The Mediterranean pattern: what defines it and what the studies say

What a Mediterranean pattern is

The idea grew out of a mid-20th-century observation: in the Seven Countries Study, populations in the Mediterranean region had lower coronary mortality, probably because of the way they ate.³ That traditional pattern was followed mainly by rural societies with few resources.¹ So it isn't a diet designed in a laboratory, but a description of how people ate in a particular place and time.

Scientific descriptions agree on a core:¹⁴

• A generous plant-based foundation: vegetables, fruit, legumes, grains (preferably whole grains), nuts, seeds and olives.
• Olive oil as the main added fat, for cooking and for dressing.
• Fish and seafood in moderate to high amounts; poultry and eggs in moderate amounts; dairy, mainly cheese and yogurt, in moderate to low amounts; little red meat, processed meat and sweets.
• Herbs, spices, garlic and onion to add flavor with less salt, and preparations such as sofrito: tomato and onion, often with garlic, cooked slowly in olive oil.
• Wine with meals, in moderation, in the classic descriptions. It's the most debated component; we come back to it at the end.

What defines it isn't the list but the whole. The pyramid of the Fundación Dieta Mediterránea, an expert consensus document intended for healthy adults aged 18 to 65, doesn't exclude any food group: according to its authors, what makes a diet healthy or not is variety, proportions, frequency and the way food is cooked. It also includes elements that aren't foods: moderate portions, cooking, eating in company, seasonal produce, regular physical activity and rest.¹ UNESCO, which inscribed the Mediterranean diet as intangible cultural heritage (the current inscription, from 2013, is shared by seven countries, Spain among them), puts the act of eating together at the center.²

Scientific reviews also need an operational definition. The 2019 Cochrane review only considered a trial “Mediterranean-style” if it met two conditions at once: a high intake of monounsaturated relative to saturated fat (from olive oil or nuts, for example) and plenty of plant foods.³ Neither the oil alone nor the vegetables alone were enough.

How studies measure it

To relate a pattern to health, it has to be turned into a number. Studies use scores that add up several food groups: the higher the score, the more closely what a person eats resembles the pattern. In cohort studies, the two most used are the Trichopoulou score, created in 2003 in a Greek cohort, and an “alternative” version created in the United States.⁷ The Spanish PREDIMED trial used another one: a short 14-item questionnaire, the MEDAS.⁵

The variety goes beyond these two. A 2018 overview of reviews (an “umbrella review”) found 34 different scores of “adherence to the Mediterranean diet” across 70 cohort studies; its table lists 18 ways of scoring alcohol alone. When the two most used scores were compared, the associations were generally similar, but the authors call for more consistent definitions so that studies can be compared.⁷ When you read that a study links “following the Mediterranean diet” to something, the useful question is which score it was measured with.

PREDIMED: what was tested, why the first article was retracted and what changed

PREDIMED is a randomized trial carried out at 11 centers in Spain, which recruited participants between 2003 and 2009. It included 7,447 people without cardiovascular disease but at high risk: men aged 55 to 80 and women aged 60 to 80, with type 2 diabetes or with at least three factors such as smoking, hypertension, abnormal cholesterol, overweight or obesity, or a family history. They were divided into three groups:⁴

• Mediterranean diet with extra virgin olive oil: sessions with dietitians and 1 liter (about 34 fl oz) of oil a week per household.
• Mediterranean diet with nuts: sessions with dietitians and 30 g (about 1 oz) a day of a mix of walnuts, hazelnuts and almonds.
• Control group: advice to reduce dietary fat.

Participants weren't asked to cut calories, and physical activity wasn't promoted. The foods were donated by olive oil and nut producers; according to the authors, they played no role in the design, analysis or publication. The primary outcome was a combination of major cardiovascular events: heart attack, stroke or death from cardiovascular causes. The trial was stopped earlier than planned, after a median of 4.8 years, because an interim analysis specified in the protocol crossed the threshold set for stopping.⁴

The first results were published in 2013. In 2017, a statistical analysis pointed out that the distribution of the groups' baseline characteristics wasn't what would be expected from random allocation. The investigators reviewed the process and found deviations: 425 people who lived with another participant were assigned to that participant's group without randomization; at one center, 467 participants were assigned by clinic rather than individually; and at another, everything indicates that the allocation tables were used irregularly. In June 2018 they retracted the 2013 article and published a new one.⁴

Random allocation is what makes it possible to attribute differences to the intervention and not to what the people in each group were like. That's why the new analysis, which according to its authors no longer relies solely on assuming that everyone was randomly allocated, adjusts for baseline characteristics and repeats the calculations without the 1,588 participants whose allocation departed, or may have departed, from the protocol. The results were similar to those of 2013. Adjusting corrects for the differences that were measured, not for those that weren't; the authors calculated that, to explain the entire result, it would take an unmeasured factor present in at least 40% of the control group and in less than 25% of each Mediterranean group, and one that multiplied the risk more than fourfold.⁴ The Cochrane review included the 2018 version but lowered its confidence in it because of these methodological problems.³

With the new analysis there were 288 primary-outcome events. Compared with the control group, the risk was about 30% lower in both Mediterranean groups (hazard ratios of 0.69 with olive oil and 0.72 with nuts). In absolute risk, the difference was 1.7 to 2.1 percentage points over five years.⁴

The same difference can be described as “30% less” or as “about 2 fewer people out of 100 over five years.” Both are correct, and it's worth keeping both in view; why the starting point changes the reading so much is something we explain in relative and absolute risk.

Separately, all-cause mortality did not differ between the groups.⁴ The Cochrane review, which analyzed these data, judged with low certainty that there was little or no effect on total mortality, cardiovascular mortality and heart attacks, and with moderate certainty that there were fewer strokes: from 24 to 14 per 1,000 people over the nearly five years of follow-up.³

Who this evidence was obtained for

A trial gives answers for the people it studies. PREDIMED's participants had a very specific profile:⁴

• Older and at high cardiovascular risk. The mean age was around 67; about half had type 2 diabetes, more than 80% had hypertension, and the mean body mass index was close to 30.⁴
• Almost all white and of European origin (about 97%), recruited in Spain.⁴
• Already eating in a similar way. According to the authors, most were already following a diet similar to the study's Mediterranean diet at baseline, and the control group also received healthy-eating advice. The biggest differences between groups were in the olive oil and nuts they were given, with small increases in fish and legumes.⁴
• With a control condition that changed along the way. In the first years, the control group received a leaflet once a year; from October 2006, the same number of sessions as the other two groups.⁴

The authors themselves point out two limits: it isn't known whether the results can be generalized to people at lower risk, nor to what extent they apply to high-risk people in other countries.⁴ Something else follows from the design: the trial compared full dietary advice plus a supplied food with different advice, so it can't separate the effect of the olive oil or the nuts from the rest of the pattern. The Cochrane review makes a similar point: because PREDIMED supplied foods in addition to advice, it isn't clear what it implies for general recommendations.³

What cohort studies and reviews add

Cohort studies follow thousands of people for years and relate how they eat to what happens to them. In the Greek cohort that gave rise to the Trichopoulou score, followed for a median of 44 months, each 2-point increase on the scale was associated with 25% lower mortality; by contrast, the individual food groups generally showed no significant association.⁶ That fits with the idea of studying the whole, but it's a single cohort and we've only been able to read its abstract.

A 2018 umbrella review brought together 27 meta-analyses of cohort studies, based on 70 studies. Almost all of them found that a higher Mediterranean score was associated with a lower risk of chronic diseases such as type 2 diabetes, cardiovascular disease and cancer. But the credibility of that evidence was rated low to moderate: new studies could change the estimates.⁷

The 2019 Cochrane review brought together 30 randomized trials with 12,461 participants. Only one large trial, PREDIMED, measured cardiovascular events in people without prior disease; most measured risk factors, such as cholesterol or blood pressure, with modest improvements and low or moderate certainty. Its conclusion is that, despite the number of studies, there is still uncertainty about the effects of the pattern and more trials are needed.³

What can be drawn from this, without stretching the data

Some practical ideas can be drawn from all of the above. They're our own inferences, not findings of the studies, nor guidance for anyone in particular:

• Think about the whole, not one ingredient. Neither the definitions nor the scores come down to a single food: they combine a plant-based foundation, type of fat, proportions and way of cooking.¹⁷ A product with the word “Mediterranean” on the package tells you nothing about what you eat as a whole; to compare specific products, it's more useful to know how to read their label.
• Wine is the most debated component. The classic descriptions include it and some scores award a point for drinking it, but in PREDIMED it was optional and only for those who already drank.⁴⁵ Spain's Ministerio de Sanidad points out that there is no safe limit for drinking alcohol.⁹ None of these studies is a reason to start drinking. If you do drink, we explain how it affects rest in alcohol and sleep.
• One change at a time. If you'd like to move closer to this pattern, you can start with a single concrete change, for example cooking with olive oil or adding legumes to one of your meals during the week. How to choose that change and keep it going is explained in the habits guide.

This article doesn't cover diet for people who already have cardiovascular disease, an area where the Cochrane review found little evidence.³ Nor does it turn the pattern into a treatment: the fact that many PREDIMED participants had type 2 diabetes or hypertension describes who was studied, not a plan for those conditions. If you have a health condition, take medication or are considering a major change in your diet, talk to a health professional first.