Nutrition, Metabolism & Cardiovascular Diseases (2011) 21, 237e244
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/nmcd
Mediterranean diet and the incidence of cardiovascular disease: A Spanish cohort ´pez a,b, M. Bes-Rastrollo a, ´lez a,*, M. Garcı´a-Lo M.A. Martı´nez-Gonza E. Toledo a,c, E.H. Martı´nez-Lapiscina a,d, M. Delgado-Rodriguez e, Z. Vazquez a, S. Benito a, J.J. Beunza a a
Department of Preventive Medicine and Public Health, Medical SchooleClinica Universitaria, University of Navarra, c/ Irunlarrea, 1 Ed. Investigacion, 31080 Pamplona, Navarra, Spain b Department of Cardiology, Medical SchooleClinica Universitaria, University of Navarra, Spain c Department of Preventive Medicine and Quality Management, Hospital Virgen del Camino, Pamplona, Navarra, Spain d Department of Neurology, Hospital de Navarra, Pamplona, Navarra, Spain e Division of Preventive Medicine and Public Health, University of Jaen, Jaen, Spain Received 31 March 2009; received in revised form 13 July 2009; accepted 20 October 2009
KEYWORDS Mediterranean diet; Nutrition; Myocardial infarction; Stroke
Abstract Background and aim: The Mediterranean diet is considered a model for healthy eating. However, prospective evidence in Mediterranean countries evaluating the relationship between this dietary pattern and non-fatal cardiovascular events is scarce. The aim of the present study was to evaluate the association between the adherence to the Mediterranean diet and the incidence of fatal and non-fatal cardiovascular events among initially healthy middle-aged adults from the Mediterranean area. Methods and results: We followed-up 13,609 participants (60 percent women, mean age: 38 years) initially free of cardiovascular disease (CVD) during 4.9 years. Participants were part of a prospective cohort study of university graduates from all regions of Spain. Baseline diet was assessed using a validated 136-item food-frequency questionnaire. A 9-point score was used to appraise adherence to the Mediterranean diet. Incident clinical events were confirmed by a review of medical records. We observed 100 incident cases of CVD. In multivariate analyses, participants with the highest adherence to the Mediterranean diet (score > 6) exhibited a lower cardiovascular risk (hazard ratio Z 0.41, 95% confidence interval [CI]: 0.18e0.95) compared to those with the lowest score (<3). For each 2-point increment in the score, the adjusted hazard ratios were 0.80 (95% CI: 0.62e1.02) for total CVD and 0.74 (0.55e0.99) for coronary heart disease. Conclusions: There is an inverse association between adherence to the Mediterranean diet and the incidence of fatal and non-fatal CVD in initially healthy middle-aged adults. ª 2009 Elsevier B.V. All rights reserved.
* Corresponding author. Tel.: þ34 948425600x6463; fax: þ34 948425649. E-mail address:
[email protected] (M.A. Martı´nez-Gonza ´lez). 0939-4753/$ - see front matter ª 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.numecd.2009.10.005
238
Introduction Food patterns represent a holistic approach to evaluate relationships between dietary habits and health outcomes [1]. The first option to appraise food patterns is to apply exploratory data analyses (e.g. principal component analysis [PCA]) to identify combinations of foods that explain a sizeable amount of total variability of food intake in the sample. This post hoc approach provides useful epidemiological information but it lacks stability and typically impedes the comparison of results across studies [2]. Alternatively, hypothesis-oriented (a priori) scores are based on available scientific evidence. A priori scores better capture the complexity of diet, and thus provide a more efficient way of controlling for confounding factors by including nutritional confounders in the score and also incorporate possible effect modification among nutritional variables. One limitation is that this approach allocates habitually equal weights to each component of the score thus implicitly assuming the same importance for every component [3]. In the context of food patterns, the Mediterranean diet, customarily operationalized using an a priori approach, is acquiring an emerging role in cardiovascular prevention [4]. It is associated with increased longevity [5] and lower mortality from cardiovascular disease (CVD) [4]. In most previous investigations only fatal events were included. Recently, a large American cohort, using an alternative definition of the Mediterranean diet [6] and a Greek cohort assessing predictors for a 5-year incidence of CVD [7] reported a favourable effect of the Mediterranean diet on the incidence of CVD. Some previous case-control studies [8,9] also found strong inverse associations between adherence to the Mediterranean diet and the incidence of CVD. Besides this, the largest studies conducted about adherence to the Mediterranean diet and cardiovascular mortality have mainly included elderly people or nonMediterranean populations [6,10]. A Mediterranean dietary pattern is uncommon in non-Mediterranean populations. We evaluated the effectiveness of the Mediterranean diet in reducing the incidence of cardiovascular events in a large cohort of Mediterranean middle-aged adults.
Methods Study population The SUN [Seguimiento Universidad de Navarra (University of Navarra follow-up)] project is an ongoing, multipurpose, prospective and dynamic cohort of university graduates conducted in Spain. The Institutional Review Board of the University of Navarra approved the study protocol. The study methods and the cohort profile have been published in detail elsewhere [11]. Briefly, beginning in December 1999, participants, who were all university graduates, were contacted periodically. Enrolment is permanently open and follow-up is conducted through mailed questionnaires every two years. Non-respondents received up to 5 additional mailings requesting their follow-up questionnaire. We assessed 15,975 participants recruited before January 2006 because they had spent enough time in the
M.A. Martı´nez-Gonza ´lez et al. study to be able to complete at least the 2-year follow-up questionnaire. Among them, the retention rate was 88.4 percent. Therefore, we had follow-up information on 14,129 participants. We excluded 306 participants who did not conform to our predefined values for total energy intake (men: <500 or >5,000; women: <400 or >4800 kcal/ day) and 214 participants due to baseline prevalent CVD (either coronary heart disease or stroke). Thus, the effective sample size was 13,609 participants. Among them 6803, 3832; and 2974 had completed the 6-year, 4-year and 2-year follow-up, respectively.
Exposure assessment A validated semi-quantitative 136-item food-frequency questionnaire [12] and updated Spanish food composition tables were used at baseline. Conformity with the traditional Mediterranean diet was appraised with a previously used [5,13] score. For each of the six potentially protective components (monounsaturated to saturated fatty acid ratio [MUFA:SFA], legumes, cereals, fruits and nuts, vegetables or fish) participants received one point if their intake was above the sample sex-specific median. Participants received one point if their intake was below the median for the two nonprotective components (total dairy products or meat / meat products). Regarding ethanol, one point was scored if consumption was 10e50 g/day for men or 5e25 g/day for women. Nine points reflected maximum adherence and zero points reflected no adherence at all. The validation study of the food-frequency questionnaire was recently repeated against four 3-day dietary records [14]. The intra-class correlation coefficients were 0.61 for MUFA, 0.75 for SFA, 0.40 for legumes, 0.71 for cereals, 0.72 for fruits, 0.81 for vegetables, 0.59 for fish, 0.84 for dairy products, 0.75 for meat, and 0.82 for ethanol. A PCA approach to empirically identify dietary patterns was also used. The 136 food items were grouped in 30 predefined food groups [15]. The Screen plot examination was used to determine the number of factors (or vectors) to be extracted. To improve interpretability, an orthogonal rotation procedure (varimax rotation) was used. The number of vectors retained was decided according to the amount of variation explained and the natural interpretation of them. Food groups that loaded >0.25 were considered to contribute to the vector. The factor score for each pattern was constructed by summing observed consumptions of the component food items weighted by their factor loadings. Thus, each participant received a factor score for each identified pattern. A higher score suggested better adherence to a certain dietary pattern. We extracted two patterns (‘‘Westernized’’ and ‘‘post hoc Mediterranean’’ scores) and categorized them in quintiles.
Assessment of other covariates The baseline questionnaire requested information about anthropometric characteristics (weight, height), health related habits (smoking status, physical activity, sedentary lifestyle), and clinical variables (use of medication, personal and family history of coronary heart disease, and other CVD). The validity of self-reported weight and body
Mediterranean diet and cardiovascular disease mass index (BMI) has been previously documented in a subsample of this cohort. The correlations between selfreported and measured weight and BMI were 0.99 and 0.94, respectively. The average relative error in weight was 1.5 percent [16]. Physical activity was assessed using a previously validated physical activity questionnaire and the Spearman correlation coefficient between the questionnaire information and objectively obtained measurements (triaxial accelerometer) was 0.51 (p < 0.001) [17]. Metabolic equivalents (METs) were estimated to yield METs-h/ week scores for each participant [18].
239 All p-values are two-tailed and statistical significance was set at the conventional cut-off of p < 0.05.
Role of the funding source This study was exclusively funded by the official public Institutes of the Spanish Government for funding health research (Instituto de Salud Carlos III ) and the Department of Health of the Navarra Regional Government. The funding sources had no role in the design, collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication.
Ascertainment of incident cardiovascular events The primary end point for the present analysis was the combined outcome measure of incident cardiovascular death, acute coronary syndromes (myocardial infarction with or without ST elevation), revascularization procedures or fatal or non-fatal stroke. Participants who reported any of these diagnoses by a physician on a follow-up questionnaire were asked for their medical records. An expert panel of physicians, blinded to the information on diet and risk factors, adjudicated the events by reviewing medical records applying the universal criteria for myocardial infarction [19] or clinical criteria for the other outcomes. A non-fatal stroke was defined as a focal-neurological deficit of sudden onset and vascular mechanism that lasted more than 24 h. Cases of fatal stroke were documented if they had evidence of cerebrovascular mechanisms. Deaths were reported to our research team by the subjects’ next of kin, work associates and postal authorities and for participants lost to follow-up, the National Death Index was used to identify deceased cohort members. Fatal coronary heart disease was confirmed by a review of medical records with the permission of the next of kin.
Statistical analysis We used Cox regression models with age as the underlying time variable. The category with the lowest adherence to the Mediterranean diet (score < 3) was used as the reference. We estimated the HRs for each 2-point increase in the score. The assumption of proportional hazards was tested using timedependent covariates. We repeated the estimation for 2-point increases in the a priori-defined Mediterranean diet by using a simple approximate Bayesian approach through information-weighted averaging, as recommended by Greenland, to obtain approximate posterior HRs [20]. In all analyses, we fitted a first Cox regression model adjusted for age, sex and total energy intake. In a second model we additionally adjusted for family history of coronary heart disease, smoking (never, past and current smokers) and physical activity (continuous). In a third model, we also adjusted for baseline BMI (<25, 25e29.9 and 30 kg/m2); a history of hypertension or use of medication for hypertension at baseline; use of aspirin; diabetes at baseline and dyslipidaemia at baseline. For linear trend tests, we treated the exposure (scores) as a continuous variable. When we assessed food items or food groups, we computed the linear trend test by assigning to each category the median of the respective quantile and treated the variable as continuous.
Results Median follow-up in our cohort was 4.9 years. During 66,577 person-years of follow-up, 100 cases of incident CVD (68 coronary acute syndromes, including 7 revascularization procedures, and 32 incident strokes) were observed. Only 8 events (4 myocardial infarctions and 4 strokes) were fatal. Baseline characteristics of the study participants according to their adherence to the Mediterranean diet are presented in Table 1. Participants in the lowest extreme (0e1) and in the upper extreme of adherence (8e9) were 5.2% and 3.4%, respectively. The most frequent scores were 4 (20.4%) and 5 points (19.2%). Baseline adherence to the Mediterranean diet slightly improved in our cohort during the recruitment period 1999e2005 (beta Z þ0.09 points/ year, p < 0.001 age- and sex-adjusted). A higher adherence was associated with a better nutrient profile and more active lifestyles, but also with older age, being an exsmoker, or having diabetes, dyslipidaemia, hypertension or a family history of coronary heart disease at baseline. A higher adherence to the Mediterranean diet was significantly associated with a lower risk for developing CVD regardless of the variables that we controlled for in multivariable-adjusted models (Table 2). Specifically, participants with the highest adherence (score 7) showed a 59 percent significantly lower risk of CVD than those with the lowest adherence (score 2). The p for linear trend was 0.07. Regarding coronary heart disease, the comparison between extreme categories of adherence (7e9 versus 0e2) rendered also similar point estimates but with wider confidence intervals. However, there was a significant trend for this association (p Z 0.04). A two-point increment in the Mediterranean-diet score was associated with a 20 percent reduction in the risk of CVD and with a 26 percent reduction in the incidence of coronary heart disease. To adopt an approximate Bayesian perspective [21], we modelled our prior ideas about the relationship between Mediterranean diet and CVD assuming a prior 95% probability on a HR between 0.69 and 0.99 (these values were the lowest of the inferior limits and the highest of the upper limits of the 95% confidence intervals of four studies estimating relative risks for cardiovascular mortality in a previous meta-analysis [4]). Under these assumptions, our posterior HR for CVD associated with a 2-point increment in the score was 0.81 (95% CI: 0.70e0.94) and for coronary heart disease it was 0.80 (95% CI: 0.68e0.93). When we repeated the analyses adjusting each component for total energy intake (residual method) and
240
M.A. Martı´nez-Gonza ´lez et al.
Table 1 Baseline characteristicsa of the SUN Project population according to adherence to the Mediterranean food pattern, 1999e2008. Adherence to the Mediterranean diet
N Sex (% women) Smoking Current smokers (%) Ex-smokers (%) Diabetes at baseline (%) Hypertension at baseline (%) Dyslipidaemia at baseline (%) Family history of CHD (%) Body mass index (kg/m2) 25e29.9 (%) 30 (%) Age (y) BMI (kg/m2) Physical activity (MET-h/week) Total energy intake (Kcal/days) Alcohol intake (g/day) Protein intake (% energy intake) Carbohydrate (% energy intake) Total fat intake (% energy intake) PUFA (% energy intake) MUFA (% energy intake) SFA (% energy intake) Ratio MUFA/SFA Olive oil consumption (g/day) Fiber intake (g/day) Vitamin C (mg/day)
Low score (0e2)
Lowemoderate score (3e4)
Moderateehigh score (5e6)
High score (7e9)
p value
2264 60
5171 61
4622 60
1552 56
27 21 0.8 7 15 6
27 27 1.7 9 21 8
25 32 2.3 12 26 8
25 37 2.4 14 33 10
<0.001
21 4.7 34 (10) 23 (3) 21 (20)
24 4.6 37 (11) 23 (3) 23 (20)
27 5.1 39 (12) 24 (3) 26 (24)
29 5.2 43 (12) 24 (3) 29 (24)
<0.001
2295 (711)
2414 (763)
2600 (761)
2679 (684)
<0.001
4 (9) 18 (4)
6 (10) 18 (4)
8 (11) 18 (3)
10 (11) 18 (3)
<0.001 <0.001
40 (7)
43 (7)
45 (7)
47 (7)
<0.001
40 (6)
38 (6)
35 (6)
33 (6)
<0.001
6 (2) 16 (3) 15 (3) 1.08 (0.18) 14 (14)
5 (2) 16 (4) 13 (3) 1.23 (0.31) 20 (17)
5 (2) 15 (4) 11 (3) 1.38 (0.34) 25 (20)
5 (2) 15 (4) 10 (2) 1.58 (0.39) 29 (20)
<0.001 <0.001 <0.001 <0.001 <0.001
18 (7) 176 (87)
25 (10) 250 (132)
33 (13) 336 (166)
40 (15) 405 (183)
<0.001 <0.001
0.001
0.001 <0.001 <0.001 0.008
<0.001 <0.001 <0.001
CHD: coronary heart disease; BMI: body mass index; MET: metabolic equivalent; MUFA: monounsaturated fatty acids; SFA: saturated fatty acids; PUFA: polyunsaturated fatty acids. a Mean (standard deviation) unless otherwise stated.
thereafter summing up the points, the associations were of similar magnitude, but they lost statistical significance, a 2-point increase in the score had HR Z 0.86 (0.68e1.09) for CVD and 0.80 (0.60e1.07) for coronary heart disease (data not shown). No interaction was apparent between the Mediterranean diet and sex. The association of each component (dichotomized at their sex-specific medians) of the Mediterranean diet with cardiovascular risk is shown in Table 3. In fully adjusted models, the only item predictive of a lower incidence of CVD was the consumption of fruits and nuts. Most point estimates were in the expected direction, with the notable exception of cereals (HR Z 1.19 and 1.36 for CVD and coronary disease, respectively). We repeated the analyses after excluding cereals from the score. Thus, the new score (range 0e8) included five potentially protective components (MUFA:SFA ratio,
legumes, fruits and nuts, vegetables and fish), two nonprotective components (dairy products and meat) and ethanol. Using this new score without cereals, the inverse association with CVD was enhanced. The fully adjusted HRs (95% confidence intervals) for moderate (4e5) and high (6) adherence were 0.74 (0.48e1.16) and 0.41 (0.22e0.76), respectively, p for trend Z 0.03. This was even more apparent for coronary disease, HR Z 0.57 (0.33e0.98) for moderate and HR Z 0.39 (0.19e0.78) for high adherence, p for trend Z 0.01. For each two additional points in this new score, the HRs were 0.75 (95% CI: 0.58e0.97) and 0.67 (95% CI: 0.49e0.92) for CVD and coronary disease, respectively (data not shown). To explore more deeply the potential effect of different types of cereals, we estimated the partial correlation coefficient (adjusted for age and sex) between white bread (the most frequently consumed item in this
Mediterranean diet and cardiovascular disease
241
Table 2 Hazard ratios (95% confidence intervals) of cardiovascular disease according to the adherence to the Mediterranean food pattern (9-point score). The SUN Project, 1999e2008. Adherence to the Mediterranean diet
Low 0e2
Lowemoderate 3e4
Moderateehigh 5e6
High 7e9
N Incident cases of cardiovascular disease Person-years Agea- and sex-adjusted HR Multivariable HR (95% CI)b Multivariable HR (95% CI)c
2264 12
5171 32
4622 43
1552 13
11491 1 (ref) 1 (ref) 1 (ref)
25584 0.63 (0.32e1.23) 0.65 (0.33e1.27) 0.64 (0.32e1.24)
22335 0.66 (0.34e1.27) 0.71 (0.37e1.39) 0.66 (0.34e1.27)
7166 0.39 (0.18e0.88) 0.43 (0.19e0.97) 0.41 (0.18e0.95)
Incident cases of coronary heart disease Agea- and sex-adjusted HR Multivariable HR (95% CI)b Multivariable HR (95% CI)c
8
24
26
10
1 (ref) 1 (ref) 1 (ref)
0.65 (0.29e1.45) 0.68 (0.30e1.52) 0.67 (0.30e1.51)
0.52 (0.23e1.18) 0.58 (0.26e1.31) 0.55 (0.24e1.24)
0.37 (0.14e0.97) 0.41 (0.16e1.08) 0.42 (0.16e1.11)
p for trend
For each þ2 points
0.04 0.08 0.07
0.78 (0.62e0.99) 0.81 (0.64e1.03) 0.80 (0.62e1.02)
0.02 0.04 0.04
0.71 (0.53e0.94) 0.74 (0.55e0.99) 0.74 (0.55e0.99)
The proportionality of hazards was tested using time-dependent covariates. The estimated hazard ratios for these covariates were 1.002 (p Z 0.85) for total cardiovascular disease and 1.006 (p Z 0.59) for coronary heart disease. a We used Cox regression models with age as the underlying time variable. b Adjusted for age, sex, family history of coronary heart disease, total energy intake, physical activity, smoking. c Additionally adjusted for BMI, diabetes at baseline, use of aspirin, history of hypertension and history of hypercholesterolemia.
group in Spain) and the cereals group. The partial correlation coefficient was 0.85. When we compared the top versus the bottom quintile of cereal consumption (data not shown), the HR for CVD was 1.30 (95% CI: 0.67e2.53) and the HR for coronary disease was 2.07 (95% CI: 0.90e4.77) after adjusting for all the above-mentioned confounders. The linear trend suggested an increased coronary disease risk across quintiles of cereal consumption (p Z 0.06). The HR for coronary disease, specifically for white bread, (top versus lowest quintile) was 1.92 (95% CI: 1.05e3.51, p for linear trend Z 0.005), after adjustment for conventional risk factors but not for total energy intake; this estimate lost significance when we adjusted the analysis for total energy intake (p for trend Z 0.08).
The PCA yielded two major vectors explaining 23 percent of total variability in food intake. Factor loading matrices are shown in Table 4 (omitted if <0.25). The first vector was represented by a high consumption of fast food, potatoes, whole-fat dairy products, processed and red meats, sauces, processed meals, commercial bakery, refined cereals, sugar-sweetened sodas and eggs and by a low consumption of low-fat dairy products. The second vector was positively correlated with the consumption of vegetables, fruits, fish, poultry, low-fat dairy, whole grains, nuts, olive oil and legumes. We labelled these two vectors as the ‘‘Westernized’’ and ‘‘post hoc Mediterranean’’ dietary patterns, respectively. The ‘‘post hoc Mediterranean’’ pattern (PHMP) showed no association with the incidence of CVD (Table 5). The correlation coefficients between these
Table 3 Hazard ratios (95% confidence intervals) of cardiovascular disease according to the intake of each of the components of the Mediterranean food pattern (9-point score). The SUN Project, 1999e2008. Components of the Mediterranean score
Adjusteda HR (95% CI) of CVD or CHD for those cut-off point Cut-off point (g/day)
Ratio MUFA/SFA Vegetables Fruits and nuts Legumes Cereals Fish Alcohol Dairy Meat/meat products
Women
Men
1.24 501 300 21 81 86 5e25 143 170
1.19 401 235 21 90 87 10e50 182 177
CVD
0.86 1.09 0.58 0.80 1.19 0.85 0.81 1.09 1.31
CHD
(0.57e1.29) (0.72e1.66) (0.38e0.89) (0.53e1.21) (0.76e1.85) (0.56e1.28) (0.51e1.28) (0.72e1.67) (0.86e2.00)
0.94 0.82 0.62 0.70 1.36 0.75 0.75 1.51 1.01
(0.61e1.47) (0.50e1.35) (0.37e1.04) (0.43e1.15) (0.80e1.33) (0.46e1.24) (0.43e1.32) (0.91e2.49) (0.60e1.70)
HR: hazard ratio, CVD: cardiovascular disease, CHD: coronary heart disease, MUFA: Mono-unsaturated fatty acids, SFA: saturated fatty acids. a Adjusted for age, sex, family history of CHD, total energy intake, physical activity, smoking, BMI, diabetes at baseline, use of aspirin, history of hypertension and history of hypercholesterolemia.
242
M.A. Martı´nez-Gonza ´lez et al.
Table 4 Empirically derived scores for food patterns: factor loading matrix. The SUN Project, 1999e2008. ‘‘Westernized’’ pattern Fast-food Potatoes High-fat dairy products Cold/processed meats Red meats Sauces Processed meals Commercial bakery Sugar-sweetened soda Refined cereals Eggs Sugar, jam, honey Low-fat dairy
‘‘Post hoc Mediterranean’’ pattern
0.55 0.51 0.49 0.49 0.49 0.49 0.45 0.42 0.40 0.38 0.36 0.30 0.26
0.39
Vegetables Fruits Fish Poultry Whole-grain cereals Nuts and seeds Olive oil Legumes
0.70 0.61 0.56 0.39 0.37 0.35 0.33 0.33
two patterns and the a priori built Mediterranean diet (Trichopoulou’s score) were 0.25 and þ0.57, respectively. However, in fully adjusted models, the ‘‘Westernized’’ pattern was associated with a higher risk of CVD (Table 5).
Discussion Our results support an inverse association between the Mediterranean diet and the incidence of CVD. This protection is consistent with previous findings [4,8,9,21e23]. The good a priori credibility of our hypothesis is likely to counterbalance the potential threat of low statistical power. The Mediterranean diet has been proved to be highly effective in reducing cardiovascular risk factors [23], the
metabolic syndrome [24,25], inflammation markers and in improving endothelial function [23]. A strong inverse association between olive oil and myocardial infarction [26] has also been reported. In contrast to the inverse association observed between the a priori definition of the Mediterranean diet and CVD, we found an unexpected non-significant positive association for the PHMP. It is possible that the post hoc approach may not have fully captured important characteristics of the traditional Mediterranean diet [3]. For example, consumption of meat (age-, sex-, energy-adjusted) was not lower but higher (184 g/day) in the upper quintile of the PHMP than in the lowest quintile (173 g/day); ethanol intake was lower in the upper quintile (5.7 g/day versus 7.6 g/day). We acknowledge the limitation related to the small number of observed incident events. However, if the Mediterranean diet is cardioprotective, it seems logical to expect a low CVD incidence in a Mediterranean population. We also acknowledge the limitation of missing information for waist circumference. Another potential limitation of our study might be related to being a non-representative sample of Spain (all participants were university graduates). However, there is no biological argument to suppose that our results might not be generalizable to other populations. In addition, our estimates were adjusted not only for sex but also for the main risk factors for CVD. More importantly, the process of scientific generalization does not need the condition of ‘‘representativeness’’. Furthermore, the pursuit of representativeness can defeat the goal of validly identifying causal relations. To maximize the validity of a cohort it is recommended that study groups be selected for homogeneity with respect to important confounders, for highly cooperative behaviors and for availability of accurate information [27]. Cereals as a group are usually considered protective. However, we found them to be associated with an increased CVD risk. The Mediterranean-diet score makes no distinction between refined cereals and whole grains. Refined cereals (such as white bread, the main staple food in Spain) are responsible for a high glycaemic load, which has been associated with reductions in HDL cholesterol, increased fasting plasma triglycerides, higher fasting insulin and increased levels of C-reactive protein [28]. In any case, it is well known that refined grains increase the glycaemic load whereas
Table 5 Hazard ratios (95% confidence intervals) of cardiovascular disease according to the adherence to empirically derived (‘‘post hoc’’) Mediterranean or Westernized food patterns. The SUN Project, 1999e2008. Adherence to the Westernized Q1 diet (quintiles) Age- and sex-adjusted HR Multivariable HR (95% CI)a
Q2
Q3
Q4
Q5
p for trend
1 (ref.) 1.40 (0.76e2.57) 1.68 (0.91e3.09) 1.75 (0.91e3.36) 2.22 (1.13e4.38) 0.02 1 (ref.) 1.36 (0.74e2.52) 1.70 (0.92e3.13) 1.79 (0.92e3.47) 2.10 (1.06e4.18) 0.02
Adherence to the ‘‘post hoc Mediterranean’’ pattern (quintiles) Age- and sex-adjusted HR 1 (ref.) 1.06 (0.55e2.05) 0.91 (0.46e1.80) 0.94 (0.48e1.85) 1.37 (0.73e2.56) 0.28 Multivariable HR (95% CI)a 1 (ref.) 1.06 (0.54e2.07) 0.84 (0.42e1.69) 0.83 (0.41e1.69) 1.14 (0.54e2.39) 0.79 Q1eQ5: Quintiles of adherence to the respective pattern. a Adjusted for age, sex, family history of coronary heart disease, physical activity, smoking, BMI, diabetes at baseline, use of aspirin, history of hypertension and history of hypercholesterolemia.
Mediterranean diet and cardiovascular disease whole grains are associated with a lower cardiovascular risk [29]. Therefore, the effects of the cereal group on CVD risk may depend on the relative contribution of refined and nonrefined grains. Nowadays in Mediterranean countries the consumption of whole grains is exceptional (89 percent of our cohort members did not consume whole grains on a daily basis and 69 percent of them did not consume them at all). Olive oil consumption continues to be a major source of MUFA in the Mediterranean countries, whereas beef is usually the main source of MUFA in the US. In fact, olive oil was the main source of MUFA, accounting for 38 percent of total MUFA intake in our cohort. The typical pattern of ethanol intake is also different between the Mediterranean area and the US and the main sources of some micronutrients in the US are vitamin supplements but not natural fresh fruits and vegetables. These contrasts highlight the usefulness of our results as being complementary to dietary patterns identified in large American cohorts as a ‘‘Mediterranean-style’’ diet [6,10,30]. Our results suggest that the traditional Mediterranean diet, as it has been usually a priori-defined, confers a substantial cardiovascular protection among initially healthy middle-aged adults.
243
[10]
[11]
[12]
[13]
[14]
[15]
Acknowledgements We thank the participants of the SUN Project for their enthusiastic collaboration, the members of the Sun Project, specially Carmen de la Fuente, for their excellent assistance and the official public Institutes of the Spanish Government for funding health research (Instituto de Salud Carlos III ) and the Department of Health of the Navarra Regional Government for their funding.
[16]
[17]
References [1] Kourlaba G, Panagiotakos DB. Dietary quality indices and human health: a review. Maturitas 2009;62:1e8. [2] Jacques PF, Tucker KL. Are dietary patterns useful for understanding the role of diet in chronic disease? Am J Clin Nutr 2001;73:1e2. [3] Trichopoulou D, Lagiou P. Mediterranean diet and cardiovascular epidemiology. Eur J Epidemiol 2004;19:7e8. [4] Sofi F, Cesari F, Abbate R, Gensini GF, Casini A. Adherence to Mediterranean diet and health status: meta-analysis. BMJ 2008;337:a1344. [5] Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adherence to a Mediterranean diet and survival in a Greek population. N Engl J Med 2003;348:2599e608. [6] Fung TT, Rexrode KM, Mantzoros CS, Manson JE, Willett WC, Hu FB. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation 2009;119:1093e100. [7] Panagiotakos DB, Pitsavos C, Chrysohoou C, Skoumas I, Stefanadis C, ATTICA Study. Five-year incidence of cardiovascular disease and its predictors in Greece: the ATTICA study. Vasc Med 2008;13:113e21. [8] Martı´nez-Gonza ´lez MA, Ferna ´ndez-Jarne E, SerranoMartı´nez M, Marti A, Martinez JA, Martı´n-Moreno JM. Mediterranean diet and reduction in the risk of a first acute myocardial infarction: an operational healthy dietary score. Eur J Nutr 2002;41:153e60. [9] Panagiotakos DB, Pitsavos C, Stefanadis C. Dietary patterns: a Mediterranean diet score and its relation to clinical and
[18]
[19]
[20]
[21]
[22]
[23]
[24]
biological markers of cardiovascular disease risk. Nutr Metab Cardiovasc Dis 2006;16:559e68. Fung TT, McCullough ML, Newby PK, Manson JE, Meigs JB, Rifai N, et al. Diet-quality scores and plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr 2005;82:163e73. Seguı´-Go ´mez M, de la Fuente C, Va ´zquez Z, de Irala J, Martı´nez-Gonza ´lez MA. Cohort profile: the ‘Seguimiento Universidad de Navarra’ (SUN) study. Int J Epidemiol 2006;35: 1417e22. Martin-Moreno JM, Boyle P, Gorgojo L, Maisonneuve P, Ferna ´ndez-Rodrı´guez JC, Salvini S, et al. Development and validation of a food frequency questionnaire in Spain. Int J Epidemiol 1993;22:512e9. Martı´nez-Gonza ´lez MA, de la Fuente-Arrillaga C, NunezCordoba JM, Basterra-Gortari FJ, Beunza JJ, Vazquez Z, et al. Adherence to Mediterranean diet and risk of developing diabetes: prospective cohort study. BMJ 2008;336: 1348e51. Ferna ´ndez-Ballart J, Pin ˜ol Jl, Zazpe I, Corella D, Carrasco P, Toledo E, et al. Relative validity of a semi-quantitative foodfrequency questionnaire in an elderly Mediterranean population of Spain. Br J Nutr (2nd review). Sa ´nchez-Villegas A, Delgado-Rodrı´guez M, Martı´nezGonza ´lez MA, De Irala-Este ´vez J, Seguimiento Universidad de Navarra Group. Gender, age, socio-demographic and lifestyle factors associated with major dietary patterns in the Spanish Project SUN (Seguimiento Universidad de Navarra). Eur J Clin Nutr 2003;57:285e92. Bes-Rastrollo M, Pe ´rez JR, Sa ´nchez-Villegas A, Alonso A, Martı´nez-Gonza ı de masa ´lez MA. Validacio ´n del peso e ´ndice corporal auto-declarados de los participantes de una cohort de graduados universitarios. [Validation of the participants’ self-reported weight and body mass index in a cohort of university graduates]. Rev Esp Obes 2005;3:352e8. Martı´nez-Go ´nzalez MA, Lo ´pez-Fontana C, Varo JJ, Sa ´nchezVillegas A, Martinez JA. Validation of the Spanish version of the physical activity questionnaire used in the Nurses’ Health Study and the Health Professionals’ Follow-up Study. Public Health Nutr 2005;8:920e7. Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2000;32:S498e504. Thygesen K, Alpert JS, White HDfor the Joint ESC/ACCF/AHA/WHF Task Force for the Redefinition of Myocardial Infarction. Universal definition of myocardial infarction. Eur Heart J 2007;28:2525e38. Greenland S. Introduction to Bayesian statistics. In: Rothman KJ, Greenland S, Lash TL, editors. Modern epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams and Wilkins; 2008. p. 334e5. Barzi F, Woodward M, Marfisi RM, Tavazzi L, Valagussa F, Marchioli R. Mediterranean diet and all-causes mortality after myocardial infarction: results from the GISSI-prevenzione trial. Eur J Clin Nutr 2003;57:604e11. De Lorgeril M, Salen P, Martin JL, Monjaud I, Delaye J, Mamelle N. Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study. Circulation 1999;99:779e85. Estruch R, Martı´nez-Gonza ´lez MA, Corella D, Salas-Salvado ´ J, Ruiz-Gutie ´rrez V, Covas MI, et al. Effects of a Mediterraneanstyle diet on cardiovascular risk factors: a randomized trial. Ann Intern Med 2006;145:1e11. Salas-Salvado ´ J, Ferna ´ndez-Ballart J, Ros E, Martı´nezGonza ´lez MA, Fito ´ M, Estruch R, et al. Effect of a Mediterranean diet supplemented with nuts on metabolic syndrome
244 status: one-year results of the PREDIMED randomized trial. Arch Intern Med 2008;168:2449e58. [25] Tortosa A, Bes-Rastrollo M, Sa ´nchez-Villegas A, BasterraGortari FJ, Nu ˜ez-Co ´rdoba JM, Martı´nez-Gonza ´lez MA. Medi´n terranean diet inversely associated with the incidence of metabolic syndrome: the SUN prospective cohort. Diabetes Care 2007;30:2957e9. [26] Ferna ´ndez-Jarne E, Martı´nez-Losa E, Prado-Santamarı´a M, Brugarolas-Brufau C, Serrano-Martı´nez M, Martı´nez-Gonza ´lez MA. Risk of first non-fatal myocardial infarction negatively associated with olive oil consumption: a case-control study in Spain. Int J Epidemiol 2002;31:474e80. [27] Rothman KJ, Greenland S, Lash TL. Validity in epidemiologic studies. In: Rothman KJ, Greenland S, Lash TL, editors.
M.A. Martı´nez-Gonza ´lez et al. Modern epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008. p. 128e47. [28] Liu S, Manson JE, Buring JE, Stampfer MJ, Willett WC, Ridker PM. Relation between a diet with a high glycemic load and plasma concentrations of high-sensitivity C-reactive protein in middle-aged women. Am J Clin Nutr 2002;75: 492e8. [29] Flight I, Clifton P. Cereal grains and legumes in the prevention of coronary heart disease and stroke: a review of the literature. Eur J Clin Nutr 2006;60:1145e59. [30] Iqbal R, Anand S, Ounpuu S, Islam S, Zhang X, Rangarajan S, et al. Dietary patterns and the risk of acute myocardial infarction in 52 countries: results of the INTERHEART study. Circulation 2008;118:1929e37.