A Spanish team measured the phenolic content of 1,239 extra virgin olive oils across two harvests. The gap between the poorest and the richest is a factor of 2,850. All of them carried the same words on the label.
That is the real olive oil story, and it comes last in every article. We debate the benefits of a product category when the variable carrying the effect swings a thousand percent from one bottle to the next, with nothing telling the buyer.
What olive oil does at cohort scale
Two American cohorts followed 92,383 adults for 28 years, over 36,856 deaths. Above 7 g of olive oil a day, half a tablespoon, all-cause mortality is 19% lower, cardiovascular mortality 19%, cancer mortality 17%, and neurodegenerative disease mortality 29%.
These are associations, not demonstrated effects, and people who pour olive oil on their vegetables do not resemble everyone else. The most useful part of the work sits elsewhere: replacing 10 g a day of margarine, butter, mayonnaise or dairy fat with olive oil is associated with an 8% to 34% lower risk of death. What the oil displaces matters as much as the oil.
The landmark trial had to be republished
PREDIMED allocated 7,447 people at high cardiovascular risk between a Mediterranean diet supplemented with extra virgin olive oil, the same supplemented with nuts, and a low-fat control diet. After a median 4.8 years the trial was stopped at an interim analysis: 288 events, with a 31% lower risk in the olive oil arm.
The story behind that figure is worth telling, because nobody prints it. The 2013 article was withdrawn, then republished in 2018 after irregularities came to light: household members enrolled without randomisation, participants assigned without randomisation at one of eleven sites, and apparently inconsistent use of randomisation tables at another.
What it invalidates: the sentence "it is a randomised trial, so the matter is settled". What it leaves standing: the authors redid the analysis without assuming everyone had been randomly assigned, and the estimate held. A result that survives its own correction says more than a result nobody ever re-examined.
Polyphenols, separated from the rest of the oil
A serious objection hangs over all of the above: olive oil is first and foremost monounsaturated fat. Any observed benefit could come from there, with polyphenols as decoration. Exactly one trial has separated the two cleanly.
EUROLIVE gave 200 men, across six centres in five countries, 25 mL a day of three olive oils identical except for phenolic content: 2.7, then 164, then 366 mg/kg. Each participant received all three, three weeks each. Same fat, same dose, only the polyphenols changed.
HDL rose in proportion to content, from 0.025 to 0.032 then 0.045 mmol/L, and the total-to-HDL cholesterol ratio fell in the same order. Oxidised LDL went from +1.21 U/L with the poor oil to -1.48 then -3.21 U/L with the other two. The dose-response is what makes the case, not the size of the numbers.
The one authorised claim, and its threshold
Europe turned that body of work into one sentence and one threshold. The 2012 regulation allows the statement that olive oil polyphenols contribute to the protection of blood lipids from oxidative stress, provided the oil contains at least 5 mg of hydroxytyrosol and its derivatives per 20 g, and the label states that the effect is obtained with 20 g a day.
Five milligrams per twenty grams works out at 250 mg per kilo. Keep that number, it is the only official figure on the subject. And note what the claim does not say: nothing about cholesterol levels, nothing about blood pressure, nothing about heart attacks. It covers the oxidation of blood lipids, which is a mechanism, not a disease avoided.
In 2025, Europe refused to go further
A consortium applied for a more ambitious claim: olive oil polyphenols lower LDL cholesterol and systolic blood pressure, therefore reducing coronary heart disease risk. EFSA reviewed the seven intervention studies submitted and ruled that a cause-and-effect relationship has not been established.
The reasoning is instructive. On LDL, one trial in people with high cholesterol showed a reduction at three weeks, not reproduced by the others, with no data beyond eight weeks and no plausible mechanism. On blood pressure, the mechanism was judged credible but the studies showed no effect. A mechanism without a result and an isolated result without a mechanism do not add up to proof.
Why "extra virgin" says nothing about polyphenols
The grade describes something else. According to the European Commission, extra virgin oil has no sensory defects, is fruity, and its acidity does not exceed 0.8%, against 2% for plain virgin oil. Free acidity measures fat degradation, not richness in phenolic compounds. They are independent axes.
Hence the result that opens this article. Across the 1,239 extra virgin oils analysed, content ranged from 1 to 2,850 mg/kg, and only 4.6% of one season's samples and 23.1% of the other's reached the 250 mg/kg of the claim. Cultivar, harvest date and orchard explain most of the gap. None of those three is required on the label.
| What the bottle says | What it tells you about polyphenols |
|---|---|
| Extra virgin | Almost nothing: acidity and taste defects, two other criteria |
| First cold pressing | Little: a process constraint, not a content |
| Best-before date | Nothing. The harvest date is what counts |
| Harvest date | A lot: phenolics degrade over time |
| Dark glass or opaque tin | A clue: light degrades phenolics |
| A content in mg/kg | Everything, which is why it is rare |
The one test you can run yourself
It is free and it happens in your throat. Freshly pressed olive oil contains oleocanthal, a compound whose pungency causes a stinging sensation in the throat like solutions of ibuprofen. The resemblance is not just a figure of speech: both molecules block the same enzymes in the prostaglandin pathway.
In practice, bitterness on the tongue and the sting at the back of the throat, the one that makes you cough after a spoonful, are the sensory signals of phenolic compounds. A perfectly smooth oil is almost always poor. The reverse is not a measurement: the sting says the oil is not poor, it does not say it clears 250 mg/kg. A cough is not an assay.
Cooking, and the result that surprises
Frying destroys most of the oil's polyphenols. Across consecutive frying cycles the drop reaches about 93% with non-breaded foods, chicken wings or anchovies, and 70% to 78% with breaded ones, with potatoes behaving like the latter.
The counterpoint comes from elsewhere. Potatoes, tomatoes, aubergines and pumpkins were deep fried, sautéed or boiled, in water or in extra virgin olive oil. Frying and sautéing raised the vegetables' total phenolic content, with oleuropein, pinoresinol, hydroxytyrosol and tyrosol transferred from the oil, while both forms of boiling lowered it.
The two results do not contradict each other: part of what the oil loses, the food gains. A 2022 review still cools the enthusiasm, judging the available data largely contradictory and recommending, for want of better, extra virgin oil at low temperature for a short time.
What this article does not say
- That polyphenols lower cholesterol. EFSA examined the question in 2025 and concluded the link is not established.
- That olive oil protects against cancer or dementia. Those figures come from a cohort: associations, among people who differ by far more than their oil.
- That 250 mg/kg is a biological threshold. It is a labelling condition, calibrated on trial doses, not a line where something switches.
- That an oil that stings is necessarily compliant. Taste screens out the poor oils, it does not certify the rich ones.
« On a bottle of olive oil, the most visible words are the ones that tell you the least. »
What Helix does
Olive oil is a clean case of what a food name fails to describe. The same product spans a thousandfold range on its active molecule, one authorised claim, one claim refused three years later, and an effect that depends mostly on what it displaces on the plate. The engine separates those layers: what was measured, on which outcome, at what dose, and what that is worth next to the rest of your stack.