Methodology

How tier calls, sub-scores, source weighting, and review cycles work — published in full so anyone can audit our calls.

On this page Tiers Sub-scores Source hierarchy Funding-source policy Review cadence Interactions model What we don't claim Flag an error

Deeper reads Editorial pipeline Funder policy How we changed our mind Bibliography

1. The four tiers

Every supplement on this site is placed in one of four tiers. The tier is the headline call: a single-letter signal of how confident the evidence is and what kind of risk profile the supplement carries.

Tier	What it means	What moves a supplement here
Tier 1	Strong evidence. Backed by multiple large clinical trials and meta-analyses. Consistent, replicated results across independent research groups.	≥1 systematic review or meta-analysis showing a meaningful effect, plus ≥1 independently-funded confirmatory RCT. Industry-only evidence cannot drive a Tier 1 call.
Tier 2	Promising / situational. Supported by clinical trials but evidence is limited to specific populations, conditions, or dosages. May not benefit everyone.	≥2 RCTs in a defined population with a consistent direction of effect, but small total N or absent meta-analysis.
Tier 3	Trending. Currently popular in wellness culture. Popularity does not equal proof — many have limited clinical evidence despite strong social media presence.	Mechanistic plausibility, animal data, or single small RCTs only. We surface these so users can see what the evidence does and doesn't say.
Tier 4	Risky / avoid. Documented safety risks including organ damage, drug interactions, or death. Regulatory warnings issued. Do not use.	Any of: FDA warning letter or recall, EMA / EFSA negative scientific opinion, multiple peer-reviewed case reports of serious adverse events, or banned/scheduled status in any major jurisdiction.

Tiers are reviewed continuously (see Review cadence). When new evidence shifts the call, the supplement moves and a dated note explains why.

2. Tier promotion flow

How and when supplements move between tiers. Tier is a function of the composite score and the citation gate. A high score alone is not enough for Tier 1; a low score alone is not enough to demote from Tier 1. Both criteria have to align.

The promotion ladder

A supplement only moves between tiers when both conditions for the new tier are satisfied at the same time:

Score band. The composite must fall inside the new tier's range (T1 ≥72, T2 60–71, T3 40–59, T4 <40).
Citation gate (T1 only). At least one pivotal PubMed-indexed study with a public or nonprofit funder. See the funder policy.
Safety floor (T1 + T2 only). No active FDA / EFSA / TGA / Health Canada warning that would render the supplement inappropriate for the general adult population. Population-specific warnings (pregnancy, paediatrics, specific drug interactions) appear inline on the supplement page rather than blocking the tier.
Replication (T1 only). The pivotal effect has been seen in at least two independent samples — not necessarily two trials, but two analyses that don't share a single research group's data.

Tier 2 → Tier 1 promotion

Triggered when a Tier 2 supplement crosses the 72-point composite threshold and a new pivotal public-funded study lands in the citation set and the replication condition is met. The promotion is logged with the triggering PMID and the recalculation reasoning.

Worked example — Psyllium husk. Sat at 70 (T2) for two years on the strength of older trials. The 2024 meta-analysis (PMID 38688104) of 27 randomised trials, partially funded by the NIH NIDDK, lifted the efficacy score by enough to push the composite to 74. Replication condition was already satisfied. Promoted T2 → T1 in April 2026 with one citation as the gate-satisfying source.

Tier 1 → Tier 2 demotion

Triggered when a Tier 1 supplement drops below the 72-point composite threshold (rare but possible after a meta-analysis revision), or when the citation gate is no longer satisfied (the gate-satisfying study is retracted, or re-classified after a funder disclosure correction). Demotion writes a row to the score-changes archive with the triggering reason.

Worked example — a hypothetical herb. Held T1 on the strength of a 2018 RCT funded by an NIA grant. In 2025, the NIA grant disclosure was corrected to "supplemented by manufacturer in-kind product donation, manufacturer participated in protocol design." Reclassified the funder type from public to industry-mixed. The remaining public-funded evidence was insufficient to satisfy the gate alone. Demoted T1 → T2 with the funder-policy reason logged.

Tier 3 / Tier 4 movement

T3 and T4 movements happen on score change alone — no gate applies. A supplement moving from T3 to T2 does not trigger the citation gate (it only kicks in for T1). A supplement moving from T2 to T3 is logged as a routine score recalculation.

Safety-driven demotion

If a major regulator (FDA, EFSA, TGA, MHRA, Health Canada) issues a contraindication or recall, the supplement is moved to T4 within one deploy cycle regardless of efficacy score, with the regulatory citation logged. The composite score is preserved in the entry's history so the change is auditable.

Worked example — kava (kava-kava). Composite efficacy and short-term safety scores would put it in T2 territory. Long-standing FDA hepatotoxicity advisory (March 2002, reaffirmed 2014) and EFSA risk assessment cap it at T4 across the whole site, with the regulatory references in the entry's safety section.

How often does this happen?

In the prior 12 months: 12 net promotions (T3→T2, T2→T1), 4 demotions, and 1 safety-driven move to T4. The full audit trail is in _archive/score-changes.csv and surfaced summarily on the homepage "What changed" feed.

3. The six sub-scores

Each supplement carries six 1–5 sub-scores. They roll up into the headline 0–100 score shown on the card. The sub-scores let you see why a supplement scores the way it does, instead of treating the headline as a black box.

Field	Letter	What 1 means	What 5 means
Efficacy	`e`	Insufficient evidence of benefit	Conclusive — meta-analytic evidence with consistent direction
Safety	`s`	Documented serious harm at typical doses	Excellent — no adverse signals across large population studies
Research depth	`r`	Minimal — <3 published trials	Extensive — >20 published trials including meta-analyses
Onset	`o`	8+ weeks before noticeable effect	Immediate / hours
Cost / value	`c`	Poor value — high cost for marginal benefit	Excellent — meaningful effect for low cost
Drug interaction risk	`d`	Severe interactions documented	None known

The headline score is a weighted combination of these sub-scores, with efficacy and safety carrying the most weight. The exact formula is in app.js under calcScore and is open to inspection.

4. Source hierarchy

Not all citations are equal. Where we draw on multiple sources for a single claim, the higher-tier source is given more weight in the tier call. The hierarchy below is what we use:

Meta-analysisPooled effect across RCTs

Cochrane systematic reviewRisk-of-bias graded

Regulator dossierFDA / EFSA / EMA / NIH ODS / Health Canada

RCT (independently funded)Direct evidence in humans

Cohort / observationalUseful for safety signals

Mechanistic / animal / anecdoteHypothesis-generating only

We pull citations from eight free authoritative sources. The full registry is in sources/registry.json.

Source	Tier	Used for
NIH Office of Dietary Supplements	1	Health-professional fact sheets, RDAs, ULs, dose recommendations
EFSA Scientific Opinions	1	EU upper-limit values and dossier conclusions
EMA HMPC monographs	1	EU regulatory positions on ~150 herbal supplements
Cochrane Library	1	Risk-of-bias-graded systematic reviews
openFDA FAERS	2	Real-world adverse-event reports for safety signals
Health Canada NNHPD	2	~600 structured monographs with approved doses and contraindications
WHO monographs on selected medicinal plants	2	Traditional-use claims from a global authority
NIH MedlinePlus / NCCIH	3	Consumer-facing sanity check on tier calls

PubMed sits alongside this stack as the deep-dive layer for primary studies that aren't summarized elsewhere. When you see a citation in an article carrying a small NIH ODS / EFSA / Cochrane / openFDA badge, that's the source the claim is anchored to.

5. Funding-source policy

Industry-funded supplement trials report effect sizes ~20–30% larger on average than independently-funded ones, after controlling for design quality. We don't pretend this away.

Every cited study is tagged with funder, funder type (public / industry / mixed / nonprofit / none_disclosed), and a competing-interest flag where the authors declared one. The full schema is in docs/citation-schema.md.
When we summarize pooled effect sizes from industry-funded trials, we apply a ~25% downward adjustment in the narrative. The adjustment is documented per supplement in the review log.
A Tier 1 call requires at least one public or nonprofit-funded confirmatory study. Industry-only evidence cannot drive a Tier 1 call.
Citations in the source list visually flag industry funding (amber pill) and competing-interest disclosures (red pill) so readers can weigh them themselves.

Why 25% and not 0%: Industry-funded trials are not worthless — many are well-designed and add real evidence. We discount rather than discard. The number is a judgment call drawn from meta-research on funding bias in supplement and pharmaceutical literature; we'll revise if our own data tracking suggests a different magnitude.

6. Review cadence

Different content carries different stakes. We review accordingly:

Content	Cadence	Triggers
Tier 4 (risky/avoid) and safety articles	14 days or on regulator alert	PubMed + openFDA FAERS + FDA MedWatch RSS + EFSA alerts
Tier 1 (strong evidence) and breakthrough articles	30 days	PubMed + Cochrane
Articles on supplements for kids	30 days	PubMed + AAP guidance + NIH ODS pediatric
Tier 2 / Tier 3 supplements	60 days	PubMed + ODS / EMA
Evergreen guides and "myth-busting" articles	90 days	PubMed

If FDA MedWatch or EFSA publishes an alert mentioning a supplement we track, that supplement jumps the queue regardless of its last-reviewed date. The full cadence policy lives at docs/cadence-policy.md.

Each supplement card and article shows its Last reviewed date so you can see how fresh the call is.

7. How interactions are modeled

The interactions system is built around three pieces:

Pairs

Explicit positive synergies — supplement combinations that work better together than either alone. Vitamin D3 + K2 for bone density. Iron + Vitamin C for absorption. Curcumin + piperine for bioavailability. These have strength ratings 1–5 to reflect how mandatory the pairing is for the effect.

Cautions

Explicit negative pair interactions with severity. Either caution (additive risk that warrants attention) or avoid (combinations to skip outright). St. John's Wort + SSRIs is an avoid for serotonin syndrome risk. Ashwagandha + thyroid medication is a caution for additive thyroid modulation.

Mechanism groups

Supplements that share a mechanism are grouped, and any two members in the same group automatically cross-flag against each other. The current groups include bleed, serotonin, sedation, stimulant, hepatotoxic, hypoglycemic, hypotensive, seizure_lowering, nephrotoxic_minerals, heavy_metal_risk, thyroid_modulator, potassium_loss, vitamin_a_overlap, estrogen_modulator, immune_stimulant, and androgenic. Adding a supplement to the right group is often more useful than enumerating every pair.

When you build a stack in your profile, the system flags conflicts in three places: a chip on each supplement card, a grouped row inside the card describing the mechanism, and a top-of-plan banner enumerating every conflict. Variants of the same base supplement (different doses or formulations of melatonin, for example) are collapsed into a single pill so the count reflects how many distinct supplements interact, not how many entries we happen to track.

Drug-supplement interactions are a separate parallel system, currently being built out as Phase 2 of our improvement roadmap.

8. What we don't claim

Supplement Score is not medical advice. We summarize evidence and provide tier calls. Individual dosing, diagnosis, and treatment decisions belong to a qualified clinician who knows your specific medical history.

Specifically, we do not:

Diagnose conditions or recommend supplements as treatment for a diagnosed condition
Provide individualized dosing — the doses shown are population-level ranges from clinical trials
Replace pharmacist or physician review of supplement-medication interactions
Endorse specific brands or manufacturers
Claim that any supplement prevents, treats, or cures any disease

Our tier calls are a starting point for an informed conversation with a clinician, not a substitute for one. If you take prescription medications, see a healthcare provider before starting any new supplement.

9. Flag an error

If you see an inaccurate claim, click the Flag inaccuracy button on the affected article or supplement card. The form requires three things:

The claim you think is wrong
What you think it should say
A citation URL — PubMed, DOI, regulator dossier, or peer-reviewed paper

We require the citation up front because it dramatically reduces noise and turns reader feedback into a signal we can actually act on. Every submission with a citation gets triaged. If you provide an email, we'll write back.

Submissions with a Tier 4 / safety angle or a drug-interaction angle are escalated to same-day review — wrong information about high-stakes content is the highest-liability content on the site, and we treat it that way.