This document reorganises nutrition around a single axis: not food groups, not servings, but how well each claim survives contact with the evidence. The result separates the handful of findings robust enough to apply to nearly everyone from the many that are genuinely individual, and from the surprising number that are simply unresolved. It covers macronutrient floors, the carbohydrate minimum, dosing for every class of fat, micronutrient bioavailability, gastrointestinal and cognitive outcomes, and the popular beliefs that recent evidence has overturned — including several that were overturned in the direction nobody expected.
Live literature search was unavailable for most sessions that produced this document. The content is drawn from a deep working knowledge of this literature, refined by a later research pass that supplied some verified quotations and intervals. Every effect estimate, confidence interval, sample size, DIAAS score and nutrient value requires independent confirmation against the primary source before publication or clinical use.
Treat all 2024–2026 figures as provisional — specifically US POINTER (JAMA 2025), KETO-CTA (JACC Advances 2025), the SSaSS intervals (NEJM 2021), the colibactin / early-onset colorectal cancer signature work (Nature 2025), and the MIND-diet RCT (NEJM 2023). Nutrient composition varies by species, cut, feed, season and preparation. Figures marked SCHEMATIC render the shape of a reported relationship, not extracted datapoints.
Vertical position encodes evidence quality, not quantity to eat. Colour depth runs with certainty: the darkest band rests on controlled trials, the amber band is a setting you choose, and the top band is where published trials contradict one another. Width encodes evidentiary weight.
Tier 2 is only useful if you know which foods discharge it. This matrix maps nine non-negotiable requirements against nine candidate foods. Read it column-wise to see how many floors a single food closes — the pattern is the point, and it is not the pattern the old pyramid implied.
The three findings that matter here. First, a small number of foods — shellfish, liver, oily fish — close almost every floor at once, which is a stronger and more defensible claim than the cardiovascular one usually made for them. Second, poultry closes only the protein floor: substituting it for red meat keeps the macronutrient and quietly opens the iron, zinc and B12 gaps. Third, legumes and leafy greens close none of these floors outright — which is not an argument against eating them, because their job is Tier 3, but it is an argument against treating them as a substitute for Tier 2. Ratings are qualitative judgements from composition data and absorption literature; verify per species and cut.
DIAAS measures true ileal digestibility of each indispensable amino acid against a reference pattern. A score of 100 means the food alone satisfies the pattern; below 100 means it is limited by a specific amino acid and must be combined with a complementary source or eaten in greater quantity. Every food discussed in this document is ranked below, so the question "which should be on top to meet the minimum protein requirement" has a direct answer.
Approximate · FAO reference pattern · 100 = reference pattern satisfied by that food alone
These are the least reliable decimals in the document. DIAAS varies substantially with processing, cooking, and whether the child or adult reference pattern is used; published values for the same food differ between sources, and fish values in particular are less commonly reported than dairy or meat. Compilations: Herreman 2020 (Food Sci Nutr); Marinangeli & House 2017 (Nutr Rev). Reconfirm any single value before quoting it.
Whether a food is a macronutrient or micronutrient source is measurable, not interpretive. The clearest test is to ask how much of each you would need to eat to hit one fixed protein target.
One per-meal protein target · approximate cooked weights
This is the answer to the macro/micro question. Non-starchy vegetables cannot function as macronutrient sources at any intake a person will actually achieve. That is not a criticism — it is a statement about which job they do.
| Food category | Macronutrient role | Signature micronutrients | Classification | Tier |
|---|---|---|---|---|
| Oily fish salmon, mackerel, sardines |
Complete protein, high DIAAS. The fat is the point: preformed EPA and DHA. | EPA/DHA, vitamin D, B12, selenium, iodine | BOTH — strongly The only common food that is both a protein source and the practical solution to several micronutrient floors. | TIER 2 |
| Shellfish oysters, mussels, clams |
Complete protein, low fat, low energy density. | Zinc (extreme), B12 (extreme), iron, selenium, iodine | MICRO — outlier The most micronutrient-dense foods in the human diet per calorie. | TIER 2 |
| Organ meat liver |
Complete protein, moderate fat. | Preformed retinol, B12, folate, copper, iron, choline | MICRO — outlier Closes retinol and copper floors nothing else reaches. Vitamin A toxicity is a real ceiling — do not eat daily. | TIER 2 |
| Lean white fish cod, haddock, tilapia |
Very high protein per calorie, minimal fat. | Iodine (cod, very high), selenium, B12 — little EPA/DHA | MACRO-leaning Excellent protein, but not the omega-3 vehicle oily fish is. | TIER 2 |
| Red meat beef, lamb |
Complete protein, high DIAAS, variable fat. | Heme iron, zinc, B12, creatine, carnitine | BOTH The most efficient iron and zinc vehicle in ordinary diets. Outcome questions sit at the apex. | TIER 2 |
| Poultry chicken, turkey |
Among the most efficient protein-per-calorie foods. ~31 g/100 g cooked breast. | Niacin, B6, selenium, phosphorus, choline | MACRO — chiefly Iron, zinc and B12 run 3–8× lower than red meat. A protein instrument, not a micronutrient one. | TIER 2 |
| Eggs and dairy | Complete protein, highest DIAAS scores of any food. | Choline (eggs), calcium and iodine (dairy), B12, retinol | BOTH Dairy is the only category that closes the calcium floor without supplementation. | TIER 2 |
| Legumes lentils, beans, chickpeas |
Meaningful protein but DIAAS 55–83, plus substantial carbohydrate and fibre. | Folate, potassium, magnesium, non-heme iron (phytate-inhibited) | MIXED The strongest plant protein after soy, and a genuine fibre vehicle. Not a Tier 2 substitute on its own. | TIER 3 |
| Non-starchy vegetables broccoli, spinach, kale |
Negligible. ~2–3 g protein and under 40 kcal per 100 g. | Vitamin C, K1, folate, potassium, carotenoids, nitrate, polyphenols, fibre | MICRO — almost purely Plus fibre, water and volume. No B12, little bioavailable iron or zinc. | TIER 3 |
| Whole fruit | Sugar and water. Minimal protein or fat. | Vitamin C, potassium, folate, anthocyanins, fibre | MICRO + fibre The whole-fruit versus juice divergence is among the better-replicated findings in the field. | TIER 3 |
| Starchy vegetables and grains potato, cassava, wheat, rice |
Genuinely a carbohydrate staple. ~17 g carbohydrate per 100 g potato. | Potassium, some B vitamins; whole grains add fibre and magnesium | MACRO — carbohydrate Grouping potatoes with broccoli obscures the only thing that matters about them nutritionally. | TIER 4 |
Reading the table. "Both" is rare. Most foods do one job well, and the old pyramid's error was arranging categories that mix jobs rather than categories that share one.
The most reproducible causal finding of the last decade is not about macronutrients at all. Hold macros constant, change only the degree of processing, and intake shifts by roughly 500 kcal per day.
Hall 2019 · ad libitum energy intake, kcal/day
Bar heights drawn to the reported ~500 kcal difference; absolute intakes approximate. Verify before citing.
0.8 g/kg/day derives from nitrogen balance, a method that under-captures losses and extrapolates linearly. Isotope tracer studies using indicator amino acid oxidation put the safe intake near 1.2 g/kg/day, and find older adults need at least as much as the young — not less.
Protein intake targets · g/kg body weight/day
The grey bar is the current official figure. Everything beyond it is post-2007 tracer and RCT literature. Higher protein does not harm healthy kidneys or bone — see the misconceptions section.
This is the most misunderstood number in nutrition, in both directions. Carbohydrate is not biochemically essential — but "not essential" is not the same as "no floor". There is a real physiological glucose demand, it is quantifiable, and it changes dramatically with metabolic state and life stage.
Grams of glucose per day · fed state versus full ketoadaptation
Why the RDA is 130 g and why that is not a requirement. The IOM set the carbohydrate EAR at about 120 g/day from average minimum brain glucose utilisation in the fed state, then the RDA at 130 g. The same report states that the lower limit of dietary carbohydrate compatible with life is essentially zero, provided adequate protein and fat are consumed. Both readings are honest: nothing is biochemically essential, and 130 g is the amount that lets the brain run on glucose alone without metabolic adaptation. Owen and Cahill's starvation-physiology work established the ketone shift; Cunnane's imaging work shows brain ketone uptake rises in proportion to blood ketone concentration. Verify the DRI wording in the 2005 report, Chapter 6.
| Population | Recommended carbohydrate | Basis | How hard is this floor? |
|---|---|---|---|
| Healthy adult | RDA 130 g/day EAR 100 g/day Biochemical minimum: 0 | Average brain glucose oxidation in the fed state. Not derived from any health-outcome trial. | SOFT Gluconeogenesis reliably covers demand. Ketogenic diets are metabolically safe short to medium term. |
| Pregnancy | RDA 175 g/day EAR 135 g/day | Maternal brain glucose plus fetal and placental demand. The fetus is an obligate glucose user. | FIRMER Ketogenic diets are cautioned mainly on animal and mechanistic grounds — human evidence is sparse and low-certainty. The caution is prudential, not proven. |
| Lactation | RDA 210 g/day | Glucose demand for lactose synthesis in milk. | FIRMER Substantial obligate draw on maternal glucose. |
| Children and adolescents | RDA 130 g/day | Same brain-glucose basis, scaled for size. Growth adds protein and energy demand, not a separate carbohydrate floor. | SOFT–MODERATE Therapeutic ketogenic diets are used in paediatric epilepsy under supervision. |
| Endurance and high-glycolytic athletes | 3–12 g/kg/day scaled to training load | Glycogen storage capacity is ~300–600 g (1,200–2,400 kcal). Carbohydrate availability limits sustained high-intensity work. | HARD, for performance Ketoadaptation raises fat oxidation but impairs exercise economy at high intensity (Burke 2017, J Physiol). |
| Type 1 diabetes | No absolute requirement | Carbohydrate counting governs insulin dosing. | CLINICAL Low-carbohydrate approaches are used by some patients but complicate insulin titration and ketoacidosis management. Physician involvement essential. |
mmol/L · the distinction is quantitative, and acidosis is the dividing line
The clinically important interaction. Anyone taking an SGLT2 inhibitor — empagliflozin, dapagliflozin, canagliflozin — should not combine it with a ketogenic diet without physician oversight. This is the single most actionable safety point in this section.
Schematic · Seidelmann 2018 · hazard ratio versus carbohydrate share of energy
Schematic. Endpoints reflect reported hazard ratios; the connecting curve is interpolated. Observational — residual confounding applies at both extremes, and the substitution finding matters more than the amount.
"Eat healthy fats" is not actionable. Below is every fat class with whether a true physiological requirement exists, the quantitative target, the basis for it, and whether the evidence is settled or contested. Two entries are genuinely essential nutrients; the rest are dosing questions of varying certainty.
| Fat class | Truly essential? | Daily target | Basis and evidence | Status |
|---|---|---|---|---|
| Total fat | No | 20–35% of energy (AMDR) | Below ~20% risks essential fatty acid and fat-soluble vitamin inadequacy and raises triglycerides on high-carb refeeding. There is no requirement for a specific total-fat percentage — only for the two essentials below. | MILDLY CONTESTED |
| Linoleic acid omega-6, C18:2 | Yes — essential | 17 g/day men 12 g/day women (AI) Deficiency prevented at ~1–2% energy | Depletion–repletion studies (foundational). The "high omega-6 is harmful" claim is not supported in humans: Marklund 2019 Circulation pooled 30 cohorts using circulating linoleic acid biomarkers (n≈68,000+) and found higher linoleic acid associated with lower CVD and all-cause mortality. Controlled feeding does not show LA raising CRP. | ESSENTIAL · harm claim refuted |
| Alpha-linolenic acid omega-3 plant, C18:3 | Yes — essential | 1.6 g/day men 1.1 g/day women (AI) | Conversion to EPA is ~5–8%; to DHA under 0.5–4%, higher in women. ALA alone does not meaningfully raise DHA status, which is why direct marine intake matters for those targeting DHA. | ESSENTIAL |
| EPA and DHA long-chain marine omega-3 | Conditionally | 250 mg/day (EFSA) +200 mg DHA in pregnancy AHA: 1–2 servings oily fish/week | Cohorts support fish; supplement RCTs conflict. REDUCE-IT (4 g EPA, high-risk statin-treated) HR 0.75 (0.68–0.83); STRENGTH, VITAL primary, ASCEND all null; Cochrane 2020 found "little or no effect". Where supplements work it appears to need high-dose EPA in high-risk patients, not general-population fish oil. | FOOD YES · PILLS CONTESTED |
| Saturated fatty acids | No | <10% of energy | Heavily contested. Hooper 2020 Cochrane: reducing SFA cut combined cardiovascular events 17% (RR 0.83, 0.76–0.90) but had little or no effect on mortality; benefit greatest when replaced by polyunsaturated fat. Siri-Tarino 2010, Chowdhury 2014, de Souza 2015 found no clear association. Astrup 2020 JACC argued against a blanket limit and drew rebuttals. | GENUINELY CONTESTED |
| Monounsaturated fat | No | No formal target | PREDIMED: Mediterranean diet plus extra-virgin olive oil HR 0.69 (0.53–0.91) or nuts HR 0.72 (0.54–0.95) for major cardiovascular events — one landmark RCT, with a randomisation irregularity that forced republication. Cohorts associate olive oil intake with lower mortality. | FAVOURABLE |
| Industrial trans fat | No — avoid | <1% of energy Target: near zero | de Souza 2015: ~21–34% higher CHD risk and higher all-cause mortality. The clearest-cut harmful fat in the field; WHO REPLACE targets global elimination. | NOT CONTESTED |
| Ruminant trans fat vaccenic acid, CLA | No | No limit needed at usual intakes | A separate question from industrial trans fat. Not clearly associated with harm at the amounts naturally present in dairy and meat. | UNCERTAIN |
| Dietary cholesterol | No | No numeric cap since 2015 | The 300 mg/day limit was removed from US guidance because dietary cholesterol affects serum LDL weakly relative to saturated and trans fat. Roughly one-third are "hyper-responders" whose serum cholesterol rises appreciably; two-thirds change little. Egg literature conflicts: Zhong 2019 JAMA found higher risk per half-egg/day; many other cohorts and short RCTs are null. | CONTESTED |
| Medium-chain triglycerides | No | No target | Rapidly absorbed and ketogenic; used in cognition research and some weight contexts. Effects modest and inconsistent. | PRELIMINARY |
| Omega-6 to omega-3 ratio | Not a real target | Do not manage the ratio | Largely superseded by absolute intakes. Because linoleic acid is not demonstrably harmful, lowering omega-6 to "fix the ratio" is unsupported. Raising omega-3 in absolute terms is the defensible lever. EFSA and most expert reviews do not endorse a ratio target. | OBSOLETE CONCEPT |
Relative LDL-raising effect by individual fatty acid
Relative magnitudes are illustrative of the established ordering, not extracted effect sizes. Stearic acid is essentially LDL-neutral because much of it is desaturated to oleic acid — which is why beef fat and cocoa butter behave differently from butterfat and coconut oil.
Why similar saturated fat behaves differently
The observational case for vegetables and fruit is strong and consistent. Nearly every attempt to isolate the responsible compound and test it has failed, sometimes harmfully. That pattern is itself one of the most informative findings in nutrition science.
Schematic · Aune 2017 · all-cause mortality
Isolated micronutrient RCTs · above 1.0 indicates harm
The failure of the supplement trials is the strongest argument for eating vegetables as food and against believing we know why they work. Four explanations remain live: the food matrix does something isolated compounds cannot; fibre and potassium carry most of the effect; vegetables displace worse foods; or the cohort signal is substantially healthy-user bias. These have very different implications, and the literature does not currently distinguish between them. What survives regardless: eat the plants, don't buy the extract.
These are two different claims carrying very different certainty. Fish delivers nutrients that are difficult or impossible to obtain elsewhere — that is biochemistry, and it is solid. Whether eating it lowers cardiovascular risk is separate, and the trials that tested the active ingredient directly did not agree.
Grams per 100 g cooked · varies with feed and season
Forest plot · cardiovascular endpoints · point estimate with 95% confidence interval
How to read this honestly. One trial found a large benefit; three found none; the pooled synthesis found essentially nothing. The single positive trial used a mineral-oil placebo that appears to have raised LDL and CRP in the control arm, which would inflate the apparent effect — a published, unresolved criticism. Two readings remain live: either the cohort fish signal reflects the whole food matrix and displacement of worse options, or it reflects residual confounding. Notably, VITAL found greater benefit in people who ate little fish at baseline, which fits a repletion model rather than a drug model.
Approximate content per 100 g cooked · relative to beef
Ratios approximate, vary by cut. Practical implication: swapping all red meat for chicken meets the protein floor but can quietly open iron, zinc and B12 gaps — particularly in menstruating women. Shellfish or oily fish close them better than either.
Two failure modes are worth separating. Some widely repeated claims are simply wrong and the correction is high-certainty. Others are popular contrarian claims that are weaker than their advocates assert. Both appear below, with the certainty of each correction stated rather than implied.
| The claim | What the evidence actually shows | Correction certainty |
|---|---|---|
| "High protein damages the kidneys" | Refuted for healthy people. Devries et al. 2018 (J Nutr, meta-analysis of RCTs) found higher protein did not adversely affect GFR in adults with normal kidney function; Van Elswyk 2018 concordant. The genuine exception is pre-existing chronic kidney disease, where higher protein can accelerate decline and restriction is a legitimate clinical tool. So: no harm in healthy kidneys, real caution in existing CKD. | ●●●●● |
| "High protein leaches calcium and harms bone" | Refuted, and the underlying theory is dead. Shams-White 2017 (AJCN) and Groenendijk 2019 found higher protein neutral-to-beneficial for bone mineral density with no increase in fracture risk. The acid-ash hypothesis — that protein acidifies blood and dissolves bone — has been refuted; increased urinary calcium reflects increased absorption, not net bone loss. | ●●●●● |
| "Protein drives mTOR, IGF-1 and shortens lifespan" | Overstated in humans. Levine 2014 (Cell Metabolism) reported higher protein associated with cancer mortality at ages 50–65 but protective over 65 — from a single 24-hour dietary recall in NHANES, with the mechanism largely from animal models. Critiques note residual confounding and poor translation. In older adults, higher protein protects against sarcopenia and frailty. | ●●●○○ |
| "Red meat causes colon cancer" | Requires the classification distinction. IARC 2015: processed meat is Group 1 (carcinogenic), unprocessed red meat is Group 2A (probably). Effect size: roughly +18% relative colorectal cancer risk per 50 g/day processed meat — but absolute lifetime risk moves from about 5% to about 6%. Mechanisms: heme iron catalysing N-nitroso compound formation, nitrosamines in processed meat, heterocyclic amines from high-temperature cooking; fibre and calcium appear protective. NutriRECS 2019 applied GRADE and argued certainty is low, provoking an unresolved methodological dispute. | ●●●○○ |
| "Eggs and dietary cholesterol cause heart disease" | Genuinely unresolved. The 300 mg/day cap was dropped from US guidance in 2015. Zhong 2019 (JAMA, n≈29,615) found higher risk per half-egg/day; many other cohorts — especially Asian — and short RCTs are null. Roughly one-third of people are hyper-responders. Remaining uncertainty is real and should not be papered over in either direction. | ●●○○○ |
| "Saturated fat causes Alzheimer's disease" | Weak and contested. Observational associations are inconsistent; the strongest signal involves the APOE4 interaction, itself inconsistent. Cohort data look impressive — MIND diet adherence associated with ~53% lower Alzheimer's rate (Morris 2015) — but the MIND-diet RCT (Barnes 2023, NEJM, ~600 participants, 3 years) found no significant cognitive benefit over a control with mild caloric restriction. B-vitamin and omega-3 cognition trials are largely null. | ●●○○○ |
| "Everyone should minimise salt" | Genuine scientific disagreement — likely a J-curve. PURE (O'Donnell 2014, NEJM) suggests both very high (>5 g sodium/day) and very low (<3 g/day) associate with higher risk. The opposing camp — TOHP long-term follow-up and DASH-Sodium — supports linear benefit and argues PURE suffers reverse causation and spot-urine error. SSaSS (Neal 2021, NEJM, n=20,995, ~4.7 years): potassium-enriched salt substitute cut stroke (RR ≈0.86), major cardiovascular events (≈0.87) and total mortality (≈0.88) without serious hyperkalaemia excess. Lowering sodium in hypertensive or high-intake people is well supported; pushing every normotensive person ever-lower is not settled. | ●●○○○ |
| "Fibre is essential for everyone" | Conditional, not absolute. Population benefits are strong, but low-FODMAP restriction reduces symptoms in IBS, meaning some fermentable fibre worsens symptoms in specific individuals. Separately, the old diverticular advice was reversed — Strate 2008 (JAMA) found no increased diverticulitis risk from nuts, seeds or popcorn, and higher fibre now associates with lower risk. | ●●●●○ |
| "Sugar is toxic" | Overstated; dose and form matter. Evidence is strongest for sugar-sweetened beverages (consistent associations with weight gain, T2D, CVD). For total added sugar the risk is more dose-dependent and confounded by total energy; isocaloric fructose substitution shows little independent harm at moderate doses but adverse effects (hepatic de novo lipogenesis, triglycerides) at high doses under hypercaloric conditions. | ●●●●○ |
| "Gluten harms everyone" | False outside specific conditions. Coeliac disease (~1%) requires strict avoidance. Non-coeliac gluten sensitivity is real for some but frequently confounded — blinded challenge studies (Biesiekierski 2013, Gastroenterology) implicate FODMAPs and amylase-trypsin inhibitors rather than gluten itself in many self-reported cases. No benefit to avoidance in the general population. | ●●●●● |
| "Soy feminises men" | Refuted. Messina 2021 (meta-analysis of clinical studies): neither soy protein nor isoflavones significantly affect total or free testosterone, estradiol or SHBG in men. | ●●●●● |
| "Frequent small meals raise metabolism / breakfast is essential" | Both refuted. The thermic effect of food scales with total intake, not meal frequency. Sievert 2019 (BMJ meta-analysis): skipping breakfast does not cause weight gain and adding it does not aid weight loss. | ●●●●● |
| "Time-restricted eating is metabolically special" | Benefit is mostly caloric. TREAT (Lowe 2020, JAMA Intern Med): 16:8 produced no significant additional weight loss versus three structured meals, with a possible lean-mass-loss signal. Meta-analyses show modest loss (~1–2 kg) attributable to reduced intake, not timing. Note also: the widely-shared 2024 conference abstract linking 8-hour windows to cardiovascular mortality was non-peer-reviewed, observational, and based on 1–2 days of recall — not established. | ●●●●○ |
| "Multivitamins and antioxidants prevent disease" | Largely refuted, and some cause harm. ATBC: β-carotene increased lung cancer in smokers. CARET: β-carotene plus retinol increased lung cancer and mortality. SELECT: vitamin E increased prostate cancer risk. Bjelakovic meta-analyses: no mortality benefit, possible harm. Partial counterpoint: COSMOS-Mind found a small but significant multivitamin benefit on global cognition in older adults — the most credible positive signal to date, modest and needing replication. | ●●●●○ |
| "Coconut oil is heart-healthy" | Refuted for lipids. Neelakantan 2020 (Circulation, 16 trials): coconut oil raised LDL-C by 10.47 mg/dL (95% CI 3.01–17.94) versus non-tropical vegetable oils, with no benefit on weight, waist or glycemia. | ●●●●● |
| "Omega-6 seed oils are inflammatory" | Not supported in humans. Marklund 2019 (Circulation, 30 cohorts, biomarker-based) found higher circulating linoleic acid associated with lower CVD and mortality. Controlled feeding does not show linoleic acid raising CRP. The oxidised-metabolite hypothesis remains mechanistic and unconfirmed as net harm. | ●●●●○ |
| "Alkaline and detox diets work" | Unsupported. Blood pH is tightly regulated at 7.35–7.45 and diet cannot meaningfully change it; the kidneys and liver perform detoxification. Any benefit comes from eating more vegetables and less processed food, not from alkalinity. | ●●●●● |
| "Artificial sweeteners cause weight gain or cancer" | Largely unsupported at normal intake. RCTs show non-nutritive sweeteners modestly aid weight control versus sugar. IARC classified aspartame Group 2B ("possibly carcinogenic") in 2023, but JECFA retained the acceptable daily intake at 40 mg/kg/day — realistic consumption sits well within that margin. | ●●●●○ |
| "LDL doesn't matter if your triglycerides and HDL look good" | The popular contrarian claim, and it is weaker than advertised. KETO-CTA (2025) followed ~100 lean-mass hyper-responders for one year and reported plaque change unrelated to LDL-C or ApoB — but it was single-arm with no control group, only one year, and investigators had low-carb ties, and independent cardiologists noted absolute progression did occur. It does not overturn the Mendelian randomisation and RCT evidence that ApoB is causal at population level. Verify all figures. | ●●○○○ |
Reading the dots. Five dots means the correction is well established and you can rely on it. Two dots means the honest answer is "nobody knows yet" — which applies as much to confident contrarian claims as to confident conventional ones.
The colon is where the fibre story is most mechanistically satisfying and most empirically frustrating — the mechanism is beautifully characterised and the supplement trials are null.
Cohort and meta-analytic estimates · approximate
The distinction that gets lost in headlines. An 18% relative increase on a 5% baseline is about one extra case per hundred people — real, worth acting on, and not the catastrophe the phrase "causes cancer" implies. Verify all figures against WCRF/AICR.
Dietary patterns look strongly protective in cohorts and mostly fail to replicate in trials. This section is where confident dietary claims about Alzheimer's disease should be treated with the most suspicion — in both directions.
Every entry here is a live dispute in the published literature. That is why the apex is drawn small and open. Anyone presenting one of these as settled — in either direction — has gone beyond the evidence.
| Contested claim | Evidence for | Evidence against | Certainty |
|---|---|---|---|
| Saturated fat causes cardiovascular disease | Hooper 2020 Cochrane: reducing SFA cut combined cardiovascular events, RR 0.83 (0.76–0.90), greatest when replaced by PUFA | Same review found little or no effect on mortality. No significant association in Siri-Tarino 2010 (RR 1.07), Chowdhury 2014, de Souza 2015. Astrup 2020 JACC argued against a blanket limit and drew rebuttals | ●●○○○ |
| Unprocessed red meat meaningfully raises risk | IARC 2015 Group 2A. Processed meat is not disputed — Group 1, ≈+18% colorectal cancer per 50 g/day. Coherent heme-iron and N-nitroso mechanisms | NutriRECS 2019 applied GRADE, judged certainty low and small absolute risk, recommending no change — provoking rebuttals and retraction calls. Ioannidis disputes the field's effect sizes generally | ●●○○○ |
| Fish consumption prevents cardiovascular disease | Consistent modest inverse cohort associations for coronary death and stroke; REDUCE-IT HR 0.75 with 4 g EPA; VITAL showed reduced MI and more benefit in low-fish consumers | STRENGTH null and stopped for futility; ASCEND null; VITAL null on its primary endpoint; Cochrane 2020 found little or no effect. REDUCE-IT's mineral-oil placebo remains disputed. PURE found benefit only in higher-risk groups | ●●○○○ |
| Sodium should be minimised in everyone | DASH-Sodium and TOHP long-term follow-up support linear benefit. SSaSS (n=20,995): potassium salt substitute cut stroke ≈14%, mortality ≈12% | PURE suggests a J-curve with elevated risk below ~3 g sodium/day. Opponents argue reverse causation and spot-urine error; PURE authors argue the trials studied hypertensive high-intake groups only | ●●○○○ |
| Dietary cholesterol and eggs raise cardiovascular risk | Zhong 2019 JAMA: higher risk per additional half-egg/day and per 300 mg cholesterol, pooled US cohorts n≈29,615 | Numerous other cohorts, especially Asian, and short-term RCTs are null. The 300 mg cap was removed from US guidance in 2015. Hyper-responder variability is real and unaccounted for | ●●○○○ |
| Diet causally prevents dementia | MIND adherence associated with ~53% lower Alzheimer's rate (Morris 2015). FINGER multidomain RCT positive on a cognitive composite. US POINTER 2025 reportedly positive | MIND-diet RCT (Barnes 2023, NEJM) null. MAPT and preDIVA null. Single-nutrient trials — omega-3, B vitamins, vitamin D — largely null. Cohort-to-trial gap is the widest in nutrition | ●●○○○ |
| Saturated fat contributes to Alzheimer's disease | Some cohort associations; an APOE4 interaction is biologically plausible and reported in places | Associations inconsistent; the APOE4 interaction itself is inconsistent; no trial evidence. Mechanism-rich and trial-poor | ●○○○○ |
| ApoB and LDL are not causal in metabolically healthy hyper-responders | KETO-CTA (2025) reported one-year plaque change unrelated to LDL-C or ApoB in ~100 lean mass hyper-responders, with baseline plaque the strongest predictor | Single-arm, no control group, one year — short for atherosclerosis. Investigator ties to low-carb advocacy. Independent cardiologists noted absolute progression occurred. Large Mendelian randomisation and RCT evidence supports ApoB causality | ●○○○○ |
| Ketogenic diets are safe beyond two years | No clear harm signal in trials up to 2 years; T2D remission data are favourable; theoretical micronutrient concerns are addressable by formulation | Evidence beyond two years is thin and dominated by observational data. Documented risks include kidney stones, possible T3 reduction, LDL elevation in a subset, and high attrition with diet convergence by 12–24 months | ●●○○○ |
| Poultry is meaningfully associated with cancer | Scattered cohort associations with specific cancers in some UK analyses; at least one recent European cohort reported higher mortality at high intake | Inconsistent across cohorts, no mechanism distinct from cooking method, unreplicated, confounded by preparation and processing. Most meta-analyses find poultry null | ●○○○○ |
| Glycemic index is clinically actionable | Mechanistically coherent and widely used in practice | OmniCarb (Sacks 2014, JAMA) found limited independent cardiometabolic effect within an already healthy diet | ●●○○○ |
| Higher protein shortens lifespan via mTOR and IGF-1 | Levine 2014 reported higher cancer mortality with high protein at ages 50–65; strong animal mTOR and IGF-1 mechanism | Reversed above age 65 in the same analysis. Single 24-hour recall, residual confounding, poor animal-to-human translation. Higher protein protects older adults against sarcopenia | ●●○○○ |
| Genotype-based diets beat standard advice | Real gene–diet interactions exist for lactase persistence, CYP1A2, FADS1/2 and BCO1 | Food4Me RCT (n=1,607): personalised advice beat generic, but genotype added nothing over phenotype and diet data. DIETFITS found no genotype interaction | ●○○○○ |
| The rise in early-onset colorectal cancer has an identified cause | A 2025 Nature signature analysis found colibactin signatures from pks+ E. coli enriched in early-onset tumours, possibly from childhood exposure | Genomic association, not demonstrated causation of the population trend. Competing candidates — ultra-processed food, early-life obesity, antibiotics, sedentary behaviour — remain speculative. The trend is substantially unexplained | ●○○○○ |
Certainty dots reflect the consistency of high-quality evidence, not the popularity of the position. Five would mean settled; nothing in this table earns more than two. Note that several entries are contested contrarian claims, not contested conventional ones — scepticism cuts both ways.
This is the empirical case against one-size-fits-all — and simultaneously the case against the commercial personalisation industry, because the same literature shows genotype-based advice failing to beat far simpler methods.
Schematic · after Zeevi 2015 and PREDICT 1
Schematic. Illustrates the reported phenomenon of large interpersonal variability, not extracted participant data.
Practical reading. Variability is real, so fixed universal ratios are poorly justified — but the validated way to personalise is measurement and response (weight, lipids and ApoB, HbA1c, glucose, and how you actually feel and adhere), not a genotype panel.
Stated as consequences of the evidence rather than as prescriptions — the entire point of the carbohydrate dial is that the setting is yours.
Very little in nutrition is both important and settled. What survives scrutiny is a short list: food form matters more than macronutrient ratio; protein requirements are higher than the RDA states and harmless to healthy kidneys and bone; only two fatty acids are truly essential; carbohydrate has no biochemical floor but several real physiological ones; fibre and whole plants carry robust signals through a mechanism nobody has isolated; and industrial trans fat is the one clearly harmful fat. Almost everything else — saturated fat causality, red meat dose, sodium targets, eggs, whether fish prevents heart disease, whether any diet prevents dementia — is genuinely unresolved, and the confident contrarian answers are no better supported than the confident conventional ones.
Grouped by topic. Items marked foundational predate the 2015 window but remain the primary basis for the claim. Items marked [verify] could not be confirmed against the primary source and should be treated as pointers.
Food form and processing
Hall KD, et al. (2019) Ultra-processed diets cause excess calorie intake and weight gain. Cell Metabolism 30(1):67–77.
Hall KD, et al. (2021) Effect of a plant-based, low-fat diet versus an animal-based, ketogenic diet on ad libitum energy intake. Nature Medicine 27(2):344–353.
de Souza RJ, et al. (2015) Intake of saturated and trans unsaturated fatty acids and risk of all cause mortality, cardiovascular disease, and type 2 diabetes. BMJ 351:h3978.
Protein requirement and quality
Humayun MA, et al. (2007) Reevaluation of the protein requirement in young men with the indicator amino acid oxidation technique. Am J Clin Nutr 86(4):995–1002. Foundational.
Rafii M, et al. (2015, 2016) Dietary protein requirement of older men / women determined by the indicator amino acid oxidation technique. J Nutr.
Morton RW, et al. (2018) A systematic review, meta-analysis and meta-regression of the effect of protein supplementation on resistance training–induced gains. Br J Sports Med 52(6):376–384.
Bauer J, et al. (2013) PROT-AGE position paper. J Am Med Dir Assoc 14(8):542–559. Deutz NEP, et al. (2014) ESPEN recommendations. Clin Nutr 33(6):929–936.
Devries MC, et al. (2018) Changes in kidney function do not differ between healthy adults consuming higher- vs normal-protein diets. J Nutr 148(11):1760–1775.
Van Elswyk ME, et al. (2018) A systematic review of renal health in healthy individuals associated with protein intake. Adv Nutr 9(4):404–418.
Shams-White MM, et al. (2017) Dietary protein and bone health. Am J Clin Nutr 105(6):1528–1543. Groenendijk I, et al. (2019) Bone Rep.
Levine ME, et al. (2014) Low protein intake is associated with a major reduction in IGF-1, cancer, and overall mortality in the 65 and younger but not older population. Cell Metabolism 19(3):407–417.
Herreman L, et al. (2020) Comprehensive overview of the quality of plant- and animal-sourced proteins based on DIAAS. Food Sci Nutr 8(10):5379–5391.
Marinangeli CPF, House JD (2017) Potential impact of the digestible indispensable amino acid score as a measure of protein quality. Nutr Rev 75(8):658–667.
Carbohydrate, ketosis and metabolic state
Institute of Medicine (2005) Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids. National Academies Press. Foundational — source of the 130 g/day RDA and the statement on minimum carbohydrate.
Owen OE, et al. (1967) Brain metabolism during fasting. J Clin Invest 46(10):1589–1595. Foundational.
Cahill GF (2006) Fuel metabolism in starvation. Annu Rev Nutr 26:1–22. Foundational.
Cunnane SC, et al. (2016) Can ketones help rescue brain fuel supply in later life? Front Mol Neurosci 9:53. Cunnane SC, et al. (2020) Brain energy rescue. Nat Rev Drug Discov.
Seidelmann SB, et al. (2018) Dietary carbohydrate intake and mortality: a prospective cohort study and meta-analysis. Lancet Public Health 3(9):e419–e428.
Dehghan M, et al. (2017) Associations of fats and carbohydrate intake with cardiovascular disease and mortality (PURE). Lancet 390(10107):2050–2062.
Lean MEJ, et al. (2018, 2019) DiRECT: primary care–led weight management for remission of type 2 diabetes. Lancet / Lancet Diabetes Endocrinol.
Hallberg SJ, et al. (2018) Diabetes Ther 9(2):583–612. Athinarayanan SJ, et al. (2019) Front Endocrinol 10:348.
Gardner CD, et al. (2018) Effect of low-fat vs low-carbohydrate diet on 12-month weight loss (DIETFITS). JAMA 319(7):667–679.
Burke LM, et al. (2017) Low carbohydrate, high fat diet impairs exercise economy and negates the performance benefit from intensified training in elite race walkers. J Physiol 595(9):2785–2807.
Norwitz NG, Budoff M, Feldman D, et al. (2025) KETO-CTA: plaque progression in lean mass hyper-responders. JACC: Advances. [verify — single-arm, no control group]
Sacks FM, et al. (2014) Effects of high vs low glycemic index of dietary carbohydrate (OmniCarb). JAMA 312(23):2531–2541.
Fats and fatty acids
Hooper L, et al. (2020) Reduction in saturated fat intake for cardiovascular disease. Cochrane Database Syst Rev 8:CD011737.
Siri-Tarino PW, et al. (2010) Meta-analysis of prospective cohort studies evaluating the association of saturated fat with cardiovascular disease. Am J Clin Nutr 91(3):535–546. Foundational.
Chowdhury R, et al. (2014) Association of dietary, circulating, and supplement fatty acids with coronary risk. Ann Intern Med 160(6):398–406.
Astrup A, et al. (2020) Saturated fats and health: a reassessment and proposal for food-based recommendations. J Am Coll Cardiol 76(7):844–857. See published rebuttals in the same journal.
Ramsden CE, et al. (2016) Re-evaluation of the traditional diet-heart hypothesis: recovered data from the Minnesota Coronary Experiment. BMJ 353:i1246. Ramsden CE, et al. (2013) Sydney Diet Heart Study. BMJ 346:e8707.
Marklund M, et al. (2019) Biomarkers of dietary omega-6 fatty acids and incident cardiovascular disease and mortality: an individual-level pooled analysis of 30 cohort studies. Circulation 139(21):2422–2436.
Neelakantan N, Seah JYH, van Dam RM (2020) The effect of coconut oil consumption on cardiovascular risk factors: a systematic review and meta-analysis of clinical trials. Circulation 141(10):803–814.
Estruch R, et al. (2018) Primary prevention of cardiovascular disease with a Mediterranean diet supplemented with extra-virgin olive oil or nuts (PREDIMED). N Engl J Med 378(25):e34. Republished after randomisation correction.
Zhong VW, et al. (2019) Associations of dietary cholesterol or egg consumption with incident cardiovascular disease and mortality. JAMA 321(11):1081–1095.
Burdge GC, Calder PC (2005) Conversion of α-linolenic acid to longer-chain polyunsaturated fatty acids in human adults. Reprod Nutr Dev 45(5):581–597. Foundational.
Marine omega-3 and fish
Bhatt DL, et al. (2019) Cardiovascular risk reduction with icosapent ethyl for hypertriglyceridemia (REDUCE-IT). N Engl J Med 380(1):11–22.
Nicholls SJ, et al. (2020) Effect of high-dose omega-3 fatty acids vs corn oil on major adverse cardiovascular events (STRENGTH). JAMA 324(22):2268–2280.
Manson JE, et al. (2019) Marine n−3 fatty acids and prevention of cardiovascular disease and cancer (VITAL). N Engl J Med 380(1):23–32.
ASCEND Study Collaborative Group (2018) Effects of n−3 fatty acid supplements in diabetes mellitus. N Engl J Med 379(16):1540–1550.
Abdelhamid AS, et al. (2020) Omega-3 fatty acids for the primary and secondary prevention of cardiovascular disease. Cochrane Database Syst Rev 3:CD003177.
Mozaffarian D, Rimm EB (2006) Fish intake, contaminants, and human health. JAMA 296(15):1885–1899. Foundational.
Mohan D, et al. (2021) Associations of fish consumption with risk of cardiovascular disease and mortality among individuals with and without vascular disease from 58 countries. JAMA Intern Med 181(5):631–649.
Chowdhury R, et al. (2012) Association between fish consumption, long chain omega 3 fatty acids, and risk of cerebrovascular disease. BMJ 345:e6698.
Hibbeln JR, et al. (2007) Maternal seafood consumption in pregnancy and neurodevelopmental outcomes in childhood (ALSPAC). Lancet 369(9561):578–585.
FAO/WHO (2011) Report of the joint expert consultation on the risks and benefits of fish consumption. See also EFSA opinions on fish and methylmercury.
Fibre, vegetables, fruit and supplements
Reynolds A, Mann J, et al. (2019) Carbohydrate quality and human health: a series of systematic reviews and meta-analyses. Lancet 393(10170):434–445.
Aune D, et al. (2017) Fruit and vegetable intake and the risk of cardiovascular disease, total cancer and all-cause mortality. Int J Epidemiol 46(3):1029–1056.
Wang DD, et al. (2021) Fruit and vegetable intake and mortality. Circulation 143(17):1642–1654.
Appel LJ, et al. (1997) A clinical trial of the effects of dietary patterns on blood pressure (DASH). N Engl J Med 336(16):1117–1124. Foundational.
Muraki I, et al. (2013) Fruit consumption and risk of type 2 diabetes. BMJ 347:f5001.
Carter P, et al. (2010) Fruit and vegetable intake and incidence of type 2 diabetes mellitus. BMJ 341:c4229.
Siervo M, et al. (2013) Inorganic nitrate and beetroot juice supplementation reduces blood pressure in adults. J Nutr 143(6):818–826.
ATBC Study Group (1994) The effect of vitamin E and beta carotene on the incidence of lung cancer. N Engl J Med 330(15):1029–1035. Foundational.
Omenn GS, et al. (1996) Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease (CARET). N Engl J Med 334(18):1150–1155. Foundational.
Klein EA, et al. (2011) Vitamin E and the risk of prostate cancer (SELECT). JAMA 306(14):1549–1556.
Bjelakovic G, et al. (2012) Antioxidant supplements for prevention of mortality. Cochrane Database Syst Rev 3:CD007176.
Baker LD, et al. (2022/2023) COSMOS-Mind / COSMOS-Web: multivitamin supplementation and cognition. Alzheimers Dement / Am J Clin Nutr. [verify]
Wallace TC, Fulgoni VL (2017) Usual choline intakes are associated with egg and protein food consumption in the United States. Nutrients 9(8):839.
Rolls BJ, et al. Controlled feeding studies on dietary energy density and energy intake. Am J Clin Nutr / Physiol Behav, various.
Meat, colon and gastrointestinal
Bouvard V, et al. (IARC Working Group) (2015) Carcinogenicity of consumption of red and processed meat. Lancet Oncol 16(16):1599–1600.
Johnston BC, et al. (2019) Unprocessed red meat and processed meat consumption: dietary guideline recommendations from NutriRECS. Ann Intern Med 171(10):756–764. See accompanying critiques.
Zheng Y, et al. (2019) Association of changes in red meat consumption with total and cause-specific mortality. BMJ 365:l2110.
Pan A, et al. (2012) Red meat consumption and mortality. Arch Intern Med 172(7):555–563. Foundational.
Aune D, et al. (2016) Whole grain consumption and risk of cardiovascular disease, cancer, and all cause and cause specific mortality. BMJ 353:i2716.
Strate LL, et al. (2008) Nut, corn, and popcorn consumption and the incidence of diverticular disease. JAMA 300(8):907–914.
Biesiekierski JR, et al. (2013) No effects of gluten in patients with self-reported non-celiac gluten sensitivity after dietary reduction of fermentable, poorly absorbed, short-chain carbohydrates. Gastroenterology 145(2):320–328.
Díaz-Gay M, Alexandrov LB, et al. (2025) Colibactin mutational signatures in early-onset colorectal cancer. Nature. [verify — citation, fold-enrichment and causal inference]
World Cancer Research Fund / AICR Continuous Update Project — colorectal cancer. Authoritative synthesis for the effect sizes cited.
Sodium
O'Donnell M, et al. (2014) Urinary sodium and potassium excretion, mortality, and cardiovascular events (PURE). N Engl J Med 371(7):612–623.
Sacks FM, et al. (2001) Effects on blood pressure of reduced dietary sodium and the DASH diet (DASH-Sodium). N Engl J Med 344(1):3–10. Foundational.
Neal B, et al. (2021) Effect of salt substitution on cardiovascular events and death (SSaSS). N Engl J Med 385(12):1067–1077. [verify intervals]
Cognition and dementia
Morris MC, et al. (2015) MIND diet associated with reduced incidence of Alzheimer's disease. Alzheimers Dement 11(9):1007–1014.
Barnes LL, et al. (2023) Trial of the MIND diet for prevention of cognitive decline in older persons. N Engl J Med 389(7):602–611.
Ngandu T, et al. (2015) A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline (FINGER). Lancet 385(9984):2255–2263.
Andrieu S, et al. (2017) Effect of long-term omega-3 polyunsaturated fatty acid supplementation with or without multidomain intervention on cognitive function (MAPT). Lancet Neurol 16(5):377–389.
Moll van Charante EP, et al. (2016) Effectiveness of a 6-year multidomain vascular care intervention to prevent dementia (preDIVA). Lancet 388(10046):797–805.
Baker LD, et al. (2025) US POINTER: structured versus self-guided lifestyle intervention and cognition. JAMA. [verify — effect size, interval, p-value and DOI unconfirmed]
Misconceptions, personalisation and methods
Messina M, et al. (2021) Neither soyfoods nor isoflavones warrant classification as endocrine disruptors: a technical review of the observational and clinical data. Crit Rev Food Sci Nutr.
Sievert K, et al. (2019) Effect of breakfast on weight and energy intake: systematic review and meta-analysis of randomised controlled trials. BMJ 364:l42.
Lowe DA, et al. (2020) Effects of time-restricted eating on weight loss and other metabolic parameters (TREAT). JAMA Intern Med 180(11):1491–1499.
Zeevi D, Korem T, Segal E, et al. (2015) Personalized nutrition by prediction of glycemic responses. Cell 163(5):1079–1094.
Berry SE, Spector TD, et al. (2020) Human postprandial responses to food and potential for precision nutrition (PREDICT 1). Nature Medicine 26(6):964–973.
Celis-Morales C, et al. (2017) Effect of personalized nutrition on health-related behaviour change: the Food4Me European randomized controlled trial. Int J Epidemiol 46(2):578–588.
Simpson SJ, Raubenheimer D (2005) Obesity: the protein leverage hypothesis. Obes Rev 6(2):133–142. Foundational.
Ioannidis JPA (2018) The challenge of reforming nutritional epidemiologic research. JAMA 320(10):969–970.