Evidence-Based Supplement Intelligence: What That Requires
The phrase is used loosely enough to be meaningless. Applied properly it requires an evidence hierarchy, a bioavailability check, and a willingness to say no.
The Short Answer
Almost every supplement company describes itself as evidence-based, which means the phrase carries no information by itself. What it should require is specific: a hierarchy that ranks outcome trials above mechanism, a check on whether the studied concentration is achievable orally, attention to whether the marker moving corresponds to a benefit, and the willingness to reach a negative conclusion. A system meeting those four is doing something; a system citing a study for each ingredient is not.
The Hierarchy That Matters
Not all evidence is equivalent, and the ranking is well established in clinical epidemiology even where supplement marketing ignores it.
Strongest: outcome trials. Randomised controlled trials measuring clinical outcomes, ideally with meta-analysis across several. This is what statin and GLP-1 evidence looks like, and almost no supplement has it.
Next: functional trials. Randomised trials measuring function rather than a laboratory marker, such as physical performance, cognitive testing or symptom scores.
Next: surrogate marker trials. Randomised trials measuring a biomarker. Useful, and the marker must be validated as predicting the outcome, which most are not.
Next: observational cohort data. Associations at population scale, subject to confounding and reverse causation.
Next: mechanistic studies in humans. Demonstrating a pathway is engaged at an achievable dose.
Weakest: cell culture and animal work. Establishes plausibility and nothing about human effect.
The critical point is that the categories are not additive. Twenty cell-culture studies do not sum to one randomised trial, and a compound with abundant mechanism and no human outcome data sits at the bottom regardless of how many papers exist.
The Surrogate Problem
A marker moving is not a benefit, and the supplement literature contains several demonstrations of exactly this.
Homocysteine. B vitamins lower it reliably. Large randomised trials lowering it did not reduce cardiovascular events.
HDL cholesterol. Niacin raises it. Adding niacin to statin therapy did not reduce events and caused adverse effects.
Antioxidant capacity. Vitamin E raises measures of it. Outcome trials were neutral with harm signals at high doses.
Beta-carotene. Raises a marker associated with lower cancer risk observationally. Supplementation increased lung cancer incidence in smokers.
These are not obscure results, and together they establish that a surrogate endpoint requires validation rather than assumption. A validated surrogate is one where changing it by that means has been shown to change the outcome, which is a demanding standard that few supplement targets meet.
The practical implication for any recommendation: if the evidence is that a compound moves a marker, the question is whether moving that marker by that route has been shown to help. For much of this field the honest answer is that nobody has tested it.
The Bioavailability Check
| Compound | Mechanism evidence | Achievable plasma level |
|---|---|---|
| Resveratrol | Abundant in vitro | Extensively conjugated; free levels far below experimental concentrations |
| Curcumin, plain | Abundant in vitro | Very poorly absorbed; enhanced formulations are what trials used |
| Quercetin | Abundant in vitro | Poor; substantial drug interactions at higher doses |
| EGCG | Abundant in vitro | Low; hepatotoxicity reports at high-dose extracts |
| NAD+ itself | Central to metabolism | Does not cross cell membranes intact at all |
| Butyrate, oral | Well characterised in the colon | Largely does not reach the colon when swallowed |
| Collagen peptides | Specific peptides proposed as signals | Prolyl-hydroxyproline appears in plasma; the interesting exception |
This check settles more questions than any other single test, and it is the one most often skipped. A compound demonstrating an effect at micromolar concentration in cell culture, where oral dosing produces nanomolar free plasma levels, is not doing in a person what it did in the dish.
It also explains a recurring frustration: the mechanism is real, the paper is real, and the product does nothing. Nothing was fraudulent, the concentration was simply unreachable.
The Quality Question, Which Is Separate
Even where evidence supports a compound, it may not support the product in front of you, and these are different questions.
Which formulation was studied? Curcumin trials used bioavailability-enhanced forms. Omega-3 trials specify EPA and DHA content, not total oil weight. Probiotic trials used named strains. A product using a different form is not the studied intervention.
Does it contain what it claims? Independent testing has repeatedly found label mismatches, particularly in probiotics, melatonin and botanical extracts.
Is it standardised? Botanical extracts vary enormously in active content, and ashwagandha withanolide content and boswellia standardisation are practical examples.
Has it been adulterated? Testosterone boosters have been found containing undeclared steroids, and weight loss products containing pharmaceuticals.
Is it degraded? Fish oil oxidises, topical vitamin C oxidises, and probiotics lose viability.
Are the doses disclosed? A proprietary blend listing twelve ingredients by total weight cannot be matched to any trial, and cannot be checked for interactions.
A compound with good evidence and a product with poor quality produces the same outcome as a compound with no evidence, which is why the two questions have to be asked separately.
The Willingness to Say No
This is the sharpest test of whether a system is evidence-based, because it costs something.
Applied honestly, an evidence hierarchy produces a short list and a long list of exclusions. Ginkgo was tested properly for cognitive decline and failed. Glucosamine and chondroitin were tested for osteoarthritis and largely failed. B vitamins for cardiovascular prevention failed. Vitamin E failed. Beta-carotene caused harm. Tribulus does not raise testosterone.
Those are findings rather than gaps, and a system that presents them as unproven rather than as tested and negative is misrepresenting the evidence in a commercially convenient direction.
The same applies to the recommendation to take nothing. Where a person's status is adequate, their stack is already appropriate, or the evidence does not support either direction, the correct output is no change or a removal. A system that always finds something to add is optimising something other than the person's health.
And it applies to admitting uncertainty. Where evidence is thin, the honest output is a well-reasoned uncertainty with a defined observation window, which is more valuable than a confident number and sells considerably worse.
What This Looks Like in Practice
Four questions handle almost any supplement claim, and they can be applied by a reader without specialist knowledge.
What kind of evidence is this? Outcome trial, functional trial, marker trial, cohort association, human mechanism, or cell culture. The category matters more than the number of papers.
Is the studied dose achievable, and is this the studied form? Check the actual active content against the trial, not the capsule weight or the ingredient name.
If it moved a marker, was that marker validated? Homocysteine and HDL are the cautionary examples.
What would change my mind? If nothing would, the position is not evidence-based regardless of what supports it.
Applied consistently, these reduce most stacks substantially and produce a set of recommendations that can be defended when challenged. That is what the phrase should mean, and the difference between a system that does this and one that cites a study per ingredient is the difference between evaluating and marketing.
The AEONNN Perspective
AEONNN's Evidence Levels are this hierarchy made operational. Level A is clinically substantiated, Level B emerging, Level C experimental and educational, and every recommendation carries its level so a member can see whether it rests on outcome data or on mechanism.
Two checks sit alongside it and do most of the pruning. The Pharmacokinetics layer asks whether the studied concentration is achievable orally, which settles the polyphenol category and explains why a real mechanism and a real paper can still produce a product that does nothing. The Quality layer asks whether this product is the studied intervention, since curcumin trials used enhanced formulations, omega-3 trials specify EPA and DHA content rather than oil weight, and probiotic trials used named strains.
The hardest requirement is the negative one. Applied honestly, the hierarchy produces a long exclusion list, ginkgo, glucosamine, B vitamins for cardiovascular prevention, vitamin E, tribulus, and those are findings rather than gaps. It also produces outputs that recommend nothing, or removal. A system that always finds something to add is optimising something other than the member's health, which is why AEONNN handles removal as a first-class action.
Pillar Matrix mapping
Database Matrix layers
- Evidence Layer (PubMed, Cochrane, ClinicalTrials.gov)
- Meta / Consensus Layer (JAMA, BMJ, specialty society positions)
- Pharmacokinetics Layer (HMDB, PubChem)
- Quality / Formulation Layer (ConsumerLab, Labdoor)
Frequently Asked
What does evidence-based actually require?
A hierarchy ranking outcome trials above mechanism, a check on whether the studied concentration is achievable orally, attention to whether a marker is a validated surrogate, and willingness to reach a negative conclusion.
Is mechanism evidence enough?
No. Cell culture and animal work establish plausibility and nothing about human effect, and twenty such studies do not sum to one randomised trial.
What is the surrogate problem?
A marker moving is not a benefit. B vitamins lowered homocysteine, niacin raised HDL, and neither reduced cardiovascular events. A surrogate needs validation rather than assumption.
Why do compounds with good mechanisms fail?
Usually bioavailability. A compound acting at micromolar concentration in cell culture, where oral dosing produces nanomolar free plasma levels, is not doing the same thing in a person.
Is a compound with evidence the same as a product with evidence?
No. Trials used specific formulations, doses and strains, and independent testing has repeatedly found label mismatches, variable standardisation and adulteration.
What is the sharpest test of an evidence-based system?
Whether it ever recommends nothing, or removal. A system that always finds something to add is optimising something other than health.
How can I check a claim myself?
Ask what kind of evidence it is, whether the studied dose and form match the product, whether any marker it moved was validated, and what would change the claimant’s mind.
Evidence and review
Any dosage ranges cited here reflect the ranges used in published human trials, not personal recommendations. Evidence in this field moves, so this article is reviewed quarterly and carries its last-updated date above. Nothing here is intended as medical advice, and supplementation should be discussed with a qualified clinician, particularly alongside prescribed medication or an existing condition.