Skip to content
Bending the Curve Trial by Trial
?

How to read the evidence · 1 of 4

Reading a Kaplan–Meier curve

A Kaplan–Meier curve estimates, over time, the proportion of people who have not yet had a particular event, such as death or progression. It drops each time events occur. Not everyone can be followed until the event: some trials end first, or people leave. Those people are censored; they stay in the calculation until they leave and are then dropped from the group still being followed, but they are not counted as events.

Two numbers are often quoted from a curve. The median is the time at which the curve reaches 50%. A landmark rate is the proportion event-free at one chosen time, such as 18 months. A curve that flattens (a plateau) can have a high landmark rate and no median at all.

Teaching example: synthetic numbers, not trial data.

0%25%50%75%100%0510152025Monthsmedian 15 months
Each step down is one or more events. A + marks a person whose follow-up ended without the event (censored): they leave the risk set without counting as an event.
Median
15 months
The first time the curve reaches 50% or below.
At 12 months (landmark)
51.4% event-free
The proportion event-free at one chosen time.
Still being followed at 12 months
9 of 20
Fewer people are left to learn from as time passes.
Data table
MonthAt riskEventsCensoredEvent-free
02000100.0%
2201095.0%
3192085.0%
4171080.0%
5160180.0%
6151074.7%
7141069.3%
8130169.3%
9121063.6%
10111057.8%
11100157.8%
1291051.4%
1380151.4%
1470151.4%
1561042.8%
1650142.8%
1841032.1%
2030132.1%
2220132.1%
2410132.1%

Reading one in a paper

  • Look at the number still at risk under the curve. When it is small, the right-hand end of the curve is uncertain.
  • Check the median follow-up. A short follow-up can make a curve look better than it will in the long run.
  • A curve is an estimate with a margin of error. Papers report confidence intervals for medians and landmark rates.

Sources and further reading

These pages explain general methods in plain language and were written for this site; the sources below are the references to read next. Some classic papers have no abstract online, so the explanations stay at textbook level.

  1. Bland JM, Altman DG. Survival probabilities (the Kaplan-Meier method). BMJ. 1998;317(7172):1572.

    PubMedDOI

  2. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343-6.

    PubMedDOI