Data edition of
Residue An evidence-graded reference for peptide science

How we grade evidence

A four-step scale, applied to the strongest human evidence we can verify for each compound and, separately, for each effect. The grade describes the quality of evidence, never whether a compound is good, safe or recommended. This is version 1.0 of the method.

A stack of plain bone and oxblood cloth-bound volumes with a brass magnifying glass in front

The scale

Every monograph carries one headline grade — the best evidence available for any use — and a table that grades each effect on its own. A compound can therefore be Grade A for one indication and Grade D for another. Tesamorelin, for example, is approved for excess abdominal fat in HIV-associated lipodystrophy, and its label says it is not indicated for weight loss; claims about weight loss would not inherit its A.

  1. A

    Grade A — Regulatory-grade human evidence

    The compound is (or was) FDA-approved for at least one indication, on the basis of adequate and well-controlled human trials. The grade applies to the approved use; other uses are graded separately.

    On Residue: Semaglutide, tirzepatide, tesamorelin, elamipretide, afamelanotide, bremelanotide; sermorelin on its historical approval. (7 compounds)

  2. B

    Grade B — Randomised human trials

    At least one published, randomized, controlled human trial supports the finding, but there is no FDA approval for that use.

    On Residue: Retatrutide, cagrilintide, thymosin alpha-1, LL-37, kisspeptin-10: published randomized trials, no approval. (5 compounds)

  3. C

    Grade C — Limited human data

    Human evidence is limited to small, uncontrolled, short or single-group studies, or trials not indexed with full methods; most support is preclinical.

    On Residue: BPC-157, ipamorelin, CJC-1295, GHK-Cu, selank, semax, epitalon, Melanotan II, DSIP: some human data, none of it decisive. (9 compounds)

  4. D

    Grade D — Preclinical only

    Evidence for giving the compound comes only from cell, tissue or animal studies. No interventional human data were found in PubMed or ClinicalTrials.gov results at our last fetch (observational measurements of naturally occurring levels may exist).

    On Residue: TB-500 (the fragment), KPV, AOD-9604, MOTS-c: no interventional human data found. (4 compounds)

Rules we apply

  • Approval beats everything, but only for the approved use. We check Drugs@FDA through openFDA and quote the Indications and Usage section of the current label from DailyMed. An accelerated approval counts as approval; we say so, and note when continued approval depends on a confirmatory trial.
  • A "B" needs randomization and a control. Open-label extensions, single-arm studies and comparisons against historical or non-randomized controls are Grade C, however positive.
  • Negative trials count. When a well-run randomized trial misses its primary endpoint, we say so in the first paragraph, not in a footnote.
  • Observational measurement is not intervention. Measuring a peptide's natural blood level tells us nothing about giving it. Such studies do not lift a compound out of Grade D.
  • Fragments are not parents. Trials of a full-length protein do not grade a short fragment of it (see TB-500 and thymosin β4).
  • Registration is not evidence. A study listed on ClinicalTrials.gov tells you that someone planned it, not what it found. Registry counts appear as context only.
  • We do not rank across trials. Two drugs tested in separate placebo-controlled trials are not compared as if they were tested against each other. Comparison pages lead with head-to-head trials where they exist and say plainly when none does.

How a claim gets onto the page

Residue is generated from a cached snapshot of public APIs. At build time a script queries PubMed E-utilities, ClinicalTrials.gov (API v2), PubChem PUG-REST and openFDA, and saves the JSON with its retrieval date. Each factual sentence in our prose carries a machine-readable citation marker that must resolve to a record in that snapshot — a PMID, an NCT number, a PubChem CID, an FDA application number or label, or a cached FDA page. If a marker does not resolve, the build fails. The current snapshot was taken on October 19, 2026.

Resolving is necessary but not sufficient, so an editor also checks each sentence against the abstract, label text or page it cites. Where the abstract does not support a statement, the statement is removed. Where two sources disagree, we report both — for example, the BPC-157 monograph notes a discrepancy between a systematic review's summary of a knee study and the study's own figures.

We cite abstracts, labels and registry records because they are freely accessible and stable. That has a cost: abstracts omit detail, and we do not report numbers that appear only in a paywalled full text we have not read.

Doses reported in trials

Some compounds have a page titled "doses in trials". These pages exist only where at least three registered interventional studies list a dose for the compound in their arms. They reproduce, verbatim, what investigators registered or published: dose, schedule, population and duration. They are a record of research design, not instructions. They omit titration advice, never convert doses between products, and are no substitute for a prescribing label or a clinician. Doses that appear only on vendor or forum sites are never included.

Limitations

  • PubMed counts depend on the query string, which we publish on each research page; they over-count where a term is ambiguous and under-count where synonyms are missed.
  • ClinicalTrials.gov text searches return some unrelated studies. We show the registry's own results and flag false matches we notice.
  • Grades summarize; they do not replace reading the study.

Independence

Residue lists no prices, links to no vendors, and carries no affiliate links or sponsored content. No manufacturer reviews or approves our pages. See disclosures and the editorial board.

Review

Every page is reviewed by the Residue Editorial Team before publication: a source-by-source check of each sentence, a re-derivation of each grade from the rules above, and a compliance check for anything resembling dosing or treatment advice. Review is editorial, not clinical; the team claims no individual medical credentials. The five stages are set out on the editorial team page.

Corrections

When we change a grade or correct a fact, the page's "last updated" and "reviewed" dates change and the change is entered in the corrections log.