Three axes, never fused into one number
A single letter hides too much. Every claim is scored on three independent axes, and all three inform the grade you see. Clinicians and pharmacists who can grade evidence this strictly are the people AI labs pay to catch AI health answers that overstate what a biomarker or mouse study shows, and AITraining.jobs, a companion site to this one, lists those remote healthcare roles and what they pay.
Axis 1Certainty of the evidence
from GRADEFollowing GRADE, we rate how confident we can be that the true effect is close to the estimate, for one specific outcome: High, Moderate, Low, or Very Low. Randomized trials start High; observational evidence starts Low; and the rating is then adjusted down for risk of bias, inconsistency, indirectness, imprecision, and publication or funding bias, and (for observational data) up for a large effect, a dose-response gradient, or confounding that would only work against the finding.
Axis 2Direction and magnitude of the effect
from GRADE + USPSTF + NutriGradeKept deliberately separate from certainty. An intervention can be high-certainty and show no effect, which is a real, publishable verdict, not a gap. So the grade always states whether the evidence points to benefit, no effect, harm, or a contested/mixed picture, alongside the effect size where one exists.
Axis 3Translation to humans
from Oxford CEBM + NIA ITPThe axis the longevity field gets wrong. We record the best species the evidence reached (human trial, human observational, mammal, lower organism, cell) and the quality of the endpoint (a hard clinical outcome, a validated surrogate, or an unvalidated biomarker). These two leaps, animal-to-human and biomarker-to-lifespan, are named on every grade and always cost certainty.
The translation ladders
Two ladders make the leaps explicit. Where a body of evidence sits on each one determines how far its certainty can be trusted for a human being.
Species tested
- S1
- Human randomized controlled trial
- S2
- Human observational (cohort, case-control, Mendelian randomization)
- S3
- Mammal (mouse lifespan, especially the replicated NIA ITP)
- S4
- Lower organism (worm, fly, yeast)
- S5
- Cell culture, in vitro, or mechanistic reasoning
Endpoint measured
- E1
- Hard clinical outcome: all-cause mortality, disability-free survival, incident disease
- E2
- Validated surrogate: trial-validated against a hard outcome (LDL, blood pressure)
- E3
- Unvalidated biomarker: epigenetic clock, NAD+ level, senescent-cell burden, telomere length, VO2max
The anti-hype caps
These are the operational form of GRADE’s indirectness rule, and they cannot be overridden by the volume of evidence.
- 01Non-human evidence caps a human claim at Very Low certainty. Unlimited mouse data does not make a human claim moderate. Animal-only therefore tops out at grade G4.
- 02An unvalidated biomarker caps a hard-outcome claim at Low. 'Raises NAD+' can be well-proven while 'extends lifespan' built only on that marker is downgraded two levels.
- 03Every intervention is graded on its best-supported relevant claim, and the caveat that most limits it, usually a species or surrogate leap, is stated in plain language.
The grade scale
Seven grades, from G1 (proven in humans) to G7 (unsupported), each mapped to a GRADE certainty level. G6 and G7 are distinct: G6 is good evidence that something does not work, G7 is not enough evidence to say. The grade never travels without its certainty.
Proven in humansHigh to moderate certainty
Human randomized trials show benefit on a hard clinical outcome (mortality, disease) or a trial-validated surrogate. The highest bar, and one very little clears for lifespan itself.
Human-proven on markersModerate to low certainty
Human trials show benefit, but only on a biomarker or short-term function, not a validated surrogate for a hard outcome. Real human data, incomplete claim.
Human signalLow certainty
The human evidence is observational only, and has not excluded confounding (healthy-user bias, reverse causation). Suggestive, not established.
Animal-provenVery low certainty for humans
Reproducible lifespan or healthspan extension in mammals (often the NIA Interventions Testing Program), with no adequate human outcome trial. Unlimited animal data cannot make a human claim certain.
Early / preclinicalVery low certainty
The best evidence is invertebrate, cell-culture, mechanistic, or a small uncontrolled early human study. Biologically interesting, humanly unproven.
Contested or refutedModerate to high certainty of no benefit
Good human evidence points to no effect or to harm, or a once-promising claim failed in human trials. This is a real finding, reported as plainly as a positive one.
UnsupportedInsufficient evidence to assess
Not enough credible evidence to judge the specific claim, or the claim sits far above the evidence behind it: anecdote, mechanism, or marketing sold as proof.
The LEV Readiness Index
The grade tells you how good the evidence for one intervention is. The LEV Proximity Index answers a different question: how close is the whole field to escape velocity? It composes three weighted inputs (the milestone ladder, the demographic rate, and the human-proof gate) into 15.6 / 100. That live composition, and the drill-down into each input, is shown on the escape-velocity page; what follows here is the scoring method beneath it: how a grade becomes a readiness weight, the ladder, and how hallmarks aggregate. Only realized, replicated results move the index.
From a grade to a readiness weight
Each intervention’s grade maps to a readiness weight and a rung on the 9-level Longevity Readiness Ladder (LRL), measured against escape velocity rather than against a single disease. The load-bearing choice is the inversion in bold: replicated mammalian lifespan (G4) outranks a human observational association (G3), because a randomized, hard-endpoint animal result establishes causation a confounded correlation cannot.
The 9-rung readiness ladder
In the spirit of NASA’s Technology Readiness Levels and their biomedical adaptations, extended with two aging-specific top rungs that nothing has reached. That empty ceiling is the point: a drug proven for one age-related disease (rung 7) is not proven to slow aging (rung 8) or extend lifespan (rung 9).
- LRL 9
Escape-velocity gradeunreached
Proven to extend healthy human lifespan at scale, fast enough to contribute to escape velocity. Not achieved by any intervention.
- LRL 8
Human aging RCTunreached
A randomized trial shows AGING ITSELF is slowed, on a multi-morbidity or composite endpoint. Not achieved by any intervention.
- LRL 7
Human outcome RCT
A randomized human trial shows benefit on a hard outcome of an age-related DISEASE (mortality, incident disease, function).
- LRL 6
Human biomarker RCT
A randomized human trial shows benefit on an aging biomarker or function, not yet a hard outcome.
- LRL 5
Mammalian lifespan
Reproducible lifespan extension in a mammal (ideally the replicated NIA ITP). Causal, wrong species.
- LRL 4
Human association
An observational human signal (cohort, Mendelian randomization); not shown to be causal.
- LRL 3
Mammalian function
Benefit on aging biomarkers or function in a mammal, short of a lifespan result.
- LRL 2
Preclinical proof
Benefit in cell culture, a lower organism (worm, fly, yeast), or by mechanism. Interesting, humanly unproven.
- LRL 1
Hypothesis or refuted
A mechanistic rationale with no demonstrated benefit, or a claim good human evidence has refuted.
From interventions to hallmarks to the index
- 1. Each hallmark’s Evidence Readiness is set by its single best-proven lever, plus a small, capped credit for having several independent decent levers. Proof dominates; breadth never does.
- 2. The weakest hallmark is the rate-limiter. Escape velocity requires addressing every damage class, so the bottleneck (currently epigenetic alterations, at 8%) matters more than the average.
- 3. No hallmark has reached rung 8 or 9, which is why the apex cannot climb into the middle of the scale no matter how many promising levers exist.
The honest caveats
The milestone completions are modeled estimates from the cited public record, not measurements. The scoring is grounded in de Grey’s LEV framework and the LEV Foundation’s mouse studies, the record life tables of the Human Mortality Database and the UN (2024), from which we compute the rate of gain ourselves, GRADE, NASA/BARDA readiness levels, the López-Otín hallmarks, and the NIA ITP. The index’s own what it is, and is not limitations are stated once, on the escape-velocity page.
Worked verdicts
Calibration examples, applied the same way every time.
Human RCT (PREDIMED) plus large cohorts on hard cardiovascular endpoints.
Human trials raise NAD+ (a biomarker); a lifespan or healthspan benefit is unproven.
Replicated mouse lifespan extension (NIA ITP); no adequate human outcome trial. Very low certainty for humans.
Human trials are largely negative or contested despite years of marketing.
A human anti-aging claim resting on animal mechanism and anecdote; regulators have warned against it.
What it is built on
Nothing here is invented. Each part is bolted to an established framework, so a methodologist can recognize and check the reasoning.
How the evidence stays current
Grades are curated by hand. The evidence underneath them is not. Every day we pull new longevity studies and preprints from Europe PMC and new human trials from ClinicalTrials.gov, match them to the relevant interventions, and stamp each with a fetch time so freshness is visible and staleness is detectable. Each intervention page shows the latest research alongside its grade, and a grade carries a reassessment trigger (for example, revisit when a pending trial reads out).
Common questions
Is any longevity intervention proven to extend human lifespan?
No drug or supplement has been shown in randomized trials to extend human lifespan. Only a G1 grade means proven human benefit on hard outcomes, and for lifespan itself essentially nothing reaches it. The closest human evidence is for exercise and treating established disease risk.
Why is a popular supplement graded G4 or lower?
Because the grade reflects human evidence. Under GRADE, evidence that tops out in animals or cells is very-low certainty for any human claim, no matter how much of it there is. A strong mouse result is a G4, not a human fact.
How can something be graded G6 when the science looks strong?
G6 is a strong finding: good human evidence that an intervention does not work or is harmful. High certainty and no benefit is a first-class verdict, reported as plainly as a positive one, not treated as a lack of evidence.
Where does the underlying research come from?
Studies and preprints from Europe PMC (which unifies PubMed, full text, and the bioRxiv/medRxiv preprint servers), and human trials from ClinicalTrials.gov. Both refresh daily, and each item links to its source.
How product and book links are chosen
Some pages link to Amazon, and Longevity Review earns a commission when a visitor buys through one. As an Amazon Associate I earn from qualifying purchases. The commission never sets a grade, and the grade decides whether there is a link at all. A product is linked only when all three hold:
- 1. The intervention it serves is graded proven in humans or proven on human markers, and its bottom line says it works for what it is used for; or the product is the measuring tool for a marker we rate as actionable.
- 2. It is sold without a prescription. We never link to a prescription drug or a pharmacy.
- 3. The evidence is about a plain, standard form anyone can buy, not one company’s product, so we describe that form and link to it rather than to a brand we have not tested. A trial run by the people who sell the product does not qualify.
That is why creatine, a heart-rate strap and a grip dynamometer have links, and NAD+ boosters, resveratrol and most other supplements say plainly that they have none. Books are different: we link the ones readers are likely to pick up and grade each against the evidence, including the ones that run ahead of it. We never show Amazon prices or stock.
Not medical advice
Longevity Review is an educational resource. Nothing here is medical advice, a diagnosis, or a recommendation to take, stop, or change any drug, supplement, or protocol. Many interventions discussed carry real risks and interactions. Talk to a qualified clinician before acting on anything you read.