Chapter 26 — Key Takeaways

One page.

⚠️ The organizing question is not "does diet matter" — it does — but FOR WHICH OUTCOME, BY HOW MUCH, AND COMPARED TO WHAT ELSE.

§26.1 — The Responsiveness Hierarchy

⚠️ A great deal of unnecessary guilt is generated by people trying to move tier 4 numbers with tier 1 interventions.

⚠️ §26.1b — One Process, Five Diseases

When intake persistently exceeds what subcutaneous fat can comfortably store, fat accumulates in liver, pancreas, muscle and viscera — and that ectopic fat produces every tier 1 number.

Fatty liver · type 2 diabetes · atherogenic dyslipidaemia · hypertension · several cancers.

⚠️ Which is why one intervention improves five conditions at once, and why §26.3 and §26.9 have the same dose-response shape. "It didn't fix three things. It reduced one exposure producing three readouts."

⚠️ Qualifications: not everyone with a condition has this process; and where you store fat is substantially genetic (Ch 24 §24.1b) — which is why waist measurement adds what BMI misses, takes ten seconds, and almost nobody does it.

✅ §26.3 — Remission, and the numbers that should be better known

DiRECT (The Lancet, 2018/2019): total diet replacement, structured reintroduction, long-term maintenance, medications withdrawn at the start. Remission ~46% at 12 months, ~36% at 24.

⚠️ AND STRONGLY DOSE-DEPENDENT ON WEIGHT LOST, at 12 months: ≥15 kg → ~86% · 10–15 kg → ~57% · 5–10 kg → ~34% · <5 kg → ~7%

⚠️ That gradient converts "try to lose some weight" into a quantified prediction. It matters more than the headline rate.

⚠️ Four qualifications: shorter duration → more likely · remission is not cure; regain means relapse · ⚠️ supervised, not self-directed — medications are withdrawn as part of the protocol · and the route matters less than the loss (surgery higher still; GLP-1s reaching the same range).

⚠️ A substantial proportion of people with type 2 diabetes have never been told remission is possible, or what magnitude of loss it would take. Both are knowable.

§26.4 — what to eat: multiple patterns work; weight loss is by far the largest lever; carbohydrate quality and total reduction, fibre, protein, no SSBs, activity. ⚠️ SAFETY: reducing carbohydrate on insulin or a sulfonylurea without adjusting the dose causes hypoglycaemia. Prescriber first.

✅ §26.5 — LDL-C, and what a settled question looks like

Convergence across: ⚠️ Mendelian randomization across many independent genes · randomized trials of multiple drug classes reducing events roughly in proportion to LDL-C lowering regardless of mechanism · familial hypercholesterolaemia · mechanistic artery-wall work.

⚠️ Different genes, different drugs, same relationship. Note the contrast with all of Part IV.

⚠️ Both halves: diet moves LDL-C meaningfully, AND declining a statin in favour of diet is choosing the smaller intervention. They are additive, not alternative.

🟢 PREDIMED — the Mediterranean pattern probably reduces events in high-risk people. ⚠️ Retracted and republished in 2018 after randomization irregularities; hold it slightly less firmly than its citation frequency suggests.

⚠️ The shape of the field: we know far more about how diet moves BIOMARKERS than EVENTS, because the trials for the second are enormous and largely unfundable.

§26.6 — ApoB and HDL

⚠️ Every atherogenic particle carries ONE ApoB — so ApoB counts particles, LDL-C measures cholesterol inside them. 🟢 ApoB is the better predictor where they diverge, and discordance is common in insulin resistance and raised triglycerides. ⚠️ Raised triglycerides with "normal" LDL-C → ask for ApoB.

⚠️ HDL-C is a MARKER, not a target — pharmacological raising has repeatedly failed. Low HDL travels with insulin resistance and adiposity, and it's those that carry the risk.

⚠️ Lp(a): inherited, an independent risk factor, ❌ does not respond to diet. Measure once in a lifetime — a raised result changes how aggressively everything else is managed.

✅ §26.7–26.8 — Blood Pressure and Sodium

DASH lowers blood pressure substantially, larger in hypertensives, ⚠️ and DASH-Sodium showed the pattern and sodium reduction are ADDITIVE.

Also: weight loss (large) · ⚠️ increasing potassium (the underrated one) · reducing sodium · alcohol · activity · sleep.

✅ ⚠️ THE SALT SUBSTITUTE TRIAL (SSaSS, NEJM 2021): potassium-enriched salt substitute in whole villages reduced STROKE, major cardiovascular events and DEATH. Randomized EVENT evidence for a dietary substitution costing almost nothing.

⚠️ CONTRAINDICATED in advanced CKD and with potassium-retaining medications — hyperkalaemia.

Sodium, where the chapter lands: ✅ reducing sodium lowers BP, dose-dependently · 🟢 population reduction reduces events · 🟡 whether normotensive people benefit from very low intakes. ⚠️ The J-curve cohorts are discounted for spot-urine estimation error and reverse causation. ⚠️ Practical version: most sodium comes from processed and restaurant food, not the salt cellar.

✅ §26.9 — Fatty Liver (MASLD)

Plus: no SSBs · less alcohol · activity independent of weight · Mediterranean pattern · 🟢 coffee is associated with less progression. ⚠️ The most reversible condition in the chapter.

⚠️ §26.10 — Cancer

✅ Established: alcohol (IARC Group 1) · excess adiposity (12+ cancers) · processed meat (Group 1), red meat (2A) · low fibre · aflatoxin. ⚠️ And the magnitude honesty: dietary factors matter and matter considerably less than not smoking; the biggest dietary lever is alcohol.

❌ What does damage: sugar feeds cancer · alkaline diets · fasting cures cancer · juicing · supplements ⚠️ (and Chapter 13's beta-carotene trials INCREASED lung cancer in smokers).

⚠️ THE HARM, specifically: delayed or abandoned treatment · malnutrition during treatment when nutritional status affects tolerance and outcomes · and the moral injury of implied blame.

⚠️ DURING TREATMENT THE GOALS INVERT: maintain weight and lean mass · adequate energy and protein · manage side effects · eat what you can tolerate · food safety if immunosuppressed. NOT restriction of any kind.

⚠️ The person most likely to be harmed by nutrition advice in this book's subject area is someone undergoing cancer treatment who has been given a restrictive diet by someone who meant well.

⚠️ §26.11 — Kidney Disease: the second inversion

⚠️ On dialysis it reverses AGAIN — protein goes UP to ~1.0–1.2 g/kg or more. ⚠️ Protein restriction's evidence is more contested than its prominence suggests, and carries real protein-energy wasting risk. ⚠️ PHOSPHATE ADDITIVES are far more bioavailable than food phosphate — someone who cuts dairy and legumes while eating processed food with additives has made it worse. Read the ingredients list.

⚠️ The pattern across §26.10 and §26.11

A person with a serious diagnosis is unusually motivated, unusually receptive, and unusually likely to be handed a restrictive diet by someone who means well — and in both cases the restriction IS the harm.

⚠️ The question is not "what should they cut out?" It is "are they eating enough, and of what?"

§26.12–26.13 — Gout and "Inflammation"

Gout: urate rises with alcohol (especially beer), fructose, purines, adiposity. ⚠️ Dietary modification produces small reductions compared with urate-lowering therapy — so "I've cut out everything and still get attacks" is the expected result of a small lever on a large problem, not an adherence failure.

"Inflammation": ⚠️ CRP responds mostly to weight loss, activity, smoking cessation, sleep and treating disease. "Anti-inflammatory diet" describes Chapter 10 §10.5's pattern with a label that hasn't earned its mechanism — and Chapter 19 §19.5 is the cautionary case.

⚠️ §26.13b — How Fast Do the Numbers Move?

Glucose: days · triglycerides: 1–2 weeks · blood pressure: 2–4 weeks · ⚠️ liver fat: weeks · LDL-C: 4–6 weeks · ⚠️ HbA1c: 8–12 WEEKS MINIMUM · fibrosis: months–years · CV events: years.

⚠️ DO NOT RECHECK HbA1c AT FOUR WEEKS. It cannot have moved, the result disappoints, and people quit on a measurement that was always going to look like that. This happens constantly.

⚠️ Rule: for early evidence something is working, measure a FAST number. To know whether it mattered, measure a SLOW one.

🧾 Economics

⚠️ Potassium salt substitute $10–25/yr — with randomized EVENT evidence, and almost nobody has heard of it. Psyllium $40–90 · sterols $80–150 · replacing SSBs with water −$700 to −$1,100 · coffee already budgeted · ⚠️ generic statin often $30–80/yr or free, doing 30–50%** — versus ⚠️ **$300–900/yr of "blood sugar," "liver detox" and "cholesterol" supplements for 🟠 to ❌.

One Thing to Remember

⚠️ Walt had been told to "watch his diet" for eleven years, and nobody had ever told him which of his numbers that would actually move.