Seed Oils: The Evidence-Based Deep Dive | Episode 28

Are seed oils (and the polyunsaturated fats they contain) quietly destroying your health… or is the outrage mostly hype and misinformation? In this deep-dive episode, Dr. Layne Norton cuts through the noise with a comprehensive, evidence-first breakdown of the science on seed oils, linoleic acid (LA/PUFA), saturated fat, LDL/ApoB, oxidation, inflammation, and cardiovascular outcomes.

Layne directly addresses the most common claims made by the anti-seed-oil crowd and shows why the totality of high-quality evidence—from mechanistic studies to randomized controlled trials and Mendelian randomization—supports replacing saturated fat with polyunsaturated fats from seed oils for better metabolic and heart health.

What’s covered in this episode:

ApoB & LDL oxidation myth: Why PUFA-enriched LDL is more prone to oxidation per particle, but why that actually misses the bigger picture (lower LDL-P/ApoB, better clearance, less overall atherogenic burden in the arterial wall)
Human RCTs: A balanced review of the major trials (Minnesota Coronary Experiment, Sydney Diet Heart, LA Veterans, Oslo Diet-Heart, Finnish Mental Hospital, STARS, etc.)—including their real strengths, weaknesses, trans-fat confounds, and what the data actually show when you filter properly
Mendelian Randomization evidence: Why lifelong lower LDL/ApoB causally reduces CVD risk regardless of how it’s lowered, and why ApoB/LDL particle number is the key driver (not just LDL-C)
The Linoleic Acid (LA) hypothesis: Does more LA → more arachidonic acid → more inflammation, T2D, CVD, and cancer? Layne examines the actual human data on tissue levels, prostaglandin synthesis, systemic inflammation markers, insulin sensitivity, and disease outcomes
Seed oil processing & refining: Hexane, neutralization, bleaching, deodorization—why the “chemical” scare tactics don’t hold up, peroxide/anisidine values actually drop during refining, and why the final product is far safer and cleaner than the fear-mongering suggests
Naturalistic fallacy & miscellaneous claims: Why “our ancestors didn’t eat seed oils” isn’t a compelling argument, plus other common talking points

Whether you’re team butter, team vegetable oil, or just tired of the tribal nutrition wars, this episode gives you the full context and references so you can make up your own mind based on science instead of social media soundbites.

I. ApoB & LDL

A. Claim by anti-seed oil people: saturated fat is better because it’s less prone to oxidation and it’s OxLDL that is the problem, not native LDL

What is true: PUFA/LA enriched more prone to oxidation per particle
What is missed: ↑ PUFA/LA —> ↓ LDL-P & ApoB 1
- Native LDL-P DOES penetrate the endothelium in a concentration dependant fashion
- Very little LDL oxidation occurs in the plasma due to the presence of anti-oxidants and fast clearance of less than a few hours. Less than 1% of plasma LDL is oxidized. 2
- The vast majority of LDL oxidation occurs in the intima where it can remain for weeks & does not have the same protection with anti-oxidants. 3
Mechanism: ↑ LDL-P —> ↑ LDL-P delivered into the arterial wall. Once in the intima, ApoB is enzymatically modified and can be retained. Once retained, oxidation of LDL can occur. This attracts macrophages which can cause the formation of foam cells. 4 Additionally, treatments that reduce enzymatic modification reduce atherosclerosis. 5
- Yes PUFA enriched LDL can be more easily oxidized once it penetrates the endothelium but this misses the broader point that PUFA DECREASES the amount of LDL that gets into the endothelium, leading to overall less OxLDL formation.
  - So per particle that ENTERS the intima, PUFA enriched particles are more prone to oxidation once trapped in the intima, but because PUFA-rich diets drastically lower ApoB, there are far fewer particles entering the arterial wall in the first place. So oxidation potential per particle inside the intima rises, but the total oxidation burden falls that’s why outcomes favor PUFA
  - A good analogy is thinking about LDL particles like loans & debt risk. PUFA might make each loan a little riskier if it defaults, meaning each LDL particle is slightly more prone to oxidation once it’s in the arterial wall. But replacing saturated fat with PUFA massively cuts the total number of loans & your overall debt load… you’ve got far fewer LDL particles circulating in the first place. So even if each ‘loan’ carries a bit more default risk, your overall balance sheet risk goes way down.
- Also PUFA/LA enriched LDL are better cleared from the plasma, this may be due to a more fluid lipid composition allowing better recognition by the LDL receptor. 6
- Saturated fat enriched LDL is more rigid and not cleared as effectively by the LDL receptor. 4
- ↑ SFA also increases the content of sphingomyelins and ceramids in LDL which promotes greater LDL aggregation. 7
So why is OxLDL higher in plasma for people who have CVD?
- This is DOWNSTREAM of heart disease and plaque development. ↑ LDL-P —> ↑ LDL retained in intima —> ↑ OxLDL in intima. Some of this OxLDL can leave the intima and enter the plasma and that is why ↑ Plasma OxLDL is associated with CVD… reverse causation.
- This is also why studies supplementing with anti-oxidants do not decrease CVD incidence, because plasma is not where LDL is typically oxidized.

II. The Human Randomized Controlled Trials

A. Those Against

Minnesota Coronary Experiment (MCE)
- Strengths: 9000+ subjects. 5 years total. Average enrollment 2 years. Food provided to participants to ensure adherence. Measured CVD outcomes
- Weaknesses: Confounded by Trans Fat inclusion as ‘PUFA.’ Enrollment was often not consecutive as these patients could check in & out of psychiatric care and no control over their diets when they were outpatients
- Findings: No differences between control diet high in SFA vs. intervention diet high in PUFA.
  - Ramsden re-analysis found an increase in all cause mortality of 22% for each 30mg/dl reduction in total cholesterol
    - Likely reverse causation
    - Mortality was actually HIGHER in the control group vs. the intervention group (32% vs. 20%) so the increased risk with reduced cholesterol was IRRESPECTIVE of which diet they were assigned to
Rose Corn Oil Trial
- Strengths: Oil Provided to participants and plasma cholesterol fell in the oil group, indicating adherence. 2 Years long. Conducted in people with ischemic heart disease
- Weaknesses: Low subject number, only 54 people total, 26 in control group, 26 in olive oil, and 28 in corn oil group. There were only 16 deaths in total (3 in control, 5 in olive oil, 8 in corn oil), only 10 cardiac related deaths (1 in control, 3 in olive oil, 6 in corn oil). The confidence intervals were MASSIVE indicating significant variability and low power. They did not do dietary recalls for total SFA or PUFA intake. Additionally, oliva oil as a comparator with more deaths than control is in opposition to TONS of data showing olive oil to be cardioprotective (even Paul will not dispute this)
Sydney Heart Study
- Strengths: 7 years total (3 year average follow up) in 458 men who had a CVD event. Baseline intake of SFA was 16% and PUFA was 6%. The intervention group was advised to increase PUFA to 15% of energy intake and reduce SFA to <10% of total energy intake. PUFA was provided from safflower oil and safflower margarine
- Weaknesses: confounded by trans fats. In fact, in safflower based margins at the time up to 25-40% of the fats were from trans fat. Overall deaths were low with 37 deaths in the intervention group and 28 in the control group
- Findings showed no significant difference in mortality but Ramsden re-analysis demonstrated a higher rate of CVD mortality in the intervention group. But once again the confidence intervals were pretty wide (1.03 to 2.8)
All of these studies had short follow-ups

B. Those for…

LA Veterans study
- Strengths: 846 men in veterans homes. Intake was controlled. Not confounded by trans fats. ~9 year average follow up. PUFA ↑ 5% —>38%. All food was provided to participants ensuring adherence
- Weaknesses: confounded by omega 3 inclusion
- Findings: 31% decrease in CHD events and a non-significant 11% reduction in mortality
Oslo Diet-Heart Study
- Strengths: 412 men post-MI. 5 year average follow up. Not confounded by trans-fats. Intervention group increased PUFA from 8—>20% of energy.
- Weaknesses: Inclusion of omega 3’s confounds outcomes.
- Findings: 47% decrease in subsequent MI events in intervention group
Finnish mental hospital study
- Strengths: 1200 subjects who were institutionalized and had meals provided to them. Crossover design increased power & reduced hospital confounding variables. Subjects did 6 years of each diet. Control diet was 3-4% PUFA & 18% SFA. Intervention diet was 14% PUFA and 9% SFA. Not confounded by omega 3 inclusion & serum fatty acid analysis demonstrated that most of the PUFA increase was linoleic acid
- Weaknesses: Subjects were not randomized individually, but rather one hospital fed one diet then another hospital fed the other. Then at year 6 they crossed over. There could be some confounding variables from the hospital. Control diet was likely higher in trans fats but this was from NATURAL trans fats present in ruminant produced butter
- Findings: during the high PUFA/low SFA periods, CHD deaths were 50% reduced compared to control diet
St. Thomas Atheroschlerosis Regression Study (STARS)
- Strengths: Measured actual angiographic progression of arterial lesions in 1200 subjects. Intervention group decreased total fat from 40% to 27%, SFA from 18% to 6%, and increased PUFA from 5% to 13% mostly from linoleic acid. Assessed blood LDL cholesterol changes (reduced 16%, but HDL unchanged)
- Was free living (but changes in LDL confirm dietary adherence). Confounded by omega 3 inclusion but omega 3 intake was small & likely had minimal impact relative to the large increase in LA
- Findings: Control diet saw progression of atherosclerosis in 39% of people vs. 16% progression in the intervention group. Only 1% of the control group saw regression but 7% of the intervention diet group saw regression. Fewer heart attacks in intervention group but was not statistically significant as it was not powered to detect MI differences
When you filter out trials confounded by trans fats and look only at controlled PUFA-for-SFA substitutions, every high-quality RCT and meta-analysis shows reduced LDL, reduced events, or both fully consistent with Mendelian and pharmacologic evidence. 8 9 10

III. Mendelian Randomization Studies

A. MR studies take advantage of genetic randomization. Allocation of genetic variants is random & therefore provides us with a way to establish causation due to the random assignment of these genes. This makes these MR studies less prone to bias from confounding variables and reverse causation.

Think of a Human RCT. Which treatment group a subject is assigned to is random. The random assignment means that any differences in various other characteristics that could introduce bias (think healthy user bias) will be randomly distributed across the groups. This allows us to assume that any differences between the treatments is due to the treatment itself & not some other characteristic(s).
MR studies work the same way, only instead of scientists randomly assigning people to different treatment groups, individuals are randomly assigned genetic variants at conception. This means that any other traits that might affect the outcomes, including lifestyle and environment will likely also be randomly distributed.
MR studies allow us to look at large numbers of subjects that have genetic variants and their effects on the exposure of various outcomes, in this case LDL/ApoB. MR studies function as de-facto lifelong human randomized controlled trials. 11

B. Very consistent association of LDL with CVD events and death. These associations are consistent regardless of the method of LDL reduction so long as ApoB is also reduced.

C. There are SOME trials that demonstrate a decrease in LDL-C without a decrease in ApoB and these trials fail to show a reduction in CVD incidence.

D. It appears to be the LDL particle number that is the most important determinant of CVD prediction. Since each LDL particle has a single ApoB, ApoB is then a proxy for LDL-P

LDL-P is not regularly measured but it, along with ApoB, & non-HDL cholesterol are all better predictors of CVD than LDL-C. This is because all ApoB containing lipids that are smaller than 70nm in diameter can cross the endothelial barrier. While smaller particles can more easily cross the barrier, they carry less overall lipid, whereas larger LDLs & VLDLs do not as easily cross the barrier but deposit more overall lipid. The net effect is that all ApoB containing particles appear to be equally atherogenic.
In Drug and MR studies where LDL-C is lowered but ApoB is not, there is no reduction in risk. (CETP inhibitors)
In studies where ApoB is normal, but LDL-C is high, risk is normal. In studies where ApoB is high but LDL-C is normal, risk is high
ApoB & LDL-P have an r>0.9 & once ApoB is accounted for in MR studies, there is little association of LDL-C with CVD, indicating it is ApoB/LDL-P that is the causative driver of CVD. 12

E. These studies are all very consistent & show that for every ~39mg/dl reduction in LDL corresponds to a 50-55% decrease in CVD risk and this is regardless of the genetic variant that lowered LDL 13

Variants that affect LDL-P clearance, VLDL synthesis, and absorption all are equally anti-atherogenic at the same dosage even though they exert their LDL-P lowering effects through different mechanisms. The dose response is the same

F. While human RCTs with statins show a less powerful risk reduction, this is due to the lifetime exposure reduction with MR studies vs. reducing exposure later in life. Each ~39mg/dl reduction in LDL is associated with a corresponding ~22% decrease in CVD event incidence. 14 These results hold REGARDLESS of the method used: ezetimibe, ileal bypass, diet & nutrition, bile acid sequesterants, statins, & PCSK9 inhibitors. 15 16 Niacin reduces risk more than expected for the decrease in LDL but niacin also decreases VLDL which contains ApoB. When calculated for the reduction in non-HDL cholesterol (ie ApoB containing particles), the reduction in risk was as expected. The only treatment that did not fall within the expected regression was CETP inhibitors which did not reduce risk. This is due to off target effects such as increasing aldosterone levels & blood pressure, additionally the drop in LDL with these inhibitors is not clinically relevant enough to reduce the risk of CVD. However, in a trial with a drug that did drop LDL by a clinically relevant amount, they did see a reduction in risk. 17 Additionally, when considered alone, genetic variants that mimic CETP inhibitors are associated with concordant changes in LDL-C and apoB. 18 When combined with variants that mimic statins, genetic variants that mimic CETP inhibitors are associated with the same decrease in LDL-C but a substantially attenuated decrease in apoB and a corresponding attenuated effect on ASCVD that is proportional to the attenuated decrease in apoB but substantially less than expected per unit LDL-C.

G. Critics will say MR studies suffer from pleiotropy, but if that were true, we’d expect discordant results across genetic pathways. Instead, we see uniform dose-response relationships across LDL receptor, NPC1L1, HMGCR, and PCSK9 variants same slope, same risk reduction per mmol LDL-C — which strongly argues causality. 19

IV. The Linoleic Acid (LA) Hypothesis

A. Inflammation

Mechanistic Hypothesis: ↑LA —> ↑ Arachidonic Acid (AA) —> ↑ Prostaglandins —> Inflammation —> ↑ T2D, CVD, & Cancer. The LA content of the western diet has increased by 75 fold in the last 150 years and this explains the increase in disease
Reality: Virtually no evidence to support this hypothesis
- Increased LA intake does NOT increase AA formation nor does reducing LA intake lower AA formation in humans in vivo. 20 21
- Changes in LA intake does not affect prostaglandin formation as these pathways are tightly regulated and already nearly at saturation at normal dietary LA intakes so that increased LA intakes simply just lead to greater tissue LA content vs. increased prostaglandin synthesis. 22 23
- Increasing LA intake does NOT increase systemic inflammation and in some cases reduces it. 24 25
- Increasing LA intake does not lead to Type 2 Diabetes and in substitution studies with saturated fat actually leads to better insulin sensitivity and less liver fat. 26 27 28 29 Higher intakes and biomarker levels of LA are also associated with LOWER rates of T2D. 30
- People with greater intake of LA have lower rates of CVD. 31
- People with greater tissue/blood LA content have lower rates of CVD. 32
- Finally, people who eat greater amounts of plant oils vs. butter have lower rates of mortality, CVD mortality, and lower cancer mortality risk. 33

V. Seed Oil Processing

A. Focuses on the processing of the plant oils as a root cause of their purported negative effects and focuses on a few steps in the process. Briefly the steps are 1) cleaning of the seeds for processing preparation 2) mechanical or chemical oil extraction 3) degumming 4) neutralization 5) bleaching 6) deodorization

Oil solvent extraction with hexane: the seeds are washed with hexane which dissolves the oil. The hexane/oil mixture is then treated with steam heat and vacuum which removes the hexane from the oil, producing crude oil. Anti-seed oil people will say that hexane is dangerous and why consume oils with hexane in them? The reality is that the final oil products have virtually no hexane in them & the amounts that are left are far below what could cause any harm. Many oils have no detectable levels of hexane, the majority are well under 1 ppm (~1mg/kg oil) with studies showing anywhere from 0 – 0.05mg/kg hexane. 34 Hexane is mostly a concern for inhalation, with actual oral ingestion far less concerning. No studies assess this in humans, but over 5000mg/kg bodyweight per day in rats (human equivalent dosage would be 811mg/kg body weight) was needed to demonstrate mild liver & nerve toxicity. 1000mg/kg body weight in rats per day (human equivalent dosage 162mg/kg/day) for ~⅛ of their lifespan caused no detectable side effects. In a 70kg human this equates to 11.35g hexane per day whereas the dosage needed to cause mild liver & peripheral nerve toxicity was ~283.8g per day. For relevance, even at the EU limit of 1ppm, a 70kg person would need to consume 11,340kg of oil PER DAY to achieve this amount.
The neutralization process. This is done to remove residual free fatty acids from the oil which can cause foul odor and taste. To achieve this, a small amount of sodium hydroxide (NaOH) & sodium carbonate (Na₂CO₃) is added to the oil. This creates a chemical reaction that produces free fatty acid soaps and water. The soap stock separates from the oil and is removed by centrifugation or settling. It’s then washed with warm water to remove residual soap and placed under vacuum drying to remove moisture. The anti-seed oil rhetoric again focuses on the chemical NaOH which in large amounts is toxic. The max amount found in the end product is 3ppm or ~0.045mg per 15g serving. The amount of NaOH needed to cause caustic injuries (the way NaOH is harmful) in a 70kg person is 700-2100mg or approximately 233kg – 700kg of oil at ONCE.
The bleaching process. Bleaching is done to remove trace impurities like chlorophyll, trace metals, & residual peroxides. The oil is heated under a vacuum to approximately 90-120 degrees celsius to prevent oxidation and small amounts of bleaching clay or activated carbon is added for 20-40 minutes to remove the impurities. The mixture is then run through a filter to remove the clay or activated carbon. Anti-seed oil people mostly focus on the heat and how it could lead to oxidation of the oils which would make them rancid and dangerous for human consumption. This bleaching process is not done by chemical means and no oxidation is involved. The heat is not high enough and since it’s done under a vacuum, no oxidation can occur since no oxygen is present. The amount of clay or activated carbon that remains in the final product is similar to what remains in wine after processing. Even at the maximum amount of clay or carbon remaining in the end product (10ppm) would require you to consume 2550kg of oil at one sitting to produce modest negative side effects
The deodorization process. This process removes volatile compounds that can negatively affect stability, flavor, and odor. The oil is heated at 180-260 degrees celsius under a deep vacuum to prevent oxidation. Once again, anti-seed oil propagandists focus on the heating process and the fact that heat can oxidize oils but the main thing they miss is that since it’s under a vacuum and no oxygen is present, oxidation doesn’t really occur. Secondly, even if no vacuum existed, the heating is only 10-90 minutes. In order to produce significant oxidation of oils you’d need to heat it for far longer. There is a small amount of trans fats that are formed during this process but it’s ~0.5%, meaning you get less than 0.075g of trans fat in a serving. This step also typically produces virtually no oxidation, as mentioned it’s under vacuum.

B. Overall, the refining process actually REMOVES oxidants and impurities.

Peroxide value (measures how much oxidation occurs) and is 3-5 in crude oil and falls to less than 0.5 in the refined oil
Anisidine value (measures aldehydes) is 5-10 in the crude oil and falls to 1-3 after refining
Conjugated dienes are 1.8-2.2nm in crude oil and 1.5-2.0nm after refining
Total polar compounds do increase from 0.1-0.2% to 0.2-0.3% in the end product but this is still far below any amount that could cause negative effects. By comparison, you get more of these compounds in a single serving of fried products than you do from 1.2kg of the refined oil
Polymerized triacylglycerols increase from <0.05% to <0.1% but once again are far below what would cause negative effects. By comparison, you get more of these compounds in a single serving of fried products than you do from 18kg of refined oil.

C. As such, the oil simply does not contain nearly enough negative by-products to even come close to causing problems even if you were guzzling as much of it as you could. This also doesn’t even touch the fact that the human outcome studies using refined oils consistently show neutral to positive effects on CVD, mortality, & metabolic health when substituted for saturated fat.

VI. Miscellaneous BS

A. Naturalistic fallacy

“Our ancestors didn’t eat seed oils.” Reality is we eat nothing like our ancestors and even our meat is modified. And high LDL is not natural. Hadza on average have LDL ~70mg/dl

References

Intro
Personal Bias
Arguments Against Seed Oils
Fatty Acid Structure
Prostaglandin Pathway
Current Health Advice
LDL Cholesterol
Specific Question
Human Trials
David Protein
Sydney Heart Study
Minnesota Coronary Experiment
Rose Corn Oil Trial
LA Veterans Study
Finnish Mental Hospital Study
STARS
Carbon Diet Coach
Analysis of Human RCTs
Mendelian Randomization
LDL Cholesterol and ApoB
Oxidation
Linoleic Acid
Seed Oil Processing
Frying in Seed Oils
Naturalistic Fallacy
Summary