital status, smoking status, parity, age at
first birth, and childhood body shape for a subset of women for
whom these data were available. A final model was tested with
additional adjustment for intakes of total fat (% of energy/d) and
1268 NEWBY ET AL
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
fiber (g/d). Last, we added a quadratic term for age to see if model
fit improved. We also tested all models without baseline energy
intake, because energy may be in the causal pathway of the
association between eating patterns and weight. Models predict-
ing weight were also adjusted for height.
Two sets of logistic regression analyses were also performed.
In the first analysis, overweight or obesity (BMI 욷 25) was the
outcome variable; in the second, obesity (BMI 욷 30) was the
outcome variable. Models were built in the same way as de-
scribed above, and odds ratios (ORs) and 95% CIs were calcu-
lated for each vegetarian group.
In a secondary analysis to check the associations between
vegetarian eating pattern and BMI, overweight, and obesity, we
reclassified women as actual lactovegetarians or actual vegans
on the basis of reported dietary intakes on the FFQ, which we
defined as zero consumption of meat, fish, and eggs or zero
consumption of meat, fish, eggs, and dairy products, respec-
tively. All regression analyses were repeated in the group of
actual lactovegetarians.
We last performed a final set of regression analyses following
the above modeling procedure among never smokers in which we
excluded current or former smokers to prevent confounding by
smoking status. Although our primary regression models above
were adjusted for smoking, residual confounding may have af-
fected our estimates. We could not discern whether ever smokers
were current or former smokers in 1987, because smoking habit
may have changed during the 10-y follow-up period when smok-
ing status was measured. All analyses were performed by using
SAS for WINDOWS, version 8.2 (SAS Institute Inc, Cary, NC).
RESULTS
A small percentage of women were semivegetarian (1.73%),
lactovegetarian (0.29%), or vegan (0.15%). The vegan women
were significantly older than the omnivorous, semivegetarian,
and lactovegetarian women (P 쏝 0.05; Table 1). The omnivo-
rous women were significantly heavier (66.6 앐 10.9 kg) than any
of the 3 vegetarian groups and also had a significantly higher
BMI (24.7 앐 3.9). The prevalence of overweight or obesity
(BMI 욷 25) was 40% among omnivores, 29% among both semi-
vegetarians and vegans, and 25% among lactovegetarians. A
higher percentage of semivegetarians (9%) and lactovegetarians
(10%) attended university compared with omnivores (5%) and
vegans (6%). Vegans had the highest percentage of never smok-
ers (72%).
As shown in Table 2, intakes of all macronutrients differed
significantly across eating patterns. Omnivores had significantly
higher energy (P 쏝 0.005) and protein (P 쏝 0.0003) intakes, and
significantly lower carbohydrate intake (P 쏝 0.001), than did any
of the 3 vegetarian groups. Omnivores also consumed a greater
percentage of energy from saturated fat and monounsaturated fat
than did all 3 vegetarian groups and had the lowest intake of fiber.
Omnivores had the highest intakes of refined grains and animal
foods and the lowest intakes of fruit and vegetables (Table 3).
Few significant differences in food group intakes were noted
among the 3 vegetarian eating groups: vegans consumed more
vegetables and less dairy products than did lacto- or semiveg-
etarians, whereas semivegetarians consumed the most fish.
In multivariate, adjusted linear regression analyses (Table 4),
women who were semivegetarian (b ҃Ҁ1.13; 95% CI: Ҁ1.42,
Ҁ0.84; P 쏝 0.005), lactovegetarian (b ҃Ҁ1.07; 95% CI: Ҁ1.81,
Ҁ0.33; P 쏝 0.005), or vegan (b ҃Ҁ1.29; 95% CI: Ҁ2.33,
Ҁ0.25; P 쏝 0.005) had a significantly lower BMI than did om-
nivores. Among the 3 vegetarian groups, vegans had the lowest
weight (b ҃Ҁ5.31; 95% CI: Ҁ7.82, Ҁ2.80; P 쏝 0.005) com-
pared with omnivores.
In multivariate, adjusted logistic regression analyses (Table
5), vegans had a significantly lower risk of overweight or obesity
(OR ҃ 0.35; 95% CI: 0.18, 0.69) than did omnivores, as did
lactovegetarians (OR ҃ 0.54; 95% CI: 0.35, 0.85) and semiveg-
etarians (OR ҃ 0.52; 95% CI: 0.43, 0.62). Semivegetarians also
had a lower risk of obesity (OR ҃ 0.46; 95% CI: 0.31, 0.66) than
did omnivores. When lactovegetarians and vegans were col-
lapsed into one group, there was a similar reduction in risk for
obesity (OR ҃ 0.46; 95% CI: 0.26, 0.83), although the odds ratio
was no longer significant when further adjusted for covariates.
In secondary analyses, we counted 158 actual lactovegetarians
and 14 actual vegans in the sample when defined on the basis of
intakes reported in the FFQ; we only analyzed lactovegetarians
because of the small number of vegans available for analysis. In
a multivariate adjusted model, actual lactovegetarians had a sig-
nificantly lower BMI (b ҃Ҁ1.68; 95% CI: Ҁ2.46, Ҁ0.90; P ҃
0.0013) and lower weight (b ҃Ҁ4.17; 95% CI: Ҁ6.01, Ҁ2.33;
P ҃ 0.0062) than did omnivores. The risk of overweight or
obesity remained significantly lower among actual lactovegetar-
ians (OR ҃ 0.48; 95% CI: 0.30, 0.78; data not shown).
The quadratic term for age was significant, so it was retained
in the final multivariate adjusted models above. Our results for all
the regression analyses were similar and remained significant
when fat and fiber were added to the model, when we omitted
energy from the model, and when all analyses were repeated
among never smokers (data not shown).
DISCUSSION
In the present study, the mean weight, BMI, and prevalence of
overweight and obesity were highest among omnivores com-
pared with semivegetarians, lactovegetarians, and vegans. In a
multivariate, adjusted logistic regression model, vegans had a
significantly lower risk of overweight or obesity than did omni-
vores, as did lactovegetarians and semivegetarians. A reduced
risk of overweight or obesity was also observed when we recat-
egorized women as actual lactovegetarians on the basis of re-
ported dietary intake on the FFQ rather than on the basis of
whether they identified themselves as a vegetarian.
Although the number of vegetarians in this population was
small, the large sample size of the study allowed us to examine
dietary associations between BMI and overweight or obesity
among subgroups of vegetarians and to detect significant effects.
We were also able to adjust for many potential confounders and
additional risk factors, including education, smoking, marital
status, parity, age at first birth, and childhood body shape. Our
results were similar and remained significant in an analysis lim-
ited to never smokers. All 3 vegetarian groups had BMIs 앒1 unit
lower than that of the omnivores, which is of a similar magnitude
to the difference reported in other studies (26, 27). Our logistic
regression results suggest that vegans may be at an even lower
risk of overweight or obesity than semi- and lactovegetarians
(65% risk reduction compared with 48% and 46%, respectively),
although the CIs between the groups overlapped. Three other
studies observed a lower mean BMI among vegans than in other
types of vegetarians (14, 22, 23). Differences in plasma lipids
BMI AND OVERWEIGHT AMONG VEGETARIANS 1269
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
among different types of vegetarians have also been reported
(14 –16). More research is needed to further examine whether
different types of vegetarian eating patterns have different effects
on weight and health.
Our study raises interesting questions about how individuals
define themselves and whether this information is useful in re-
search even if it does not completely reflect “truth.” In this case,
our questionnaire asked participants to self-identify themselves
as omnivorous, semivegetarian, lactovegetarian,orvegan, and
these terms were specifically defined for them. When we looked
at intakes of major food groups across the 3 vegetarian groups on
the basis of the FFQ, however, we found that in none of the
self-defined vegetarian groups, including vegans, was mean an-
imal product intake actually zero. We were able to analyze actual
lactovegetarians, and these individuals showed a similar relation
with BMI and risk of overweight or obesity as did self-reported
lactovegetarians and vegans. Therefore, that self-reported veg-
etarians consumed some animal foods may not be biologically
meaningful with regard to obesity and weight, because the food
and nutrient intakes of vegetarians differed significantly from
those of omnivores despite some animal product consumption,
and a significant protective effect was still observed.
Our finding that self-identified vegetarians consume meat is
not unusual (25, 33). A study of 13 313 Americans aged 욷6y
found that only 36% of self-defined vegetarians reported con-
suming no meat (33). In that study, the investigators further
stratified the self-defined vegetarians and nonvegetarians into
those who did and did not report consuming meat on their diet
recalls. They found that self-defined vegetarians aged 욷20 y
had a lower BMI than did self-defined nonvegetarians, re-
gardless of whether they ate meat (33); this result supports our
finding.
TABLE 1
Sample characteristics of 55 459 self-identified omnivorous, semivegetarian, lactovegetarian, and vegan women participating in the Swedish
Mammography Cohort
Characteristic Sample size
Omnivores
(n ҃ 54 257)
Semivegetarians
(n ҃ 960)
Lactovegetarians
(n ҃ 159)
Vegans
(n ҃ 83) P
1
Age (y) 55 459 52.5 앐 9.7
2,a,b
53.6 앐 9.8
a
51.1 앐 9.5
b
54.8 앐 9.5
c
쏝 0.05
Weight (kg) 55 459 66.6 앐 10.9
a
63.6 앐 10.0
b
64.0 앐 10.9
b
62.4 앐 10.7
b
쏝 0.05
Height (m) 55 459 1.64 앐 0.1 1.64 앐 0.1 1.65 앐 0.1 1.64 앐 0.1 0.14
BMI (kg/m
2
) 55 459 24.7 앐 3.9
a
23.6 앐 3.5
b
23.4 앐 3.5
b
23.3 앐 3.8
b
쏝 0.005
Overweight [n (%)] 55 459 16 285 (30) 228 (24) 33 (21) 19 (23) 쏝 0.0001
Obese [n (%)] 55 459 5220 (10) 49 (5) 7 (4) 5 (6) 쏝 0.0001
Education [n (%)] 54 570 쏝0.0001
Less than high school 42 725 (80) 655 (70) 110 (70) 59 (74)
High school 3738 (7) 81 (9) 14 (9) 9 (11)
University 2522 (5) 87 (9) 16 (10) 5 (6)
Other training 4404 (8) 120 (12) 17 (11) 8 (9)
Marital status [n (%)] 54 957 쏝 0.0001
Single 3156 (6) 89 (9) 15 (9) 7 (9)
Married 38 163 (71) 564 (59) 86 (54) 51 (63)
Cohabitating 3458 (6) 51 (5) 14 (9) 2 (3)
Divorced 4623 (9) 149 (17) 31 (20) 14 (16)
Widowed 4369 (8) 96 (10) 12 (8) 7 (9)
Smoking status [n (%)]
3
31 473 쏝 0.0001
Never 18 479 (60) 409 (68) 61 (66) 33 (72)
Past 7619 (25) 153 (24) 26 (28) 11 (24)
Current 4628 (15) 46 (8) 6 (6) 2 (4)
Age at birth of first child [n (%)]
3
31 351 0.004
울 24 y 18 102 (59) 309 (54) 48 (58) 25 (55)
25–30 y 10 168 (33) 215 (38) 26 (32) 19 (41)
쏜 30 y 2382 (8) 47 (8) 8 (10) 2 (4)
Parity [n (%)]
3
34 825 0.18
0 2521 (7) 65 (10) 10 (10) 1 (2)
1–2 19 805 (58) 374 (58) 54 (54) 28 (56)
3–4 10 196 (30) 179 (28) 31 (31) 16 (32)
쏜 4 1502 (4) 33 (5) 5 (5) 5 (10)
Body shape at age 10 y [n (%)]
3,4
35 024 0.14
1–2 18 365 (54) 341 (52) 52 (50) 24 (48)
3–4 12 796 (37) 268 (41) 47 (45) 19 (38)
욷 4 3052 (9) 48 (7) 5 (5) 7 (14)
1
P values were obtained by using Tukey’s honestly significant differences test for continuous variables (age, weight, height, and BMI) and the chi-square
test for categorical variables (overweight, obesity, education, marital status, smoking status, parity, age at first birth, and body shape at age 10 y). For continuous
variables, groups with different superscript letters are significantly different, P 쏝 0.05.
2
x 앐 SD (all such values).
3
Data were available only for women who participated in the follow-up study in 1997 (n ҃ 38 984); missing data are not included in the calculations.
4
Body shape was assessed by use of a pictogram, which asked subjects to rank their body size on a scale ranging from 1 (smallest) to 7 (largest).
1270 NEWBY ET AL
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
We found that food and nutrient intakes across the 3 groups of
vegetarians showed many significant differences. Notably, all
the vegetarian groups had higher intakes of fruit, vegetables, and
fiber and lower intakes of fat and protein. Studies of vegetarians
in the United Kingdom (16), Israel (34), Canada (29), and the
United States (33) also reported higher intakes of fiber and car-
bohydrate and lower intakes of protein and saturated fat by veg-
etarians than by omnivores. A higher intake of carbohydrates and
fiber among vegetarians is expected, because plant-based foods
are composed mainly of carbohydrates. That vegetarians are
leaner and have a reduced risk of overweight or obesity despite
higher total carbohydrate intake points to the importance of dif-
ferentiating between types of carbohydrate when selecting diets,
including weight-loss diets. Current fad diets that emphasize low
carbohydrate intakes ignore the fact that whole and refined car-
bohydrate foods evoke different metabolic responses, thereby
exerting different effects on appetite and energy intake. This
study and others (5, 7) suggest that a high-carbohydrate diet may
be protective against obesity if the carbohydrates come from
fiber-rich foods such as fruit, vegetables, and whole grains.
Nutritional studies that examine eating patterns, rather than single
nutrients, cannot determine the biological mechanisms responsible
for observed effects. In fact, several different mechanisms involving
both macro- and micronutrients are likely responsible for the lower
risk of obesity and smaller BMI that we observed. In an elegant study
of 38 000 participants in EPIC-Oxford, Spencer et al (22) found that
TABLE 2
Energy and macronutrient intakes among 55 459 self-identified omnivorous, semivegetarian, lactovegetarian, and vegan women participating in the
Swedish Mammography Cohort
1
Energy or macronutrient (intake/d) Omnivores (n ҃ 54 257) Semivegetarians (n ҃ 960) Lactovegetarians (n ҃ 159) Vegans (n ҃ 83) P
Energy (kJ)
2
5766 앐 8
a
5183 앐 50
b
5067 앐 117
b
4786 앐 167
b
쏝 0.005
Carbohydrate (% of energy)
3
50.9 앐 0.1
a
57.3 앐 0.2
b
59.8 앐 0.5
c
62.7 앐 0.6
d
쏝 0.001
Protein (% of energy)
3
16.3 앐 0.0
a
14.7 앐 0.1
b
13.5 앐 0.2
c
12.4 앐 0.2
d
쏝 0.005
Fat (% of energy)
3
30.7 앐 0.0
a
26.0 앐 0.2
b
25.2 앐 0.4
b
23.0 앐 0.6
c
쏝 0.05
Saturated fat (% of energy)
3
13.0 앐 0.0
a
11.4 앐 0.1
b
11.1 앐 0.2
b
9.0 앐 0.3
c
쏝 0.0001
Monounsaturated fat (% of energy)
3
11.2 앐 0.0
a
8.8 앐 0.1
b
8.4 앐 0.2
b,c
8.1 앐 0.2
c
쏝 0.05
Polyunsaturated fat (% of energy)
3
4.4 앐 0.0
a
3.9 앐 0.0
b
3.7 앐 0.1
b
4.1 앐 0.1
a,b
쏝 0.0001
Fiber (g)
3
17.0 앐 0.1
a
20.9 앐 0.1
b
22.4 앐 0.3
c
23.0 앐 0.5
c
쏝 0.0001
Alcohol (g)
3
3.2 앐 0.0
a
2.7 앐 0.1
b
1.8 앐 0.2
c
2.0 앐 0.3
b,c
쏝 0.005
1
All values are x 앐 SE. Values in a row with different superscript letters are significantly different, P 쏝 0.05 (Tukey’s honestly significant differences
test).
2
Adjusted for age.
3
Adjusted for age and total energy intake.
TABLE 3
Food group intakes for 55 459 self-identified omnivorous, semivegetarian, lactovegetarian, and vegan women participating in the Swedish Mammography
Cohort
1
Food group (servings/d) Omnivores (n ҃ 54 257) Semivegetarians (n ҃ 960) Lactovegetarians (n ҃ 159) Vegans (n ҃ 83) P
Fruit
2
1.5 앐 0.0
a
2.0 앐 0.0
b
2.2 앐 0.1
b
2.2 앐 0.1
b
쏝0.0001
Vegetables
3
1.7 앐 0.0
a
2.5 앐 0.0
b
2.6 앐 0.1
b
3.2 앐 0.1
c
쏝0.0001
Potatoes
4
0.81 앐 0.00
a
0.66 앐 0.01
b
0.74 앐 0.03
a,b
0.86 앐 0.02
a
쏝0.0005
Legumes
5
0.05 앐 0.0
a
0.05 앐 0.0
a,b
0.06 앐 0.0
a,b,c
0.08 앐 0.1
c
쏝0.05
Cereal
6
0.27 앐 0.00
a
0.40 앐 0.03
b
0.40 앐 0.03
b
0.22 앐 0.04
a
쏝0.005
Whole grains
7
1.9 앐 0.1
a
2.2 앐 0.0
b
2.2 앐 0.1
b
2.0 앐 0.2
a,b
쏝0.05
Refined grains
8
0.88 앐 0.0
a
0.60 앐 0.0
b
0.58 앐 0.1
b
0.61 앐 0.1
b
쏝0.01
Eggs 0.19 앐 0.0
a
0.16 앐 0.0
b
0.06 앐 0.02
c
0.04 앐 0.02
c
쏝0.0001
Dairy products
9
4.9 앐 0.0
a
4.3 앐 0.1
b
4.1 앐 0.2
b
2.7 앐 0.2
c
쏝0.0003
Meat
10
1.1 앐 0.0
a
0.1 앐 0.0
b
0.2 앐 0.0
b
0.3 앐 0.1
b
쏝0.0001
Poultry
11
0.06 앐 0.0
a
0.02 앐 0.0
b
0.01 앐 0.1
c
0.02 앐 0.1
b,c
쏝0.05
Fish
12
0.28 앐 0.00
a
0.34 앐 0.01
b
0.07 앐 0.01
c
0.15 앐 0.02
d
쏝0.05
1
All values are x 앐 SD. Values in a row with different superscript letters are significantly different, P 쏝 0.05 (Tukey’s honestly significant differences
test).
2
Includes apples, pears, bananas, and citrus fruit.
3
Includes lettuce, root vegetables, spinach, tomatoes, and white cabbage.
4
Includes boiled potatoes, French fries, and fried potatoes.
5
Includes brown beans and pea soup.
6
Includes ready-to-eat (cold) cereal and muesli.
7
Includes hard (crisp) bread, porridge, and whole-meal bread.
8
Includes white bread, rice, pancakes, and spaghetti.
9
Includes milk, sour milk, butter, cheese, and margarine (both full-fat and reduced-fat products).
10
Includes meat, stews, casseroles, minced meat, sandwich meats, bacon, sausage, black pudding, liver, and paˆté.
11
Includes chicken.
12
Includes fatty fish (eg, salmon, mackerel, and herring), other fish, and shellfish.
BMI AND OVERWEIGHT AMONG VEGETARIANS 1271
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
high protein (as a percentage of energy) and low fiber intakes
were the dietary factors most strongly associated with increasing
BMI and that differences in macronutrient intakes accounted for
about one-half the difference in mean BMI between vegans and
meat-eaters. Another study of 5292 vegetarians in the United King-
dom found that differences in BMI between meat-eaters and vege-
tarians were partly, but not wholly, explained by differences in
animal fat, fiber, and alcohol intakes (24). Likewise, our results
remained similar and significant when we adjusted our models for
total fat and fiber intakes, which suggests that additional, significant
dietary components of a vegetarian diet are protective.
Because differences in foods and nutrients essentially define
the different vegetarian eating patterns, and may be in the causal
pathway between diet and body weight, it is arguable whether
some dietary variables should be adjusted for in the analysis. For
example, it is likely that the high fiber consumption among veg-
etarians is partly responsible for the lower energy intakes ob-
served in some studies (21, 22, 33) and ours, because fiber con-
tributes to greater satiation and satiety and hence decreased
energy intakes within and between meals (35). In the present
study, energy intakes were higher and fiber intakes were lower as
more animal products were included in the diet. Low energy
intakes among vegetarian groups may also be because we used a
short questionnaire designed for the general population that did
not specifically include vegetarian foods (eg, soy products, meat-
less products, tofu, and hummus). However, estimated energy
intakes among all groups were low in the present study, likely
because of the underreporting that is common with the FFQ
method. It is controversial as to whether underreporting is asso-
ciated with dietary composition (36, 37), although there did not
appear to be distortion in macronutrient intakes expressed as a
percentage of total energy in our study. Omitting total energy
from our models did not appreciably change our regression re-
sults. One review notes that the probability of underreporting
increases as BMI increases (37), although we do not expect that
underreporting would affect self-report of vegetarian status.
Our study has several limitations, including its cross sectional
design. Although data on weight from the follow-up study of the
TABLE 4
Linear regression coefficients (

) and SEs showing the association
between self-identified vegetarian eating patterns and BMI and weight
among 55 459 women participating in the Swedish Mammography Cohort
Self-identified vegetarian
eating pattern
1

(SE)
BMI (kg/m
2
) Weight (kg)
2
Semivegetarian
3
Adjusted for age and energy Ҁ1.30 (0.12)
4
Ҁ3.47 (0.33)
4
Multivariate adjusted
5
Ҁ1.13 (0.15)
4
Ҁ3.29 (0.36)
6
Lactovegetarian
7
Adjusted for age and energy Ҁ1.25 (0.30)
4
Ҁ3.22 (0.80)
4
Multivariate adjusted
5
Ҁ1.07 (0.38)
4
Ҁ2.74 (0.90)
4
Vegan
8
Adjusted for age and energy Ҁ1.76 (0.41)
4
Ҁ4.75 (1.06)
4
Multivariate adjusted
5
Ҁ1.29 (0.53)
6
Ҁ5.31 (1.28)
4
1
All 3 eating patterns were tested in the same model with omnivores as
the reference group.
2
Models were further adjusted for height.
3
Semivegetarians were defined as women who considered themselves
mostly lactovegetarian but who sometimes consumed fish or eggs.
4
P 쏝 0.005.
5
Further adjusted for age squared, alcohol intake, education, marital
status, smoking status, parity, age at first birth, and childhood body shape.
6
P 쏝 0.05.
7
Lactovegetarians were defined as women who did not consume meat,
fish, or eggs.
8
Vegans were defined as women who did not consume meat, fish, eggs,
or dairy products.
TABLE 5
Odds ratios (ORs) and 95% CIs showing the association between self-identified vegetarian eating patterns and overweight and obesity among 55 459
women participating in the Swedish Mammography Cohort
Self-identified vegetarian eating pattern
1
Overweight or obese (BMI 욷 25) Obese (BMI 욷 30)
OR 95% CI OR 95% CI
Semivegetarian
2
Adjusted for age and energy 0.55 0.48, 0.64 0.47 0.35, 0.62
Multivariate adjusted
3
0.52 0.43, 0.62 0.46 0.31, 0.66
Lactovegetarian
4
Adjusted for age and energy 0.51 0.35, 0.73 0.46 0.26, 0.83
Multivariate adjusted
3
0.54 0.35, 0.85 0.64 0.34, 1.19
Vegan
5
Adjusted for age and energy 0.51 0.31, 0.82 0.52 0.21, 1.28
Multivariate adjusted
3
0.35 0.18, 0.69 0.64 0.29, 1.40
Lactovegetarian ѿ vegan
6
Adjusted for age and energy 0.50 0.38, 0.68 0.46 0.26, 0.83
Multivariate adjusted
3
0.47 0.32, 0.68 0.64 0.29, 1.41
1
All 3 eating patterns were tested in the same model with omnivores as the reference group.
2
Semivegetarians were defined as women who considered themselves mostly lactovegetarian but who sometimes consumed fish or eggs.
3
Multivariate models were further adjusted for age squared, alcohol intake, education, marital status, smoking status, parity, age at first birth, and childhood
body shape.
4
Lactovegetarians were defined as women who did not consume meat, fish, or eggs.
5
Vegans were defined as women who did not consume meat, fish, eggs, or dairy products.
6
Because of the small numbers of obese lactovegetarians (n ҃ 7) and obese vegans (n ҃ 5), the 2 groups were collapsed into 1 category.
1272 NEWBY ET AL
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
Swedish Mammography Cohort are available, these measures
are 앒10 y away from baseline and are available only for a smaller
sample of women who participated in the second wave of the
study (앒65% of our baseline study sample). In addition, the FFQ
in 1997 was modified and excluded our primary exposure vari-
able on how women defined their eating pattern. Furthermore, we
had no data on changes in weight and changes in diet during the
10-y follow-up period.
Cross-sectional studies of obesity and body weight cannot
determine causality, because it cannot be determined whether
diet led to obesity or whether obesity led to changes in diet. In our
study, reverse causality seems unlikely, because we would have
observed positive associations between vegetarian patterns and
BMI or overweight or obesity, which would most likely indicate
that overweight women switched to a vegetarian diet in an at-
tempt to lose weight. Also, individuals usually choose to become
vegetarians earlier in life than the age of the women in this study
(앒52 y on average), so it seems unlikely that the women in this
study would have become vegetarians simply to lose weight.
Because our study showed inverse relations between vegetarian
eating patterns and overweight or obesity and our results are
supported by other studies, our findings may reflect true associ-
ations; however, prospective research studies and intervention
studies are needed to confirm our findings.
An additional limitation of our study is that it only included
older women. One review, however, found that vegetarians of all
ages and both sexes are leaner than omnivores (26), which sug-
gests that our results may be generalizable to younger women and
to men. Last, although we were able to adjust for some confound-
ers, we were not able to adjust for physical activity. A large study,
however, found that differences in BMI among vegetarian
groups remained significant in models adjusted for physical ac-
tivity and nondietary lifestyle factors (including smoking and
physical activity) and accounted for 쏝5% of the differences in
BMI among these dietary groups (22).
In conclusion, even if vegetarians consume some animal prod-
ucts as part of their diet, our results suggest that self-identified
semivegetarian, lactovegetarian, and vegan women have a lower
risk of overweight and obesity than do omnivorous women. Ad-
ditional studies focusing on single nutrients are needed to eluci-
date the precise dietary mechanisms responsible for this associ-
ation, and prospective or intervention studies that adjust for
physical activity and other confounders are also needed. The
advice to consume more plant foods and less animal products
may help individuals control their weight.
We thank Christoph Weismayer and Niclas Hayes at the Karolinksa In-
stitute for assistance in running statistical programs and Paul Jacques and
Martha Morris for helpful comments on the analysis.
PKN was responsible for the design and analysis of this report and drafted
the manuscript. KLT contributed to the analysis. AW contributed to the
original design and data collection for the Swedish Mammography Cohort.
All authors made critical comments during the preparation of the manuscript
and fully accept responsibility for the work. No author had a financial interest
or professional or personal affiliation that compromised the scientific integ-
rity of this work.
REFERENCES
1. Seidell JC. Obesity: a growing problem. Acta Paediatrica Suppl 1999;
88:46 –50.
2. Newby PK, Tucker KL. Empirically derived eating patterns using factor
or cluster analysis: a review. Nutr Rev 2004;62:177–203.
3. Togo P, Osler M, Sorensen TI, Heitmann BL. Food intake patterns and
body mass index in observational studies. Int J Obes Relat Metab Disord
2001;25:1741–51.
4. Steffen LM, Jacobs DR Jr, Murtaugh MA, et al. Whole grain intake is
associated with lower body mass and greater insulin sensitivity among
adolescents. Am J Epidemiol 2003;158:243–50.
5. Liu S, Willett WC, Manson JE, Hu FB, Rosner B, Colditz G. Relation
between changes in intakes of dietary fiber and grain products and
changes in weight and development of obesity among middle-aged
women. Am J Clin Nutr 2003;78:920–7.
6. Koh-Banerjee P, Rimm EB. Whole grain consumption and weight gain:
a review of the epidemiological evidence, potential mechanisms and
opportunities for future research. Proc Nutr Soc 2003;62:25–9.
7. Newby P, Muller D, Hallfrisch J, Qiao N, Andres R, Tucker K. Dietary
patterns and changes in body mass index and waist circumference in
adults. Am J Clin Nutr 2003;77:1417–25.
8. Zemel MB, Thompson W, Milstead A, Morris K, Campbell P. Calcium
and dairy acceleration of weight and fat loss during energy restriction in
obese adults. Obes Res 2004;12:582–90.
9. Zemel MB. Role of calcium and dairy products in energy partitioning
and weight management. Am J Clin Nutr 2004;79(suppl):907S–12S.
10. Rouse IL, Armstrong BK, Beilin LJ. The relationship of blood pressure
to diet and lifestyle in two religious populations. J Hypertens 1983;1:
65–71.
11. Appleby PN, Davey GK, Key TJ. Hypertension and blood pressure
among meat eaters, fish eaters, vegetarians and vegans in EPIC-Oxford.
Public Health Nutr 2002;5:645–54.
12. Reddy S, Sanders TA. Lipoprotein risk factors in vegetarian women of
Indian descent are unrelated to dietary intake. Atherosclerosis 1992;95:
223–9.
13. Krajcovicova-Kudlackova M, Simoncic R, Bederova A, Ondreicka R,
Klvanova J. Selected parameters of lipid metabolism in young vegetar-
ians. Ann Nutr Metab 1994;38:331–5.
14. Toohey ML, Harris MA, DeWitt W, Foster G, Schmidt WD, Melby
CL. Cardiovascular disease risk factors are lower in African-
American vegans compared to lacto-ovo-vegetarians. J Am Coll Nutr
1998;17:425–34.
15. Key TJ, Fraser GE, Thorogood M, et al. Mortality in vegetarians and
non-vegetarians: a collaborative analysis of 8300 deaths among 76,000
men and women in five prospective studies. Public Health Nutr 1998;
1:33– 41.
16. Appleby PN, Thorogood M, Mann JI, Key TJ. The Oxford Vegetarian
Study: an overview. Am J Clin Nutr1999;70(suppl):525S–31S.
17. Lin CL, Fang TC, Gueng MK. Vascular dilatory functions of ovo-
lactovegetarians compared with omnivores. Atherosclerosis 2001;158:
247–51.
18. Fraser GE. Associations between diet and cancer, ischemic heart disease,
and all-cause mortality in non-Hispanic white California Seventh-day
Adventists. Am J Clin Nutr 1999;70(suppl):532S–8S.
19. Key TJ, Fraser GE, Thorogood M, et al. Mortality in vegetarians and
nonvegetarians: detailed findings from a collaborative analysis of 5
prospective studies. Am J Clin Nutr 1999;70(suppl):516S–24S.
20. Appleby PN, Key TJ, Thorogood M, Burr ML, Mann J. Mortality in
British vegetarians. Public Health Nutr 2002;5:29–36.
21. Phillips F, Hackett AF, Stratton G, Billington D. Effect of changing to a
self-selected vegetarian diet on anthropometric measurements in UK
adults. J Hum Nutr Diet 2004;17:249–55.
22. Spencer EA, Appleby PN, Davey GK, Key TJ. Diet and body mass index
in 38 000 EPIC-Oxford meat-eaters, fish-eaters, vegetarians and vegans.
Int J Obes Relat Metab Disord 2003;27:728 –34.
23. Key T, Davey G. Prevalence of obesity is low in people who do not eat
meat. BMJ 1996;313:816 –7.
24. Appleby PN, Thorogood M, Mann JI, Key TJ. Low body mass index in
non-meat eaters: the possible roles of animal fat, dietary fibre and alco-
hol. Int J Obes Relat Metab Disord 1998;22:454 – 60.
25. Brathwaite N, Fraser HS, Modeste N, Broome H, King R. Obesity,
diabetes, hypertension, and vegetarian status among Seventh-Day Ad-
ventists in Barbados: preliminary results. Ethn Dis 2003;13:34 –9.
26. Key TJ, Davey GK, Appleby PN. Health benefits of a vegetarian diet.
Proc Nutr Soc 1999;58:271–5.
27. Dwyer JT. Health aspects of vegetarian diets. Am J Clin Nutr 1988;48:
712–38.
28. Kennedy ET, Bowman SA, Spence JT, Freedman M, King J. Popular
BMI AND OVERWEIGHT AMONG VEGETARIANS 1273
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from
diets: correlation to health, nutrition, and obesity. J Am Diet Assoc
2001;101:411–20.
29. Barr SI, Broughton TM. Relative weight, weight loss efforts and
nutrientintakes among health-conscious vegetarian, past vegetarian
and nonvegetarian women ages 18 to 50. J Am Coll Nutr 2000;19:
781– 8.
30. Michels KB, Holmberg L, Bergkvist L, Wolk A. Coffee, tea, and caf-
feine consumption and breast cancer incidence in a cohort of Swedish
women. Ann Epidemiol 2002;12:21– 6.
31. Kuskowska-Wolk A, Karlsson P, Stolt M, Rossner S. The predictive
validity of body mass index based on self-reported weight and height. Int
J Obes 1989;13:441–53.
32. James PT, Leach R, Kalamara E, Shayeghi M. The worldwide obesity
epidemic. Obes Res 2001;9(suppl):228S–33S.
33. Haddad EH, Tanzman JS. What do vegetarians in the United States eat?
Am J Clin Nutr 2003;78(suppl):626S– 632S.
34. Levin N, Rattan J, Gilat T. Energy intake and body weight in ovo-lacto
vegetarians. J Clin Gastroenterol 1986;8:451–3.
35. Stubbs RJ, Prentice AM, James WP. Carbohydrates and energy balance.
Ann N Y Acad Sci 1997;819:44– 69.
36. Willett WC. Nutritional epidemiology. 2nd ed. New York: Oxford Uni-
versity Press, 1998.
37. Livingstone MB, Black AE. Markers of the validity of reported energy
intake. J Nutr 2003;133(suppl):895S–920S.
1274 NEWBY ET AL
by guest on June 7, 2013ajcn.nutrition.orgDownloaded from