The editorial pipeline

Discovery

Each Monday, a literature scan runs against PubMed, the Cochrane Library, ClinicalTrials.gov, bioRxiv, and the EFSA/FDA bulletin feeds. The query set is keyed to the 734 supplements currently in the database, plus a watch list of 60 emerging compounds.

The scan returns roughly 250–400 candidate items per week. About 80% are filtered out at this stage: animal-only studies, single-case reports, conference abstracts without full data, and articles in non-indexed journals.

Triage

An editor reads each surviving abstract and classifies it as: pivotal (changes a score), confirmatory (reinforces an existing score), null (no change but worth citing), retraction-watch, or signal-only (interesting but not yet actionable).

Pivotal items are routed for full-text review. Everything else is filed against the relevant supplement entry as a citation, with the classification logged.

Full-text review

The full text is read alongside the protocol (where available). The reviewer extracts: study design, sample size, blinding, primary endpoint, effect size with confidence interval, dropout, and funder. They write a one-paragraph extraction that lives in the citation record.

Funder type is set explicitly: public, nonprofit, industry-sponsored, mixed, or undisclosed. Tier-1 supplements require a public or nonprofit pivotal study; industry-only evidence keeps a supplement at tier 2 even if the effect size is large.

Independent second-read

A second editor reads the same paper without seeing the first reviewer's extraction. They produce their own one-paragraph summary and effect-size estimate.

If the two summaries diverge meaningfully on direction, magnitude, or tier implication, both reviewers meet (async, in writing) and either reach consensus or escalate to a third reviewer. Disagreements are logged so we can audit our own reliability over time.

Score recalculation

The reviewer applies the formula e×7 + s×4 + r×3 + o×2 + c×2 + d×2 with the new evidence input. The composite score (0–100) and the tier (T1 ≥72, T2 60–71, T3 40–59, T4 <40) are recomputed.

Any movement of more than 4 points or any tier change writes a row to _archive/score-changes.csv with the citation, the reviewers, and a one-sentence rationale.

Tier-1 gate (if applicable)

Any supplement promoted to tier 1 must satisfy the citation gate: at least one pivotal study with a public or nonprofit funder, indexed by PubMed, with a PMID stored in the entry. The gate runs as a CI check (scripts/tier1_gate.py) on every deploy. A failure blocks the deploy.

Publication

The change is committed, the score updates, and the supplement's "last reviewed" date is bumped. The citation appears in the bibliography and (where the study is open-access) in the supplement's per-page references.

For pivotal updates, an entry is added to the homepage "What changed this week" feed and surfaced in the New this week callout for seven days.

Reader feedback loop

Every supplement page carries a "Report inaccuracy" link. Submissions land in our editorial inbox, are triaged within 3 business days, and re-enter at step 3 (full-text review) when a reader provides a citation we haven't already considered.