# Survival Curve — Situational Evidence

Use for time-to-event probability with censored observations. Do not substitute a smoothed line chart.

## Required

- Draw Kaplan–Meier estimates as steps when using Kaplan–Meier data.
- Mark censoring and explain the censor symbol.
- Label the event, time unit, cohort, and probability scale.
- Start probability at 1 or 100% unless the analysis explicitly begins after baseline.
- Use the selected colourmap consistently across cohorts and provide a legend or direct labels.
- Include numbers at risk when interpretation or clinical reporting requires them.

## Allowed choices

- Show confidence bands with restrained opacity when intervals are available.
- Use direct end labels for a small number of well-separated curves.
- Add a median-survival guide only when the median is estimable and relevant.
- Truncate the time axis when later estimates are too sparse, but disclose the rule and retain the risk table.

## Forbidden

- Do not smooth steps into continuous curves.
- Do not infer censoring, confidence intervals, medians, or hazard ratios from an image.
- Do not let a confidence band obscure peer curves or censor marks.
- Do not report a median when the curve never crosses 0.5.