A browser-based tool for meta-analysis of restricted mean survival time (RMST) differences across trials, avoiding the proportional-hazards assumption inherently required by standard hazard-ratio synthesis.
Live dashboard: https://mahmood726-cyber.github.io/rmstmeta/
- Pools pre-computed RMST differences across studies (DerSimonian-Laird random-effects).
- Optional Hartung-Knapp-Sidik-Jonkman adjustment and prediction intervals.
- Leave-one-out sensitivity, funnel plot, Egger's test.
- ~1200 lines of dependency-free JavaScript — works offline.
Open index.html (or rmst-meta.html) in any modern browser. No build step.
For local development:
python -m http.server 8000
# then open http://localhost:8000/python -m pytest -q # Python smoke + parity tests
Rscript validate.R # numerical parity against R metafor (requires R on PATH)If Rscript is not on PATH, set RSCRIPT=/full/path/to/Rscript before running the parity gate.
| Path | Purpose |
|---|---|
rmst-meta.html |
the dashboard (main artifact) |
index.html |
landing page |
tests/ |
pytest smoke + R-parity tests |
validate.R |
R reference implementation for parity |
manuscript_stat_med.md |
Stat Med submission manuscript |
e156-submission/ |
E156 micro-paper bundle |
E156-PROTOCOL.md |
project metadata (E156 entry #148) |
See manuscript_stat_med.md for full methodological reference. The tool implements the RMST framework of Uno et al. 2014 (doi:10.1200/JCO.2014.55.2208).
See LICENSE (MIT).