How we know what we show
NephTox is a citation-grounded reference, not a content aggregator. Every incidence figure and mechanism claim ties to a primary source. This page explains how the atlas is assembled, verified, and kept current — the “how we know” layer behind 263 agents and 1,334 unique references.
Citation-grounded
Every quantitative claim (incidence, mechanism, dosing threshold) links to a real PubMed record by PMID. No invented citations. Curated conference abstracts that carry no PMID are the documented exception and are flagged as non-PubMed in their own block.
Adversarially checked
Citations are verified against the live PubMed record — title, authors, journal, and year — so a reference that doesn't resolve, or resolves to something else, is caught rather than shipped.
Freshness-tracked
Each profile carries the year-span of its evidence, and a freshness signal (up to date → dated) is derived from the newest reference so you can see at a glance how current the underlying literature is. A scheduled PubMed feed surfaces new onco-nephrology papers as they publish.
Structured & typed
Content lives in a typed dataset (injury signatures drawn from a 14-lesion taxonomy, nephron segments, severity, reversibility, drug class), so the same fact renders consistently across the drug page, the atlas, the dosing tools, and the data-visualization hub.
Human-in-the-loop
New agents pass through an auto-research draft → human approval queue before publishing; database overrides merge over the static seed only after review. The static, citation-grounded seed is always the source of truth.
Degrades gracefully
The public site renders fully from the static seed with zero environment variables — no login, database, or API key required. Optional services (accounts, alerts, the AI Oracle) enhance it but never gate the core reference.
Evidence, graded honestly
Not all evidence is equal, and NephTox does not pretend otherwise. References are typed by study design (meta-analysis, RCT, cohort, case series, pharmacovigilance, mechanistic), and pharmacovigilance signals (FAERS reporting odds ratios) are labeled as reflecting disproportionate reporting — not incidence or proven causation. Where the renal signal for a newer agent is extrapolated from its class rather than established directly, that is stated.
Explore the full bibliography on the evidence base, see the field's evolution on the insights hub, or read the guidelines the atlas is aligned to.
NephTox is medical-education content — not medical advice. It supports, but does not replace, the judgement of a qualified clinician and confirmation against current primary sources.