1. Introduction
The Mediterranean diet (MedDiet) was first defined by Ancel Keys as being low in saturated fat and high in vegetable oils, observed in Greece and Southern Italy during the 1960s [
1]. In the Seven Countries Study this dietary pattern was associated with reduced risk of coronary heart disease (CHD) compared to northern European countries and the United States after 25 years follow-up [
2,
3]. Over the past several decades the study of the MedDiet has advanced, and the definition originally introduced by Keys has evolved and varied. There are a number of ways to define a dietary pattern, including general descriptions, dietary pyramids,
a priori scoring systems,
a posteriori dietary pattern formation, or by food and nutrient content [
4,
5,
6,
7,
8,
9].
Of these,
a priori scoring systems have gained most popularity in the past two decades as they simplify analysis of adherence to the diet in relation to primary outcomes [
10]. Dietary intake is separated into selected food groups related to health outcomes and points are awarded for higher intakes of health-promoting foods and lower intakes of health-harming foods, to calculate a single adherence score. However there are several
a priori Mediterranean diet scores (MDS) with different scoring criteria [
11,
12,
13,
14]. Sofi
et al. [
10] recently compared data from 26 cohort studies utilising some form of MDS, and noted the large range of cut-offs for major food groups such as cereals, even amongst similar populations. When compared on the same nutritional data, 10 different
a priori MDS resulted in a mean adherence ranging from 22.7% to 87.7%, with poor correlation between most indices [
14]. This implies the defining aspects of the MedDiet used to calculate these scores are widely different. Similarly, there are large differences between studies using gram intakes of foods and nutrient content as descriptions/adherence scores. For example, the Greeks in the Seven Countries study consumed an average of 191 g/day of vegetables, in the Prevención con Dieta Mediterránea (PREDIMED) study participants in the intervention group consumed approximately 350 g/day, and the Greeks enrolled in EPIC consumed over 500 g/day [
11,
15,
16].
Marinez-Gonzalez
et al. [
1] suggest that “the very definition of the MedDiet is not a minor issue” (p 10), and point out that two prominent randomised trials investigating health effects of the MedDiet used interventions not fully in line with traditional ideas of the diet, such as the high oil content. A systematic review of intervention trials investigated the relationship between the MedDiet and health outcomes; the authors concluded that there is good evidence the diet improves the lipid profile, endothelial function and blood pressure, but that one of the most limiting factors to drawing conclusions was discrepancies in how the MedDiet had been defined and formulated [
17].
Differences in definitions could be limiting our understanding of the mechanisms by which the MedDiet confers its health benefits. The biological actions of key nutritional components of the MedDiet, such as specific fatty acids, have been studied with promising, although somewhat inconsistent results [
18]. One reason for this may be differences in dose of foods and nutrients between studies. Additionally it is difficult to formulate new MedDiets for intervention studies which are consistent with previous studies, as there is little consistency on which to base these new diets. One potential approach to address such problems is to form a more universal definition by calculating an average quantity of foods and nutrients from previous MedDiets, which would combine traditional and modern examples from relevant studies and provide a benchmark profile of the MedDiet. This definition could be used in future to design intervention MedDiets or MDS which are comparable to other studies. Our objective was to collate information from a range of studies to form a more comprehensive and quantitative definition of the MedDiet than presently exists, by summarising existing definitions and calculating the mean amounts of foods and nutrients.
3. Discussion
The MedDiet has been described similarly for the past five decades, and several pyramids represent the general principles. However, this review found that studies vary considerably when defining the amounts of foods in grams and/or nutrients constituting the MedDiet, although less so when the nutrient profiles are compared.
Dietary constituents of the MedDiet may reduce the risk of CVD and cancer in a dose dependent manner, highlighting the need for greater consistency between studies in the amount of foods and nutrients administered as part of a MedDiet. Sofi
et al. [
37] reviewed the dietary data of the Greek component of the EPIC study, and using segmented logistic regression models evaluated the dose-response relationship between intakes of the nine components of the MDS and overall mortality. There appeared to be an increased risk reduction at two threshold levels for intakes of fruits and nuts, meat and meat products, ethanol, vegetables, cereals and dairy [
37].
Evidence from the present study shows considerable variation in quantity of MedDiet components. The intake of olive oil ranged from 15.7 to 80 mL/day, legumes from 5.5 to 60.5 g/day, vegetables from 210 to 682 g/day and fruits and nuts from 109 to 463 g/day amongst studies. A 5-fold difference in olive oil intake and 10-fold difference in legume intake could have significant implications for specific and all-cause mortality risk. Menotti
et al. [
2] used Seven Countries Study data to examine whether modest variations in food intake predicted changes in CHD death rate. The daily increase for oils (30 g), legumes (30 g), all vegetables (+ 20%, 189 g) and all vegetable food (+ 25%, 237 g) all predicted decreased death from CHD (by 18%, 28%, 28%, and 32%, respectively). Amongst the Spanish population of the EPIC study, for each 10 g increase in olive oil, the hazard ratio was 0.93 for risk of all-cause mortality (95% CI 0.90–0.97) [
38]. Furthermore, sub-analyses from the PREDIMED study showed after 3 months on the intervention, C-reactive protein was significantly decreased in the extra virgin olive oil-enriched arm, but not the nut-enriched arm [
40]. After 12 months, a 24 g/day increase in extra virgin olive oil resulted in a 0.3 μg/L decrease in TNF-α receptor 60 concentration, and a 62.7 g/day increase in vegetable intake resulted in a 0.2 μg/L decrease (
p < 0.05) [
41]. The variety of olive oil intake alone seen across different studies could affect whether the study finds significant effects of the MedDiet.
Quantity of foods appears to impact health outcomes, and forms the basis for most
a priori MDS scoring criteria [
11]. Meta-analytic evidence has shown those consuming more vegetables, fruits/nuts, legumes, cereals and fish, less dairy and meat/poultry and who have a higher MUFA:SFA and consume moderate amounts of ethanol have better cardiovascular and cognitive health than those consuming less [
42,
43]. However, the quantity used to define cut-offs varies between studies; when the 9-point MDS score was first used in 1995 the cut-off for vegetable intake was 303 g/day for men, and when used again in 2003 this increased to 550 g/day [
8,
11]. Differences of such magnitude are likely to substantially alter intakes of bioactive nutrients. Furthermore any subtle improvements in health with increasing intakes may be lost when only one cut-off point is used. There have been recent attempts to improve these scores—Sofi
et al. [
10,
37] in their work have proposed scores with multiple cut-offs and using weighted mean cut-offs from a number of studies. While these newer scores are probably improvements on existing MDS’s, they are still limited by a number of factors, such as failing to recognise major foods like nuts, and differences between studies as to foods are included into each food group. An average nutrient content may be more useful as a basis for forming
a priori scoring systems.
In this review, nutrient content was found to be more consistent across different studies than food quantity. Different foods can provide similar nutrients which allows for preservation of unique foods and dishes observed amongst the different Mediterranean countries while retaining the mechanistic effects of the nutrients and bioactive compounds. Thus there is a distinct advantage to defining the diet by nutrients rather than foods. There are currently no
a priori based scores which use nutrient content exclusively [
11,
19]. Consumption of fatty acids as a percentage of total energy intake, protein, the MUFA to SFA ratio and fibre, vitamins C and E, minerals including selenium and potassium, folate, β-carotene, antioxidant or phytosterol content may be useful nutrients to consider in defining the diet, as these nutrients are consistently implicated as combining for anti-CVD, anti-cancer, anti-aging effects and preventing cognitive decline [
36,
42,
44,
45,
46]. According to this review, on average PUFA intake contributed 4.9% total energy, MUFA 18.4%, SFA 9.0%, the MUFA:SFA was 2.0, fibre intake was 33 g/day, vitamin C 225 mg/day and folate 508.2 μg/day. Notably, it was not possible to derive detailed information on the content of nutrients such as selenium, vitamin E, beta-carotene, long chain omega-3 PUFA or other bioactives such as plant sterols. Expressing nutrient intake as a percentage of total energy is recommended, as those consuming more energy will usually consume more nutrients [
47].
Previously, Sauro-Calixto
et al., formed a definition of the MedDiet based on nutrient intakes of the Spanish population in 1964 [
9]. This definition focused only on four biologically active components of the diet; fibre, total daily antioxidant capacity, MUFA:SFA and phytosterol content. The MUFA:SFA suggested to define the diet was 1.6–2.0, compared to 2.0 in the present study. This appears to be a consistent element of the MedDiet. A defining dietary fibre intake of 41–62 g/day compares to an average intake of 33 g/day found in the present study, with a large variation amongst both interventions and observations [
15,
16]. It is possible that fibre intake has been too low in recent interventions. The other two components of this definition are rarely considered in studies, total daily antioxidant capacity and phytosterol intake. From the four studies included in the review investigating the total flavonoid content, intake is likely to be at least 79 mg/day with an average of approximately 350 mg. Estimates for flavonoid intakes ranged from 79 to 670 mg/day, depending on population studied and whether chemical analysis or databases were used. Indeed there are so many methods for determination of flavonoids that it is not possible to compare studies. Standardization of practices for determination of flavonoids is necessary before we can accurately compare different MedDiets and calculate an approximate range or average [
15,
24].
Because servings of foods tend to be better received than nutrients or grams in public health, we calculated the number of standard Australian serves provided on average by the MedDiet [
30]. Based on the average gram content, the MedDiet provides approximately seven serves of bread, four serves of cereals, five serves of vegetables, 1.5 serves of potato, 1.5 serves of fruit, 0.5–0.75 serves of meat, 0.5 a serve of cheese, and one serve of dairy per day, as well as one serve nuts and three serves of legumes and fish per week. Popular Mediterranean pyramids recommend at least 3–4 weekly serves of nuts, and at least three daily serves of fruit, and usually fewer serves of potato. Considering the averages were based primarily on observation studies these difference are understandable—there appears to be a mismatch between the reality of what Mediterranean populations are eating and pyramid definitions of the MedDiet. This is one limitation of this review, which did not attempt to distinguish between definitions of the MedDiet based on whether they came from observations of diet, or intervention diets.
This review was limited by several other factors. Limited reporting of key nutrient or bioactive molecules has already been mentioned. Only four of 12 food groups had gram values from all 15 data sets. Only energy and fibre was provided by all eight studies, and seven provided the per cent contribution to total energy intake from the macronutrients. Three reported amounts of fats, protein and carbohydrates in grams, and only four reported on other key nutrients including calcium, potassium, phosphate, magnesium, sodium, folate and vitamins A, E and C. There was rarely information provided on sugar, sources of sugar (e.g., desserts or sweets) or wild greens and other herbs, known sources of antioxidants. The average values must be interpreted with some caution.
There were inconsistencies in classification of food groups. For example, the fruit and nuts group consisted only of fruits for intakes reported by Varela-Moreiras
et al. [
23] and Alberti-Fidanza
et al. [
39]. Potentially, separation of fruits and nuts would be worthwhile, as nuts appear to have an independent role in health [
48]. Little information was given on the diet formation when administered as an intervention. It was often unclear whether there was consideration for origin of the diet, which (if any) previous research it had been modelled on, and where and how foods were sourced.
It may be becoming increasingly important to distinguish between observed modern MedDiets and the traditional model based primarily on the Cretan, Greek and Southern Italian diets of 1960 and prior, as countries move towards a more Westernised eating pattern. Most pyramids and general descriptions are still based on traditional practices. However the intervention diets included in the review were in some cases “inspired” by the MedDiet but had distinct differences to typical traditional diets. Ambring
et al. [
33] formulated an intervention diet with less calories and total fat than the control diet, despite the traditional MedDiet typically being high in energy and moderate to high in total fat. Several of the observations were of modern MedDiets, for example Spain in the early 2000s [
23]. If the uniqueness of the traditional diet is lost, the longevity and protection against CHD observed in the Seven Countries Study may also be lost. However, arguably it is prudent to include modernised MedDiets in the definition, which incorporate new health research and allow for changes in food supply and habits, such as was done in the formation of the MDF pyramid [
5]. It may no longer be possible to follow a traditional diet for most populations, especially outside of Mediterranean countries.
4. Materials and Methods
To define the MedDiet a literature review was performed. Databases PubMed [
49], MEDLINE [
50], Science Direct [
51], Academic Search Premier [
52] and the University of South Australia Library Catalogue [
53] were searched using the following search terms; “Mediterranean diet”, “Mediterranean dietary pattern”, Mediterranean, and content, nutrient *, “nutrient content”, definition, define *, pyramid, and “number of serve *”, flavonoid *. Definitions were classified into one of four categories for the purpose of this review: (1) general descriptive definitions; (2) diet pyramids/numbers of servings of key foods and serving size; (3) grams of key foods/food groups; and (4) nutrient and flavonoid content. Total intakes of other phytochemicals, including total polyphenols, phytosterols and carotenoids, were reported by
n < 2 papers which was deemed insufficient information to draw conclusions from. Papers were included in the review if they defined the MedDiet in at least two of the above four ways. This included studies reporting the dietary intake of a Mediterranean population in an observational capacity, presenting a Mediterranean menu designed based on evidence, or studies using a Mediterranean diet as an intervention where the grams of foods and/or nutrient content was reported. All studies were published in English. There was no restriction on study design, date of publication or sample characteristics. MedDiets used for weight loss purpose were excluded due to caloric and food restrictions. Studies reporting less than five nutrients or food group quantities were excluded. Where the same MDS had been applied using identical cut-off values in separate articles, only the original paper was included.
To define the diet by quantity of foods in grams or milllitres, the mean intake was recorded for major foods or food groups. Twelve groups were included (bread, all cereals, legumes, potatoes, all vegetables, fruits/nuts, meat/meat products, cheese, all dairy, eggs, olive oil and fish), based on available data. For the all cereals, all vegetables and all dairy groups, if not originally reported the sum of individual components was used; for all cereals, bread and cereals were combined, for all vegetables, potatoes and other vegetables were combined and for all dairy, cheese, yoghurt and milk were combined. Using standard Australian serving sizes [
30] the gram value was converted to numbers of serves to provide a defining range or number of servings for key food groups. To define the diet by nutrient intake, the mean of all studies reporting at least five of the following was calculated; energy, total fat, SFA, MUFA, protein, percentage energy contributions from total fat, SFA, MUFA, PUFA, protein, carbohydrate, MUFA:SFA, fibre, vitamin C, folate and potassium. The percentage energy contributions of macronutrients and the MUFA to SFA ratio were calculated based on total gram value where not originally reported.