Loading…
We use essential cookies (authentication, your saved goals/stack) by default. With your permission we’ll also enable privacy-respecting analytics (Vercel Web Analytics, anonymous load-time metrics) and error-replay diagnostics (Sentry — DOM snapshots only when an error fires) so we can fix bugs faster. Learn more about cookies
SupStack reads the published research, weights it by quality, and translates the conclusions into a 0–10 evidence score and a verdict word — so you can decide what's worth trying without wading through the literature yourself.
Finds and summarizes the actual studies behind each supplement, so you don't have to dig through journals yourself.
Presents key information — dosing, effects, safety — in a clear format so you can quickly understand what matters.
Compare options side-by-side and match supplements to your goals so you can figure out what's worth trying.
25 curated, evidence-based supplement stacks (protocols) for specific conditions — with dosage, timing, contraindications, and PubMed citations. All reviewed for FDA DSHEA compliance.
Not every supplement makes the cut. Here's how I decide what to include in SupStack:
Have a supplement you'd like to see added? Suggest it in the library and it'll be considered for future updates.
Each supplement gets a simple score from 1-10 based on how strong the research is. Here's what the numbers mean:
Multiple high-quality meta-analyses and RCTs with consistent, significant results. Effect sizes are meaningful and replicated across populations.
Several RCTs and at least one meta-analysis showing positive effects. Some inconsistency in results but overall evidence is supportive.
Limited RCTs with mixed results or primarily observational studies. Promising but more research needed.
Mostly animal studies, mechanistic research, or very limited human trials. Theoretical benefits not yet confirmed.
When the research is too thin to grade honestly — for example a compound studied only in cells or animals, with no human trials — we don't assign a number at all. A misleadingly precise score is worse than none, so these entries show this note instead and are ranked last. The underlying studies are still listed in full so you can judge them yourself.
Each supplement and outcome gets a one-line verdict word — Likely helps, Mixed evidence, Limited support, and so on. The verdict word is the primary evidence signal across the whole site; the 0-10 score sits beside it as a secondary, depth-of-evidence detail. The two answer different questions:
0-10 evidence score
Our editorial rating of how strongly the research supports this supplement for its main use. Reviewers weigh the amount and quality of evidence — meta-analysis and RCT count, sample size, and consistency — but the score reflects that expert judgment, not a fixed formula.
“How strong is the evidence it works?”
Verdict word
What the studies actually concluded. Computed from the share of studies whose effect sizes indicate the supplement helped vs. didn’t help.
“What does the evidence say?”
The verdict cutoffs are calibrated to be confident-sounding without over-promising — a supplement only earns “Likely helps” when at least seven out of ten graded studies came down on the positive side.
≥70% of graded studies showed benefit
A clear majority of studies with structured effect-size data found a positive effect. The evidence is consistent enough that the bottom line is unlikely to flip with new data.
50–69% of graded studies showed benefit
About half the studies that measured an effect found one. Promising direction, but not unanimous — new well-designed trials could shift the picture.
30–49% of graded studies showed benefit
Studies are split. Some found a benefit, others did not. Often points to a population-specific or dose-specific effect rather than a universal one.
<30% of graded studies showed benefit
Most studies did not find a clear positive effect. The supplement may still help in narrow contexts, but the broad claim isn’t well-supported.
Total studies ≥3 but graded <3
There are studies on the supplement, but most are mechanism / observational / animal work rather than randomised trials with measurable clinical outcomes. Treat conclusions as provisional.
<3 total studies
There aren’t enough studies yet to pool a verdict. Browse the underlying papers for individual findings.
“Graded studies” are the subset with structured effect-size data — typically RCTs and meta-analyses, plus observational studies that report a measurable change. Studies without that structured data contribute to the total count but don’t move the verdict either way.
The evidence score combines several factors to give you a quick sense of how solid the research is:
Randomized controlled trials (RCTs) and systematic reviews carry the most weight. Animal studies and anecdotal reports are noted but don't drive the score.
A statistically significant result doesn't always mean much in practice. The effect size tells you if the benefit is actually noticeable.
One study is interesting. Multiple studies showing the same thing? That's more convincing.
Who funded the study? Was it double-blind with a placebo? These details affect how much you can trust the results.
Every claim on SupStack is backed by real data. Here's the scale of what's behind the site:
550
Supplements
58,290+
PubMed Studies
740+
Drug Interactions
4,400+
Medical Claims Verified
Pulling studies off PubMed is just the start. Every supplement goes through a multi-step validation pipeline before anything goes live:
Studies are pulled from PubMed using supplement names, aliases, and scientific names. Each study is matched, ranked by quality, and tagged with DOI links so you can read the originals.
An automated audit checks every supplement across 11 categories: dosages against clinical ranges, safety ratings against reported side effects, drug interactions against known contraindications, and more. Issues get flagged by severity — critical, warning, or informational.
After enrichment, the actual study findings are compared against every supplement's claims. Safety keywords are scanned for concerns, dosage mentions are extracted and compared, and positive vs. negative findings are tallied against evidence scores. This is how we catch things like a safety rating that's too generous or an evidence score that doesn't match the research.
The pipeline runs on every update. New studies get pulled, validations re-run, and scores get adjusted. If a meta-analysis comes out that changes the picture, the data catches up.
A drug like metformin has tens of thousands of papers — listing them all would bury the signal in noise. So each supplement card shows a curated set of landmark studies: the handful that actually determine the verdict, each summarized and linked to PubMed. Everything else is still indexed and counted (and reachable in the Research library) — just not hand-summarized.
For each outcome a compound is used or studied for, we include the highest available tier of evidence or a field-defining study — chosen so the set as a whole spans the major outcomes, leads with the strongest designs, and includes the best contradicting evidence. A study qualifies if it is:
Every cited study must clear the same gates: a real, PubMed-resolvable ID; a study type that matches the actual abstract (no animal study dressed up as a human trial, no design paper passed off as results); and a finding grounded in that abstract. Crucially, including the strongest negative result is mandatory when one exists — so the evidence you see isn't cherry-picked. The full policy is public in our repository's STUDY-CURATION.md.
Knowing if something works is only half the picture. Each supplement also gets a safety rating:
Well-established safety profile with minimal side effects at recommended doses. Safe for most adults.
Generally safe but with notable drug interactions or contraindications for specific populations.
Significant potential for side effects or interactions. Medical supervision recommended.
Pick what you're trying to improve, and SupStack shows you which supplements have research supporting that goal, ranked by how well they match.
Each supplement is scored for how well it addresses each of your goals. Your match score is the highest of those — so a supplement that's an excellent fit for even one of your goals ranks as highly as if it fit all of them. We deliberately don't dilute a strong single-goal match by averaging it against your other goals.
When matches tie, we break the tie by whether the supplement has helped, been unclear, or not helped for your goals, and then by its overall evidence score.
How the "Head-to-head ranking" on each goal page is built — end-to-end, from indexed RCTs to the headline number you see.
esearch query per supplement to find primary RCTs, meta-analyses, and systematic reviews not already in our index. Filtered by "humans"[MeSH] and the bucket synonyms.temperature: 0.Every pooled estimate is automatically tagged with a confidence tier based on study count, total sample size, heterogeneity (I²), and the relative width of the 95% confidence interval:
The ranking sorts by tier first, then effect within tier — so a single small preliminary study with a large effect doesn't outrank a moderate-confidence pool with hundreds of subjects.
We test the pipeline against well-established meta-analyses. The sleep onset latency bucket is calibrated against:
Our population-stratified estimate for melatonin in primary insomnia (−6.4 min, 95% CI −7.2 to −5.6, k=4 RCTs, n=6,605) sits squarely within these published ranges.
Each ranked supplement on a goal page shows:
Anything underlined with a dotted line in the expanded panel has a hover-tooltip definition (Pooled effect, 95% CI, Heterogeneity, Hedges' g, Cohen's U3, etc.).
Research tells you what works on average — but you're not average. SupStack lets you run structured personal experiments to find out if a supplement actually works for you.
Pick a supplement you haven't started yet, select a goal (e.g. better sleep), and SupStack generates a protocol with recommended dose, timing, and duration based on the research.
You start with a baseline survey before taking anything. Then you follow the protocol and do periodic check-ins. At the end, SupStack compares your responses to your baseline and gives you a personal verdict: helped, no clear effect, or made things worse.
Most people try a supplement and “feel like” it helps — or doesn't. A structured experiment with a real baseline removes guesswork. It's not a clinical trial, but it's far better than vibes.
Important: Experiments are designed for supplements you haven't started yet. If you're already taking something, the baseline won't reflect your true starting point, and the results won't be meaningful.
The scoring system and goal matching are works in progress. If you have ideas for how to improve them — better ways to weight the research, additional factors to consider, or goals that should be added — I'd genuinely love to hear them.
This project gets better with feedback. Reach out at feedback@supstack.me.
Supplement research comes with its own vocabulary. Throughout the site, hover over underlined terms for quick definitions. Here are some common ones:
How much of a supplement actually gets absorbed and used by your body.
How long a substance stays active in your system before being eliminated.
Natural compounds that help your body adapt to stress and maintain balance.
Substances that enhance cognitive function like memory, focus, or creativity.