Was the PREDIMED trial retracted?

Yes, in 2018, and it was republished the same day. The reason was protocol deviation rather than fraud: some household members were enrolled without randomisation and assigned to their relative's arm, and randomisation failed at two of eleven sites. Both halves of what followed matter. The event counts were identical and the hazard ratios moved by 0.01 or not at all. The causal claim weakened, because the new analysis leans on propensity adjustment rather than randomisation alone.

Is the Mediterranean diet good for weight loss?

Only if you restrict calories, and the trials make this unusually clear. PREDIMED did not advise energy restriction and produced essentially no weight change over five years while still reducing cardiovascular events. DIRECT did restrict energy and produced 4.4 kg over two years. The pattern's strength is events, not weight.

Does it actually prevent heart attacks?

It is the only dietary pattern with randomised evidence on hard cardiovascular events, and the evidence is real and qualified. Two trials in Spain found roughly a 28 to 31 percent relative reduction. Both supplied free olive oil to the intervention arms and both were funded in part by olive oil industry bodies, the landmark trial had a randomisation failure, and in the secondary prevention trial the benefit was confined to men.

Which part of it is doing the work?

Nobody knows, because no trial has isolated a component. It was tested as a whole pattern, and in the two biggest trials the thing that was actually handed to participants for free was olive oil or nuts. That is a reason to adopt the pattern rather than to buy one ingredient.

Does it work outside the Mediterranean?

Untested at the level of hard outcomes. PREDIMED and CORDIOPREV both ran in Spain, where the baseline diet was already closer to the intervention than a typical North American or Northern European one. That cuts in an interesting direction: the gap being closed there was smaller, so a larger change elsewhere might do more, or the population might respond differently. Neither has been measured.

Whole-food4 human studies cited

Mediterranean

Written by Aaron CuhaReviewed Sep 2026

Also called: Med diet, MedDiet, PREDIMED pattern

The only pattern with randomised evidence on hard cardiovascular events, and its landmark trial was retracted and republished. The effect estimate survived intact. The strength of the causal claim did not.

What it actually is

Extra virgin olive oil as the main fat, high vegetables, legumes, nuts and fish, moderate wine with meals, low red and processed meat, low refined sugar. In the trials it is not a weight loss diet and was never presented as one: PREDIMED did not advise energy restriction at all and produced essentially no weight change over five years.

The mechanism its advocates propose

A mixture rather than a single lever. Monounsaturated fat replacing saturated fat improves the lipid profile, polyphenols in olive oil and wine are proposed to act on endothelial function and oxidation, nuts supply unsaturated fat and fibre, and oily fish supplies long-chain omega-3. Because it is a whole pattern rather than one intervention, no trial has isolated which part carries the effect, and the two biggest trials both supplied free olive oil.

What happened when calories were controlled

Strongest design available: prescribed, self-reported

Participants were told what to eat and reported back. Under-reporting is the norm rather than the exception, so the actual intake in each arm is unknown.

Rarely, and the distinction determines what a trial means. PREDIMED explicitly did not restrict calories and got cardiovascular benefit without weight loss. DIRECT did restrict them and got weight loss. Conflating those two is the most common error made with this pattern.

DIRECT, Shai 2008, is the calorie-restricted one: 322 adults, two years, three arms. Mean weight loss low fat 2.9 kg, Mediterranean 4.4 kg, low carbohydrate 4.7 kg. Adherence was the best in this whole evidence base, 84.6 percent at two years. Among the 36 participants with diabetes, glucose and insulin changes were more favourable on Mediterranean than low fat (p < 0.001 for the interaction), and n = 36 for that subgroup.
PREDIMED provided free extra virgin olive oil or free mixed nuts to the intervention arms and gave the control arm advice to reduce fat plus small nonfood gifts. That asymmetry matters: the intervention was food, the control was advice.
In the Ge 2020 network meta-analysis of 121 trials and 21,942 participants, weight loss diminished at twelve months across every macronutrient pattern and every named diet, and the cardiovascular risk factor benefits essentially disappeared at twelve months for all interventions except the Mediterranean diet. That is one finding within one analysis and is reported without inflation.

What happened when people just ate it

This is the one pattern where free-living trials measured events rather than markers, and that is why it sits apart. PREDIMED randomised 7,447 people at high cardiovascular risk and followed them a median 4.8 years to a composite of myocardial infarction, stroke or cardiovascular death. CORDIOPREV randomised 1,002 patients who already had coronary heart disease and followed them seven years, finding 87 primary events on Mediterranean against 111 on low fat, with adjusted hazard ratios from 0.719 to 0.753. The Lyon Diet Heart Study, in secondary prevention after a first heart attack, reported 14 cardiac deaths plus non-fatal infarctions against 44 on a prudent Western diet. That last effect size, a risk ratio as low as 0.28, is not reproduced by any modern trial and the study predates current statin and revascularisation standards, so treat it as historically important rather than as the expected effect.

Protein, and whether it confounds the result

Protein is moderate and mostly from fish, legumes and dairy, and the pattern was never designed around it. None of the cardiovascular trials matched or reported protein intake, and none measured lean mass. If you are combining this pattern with a weight loss drug or a serious training programme, this pattern gives you no protein guidance at all and you will need to add it yourself.

The measured intakes behind the protein figures are on the protein page.

What reliably moves

Marker	Direction	From
Major cardiovascular events	Down, in the only pattern with randomised event data	PREDIMED republished hazard ratios 0.69 for olive oil and 0.72 for nuts. CORDIOPREV 0.719 to 0.753 in secondary prevention. Both estimates come from trials with olive oil industry funding and free olive oil supplied to the intervention arms.
Body weight	Barely, or not at all, depending on the trial	PREDIMED did not advise energy restriction and produced essentially no weight change over five years. DIRECT, which did restrict energy, produced 4.4 kg over two years. The pattern is not a weight loss intervention unless calories are restricted.
Systolic blood pressure	Down, and better than vegan on this one marker	In the only head-to-head against a low fat vegan diet, Mediterranean won on systolic pressure by 5.9 mmHg (p = 0.02) while losing on weight and LDL. That trial was run by a plant-based advocacy organisation, which published the result against its own pattern.
Cardiovascular risk factors at twelve months	Retained, when others were not	In a 121-trial network meta-analysis, risk factor benefits essentially disappeared by twelve months for every intervention except this one. One finding, one analysis.

Long term, and hard outcomes

The best long-term data of any pattern here, and it comes with a retraction that has to be reported. PREDIMED's 2013 publication was retracted in 2018. The reason was protocol deviation, not fraud: household members were enrolled without randomisation and assigned to the same arm as their relative, some participants at one of eleven sites were assigned without randomisation, and randomisation tables appear to have been used inconsistently at another. A subset of the trial was therefore not properly randomised, which breaks the assumption a causal reading depends on. The republication is the part most people get wrong in both directions. The event counts are identical: 96, 83 and 109 in the three arms, 288 total. The hazard ratios moved by 0.01 or not at all, from 0.70 to 0.69 for olive oil and 0.72 to 0.72 for nuts, and held after omitting 1,588 participants whose assignments were known or suspected to have departed from protocol. So the effect estimate survived. What did not survive was the strength of the causal claim: the 2018 analysis is adjusted for baseline characteristics and propensity scores rather than resting on randomisation alone, which places part of the evidence closer in character to an adjusted observational comparison. The abstract wording shifted accordingly, from the diet reduced the incidence to the incidence was lower among those assigned. Both halves belong in any honest report of this trial, and most reports carry only one of them.

Citations

Review2018
Retraction and republication: primary prevention of cardiovascular disease with a Mediterranean diet
New England Journal of Medicine
The retraction notice for the 2013 PREDIMED paper. Protocol deviations: household members enrolled without randomisation, some participants at one site assigned without randomisation, and apparent inconsistent use of randomisation tables at another.
Human2018
Primary prevention of cardiovascular disease with a Mediterranean diet supplemented with extra-virgin olive oil or nuts
New England Journal of Medicine
The republished analysis. 7,447 high-risk adults, median 4.8 years, 288 primary events. Olive oil HR 0.69 (0.53 to 0.91), nuts HR 0.72 (0.54 to 0.95), holding after omitting 1,588 participants with suspected protocol departures. Adjusted for baseline characteristics and propensity scores rather than relying on randomisation alone.
Human2022
Long-term secondary prevention of cardiovascular disease with a Mediterranean diet and a low-fat diet (CORDIOPREV): a randomised controlled trial
The Lancet
1,002 patients with established coronary heart disease, seven years. 87 events on Mediterranean against 111 on low fat, adjusted HR 0.719 to 0.753. The effect was concentrated in men (HR 0.669); no difference found in the 175 women. Funded in part by olive oil industry bodies.
Human1999
Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study
Circulation
Secondary prevention after a first myocardial infarction, mean 46 months. Cardiac death plus non-fatal infarction 14 against 44 events, p = 0.0001, with adjusted risk ratios 0.28 to 0.53. Very small event counts and predates modern statin and revascularisation care.
Human2008
Weight loss with a low-carbohydrate, Mediterranean, or low-fat diet
New England Journal of Medicine
DIRECT. 322 adults, two years, and the Mediterranean arm here was calorie restricted, unlike PREDIMED. Weight loss 2.9 kg low fat, 4.4 kg Mediterranean, 4.7 kg low carbohydrate. Adherence 84.6 percent at two years, the best in this evidence base.
Review2020
Comparison of dietary macronutrient patterns of 14 popular named dietary programmes for weight and cardiovascular risk factor reduction in adults: systematic review and network meta-analysis
BMJ
121 trials, 21,942 participants. Weight loss diminished at twelve months across all macronutrient patterns and all named diets, and cardiovascular risk factor benefits essentially disappeared at twelve months for all interventions except the Mediterranean diet.

What people report

These are uncontrolled self-reports, not evidence. They are here because they tell you what to expect and what to watch for, which the trial literature does not. They cannot tell you whether anything works.

Easiest pattern to sustain socially and while travelling, reported more consistently than for any other pattern here, and matched by the best adherence figures in the trial data.
No hunger and no elimination phase, which is why people describe it as a way of eating rather than a diet.
Slow or absent weight loss, which disappoints people who came to it for that and is exactly what the trials predict.
Cost of extra virgin olive oil, nuts and fish is the most common practical complaint.
Frequent uncertainty about what counts, since the pattern is defined by a set of foods rather than a rule, and commercial versions vary widely.

Sources: Trial adherence data, which is unusually strong here at 84.6 percent completion at two years in DIRECT, plus general nutrition forums. Uncontrolled self-report.

Who this is wrong for

Nobody in particular, which is unusual on this page. No specific nutrient deficiency is characteristic of the pattern.
Anyone expecting weight loss without restricting calories. PREDIMED produced essentially none over five years, and the pattern is energy dense and high in fat.
Anyone for whom the wine component is a problem. It is part of the tested pattern and is not a required part of it, and no trial has isolated its contribution.
Anyone reading the results as applying to themselves without noting the populations. PREDIMED and CORDIOPREV both ran in Spain, where baseline habits were already closer to the intervention than a typical North American or Northern European baseline. CORDIOPREV's benefit was confined to men, 67 of 414 against 94 of 413, with no difference found in the 175 women.

Questions

Was the PREDIMED trial retracted?: Yes, in 2018, and it was republished the same day. The reason was protocol deviation rather than fraud: some household members were enrolled without randomisation and assigned to their relative's arm, and randomisation failed at two of eleven sites. Both halves of what followed matter. The event counts were identical and the hazard ratios moved by 0.01 or not at all. The causal claim weakened, because the new analysis leans on propensity adjustment rather than randomisation alone.
Is the Mediterranean diet good for weight loss?: Only if you restrict calories, and the trials make this unusually clear. PREDIMED did not advise energy restriction and produced essentially no weight change over five years while still reducing cardiovascular events. DIRECT did restrict energy and produced 4.4 kg over two years. The pattern's strength is events, not weight.
Does it actually prevent heart attacks?: It is the only dietary pattern with randomised evidence on hard cardiovascular events, and the evidence is real and qualified. Two trials in Spain found roughly a 28 to 31 percent relative reduction. Both supplied free olive oil to the intervention arms and both were funded in part by olive oil industry bodies, the landmark trial had a randomisation failure, and in the secondary prevention trial the benefit was confined to men.
Which part of it is doing the work?: Nobody knows, because no trial has isolated a component. It was tested as a whole pattern, and in the two biggest trials the thing that was actually handed to participants for free was olive oil or nuts. That is a reason to adopt the pattern rather than to buy one ingredient.
Does it work outside the Mediterranean?: Untested at the level of hard outcomes. PREDIMED and CORDIOPREV both ran in Spain, where the baseline diet was already closer to the intervention than a typical North American or Northern European one. That cuts in an interesting direction: the gap being closed there was smaller, so a larger change elsewhere might do more, or the population might respond differently. Neither has been measured.