How letters get
assigned and updated.

Publicly we present grades as peptide × outcome because it reads cleanly. Internally the grading unit is narrower:

“BPC-157 for tendon healing” is not one evidentiary object if the underlying studies differ materially by population, route, or time window. Injectable BPC-157 in adults with musculoskeletal injury is not the same grade object as oral BPC-157 in a rodent gut-inflammation model.

Operational rules

• Always score against an InternalGradeKey, even if the public page collapses to a simpler peptide × outcome label.
• Only collapse multiple internal keys into one public row when evidence is directionally aligned and the collapse does not hide a weaker population, route, or time horizon behind a stronger one.
• If route, population, or time horizon changes the evidentiary story, maintain separate internal records and state the qualifier on the public page.

These thresholds are bands, not formulas. The letter is the editor’s defensible synthesis of the six sub-scores, the hard caps below, and the sign-off rules.

No peptide × outcome receives a letter grade unless the underlying evidence set clears a minimum count of qualifying studies. Below the threshold, the outcome carries Pending until editorial backfill completes.

Rationale: with fewer than 3 studies you cannot assess consistency, and consistency — replication, directional agreement, or deliberate contradiction — is the core evidence-grading judgment. Two studies can agree by chance; three is the smallest count where a pattern becomes visible. Aligns with GRADE and Cochrane norms for when certainty assessment becomes meaningful.

What counts as a qualifying study

• Peer-reviewed and published in an indexed journal, or a registered clinical trial with posted results. Preprints, conference abstracts without full publication, vendor white papers, and marketing collateral do not count.
• About the relevant InternalGradeKey — the right peptide, outcome, population, route, and time horizon. A rodent oral study does not count toward a human injectable count.
• Independent of other counted studies when assessing replication. Follow-on papers from the same lab on the same cohort count as one unless the design is materially different.
• Not retracted or flagged with unresolved integrity concerns at the time of grading.

Systematic reviews and meta-analyses count as one study each toward the threshold but strengthen the grade through sub-score 02.

Additional quality gates on top of the count

• A or B still require at least one well-powered controlled human trial. 50 animal studies and zero human trials cannot clear B.
• F still requires decisive evidence of null effect, harm, or medical regulatory rejection, and must be supported by at least 2 studies showing null/harmful effect — a single trial cannot falsify.
• Study-type weighting (RCT > cohort > case series > animal > in vitro) is handled by the sub-scores, not the threshold.

Pending vs Insufficient

Operational rule: when the 3-study threshold is not met, default to Pending if the gap is expected to be fillable through ordinary literature search, and Insufficient if the claim itself is the problem (vague indication, unverifiable sourcing, absent primary literature by design).

Promotion out of Pending

• Threshold met + rubric supports a letter → assign the letter, update lastUpdated, log a grade-history entry with fromGrade: “Pending”.
• Threshold met but rubric caps low (e.g. qualifying studies are all animal-only for a human-outcome claim and sub-score 03 is capped at 1) → assign the letter grade the rubric supports; use Insufficient only if the grading unit itself is still unworkable.
• Editorial review confirms literature is structurally absent → convert Pending → Insufficient with a short rationale.

Promotions out of Pending follow the same sign-off rules as any other grade assignment (see §12). The initial Pending → letter transition is treated as a one-letter change for sign-off purposes.

Why Pending and not a lower letter

The tempting shortcut is to call a 0–2-study outcome a D (“weak evidence”). The problem: D is a grading judgment, and grading requires the evidence set to be large enough to judge. Calling something D on 1 study claims more confidence in the negative than the data supports — it says “we looked and the evidence is weak,” when the honest statement is “we have not completed the review.” Pending forces that honesty.

Claim audits and peptide grades are related but not identical. A claim audit evaluates a public assertion using the research protocol. The grading layer only acts on structured outputs from that process, not on free-form prose.

Every claim audit that touches a peptide grade must emit one or more EvidenceConclusion objects:

Operational rules

• A claim audit never changes a letter grade directly. It produces EvidenceConclusion records that may trigger reassessment.
• OVERSTATED / FALSIFIED / WITHDRAWN normally imply gradeImpact = “mandatory_reassessment” when they touch a central sub-claim or decisive cited source.
• CONTESTED implies mandatory_reassessment when comparable-quality direct support and contradiction exist for the same internal grading unit.
• VALIDATED may justify possible_upgrade, but never auto-upgrades. The rubric still has to be re-run.
• DEPENDENT / SPECULATIVE usually imply none, unless the peptide page is currently presenting them as direct support. In that case the grade may stay unchanged while the page wording is corrected.

Every grade rolls up six weighted sub-scores, each rated 1–5 with a written justification visible on the peptide page.

The first three carry the most weight. Efficacy and human directness are what the grade is fundamentally about. Safety and regulatory context matter, but a perfectly safe molecule with no demonstrated effect still earns a low grade.

These rules are not optional. They exist to stop the most common grading inflation errors.

Sub-score 03 · Effect vs placebo

• 03 ≥ 3 requires controlled human outcome data in the relevant InternalGradeKey.
• 03 = 2 is the ceiling when some direct human outcome signal exists, but it is uncontrolled, retrospective, open-label, or otherwise not comparator-based.
• 03 = 1 is the default when no direct controlled human efficacy evidence exists for the relevant grading unit.
• Animal-versus-sham findings may strengthen sub-score 01 and narrative context, but do not raise sub-score 03 above what human evidence allows.

Sub-score 05 · Side effect profile

• Cumulative direct human exposure under 50 people, or follow-up under 30 days → 05 cannot exceed 3.
• No direct human exposure data, safety case mostly animal toxicology or mechanistic inference → 05 cannot exceed 2.
• “No adverse events reported” in a tiny pilot is an early tolerability signal, not a mature safety profile.
• Any credible serious adverse event signal, unresolved integrity problem, or major uncertainty about the administered material caps the score at 1–2 pending review.

Sub-score 06 · Regulatory status

Track these dimensions separately in page notes and internal data:

• medicalApprovalStatus — approval, non-approval, or review status for the studied use
• safetyWarningStatus — warning letters, withdrawals, safety alerts, refusals
• availabilityStatus — compounding, legal-access, or supply-chain status
• sportsProhibitedStatus — WADA or other sports-governing-body status

• Sub-score 06 primarily reflects medical-review and safety-regulatory context.
• WADA status is compliance information for athletes. It does not imply efficacy or clinical danger by itself.
• Legal availability or compounding restrictions alone do not push a grade to F.
• Only safety- or efficacy-driven medical regulatory action can materially support a downgrade toward F.

There is no rigid formula. The editorial heuristic:

This boundary must be applied consistently:

1. D

   We looked, and what exists is weak

   At least 3 qualifying studies and the rubric supports a weak-evidence letter. Animal data, mechanistic papers, uncontrolled human case series, or anecdotal evidence that is inspectable and clearly limited.

2. Pending

   We have not finished looking

   Evidence set has fewer than 3 qualifying studies and editorial backfill is reasonably expected to close the gap.

3. Insufficient

   Evidence set is structurally unworkable

   Target cannot yet be meaningfully assessed — claim underspecified, route/population/time horizon unclear, literature effectively absent by design, or decisive sources unverifiable.

An internal peptide grade is re-evaluated when ANY of the following occurs:

When a trigger fires, walk through this:

1. § 01

   Confirm the trigger is in scope

   Read the new evidence, claim-audit handoff, retraction notice, or regulatory action. Confirm it concerns the peptide, outcome, population, route, and time horizon under consideration. A new oral rodent study does not automatically change an injectable human grade.

2. § 02

   Import the EvidenceConclusion if the trigger came from a claim audit

   Do not translate prose by hand when a structured claim-audit output exists. Identify the sub-claim, affected internal grade key, touched sub-scores, and whether the outcome is wording-only, possible band move, or mandatory reassessment. If fields are missing, send it back for completion before changing the grade.

3. § 03

   Re-score only the affected sub-scores, applying the hard caps

   Identify which sub-score(s) the new evidence actually touches. Re-score only those first. Apply the caps in §07 — do not let narrative enthusiasm override them. Write updated justifications in plain language. Examples: a placebo-controlled human RCT typically touches 02 and 03; longer follow-up touches 04 and maybe 05; a safety warning affects 05 and 06.

4. § 04

   Re-roll the letter grade

   Apply the heuristic in §08. The letter may change up, down, or stay the same. A single positive Phase 2 trial can plausibly move C → B if well-powered and clinically meaningful; it does not move C → A on its own. If the internal grade key needs to be split by route/population/time horizon first, do that, then re-roll each resulting grade separately.

5. § 05

   Determine whether the publication packet changes

   Four common outcomes: wording correction only (update copy, citations, lastReviewed; no grade-history entry); grade unchanged with updated sub-scores (update notes and lastReviewed, no editorial note required); grade up or down by one letter (add Grade history entry, editor + second-editor review required); grade change of two+ letters or any move to/from A or F (Grade history entry, full editorial-board review, publish a CHANGE NOTE).

6. § 06

   Update related artifacts

   Grade changes ripple. Update the peptide's topGrade if this was the top outcome and the grade moved. Update any /protocols/[slug] page that includes the peptide as a component. Update any /claims/[slug] page that depends on the affected sub-claim. Update the home-page carousel and featured-peptides section if featured. Regenerate sitemap.xml on the next build.

7. § 07

   Editorial sign-off

   Wording correction only → author editor. No grade change with sub-score updates → author editor. One-letter change → author editor + one second editor. Two+ letter change or any move to/from A or F → author editor + second editor + clinical advisor. Retraction-, integrity-, or safety-driven change → same chain plus same-day publication once verified.

The grading layer needs clear authority boundaries:

Every grade change must leave a trail. The pattern:

We do not yet store this in the codebase — currently each peptide outcome has only lastUpdated. Next step: extend the OutcomeGrade type with an optional history: GradeHistoryEntry[] field, render the history on the peptide page as a small Grade history panel under the grade matrix, and surface it in the Drug JSON-LD.

Three distinct operations, three different workflows:

These never count as supporting evidence regardless of how persuasive they sound:

• Influencer testimonials
• Vendor marketing copy
• Forum or social-media anecdotes
• Vendor-funded white papers without peer review
• “It worked for me” reports
• Mechanistic plausibility absent direct evidence
• Review articles cited as if they were decisive primary efficacy evidence

These can be cited as context but they cannot move sub-scores 02, 03, 04, or 05 in a positive direction.

A multi-compound protocol is not a peptide × outcome pair. Borrowing the A–F peptide letters for a protocol would imply direct regimen-level human-outcome evidence that almost never exists.

Protocols use a distinct label set (see /protocols):

Public label: BPC-157 × tendon healing

Internal grade key, simplified: BPC-157 × tendon healing × adults with musculoskeletal injury × injectable/systemic exposure × acute/subacute healing window

Current state: Grade C, sub-scores 4 / 2 / 1 / 2 / 2 / 1, top outcome.

Why C and not B

The B rubric requires at least one well-powered controlled human trial plus sub-scores 02 and 03 both ≥ 3. BPC-157 × tendon healing does not meet that bar. Strong and replicated animal evidence plus sparse uncontrolled human signal is a canonical C, not a B.

What would move it to B

A positive, well-powered controlled Phase 2 or Phase 3 human trial showing clinically meaningful benefit in the relevant grading unit could plausibly move 02 to 3–4 and 03 to 3, producing a likely C → B reassessment if safety remains acceptable.

What would move it to A

At least one additional independent high-quality controlled human trial pointing the same way, plus stronger longer-term human safety data. A single positive Phase 2 trial would not be enough.

What would move it down

A decisive high-quality null RCT, a credible serious safety signal, or a safety-/efficacy-driven medical regulatory action could force reassessment toward D or F depending on how conclusive the new evidence is.