Butter VS. Seed Oils: Which is Worse for Your Health? | Educational Video | Layne Norton PhD

The substitution analysis from the study isolates the effect of replacing butter fat with plant oil fat while holding total fat intake constant. By looking at 10-gram exchanges, the researchers mimicked realistic dietary swaps. The protective association persisted across multiple confounders including physical activity, BMI, smoking, alcohol, and diet quality. Norton emphasizes that this is not an argument to guzzle oil or add it to an already hypercaloric diet; rather, it's about what you replace when choosing between butter and plant-based fats. He notes that the cumulative evidence from replacement trials and cohort studies consistently indicates neutral or beneficial effects of polyunsaturated fats over saturated fats on insulin sensitivity, liver fat, metabolic health, CVD, and cancer. The substitution approach directly addresses the criticism that prior research compared isolated fatty acids, not real foods.

Norton explains that the refining process for seed oils involves removing free fatty acids and other impurities that would otherwise promote oxidation. While heating during processing can transiently increase oxidation, the overall effect of the multi-step refining is a purification that leaves an oil with very low oxidative status. He points out that demonizing a food based on how it's made, rather than on what happens when humans actually consume it, is a logical error. If the claimed oxidative damage were real and clinically meaningful, we would see it in human mortality and disease studies; instead, plant oils show protective associations. Therefore, processing fears are not supported by hard endpoints.

Norton highlights that the anti-seed oil community points to isolated animal studies with supraphysiologic doses of linoleic acid that may increase inflammatory markers. However, in human randomized controlled trials and observational cohorts where saturated fat is replaced by polyunsaturated fat at realistic intakes, the opposite pattern emerges: neutral or improved outcomes. He underscores that epidemiology has its limitations, but when multiple large cohorts across different populations and with extensive confounder control all point in the same direction, and no human data show harm, the mechanistic scare stories are not just incomplete—they're misleading. This protocol is essentially a mental model for evidence evaluation.

The researchers used two statistical models: one that adjusted for basic confounders (showing butter raised mortality risk by up to 40%) and a more fully adjusted final model (showing 15% increased risk). They performed substitution analyses, replaced 10g increments of butter with plant oil, and found 17% lower total and cancer mortality. Subgroup analyses by age, BMI, healthy eating index, smoking, and other confounders consistently showed butter linked to higher mortality. The study broke down plant oils into corn, safflower, canola, soybean, and olive oil. Corn oil showed a non-significant 9% increased mortality risk in the final model (versus 15% significant increase in the less adjusted model), but still far less harmful than butter. All other plant oils showed protective trends, with canola, soybean, and olive oil reaching statistical significance.

Norton investigated the processing steps: refining removes free fatty acids, impurities, and already oxidized lipids. While heating can increase oxidation, the final product's oxidative status changes by less than 1%. The overall net effect of refining is removal of oxidized molecules, so the oil may have lower oxidation than what one would get from less processed sources. He argues that if processing did cause harmful oxidation, that should manifest in human outcome data, but it doesn't. Comparisons to motor oil or chemical solvents ignore that these steps are designed to purify, not contaminate.

He systematically dismantles the argument that seed oils are uniquely deleterious by pointing out that if they were, we would at minimum see no difference compared to butter. Instead, data consistently go in the opposite direction: lower mortality, lower cancer risk, and lower cardiovascular disease risk. He highlights that even across dozens of subgroup analyses and confounder adjustments, the directionality remains the same. The burden of proof, he asserts, lies on those making the claims, and they have not produced any large-scale human studies with negative outcomes for seed oils.

Norton explains that using the same statistical standard across models, butter always emerged as far more harmful. The protective trends for safflower, canola, soybean, and olive oil were consistent. He notes that even if corn oil's 9% increase were real, it pales in comparison to butter's 15–40% increased risk, so steering people toward butter based on corn oil concerns would be unjustified. The study also broke down butter usage (frying, added to foods) and found similar patterns. This granularity strengthens the overall conclusion that plant oils, as a category, outperform butter on hard outcomes.