Disclaimer

What this is and isn't.

glio.ai is a research and educational tool. It is not a medical, regulatory, or statistical authority. Read this page before relying on the tool's output for any decision that matters.

Short version This site provides AI-generated pattern matching against a small, manually curated database. It is not a substitute for biostatistical review, regulatory advice, or clinical judgment. If you are making a real trial-design or treatment decision, consult licensed professionals.

This tool is not

  • A substitute for biostatistical or regulatory advice
  • A clinical decision-support tool
  • A predictor of trial approval probability or commercial success
  • A replacement for peer-reviewed scientific consultation
  • A diagnostic or treatment recommendation tool
  • A licensed medical device

If you are designing a real trial

Consult, at minimum:

  • A licensed biostatistician
  • Regulatory affairs counsel (FDA, EMA, or your jurisdiction's equivalent)
  • An institutional review board
  • Domain experts in your specific cancer indication and mechanism of action

The output of glio.ai may be a useful starting point for those conversations. It is not a substitute for them.

About the AI

  • All analyses are generated by Anthropic's Claude based on a curated database of 31 historical trials.
  • AI outputs are not deterministic. The same input may produce slightly varied results.
  • The AI may make errors, omit relevant context, or generate plausible-sounding analysis that is not fully grounded in the underlying data.
  • Always verify the underlying claims (trial names, hazard ratios, p-values) against the cited primary sources.
  • The AI does not have access to information beyond the curated database. It cannot pull live data from PubMed, ClinicalTrials.gov, or any external system.

About the database

  • The 31 trials in the database were curated by one person from publicly available sources.
  • 31 trials is not a representative sample of the full GBM drug development landscape.
  • Pattern recognition across 31 cases is suggestive, not statistically powered prediction.
  • The author has done their best to verify each entry against primary publications, but errors are possible.
  • Trial outcomes can be reinterpreted as new data emerges. The database reflects published interpretations as of the last update date.
  • See the methodology page for full curation details and limitations.

Reporting errors

If you spot a factual error in the database, an incorrect citation, or a misleading analysis output, please report it by email to aklimahos@gmail.com or via the GitHub repository. Honest, specific reports help the tool improve.

Last updated: April 2026. This disclaimer may change as the tool evolves. Significant changes will be noted on the about page changelog.