The Supplement Cost-per-Studied-Dose & Evidence Dataset
Free and open (CC BY 4.0). 72 supplement products across 16 categories, each with its elemental dose, cost per studied daily dose, third-party certification, an evidence grade, and the PubMed IDs behind that grade. Built for journalists, researchers, developers, and anyone who wants supplement data that's actually structured and sourced.
This is our May 2026 release (v1.0). The live site now tracks more than 1,000 products. For current testing counts, see the tested database.
May 2026 release · v1.0
Ingredient Master Dataset — every layer, joined
The dataset above is product-level. This one is ingredient-level: one row per ingredient (27 of them) that fuses five separate data layers into a single record — so you can see, for any ingredient at a glance, what it costs to dose it properly, what the studied dose is, how often it shows up in FDA adverse-event reports, how many people are actually deficient, and whether a third-party-tested option even exists.
- Cost per studied dose — min, median, and our top pick per ingredient (our proprietary normalization)
- Clinical dose target — the studied dose (PubMed-cited), present for 13/27
- FDA CAERS/FAERS safety signal — total adverse events, serious-event rate, top reaction, present for 19/27
- NHANES deficiency prevalence — who's actually deficient, present for the 5 nutrients NHANES measures
- Certification availability — whether any third-party-tested product exists in the category at all
Coverage is uneven by design — we only join a layer where the source actually has the ingredient (NHANES measures 5 nutrients; CAERS covers 19 categories). Blank cells mean "no public data," not zero.
May 2026 release · v1.0
Condition × Ingredient Evidence Crosswalk
The map of what has been studied for what: 29 ingredients × 66 research topics, 137 links. Each link carries a relationship type — is this a nutrient studied in a deficiency, a dose used in studies for the topic, or a pair we've reviewed in depth — plus an evidence grade where one exists, the dose for that use, and the PubMed IDs behind it. Every row traces to a source: a PMID-bearing dataset or a researched page on this site. No invented pairs, no made-up grades.
⬇ Crosswalk CSV ⬇ Crosswalk JSON
Both granular and canonical condition IDs are included, so you can roll up (e.g. always-tired + energy-fatigue → Fatigue) or keep the specific page-level link.
More crosswalk tables
The supplement graph, sliced by other axes — every row carries its source. All CC BY 4.0.
Form × Bioavailability
39 forms across 18 ingredients — which form of each ingredient actually absorbs (glycinate vs oxide, methylfolate vs folic acid, ubiquinol vs ubiquinone), what it's best for, what to avoid it for, dose, and side-effects.
⬇ CSV ⬇ JSONWhy this dataset exists
Supplement information is scattered: clinical evidence sits in PubMed, product labels sit in the NIH DSLD, certifications sit across USP and NSF, and prices sit on retailer pages. Nobody joins them. So a simple question — "which product gives me the clinically-studied dose for the lowest cost per day, and how good is the evidence?" — has no single answer.
This dataset is that join. The column most people come for is cost_per_day_usd: not cost per pill or per serving, but the cost to take the dose actually used in trials. A $10 bottle and a $30 bottle routinely flip rankings once you normalize to the studied dose — and this is the only public dataset we know of that computes it across categories.
What's inside
- 72 products across 16 categories (magnesium glycinate, omega-3, vitamin D3, CoQ10, creatine, iron bisglycinate, methylfolate, B12, and more)
- Evidence grades on every row — 53 graded strong (systematic reviews / large RCTs), 19 moderate
- PubMed IDs linking each category's grade to its underlying trials
- Cost-per-studied-dose normalized to the clinically-studied dose, plus elemental dose, certification, and ASIN
Download
⬇ CSV (spreadsheet-ready) ⬇ JSON (with field definitions & metadata)
Prices in this dataset come from the catalog review of March 2026 and may be out of date — see our data-freshness policy. The retailer price at checkout is the price you pay.
Column reference
| Column | Description |
|---|---|
name | Full product name as listed by the manufacturer |
brand | Brand / manufacturer |
category | Supplement category slug (e.g. magnesium-glycinate, omega-3) |
form | Specific form (e.g. bisglycinate, ubiquinol, rTG triglyceride) |
dose_per_serving | Elemental / active dose per serving — not compound weight |
dose_unit | mg, mcg, IU, g |
serving_size | Physical serving description |
servings_per_container | Servings per container |
price_usd | Retail price (Amazon US) |
cost_per_day_usd | Cost to take the clinically-studied daily dose — our core normalization metric |
third_party_certification | USP, NSF, IFOS, Informed Sport, etc. (or None) |
evidence_grade | strong / moderate / limited — by highest available evidence tier |
primary_use | Primary clinical use case(s) |
pubmed_ids | PubMed IDs supporting the category evidence grade |
editorial_pick | Our pick label (best-value, quality, budget) if any |
amazon_asin | Amazon product identifier |
How we built it
Doses are reported as elemental/active ingredient per serving (elemental magnesium, not magnesium glycinate compound weight; combined EPA+DHA, not "fish oil"). Cost-per-day is calculated as (price ÷ servings per container) × (clinical dose ÷ dose per serving). Evidence grades reflect the highest available tier — strong for systematic reviews and large RCTs, moderate for individual RCTs or strong observational data, limited for preliminary or mechanistic evidence. Full detail on our methodology page.
License & how to cite
This dataset is released under Creative Commons Attribution 4.0. You're free to use it commercially, remix it, and build on it — just credit the source. We'd genuinely love to see what you make.
Cite as:
Verified Supplement Data. (2026). Cost-per-Studied-Dose & Evidence Dataset (Version 1.0) [Data set]. https://doi.org/10.6084/m9.figshare.32917331
Found an error or have a product to add? Email us — corrections are logged on our corrections page.
Related
- Methodology — how we collect, normalize, and grade
- llms.txt & JSON API — machine-readable access for AI agents
- Deficiency statistics — our other open data