What this is
An independent research instrument for diet–disease evidence. It pools published study effects into an association ledger — every exposure→disease edge with its pooled effect, confidence interval, GRADE certainty, and heterogeneity — then computes what the studies collectively support: agreement verdicts, disease-level rollups, and the specific next study that would settle each open dispute.
How it works
- Extraction gates — study records pass fail-closed checks (span-grounded numbers, CI sanity, mediator-adjustment flags) before entering the ledger.
- Verdict ladder — pooled edges are judged on the prediction interval, not the confidence interval: strong consensus, average agreement, consensus-on-no-effect, contested, unsettled, or sparse.
- Gap engine — for every unsettled edge, the instrument names the study that would move the needle: more power, a hidden moderator, or a scope split.
- Fail-closed throughout — a claim no data can defend is a claim not shown. Contested strands stay visible but are excluded from guidance.
Data status
The corpus blends live extracted records with demo values shaped like the literature. Society-level figures (prevalence, costs, attributable fractions) are clearly labeled where they are placeholder endpoints, and every number routes through a verification queue before any public claim. The instrument is designed to downgrade its own claims — a field that cannot downgrade cannot be trusted to assert.
Disclaimer
This is not a government website. Independent research instrument. Not affiliated with NIH, HHS, USDA, FDA, or any U.S. federal or state agency. Seeded NIH grant figures are public-record excerpts, not official agency products. It does not guarantee results and can be wrong. Not medical advice.