Let me be direct about something that is rarely said in functional medicine circles: the metabolic typing systems that dominated the field from the 1970s through the early 2000s were built on a genuinely important clinical observation, and they wrapped it in a claim they could not support.
The observation was sound, as far as it went. Human metabolism is not uniform, and people respond very differently to the same meal — that part is now well evidenced. What has never been tested is whether each person has a stable balance of food on which they do best over the long term. That is the plausible part, and the untested one, and this framework treats it as a starting hypothesis rather than a finding.
The failure was not the questionnaire. It was the claim that a laboratory measurement had identified an underlying metabolic type. That is the part that never had the evidence behind it, and it is the part that is still repeated. Blaming the questionnaire is the wrong diagnosis, and it leads to the wrong fix — replacing a self-report instrument with a panel of markers that have never been shown to predict what the panel is being used to predict.
What this is, and what the evidence for it actually is
Metabolic Natures is a framework for deciding where to start, not a test result.
What is well supported: people respond very differently to the same meal. One study tracked 800 people across nearly 47,000 meals and found high variability in blood glucose after identical meals DOI; another, of 1,002 UK adults, found the same for glucose, insulin and blood fats — several of its authors work for a commercial nutrition company, which we note as we would for any study DOI.
What is not established: that any one person does better long-term on one particular balance of food. When four named diets were compared for a year, weight loss tracked how well people stuck to the plan (r = 0.60) and essentially not at all which plan it was (r = 0.07, not significant). DOI The balance you can actually sustain is the one that works.
The specific categories are not validated by controlled trials, and we do not claim they are. No trial has tested whether assigning someone to an oxidative or autonomic category, and matching a macronutrient ratio to it, produces a better result. The closest test of that shape of argument is DIETFITS — 609 adults, healthy low-fat against healthy low-carbohydrate, twelve months, a difference of 0.7 kg between arms, and neither a genotype pattern nor baseline insulin secretion predicted which diet suited whom. DOI DIETFITS did not test oxidation rate or autonomic tone, so it is not a direct refutation — but it tested the logic, and the logic did not hold up.
So what are we claiming? That a structured way of recognising your own long-standing pattern, adjusted from what your test findings currently show, gives you an approach you can recognise yourself in and therefore sustain. And one thing more, which is the part that can be wrong: we write down, before your protocol starts, what we expect to change and by when — and we check it at retest. That prediction can come back not met. Nothing else in the framework can.
And what we are not selling. Metabolic Natures recommends no proprietary formula, and no product is required for the method to work. The lineage this framework comes from did the opposite — it defined its own efficacy as depending on a sole-source supplement range. That is named here because the same standard has to apply to a tradition we came from as to one we did not.
Read the corrections to the book, dated →"The problem was never that the instrument was a questionnaire. It was that the answer was presented as a measurement."
A brief history of where metabolic typing came from
The evidence the older systems actually rested on
This is checkable, and anyone can check it. Healthexcel — the most developed of the metabolic typing systems, and the one my own work descends from — published a bibliography for its framework. It lists roughly twenty sources across ten claimed foundations.
Every one of them is a book or an unpublished theoretical paper. There is not a single peer-reviewed trial in it, and the most recent entry is from the year 2000. The foundational items run Pottenger 1919, Harrower 1932, Lee 1946, Revici 1961, Bieler 1965, Watson 1972, Kelley 1976, Wolcott's unpublished theoretical model of 1983, Schenker 1989 and Wiley 1989. Two of the ten foundations are the blood type diet and Ayurveda.
That is not a hidden weakness someone had to dig out. It is the system's own reference list. A body of twentieth-century clinical observation is a legitimate thing to have; presenting it as a measured mechanism is not.
The study usually produced in support — read as written
The closest thing to a trial of metabolic typing is an undergraduate dissertation: University of Hull, April 2010, thirty-five participants from British Military Fitness over ten weeks, metabolic typing against Eatwell guidance. The metabolic typing arm lost 2.2 kg to 1.66 kg, with better body fat and waist figures.
The between-group interaction effect was not significant. Read as written, it is a null result on its primary comparison. Both arms also had a nutritionist, weekly diaries and phone support, and the metabolic typing arm differed in food quality as well, so the typing was never the isolated variable. It is a small, unblinded, confounded undergraduate study, and it is regularly cited as though it were a win. That is a check anyone can run.
And the one properly powered test of the underlying idea
DIETFITS randomised 609 adults to a healthy low-fat or a healthy low-carbohydrate diet for twelve months. The between-group difference was 0.7 kg. Neither a three-SNP genotype responsiveness pattern nor baseline insulin secretion predicted which diet suited whom. DOI
DIETFITS did not test oxidation rate or autonomic tone, so it is not a direct refutation of this framework or any other. But it tested the shape of the argument — measure a biological marker, match a macronutrient ratio to it, get a better result — and that shape has now been tested properly once, and it did not hold. “They measured the wrong marker” is only an answer if there is a reason why this marker is different, and at present there is no such reason that has been tested.
What questionnaires can and cannot tell you
None of this is an argument against clinical intake forms, and it is not an argument for them either. It is an argument about what an instrument is entitled to claim.
A questionnaire written in trait language — I tend to run warm, red meat leaves me feeling strong, I can comfortably skip a meal — asks about enduring disposition. It is the closest available proxy for long-standing pattern, and it is the instrument that lets a person recognise themselves, which matters more than it sounds. It is not a measurement of constitution. Nothing measures constitution. Self-report is contaminated by current state: someone fifteen years into a diet that does not suit them has adapted, imperfectly, and answers from inside that adaptation.
The functional panels have the opposite profile. Bicarbonate, triglycerides, total protein and the rest are objective and repeatable, and they are unambiguously a snapshot of the present. They move with recent diet, hydration, illness and training load. A panel taken today reports today.
So the tests are not a window onto an underlying type, and treating them as one is the same error the older systems made, in newer clothes. Sustained HPA axis activation, to take the most common case, pushes both instruments in the same direction at once — it produces sympathetic indicators on the questionnaire and on the panel. When that is happening there is no clean reading available from either side, and the honest position is that the type is provisional until the constraint clears.
What changes, then
| Problem area | Legacy metabolic typing | Metabolic Natures approach |
|---|---|---|
| What the instruments claim | The system identifies your metabolic type, from a questionnaire or from a panel depending on the version. | Neither instrument identifies a type. The questionnaire gives a starting hypothesis; the panels describe current function. Both limits are stated to the client. |
| When the two disagree | Resolved silently, usually in favour of whichever instrument the system was built around. | Recorded side by side, with the direction of the conflict named. Current function governs the protocol until the constraint clears. |
| Blocking factors | Acknowledged but not systematically integrated. | Treated as making both instruments less trustworthy at once, not one reliable reading and one distorted one. |
| How results are reported | A single type, assigned from whichever axis scored highest, with near-ties resolved invisibly. | Both axes, always — an oxidative position and an autonomic one, each with its strength shown, so how close to the middle you sit is visible rather than hidden. |
| Evidence claimed | A measured mechanism, supported by a bibliography of books and unpublished papers. | Clinical observation for the categories, stated as such; adherence for the mechanism, which has trial evidence behind it. The distinction is stated on every page. |
| What could show it wrong | Nothing. Foods were said to have opposite effects depending on which control system dominated, so any outcome could be explained after the fact. | A prediction written before the protocol begins, with named clearance markers and a stated window, reviewed at retest — including the outcome where the constraint cleared and the predicted shift did not happen. |
The part that can come back not met
Everything above is a subtraction. Here is the one addition, and it is the only part of the framework capable of generating evidence.
Before a protocol starts, four things are written down: the pattern the questionnaire identified, the constraint the findings identified, the specific markers that would show that constraint clearing — with their current values, their units and the direction of change expected — and the shift in food preference and food response predicted to follow, in a stated direction and within a stated window.
Clearance is then called against a retest of those markers and nothing else. Not symptoms, not how someone presents at review, and above all not the predicted shift itself — because if clearance were defined by the client reporting the change, the prediction could never come back not met. That escape hatch has to be nailed shut in the definition rather than left to good intentions.
There are four possible outcomes, not two: cleared, partly cleared, not cleared, and cleared but the predicted shift did not happen. The fourth is the falsifying case. Most frameworks of this kind have no box for it. If the named markers have not moved after two retest cycles, the answer is not to wait longer — it is that the constraint was misidentified or is not modifiable by this protocol, and that gets recorded as such.
Written down fifty clients in, that is fifty prospectively recorded observations, each an n of one. It does not validate the categories and must never be presented as though it does. It is the difference between no evidence and no evidence yet, and here is how we would know.
What Metabolic Natures is not claiming
It's worth being clear about the limits of the framework, because overclaiming is what damaged earlier metabolic typing systems' credibility.
Metabolic Natures is not claiming that the six categories have been validated. No controlled trial supports assigning people to oxidative or autonomic categories, or matching macronutrient ratios to those categories, and I will not tell you otherwise. The categories carry forward a body of clinical observation, and that is what they are.
It is not claiming that a laboratory panel reveals your constitutional type. It is not claiming that the calcium-to-phosphorus ratio, the eosinophil-to-basophil ratio, or any Krebs-cycle signature predicts which balance of food will suit you. Those claims come from the unpublished literature of the 1970s and 1980s, no validation of any of them has been found, and they should not be presented as established. Where they still appear on this site, they are being removed.
It does not explain all of health, it does not override clinical presentation, and it does not remove the need for a thorough investigation across the full range of HIDDEN stressors — hormonal, immune, digestive, detoxification, nervous system and energetic.
What it does claim is narrower and, I think, more useful. A structured way of recognising your own long-standing pattern, adjusted from what your findings currently show, gives you an approach you can recognise yourself in and are therefore more likely to sustain — and sustained adherence is the thing with trial evidence behind it. Then one prediction, written down before we start and checked at retest, which can come back not met.
That is a smaller claim than the field is used to making. It is also one I can defend.