Cooking Oils, Heat and Health

Aldehydes in heated oils, oil by oil — what has actually been measured

Public science education · every number traceable to a named study · figures recreated from open-licence (CC BY 4.0) sources · compiled 07 October 2026

1. The short version

Heat cooking oil and its unsaturated fats break down into small reactive molecules called aldehydes. One of them, 4-HNE, is the most studied. It forms from linoleic acid — so the more linoleic acid an oil contains, the more 4-HNE it tends to make when hot. Fried food absorbs these compounds, and you eat them.

That is well established and reproduced across many laboratories. What follows from it is more limited than headlines suggest. Animal studies show repeatedly heated oils raise inflammatory markers. Human evidence linking these aldehydes to high blood pressure and artery disease is observational — it shows association, not cause. And the popular idea that antioxidant supplements protect against any of this has the weakest support of anything on this page: the largest trials found no benefit at common doses, and for some supplements possible harm.

4-HNE
Best-studied toxic aldehyde
formed from linoleic acid
8 oils
with measured 4-HNE
after 5 h at 185 °C
≈11×
spread between the lowest
and highest in that test
0
human studies showing that
eating oxidised oil causes wrinkles

2. What happens when oil gets hot

Oil is mostly triglycerides: a glycerol backbone carrying three fatty acids. The number of double bonds in those fatty acids decides how easily heat breaks them apart.

Heating starts a chain reaction called lipid peroxidation, in two stages:

StageWhat formsWhy it matters
PrimaryHydroperoxides Unstable, very short-lived. They break down almost as fast as they form, so they are not what reaches you in meaningful amounts.
SecondaryAldehydes — MDA, 4-HNE, 4-HHE These are the molecules of concern. More stable than hydroperoxides, absorbed by fried food, reactive towards proteins and DNA.
A common misunderstanding worth correcting. You often read that “free radicals from heated oil enter the body.” Free radicals are far too short-lived for that. What enters the body is the secondary product — the aldehydes. That distinction is the difference between a vague worry and a testable mechanism.
Chain of events from heated oil to health markers
Figure 1. The chain from heating oil to measurable health markers. Evidence strength falls as you move down: the chemistry at the top is solid; the human cardiovascular outcome at the bottom rests on observational data. Source: Leong 2021, Frontiers in Nutrition (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)

3. Which fat becomes which aldehyde

Which fatty acid becomes which aldehyde
Figure 2. Each aldehyde has its own parent fat. Oleic acid — the main fat in olive oil and high-oleic canola — cannot form 4-HNE at all; it breaks down into nonanal and 2-decenal. So for high-heat cooking the useful question is “how much linoleic acid?”, not “how much oleic acid?” Sources: Grootveld 2022 precursor table; Ma 2020 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)

4. The oil-by-oil comparison

This is the core of the page: what has actually been measured, oil by oil. Two things to keep in mind. First, values are only comparable within the same test — different temperatures, times and methods give different numbers. Second, “not measured” means nobody has published that measurement, which is itself worth knowing.

4.1 4-HNE after 5 hours at 185 °C (the most direct comparison)

Measured 4-HNE for eight oils at 185 C
Figure 3. Eight oils, same temperature, same duration. The ordering tracks linoleic acid content: coconut and high-oleic sunflower lowest, grape seed highest. Note the two olive oils sit above high-oleic sunflower oil despite similar linoleic acid, which is why antioxidant content matters too. Sources: Liu 2017 (coconut, palm, safflower, grape seed); Yuan 2018 (olive, sunflower) (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)
OilLinoleic acid
(% of fat)
4-HNE, µg/g
185 °C, 5 h
Source
Coconut oil1–2.510.1Liu 2017
High-oleic sunflower oil2.1–1712.5Yuan 2018
Light olive oil (two samples)3.5–2118.0 / 25.0Yuan 2018
Palm oil9–1228.8Liu 2017
Regular sunflower oil45.4–7440.0Yuan 2018
Safflower oil70–7847.5Liu 2017
Grape seed oil≈ 70113.3Liu 2017
Soybean oil48–5936.9 Bastos 2017 — 180 °C, continuous heating (duration not stated in the abstract)
Canola / rapeseed oil15–30not on this test Formation rate is 6× lower than soybean’s (Figure 4); no 5-hour value published
High-oleic canola oil6.6–14.4not measured Petersen 2013 measured 4-HNE in this oil during 32 h frying, but the numbers are paywalled; a review citing it reports only that sunflower was highest and palm lowest. Flagged as an unresolved gap.
High-oleic soybean oil1.3–3.5not measured No numeric aldehyde value of any kind was found in the literature for this oil
Two honest caveats about that table. First, the coconut, palm, safflower and grape seed values come from a university master’s thesis, not a peer-reviewed journal article — it is the only source that measured these four oils this way, and it is flagged rather than passed off as journal evidence. Second, soybean’s 36.9 µg/g was measured at 180 °C rather than 185 °C, so it is indicative of the same position in the ranking, not a like-for-like number.

4.2 How fast each oil makes 4-HNE

A different way to measure the same thing: how much 4-HNE appears per hour of heating. This covers oils the 5-hour test did not include.

4-HNE formation rate at 185 C by oil
Figure 4. Rate of 4-HNE formation at 185 °C. Corn, soybean and lard — all high in linoleic acid — form it roughly 30 times faster than beef tallow and about 6 times faster than canola. Canola’s lower rate reflects its lower linoleic acid and its much lower linolenic acid. Source: Liu 2014 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)
Fat or oilLinoleic acid (%)4-HNE formation rate
(ng per g oil per hour, 185 °C)
Corn oil34–65.61,473
Lard (animal fat)—1,306
Soybean oil48–591,247
Peanut oil4–43821
Canola / rapeseed oil15–30209
Beef tallow (animal fat)—40

Canola is the interesting case. Its rate is six times lower than soybean’s, and in the original work it produced the lowest 4-HNE of the plant oils tested — lower even than peanut oil, despite peanut oil having more linoleic acid. Canola also produced the highest TBARS value, a different oxidation marker, which the authors attributed to its 9.4 % linolenic acid. Different markers rank the same oils differently; that is a real property of the chemistry, not a contradiction.

Linoleic acid content of common cooking oils
Figure 5. Linoleic acid content — the single most useful guide, because it varies enormously between oils, from about 1 % in coconut oil to nearly 80 % in safflower oil. Bars show the range reported across sources. Sources: Abrante-Pascual 2024 review Table 1; Wroniak 2021; Zhou 2024; Warner & Gupta 2005; Yuan 2018 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)

4.3 Other measurements of the same pattern

4-HNE is not the only aldehyde, and different studies measure different things. Every one of them puts the linoleic-rich oils at the top.

Aldehyde classes in soybean oil over 90 minutes
Figure 6. Soybean oil at 180 °C, aldehyde classes tracked over 90 minutes by NMR. The “4-hydroxy/4-hydroperoxy-2-alkenals” class is the one that contains 4-HNE and 4-HHE; in this study it is measured as a class, not as individual compounds. Note that the unsaturated aldehyde classes rise far faster than the hydroxyalkenals — they are the bulk of the load, even though the hydroxyalkenals are the more reactive. Source: Wann 2021 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)
Total toxic aldehydes by oil
Figure 7. Total toxic α,β-unsaturated aldehydes after 90 minutes at 180 °C. Sunflower and corn lead; olive is lower. Note that canola sits above extra-virgin olive oil here — rankings shift with the metric, which is why no single number tells the whole story. Source: Moumtaz 2019 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)
MeasurementConditionsResult (worst first)Source
Total toxic α,β-unsaturated aldehydes180 °C, 90 min Sunflower 17.2 > corn 15.7 > canola 11.5 > olive 10.6 > algae oil 4.8
mmol per mol fatty acid
Moumtaz 2019
Total unsaturated aldehydes180 °C, 90 min Sunflower 17.22 > corn 15.67 (Moumtaz) > corn 10.10 > soybean 9.67 (Wann) > soybean 6.95
mmol per mol fatty acid — two independent studies agree that soybean and corn are the worst plant oils after sunflower
Wann 2021; Moumtaz 2019
p-Anisidine value175 °C, 18 h, 36 cycles Rice bran 307 > soybean 262 > palm 192 > high-oleic canola (cold-pressed) 179 > high-oleic canola (hot-pressed) 152Zhou 2024
Aldehydes in frying fumes180 °C, 2 h Soybean 3,655 > olive 2,453 ≈ palm 2,197
µg per m³ of air
Chiang 2022
Acrolein in frying fumes180 °C, 2 h Soybean 674 > olive 150 > palm 73
µg per m³ — soybean emits about 9× more than palm
Chiang 2022
Hexanal in frying fumes180 °C, 2 h Soybean 1,033 > palm 636 > olive 550
µg per m³
Chiang 2022
Nonanal in frying fumes180 °C, 2 h Olive 643 > palm 444 > soybean 222
µg per m³ — the oleic-derived aldehyde dominates in olive oil, as the chemistry predicts
Chiang 2022
p-Anisidine value after 18 h frying, three oils
Figure 8. p-Anisidine value — a bulk measure of secondary oxidation products including aldehydes — after 18 hours of deep-frying at 180 °C, plotted against each oil’s linoleic and linolenic acid content. Regular canola, highest in both, exceeded the limit of quantification; high-oleic canola, with less than half the linoleic acid, was far lower; palm olein lowest. The 200 shown for regular canola is a floor value, not an exact reading. Source: Wroniak 2021 (Figure recreated from data published under CC BY 4.0; adapted, not reproduced.)
Read those rows together and a coherent picture emerges. Linoleic-rich oils (soybean, sunflower, corn, grape seed, safflower) produce the most 4-HNE and the most aldehydes generally. Oleic-rich oils (olive, high-oleic sunflower, high-oleic canola, palm) produce much less — and where they do produce aldehydes, they produce different ones, such as nonanal from oleic acid, which is far less reactive than 4-HNE.

5. Three aldehydes worth knowing

NameComes fromWhat it does
MDA
malondialdehyde
Fat oxidation generally; also made in small amounts by the bodyA dialdehyde. Links proteins together and can damage DNA; the commonest marker of oxidative stress in research. Considered the more mutagenic of the three.
4-HNE
4-hydroxy-2-nonenal
Linoleic acid (n-6) The best-studied toxic aldehyde in cooking oil. Reacts very fast with sulfur groups on proteins, and with lysine and histidine. Active at low micromolar concentrations. Considered the more toxic, and can attract immune cells and promote cytokine release.
4-HHE
4-hydroxy-2-hexenal
Linolenic acid (n-3) Close chemical relative of 4-HNE, from the n-3 family. Detected in heated rapeseed and linseed oils; in one study it was detected only in those two and in none of the other oils tested.

6. What the animal studies found

Seven controlled feeding studies gave animals repeatedly heated oils, then measured inflammation markers. Exposure ranged from three to twenty-four weeks.

Oil fedDurationAnimalWhat rose
Repeatedly heated palm oil24 weeksRatsVCAM-1
Repeatedly heated soybean oil24 weeksRatsICAM-1, VCAM-1
Repeatedly heated virgin coconut oil24 weeksRatsICAM-1, VCAM-1, C-reactive protein
Used cooking oil10 weeksRatsIL-6, TNF-α
Cyclic fatty acid monomers + soybean oil28 daysRatsIL-6
Moderately/highly heated soybean oil21 daysBroiler chickensIL-22
Repeatedly heated canola oil4 weeksMiceTNF-α, IL-1β, IL-6, MCP-1, IFN-γ

VCAM-1 and ICAM-1 are adhesion molecules — proteins that help immune cells stick to blood vessel walls, an early step in artery disease. In the original work the authors also reported that adhesion-molecule expression correlated positively with blood pressure.

How much weight this deserves. These are animals, not people; the studies are small; and three of the seven come from one research group using one rat strain, so they are not seven independent confirmations. The direction is consistent and the mechanism is plausible, but this is not proof that eating reheated oil causes heart disease in humans.

7. What we know about people

FindingEvidence typeStrength
People with high blood pressure tend to have higher TNF-α and MDA and lower antioxidant capacity ObservationalModerate Association only
Oxidised LDL is taken up by immune cells to form “foam cells”, which are cytotoxic and drive plaque formationLaboratory mechanism Strong But a mechanism is not an outcome
Dietary oxidised oil causes skin wrinkles— No evidence found Skin ageing from oxidative stress is mainly UV-driven, a different process

8. The antioxidant question

If aldehydes cause oxidative damage, antioxidant supplements should help. This is where the evidence is most counter-intuitive.

Supplement testedWhat the trials found
Beta-caroteneAssociated with increased all-cause mortality (RR 1.07, 95 % CI 1.02–1.11)
Vitamin AAssociated with increased all-cause mortality (RR 1.16, 1.10–1.24) — the largest signal
Vitamin ESmall increase in all-cause mortality (RR 1.04, 1.01–1.07); in a large prostate-cancer trial, 400 IU/day raised prostate cancer incidence
Vitamin C, seleniumNo significant effect on mortality; selenium showed a small protective association in one analysis and is still under study

Two qualifications keep this honest. These mortality signals appeared in the lower-bias subset of trials; pooled across all 68 trials the result was not statistically significant. And a later reanalysis argued that most original trials were not designed to test mortality at all, and that benefits may exist in people who were malnourished to begin with.

What is not in dispute. Foods rich in antioxidants — vegetables, fruit, polyphenol-rich foods — are consistently associated with better health. Extracting those compounds into high-dose pills has not reproduced the benefit. The whole food and the isolated pill are not interchangeable.

9. What to actually do

  1. Don’t reuse frying oil. Heating history matters at least as much as oil choice. Use fresh oil and discard it rather than topping up the same batch.
  2. For high heat, choose an oil low in linoleic acid: coconut, palm, olive, high-oleic sunflower, high-oleic canola. For dressings and low-heat uses the choice matters far less.
  3. Avoid very-high-linoleic oils for deep frying: safflower, grape seed, regular sunflower, corn and soybean are the worst performers in every measured comparison here.
  4. Don’t rely on antioxidant supplements for this. Get antioxidants from food.
  5. Ventilate the kitchen. Frying fumes carry measurable aldehydes including acrolein; soybean oil emitted about nine times more acrolein than palm oil in one study.
  6. Be sceptical of hype in both directions. “Seed oils are poisoning you” overstates observational data. “Only calories matter” ignores a mechanism measured directly and repeatedly.

10. How strong is each claim?

StatementStrengthBasis
Heating oil produces aldehydes, and fried food absorbs themStrongMany direct laboratory measurements
4-HNE comes from linoleic acid; oleic acid cannot form itStrongReaction chemistry and precursor mapping
Oils high in linoleic acid make more 4-HNE when hotStrongMeasured, and reproduced across studies and methods
Linoleic acid content alone predicts the amountModerateNecessary but not sufficient; antioxidants also matter
Repeatedly heated oils raise inflammatory markers in animalsModerateSeven feeding studies, but clustered and all in animals
Aldehyde intake contributes to high blood pressure and artery disease in peopleModerate to weakObservational and mechanistic; causation not established
Antioxidant supplements protect against thisNot supportedLarge trials show no benefit, some harm
Eating oxidised oil causes skin wrinklesNo evidence foundNot studied directly

11. Sources

  1. Leong X-F. Lipid Oxidation Products on Inflammation-Mediated Hypertension and Atherosclerosis: A Mini Review. Front Nutr 2021;8:717740. doi:10.3389/fnut.2021.717740 (CC BY 4.0)
  2. Yuan J, Shoeman DW, Csallany AS. Formation of 4-Hydroxy-2-Trans-Nonenal, a Toxic Aldehyde, in Thermally Treated Olive and Sunflower Oils. J Am Oil Chem Soc 2018;95(7):813–823. doi:10.1002/aocs.12098
  3. Liu X. The effect of temperature and heating time on the formation of alpha,beta unsaturated hydroxyaldehydes in various vegetable oils and fats. M.S. thesis, University of Minnesota, 2014. hdl.handle.net/11299/166843 — formation-rate table
  4. Liu A. Temperature Dependence of 4-Hydroxy-2-Trans-Nonenal (HNE), a Toxic Aldehyde, in Coconut, Palm, Safflower and Grape Seed Oils. M.S. thesis, University of Minnesota, 2017. hdl.handle.net/11299/193414 — 185 °C/5 h values for those four oils
  5. Bastos LCS, de Almeida Costa EA, Pereira PAP. Development, validation and application of an UFLC-DAD-ESI-MS method for determination of carbonyl compounds in soybean oil during continuous heating. Food Chem 2017;218:518–524. doi:10.1016/j.foodchem.2016.09.036
  6. Wann AI, Percival BC, Woodason K, Gibson M, Vincent S, Grootveld M. Comparative 1H NMR-Based Chemometric Evaluations of the Time-Dependent Generation of Aldehydic Lipid Oxidation Products in Culinary Oils Exposed to Laboratory-Simulated Shallow Frying Episodes. Foods 2021;10(10):2481. doi:10.3390/foods10102481 (CC BY 4.0) — soybean aldehyde classes
  7. Moumtaz S, Percival BC, Parmar D, Grootveld KL, Jansson P, Grootveld M. Toxic aldehyde generation in and food uptake from culinary oils during frying practices. Sci Rep 2019;9:4125. doi:10.1038/s41598-019-39767-1 (CC BY 4.0)
  8. Chiang K-M, Xiu L, Peng C-Y, Lung S-CC, Chen Y-C, Pan W-H. Particulate matters, aldehydes, and polycyclic aromatic hydrocarbons produced from deep-frying emissions. npj Sci Food 2022;6:28. doi:10.1038/s41538-022-00143-5 (CC BY 4.0)
  9. Wroniak M, Raczyk M, Kruszewski B, Symoniuk E, Dach D. Effect of Deep Frying of Potatoes and Tofu on Thermo-Oxidative Changes of Cold Pressed Rapeseed Oil, Cold Pressed High Oleic Rapeseed Oil and Palm Olein. Antioxidants 2021;10(10):1637. doi:10.3390/antiox10101637 (CC BY 4.0)
  10. Zhou Z, Gao P, Zhou Y, Wang X, Yin J, Zhong W, Reaney MJT. Comparative Analysis of Frying Performance: Assessing Stability, Nutritional Value, and Safety of High-Oleic Rapeseed Oils. Foods 2024;13(17):2788. doi:10.3390/foods13172788 (CC BY 4.0)
  11. Grootveld M. Evidence-Based Challenges to the Continued Recommendation and Use of Peroxidatively-Susceptible Polyunsaturated Fatty Acid-Rich Culinary Oils for High-Temperature Frying Practises. Front Nutr 2022;8:711640. doi:10.3389/fnut.2021.711640 (CC BY 4.0)
  12. Abrante-Pascual S, Nieva-Echevarría B, Goicoechea-Oses E. Vegetable Oils and Their Use for Frying: A Review of Their Compositional Differences and Degradation. Foods 2024;13(24):4186. doi:10.3390/foods13244186 (CC BY 4.0)
  13. Petersen KD, Jahreis G, Busch-Stockfisch M, Fritsche J. Chemical and sensory assessment of deep-frying oil alternatives for the processing of French fries. Eur J Lipid Sci Technol 2013;115(8):935–945. doi:10.1002/ejlt.201200375 — reported here via secondary citation only; full text paywalled
  14. Warner K, Gupta M. Potato Chip Quality and Frying Oil Stability of High Oleic Acid Soybean Oil. J Food Sci 2005;70(6). doi:10.1111/j.1365-2621.2005.tb11462.x
  15. Ma L, et al. The effect of heating on the formation of 4-hydroxy-2-hexenal and 4-hydroxy-2-nonenal in unsaturated vegetable oils. Food Chem 2020;321:126603. doi:10.1016/j.foodchem.2020.126603 — cited only; all rights reserved, not reproduced
  16. Schaur RJ, Siems W, Bresgen N, Eckl PM. 4-Hydroxy-nonenal — a bioactive lipid peroxidation product. Biomolecules 2015;5(4):2247–2337. doi:10.3390/biom5042247
  17. Bjelakovic G, et al. Mortality in Randomized Trials of Antioxidant Supplements for Primary and Secondary Prevention. JAMA 2007;297(8):842–857; erratum 2008;299(7):765–766. Reanalysed in Biesalski HK et al. Nutrients 2010;2(9):929–949. doi:10.3390/nu2090929
  18. Klein EA, et al. Vitamin E and the Risk of Prostate Cancer: Updated Results of SELECT. JAMA 2011;306(14):1549–1556

12. Licensing and reuse

Every figure on this page was drawn by us from data published in open-access papers, not copied from them. Seven sources are licensed CC BY 4.0, which permits reuse and adaptation with attribution: Leong 2021, Moumtaz 2019, Chiang 2022, Wroniak 2021, Zhou 2024, Grootveld 2022, Abrante-Pascual 2024 and Wann 2021. Attribution appears in each figure caption and in the source list above.

Some sources are all rights reserved and were used for citation only, with no figures or tables reproduced: Ma 2020 (Food Chemistry, Elsevier), Viau 2016 (Eur J Lipid Sci Technol, Wiley) and Petersen 2013 (Eur J Lipid Sci Technol, Wiley). The two University of Minnesota theses are openly available through the university repository.

About this page. A public-education summary built from peer-reviewed sources and two openly available theses. Claims that could not be traced to a verifiable source were removed rather than softened, which is why some familiar assertions do not appear — notably any claim that oleic acid content protects an oil from forming 4-HNE (oleic acid is not a 4-HNE precursor), and any oil ranking produced by estimation rather than measurement.

Limits. Some full texts are paywalled; where only the abstract or a secondary citation could be verified this is stated in the text. Values from different studies use different temperatures, times and methods and are not all directly comparable; the conditions are given with every figure. Fatty-acid percentages vary by variety and processing. Animal findings are labelled as animal findings. This is general information, not medical or dietary advice for an individual.