RMST — Restricted Mean Survival Time (area under S(t) to τ)
VerifiedAdvanced & specialized
Run this test straight away on a free built-in teaching dataset — no data of your own needed — or bring your own. Either opens the guided workspace: variable setup, assumption diagnostics, results with effect sizes and confidence intervals, figures, and APA-ready reporting.
Loading teaching datasets…
Or use your own dataset
Loading your datasets…
Area under the Kaplan-Meier curve from 0 to a pre-specified restriction time τ — interpreted as 'expected event-free time within the τ-year window'.
The engine reports per-group RMST(τ), RMST difference + ratio between groups with 95% CIs (Greenwood plug-in delta method + bootstrap validation), and KM curves with the τ marker. The reviewer-friendly alternative to the hazard ratio when proportional hazards fails — RMST is interpretable WITHOUT requiring PH or any specific hazard pattern.
Worked example
How much longer, on average, do patients survive up to 3 years?
Restricted mean survival time — the average event-free time up to a 36-month horizon — is compared between two arms; an assumption-light alternative to the hazard ratio.
The new arm gained about 5.2 more event-free months (RMST 29.0 vs 23.8 months; difference = 5.2).
The restricted mean survival time was 5.2 months longer in the new arm (29.0 vs 23.8 months).
Try it yourself: Load this ready-made sample and follow the run above.
When to use it
- Two-group RMST when PH is violated or not assumedImmunotherapy vs chemotherapy for stage-IV cancer: KM curves cross at 18 months (chemo better early, immunotherapy better late), so HR is meaningless.
- Single-group RMST estimateCancer registry: 3-year RMST for newly-diagnosed Stage-III patients = 28.3 months (95% CI 27.1, 29.5) — cleaner summary than median survival when ~40% are still alive at 3 years.
When NOT to — use instead
- PH assumption holds and HR is interpretableWhen PH holds, the Cox HR is a sufficient summary and is more standard. → Cox Proportional Hazards \u2014 multivariable survival with adjusted HRs
- Competing risks present (informative)Standard KM-based RMST overestimates survival when competing events exist. → Competing Risks \u2014 Fine-Gray + Aalen-Johansen CIF (multi-cause survival)
- Tiny sample (< 50 per group)RMST CIs are wide and unreliable on tiny samples. → Kaplan-Meier + log-rank \u2014 survival analysis with censoring
- Need covariate-adjusted estimateStandard RMST is a marginal estimand. → Cox Proportional Hazards \u2014 multivariable survival with adjusted HRs
Hypotheses
Parameter tested: RMST_g(τ) per group; RMST difference + ratio with 95% CI.
Assumptions (and what to do if they fail)
Check: Compare τ against per-group max times.
If violated: τ beyond the observed support forces the KM estimator to extrapolate flat — biasing the area estimate.
Check: Per-group censoring rates within 30 pct pts.
If violated: Informative censoring biases the area under Ŝ(t).
Check: Per-group events ≥ 10.
If violated: Too few events make Greenwood SE unstable; bootstrap CI is safer.
Ready to run a RMST — Restricted Mean Survival Time (area under S(t) to τ) on your own data?
Guided setup, automatic assumption checks, effect sizes, figures and an APA write-up.
Run this test →