safety.viz v1.7.0
See it move: the annotated v1.7.0 demo walks the new chart with captures and try-it-yourself steps against the live demo.
The gallery learns to answer "how long until…". A thirteenth renderer brings Kaplan–Meier time-to-event displays to the safety portfolio: step curves, confidence bands, and the at-risk table the FDA Safety Tables & Figures guide mandates beneath every time-to-event plot. No existing API is removed or renamed.
The Time-to-Event Explorer ships marked Experimental, and stays marked until an external clinical review confirms the Kaplan–Meier implementation. That is a deliberate withholding of confidence rather than a formality: until that review lands, safety.viz does not assert that these curves, confidence bands and at-risk counts are correct, so treat the estimates as provisional. The badge is on the gallery card and on every one of the renderer's pages, and it comes off in a later release under obot.roadmap#182.
What's new
- Time-to-Event Explorer — a new renderer for Kaplan–Meier safety displays (obot.roadmap#161, #128, PRs #129 and the sv#131 review rework). You compose the endpoint yourself, from the event data: flexible multiselect filters over the adverse events (body system, preferred term, seriousness, severity — configurable per study) define what counts as a qualifying event, and the chart shows time to each participant's first qualifying event, censored at end of follow-up from the population data. No endpoint list is hard-coded — the important events vary from study to study; configured one-click presets are a natural later release on the same filter state. Step curves by treatment group with censoring tick marks, pointwise 95% confidence bands (the
survival::survfitdefault family, cross-validated against it), and the at-risk / cumulative-events strip table — all derived from one estimator pass, so the table cannot disagree with the curve, and drawn without intro animation so every frame the chart shows is an estimate, never a transition. It consumes ADAE-shaped event records plus an ADSL-shaped population extract, defaults to cumulative incidence (1 − KM) with the estimator always named on the axis, and states plainly — in-app and in the clinical guide — the fixed derivation rule and where that estimator overreads risk. Ships Experimental pending the external clinical review above, and pending review of the design decisions. Try it live.
Also in this release
- NEWS.md becomes the running release log — this file; unreleased work now accumulates under a
(Upcoming)heading per the program-wide convention (#125, #127, convention: obot.roadmap#155). - Evidence baselines for every module refreshed on the canonical Linux environment as part of the time-to-event landing.
- Release prep:
dist/safety.viz-1.7.0/vendored, e2e fixtures repointed to the new bundle.
1 257 unit + 255 browser tests pass; evidence:check, requirements:check, build:check-dist, prettier and the site build all clean.
This release was drafted by Claude Code using Opus 5 and reviewed by @jwildfire
