The Physician's Longevity Supplement Framework: An Evidence Tiering

Ask a careful clinician for their longevity stack and, if they're honest, they won't give you a list. They'll give you a filter — the set of questions each candidate has to pass — and then apply it in front of you. The list is the output. The filter is the skill, and it's the only part worth learning, because next year's compounds and next year's trials will be different and the filter will still work.

Here is that filter, laid out as three axes. Run any supplement — the one you take now, the one in tomorrow's headline — through all three.

Axis 1 — How Strong Is the Evidence, Actually?

Not "is there a study." There is always a study. The question is what kind, and clinicians have a standard scale for it: the Oxford levels of evidence, which rank a claim by study design.

The uncomfortable truth this table exposes: most of the longevity supplement market lives at levels 4 and 5 while being marketed as though it lived at 1. The mechanism sounds like a level-1 fact. It isn't one.

Two examples of reading the axis correctly.

Creatine sits at 1a — a systematic review of a huge trial base confirming safety and efficacy.

^[1]

You can recommend it with genuine confidence because of how it is known, not how it sounds.

Taurine produced one of the most exciting animal results in the field: restoring it extended lifespan in mice and improved healthspan markers in monkeys.

^[9]

But that's Oxford level 5 for a human longevity claim — the outcome trial hasn't reported. Exciting and level 5 are not a contradiction; they're just two different facts that the marketing collapses into one.

The discipline of Axis 1 is refusing to let a compelling mechanism promote itself up the evidence ladder.

Axis 2 — Risk-Benefit, for This Person

Here is where a framework decisively beats a list, because the same compound sits in different tiers for different people.

Take berberine. The meta-analysis is strong — it reliably improves the components of metabolic syndrome.

^[4]

For a patient with rising fasting glucose, borderline lipids, and central adiposity, that makes berberine a strong recommendation: a real problem, a well-matched tool, a benefit you'll see on the next panel.

For someone whose HbA1c, lipids, and insulin are already clean, the same compound is a pointless gamble — a molecule that meaningfully alters the gut microbiome, taken to fix a problem they don't have, for a longevity benefit that has never been demonstrated in metabolically healthy people. Same evidence, opposite decision, because the patient is different.

This is why "what's your stack?" is close to meaningless as a question. A useful answer requires knowing the person's bloodwork, medications, and goals. The clinician's version of Axis 2 asks three things:

1. What's the realistic benefit for this person? A positive trial with a number-needed-to-treat of 200 means most individuals gain nothing even though the average moved.
2. What's the downside, and is it reversible? Creatine's worst case is water weight. An immunosuppressant's worst case is not. The asymmetry should dominate the decision.
3. What does it interact with? A supplement is only benign in isolation, and almost nobody takes them in isolation.

Omega-3 shows the payoff of dosing Axis 2 well. At 1 g/day in a general population, VITAL missed its primary endpoint.

^[2]

But in patients with elevated triglycerides on statin therapy, a high-dose prescription formulation cut cardiovascular events substantially.

^[3]

Right compound, right dose, right patient — the entire game is in the matching, not in whether omega-3 is "good."

Axis 3 — Can You Measure Whether It's Working?

This is the filter clinicians lean on hardest and consumers ignore most, and it's the most practically useful of the three.

If no biomarker can tell you whether a compound is doing its job, you are running an experiment whose result you can never read.

Sort the field by this axis and it splits cleanly.

Measurable — you can confirm the effect:

• Berberine → fasting glucose, HbA1c, lipid panel
• Omega-3 → triglycerides, omega-3 index
• Anything metabolic → a follow-up panel closes the loop

Partly measurable — you can confirm the mechanism, not the outcome:

• NAD+ precursors raise NAD+ — Martens confirmed nicotinamide riboside does it reliably.

^[7]

But NAD+ is the marker, not the goal. The strongest NMN outcome — improved insulin sensitivity — was confined to prediabetic women.

^[6]

You can prove your NAD+ went up. You generally cannot prove anything downstream changed.

Unmeasurable — you're flying blind:

• Most senolytics, at the consumer level. The mechanism is real, but no routine test tells you your senescent-cell burden fell. You take it on faith and re-dose on faith.

Urolithin A is the instructive exception that proves why Axis 3 matters: its trials didn't just track a surrogate, they measured mitochondrial function in muscle and improved functional endpoints.

^[5]

That's rare, and it's precisely why urolithin A grades better than most of its neighbours — not because the mechanism is fancier, but because it was measured in the tissue that matters.

Running the Framework: Three Worked Cases

The three axes combine into a single verdict. Watch them interact.

Creatine → Evidence 1a, risk trivial and reversible, benefit measurable via strength and function. Passes all three. Recommend broadly.

Resveratrol → Mechanism dazzling, but a clean human trial found no metabolic benefit,

^[8]

so evidence is negative at level 1b where it's been tested. Decline — this is the case that teaches Axis 1 to overrule a good story.

Rapamycin → Evidence for the mechanism is the strongest in the field, with human immune data,

^[10]

but Axis 2 is decisive: it's a prescription immunosuppressant with an irreversible downside profile, no human longevity outcome trials, and dosing extrapolated from animals. Not a stack decision at all — a supervised medical one. The framework doesn't ban it; it correctly routes it out of the DIY category.

Why the Framework Beats the List

Every January brings a new longevity headline: a compound, a clock, a trial. A list can't tell you what to do with it. A framework can — before you've finished reading the abstract:

• What Oxford level is this result? (Usually lower than the headline implies.)
• For whom is the risk-benefit actually favourable? (Rarely everyone.)
• Can I measure whether it worked in me? (Often not — and that's the tell.)

Three questions, and most of the year's hype sorts itself. That's the entire value of thinking like a clinician instead of a collector.

One Non-Negotiable: Verify What's in the Bottle

The framework assumes the capsule contains what the label says. That assumption is frequently false — analysis of products sold online found label inaccuracy was the norm in the category tested.

^[11]

No amount of evidence-tiering survives a mislabelled product. Third-party certificates of analysis are the price of entry, especially for compounds you cannot identity-check by eye.

From Framework to Stack

This is the reasoning layer. The output — the specific compounds, sorted into tiers with doses — is in The Ultimate Longevity Stack 2026, and the assembled daily version is on the protocol page. If you'd rather see the framework applied to the maximalist regimens that inspired it, that's our data-driven blueprint analysis.

And the meta-point, worth more than any single recommendation: learn the filter, not the list. The compounds will change. Good clinical reasoning doesn't.

This article is educational and not medical advice. It describes how clinicians reason about evidence; it is not a substitute for individualised care. Discuss any supplement or medication with your own healthcare provider.