Which oils are safe to cook with at high heat?

What a smoke point is. It is the temperature at which an oil starts to give off visible smoke. Part of that smoke is acrolein, a reactive compound formed when the glycerol in fats breaks down with heat and when polyunsaturated fats oxidize⁹. Smoking oil is a sign you have overheated it. But oils begin to degrade well before they smoke.

Stability is what counts. Heating in air triggers oxidation, hydrolysis and polymerization. The more unsaturated fatty acids an oil has, the faster it oxidizes, and high temperature, many fryings, free fatty acids and metals all speed the process¹⁰. Oxidising polyunsaturated fats form aldehydes, some of them toxic, which pass into the fried food¹¹.

What frying tests show.

• Olive oils lasted 24 to 27 hours of simulated home frying before total polar compounds (a measure of breakdown) hit the 25% legal limit. A vegetable-oil blend lasted 15 hours¹.
• In 80 repeated fryings of potato chips, soybean oil built up the most oxidation products of four fats tested (refined coconut oil, soybean oil, pure olive oil, shortening), and each oil's aldehydes reflected its fatty-acid makeup⁸.
• In kitchen fume tests, sunflower oil released the most aldehydes whatever the method. Palm and rapeseed oil released less. Deep frying released more than pan frying, and pan frying more than stir-frying⁷.

So oils rich in monounsaturated fat (olive) or saturated fat (coconut, palm) are more stable than oils rich in linoleic acid (sunflower, soybean). Ghee and tallow follow the same logic, but we found no head-to-head tests of them. Typical smoke points differ between refined and unrefined versions; see olive oil, avocado oil and seed oils.

Does it matter for health? After a meal rich in fat that had been used for deep frying in a fast-food restaurant, ten men's arteries widened less than after the same meal made with unused fat⁴. A Spanish cohort of 40,757 adults found no link between fried food and heart disease, whether the frying oil was olive or sunflower¹², yet a meta-analysis found higher fried-food intake went with 28% more major cardiovascular events⁶. Fumes matter too: among women who cooked, poor ventilation and waiting for oil to smoke went with more lung cancer^2,3.

Practical rules. Match the oil to the job. Heat it until it shimmers, not until it smokes. Ventilate. Use fresh oil, and don't keep reusing it.

Head-to-head frying studies

In a Portuguese study, oil was sampled every three hours of simulated domestic frying until total polar compounds reached the 25% legal limit. All olive oils lasted 24 to 27 hours, against 15 for a commercial vegetable-oil blend. The extra-virgin oil showed the least oxidation and hydrolysis and the most minor antioxidants, and the vegetable oil, despite more vitamin E, oxidized far faster¹. A Korean study fried potato chips 80 times in refined coconut oil, refined soybean oil, pure olive oil and vegetable shortening. Polyunsaturated fatty acids fell in every oil. Soybean oil had the highest conjugated dienes (21.8 mmol/L) and p-anisidine value (47.7), while total polar compounds ended between 8.1% and 9.5% in all four, under the rejection limit⁸. A Taiwanese team compared four oils (palm, rapeseed, sunflower, soybean), three methods and two foods in an ordinary kitchen: deep frying gave the most aldehydes, stir-frying the fewest, and sunflower oil the most of any oil⁷. These are chemistry measurements from simulations, not measurements of harm.

What eating or breathing the breakdown products does

Ten men had flow-mediated brachial artery dilation measured before and four hours after three meals. After the meal made with used restaurant frying fat, dilation fell from 5.9% to 0.8% (P = 0.0003); it did not change after the same meal with unused fat or after a low-fat meal⁴. The study was tiny, single-meal, and did not standardize the oil. A 2020 review argues that polyunsaturated oils heated for frying generate cytotoxic aldehydes and epoxy-fatty acids, then concedes the "urgent and unmet" need for epidemiological trials linking intake of these compounds to disease¹¹. A 2015 review likewise puts most suspicion on abused frying oils, not on frying in general⁵.

For fumes, a Taiwanese case-control study of 131 nonsmoking women with lung cancer, 252 hospital controls and 262 community controls found adjusted odds ratios of 2.0 to 2.6 for women who waited for oil to emit fumes before cooking and 3.2 to 12.2 for those without a fume extractor². A meta-analysis of 23 observational studies (9,411 cases) found cooking women had about double the odds of lung cancer (OR 1.98 in nonsmokers), poor ventilation raised odds (OR 1.20), and stir-frying was linked to more risk (OR 1.89) than deep frying (OR 1.41, not significant). Heterogeneity was high³. The stir-frying finding runs opposite to the aldehyde-emission ranking above, so emissions do not map neatly onto disease, and most of this research concerns hot wok cooking in Asia.

Fried foods and heart disease

In 40,757 Spanish adults followed for a median 11 years, there were 606 coronary events. The hazard ratio for the highest versus lowest quarter of fried-food intake was 1.08 (0.82 to 1.43), with no difference between olive and sunflower oil users, and none for all-cause mortality¹². A meta-analysis pooled 17 studies: highest versus lowest intake went with relative risks of 1.28 (1.15 to 1.43) for major cardiovascular events, 1.22 (1.07 to 1.40) for coronary heart disease and 1.37 (1.07 to 1.75) for heart failure, but 1.03 (0.96 to 1.12) for all-cause mortality. Heterogeneity was high (I² 82%), and the authors flag recall and misclassification bias⁶. Fried food may be a marker for fast food, chips and calories rather than for the oil.

Where the evidence conflicts or is missing

We found no trial comparing health outcomes of different frying oils. Peer-reviewed head-to-head tests exist for olive, coconut, palm, rapeseed, soybean and sunflower oils, but not for ghee, tallow or avocado oil in the papers we reviewed. Published smoke points vary by refining and test method, so treat them as rough guides.

Bottom line

Confidence is 3 of 5. The ranking of oils by stability is consistent across several laboratory and kitchen studies, and good practice (moderate heat, ventilation, fresh oil) is sensible. Whether the choice of oil changes disease risk is unknown.