Professional Documents
Culture Documents
& 2006 Nature Publishing Group All rights reserved 0307-0565/06 $30.00
www.nature.com/ijo
ORIGINAL ARTICLE
Adherence to a Mediterranean dietary pattern and
weight gain in a follow-up study: the SUN cohort
A Sanchez-Villegas1,2, M Bes-Rastrollo2, MA Martnez-Gonzalez2 and L Serra-Majem1
1
Department of Clinical Sciences, University of Las Palmas de Gran Canaria, Las Palmas, Spain and 2Department of
Preventive Medicine and Public Health, University of Navarra, Navarra, Spain
Introduction: The promotion of Mediterranean Diets has generated some doubts, because of the concern that its high fat
content might lead to the development of obesity.
Methods: Longitudinal analysis of 6319 participants in the SUN cohort study. We used a validated semiquantitative food
frequency questionnaire (136 items). Baseline adherence to a Mediterranean dietary pattern (MDP) was assessed using a score
(score_1) including vegetables, fruits, cereals, nuts, pulses, fish, olive oil and moderate consumption of red wine (positively
weighted), whereas meat and dairy products were negatively weighted. We assessed the association between the overall
baseline adherence to the MDP (score_1) and subsequent weight change after 28 months of follow-up. We also built another
score (score_2) to assess changes in diet during follow-up and appraised the association between the joint exposure to both
scores and weight change during follow-up.
Results: Participants in the first quartile of score_1 (lowest baseline adherence to MDP) showed a higher weight gain ( 0.73 kg)
than those in the top quartile ( 0.45 kg). The results indicated an inverse doseresponse relationship (P for trend 0.016). A
similar inverse association was apparent when we used change in adherence to the MDP (score_2). However, both inverse
associations did not remain statistically significant after adjusting for relevant confounders. Consumption of dairy products was
inversely associated with weight gain.
Conclusions: Although participants increased their average weight during the follow-up period, weight increments were smaller
among those with a higher adherence to an a priori defined MDP. Results did not remain statistically significant after
multivariate adjustment.
International Journal of Obesity (2006) 30, 350358. doi:10.1038/sj.ijo.0803118; published online 11 October 2005
Keywords: Mediterranean dietary pattern; olive oil; overweight; weight gain; cohort study
Introduction and lifestyle factors such as diet and physical activity pattern
in the Mediterranean area could play an essential role in the
Obesity is a growing health problem around the world. By development of obesity.2
2025, it has been estimated that the prevalence will be On the other hand, Mediterranean diets have been shown
440% in the US, 430% in England and 420% in Brazil.1 to be protective against mortality for all causes or the
Although the obesity rates are dramatically increasing in development of coronary heart disease, stroke, hypertension
Western countries, a rapid increase has been also observed in or cancer.36 A triangle (pyramid) has become the standard
developing countries. Moreover, the prevalence of obesity is pictorial way to represent food-based dietary guidelines. The
higher in Mediterranean countries (i.e. prevalence among base of the pyramid shows foods which should be consumed
Greek women is around 38%) than in Nordic countries (i.e. most frequently and the top includes those to be consumed
prevalence among Danish women is around 6%).2 The rarely. In this context, the food pyramid created by the US
reason for the reported difference is not known, but genetic Department of Agriculture (USDA) has been proposed to be
replaced by other models, including the Mediterranean Diet
Pyramid.5 In the Mediterranean Pyramid, nonrefined (i.e.
Correspondence: Dr A Sanchez-Villegas, Department of Clinical Sciences, whole-grain) cereals, fruits, vegetables and legumes are the
Center for Health Sciences, University of Las Palmas de Gran Canaria, PO Box base, whereas red meat is in the top, olive oil is closer to
550, Las Palmas de Gran Canaria, Las Palmas 35080, Spain.
the base than to the top (see http://www.oldwayspt.org/
E-mail: asanchez@dcc.ulpgc.es
Received 22 March 2005; revised 24 June 2005; accepted 20 July 2005; pyramids/med/p_med.html and http://www.nut.uoa.gr/
published online 11 October 2005 english/Greekguid.htm#diatr9).
Mediterranean dietary pattern and weight gain
A Sanchez-Villegas et al
351
However, the increased likelihood to obtain a high frequency nutrient composition of specified portion size,
percentage of energy from fats (specifically monounsatu- where frequencies were measured in nine categories (6 per
rated fatty acids such as oleic acid) when populations follow day/46 per day/23 per day/1 per day/56 per week/24 per
this dietary pattern generates doubts regarding its effect on week/1 per week/13 per month/never or almost never for
weight gain and obesity. each food item).
The Seguimiento Universidad de Navarra Follow-up Uni- All dietary intakes were adjusted for total energy intake
versity of Navarra (SUN Project) is a prospective dynamic using the residuals method.11
cohort study designed to establish the association between The MDP was defined according to an a priori defined
the adherence to a Mediterranean dietary pattern (MDP) and overall score previously used in the scientific literature.4,1216
several diseases and health conditions like overweight, This score was calculated according to the tertile distribution
obesity or weight change along time.7 The aim of the for the consumption of several components of the Mediter-
present prospective research was to evaluate the potential ranean diet. Eight of them were positively weighted (cereals,
relation between compliance with traditional Mediterranean vegetables, fruits, legumes, fish, nuts, olive oil and moderate
habits and subsequent weight maintenance and changes. red wine consumption) and two others were negatively
weighted (meat/meat products and whole-fat dairy pro-
ducts). The lowest tertile was coded as 1, the medium tertile
Methods as 2 and the highest tertile as 3 in the case of the eight
beneficial (positive) components. For the two negatively
The SUN Project is a Spanish prospective cohort study weighted items (meat/meat products and whole-fat dairy
designed in collaboration with the Harvard School of Public products), we inversely ranked the score, with 1 representing
Health that uses a similar methodology as large American the highest and 3 representing the lowest tertile. Finally, we
cohorts such as the Nurses Health Study and the Health summed up the 10 tertile values for each participant,
Professionals Follow-up Study.7 obtaining a score (score_1, hereafter) ranging from 10 to 30.
Information is collected using self-administered question- Red wine consumption was defined as moderate red wine
naires sent by postal mail every 2 years. The follow-up rate consumption. For scoring moderate red wine consumption,
exceeds 90% for the first 2-year period. The recruitment of a transformation centered at the level of consuming 20 g/day
participants started in December 1999 and it is permanently for men (20|20-absolute wine intake|) and 10 g/day for
ongoing as this is a dynamic cohort study. All participants women (10|10-absolute wine intake|) was used to obtain the
are university graduates. For the present analyses we highest value for men consuming 20 g/day or women
included participants who had already been followed up consuming 10 g/day, and progressive lower values as the
for at least 2 years (n 7908). Those participants who consumption was lower or higher than these values. The
reported extremely low or high values for total energy intake continuous variable thus built was then categorized in
(less than 800 kcal/day in men and 600 kcal/day in women or tertiles and included as a positively weighted component
more than 4200 kcal/day in men and 3500 kcal/day in in the pattern.
women), subjects with missing values in the variables of Additionally, the 2-year follow-up questionnaire (Q_2,
interest in the analysis and participants with biological hereafter) included several items regarding changes in the
implausible values for the weight or height values were main dietary habits (consumption in fruits, vegetables,
excluded. Finally, data from 6319 participants remained meat/meat products, fish, olive oil and alcohol). These items
available for the analysis. were combined to assess whether the adherence to the MDP
The study was approved by the Human Research Ethical had increased or not during the follow-up period. Thus, we
Committee at the University of Navarra. Voluntary comple- built another score (score_2, hereafter) using these changes.
tion of the first self-administered questionnaire was con- For any increase in the consumption of fruit, vegetables,
sidered to imply informed consent. olive oil, fish or alcohol we summed 2 points, for no change
in each of these items we summed 1 point and for any
reduction we summed no point. On the other hand, for any
Assessment of exposure: A priori MDP increase in the consumption of meat/meat products we
Dietary exposures were assessed through a self-administered summed no point, for no change we summed 1 point and for
questionnaire (baseline questionnaire) including a 136-item any decrease we summed 2 points. Therefore, score_2 ranged
food frequency questionnaire (FFQ) previously validated in from 0 to 8. The new score was further categorized in two
Spain.8 This questionnaire was included in the baseline groups (subjects with p4 points versus subjects with 44
assessment (Q_0, hereafter). points).
Nutrient intake scores were computed using an ad hoc
computer program specifically developed for this aim. A
trained dietitian updated the nutrient data bank using the Assessment of other variables
latest available information included in food composition The baseline questionnaire also included different questions
tables for Spain.9,10 Nutrient scores were calculated as related to lifestyle, with 46 items for men and 54 items for
Gender (%)
Men 56.1 43.0 36.9 34.2
Lipid intake (%E) 39.7 (5.9) 37.7 (6.4) 36.6 (6.4) 34.4 (6.4)
PUFA (%E) 5.7 (1.7) 5.3 (1.6) 5.3 (1.6) 5.1 (1.5)
MUFA/SFA ratio 1.1 (0.2) 1.2 (0.3) 1.4 (0.3) 1.5 (0.4)
Omega-3 (%E) 0.3 (0.2) 0.4 (0.3) 0.5 (0.4) 0.5 (0.3)
Trans fatty acids (%E) 0.3 (0.3) 0.2 (0.3) 0.2 (0.3) 0.1 (0.2)
Fiber intake (g/day) 20.3 (8.5) 24.6 (9.8) 29.1 (12.2) 34.3 (12.6)
Calcium intake (mg/day) 1235.4 (479.7) 1222.0 (486.3) 1212.6 (486.3) 1244.9 (436,3)
Energy-adjusted whole fat dairy products (g/day) 329 (225.2) 249 (197.8) 184 (159.0) 127 (123.7)
Energy intake (kcal/day) 2569 (636.6) 2386 (647.3) 2323 (645.0) 2301 (600.3)
a
Mean and s.d. unless otherwise stated. BMI: body mass index, MET: metabolic equivalents, CHO: carbohydrates, PUFA: polyunsaturated fatty acids, MUFA:
monounsaturated fatty acids, SFA: saturated fatty acids.
Although the point estimates suggested a beneficial effect highest quartile had an increment 0.17 kg/m2 (P 0.06).
of a high consumption of fruit, nuts and olive oil on weight In addition, the results for BMI also indicated a statistically
gain, the results were not statistically significant. significant inverse doseresponse relationship (P for tren-
Table 3 shows that although, on average, participants d 0.026). However, this association did not remain statis-
increased their weight and their BMI throughout follow-up, tically significant after adjusting for relevant confounding
weight increments were smaller among participants with a factors. Adjustment in model 2 (Table 3) includes energy
higher adherence to an MDP. Subjects with the lowest intake; when we alternatively controlled for baseline weight
baseline adherence to an MDP (quartile 1 of score_1) showed (instead of energy and BMI) this alternative adjustment did
weight gain 0.73 kg, whereas subjects with the highest not materially affect our results (data not shown).
adherence (quartile 4) experienced weight gain 0.45 kg When we analysed 1763 prevalent cases of overweight or
(P 0.05). The results indicated a statistically significant obesity reported in the baseline questionnaire (defined as
doseresponse relationship (P for trend 0.016). However, BMIX25 kg/m2 in Q_0), the results did not substantially
although the trends were similar, these associations did not differ. No statistically significant association between the
remain statistically significant in the multivariate model. adherence to the MDP and weight gain was found in the
Subjects belonging to the first quartile of score_1 showed regression models (mean changes [kg] 0.21, 0.49,
an increment 0.26 kg/m2 in BMI, whereas subjects in the 0.09, 0.37 for quartiles 14; P for trend 0.950).
Components of score_1 adjusted Weight changea (kg) 95% CI Weight changeb 95% CI Weight changec 95% CI
for energy (tertiles: T1T3) (mean) (kg) (mean) (kg) (mean)
T1: lowest tertile, T2: medium tertile, T3: highest tertile. aCrude. bAdjusted for age, gender, baseline BMI, smoking, physical activity, alcohol consumption, energy
intake, change in dietary habits (fruits, vegetables, meat/meat products, fish, olive oil and alcohol) and physical activity from Q_0 to Q_2. cAdditionally adjusted for
all the other food items shown in the table. Weight change: weight (kg) in Q_2 minus weight (kg) in Q_0. s.d.: standard deviation.
95% CI: 95% confidence interval. Model 1: crude. Model 2: adjusted for baseline age, gender, BMI, smoking, physical activity, alcohol and energy intake. Model 3:
additionally adjusted for changes in dietary habits (fruits, vegetables, meat/meat products, fish, olive oil and alcohol) and physical activity during follow-up. Weight
change: weight (kg) in Q_2 minus weight (kg) in Q_0. BMI: Body Mass Index (kg/m2). BMI change: Body Mass Index (kg/m2) in Q_2 minus Body Mass Index
(kg/m2) in Q_0.
Table 4 shows the joint exposure to baseline adherence to a Although participants increased their average weight and
MDP (score_1) and to changes in the adherence during BMI during follow-up, crude increments were lower among
follow-up (score_2). Participants with lower baseline adher- those with a higher adherence to an a priori defined MDP
ence to a MDP and with no improvement during follow-up and among those who increased their adherence to a MDP
(score_2p4) were considered as the reference category. A during follow-up.
higher baseline adherence to a MDP together with an In any case, beyond controversies, the MDP did not show
increase in adherence during follow-up led to lower weight an association with a higher likelihood of weight gain,
gain ( 0.36 kg, whereas weight gain was 0.80 kg in the overweight or obesity in the SUN cohort.
reference category; P 0.02). The results were similar when Similarly, a recent study found a protective and statisti-
we assessed change in BMI ( 0.13 versus 0.28 kg/m2; cally significant inverse association between the MDP and
P 0.04). Nevertheless, the results were not statistically obesity. However, the results reported from that study
significant after adjusting for relevant confounders. acknowledged that they were not conclusive because of the
We included 4556 subjects without prevalent overweight cross-sectional nature of the design. In addition, that study
or obesity at baseline to assess the association between did not include olive oil consumption in the Mediterranean
adherence to a MDP and the risk of incident overweight/ score.15
obesity during follow-up. During 2 years of follow-up, we The scientific literature about this topic is not abundant,
observed 353 new cases of overweight/obesity. No significant but to our knowledge some small clinical trials are consistent
association was observed when we fit either crude or with our findings.1921 They found that the Mediterranean
multivariate nonconditional logistic regression models with diet is a good alternative diet for weight loss compared with a
quartiles of baseline adherence to a MDP (score_1) as the low-fat diet, both controlled for energy.19 Fernandez de la
exposure and incident overweight/obesity as the outcome Puebla et al.20 found that an isocaloric replacement of a
(Table 5). saturated fat-rich diet by a Mediterranean diet decreases total
We also assessed different interactions (effect modifica- body fat in hypercholesterolemic males. Flynn et al.21
tion) in the association between adherence to a MDP and the reported that a Mediterranean diet is at least as effective as
risk of overweight/obesity, but none of them was statistically any diet published, in patients attending a weight-loss
significant (gender*age; gender*baseline BMI, age*baseline programme.
BMI; baseline BMI*MDP; gender*MDP, age*MDP and phy- The promotion of olive oil consumption has been a
sical activity*MDP). notorious issue for public health nutrition. Although its
benefits on health have been largely tested, the composition
of this Mediterranean component (100% fat) has led to some
criticism about recommending virgin olive oil consumption
Discussion to protect from cardiovascular disease because it is presumed
to increase the risk of overweight/obesity. Furthermore,
To our knowledge, this is the first time that a prospective Ferro-Luzzi et al.22 criticized the current nutritional advice
cohort study analyses the effect of the adherence to a MDP applied in Greece and based in the Mediterranean Diet
on weight or BMI change. Pyramid.23 A crucial discussion matter was that the high
Model 1: crude. Model 2: adjusted for age and gender. Model 3: additionally adjusted for baseline BMI, smoking, physical activity, alcohol consumption and energy intake. BMI: Body Mass Index (kg/m2).
+0.45 (+0.06 to +0.84)
+0.52 (+0.13 to +0.90)
Increase (44)
area in the last 30 years is a strong reason for concern
750
regarding the threating obesity epidemic. The high fat
BMI change: Body Mass Index (kg/m2) in Q_2 minus Body Mass Index (kg/m2) in Q_0. Weight change: weight (kg) in Q_2 minus weight (kg) in Q_0. 95% CI: 95% confidence interval.
content based on olive oil consumption of the Greek diet
Quartile 4 (X22)
gain.
The usual definition of MDP involves whole milk products
No change/decrease (p4)
Model 1
Model 2
Model 3
Quartile 1 (o18) Quartile 2 (1819) Quartile 3 (2021) Quartile 4 (X22) P for trend
95% CI: 95% confidence interval. aAdditionally adjusted for baseline BMI, smoking, physical activity, alcohol and energy intake and change in dietary habits (fruits,
vegetables, meat/meat products, fish, olive oil and alcohol) and physical activity from Q_0 to Q_2.
acknowledge this potential limitiation. However, it is needed Mediterranean countries, particularly among younger gen-
to take also into account that the major fat-related feature erations.37
generally used to define a MDP is not the total fat intake
(which is still high in absolute terms in our overall sample)
but the ratio of monounsaturated to saturated fatty acids Acknowledgements
(MUFA/SFA ratio), and we did find that the distribution of
the MUFA/SFA ratio monotonically increased from 1.1 to 1.5 We are indebted to the participants of the SUN Study for
across successive quartiles of the MDP. their continued cooperation and participation. The Depart-
One of the strengths of our study is its ability to adjust for ment of Health of the Navarre Regional Government and
changes in dietary habits during follow-up to eliminate the Spanish Ministry of Health (Fondo de Investigaciones
possible effect of reverse causality in the reported associa- Sanitarias, projects 01/0619 and 04/042241 and G03/140,
tions. The inclusion of changes in fruits, vegetables, alcohol, Red Tematica de Dieta y Enfermedad Cardiovascular) are
fish, meat/meat products and olive oil consumption in the gratefully acknowledged for supporting the present study.
models could be considered as a proxy of change in the We would like to thank other members of the SUN Study
adherence to a MDP. Group: M Segu-Gomez, A Alonso, RM Pajares, JA Martinez,
Additionally, we have to consider that Mediterranean M Delgado-Rodrguez, M Serrano-Martnez, M Marques, A
lifestyle includes the regular practice of physical activity. Marti, M Munoz, F Guillen-Grima and I Aguinaga and our
Besides this, epidemiological evidence supports a detrimen- advisors from the Department of Nutrition of the Harvard
tal role of sedentary lifestyles on increasing obesity epidemic School of Public Health: A Ascherio, W Willett and FB Hu,
in Mediterranean countries.35,36 For that reason, we took who helped us to design the SUN cohort study.
into account not only baseline physical activity but also
changes in physical activity patterns during follow-up. Thus,
we do not think this factor has confounded the reported References
associations.
Among participants initially free of overweight/obesity, a 1 Kopelman PG. Obesity as a medical problem. Nature 2000; 404:
635643.
higher adherence to an MDP was not associated with a
2 York DA, Rossner S, Caterson I, Chen CM, James WP, Kumanyika
higher likelihood of becoming overweight/obese. On the S et al. AHA Conference preceedings. Prevention Conference VII.
other hand, when we restricted the analyses to either Obesity, a worlwide epidemia related to heart disease and stroke.
subjects with a baseline BMIX25 kg/m2, the results regarding Group I: Worlwide demographics of obesity. Circulation 2004;
110: e463e470.
the association between the adherence to an MDP and
3 Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adher-
weight gain did not change. Thus, it does not seem that a ence to a Mediterranean diet and survival in a Greek population.
previous status of overweight/obesity modifies the associa- N Engl J Med 2003; 348: 25992608.
tion between adherence to an MDP and subsequent weight 4 Martnez-Gonzalez MA, Fernandez-Jarne E, Serrano-Martnez M,
Marti A, Martinez JA, Martn-Moreno JM. Mediterranean diet and
change.
reduction in the risk of a first acute myocardial infarction: an
The results of this study do not support the hypothesis operational healthy dietary score. Eur J Nutr 2002; 41: 153160.
that the reported increases of obesity rates in Mediterranean 5 Willett WC. Eat, Drink, and be Healthy. Simon & Schuster Source:
countries are attributable to the adherence to an MDP. An New York, 2001.
6 Bosetti C, Gallus S, Trichopoulou A, Talamini R, Franceschi S,
alternative explanation could be that the increasing obesity
Negri E et al. Influence of the Mediterranean Diet on the risk of
rates may instead be related to declining physical activity in cancers of the upper aerodigestive tract. Cancer Epidemiol
the population and to the abandonment of MDP observed in Biomarkers Prev 2003; 12: 10911094.