Slide 1
Survival analysis measures time until a defined event
Survival analysis studies the time until a defined event. The event may be death, disease recurrence or another outcome specified by the study.
More to Read
The word survival is broader than staying alive in this statistical setting. Before interpreting a graph, readers need to know the event and the point from which time was measured.
Slide 2
The curve estimates the chance of remaining event-free
A survival curve estimates the probability of remaining free of the event through time. A Kaplan–Meier curve drops in steps when observed events occur.
More to Read
The size of a step depends on how many people are still being followed and remain event-free at that time. The same number of events can therefore produce different-sized steps at different stages.
Slide 3
Observation can end before the event occurs
A participant may leave observation before experiencing the event. Censoring records that the event-free time is known only up to a particular point.
More to Read
Censoring is not the same as recording an event or assuming permanent survival. It preserves the information available before follow-up ends while acknowledging that the later outcome is unknown.
Slide 4
Median survival marks the halfway probability
Median survival is the time when the estimated event-free probability reaches one half. It may not be reached during the available follow-up.
More to Read
The median is not the average of every participant’s lifetime and does not predict an individual’s outcome. Two groups can share a median while their curves differ elsewhere in time.
Slide 5
Later estimates often rest on fewer people
Late parts of a curve may contain much less information than early parts. Numbers at risk and uncertainty intervals help show that difference.
More to Read
Methods also rely on assumptions about censoring. If losing contact is related to prognosis, simply treating those losses like routine study endings can mislead the estimate.
Sources
- Time-To-Event Data Analysis
Columbia University Mailman School of Public Health
- How to read a Kaplan-Meier survival plot
The BMJ