Body and habits
Nutrition fundamentals
Energy balance, macronutrients and food quality — plus the skill the field really demands: reading observational evidence without being fooled by it.
Start this topic
Teach me nutrition fundamentals using learn.rapold.ioPaste it into any capable agent. It asks what you already know before it teaches anything.
What this subject is
Nutrition fundamentals covers four things and one skill. The four things: energy balance, correctly stated; the macronutrients and what is actually established about protein, fat and carbohydrate; food quality, including the processing question; and micronutrients, where the distinction between correcting a deficiency and supplementing an adequately nourished person carries most of the weight. The skill is reading the field. Nutrition is unusually hard to read for reasons that are structural rather than accidental: everyone is exposed, so there is no unexposed control group; diets arrive as correlated bundles, so single-nutrient effects are hard to isolate; the outcomes take decades, so trials measure surrogates instead; blinding is usually impossible; adherence decays; and the exposure is measured by asking people what they ate, which they cannot accurately say. On top of that sit commercial interests larger than the research budget and a public appetite for simple instructions. The result is a literature in which genuine, well-replicated findings sit next to reversals, and in which the same word — "study" — covers a metabolic-ward crossover and an online questionnaire. The purpose of this package is to make that distinction automatic.
What the package holds
Curated scaffolding your agent loads before it researches, so it starts from vetted ground rather than a cold search.
93
tier-classified sources
18
mapped concepts
7
named controversies
9
documented misconceptions
- Tier 1: 41
- Tier 2: 43
- Tier 3: 3
- Tier 4: 6
The questions and claims below are quoted from the package files.
Where the field disagrees
Each one carries real proponents on more than one side, so your agent cannot quietly pick a winner.
Is obesity primarily driven by the hormonal response to diet composition, or by energy intake within a regulated system?
0 named positions · The strong form is largely unsupported by controlled feeding experiments. Weak forms are uncontroversial and shared. The remaining genuine disagreement is about how much of population-level obesity the composition pathway explains, which no ward study can settle because it is a question about the food environment over decades.
Should saturated fat intake be limited, and is a nutrient-level target the right instrument?
0 named positions · Genuinely contested at the level of instrument and magnitude, not at the level of whether the replacement effect exists. A defensible summary: replacing saturated fat with unsaturated fat modestly reduces cardiovascular events; replacing it with refined carbohydrate probably does not; total mortality effects are unproven; and food-based advice may serve people better than a nutrient percentage.
Is industrial processing itself the causal variable, or a proxy for properties that could be measured directly?
0 named positions · Open, and the most consequential open question in applied nutrition right now, because policy is being written on a category whose causal content is unidentified. The productive framing is not category versus mechanism but which measurable properties — energy density, eating rate, protein and fibre content, hyper-palatable combinations — carry the effect.
Should evidence-grading systems developed for drug trials be applied strictly to nutritional evidence?
2 named positions · Unresolved and important. The most useful position for a learner is to hold both: certainty grading is right that nutritional evidence is weaker than the advice built on it implies, and its critics are right that "low certainty" is not the same as "no reason to act".
Is self-reported dietary intake fit for the causal purposes it is put to?
0 named positions · Both are partly right, and the practical rule is more useful than the verdict. Treat rankings across large intake contrasts as informative and absolute intake figures from questionnaires as uninterpretable. Where a finding depends on precise quantities of a single nutrient, treat it as a hypothesis. Where it depends on a broad pattern contrast replicated across instruments and populations, treat it as evidence.
Myths the package corrects
Widely held claims with the evidence that settles or bounds them.
“A calorie is a calorie, so food quality does not matter — all that counts is the number.”
debunked-as-stated
Both halves have to be held at once, and the precise formulation is short. As accounting, a calorie is a calorie: body energy stores change with the difference between energy absorbed and energy expended, and no dietary pattern suspends this. As causation, calories are not interchangeable, because the same energy delivered in different forms produces different amounts of subsequent eating, different lean and fat partitioning, and different adherence. The experiment that separates these cleanly is an inpatient overfeeding trial in which the same excess energy was given at low, normal and high protein: fat gain was essentially identical across groups and tracked excess energy, while lean mass and energy expenditure differed with protein (bray-2012). Energy determined fat storage; composition determined much of everything else. On the intake side, energy density is decisive — people eat a roughly constant weight of food, so energy per gram drives energy intake with no subjective sense of eating more (rolls-1999) — and isoenergetic portions of different foods produce very different fullness and subsequent intake (holt-1995). The mainstream scientific position was never the naive one, and says so explicitly: intake is regulated by the brain in response to food environment, palatability and energy density, not set by conscious arithmetic (hall-2022-ebm, hall-guo-2017).
“Eating fat makes you fat — dietary fat is stored as body fat, so cutting fat is how you lose weight.”
debunked-as-stated
The controlled comparison exists and it is unusually good. DIETFITS randomised six hundred and nine adults for twelve months to a healthy low-fat or a healthy low-carbohydrate diet, both taught with intensive support and both emphasising whole foods and minimal added sugar and refined grain. Mean weight change differed by well under a kilogram and was not statistically significant, and two prespecified predictors — a multi-locus genotype pattern and baseline insulin secretion — failed to identify who would do better on which diet (gardner-2018-dietfits). The same result appears across the comparison literature (dansinger-2005, sacks-2009-poundslost, johnston-2014) and survives formal certainty grading (naude-2022). On mechanism, metabolic-ward studies test the strong insulin claim directly: cutting carbohydrate lowered insulin substantially and produced slightly less body fat loss than isocaloric fat restriction (hall-2015), and an isocaloric ketogenic diet raised energy expenditure by roughly a tenth of the predicted magnitude while slowing fat loss (hall-2016-keto). Neither macronutrient is uniquely fattening at matched energy.
“Ultra-processed food is a problem only because it is energy dense — the processing itself is irrelevant, and a calorie-controlled diet of packaged food is equivalent to one of whole food.”
contested-correction
The causal probe is an inpatient randomised crossover. Twenty adults lived in a metabolic ward and ate ad libitum for two weeks on an ultra-processed diet and two weeks on a minimally processed diet, in randomised order, with menus matched for presented calories, energy density, macronutrients, sugar, sodium and fibre. On the ultra-processed diet participants spontaneously ate around five hundred kilocalories per day more and gained weight; on the minimally processed diet they lost it (hall-2019-upf). So processing is not irrelevant: with the standard nutrient variables matched, something about the ultra-processed menu still caused people to eat substantially more. That is the refutation of the first claim. The refutation of the mirror is in the same study's limits and in the surrounding critique. The trial is twenty people over four weeks on one pair of menus, unblinded by necessity, and it does not identify which property produced the effect. The category itself is contested: trained coders classify the same foods inconsistently, and it places wholemeal packaged bread and confectionery in the same group (gibney-2017). Plausible operative properties — energy density, eating rate, texture, protein and fibre content, hyper-palatable combinations — are measurable directly and may carry the whole effect (forde-2020, fazzino-2019, rolls-1999). The cohort evidence associating ultra-processed intake with disease is consistent but observational, with the usual confounding by socioeconomic position and health behaviour (srour-2019).
“A new study proves that eating food X causes disease Y — the researchers followed thousands of people for years and found a significant increase in risk.”
debunked
Six checks, which together are the transferable skill this package exists to deliver. First, design. A prospective cohort establishes association and temporality, not causation; only randomisation removes unmeasured confounding, and the word "proves" is almost never available. Second, exposure measurement. Intake is usually captured by food-frequency questionnaire, which is built to rank people by habitual relative intake and cannot deliver absolute grams or calories; measured against doubly labelled water, self-report under-reports substantially and the error is correlated with body weight and with the exposures under study rather than random (schoeller-vansanten-1982, lichtman-1992, subar-2003, freedman-2014). Correlated error can bias an association in either direction, so "the error would only weaken it" is an assumption, not a defence. Third, effect size. Typical diet-disease hazard ratios sit between about 1.1 and 1.4, and confounding by socioeconomic position, smoking, physical activity and general health behaviour can produce differences of that size unaided — which is why Bradford Hill put strength of association first among his considerations and why the smoking case, with relative risks an order of magnitude larger, is not a template for this field (hill-1965, ioannidis-2018). Fourth, comparator. Eating less of one thing means eating more of something else; the substitution usually determines the answer, which is why cohorts and replacement trials on saturated fat appear to disagree while measuring different contrasts (siri-tarino-2010 beside mozaffarian-2010). Fifth, endpoint. A change in cholesterol, insulin sensitivity or a glucose curve is a hypothesis about health, not health. Sixth, absolute risk and provenance: a twenty percent relative increase on a small baseline is a small absolute change, and the coverage may be describing a press release. The corrective is triangulation — agreement across designs whose biases point in different directions, not the accumulation of more studies of the same design (lawlor-2016).
“The studies show people who eat more antioxidants, or who have higher vitamin D, are healthier — so taking the supplement gets you the same benefit.”
debunked
This is the field's most complete and most repeated reversal, and it has happened in four separate nutrients. Higher carotenoid intake and blood levels were consistently associated with lower lung cancer risk; randomised beta-carotene supplementation increased lung cancer incidence in male smokers (atbc-1994), and an independent trial in smokers and asbestos-exposed workers was stopped early for higher cancer incidence and higher mortality in the supplemented arm (omenn-1996-caret). Randomised vitamin E increased prostate cancer incidence in healthy men (klein-2011-select). Pooled across trials, antioxidant supplements did not reduce mortality, and beta-carotene and vitamin E were associated with increased mortality in the trials at lowest risk of bias — a detail worth noting, since separating trials by quality changed the answer (bjelakovic-2012). Vitamin D repeated the pattern on a larger scale: low circulating vitamin D is associated with a very wide range of diseases, and randomised supplementation did not reduce invasive cancer or major cardiovascular events over around five years in more than twenty-five thousand adults (manson-2019-vital), did not reduce progression to type 2 diabetes in high-risk adults (pittas-2019-d2d), and did not reduce fractures in adults not selected for deficiency (leboff-2022). The likeliest explanation for the association is reverse causation and confounding: low vitamin D is a marker of illness, inactivity, adiposity and time indoors (autier-2014). Three mechanisms explain the general failure. The nutrient in food arrives with a matrix and at physiological dose; the supplement arrives isolated and often at a much higher one. The observational association may be produced by everything else that travels with a diet rich in that nutrient. And people who take supplements differ systematically from people who do not.
Learning paths
- fundamentals
- how-to-read-nutrition-evidence
- energy-balance-and-appetite
- macronutrients
- food-quality-and-processing
- micronutrients-and-supplementation
- dietary-patterns-and-the-diet-wars
- guidelines-policy-and-conflicts-of-interest
Domains
- nutrition science and dietetics
- energy metabolism and human physiology
- nutritional epidemiology and study design
- food science, food technology and processing
- appetite regulation and behavioural science
- public health nutrition, guidelines and policy
- evidence appraisal and epistemics

