Methodology — the audit document.

This is the audit document. Every grading decision Stack makes — evidence tiers, regulatory tiers, field reports, perspectives, calibration thresholds, supply-chain attribution — has a rule, and the rules live here. If you trust Stack's verdicts, you should be able to read this page and re-derive them yourself.

Stack is a discussion guide, not a prescription. Methodology transparency is what we owe a reader who is choosing whether to trust the discussion.

Each peptide and each specific claim is tagged with one of three legacy tiers (Established / Emerging / Experimental) plus a per-claim letter grade (A / B / C / D). The letter grade is the one we lean on now; the legacy tier is preserved for backwards-compatibility with prior copy.

Grade A — Strong evidence. Multiple human RCTs (n > 100 per arm), consistent direction of effect, registered trials, peer-reviewed primary publications. Example: semaglutide weight loss in STEP trials.

Grade B — Moderate evidence. At least one human RCT plus consistent observational data, or multiple high-quality observational studies with mechanistic plausibility.

Grade C — Limited evidence. Animal RCTs only, single small human study, or post-hoc community reports with strong mechanistic argument. The mechanism is plausible but the human data is sparse. Most peptide claims live here.

Grade D — Mechanistic / anecdotal only. No human studies, only theoretical pathway logic or non-systematic field reports. We surface these explicitly so they're not confused with stronger evidence.

FDA-approved — the peptide is the active ingredient of an FDA-approved drug. Example: semaglutide (Ozempic, Wegovy), tirzepatide (Mounjaro, Zepbound), tesamorelin (Egrifta).

FDA Cat. 1 — eligible for compounding under Section 503A by licensed pharmacies. The compounding pharmacy operates under prescription, the bulk drug substance is on the 503A Bulks List.

Pending PCAC review (Jul 2026) — formerly Cat. 2, removed from significant-safety-concerns on Apr 15 2026 by FDA, scheduled for PCAC advisory committee review July 23-24 2026. Removal does NOT mean approved for compounding — PCAC must still review and recommend addition to the 503A Bulks List. Affects: BPC-157, KPV, MOTS-c, DSIP, Epitalon, Semax, TB-500.

Pending PCAC review (Feb 2027) — second-batch peptides on FDA's reclassification track. Affects: GHRP-2, GHRP-6, Kisspeptin-10, PEG-MGF, GHK-Cu (injectable).

Research-grey — sold as 'research only / not for human consumption'. Not FDA-approved, not on the 503A Bulks List, no purity testing required by law. Most distributors operate in this tier today. Stack flags this tier explicitly so a user can see what they're entering.

Every reference has four metadata tags beyond URL + label: study type, funding source, last-verified date, retraction status.

Study type: RCT (randomized controlled trial) > observational > case-report > review > mechanistic > anecdotal > regulatory-document. Visual weighting on the reference card reflects this hierarchy — RCTs render bigger and bolder than anecdotal reports.

Funding source: NIH / industry / self-funded / mixed / academic / unknown. When public metadata doesn't disclose funding, we mark it 'unknown' rather than guessing. Industry-funded studies are flagged so a reader can weight them appropriately.

Last-verified date: when Stack last checked the reference URL resolves and the cited paper still exists in its original form. Stale (>180 days) freshness stamps render in warm-amber as a flag to re-check.

Retraction status: active / questioned / retracted. We cross-reference cited PubMed papers against Retraction Watch and PubPeer. A retracted paper triggers a banner; a questioned paper renders with a warning chip.

Stack grades CLAIMS, not whole peptides. 'BPC-157 promotes tendon repair' might be a Grade C claim while 'BPC-157 reduces gut inflammation' might be Grade D — same peptide, different evidence base. The legacy peptide-wide evidence tier (Established / Emerging / Experimental) is a summary; the per-claim grades are the substance.

Each claim cites supporting references by index and (when applicable) excluding references — the studies we considered but did not weight (wrong population, retracted, methodology too weak). Showing what we excluded is part of the discipline.

Field reports are distilled themes from real user communities (Reddit, peptide forums, podcaster transcripts, peptide-curious creators). They're paraphrased, not quoted — the source is named and the volume is honest, but no individual identifying details are reproduced.

Inclusion threshold: a theme appears in field reports only when it's surfaced by at least three independent community sources OR by one author with deep on-topic credibility. Single-anecdote claims are not surfaced.

When a peptide is contested, Stack surfaces multiple named perspectives: a clinician view, an RCT/literature view, a biohacker view, a regulatory view. Each perspective is attributed to a real named source (a published clinician, a specific paper, a named creator, a named regulatory document) — never anonymous, never aggregated.

This is the substrate-grounding move: claims point back to specific named entities you can verify, not to a homogenized 'experts say' voice.

Stack will eventually surface aggregate user-outcome data alongside literature claims — but only when N is large enough to be honest. The thresholds:

N < 10 — not displayed. The sample is too thin to claim anything.

10 ≤ N < 30 — labeled 'Early signal' with explicit thinness disclosure. Read directionally only.

30 ≤ N < 100 — labeled 'Emerging (Stack data)' with N + methodology + Wilson confidence interval visible.

N ≥ 100 — labeled 'Established (Stack data)' as a primary editorial claim that can be compared against literature.

Enforcement is server-side in the `peptide_calibration_stats` SQL view (HAVING clause). The UI N-gate is a redundant defense — the privacy + honesty discipline lives in the database.

For each peptide where public data supports it, Stack maps the supply chain across five layers: API manufacturers (the chemical synthesis houses), pharma brands (FDA-approved end products), 503A compounding pharmacies (US legal-prescription path), Mexico channels (COFEPRIS-licensed clinics + compounding pharmacies + distributors), and research-grey channels (the 'for research only' market).

Every supply-chain entry cites a public source URL and a last-verified date. Accreditation status is captured where it exists: PCAB for compounding pharmacies, FDA registration + Green List for API manufacturers, FDA approval year for pharma brands, COFEPRIS reference number for Mexico channels.

Stack does NOT make quality claims about specific vendors. Whether a particular research-grey distributor's product is actually pure is not something Stack can verify without a lab budget. The supply-chain layer is factual public-source attribution — knowing who is in the chain — not endorsement of any link in it.

Stack accepts no vendor advertising, no affiliate revenue, no compounding-pharmacy sponsorship, no peptide-distributor referral fees. Editorial output is independent of any commercial relationship. The only revenue path on Stack's roadmap is verified-listing fees from licensed compounding pharmacies that pass an independent verification check — and those listings are clearly marked as paid placement when they ship.

Per-reference funding tags surface industry-funded studies so readers can weight them appropriately. We do not exclude industry-funded studies — they're often the largest and most rigorous trials available — but we mark them.

Stack is bilingual EN/ES from day one. Spanish copy is human-edited not machine-translated for editorial sections. Russian-language sources (the Semax, Selank, Cortexin literature) are translated by domain-aware editorial pass and marked with a Russian-language chip on the reference card so readers see the translation chain.

Stack does not prescribe doses. Stack does not recommend specific vendors or sources. Stack does not claim to verify product purity. Stack does not interpret your individual lab results. Stack does not replace a clinician.

Stack maps the territory worth discussing with a qualified provider, with public-source attribution and the discipline to mark what we don't know.

Regulatory archive: weekly Wayback-snapshot cron (Sundays 04:00 UTC) preserves every cited regulatory document so the link won't rot. Pulse digest: weekly publication summarizing the week's regulatory + scientific peptide news with primary-source citations. Per-reference last-verified date is updated when Stack confirms the URL still resolves and the cited paper still exists.