Research & Evidence Method

The four tiers

What each label means

Strong

Established in humans

Multiple randomised controlled trials with consistent direction of effect, or approval by a major regulator on the basis of such trials. Adverse effects are characterised well enough to appear on a label.

Not a claim of safety for everyone : colistin sits here and is meaningfully nephrotoxic. It is a claim that the effects, good and bad, are known.

Mixed

Human data exist but disagree

Trials have been run in people, but they are small, short, conflicting, confined to one region's literature, or measure surrogate endpoints rather than outcomes that matter.

This is the most commonly misread tier. "Mixed" means the question is genuinely open, not that it leans positive.

Limited

Isolated or low-quality human data

One or two human studies, typically unblinded, uncontrolled, very small, or authored by parties with a direct commercial interest. Also used where development was abandoned after early-phase work.

A discontinued programme is informative. Compounds are usually dropped because they did not work, not because nobody noticed them.

Preclinical

No reliable human evidence

Cell culture and animal work only. The mechanism may be well described and the animal results striking; neither tells you what happens in a person, at an achievable exposure, over a meaningful period.

Most peptides discussed enthusiastically online sit here. That is the single most useful thing this site can tell you.

Assignment process

Five steps, in this order

Find the human trials first
Search trial registries before the literature. Registered studies that never published are themselves a finding.
Check registry against publication
Compare the pre-registered primary endpoint with the one reported. Divergence is downgraded.
Weight by design, not by result
A null result from a good trial outranks a dramatic result from a poor one.
Look for independent replication
Work concentrated in a single group or institution is capped below Strong, however voluminous.
Record the date and revisit
Tiers carry a review date. They move down as often as up.

Evidence pyramid from in vitro work through animal models and early trials to systematic reviews

Pattern recognition

Six ways a claim goes wrong

These account for most of the distance between what a study showed and what gets repeated about it.

Species laundering

A result in rats is restated without the species. "Accelerates tendon healing" is true of the rodent study and unestablished in people, but only one version travels.

Dose disconnection

Concentrations that work in a dish are frequently unreachable in living tissue. An in-vitro effect at 100 µM says little if plasma levels peak three orders of magnitude lower.

Surrogate substitution

A marker moves, so the outcome is assumed to follow. Raising IGF-1 is not the same as improving body composition, strength, or anything a person would notice.

Citation laundering

A review cites a review that cites a preprint. Follow the chain to the primary source and the confident claim frequently dissolves into a hedge.

Survivorship in the literature

Negative results go unpublished. A compound with five positive small studies may also have had eleven that never appeared.

Mechanism as proof

A plausible pathway is treated as evidence of effect. Most drug candidates with excellent mechanistic stories still fail in phase II.

Worked examples

The method applied

Four entries and the specific reasoning behind each tier. Reviewed against the criteria above.
Entry	Tier	Deciding factor
Tirzepatide	High-quality	Multiple large randomised phase III trials with pre-registered endpoints, independent replication, and regulatory approval across several jurisdictions.
Thymosin α1	Mixed	Genuine human trials across several indications, but results vary in direction and quality, and approval is regional rather than universal.
CJC-1295	Limited	Early-phase human pharmacokinetic data exist, but development was discontinued and no efficacy outcome was ever established.
BPC-157	Preclinical	Extensive rodent literature concentrated in a small number of affiliated groups; no published randomised controlled trial in humans.