Kaplan–Meier Curve

A non-parametric estimate of the survival function. The curve steps DOWN at each event; participants with incomplete follow-up are CENSORED (shown as tick marks) and contribute data up to the moment they leave.

Two survival curves — treatment versus control — descending over time, with censoring ticks along each. The vertical gap between the curves at any time is the treatment effect; the treatment curve here stays consistently above control, so more treated patients remain event-free.

The y-axis is the probability of still being event-free and starts at 1.0; each downward step is an event. Ticks mark censored patients (lost to follow-up or event-free at the end), not events. Compare the whole curves over time — never just survival at one arbitrary point.

Time-to-event outcomes — death, ROSC, re-attendance — are the currency of many EM and critical-care trials. KM shows when events happen, not just how many, and uses all available follow-up rather than discarding patients with incomplete data.

• Steps down = events · ticks = censored
• Compare curves with the log-rank test
• Reads survival over time, not a single endpoint

PARAMEDIC-2 (NEJM 2018) — 8007 out-of-hospital cardiac arrests, adrenaline 1 mg vs placebo. 30-day survival was higher with adrenaline: 3.2% vs 2.4% (adjusted OR 1.47, 95% CI 1.09–1.97). But favourable neurological outcome at discharge did not differ: 2.2% vs 1.9% — more survivors in the adrenaline arm had severe neurological impairment. A survival difference on a KM curve says nothing about the quality of survival — always report patient-important outcomes (here, neurological status), not just time-to-event.

• Ignoring censoring — treating censored patients as if they had the event (or as if lost data didn’t exist).
• Comparing survival at a single time point instead of the whole curve.
• Assuming proportional hazards when the curves cross.