Fruit and vegetable intakes and subsequent changes in body weight in

European populations: results from the project on Diet, Obesity, and

Genes (DiOGenes)14
Brian Buijsse, Edith JM Feskens, Matthias B Schulze, Nita G Forouhi, Nicholas J Wareham, Stephen Sharp, Domenico Palli,
Gianluca Tognon, Jytte Halkjaer, Anne Tjnneland, Marianne U Jakobsen, Kim Overvad, Daphne L van der A,
Huaidong Du, Thorkild IA Srensen, and Heiner Boeing

INTRODUCTION

Fruit and vegetables have a low energy density because of their

high content of water, low content of energy (1), and high content of
dietary fiber, which is considered to increase satiety and reduce
feelings of hunger (2). Finally, they also contain flavonoids, a group
of nonnutritive phytochemicals that may have antiobesity effects
(3). These properties make fruit and vegetables appealing foods to
be included in public health messages that combat obesity, either as
advice for preventing weight gain or in weight-reduction programs.
Dietary patterns characterized by high amounts of fruit and
vegetables relate to smaller weight gains and to a lower risk of
overweight and obesity in prospective studies conducted in adult

202

populations with different cultural backgrounds (412). However, not all studies are consistent (13), and, because these studies
focused on patterns of food group intake rather than on intakes of
single food groups, conclusions about whether fruit and vegetables per se slow down weight gain cannot be inferred directly
from these findings.
Cross-sectional studies that specifically addressed the relation of fruit and vegetable intake with body weight often yielded
null results (14). However, the temporality of this relation
cannot be addressed in cross-sectional studies because it is
possible that persons who have gained weight change their food
consumption patterns. In prospective studies, high intakes of
fruit (15, 16), vegetables (17), or both combined (8, 18, 19) were
related to smaller weight gains. However, some prospective studies
reported that fruit and vegetable intake was not associated with
change in body weight (20, 21) or waist circumference (22),
1
From the German Institute of Human Nutrition (DIfE) Potsdam-Rehbrucke, Department of Epidemiology, Nuthetal, Germany (BB, MBS, and
HB); the Division of Human Nutrition, Wageningen University, Wageningen, Netherlands (EJMF); the Department of Public Health Nutrition, Technical University Munich, Munich, Germany (MBS); the Medical Research
Council Epidemiology Unit, Elsie Widdowson Laboratories, Cambridge,
United Kingdom (NGF, NJW, and SS); the CSPO-Scientific Institute of
Tuscany, Molecular and Nutritional Epidemiology Unit, Florence, Italy
(DP and GT); the Danish Cancer Society, Institute of Cancer Epidemiology,
Copenhagen, Denmark (JH and AT); the Departments of Clinical Epidemiology (MUJ and KO) and of Cardiology (KO), Aarhus University Hospital,
Aalborg, Denmark; the National Institute for Public Health and the Environment, Bilthoven, Netherlands (HD and DLvdA); and the Institute of Preventive Medicine, Copenhagen University Hospital, Centre for Health and
Society, Copenhagen, Denmark (TIAS).
2
None of the sponsors of DiOGenes (Diet, Obesity, and Genes) had a role
in the study design, data collection, analysis, or interpretation of the data in
the written report or in the decision to submit for publication. This integrated
program was set up to target the issue of the obesity problem from a dietary
perspective, seeking new insights and new routes to prevention.
3
Supported by the European Committee, contract number FP6-513946.
The parties of the project are listed on the website of the project (http://www.
diogenes-eu.org/).
4
Address correspondence to B Buijsse, German Institute of Human Nutrition Potsdam-Rehbrucke, Department of Epidemiology Arthur-ScheunertAllee 114-116, 14558 Nuthetal, Germany. E-mail: brian.buijsse@dife.de.
Received December 19, 2008. Accepted for publication April 7, 2009.
First published online May 20, 2009; doi: 10.3945/ajcn.2008.27394.

Am J Clin Nutr 2009;90:2029. Printed in USA. 2009 American Society for Nutrition

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

ABSTRACT
Background: High fruit and vegetable intakes may limit weight
gain, particularly in susceptible persons, such as those who stop
smoking.
Objective: The objective was to assess the association of fruit and
vegetable intake with subsequent weight change in a large-scale
prospective study.
Design: The data used were from 89,432 men and women from 5
countries participating in the European Prospective Investigation
into Cancer and Nutrition (EPIC). The association between fruit
and vegetable intake and weight change after a mean follow-up of
6.5 y was assessed by linear regression. Polytomous logistic regression was used to evaluate whether fruit and vegetable intake relates
to weight gain, weight loss, or both.
Results: Per 100-g intake of fruit and vegetables, weight change
was 214 g/y (95% CI: 219, 29 g/y). In those who stopped smoking
during follow-up, this value was 237 g/y (95% CI: 258, 215 g/y;
P for interaction , 0.0001). When weight gain and loss were analyzed separately per 100-g intake of fruit and vegetables in a combined model, the odds ratios (95% CIs) were 0.97 (0.95, 0.98) for
weight gain 0.5 and ,1 kg/y, 0.94 (0.92, 0.96) for weight gain 1
kg/y, and 0.97 (0.95, 0.99) for weight loss 0.5 kg/y. In those who
stopped smoking during follow-up, the odds ratios (95% CIs) were
0.93 (0.88, 0.99), 0.87 (0.81, 0.92), and 0.97 (0.88, 1.07), respectively (P for interaction , 0.0001).
Conclusions: Fruit and vegetable intake relates significantly, albeit
weakly inversely, to weight change. For persons who stop smoking,
high fruit and vegetable intakes may be recommended to reduce the
risk of weight gain.
Am J Clin Nutr 2009;90:2029.

FRUIT AND VEGETABLE INTAKE AND WEIGHT CHANGE

which makes the overall view of the role of fruit and vegetable in
the prevention of weight gain less consistent.
In the present study, we assessed whether fruit and vegetable
intake is associated with subsequent changes in body weight
within a multicenter European prospective cohort study. Because
of our large sample size, we were able to evaluate whether the
relation between fruit and vegetable consumption and weight
change was stronger in situations in which persons are susceptible
to weight gain, with smoking cessation the most well-known
factor (2325).
SUBJECTS AND METHODS

Study population

tionnaires (FFQs) (27). In a random sample of 8% from each

EPIC cohort, dietary intakes were also assessed with a highly
standardized 24-h recall by using EPIC-SOFT (29). The data derived with this method were used to account for differences between national FFQs and to reduce potential measurement error
introduced by the FFQs (30, 31). Dietary intakes of fruit and
vegetables and other food groups from the 24-h recall were regressed on the FFQ values in a sex- and center-specific linear
calibration model while adjusting for age, body weight, height, and
season in which the FFQ was conducted (30, 32); these calibrated
values will be used throughout this article unless otherwise stated.
The total intake of fruit and vegetables (g/d) was considered
the main exposure. The food group fruit included all fresh fruit
and as well as dried fruit, canned fruit, and fruit compote.
Vegetables included leafy vegetables, fruiting vegetables, root
vegetables, cabbages, mushrooms, grain and pod vegetables,
onion and garlic, stalk vegetables and sprouts, legumes, mixed
salads, and mixed vegetables. Potatoes were not considered as
vegetables because they are often consumed as a starch component rather than as vegetables. Olives were not considered as
fruit because of their fat content. Finally, fruit and vegetable
juices as well as tomato sauces were not included in the fruit and
vegetable groups because of differences in food matrixes and
lower contents of dietary fiber. Energy intake was calculated by
using national food composition tables that were standardized in
the EPIC Nutrient Database (33).
Assessment of anthropometric measures and weight change
Body weight and height were measured at baseline by trained
staff in all 6 cohorts, with participants wearing no shoes and
either light indoor clothing (UK-Nor, NL-Doe, and NL-AmMa)
or underwear (IT-Flo, GER-Pot, and DK-CopAa). Anthropometric measurements at follow-up were either performed by
trained staff following the same protocol as during baseline
examinations (UK-Nor and NL-Doe) or were self-reported by the
participants in the follow-up questionnaire (IT-Flo, NL-AmMa,
GER-Pot, and DK-CopAa).
Differences in body weight measurement across the different
cohorts were corrected by subtracting 1 kg from a persons
weight in case of measurement in light indoor clothing (34). Selfreported body weight at follow-up was corrected for potential
underreporting by using a sample of a health-conscious UK
cohort among whom both measured and self-reported body
weights were known (34). Specifically, measured body weight
was predicted from self-reported weight, age, and sex by using
linear regression models. BMI was calculated by dividing weight
(kg) by the square of height (m).
The outcome in the current analysis was change in body
weight. The mean duration of follow-up between cohorts ranged
from 3.7 y in UK-Nor to 10.0 y in NL-AmMa. Because of this
difference, the annual change in body weight was studied. This
was calculated for each participant by subtracting baseline body
weight from follow-up body weight and dividing the difference in
grams by the duration of follow-up in years. Thus, negative values
indicated weight loss and positive values indicated weight gain.

Consumption of fruit and vegetables at baseline

Assessment of covariate data

Habitual dietary intakes over the past year were assessed at

baseline by means of country-specific food-frequency ques-

In each center, occupational and leisure-time physical activity

were assessed by questionnaire (27, 35). Occupational physical

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

The project on Diet, Obesity, and Genes (DiOGenes) is a

multidisciplinary European project that focuses on dietary
components and genetic and behavioral factors in the prevention of weight gain (26). This integrated program was set up to
target the issue of the obesity problem from a dietary perspective, seeking new insights and new routes to prevention. The
studies performed within DiOGenes use either an experimental
or an observational design. The current analysis is an observational study and uses data from 5 countries participating within
the European Prospective Investigation into Cancer and Nutrition
(EPIC) (27), ie, Denmark (Aarhus and Copenhagen, n = 39,909;
DK-CopAa), Germany (Potsdam, n = 16,307; GER-Pot), United
Kingdom (Norfolk, n = 12,808; UK-Nor), Italy (Florence, n =
9297; IT-Flo), and Netherlands (Doetinchem, Amsterdam, and
Maastricht). Because of differences in the assessment of followup anthropometric measures in the Netherlands, the cohort of
Doetinchem (n = 4200; NL-Doe) and the cohort of Amsterdam/
Maastricht (n = 6911; NL-AmMa) were treated as separate cohorts; thus, a total of 6 cohorts were included in this study.
Of the 146,543 initial participants who attended the baseline
examination in 19921998, 102,346 (69.8%) participated at
follow-up. We excluded women who were pregnant at either
baseline or follow-up (n = 133), participants who provided no
information on baseline diet (n = 113), who were in the centerspecific top 1% and bottom 1% of the distribution of the ratio
of reported energy intake over estimated energy requirement
[basal metabolic rate calculated from age, sex, body weight
and height (28) 1.55; n = 1803], who had no information on
anthropometric measures at baseline (n = 964), or who had no
follow-up information on anthropometric measures or followup time (n = 1058). Furthermore, we excluded those with
unrealistic anthropometric measures at either baseline or followup (n = 165), ie, who had heights ,1.3 m, body mass indexes
(BMIs; in kg/m2) ,16, waist circumferences ,40 or .160 cm,
or the combination of waist circumferences ,60 cm and BMIs
.25. Participants who reported extreme changes in annual weight
or waist circumference (ie, .5 kg or .7 cm annually, respectively) were also excluded (n = 166). We finally excluded
those with baseline cardiovascular disease, diabetes mellitus, or
cancer (n = 8512). The current analysis is therefore based on
89,432 participants, 52,307 of whom (58%) were women.

203

204

BUIJSSE ET AL

Statistical analysis
We first evaluated the shape of the association of fruit and
vegetable intake with weight change by using spline regression
with knots placed at quartiles of the distribution (37). After
confirming a linear relation, we chose to analyze fruit and vegetable intake as a continuous variable.
The association of fruit and vegetable intake with weight change
was first assessed by linear regression methods. The difference
in weight change and corresponding 95% CI by intake of fruit
and vegetables was estimated with annual weight change as the
dependent variable. In polytomous logistic regression models
(PROC CATMOD; SAS Institute Inc, Cary, NC), annual weight
change was classified as minor weight gain (0.5 to ,1 kg/y),
major weight gain (1 kg/y), weight loss (0.5 kg/y), or weight
stable ( 20.5 to ,0.5 kg/y weight change). Odds ratios and
their 95% CIs of weight gain and weight loss were calculated
per 100-g intake of fruit and vegetables per day.
All analyses were adjusted for age (continuous), sex, cohort (5
dummy variables with DK-CopAa as the reference), duration of
follow-up (years), and baseline weight (continuous). In multivariable-adjusted models, we also included level of education
(dummy variables for none/primary school, secondary school,
technical/professional school, missing values; university degree =
reference), height at baseline (continuous), smoking behavior
(dummy variables for stable smokers, starters, quitters, and unknown smoking status; nonsmokers = reference), physical activity (inactive, moderately active, active, missing; moderately
inactive = reference), alcohol intake (dummy variables for 0,
.1830, .3060, and .60 g/d; because of higher intakes in
men, intakes .60 g/d in this group was split as .6096 and

.96 g/d; .018 g/d = reference), and, in women, postmenopausal status and postmenopausal hormone use (both yes
or no). In additional models we also adjusted for total energy
intake (continuous).
Stratified analyses were conducted by sex, cohort, age, smoking
behavior, and follow-up duration (DiOGenes-wide median split
at 5.5 y). Effect modification was statistically tested by entering
cross-product [fruit and vegetable intake times sex, cohort in
categories, age in 2 categories, BMI in 2 categories, follow-up
duration in 2 categories (median split), and smoking in categories] terms along with main effects in the multivariable model
(model C). We observed effect modification by cohort in the
linear regression analysis (P , 0.0001), but not in the polytomous logistic regression analysis (P = 0.25). The effect modification was due to a positive association in the UK-Nor cohort
[25 g/y weight change per-100 g fruit and vegetable (95% CI: 3
47 g/y) in adjusted analyses], whereas nonsignificant or inverse
associations were observed in the other cohorts. After the UKNor data were omitted from the linear regression analysis, the
effect modification by cohort was no longer evident (P = 0.12).
Because we could not explain why the relation in UK-Nor deviated from that of other cohorts, the interaction of this cohort
was taken into account in the linear regression analyses by entering a cross-product term between UK-Nor (dummy variable
with value 1 indicating UK-Nor, value 0 if not) and fruit and
vegetable intake (continuous) along with main effects.
Finally, we assessed the effect of excluding participants with
incident cardiovascular diseases, diabetes, or cancer during followup as well as those with missing data on covariates. All analyses
were conducted by using SAS software (version 9.1; SAS Institute Inc).
RESULTS

Baseline characteristics and weight change

The median intake of fruit and vegetables was 324 g/d in men
and 377 g/d in women. Men and women in IT-Flo had the highest
median intake (600 and 515 g/d, respectively), whereas men and
women in NL-AmMa had the lowest intake (median for both:
265 g/d). Those with higher fruit and vegetable intakes were
less likely to smoke, had lower intakes of alcohol, and higher
intakes of calories from carbohydrates (Table 1). Women with
high fruit and vegetable intakes were more likely to engage in
moderate physical activity.
After differences in the assessment of weight between cohorts
were adjusted for, men and women gained weight over time in all
cohorts, with an overall mean weight change of 330 g/y. In men, it
ranged from 330 g/y in IT-Flo to 439 g/y in NL-AmMa and
in women from 208 g/y in DK-CopAa to 447 g/y in NL-Doe.
Participants who stopped smoking during follow-up (6.6% of the
study population) gained on average the most weight (mean
weight change adjusted for age, sex, cohort, and years of followup: 750 g/y; 95% CI: 729, 771 g/y), whereas participants who
started smoking (1.6%) had the smallest average weight gain
(adjusted mean weight change: 50 g/y; 95% CI: 10, 90 g/y).
Fruit and vegetable intakes and weight change
Spline regression was used to assess the shape of the association of fruit and vegetable intake with annual weight change

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

activity was classified as sedentary occupation, standing occupation, manual work, heavy manual work, unemployed, or missing.
Type and duration of different leisure-time physical activities
were ascertained at baseline, and reported activities included
walking, cycling, gardening, sports, household work, home repair, stair climbing, and vigorous recreational activity. Each type
of activity was converted into metabolic equivalent-hours (MET-h)
(36). A variable of total physical activity was created in which
participants were cross-classified on the basis of sex-specific quartiles of leisure-time activity and on categories of occupational work.
This variable was coded as inactive, moderately inactive, moderately active, and active. Missing values were assigned to 3319
participants (3.7%).
Information on smoking habits was collected by questionnaire
at baseline and follow-up. Smoking behavior was defined in 4
categories: nonsmokers (never smokers and those who stopped
smoking before baseline), stable smokers (those who reported
smoking at baseline and follow-up), stopped smokers (those who
stopped smoking during follow-up), and started smokers (those
who started smoking during follow-up). Smoking status was
unknown for 1440 participants (1.6%).
Information on highest achieved education [missing values
were assigned for 1579 participants (1.8%)], alcohol consumption, and, in women, postmenopausal status and postmenopausal
hormone use was provided by questionnaire at baseline. Similarly, the prevalence and incidence of diabetes, cardiovascular
disease, and cancer were assessed by questionnaire, interview,
and/or hospital discharge information.

205

FRUIT AND VEGETABLE INTAKE AND WEIGHT CHANGE

TABLE 1
Characteristics of the study population by sex and consumption by quintile (Q) of fruit and vegetable intake1
Fruit and vegetable intake
Men and women
(n = 89,432)
Characteristic

Women
(n = 52,307)

Q1

Q3

Q5

Q1

Q3

Q5

Q1

Q3

Q5

17,886
147
(60, 265)
52.6 6 9.13

17,886
307
(176, 460)
53.5 6 8.7

17,886
510
(331, 731)
53.1 6 7.7

7425
115
(43, 229)
53.9 6 7.9

7425
284
(155, 448)
53.6 6 8.5

7425
466
(282, 662)
53.8 6 7.7

10,461
168
(76, 290)
51.3 6 9.8

10,461
332
(200, 500)
53.5 6 8.9

10,461
535
(356, 770)
52.9 6 7.7

28.9

23.5

31.4

32.7

21.6

26.8

26.5

25.2

32.9

38.8

37.9

27.0

35.3

32.8

23.4

44.2

41.2

28.7

12.3
18.5
20 (11, 37)

10.9
25.1
14 (7, 24)

17.3
23.6
10 (4, 19)

9.1
21.1
27 (17; 47)

9.9
33.7
23 (15, 37)

13.9
34.6
21 (12, 32)

16.1
12.1
12 (7, 20)

12.1
18.5
10 (5, 17)

18.2
19.7
8 (3, 14)

56.5
31.4
8.9
2.2

76.0
14.8
5.6
1.6

77.4
13.2
6.3
1.6

54.1
33.9
9.1
2.2

71.3
18.0
7.0
2.1

77.1
13.3
6.0
1.8

59.5
29.3
8.2
1.8

77.8
13.7
4.8
1.5

77.7
13.0
6.5
1.4

25.3
26.9
31.2
10.9
53.9

21.4
28.3
35.7
10.6
58.9

19.1
26.6
42.7
10.8
58.7

26.8
26.4
31.8
11.1
NA

26.6
27.2
29.8
11.6
NA

28.1
29.2
30.8
10.1
NA

22.1
28.5
31.8
10.6
53.9

18.0
28.5
39.7
10.4
58.9

14.5
24.9
48.7
11.3
58.7

22.8
67.2 6 11.5
163.5 6 6.3
25.1 6 4.1
80.4 6 10.5

18.5
66.8 6 11.5
162.7 6 6.4
25.2 6 4.1
79.1 6 10.4

21.3
22.8
75.9 6 13.3 72.9 6 13.4
171.1 6 8.9 168.5 6 9.1
25.9 6 3.8 25.6 6 3.8
89.8 6 12.0 85.9 6 12.1

2273
36.3
15.0
40.2

6
6
6
6

18.5
NA
71.1 6 13.3 80.6 6 11.8
166.5 6 8.8 175.6 6 6.6
25.6 6 3.9
26.1 6 3.4
83.4 6 12.1 94.7 6 9.8

441 2145 6 459 2150 6 438

3.2 35.8 6 3.7 34.3 6 3.4
1.1 15.1 6 1.3 15.5 6 1.6
4.2 43.2 6 4.0 46.0 6 3.5

20 6 5

22 6 5

26 6 5

2538
36.6
14.6
38.8

6
6
6
6

NA
81.5 6 11.4
176.5 6 6.7
26.1 6 3.2
94.2 6 9.4

NA
21.3
81.6 6 11.5 67.8 6 11.5
175.7 6 6.8 163.6 6 6.5
26.4 6 3.3
25.4 6 4.3
93.8 6 9.6
81.7 6 10.9

275 2610 6 313 2681 6 306

3.3 37.0 6 3.5 35.4 6 4.4
1.0 14.7 6 1.1 15.1 6 1.4
4.1 40.8 6 3.8 43.6 6 4.1

21 6 4

25 6 5

29 6 5

1835
35.6
15.4
42.3

6
6
6
6

233 1865 6 231

3.1 35.0 6 3.3
1.2 15.2 6 1.4
3.9 44.9 6 3.4

18 6 4

21 6 4

1929
34.1
15.6
46.8

6
6
6
6

234
2.8
1.6
3.0

24 6 4.8

NA, not applicable; HRT, hormone replacement therapy. P values for differences in characteristics across quintiles of fruit and vegetable intake were
based on chi-square tests for discrete variables or, for continuous variables, on modeling median values as one continuous variable in linear regression
analyses; because of the large sample size, all P values were significant (P , 0.05), with the exception of age in the total study population.
2
Median unadjusted values (25th percentile, 75th percentile) based on 24-h dietary recalls in a subsample of n = 6790.
3
Mean 6 SD (all such values).
4
Percentages do not add up to 100 because of missing values.
5
Median values (25th percentile; 75th percentile) based on calibrated intake values.

(Figure 1). Although the spline indicated a weaker relation toward higher intakes of fruits and vegetables, the nonlinearity
was not statistically significant (P = 0.72) and led us to conclude
that the assumption of a linear effect of fruit and vegetable intake is acceptable.
Fruit and vegetable intake was weakly inversely associated
with weight change (Table 2). Additional adjustment for baseline body weight compared with the adjusted model with age,
sex, cohort, and years of follow-up attenuated the association
particularly for studying vegetable intake. Because the mean

weight change was positive, which indicated that participants

weight increased over time, the results of the multivariable-adjusted model indicated a 14-g/y smaller weight gain per difference in 100 g fruit and vegetables (95% CI: 219, 29 g/y).
After the customary level of statistical significance of 5% was
applied, the association between fruit and vegetable intake and
weight change differed by sex, age, BMI, smoking behavior, and
length of follow-up (Table 3). However, the adjusted mean
weight change was similar for strata of most of these variables.
The largest difference was observed between strata of smoking

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

No. of subjects
Fruit and vegetable
intake (g/d)2
Age (y)
Education (%)4
No school degree or
primary school
Technical or
professional school
Secondary school
University degree
Alcohol intake (g/d)5
Smoking behavior (%)4
Nonsmokers
Stable smokers
Stopped smokers
Started smokers
Total physical
activity (%)4
Inactive
Moderately inactive
Moderately active
Active
Postmenopausal
(% women)
HRT (% women)
Weight (kg)
Height (cm)
BMI (kg/m2)
Waist circumference
(cm)
Daily dietary intake
Total energy (kcal)
Fat (% of energy)
Protein (% of energy)
Carbohydrates (%
of energy)
Fiber (g)

Men
(n = 37,125)

206

BUIJSSE ET AL

behavior. The strongest inverse relation was seen in those who

stopped smoking during follow-up, whereas this was less pronounced in nonsmokers and no association was seen in stable
smokers and those who started smoking during follow-up.
Analysis of observed rather than calibrated fruit and vegetable
intakes resulted in weaker associations. Per 100 g of observed
fruit and vegetable intakes, weight change was 28 g/y (95% CI:
212, 24 g/y) after multivariable adjustments (model C from
Table 2). For the separate intake of fruit and vegetables, these
estimates were 28 (95% CI: 212 to 24) and 211 (95% CI:
218 to 25) g/y, respectively. Multivariable-adjusted weight

TABLE 2
Association between fruit and vegetable intake, combined and separately,
and subsequent weight change in linear regression analyses1
Weight change (95% CI)

Model A2
Model B3
Model C4

Fruit and vegetables

Fruit

Vegetables

215 (220, 210)

212 (217, 27)
214 (219, 29)

g/y
217 (223, 211)
214 (220, 28)
216 (222, 210)

230 (244, 216)

219 (233, 25)
219 (233, 24)

1
Values are coefficients for the linear regression of weight change (g/y)
on fruit and vegetable intake (per 100-g/d increase in intake) in the total
study population (n = 89,432).
2
Adjusted for age, sex, cohort, product term UK-Nor (United Kingdom, Norfolk cohort) fruit/vegetable intake (for details see text), and years
of follow-up (continuous).
3
As for model A, with additional adjustment for baseline weight (continuous).
4
As for model B, with additional adjustment for baseline height (continuous), change in smoking status (dummy variables), baseline total physical activity (dummy variables), education (dummy variables), alcohol intake
(dummy variables), and, in women, postmenopausal status (yes or no) and
postmenopausal hormone use (yes or no).

DISCUSSION

In this large-scale prospective cohort study, fruit and vegetable

intake was significantly, albeit weakly, related to smaller weight
gains. Our findings suggest that, within the range of fruit and
vegetable intake we observed, an increase in fruit and vegetable
intake slows down weight gain to a small extent. Those who
stopped smoking may benefit the most from a high fruit and
vegetable intake.
The strengths of the present study include its large size, its
prospective design, and the wide variation in fruit and vegetable
intake and weight change, which made it possible to detect weak
associations and to perform stratified analyses with only losing
a small proportion of statistical power.
This study also has limitations that warrant attention. First, the
assessment of body weight at baseline and follow-up differed
across centers. Although at baseline all participants were weighed
at the centers by trained staff, local study protocols about the
participants clothing differed between cohorts. At follow-up,
participants were weighed at the study site in 2 cohorts, whereas
in other cohorts participants self-reported their weight in the
follow-up questionnaire. We corrected for these differences in

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

FIGURE 1. Spline model for describing the shape of the association of

the intake of fruit and vegetables with weight change. The fifth percentile of
fruit and vegetable intake was set as the reference. P for nonlinearity = 0.72.
Adjusted for age (continuous), sex, cohort (dummy variables), years of
follow-up (continuous), baseline weight (continuous), baseline height
(continuous), change in smoking status (dummy variables), baseline total
physical activity (dummy variables), education (dummy variables), alcohol
intake (dummy variables), and, in women, postmenopausal status (yes or no)
and postmenopausal hormone use (yes or no).

change in participants who stopped smoking during follow-up

was 220 g/y (95% CI: 233, 27 g/y) per 100 g of observed
intake of fruit and vegetables.
After participants with incident chronic diseases were excluded, the relation became somewhat stronger: per 100-g increase in calibrated fruit and vegetables, adjusted weight change
(model C) in the total study population was on average 215 g/y
(95% CI: 220, 210 g/y). In those who stopped smoking, this
estimate was 246 (95% CI: 268, 223 g/y).
We also evaluated the role of fruit and vegetable intake on
weight gain and weight loss separately. Weight change was
categorized as minor weight gain (0.5 to ,1 kg/y), major
weight gain (1 kg/y) weight loss (0.5 kg/y), or weight stable
(reference, weight change . 20.5 to ,0.5 kg/y). Adjusted odds
ratios for weight gain and loss were calculated in a combined
model by using polytomous logistic regression analysis (Table 4).
The percentage of those with a minor weight gain was 23%,
those with a major weight gain was 15%, those who lost weight
was 11%, and those who had a stable weight was 51%. After
multivariable adjustments, a daily increase of 100 g fruit and
vegetables was related to slightly lower odds for weight gain
and, unexpectedly, weight loss. This relation was again most
pronounced in those who stopped smoking during follow-up,
with odds ratios (95% CI) of 0.93 (0.88, 0.99) for minor weight
gain, of 0.87 (0.81, 0.92) for major weight gain, and of 0.97
(0.88, 1.07) for weight loss. Nonsmokers and stable smokers
also had a significantly lower odds of weight gain, and stable
smokers also had a significantly lower odds of weight loss,
whereas started smokers had higher odds of weight loss.
Finally, all analyses were repeated after participants with
missing information on either physical activity, smoking behavior, education, or hormone replacement therapy were excluded (n = 6189). We also evaluated whether our findings
changed when weight change was calculated from uncorrected
body weights. Finally, we adjusted all analyses for total energy
intake. However, these sensitivity analyses all yielded essentially similar results (data not shown).

207

FRUIT AND VEGETABLE INTAKE AND WEIGHT CHANGE

TABLE 3
Association between fruit and vegetable intake and subsequent annual weight change in stratified, multivariable-adjusted
linear regression analyses1
Variable

Weight change
(95% CI)3

P for
interaction

g/y

g/y

358
333

27 (215, 1)
212 (218, 25)

0.02

346
354

215 (221, 29)

213 (225, 21)

0.05

329
368

211 (216, 25)

217 (225, 29)

0.01

351
345

215 (222, 29)

214 (222, 26)

,0.0001

387
288
306
356

213
25
237
24

,0.0001

(219, 27)
(217, 6)
(258, 215)
(243, 35)

Values are coefficients of the linear regression of weight change (g/y) on fruit and vegetable intake (per 100-g/
d increase in intake) stratified by sex, baseline age, follow-up duration, baseline BMI, and smoking behavior.
2
Mean fruit and vegetable intakes based on 24-h dietary recalls in a subsample of n = 6790, adjusted for cohort and
(mutually for) age, sex, follow-up duration, BMI, and smoking.
3
Adjusted for age, sex, cohort, product term UK-Nor (United Kingdom, Norfolk cohort) fruit/vegetable intake (for
details see text), years of follow-up (continuous), baseline weight (continuous), baseline height (continuous), change in
smoking status (where applicable dummy variables), baseline total physical activity (dummy variables), education (dummy
variables), alcohol intake (dummy variables), and, in women, postmenopausal status (yes or no) and postmenopausal
hormone use (yes or no).

methods by cohort, although the results from analyses before and

after these corrections were essentially similar. Second, we found
heterogeneity in results across cohorts, which was due to one
cohort in particular. One explanation may be that the mean
duration of follow-up in this cohort (3.7 y, UK-Nor) was too short
to get an accurate figure for long-term weight change, which is
also supported by data from other prospective studies (see below).
From the EPIC-Potsdam cohort, it is known that the strongest

predictors for short-term weight gain are prior weight loss and
prior weight cycling, whereas prior weight gain was not predictive (38). Third, we only relied on baseline fruit and vegetable consumption and therefore assume a stable consumption
over time. Finally, we cannot exclude the possibility that the
association between fruit and vegetable intake and weight change
is partly explained by confounding factors that were incompletely
accounted for.

TABLE 4
Multivariable-adjusted odds ratios (95% CIs) for gaining and losing weight per 100-g increase in fruit and vegetable intake in polytomous logistic regression
analysis according to smoking status1
Odds ratio (95% CI)

All (n = 89,432)3
Smoking behavior
Nonsmokers (n = 63,675)
Stable smokers (n = 16,837)
Stopped smokers (n = 5916)
Started smokers (n = 1564)

Weight stable
 20.5 to ,0.5 kg/y

Weight gain
0.5 to ,1.0 kg/y

Weight gain
1.0 kg/y

Weight loss
0.5 kg/y

0.97 (0.95, 0.98)

0.94 (0.92, 0.96)

0.97 (0.95, 0.99)

1
1
1
1

0.97
0.96
0.93
1.07

0.97
0.95
0.87
1.15

0.99
0.95
0.97
1.13

(0.95,
(0.93,
(0.88,
(0.94,

0.99)
1.00)
0.99)
1.22)

(0.95,
(0.90,
(0.81,
(0.99,

1.00)
1.00)
0.92)
1.34)

(0.97,
(0.91,
(0.88,
(1.02,

1.02)
0.99)
1.07)
1.26)

P for interaction2

,0.0001

Adjusted for age, sex, cohort, product term UK-Nor (United Kingdom, Norfolk cohort) fruit/vegetable intake (for details, see text), years of follow-up
(continuous), baseline weight (continuous), baseline height (continuous), change in smoking status (where applicable dummy variables), baseline total
physical activity (dummy variables), education (dummy variables), alcohol intake (dummy variables), and, in women, postmenopausal status (yes or no)
and postmenopausal hormone use (yes or no).
2
P for interaction based on modeling smoking behavior as a categorical variable fruit and vegetable intake in participants with known information on
smoking (n = 87,992).
3
Includes also 1440 participants with an unknown smoking status.
1

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

Sex
Men (n = 37,125)
Women (n = 52,307)
Baseline age
,60 y (n = 69,353)
60 y (n = 20,079)
Follow-up duration
,5.5 y (n = 43,235)
5.5 y (n = 46,197)
Baseline BMI
,25 kg/m2 (n = 42,819)
25 kg/m2 (n = 46,613)
Smoking behavior
Nonsmokers (n = 63,675)
Stable smokers (n = 16,837)
Stopped smokers n = 5916)
Started smokers (n = 1564)

Mean fruit and

vegetable intake2

208

BUIJSSE ET AL

was designed to study the effects of a diet low in fat and high in
fruit, vegetables, and whole grain on cancer, and the PPT was
designed to test the effects of a high-fiber, high-fruit-and-vegetable, low-fat diet on adenomatous polyp recurrence. In the WHI,
women in the intervention group did not gain weight over
a period of 7.5 y compared with controls (40), whereas the intervention group in the PPT had, on average, a 0.7-kg lower body
weight after 4 y compared with a 0.3-kg weight gain in controls
(41). In summary, these findings suggest a weak effect of an
increased fruit and vegetable intake on weight control.
Our findings suggest that persons who quit smoking are likely
to benefit the most from an increased intake of fruit and vegetables. No other prospective studies have reported a similar result
(8, 1521). Also, a biological plausible explanation why stopped
smokers would specifically benefit from an increased fruit and
vegetable intake is lacking; it might be that other groups at increased risk of weight gain also benefit more from higher fruit and
vegetable intakes. However, if our finding is confirmed, it may
have important implications because smoking cessation is accompanied by weight gain (2325), which may prevent smokers
from quitting. Increasing the intake of fruit and vegetables is one
strategy for limiting weight gain in those who stop smoking.
In conclusion, although fruit and vegetable intake was weakly
related with weight change in the present study, the findings have
public health relevance and support initiatives to increase the
intake of fruit and vegetables. Our findings are also of relevance
to the massive smoking cessation initiatives in Europe, which
suggest that adding a fruit and vegetable component to programs
directed at stopping cigarette smoking may help to limit weight
gain in those who stop smoking.
The authors responsibilities were as followsBB: performed the data
analysis and wrote the manuscript; MBS, HB, EJMF, and TIAS: contributed
to the conception and design of the study; HB: helped with the interpretation
and contributed to the writing of the manuscript; SS: provided statistical expertise and gave critical comments on the manuscript; EJMF, MBS, NGF,
NJW, SS, DP, GT, JH, AT, MUJ, KO, DLvdA, HD, and TIAS: helped with
the interpretation of the results and gave critical comments on the manuscript;
and EJMF, DP, NJW, AT, KO, TIAS, and HB: contributed to providing funding. TIAS has industrial collaborations as listed at http://www.ipm.regionh.
dk/person/tias/Disclosures.html. None of the other authors declared a conflict
of interest.

REFERENCES
1. Drewnowski A, Almiron-Roig E, Marmonier C, Lluch A. Dietary energy density and body weight: is there a relationship? Nutr Rev 2004;62:
40313.
2. Rolls BJ, Ello-Martin JA, Tohill BC. What can intervention studies tell
us about the relationship between fruit and vegetable consumption and
weight management? Nutr Rev 2004;62:117.
3. Hughes LA, Arts IC, Ambergen T, et al. Higher dietary flavone, flavonol, and catechin intakes are associated with less of an increase in BMI
over time in women: a longitudinal analysis from the Netherlands Cohort Study. Am J Clin Nutr 2008;88:134152.
4. Quatromoni PA, Copenhafer DL, DAgostino RB, Millen BE. Dietary
patterns predict the development of overweight in women: The Framingham Nutrition Studies. J Am Diet Assoc 2002;102:123946.
5. Newby PK, Muller D, Hallfrisch J, Qiao N, Andres R, Tucker KL.
Dietary patterns and changes in body mass index and waist circumference in adults. Am J Clin Nutr 2003;77:141725.
6. Schulz M, Nothlings U, Hoffmann K, Bergmann MM, Boeing H.
Identification of a food pattern characterized by high-fiber and low-fat
food choices associated with low prospective weight change in the
EPIC-Potsdam cohort. J Nutr 2005;135:11839.
7. Newby PK, Weismayer C, Akesson A, Tucker KL, Wolk A. Longitudinal changes in food patterns predict changes in weight and body mass

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

In the present study, a higher fruit and vegetable intake was

weakly related to a smaller weight gain, which, overall, suggested
a 14-g smaller weight gain per year for a daily higher intake of
100 g fruit and vegetables. This seems to not be a particularly
large benefit. Considering a difference of 350 g/d between the
highest and lowest quintiles of fruit and vegetables, low fruit and
vegetables intakes would roughly account for 17% of the
overall weight change of 294 g/y observed in those with low
intakes.
Note that we found a greater reduction in weight gain in relation to fruit and vegetable intake among those who stopped
smoking during follow-up, which suggests a 37 g smaller weight
per year for each 100-g fruit and vegetable intake per day. As
expected (2325), this group also gained most weight over time.
Considering a 350-g/d mean difference between the highest and
lowest quintiles of fruit and vegetable intake and an average
weight change of 815 g/y in stopped smokers in the lowest
quintile of fruit and vegetable intake, low intakes would still
account for 16% of the weight gain in this group. Although the
benefit from fruit and vegetables according to our findings seems
small, they still have public health relevance. Assuming a stable
increase in body weight in stopped smokers, an increase of 2
portions of 80 g fruit and vegetables per day would lead to
0.5-kg smaller weight gain over a 10-y period in this group.
Other prospective cohort studies that reported on the relation
between fruit and vegetable intake and weight change varied by
how weight change was defined, their statistical methods, and
study populations. Also, there was large variation in the duration
of follow-up across studies. In nonsmokers of the EPIC-Potsdam
Study (21), who partly also contribute to the present study, fruit
and vegetable intakes were not related to weight change after 2 y
of follow-up. Likewise, in British adults, no association was seen
for weight change after 4 y (20). However, studies with a longer
duration of follow-up reported inverse relations of fruit and
vegetable intake with weight change (8, 1519). Their results
suggest smaller weight gains with either higher fruit (15, 16),
vegetable (17), or combined fruit and vegetable (8, 18, 19)
intakes. In the Nurses Health Study, the second largest study
after the present study, women who reported the largest increase
in fruit and vegetable intake over a 10-y period (mean increase:
1.9 portions) had a 24% lower risk of developing obesity than did
those who reported a decrease of, on average, 1.3 portions (18).
Taken together, the mixed results from prospective studies may
be due to differences in follow-up duration. Studies with a followup duration longer than 4 y showed inverse relations.
An important question is whether a higher fruit and vegetable
intake by itself, ie, without any other intervention, lowers ad
libitum energy intake by an amount that results in smaller weight
gains. Intervention studies that assessed the effects of higher fruit
and vegetable intakes without specifically aiming to lower energy
intake and reduce weight gain are limited (2). The only randomized controlled trial that specifically aimed to increase fruit
and vegetable intakes without any other concurrent intervention
tested the effects on blood pressure (39). Although fruit and
vegetable intakes increased more in the intervention than in the
control group, body weights remained the same in both groups
after 6 mo (39). Two other randomized intervention studies that
aimed to increase fruit and vegetable intake, but also intervened
in other dietary factors, were the Womens Health Initiative
(WHI) (40) and the Polyp Prevention Trial (PPT) (41). The WHI

FRUIT AND VEGETABLE INTAKE AND WEIGHT CHANGE

8.

9.
10.
11.
12.
13.
14.
15.

17.
18.
19.
20.

21.
22.
23.
24.

25. John U, Hanke M, Rumpf HJ, Thyrian JR. Smoking status, cigarettes per
day, and their relationship to overweight and obesity among former and
current smokers in a national adult general population sample. Int J
Obes (Lond) 2005;29:128994.
26. Saris WH, Harper A. DiOGenes: a multidisciplinary offensive focused
on the obesity epidemic. Obes Rev 2005;6:1756.
27. Riboli E, Hunt KJ, Slimani N, et al. European Prospective Investigation
into Cancer and Nutrition (EPIC): study populations and data collection.
Public Health Nutr 2002;5:111324.
28. Food and Agriculture Organization/World Health Organization/United
Nations University. Energy and protein requirements. World Health
Organ Tech Rep Ser 1985;724.
29. Slimani N, Kaaks R, Ferrari P, et al. European Prospective Investigation
into Cancer and Nutrition (EPIC) calibration study: rationale, design and
population characteristics. Public Health Nutr 2002;5:112545.
30. Rosner B, Gore R. Measurement error correction in nutritional epidemiology based on individual foods, with application to the relation of
diet to breast cancer. Am J Epidemiol 2001;154:82735.
31. Ferrari P, Kaaks R, Fahey MT, et al. Within- and between-cohort variation in measured macronutrient intakes, taking account of measurement
errors, in the European Prospective Investigation into Cancer and Nutrition study. Am J Epidemiol 2004;160:81422.
32. Kaaks R, Riboli E, van Staveren W. Calibration of dietary intake
measurements in prospective cohort studies. Am J Epidemiol 1995;142:
54856.
33. Slimani N, Deharveng G, Unwin I, et al. The EPIC nutrient database
project (ENDB): a first attempt to standardize nutrient databases across
the 10 European countries participating in the EPIC study. Eur J Clin
Nutr 2007;61:103756.
34. Haftenberger M, Lahmann PH, Panico S, et al. Overweight, obesity and
fat distribution in 50- to 64-year-old participants in the European Prospective Investigation into Cancer and Nutrition (EPIC). Public Health
Nutr 2002;5:114762.
35. Haftenberger M, Schuit AJ, Tormo MJ, et al. Physical activity of subjects
aged 50-64 years involved in the European Prospective Investigation into
Cancer and Nutrition (EPIC). Public Health Nutr 2002;5:116376.
36. Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical
activities: classification of energy costs of human physical activities.
Med Sci Sports Exerc 1993;25:7180.
37. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 1995;6:35665.
38. Kroke A, Liese AD, Schulz M, et al. Recent weight changes and weight
cycling as predictors of subsequent two year weight change in a middleaged cohort. Int J Obes Relat Metab Disord 2002;26:4039.
39. John JH, Ziebland S, Yudkin P, Roe LS, Neil HA. Effects of fruit and
vegetable consumption on plasma antioxidant concentrations and blood
pressure: a randomised controlled trial. Lancet 2002;359:196974.
40. Howard BV, Manson JE, Stefanick ML, et al. Low-fat dietary pattern
and weight change over 7 years: the Womens Health Initiative Dietary
Modification Trial. JAMA 2006;295:3949.
41. Lanza E, Schatzkin A, Daston C, et al. Implementation of a 4-y,
high-fiber, high-fruit-and-vegetable, low-fat dietary intervention: results
of dietary changes in the Polyp Prevention Trial. Am J Clin Nutr 2001;
74:387401.

Downloaded from ajcn.nutrition.org by guest on April 19, 2016

16.

index and the effects are greatest in obese women. J Nutr 2006;136:
25807.
Sanchez-Villegas A, Bes-Rastrollo M, Martinez-Gonzalez MA,
Serra-Majem L. Adherence to a Mediterranean dietary pattern and
weight gain in a follow-up study: the SUN cohort. Int J Obes (Lond)
2006;30:3508.
Schulze MB, Fung TT, Manson JE, Willett WC, Hu FB. Dietary patterns and
changes in body weight in women. Obesity (Silver Spring) 2006;14:144453.
McNaughton SA, Mishra GD, Stephen AM, Wadsworth ME. Dietary patterns throughout adult life are associated with body mass index, waist circumference, blood pressure, and red cell folate. J Nutr 2007;137:99105.
Okubo H, Sasaki S, Murakami K, et al. Three major dietary patterns are
all independently related to the risk of obesity among 3760 Japanese
women aged 18-20 years. Int J Obes (Lond) 2008;32:5419.
Esmaillzadeh A, Azadbakht L. Major dietary patterns in relation to
general obesity and central adiposity among Iranian women. J Nutr
2008;138:35863.
Togo P, Osler M, Sorensen TI, Heitmann BL. A longitudinal study of
food intake patterns and obesity in adult Danish men and women. Int J
Obes Relat Metab Disord 2004;28:58393.
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;25:174151.
Drapeau V, Despres JP, Bouchard C, et al. Modifications in food-group
consumption are related to long-term body-weight changes. Am J Clin
Nutr 2004;80:2937.
Nooyens AC, Visscher TL, Schuit AJ, et al. Effects of retirement on
lifestyle in relation to changes in weight and waist circumference
in Dutch men: a prospective study. Public Health Nutr 2005;8:126674.
Kahn HS, Tatham LM, Rodriguez C, Calle EE, Thun MJ, Heath CWJ.
Stable behaviors associated with adults 10-year change in body mass index
and likelihood of gain at the waist. Am J Public Health 1997;87:74754.
He K, Hu FB, Colditz GA, Manson JE, Willett WC, Liu S. Changes in intake
of fruits and vegetables in relation to risk of obesity and weight gain among
middle-aged women. Int J Obes Relat Metab Disord 2004;28:156974.
Vioque J, Weinbrenner T, Castello A, Asensio L, Garcia de la Hera M.
Intake of fruits and vegetables in relation to 10-year weight gain among
Spanish adults. Obesity (Silver Spring) 2008;16:66470.
Parker DR, Gonzalez S, Derby CA, Gans KM, Lasater TM, Carleton
RA. Dietary factors in relation to weight change among men and women
from two southeastern New England communities. Int J Obes Relat
Metab Disord 1997;21:1039.
Schulz M, Kroke A, Liese AD, Hoffmann K, Bergmann MM, Boeing H.
Food groups as predictors for short-term weight changes in men and
women of the EPIC-Potsdam cohort. J Nutr 2002;132:133540.
Halkjaer J, Sorensen TI, Tjonneland A, Togo P, Holst C, Heitmann BL.
Food and drinking patterns as predictors of 6-year BMI-adjusted changes
in waist circumference. Br J Nutr 2004;92:73548.
Williamson DF, Madans J, Anda RF, Kleinman JC, Giovino GA, Byers
T. Smoking cessation and severity of weight gain in a national cohort. N
Engl J Med 1991;324:73945.
Flegal KM, Troiano RP, Pamuk ER, Kuczmarski RJ, Campbell SM. The
influence of smoking cessation on the prevalence of overweight in the
United States. N Engl J Med 1995;333:116570.

209