The Mediterranean pattern has better randomized evidence behind it than any other named diet. It also produced the most awkward episode in modern nutrition research: PREDIMED, the trial that made its reputation, was retracted by the New England Journal of Medicine in 2018 and republished the same day with a weaker claim. Most articles about this diet skip that entirely. It is the most interesting thing about the trial, and understanding it tells you more about how nutrition evidence works than any list of foods.
What PREDIMED set out to test
PREDIMED enrolled 7,447 adults in Spain, aged 55 to 80, none of whom had cardiovascular disease at entry but all of whom carried risk factors — type 2 diabetes, or at least three of smoking, hypertension, high LDL, low HDL, overweight and family history.
Participants were assigned to one of three arms: a Mediterranean pattern supplemented with free extra-virgin olive oil, the same pattern supplemented with free mixed nuts, or a control group advised to reduce dietary fat. The primary endpoint was a composite of myocardial infarction, stroke and cardiovascular death. Median follow-up was about 4.8 years.
This design matters. The control arm was advised, not fed, and adherence to low-fat advice was poor. PREDIMED therefore compared an intensively supported diet plus free food against ordinary dietary counseling — not one diet against another under equal conditions.
The retraction, and what the corrected analysis found
In 2017 the anesthetist John Carlisle published an analysis of baseline data in 5,087 randomized trials, looking for distributions too tidy or too odd to be plausible if allocation had truly been random. PREDIMED was among the trials whose baseline tables looked wrong.
The investigators re-examined their own records and found three genuine failures, described in the journal’s retraction and republication notice:
- At one of 11 sites, entire clinics were allocated as blocks rather than individual participants randomized.
- Household members of enrolled participants were assigned to the same arm as their relative, without randomization.
- At another site, randomization tables appear to have been used inconsistently.
Together this affected 1,588 of the 7,447 participants — roughly one in five. The 2013 paper was retracted.
The authors reran everything. They excluded the 425 participants from the block-allocated site, treated household members and clinic clusters as correlated rather than independent, and published the corrected analysis as a new paper in June 2018.
The numbers barely moved. Major cardiovascular events were recorded in 288 participants. Compared with the control arm, the hazard ratio was 0.69 in the olive oil group and 0.72 in the nut group — the same rough 30 percent difference reported in 2013.
What changed was the language. The original paper described the Mediterranean diet as reducing events. The republished version reports that the incidence of events was lower among those assigned to the Mediterranean arms, and stops there. That distinction is not lawyerly hedging. With randomization compromised for a fifth of the sample, the trial can no longer fully support the causal reading, however stable the estimate looks. As Harvard’s Nutrition Source noted at the time, the interpretation survived largely intact while the strength of inference did not. Independent researchers gave their own read in the Science Media Centre round-up published alongside the correction.
A claim this blog got wrong
The previous version of this article, written in 2013, told readers the Mediterranean diet “slashed cardiovascular risk factors by 30 percent.” That was wrong twice over. PREDIMED measured cardiovascular events, not risk factors, and the paper it relied on was later retracted. We are withdrawing that sentence.
The same post claimed a resveratrol supplement was equivalent to ten glasses of red wine. No supplement has been shown to reproduce the outcome of a dietary pattern, and equivalence claims of that kind cannot be substantiated. That is withdrawn too.
The trial that came after
The more useful development is a trial that was never retracted. CORDIOPREV, published in The Lancet in 2022, randomized 1,002 adults who already had coronary heart disease to either a Mediterranean pattern or a low-fat pattern and followed them for seven years.
Major cardiovascular events were recorded in 87 participants in the Mediterranean arm and 111 in the low-fat arm — 28.1 versus 37.7 per 1,000 person-years. Adjusted hazard ratios ranged from 0.72 to 0.75 depending on the model. In men the difference was clear (16.2 percent versus 22.8 percent of participants). In women, who made up 17.5 percent of the trial, no between-group difference was detected.
Two independent trials in Spanish populations landing near the same estimate is a meaningful replication. It is still two trials, in one country, in people at elevated risk.
What the pattern actually consists of
Stripped of romance, the traditional Mediterranean pattern is a set of proportions rather than a menu.
| Component | Traditional practice |
|---|---|
| Primary fat | Extra-virgin olive oil, used generously; PREDIMED supplied about one liter per week per household |
| Vegetables and fruit | Several servings daily, largely seasonal and locally grown |
| Legumes | Three or more servings a week, often as the main protein of a meal |
| Nuts | PREDIMED’s nut arm supplied 30 g daily: 15 g walnuts, 7.5 g almonds, 7.5 g hazelnuts |
| Fish | Three or more servings a week, including oily fish |
| Red and processed meat | Occasional, in small portions |
| Wine | Optional, with meals; not a component anyone should start for health reasons |
The nut arm is worth its own reading, since nuts appear independently in the cohort literature — covered in our piece on nuts and long-term health.
Where the evidence is weaker than advertised
Four limitations deserve stating plainly.
The trials were Spanish. Both PREDIMED and CORDIOPREV recruited people already living in a Mediterranean food culture. Adopting the pattern is a smaller change there than in Ohio, and the trials cannot tell you what a larger change does.
Food was donated by industry. PREDIMED’s olive oil came from Spanish olive-oil bodies, and its nuts from the California Walnut Commission, Borges and Morella Nuts. The public funding was independent, but the trial could not have run in that form without the commodity groups whose products it tested.
The control arms were weak. Advice to eat less fat is a low bar. Some of the difference between arms reflects the intensity of support, not the composition of the food.
Wine gets a free pass it has not earned. Moderate wine is part of the traditional pattern, but the observational evidence has weakened as studies have improved. The National Academies’ 2025 review of alcohol and health rated the mortality evidence only moderate in certainty, rated the association with breast cancer as increased risk, and singled out abstainer bias — lumping former drinkers in with lifelong non-drinkers — as a persistent flaw. Nothing in PREDIMED tested alcohol.
The general reasons diet studies are hard to read are set out in our overview of diet and chronic disease.
Cost and practicality
Extra-virgin olive oil, oily fish and nuts are the three most expensive lines on this list, and olive oil prices have been volatile since the 2022–2024 Spanish droughts. The pattern is not automatically cheap.
Where it can be made cheaper: legumes are among the least expensive sources of protein in any American supermarket, dried more so than canned. Frozen vegetables are nutritionally comparable to fresh and cost less. Canned sardines and mackerel deliver the same oily fish as fresh salmon at a fraction of the price. Nuts are cheapest bought raw and in bulk rather than roasted and packaged.
Where it cannot: if olive oil is out of budget, a neutral high-oleic oil is not the same food. In that case the honest answer is that you are following most of the pattern, not all of it.
Common questions
Does the retraction mean the Mediterranean diet doesn’t work?
No. It means the strongest single trial supports its conclusion less firmly than it originally claimed. The estimate held up after correction, CORDIOPREV pointed the same way in a separate population, and the USDA’s systematic review of dietary patterns graded the wider observational evidence as strong. What the episode removes is the certainty, not the finding.
Do I have to drink wine?
No, and no trial has tested it. Wine appears in descriptions of the traditional pattern because it was part of the culture, not because it was isolated and shown to do anything. Nobody who does not currently drink should start.
Is olive oil better than other oils?
Extra-virgin olive oil is what PREDIMED supplied, so it is what was tested. Refined olive oil contains far fewer polyphenols. Whether the difference between extra-virgin olive oil and other unsaturated oils matters at the level of clinical outcomes has not been settled by any trial.
How does this compare with other longevity diets?
The traditional Okinawan pattern was very different — far lower in fat and protein, built on sweet potato — yet the populations following both were long-lived. That comparison, and the serious challenges to the underlying longevity data, are covered in our piece on the Okinawan diet.
What to take from it
The Mediterranean pattern has the best randomized evidence in nutrition, and that evidence is still thinner than the marketing around it. Two trials in Spain, one of them retracted and rebuilt, is a genuine result and a modest one. The components most people can act on — more legumes, more vegetables, more fish, fewer processed meats — are also the cheapest parts of the pattern.
For the days when the vegetable half of that does not happen, a concentrated greens powder is a reasonable convenience rather than a replacement. Greens Plus has been making them since 1993.
About the Author
