Chapter 9 — Fat: Saturated, Unsaturated, Trans, Omega-3/6 — What the Evidence Says About Each Type

The Hook: Back to the folder

Two chapters into this book, Theo Vasquez put a green cardboard folder on my desk and asked which of us was lying.

The pair he'd highlighted in two colours was about saturated fat:

"Saturated fat is the primary dietary driver of heart disease. Authorities continue to recommend limiting intake to under 10% of calories."

"Meta-analysis finds no significant association between saturated fat intake and cardiovascular mortality."

In Case Study 1 of Chapter 1 I showed him why both articles were accurate — the meta-analysis compared people eating more saturated fat to people eating less in their ordinary lives, and what those people ate instead was largely refined carbohydrate. The guidance rests on studies where saturated fat was replaced with a specified alternative, usually polyunsaturated fat. Different questions. Different answers. Both correct.

He wrote "compared to what?" on an index card and stuck it inside a kitchen cabinet.

This chapter is where I owe him the full answer.

And I want to be honest up front about why it's the hardest chapter in this book to write. Dietary fat is the topic where nutrition science was most confident, most influential, and most wrong about some things — and where the correction has been so enthusiastically over-applied that a large audience now believes the opposite of the truth with equal certainty.

There are two loud positions. One says saturated fat is the primary cause of heart disease. The other says saturated fat has been exonerated and the whole thing was a fraud.

Both are wrong, and the honest answer sits in a place neither camp finds satisfying — which is that the question, asked without a comparator, has no answer at all.

🏃 Fast Track: §9.2 (the four types), §9.5 (the substitution question), and §9.12 (what to do). Twenty-five minutes.

🔬 Deep Dive: §9.4 (the honest history), §9.6 (saturated fat), §9.8 (omega-3 and omega-6), and §9.10 (what LDL actually is) are where students, clinicians, and anyone arguing about this should spend real time. §9.10 is load-bearing for Chapter 26.

9.1 What a fat actually is

Two paragraphs of chemistry, and they earn their place because the whole chapter turns on one word.

Dietary fat is mostly triglyceride: three fatty acid chains attached to a glycerol backbone (Chapter 6 §6.5). The fatty acid chains are what differ between fats, and they differ in two ways: length and saturation.

Saturation refers to hydrogen. A fatty acid chain is a string of carbon atoms with hydrogens attached. If every carbon carries its full complement of hydrogen — no double bonds between carbons — the chain is saturated: saturated with hydrogen. One double bond makes it monounsaturated. Two or more make it polyunsaturated.

That's it. That's what the word means. It is not a claim about how the fat behaves in your body, and the fact that it sounds like one — saturated, heavy, clogging — has done a remarkable amount of rhetorical work for fifty years.

The practical consequence of saturation is physical: double bonds put a kink in the chain, so the molecules pack together less neatly. Which is why saturated fats are solid at room temperature (butter, lard, coconut oil) and unsaturated fats are liquid (olive, canola, sunflower).

And no food is one type. Butter is roughly 63% saturated and 26% monounsaturated. Olive oil is about 14% saturated. Lard has more monounsaturated fat than saturated. Talking about "saturated fat foods" and "unsaturated fat foods" is already an approximation, and a surprising number of arguments depend on forgetting that.

9.2 The four types, briefly

Before the history, the current state of play, so you know where we're going.

Note that three of four are relatively settled and the argument is about the fourth. That's worth remembering when someone tells you "everything we knew about fat was wrong."

9.3 Trans fats: the one where the system worked

Start here, because it's the clearest case and because it's a genuinely encouraging story that almost nobody tells.

Where they came from. Partial hydrogenation — adding hydrogen to liquid vegetable oils to make them semi-solid — was industrially transformative. It produced margarine and shortening: cheap, shelf-stable, spreadable, ideal for baking. And it was actively promoted as the heart-healthy alternative to butter, because it replaced saturated animal fat with something derived from vegetable oil.

That was the mainstream nutrition position for decades. People switched to margarine on medical advice.

What went wrong. Evidence accumulated from the 1990s onward that industrial trans fats did something no other fat does: they raise LDL cholesterol and lower HDL cholesterol simultaneously, alongside effects on inflammation and endothelial function. On cardiovascular risk, trans fats were substantially worse than the saturated fat they replaced.

What happened next. Regulatory action, slowly. Labelling requirements. Local bans. Then in 2015 the FDA determined that partially hydrogenated oils were no longer "generally recognized as safe" (GRAS), with phase-out following. The WHO launched REPLACE, a global initiative to eliminate industrial trans fats from the food supply.

Industrial trans fats are now largely gone from the food supply in many countries.

💡 Aha moment. This story is used by both camps and neither uses it well.

The critics' version: "They told us margarine was healthy and it was killing us — you can't trust nutrition guidance." True as far as it goes, and it should genuinely lower your confidence in confident dietary advice.

What the version omits: the field detected its own error, quantified it, published it, and drove regulation that removed the substance from the food supply. That took too long and cost real lives. It also worked — and it's roughly the only nutrition intervention in modern history that functioned like a public health success.

The correct lesson is neither "trust nothing" nor "trust everything." It's that dietary advice is a moving estimate that occasionally reverses, and that a field which publishes its reversals is more trustworthy than one that never reports any.

(A footnote for completeness: small amounts of naturally occurring trans fats exist in ruminant fat — beef, lamb, dairy. These are chemically distinct from the industrial ones, are consumed in much smaller quantities, and the evidence does not show the same harms. The verdict above is about industrial trans fats.)

9.4 The diet-heart hypothesis: an honest history

Now the contested part, told without either camp's editing.

The 1950s–60s. Cardiovascular disease was rising sharply in industrialized countries and nobody knew why. Ancel Keys, a physiologist at Minnesota, proposed that dietary saturated fat raised blood cholesterol, which drove atherosclerosis. He designed the Seven Countries Study — a long-running prospective cohort across Finland, Greece, Italy, Japan, the Netherlands, the United States, and Yugoslavia — and it found associations broadly consistent with the hypothesis.

The critique, and what's true in it. You will encounter the claim that Keys "cherry-picked seven countries out of twenty-two" to fit his theory. This deserves precision, because it's the load-bearing accusation in a great deal of contrarian writing and it's a garbled version of something real.

The twenty-two-country figure comes from a different analysis — a critique by Yerushalmy and Hilleboe using available international food-supply and mortality statistics, which showed a weaker association than Keys's data suggested. Keys's Seven Countries Study was a separately designed prospective cohort, and the countries were chosen for practical and methodological reasons — available collaborators, contrasting dietary patterns, feasible data collection — rather than selected from twenty-two after seeing the results.

So: was there a legitimate methodological criticism? Yes — the ecological data was weaker and more confounded than the confidence of the era implied, and Keys was a forceful, sometimes dismissive advocate who did not handle disagreement gracefully. Was it fraud, or selection of countries to fit a conclusion? The evidence does not support that, and repeating it as established fact is exactly the kind of confident overstatement this book is against — including when it's aimed at a target I might otherwise be sympathetic to.

The overstatement that did happen. By the 1980s, the hypothesis had hardened into guidance far more confident than the evidence warranted, and — crucially — it was communicated as "reduce total fat" rather than "replace saturated fat with unsaturated fat." The food industry responded by removing fat and adding sugar and refined starch (Chapter 7 §7.8). That was a real failure with real consequences.

The Sugar Research Foundation episode (Chapter 1 §1.5) sits in this period, and it's real: industry funding shaped the framing of an influential review, undisclosed. It is also routinely asked to carry far more than it can — it does not establish that sixty years of subsequent independent research across dozens of countries was fabricated.

The trials that complicated things. Two long-buried datasets were recovered and re-analyzed decades later: the Minnesota Coronary Experiment and the Sydney Diet Heart Study. Both had replaced saturated fat with vegetable oil high in linoleic acid. Both successfully lowered blood cholesterol, and neither showed the expected mortality benefit — with some analyses suggesting worse outcomes in certain subgroups.

These are genuinely important and genuinely contested. The trials were old, incompletely reported, used oils that may have contained trans fats, and had methodological limitations that make firm conclusions difficult. They should reduce your confidence. They do not overturn the entire body of evidence, and how much weight to give them is a live disagreement among serious people.

9.5 The substitution question, formalized

This is the section that resolves the chapter.

"Is saturated fat bad for you?" is not a hard question. It is a malformed one — it has no answer, in the way "is a number large?" has no answer.

You cannot remove saturated fat from a diet and replace it with nothing. Something takes its place. And the answer changes completely depending on what.

Here is what the evidence actually supports, laid out as substitutions:

Read that table twice. It explains Theo's contradictory headlines completely. It explains why the 1980s guidance failed. And it explains why both camps can cite real evidence indefinitely: the people saying saturated fat is harmful are describing rows one and two, and the people saying it's been exonerated are describing row four.

🔍 Why this works. The reason the substitution frame is so powerful is that it converts an unanswerable question into an answerable one by supplying the missing variable. Consider the same logic elsewhere: "Is it bad to quit your job?" has no answer. "Is it bad to quit your job for a better-paid one?" does. "Is it bad to quit your job for no job?" does too, and the answers differ.

Nutrition spent fifty years arguing about the first form of the question. The moment you insist on the second form, most of the argument dissolves — and what's left is a genuine and much narrower disagreement about the size of the effects.

🔄 Check your understanding. A study reports that replacing 5% of energy from saturated fat with polyunsaturated fat is associated with reduced coronary events. A commentator responds: "But another meta-analysis found no association between saturated fat and heart disease." Are these in conflict?

Answer

No. They're answering different questions, and this is exactly Theo's folder.

The first is a substitution analysis: saturated fat versus a specified replacement. The second is a comparison of habitual intakes — people who eat more saturated fat versus people who eat less, with whatever they actually ate instead, which in most Western cohorts across those decades was substantially refined carbohydrate.

Both findings can be true simultaneously, and both are. The apparent contradiction exists only because the second study's comparator was never stated in the headline.

🧩 Productive struggle. Five minutes on this before reading on. It's the hardest problem in Part II and it's worth the effort.

A country runs a successful public health campaign: "Cut your saturated fat." Over fifteen years, average saturated fat intake falls by a third. Cardiovascular death rates do not improve, and rates of type 2 diabetes rise substantially.

Four people offer explanations. Which are defensible, and which are you missing?

• A: "This proves saturated fat was never the problem."
• B: "The campaign was too small to detect an effect."
• C: "Something else got worse over those fifteen years and masked the benefit."
• D: "The advice was right but the food supply implemented it wrongly."

What I'd say

D is the strongest, and it's the actual history. "Cut saturated fat" was implemented by the food industry as "cut fat, add sugar and refined starch" (Chapter 7 §7.8). Per §9.5's table, replacing saturated fat with refined carbohydrate produces no cardiovascular benefit and plausibly worse metabolic outcomes — which predicts exactly this result, including the diabetes rise. The advice failed because it specified a subtraction and not a substitution.

C is also defensible and compounds D: over any fifteen-year window in an industrializing food environment, ultra-processed food availability, portion sizes, physical activity, and sugar-sweetened beverage intake are all moving, and each could mask an effect.

B is weak but not stupid. A one-third reduction in a single nutrient producing a modest effect on a multi-causal outcome, in an unblinded population-level intervention with imperfect adherence, is genuinely hard to detect. That's Chapter 1 §1.2's small-signal-large-noise problem at national scale.

A is the one that fails, and it's the one most people reach for. It's the ecological version of the unspecified-comparator error — treating a population-level "before and after" as though the only thing that changed was the exposure of interest. Notice that A is the skeptical position, and that being skeptical did not protect it from making the same inferential mistake as the guidance it's attacking.

What most people miss: that D and A are describing the same data, and that the choice between them is not settled by the data. That's why this argument has run for forty years.

9.6 Saturated fat: where the evidence actually lands

So, carefully:

What is well established: saturated fat intake raises LDL cholesterol relative to polyunsaturated fat. This is one of the most reproducible findings in nutrition — controlled feeding studies demonstrate it consistently, and the mechanism is understood.

What is well established: LDL is causally involved in atherosclerosis. This isn't seriously contested among cardiovascular researchers, and the evidence is unusually strong — it comes from cohorts, from randomized drug trials across multiple mechanisms, and from Mendelian randomization (Chapter 2 §2.10) showing that people genetically predisposed to lower LDL have less cardiovascular disease across their whole lives. §9.10 unpacks this.

Where it gets contested: whether dietary saturated fat reduction produces the expected outcome benefit at the population level. Meta-analyses of trials reducing saturated fat have generally found modest reductions in cardiovascular events, with the benefit clearest where saturated fat was replaced by polyunsaturated fat and weaker or absent where it was replaced by carbohydrate.

Cochrane reviews on reducing saturated fat have generally concluded there is a reduction in cardiovascular events, of modest size, with the certainty of evidence rated moderate at best.

🔬 Claim → Evidence → Verdict

The claim: "Saturated fat causes heart disease and should be limited to under 10% of calories."

Where it comes from: A real, reproducible mechanism (saturated fat raises LDL relative to unsaturated fat), a strongly supported causal role for LDL, and trial evidence showing modest event reduction when saturated fat is replaced with polyunsaturated fat.

What the evidence actually shows: The mechanism holds. The outcome effect is real, modest, and entirely dependent on the comparator. Trials replacing saturated fat with polyunsaturated fat generally show benefit; trials or cohorts where it's replaced by refined carbohydrate generally don't. The certainty is moderate rather than high, and the effect size is smaller than the guidance's confidence implies.

📉 Evidence quality: Rung 6–7 for the LDL mechanism (strong). Rung 6–7 for outcomes (moderate certainty, modest effect, comparator-dependent).

Verdict: 🟡 Unclear / it depends — and the "it depends" is not a dodge, it's the actual finding. The recommendation that survives is not "eat less fat." It's "replace saturated fat with unsaturated fat" — which is a different instruction, and the one the 1980s failed to give.

🔬 Claim → Evidence → Verdict

The claim: "Saturated fat has been exonerated. The whole thing was based on fraud and industry money, and butter is back."

Where it comes from: Genuine grievances: the low-fat era's failure, the real Sugar Research Foundation episode, the null meta-analyses of habitual intake, and the recovered Minnesota and Sydney trial data. Every one of these is a real thing, and anyone dismissing them wholesale is not being honest.

What the evidence actually shows: The claim overshoots on every count. The null meta-analyses are the unspecified-comparator problem (§9.5), not exoneration. The LDL mechanism has not been overturned and the causal role of LDL has, if anything, strengthened. The recovered trials are genuinely complicating and genuinely limited. And "it was fraud" requires that six decades of independent research across dozens of countries with incompatible agricultural industries all produced the same error — which is the argument Chapter 1 §1.5 already rejected.

📉 Evidence quality: Real grievances, invalid inference.

Verdict: 🟠 Probably false. Saturated fat has been de-escalated, not exonerated. The honest summary is: the effect is real, modest, comparator-dependent, and was oversold for thirty years — which is a much less satisfying sentence than either camp is selling.

9.7 Dietary cholesterol: the eggs answer

The other pair in Theo's folder, and this one has a cleaner resolution.

The old logic: cholesterol in food raises cholesterol in blood, and blood cholesterol causes heart disease, therefore limit dietary cholesterol. A specific limit — 300 mg/day — was standard guidance for decades.

What we learned: the liver manufactures most of the cholesterol in your blood, and it adjusts production in response to intake. Eat more, make less. The compensation is imperfect but substantial, and for most people dietary cholesterol has a much weaker effect on blood cholesterol than saturated fat does.

The 2015–2020 Dietary Guidelines for Americans removed the specific 300 mg limit, noting that cholesterol was no longer a "nutrient of concern for overconsumption" — while still advising that intake be as low as practicable within a healthy pattern.

Three genuine complications, because this got over-celebrated:

1. Hyper-responders exist. A minority of people show substantial blood cholesterol responses to dietary cholesterol. If your LDL is high and you eat a lot of eggs, it is worth testing rather than assuming.
2. Cholesterol-rich foods often travel with saturated fat. The problem with a fry-up was rarely the egg.
3. People with type 2 diabetes appear in some analyses to show less favourable associations with higher egg intake, and this remains unsettled.

🔬 Claim → Evidence → Verdict

The claim: "Dietary cholesterol raises your blood cholesterol, so limit eggs."

Where it comes from: A plausible mechanism and decades of official guidance built on it.

What the evidence actually shows: Hepatic regulation substantially compensates. Dietary cholesterol's effect on blood lipids is modest for most people and much smaller than saturated fat's. The specific numeric limit was dropped from US guidance in 2015–2020.

📉 Evidence quality: Well studied; the compensation mechanism is established.

Verdict: 🟠 Probably false as a general rule — with real exceptions (hyper-responders, and possibly people with type 2 diabetes). For most people, eggs are a cheap, nutrient-dense protein source and the fifty-year argument was mostly about the bacon.

(Ruth Kaminsky was told to cut eggs and cheese eight years ago, on this reasoning, and was still following it. Chapter 8's Case Study 1 is what that cost.)

9.8 Omega-3 and omega-6

The most confusing corner of the chapter, so let's be orderly.

Both are polyunsaturated. The numbers refer to where the first double bond sits in the chain. Two of them are essential — you cannot make them:

• Linoleic acid (LA) — omega-6. Abundant in vegetable oils, nuts, seeds.
• Alpha-linolenic acid (ALA) — omega-3. In flax, chia, walnuts, canola, soy.

From ALA your body can make the longer-chain omega-3s — EPA and DHA — but conversion is poor: commonly estimated at something like 5–10% for EPA and under 1–5% for DHA, varying with sex, genetics and background diet. Which is why oily fish, which supplies EPA and DHA directly, is treated as a separate recommendation from ALA-rich plants.

The ratio argument

The claim: humans evolved eating roughly equal omega-6 and omega-3; modern diets are 15:1 or 20:1; omega-6 is converted to arachidonic acid, which produces pro-inflammatory mediators; therefore modern diets are inflammatory and omega-6 should be reduced.

What's true: the ratio has shifted enormously with industrial vegetable oil production. Arachidonic acid is a precursor to inflammatory mediators. The pathways compete for the same enzymes.

What the evidence shows: the inference doesn't hold up well in humans. Increasing dietary linoleic acid does not reliably increase arachidonic acid in tissue — the conversion is tightly regulated. Cohort studies and biomarker studies generally find higher linoleic acid intake associated with lower cardiovascular risk, not higher. And trials substituting linoleic acid for saturated fat generally show benefit.

The honest residual: the ratio hypothesis is not absurd, the omega-3 side of it is genuinely supported (most people would benefit from more EPA and DHA), and the absolute amount of omega-3 may matter more than the ratio. Chapter 19 takes the seed oil argument apart in full.

Do fish oil supplements work?

This is where a genuinely interesting mess lives, and it's worth stating honestly.

• VITAL (large trial, ~1 g/day omega-3, primary prevention): largely null for major cardiovascular events overall, with some signals in subgroups.
• ASCEND (people with diabetes, 1 g/day): null for serious vascular events.
• REDUCE-IT (high-dose purified EPA, 4 g/day, high-risk patients on statins): substantial benefit — but the trial used mineral oil as placebo, and the placebo group's LDL and inflammatory markers rose, which has generated serious ongoing debate about how much of the apparent benefit is real.
• STRENGTH (high-dose EPA/DHA, corn oil placebo, similar population): null.

🔬 Claim → Evidence → Verdict

The claim: "Fish oil supplements prevent heart disease. Everyone should take them."

Where it comes from: Strong observational associations between fish consumption and cardiovascular outcomes, a plausible mechanism, and one large trial (REDUCE-IT) showing substantial benefit at high dose.

What the evidence actually shows: At ordinary supplement doses (~1 g/day) in general populations, large trials have been largely null. At high doses in high-risk patients, the two major trials disagree (REDUCE-IT positive, STRENGTH null), and the difference may be the formulation, the population, or the placebo. Meanwhile, eating fish remains associated with better outcomes — which may reflect the fish, what it displaces, or who eats it (Chapter 2).

📉 Evidence quality: Multiple large RCTs, inconsistent, with an unresolved placebo controversy.

Verdict: 🟡 Unclear / it depends. Not supported as general prophylaxis for a healthy person. A reasonable option for people who eat no fish (Walt's case, Chapter 16), and a genuine prescription decision at high dose in specific high-risk patients — with a cardiologist, not a supplement aisle.

🧾 Cost check. The omega-3 recommendation is one of the few places in this book where I suggest a supplement might be reasonable, so it's worth pricing the alternatives honestly.

Route to omega-3	Cost	EPA+DHA delivered
Tinned sardines, 2× week	~$2.40/week = **~$125/year**	Substantial, plus calcium, vitamin D, protein, B12
Tinned mackerel, 2× week	~$3.00/week = **~$156/year**	Substantial
Frozen salmon, 2× week	~$8/week = **~$416/year**	Substantial
Standard fish oil capsules, 1 g/day	~$8–$15/month = ~$96–$180/year	Modest; large trials at this dose are null
Algae-derived EPA/DHA	~$20–$35/month = ~$240–$420/year	Modest; the option for people who eat no fish at all
High-dose prescription omega-3	Prescription	The REDUCE-IT dose — a clinical decision

Two things stand out. Sardines are the cheapest route to omega-3 in existence and deliver several nutrients the capsule doesn't — and they're shelf-stable, require no cooking, and cost about the same per year as the capsules that the large trials found null.

And the algae option, at $240–$420, is genuinely defensible for someone who eats no fish at all — which is the narrow case where I'd suggest it. That's what Walt is, and Chapter 16 revisits it.

9.9 Monounsaturated fat and the pattern with the best evidence

Short section, and it points at the strongest thing in this chapter.

PREDIMED — a large Spanish randomized trial — assigned participants at high cardiovascular risk to a Mediterranean dietary pattern supplemented with either extra-virgin olive oil or mixed nuts, versus a control advised to reduce fat. It found reduced major cardiovascular events in the Mediterranean arms.

You met its history in Chapter 2 §2.10: irregularities in randomization at some sites came to light, the paper was retracted and republished with corrected analysis, and the main conclusions largely held.

What's notable is what PREDIMED was not. It was not a low-fat trial. The Mediterranean arms were higher in total fat than the control. The fat was olive oil and nuts — monounsaturated and polyunsaturated — and the pattern also delivered vegetables, legumes, fish, and less red and processed meat.

Which is why the honest conclusion of this chapter is about patterns rather than about a macronutrient, and why Chapter 10 exists.

9.10 What LDL actually is, and why it matters

A short technical section that Chapter 26 will lean on hard.

Cholesterol doesn't dissolve in blood. It travels inside particles — lipoproteins — which are essentially cargo containers with a protein wrapper.

A crucial distinction: "LDL cholesterol" (LDL-C) measures the cholesterol carried inside LDL particles. ApoB measures the number of particles, because each LDL, VLDL and Lp(a) particle carries exactly one apolipoprotein B.

And it's increasingly clear that particle number predicts risk better than cholesterol content does — because it's the particles that enter and get retained in the artery wall. This matters clinically: some people have normal LDL-C carried in a large number of small particles, and their risk is higher than their standard lipid panel suggests. ApoB testing is inexpensive, not routine in many settings, and worth asking about.

And "good and bad cholesterol" is a poor model. HDL's protective association is robust in observational data, but drugs that raise HDL have repeatedly failed to reduce events, and Mendelian randomization has not supported a simple causal role. HDL is better read as a marker of metabolic health than as a lever.

📊 Diagram (described). Picture the wall of an artery in cross-section, and think of it as a three-layer road: a smooth inner surface (the endothelium, one cell thick), a thicker structural layer beneath it (the intima and media), and blood flowing over the top.

Now picture the blood carrying an enormous number of small spheres — LDL particles, each wrapped in one molecule of apolipoprotein B, each carrying cholesterol as cargo. Most of them simply flow past. But the endothelium is not a sealed barrier; particles cross into the wall and cross back out, continuously, in both directions.

The problem starts when a particle enters and doesn't leave. Inside the wall, LDL particles can be retained — snagged by structures in the intima — and once retained, they are modified and oxidized. Immune cells arrive to clear them, engorge themselves, and become foam cells. Foam cells accumulate into a fatty streak, the fatty streak becomes a plaque, the plaque grows over decades under a fibrous cap, and the clinical event happens when that cap ruptures.

Two things follow from the picture, and they're the reason this section exists.

First: it's a numbers game. The more particles flowing past, the more entries, and the more retentions — which is why particle count (ApoB) predicts risk better than cholesterol cargo (LDL-C). Two people with the same LDL-C can be carrying very different numbers of particles, and the one with more is having more encounters with the wall every second of their life.

Second: it's cumulative. This is a process running for decades before anything is felt. That's why Mendelian randomization is so powerful here — people genetically predisposed to slightly lower LDL have that small advantage from conception, and their lifetime risk reduction is far larger than a comparable reduction started at fifty. Exposure is dose × time, and the time term is doing most of the work.

⚠️ When to see a professional. Lipid management is a clinical decision, not a dietary one. If your LDL or ApoB is elevated, that's a conversation with a physician about your total risk — family history, blood pressure, smoking, diabetes, Lp(a) — and about whether medication is warranted. Diet moves LDL modestly; statins and related drugs move it a great deal, and the evidence for them on hard outcomes is among the strongest in medicine. A book about food should be clear about the limits of food.

9.11 What to actually cook in

The practical question, and a preview of Chapter 19.

Two properties matter, and people usually track the wrong one.

Smoke point is the temperature at which an oil visibly smokes. It's what everyone quotes, and it's the less important of the two.

Oxidative stability — resistance to forming oxidation products when heated — matters more. And it tracks saturation: more double bonds means more sites for oxidation. So highly polyunsaturated oils are the least stable, and saturated and monounsaturated oils the most.

This produces a mildly counterintuitive result: extra-virgin olive oil performs better under household cooking conditions than its modest smoke point suggests, because its monounsaturated profile and its polyphenol antioxidants give it good oxidative stability.

🔬 Claim → Evidence → Verdict

The claim: "Coconut oil is a health food — the MCTs are metabolized differently and it raises 'good' cholesterol."

Where it comes from: Genuine facts, applied loosely. Medium-chain triglycerides genuinely are metabolized differently — absorbed via the portal vein rather than as chylomicrons, more readily oxidized. And coconut oil does raise HDL.

What the evidence actually shows: Coconut oil is roughly 80–90% saturated fat, mostly lauric acid, which is a 12-carbon fatty acid and behaves metabolically much more like a long-chain fat than like the true MCTs (C6–C10) that the research is about. Controlled trials consistently find that coconut oil raises LDL relative to unsaturated plant oils — the American Heart Association issued an advisory to this effect. The HDL rise is real and, per §9.10, not the lever it's presented as.

📉 Evidence quality: Consistent controlled feeding trials on lipids; no outcome trials supporting the health claims.

Verdict: 🟠 Probably false as a health food. It is a fat with a distinctive flavour that is useful in some cooking, and there is nothing wrong with using it for that. It is not a cardiovascular intervention, and the MCT argument is about a different set of molecules.

🔄 Check your understanding. Someone shows you two oils and asks which is "better for high-heat cooking": refined avocado oil (smoke point ~270°C, mostly monounsaturated) or extra-virgin olive oil (smoke point ~190–210°C, mostly monounsaturated with polyphenols). What's the honest answer?

Answer

For genuinely high-heat applications — searing, deep frying, wok cooking — refined avocado oil has the advantage, and the smoke point is the relevant property there.

For almost everything a home cook actually does — sautéing, roasting at 180–200°C, frying an egg — the difference is largely irrelevant, and extra-virgin olive oil performs better than its smoke point suggests because oxidative stability is the property that matters and its polyphenols contribute to it.

The honest answer also includes: avocado oil costs three to five times as much, the adulteration rate in the avocado oil market has been a documented problem, and the population with the best cardiovascular trial evidence (PREDIMED, §9.9) cooked in olive oil.

The deeper point: "which oil is best for high heat" is a question people ask instead of the question that matters, which is what fat am I using for the 95% of my cooking that isn't high heat? Optimizing the rare case while the default goes unexamined is a recurring shape.

9.12 What to actually do

Six things, in descending order of evidential strength.

1. Avoid industrial trans fats. Mostly done for you. Check ingredient lists for "partially hydrogenated" if you're in a jurisdiction where they persist.

2. Replace, don't remove. The instruction the 1980s got wrong. Swap butter, lard, and coconut oil for olive, canola, or other unsaturated oils where it doesn't cost you the dish. This is the best-supported dietary fat recommendation in existence.

3. Eat fish, especially oily fish, if you eat fish. Around two servings a week is the common recommendation. Salmon, mackerel, sardines, herring — sardines being by far the cheapest.

4. If you don't eat fish, address omega-3 deliberately. ALA from walnuts, flax, chia, canola, plus consider an algae-derived EPA/DHA supplement — which is the one supplement recommendation in this chapter, and it's conditional.

5. Don't fear whole-food fats. Nuts, seeds, olive oil, avocado, oily fish, and eggs are among the most consistently health-associated foods in the literature. The fat argument was never about these.

6. Cook in olive oil by default. Cheap, stable, versatile, and it's what the population with the best trial evidence cooks in.

🍽️ On your plate — what I told Theo. His triglycerides were 186 and his HDL 38, which together point at metabolic dysfunction rather than at butter (Chapter 26). We changed three things: cook in olive oil instead of butter, sardines or salmon twice a week, and keep the eggs — which he had been avoiding since 2016 on advice that had been withdrawn.

Then I told him the thing that mattered most: his fat intake was not his problem. His alcohol was 190 kcal a day, his fiber was 14 g against a target of 38, and he was 490 kcal a day in surplus. The saturated fat question that had consumed two colours of highlighter was, for him, close to the least important variable on the page.

What we don't know

We don't know the size of the saturated fat effect with precision, and honestly we may never. The definitive trial — decades long, adequately powered for hard outcomes, with maintained dietary separation — has never been run and probably can't be (Chapter 1 §1.2). What we have is a strong mechanism, moderate-certainty trial evidence of modest effect, and a comparator problem that makes observational data hard to interpret.

We also don't know why REDUCE-IT and STRENGTH disagree. The mineral oil placebo issue is a plausible explanation and it is not settled, and until it is, high-dose omega-3 sits in genuine uncertainty.

🪞 Learning Check-In

Third of these. Two minutes.

On this chapter specifically:

• Did you notice yourself wanting a verdict? This chapter's central answer is "it depends on the comparator," which is genuinely the finding and genuinely unsatisfying. Notice the pull toward wanting me to just say whether butter is bad. That pull is what both camps are selling to.
• Did you find yourself more sympathetic to one camp? Most readers do, and it usually tracks what they believed beforehand rather than what they just read. Which of the two 🔬 verdicts in §9.6 did you find easier to accept?
• Chapter 5 asked whether you'd made the Claim Filter card. If you did — did you use it on anything in this chapter? §9.4's "Keys cherry-picked seven countries" claim is a good candidate, and it's one where the skeptical position is the one that fails.

One thing to carry into Chapter 10: you now have the substitution frame. Chapter 10 is the diet wars, and every single one of them is a substitution argument in disguise. Try to spot which comparator each diet is implicitly claiming before I tell you.

Spaced Review

Answer before reading on.

1. (Chapter 3) Why does dietary fat reach the bloodstream by a different route than carbohydrate and protein, and why does that matter here?

Long-chain fat is packaged into chylomicrons and enters via the lymphatic system, joining the bloodstream near the heart — bypassing the liver's first pass (Ch 3 §3.4). Carbohydrate and amino acids go via the portal vein. It matters here because chylomicrons are one of the lipoproteins in §9.10, and because it's why blood triglycerides rise for hours after a fatty meal.

2. (Chapter 2) The recovered Minnesota and Sydney trials lowered cholesterol without improving mortality. Which Chapter 2 concept does that illustrate, and what's the honest response?

Surrogate versus hard endpoint (§2.7). Lowering a marker is not the same as preventing an outcome — the beta-carotene lesson in a different setting. The honest response is that these trials should reduce your confidence without overturning the whole evidence base, because they were old, incompletely reported, and methodologically limited. Holding both is the skill.

3. (Chapter 8) Ruth was told to cut eggs and cheese eight years ago. Using §9.7, say what was wrong with that advice and what it cost her.

The dietary cholesterol rationale has substantially weakened — hepatic regulation compensates, and the 300 mg limit was dropped from US guidance in 2015–2020. What it cost her: eggs and cheese were two of the three protein sources she liked and could chew, and removing them helped push a sarcopenic 79-year-old to 0.72 g/kg. Advice gets given once and followed forever; guidance changes quietly.

Project Checkpoint: Your Fat Audit

Component nine. Use your three-day diary.

Step 1 — total and percentage.

Fat % of energy = (grams × 9 ÷ total kcal) × 100

Theo: 137 g × 9 = 1,233 ÷ 3,160 = 39%. The AMDR for fat is 20–35% — he's above it, though this matters far less than his total intake does.

Step 2 — the split that matters. Estimate roughly how much of your fat is saturated versus unsaturated. You don't need precision; sorting your sources is enough:

Step 3 — the substitution question, applied to yourself. For your two largest saturated fat sources, ask: if I reduced this, what would replace it? Write the actual answer. If the honest answer is "refined carbohydrate," the change is not worth making (§9.5).

Step 4 — omega-3. How many servings of oily fish did you eat in the last week? If the answer is zero, note whether you have any deliberate ALA source (walnuts, flax, chia, canola).

Step 5 — what you cook in. Name it. Then check it against §9.11. This is often the single easiest change in the entire audit — swapping the cooking fat requires no behavioural change at all, just a different bottle.

Step 6 — one change, with a likelihood rating. Best candidates: 1. Change your cooking oil — costs nothing behaviourally 2. Add oily fish twice a week (sardines are ~$1.20 a tin) 3. Swap one saturated-fat snack for nuts

Non-tracking alternative. Skip the grams. Just answer three questions: What do I cook in? How much oily fish did I eat this week? If I cut my biggest saturated-fat source, what would take its place? Those three answers contain most of the actionable content.

Next checkpoint (Chapter 10): your diet-history autopsy — every diet you've tried, why it ended, and what that reveals.

Chapter Summary

Saturation is about hydrogen and double bonds. No double bonds = saturated = solid. It is not a claim about how the fat behaves in your body, and no food is one type — butter is 63% saturated and 26% monounsaturated.

The four types:

The substitution table — the chapter in one place:

The history, honestly: Keys's hypothesis was oversold, the "cherry-picked seven countries" charge is a garbled version of a different critique, the Sugar Research Foundation episode is real and over-extended, the recovered Minnesota and Sydney trials genuinely complicate things, and none of it adds up to exoneration.

Dietary cholesterol: the liver compensates; the 300 mg limit was dropped in 2015–2020; hyper-responders exist. The argument was mostly about the bacon.

Omega-3/6: LA and ALA are essential; ALA→EPA→DHA conversion is poor. The ratio hypothesis is weakly supported — higher LA is generally associated with lower risk. Fish oil supplements: large trials at ordinary doses are null; high-dose trials disagree (REDUCE-IT vs STRENGTH, with an unresolved placebo controversy).

LDL vs ApoB: ApoB counts particles and predicts risk better than LDL-C. HDL is a marker, not a lever — drugs that raise it haven't reduced events.

Cooking: oxidative stability matters more than smoke point, and it tracks saturation. Olive oil is the default. Flax and walnut oils: dressings only.

This chapter's verdicts:

The one thing to remember: "Is saturated fat bad?" is not a hard question — it's a malformed one. Supply the comparator and the fifty-year argument mostly dissolves.

And a closing note on temperament, because this is the chapter where it matters most. Both camps in the fat wars are defending something true and attacking something nobody said. The mainstream position is right that the LDL mechanism holds and that replacing saturated with unsaturated fat reduces events; it was wrong to communicate that as "eat less fat" and wrong to hold it with more confidence than the evidence carried. The contrarian position is right that the low-fat era failed, that the guidance was oversold, and that industry influence was real; it is wrong to convert those grievances into exoneration.

If you finish this chapter irritated with both, you have understood it.

What's Next

You now have all three macronutrients and, more importantly, the substitution frame.

Chapter 10 puts them head to head: low-carb versus low-fat versus Mediterranean versus vegan versus paleo versus carnivore, judged by Chapter 2's standards rather than by which side has better content.

What do the head-to-head trials show? (Much less difference than either camp claims.) What survives at two years? And what do the winning diets have in common — which turns out to be the actual answer, and the threshold concept the whole of Part II has been building toward.

Adherence beats composition. But you should see the evidence before you believe me.