Methodology deep dive Evidence tiers Reviewed May 2026

What "strong" actually means.

Five tiers. Each with an explicit definition, a study-count threshold, real named examples from the database, and the conditions under which a row moves up or down. No aggregated "trust score". No proprietary letter grade.

Adapted from the GRADE framework. Counts are current as of this build. Last reviewed May 26, 2026.

Back to the methodology → Browse rated claims

A tier is a claim about evidence quality, not about a supplement's value.
A "strong" caution is just as honest as a "strong" synergy. We sort by what we know.

The five tiers · Movement between tiers

The five tiers

Defined, with criteria.

Each tier has explicit criteria and real named examples from the database. No tier is a verdict on a product; each is a measure of how confident the underlying claim is.

Strong

Multiple high-quality RCTs or meta-analyses point the same way.

Criteria. At least two of the following: a Cochrane review, a published meta-analysis of three or more RCTs, two or more independent RCTs with consistent direction of effect, or strong mechanistic evidence backed by at least one confirmatory RCT.

Real examples. Vitamin C + Iron synergy · Iron + Calcium conflict · Iron + Zinc conflict · Zinc + Copper caution · Vitamin E + Vitamin C synergy · CoQ10 + Fish Oil synergy · 5-HTP + L-Tryptophan contraindication · Warfarin + Vitamin K2 conflict · Warfarin + St. John's Wort contraindication · Levothyroxine + Calcium timing.

Moderate

Consistent evidence from quality studies, with open questions.

Criteria. At least one RCT plus supporting observational or mechanistic data; or multiple smaller RCTs that agree on direction but leave dose, duration, or subpopulation effects unresolved.

Real examples. Vitamin D3 + Vitamin K2 synergy · Calcium + Magnesium timing · Magnesium Glycinate + L-Theanine synergy · Atorvastatin + CoQ10 Ubiquinol synergy · Warfarin + Vitamin E caution.

Moderate is the largest tier in the database. Most well-known supplement-supplement pairings live here.

Emerging

Promising signals, not yet confirmed at scale.

Criteria. One small RCT, multiple pilot studies, or strong mechanistic evidence without large-scale clinical confirmation. Labelled "emerging" so hope is not mistaken for proof.

Real examples. Ashwagandha + Rhodiola Rosea synergy · Ashwagandha + Magnesium Glycinate synergy · Ashwagandha + L-Theanine synergy · L-Theanine + Magnesium Glycinate synergy · Boron + Vitamin D3 synergy.

Limited

Sparse or conflicting evidence.

Criteria. Only a handful of studies exist, the studies disagree, or the designs were weak (no controls, very small N, single-site without replication). Rows here often describe plausible interactions where the literature simply has not caught up.

Why this tier is intentionally small. The temptation when building a database is to promote weak rows to "emerging" so the catalogue looks complete. We resist that. Rows that do not clear the bar stay marked "limited" until new studies justify a move.

Traditional

Centuries of use, minimal modern science.

Criteria. Long-standing use in established systems (Ayurveda, TCM, folk pharmacopoeia) with limited or no replicated randomized evidence in modern populations. This tier signals "people have used this for a long time" and nothing stronger.

How to read it. Traditional does not mean safe and does not mean effective. It means the evidence stream is different. Kept as a separate tier so it cannot be conflated with strong RCT data.

Shape of the database

Tier distribution, supplement-supplement rows.

Of the supplement-supplement interactions in interactions_database.json, here is how the evidence sorts.

Row counts by tier

Strong: 83 rows
Moderate: 177 rows
Emerging: 79 rows
Limited: 4 rows

The shape is intentional. Strong sits below moderate because the bar for "strong" is explicit (Cochrane or meta-analysis or multiple independent RCTs). Limited sits very small because we will not promote rows we cannot defend.

Visual

Strong 83

Moderate 177

Emerging 79

Limited 4

Bar widths are proportional to row counts within the supplement-supplement set. Traditional rows are tracked separately and listed on relevant supplement pages.

How to read a tier

Decision rules, plain English.

A tier is not a verdict on whether to act. It is a measure of how confident you should be that the effect is real. Here is how we would use each one.

Strong synergy. Default to taking them together (or in the recommended order). Skipping the pairing costs you measurable benefit.
Strong conflict or contraindication. Follow the recommendation without negotiation. These are pairs we are confident enough about to assume meaningful real-world risk.
Moderate. The direction is probably right; the magnitude may be smaller than headlines suggest. Apply the recommendation but do not reorganize your life around it.
Emerging. Treat it as a tilt, not a rule. Reasonable people can differ on whether the signal justifies the action.
Limited. Read the row, note the uncertainty, and discount it accordingly. Labelled "limited" so you can decide rather than us hiding the weakness.
Traditional. Useful as cultural and historical context. Not a substitute for randomized evidence.

Movement between tiers

When a tier moves, and what it looks like.

Tiers are revisable. The conditions for moving up and moving down are not symmetric, and both are public.

Upgrade

When a row moves up.

Two scenarios trigger an upgrade: a new meta-analysis or Cochrane review confirms a result previously rated moderate; or two new independent RCTs replicate an emerging result.

Upgrades require positive evidence, not the absence of negative evidence. "No contradiction yet" does not move a row up.

Downgrade

When a row moves down.

A high-quality meta-analysis published since the previous rating contradicts the direction of effect; a strongly-cited foundational study fails to replicate at scale; new safety data shifts the recommendation.

Downgrades are public in release notes. We do not quietly keep the old rating because users (or marketing) prefer it.

Real example

The Vitamin D3 + Vitamin K2 synergy was widely cited as "strong" in earlier popular writing. The 2023 meta-analysis of 14 RCTs found that vitamin K supplementation did not significantly prevent vascular calcification. The row is now rated moderate. The bone-density story still holds, but the vascular claim is no longer treated as established. The caveat is in the mechanism field on the live row.

Why no aggregate "trust score"

Per-claim, not per-supplement.

Evidence is deliberately not rolled up into a single supplement-level grade.

A single supplement can have strong evidence for one use, moderate for a second, and none for the third. Compressing that into one letter grade discards information that matters to your specific decision.

Magnesium glycinate has strong evidence for sleep onset, moderate evidence for blood pressure reduction, and emerging evidence for migraine prophylaxis. The right summary is three numbers, not one.

Tiers attach to claims, not to products. We will never publish a NutriStack score-out-of-100 for any individual supplement, because that number cannot be right for everyone.

See tiers in context

Every supplement page shows the tier on every claim.

The research site renders tiers per-claim, not per-product, so a moderate-evidence use does not get hidden behind a strong-evidence headline.

Open the research site → How tiers feed the score

NutriStack is an informational and organizational tool, not a medical service, and not a substitute for professional advice. Always consult a qualified healthcare professional before starting, stopping, or changing any supplement or medication.