Is Olive Oil Good For You? Yes — But Not Compared To The Oils You’ve Been Told To Fear

Short answer: yes, probably, and we are rating it Supported.

But the famous study everyone cites for it was withdrawn from the New England Journal of Medicine and published again with different numbers — and the largest analysis ever done contains one sentence that quietly deflates the marketing. Neither fact tends to make it into the articles.

The trial you have heard of, and what happened to it

PREDIMED is the reason olive oil has a reputation backed by more than vibes. It was a proper randomised trial: 7,447 Spaniards aged 55 to 80, all at high cardiovascular risk but none with heart disease yet, followed for a median of 4.8 years [1].

In 2013 it reported that a Mediterranean diet with extra-virgin olive oil cut major cardiovascular events. It became the citation. It is on packaging.

Then in 2018 the authors withdrew it.

Their own account of why, and it is worth reading in their words: they found “protocol deviations, including enrollment of household members without randomization, assignment to a study group without randomization of some participants at 1 of 11 study sites, and apparent inconsistent use of randomization tables at another site” [1].

Randomisation is the entire machinery of a trial. It is what makes the groups comparable, and it is the reason a randomised study can claim cause where a survey cannot. If some people were assigned by household — a whole family into one arm — then the groups were not assembled the way the analysis assumed.

So they retracted the paper and published a new one, with, in their words, “revised effect estimates based on analyses that do not rely exclusively on the assumption that all the participants were randomly assigned” [1].

And now the half that the sceptics leave out

If we stopped there you would conclude the olive oil evidence collapsed. It did not.

The re-analysis still found a benefit. Compared with the control diet, the hazard ratio was 0.69 for the olive oil group (95% confidence interval 0.53 to 0.91) [1] — roughly a third fewer events, with the whole plausible range sitting below 1.

More convincing still: they reran it after throwing out 1,588 participants whose assignment was known or suspected to have gone wrong. “Results were similar” [1].

That is close to the best a damaged trial can do. The finding does not depend on the broken part.

What this episode should change is not whether you believe the result. It is how you treat anyone who cites the 2013 paper without mentioning any of this — which, six years on, is most people citing it.

Look at what the three groups actually got

Here is the part that matters more than the retraction, and it was never hidden. It is in the abstract. It just does not suit a headline about olive oil.

Two things jump out.

The nuts did just as well. Events were 3.8% on olive oil and 3.4% on nuts [1], with adjusted hazard ratios of 0.69 and 0.72 — overlapping so heavily that the trial cannot tell them apart. PREDIMED is not evidence that olive oil is special. It is evidence that a Mediterranean diet with a generous fat source beats being told to eat less fat.

And only two of the three groups got food. The intervention arms had extra-virgin olive oil or nuts delivered to them, free, plus quarterly educational sessions. The control group got advice to reduce dietary fat and “small nonfood gifts” [1].

Free groceries and regular contact with people invested in your success is not nothing. It is not a flaw the authors hid — it is the practical reality of running a diet trial — but it means the comparison was never purely about what was on the plate.

The sentence nobody quotes

Away from the trial, the observational evidence on olive oil is genuinely large and genuinely consistent.

The biggest analysis followed 60,582 women and 31,801 men for 28 years, recording 36,856 deaths [2]. People eating the most olive oil — more than half a tablespoon a day — had a 19% lower rate of death from any cause than those who never or rarely ate it (hazard ratio 0.81, 95% CI 0.78 to 0.84). Lower cardiovascular death, lower cancer death, and the largest gap of all in neurodegenerative death, at 0.71.

A separate meta-analysis pooling hundreds of thousands of people found the same direction: per extra 25 g a day, 16% less cardiovascular disease, 22% less type 2 diabetes, 11% lower all-cause mortality [3]. Cancer, notably, showed nothing — a relative risk of 0.94 with a range crossing 1 [3].

Now the sentence. In that 92,000-person analysis, the researchers also asked what happens when you swap other fats for olive oil. Replacing margarine, butter, mayonnaise or dairy fat with it was associated with 8 to 34% lower risk of death [2].

And then: “No significant associations were observed when olive oil was compared with other vegetable oils combined.” [2]

Read that carefully, because it is the practical answer to the question you probably came here with. Olive oil looks good against butter. Against canola, sunflower and the rest, this dataset could not find a difference at all.

Which is the third time we have landed here this week

We have now spent three verdicts on cooking oil, from three separate literatures, and they keep arriving at the same place.

On seed oils, the inflammation mechanism failed when tested, and the trials that swapped saturated fat for vegetable oil never produced a mortality benefit.

On canola, the molecule everyone fears was bred out fifty years ago — and in thirteen head-to-head trials canola lowered LDL cholesterol more than olive oil did.

And here, the largest cohort analysis available cannot separate olive oil from other vegetable oils on mortality.

Three literatures, three methods, one conclusion: among unsaturated cooking oils, which one you buy matters far less than the marketing on either side implies. What moves the numbers is replacing butter, margarine and the fats inside processed food. After that, you are choosing on taste and price, and that is a perfectly good way to choose.

What we are rating, and why not Established

Supported: olive oil is good for you in the sense that matters — higher intake tracks lower cardiovascular disease, lower diabetes and lower all-cause mortality across very large populations, and the one randomised trial, damaged and repaired, still points the same way.

Not Established, for three reasons we will state rather than bury. The trial was retracted and its control arm differed in more than diet. It tested a dietary pattern, with nuts doing equally well. And the cohort data rests on food frequency questionnaires — people recalling what they ate, every four years [2] — in populations where buying olive oil in 1990s America travelled with a good deal else.

Not supported at all: cancer. The pooled estimate is a flat null [3], and the same meta-analysis notes that only its diabetes result was statistically homogeneous — meaning the studies behind the other outcomes disagreed with each other more than you would like.

What to actually do

Use olive oil. It is a good oil. Nothing on this page is an argument against it, and if you enjoy it, that is reason enough.

Do not pay for the story. The gap that the evidence actually shows is between olive oil and butter, margarine and mayonnaise — not between olive oil and the bottle next to it [2].

Half a tablespoon a day is where the top group sat. That is the intake associated with the lowest mortality in the largest cohort [2] — a modest amount, not a regime.

And treat “a randomised trial proved it” as a claim to check, not a conversation-ender. This is the most cited diet trial of the century and it was withdrawn and rebuilt. That is science working — the correction is a feature — but only if somebody tells you it happened.

One thing we have left out deliberately: the claim that a compound in olive oil works like ibuprofen. That rests on enzyme experiments in glassware with no human outcome evidence behind it, which is exactly the error we corrected on our sucralose page. If that changes, we will write it up.

This is journalism, not medical advice. Cardiovascular risk is a conversation for a clinician who knows your numbers.