Home / Statistics / Survival analysis - Kaplan Meier curve
Statistics
Survival analysis - Kaplan Meier curve
Reading and constructing Kaplan-Meier survival curves.
6 questions1 source pages1 fact-check flags
Images appear with the first
question taken from each source page — tap a question to open it.
6 questions
Q1What is a Kaplan-Meier curve?▸
Non-parametric estimate of the survival function
Visual representation of the cumulative probability of an event at a specific time (a type of cohort study with outcome plotted over time)
Commonly used in joint registry data for implant survivorship; X axis is time, Y axis is cumulative survival probability
The event is represented by a downward step in the curve
Q2What is censoring in a Kaplan-Meier analysis?▸
Subjects who drop out of the study for reasons other than failure
Right censored: items that have not yet failed
Left censored: items that failed before the start of the test
Interval censored: items that have failed but the interval is uncertain
Q3How are survival probabilities calculated in a Kaplan-Meier curve?▸
For each interval, survival = number of patients surviving / number at risk
Censored patients are not included in the denominator
Successive conditional probabilities are multiplied to give the cumulative probability
Large vertical step down = many deaths/failures; large horizontal step = few deaths
Q4What is the difference between a life table and a Kaplan-Meier curve?▸
Life table divides time into regular intervals and calculates survival at each interval (actuarial method)
Kaplan-Meier recalculates the survival rate each time a failure occurs (product limit method)
Q5How do you compare survival between studies?▸
Need 95% confidence interval (1.96 SD from the mean)
Upper line represents censored data if they survived; lower line assumes all censored data died
If CI overlap >25%, not statistically significant; use the Cox proportional hazard test or log rank test
CI widens on the right side as population size decreases
Q6Define conditional survival, median survival time and the limitations of KM curves.▸
Conditional survival probability: chance of surviving a specific time frame between failures
Unconditional survival probability: chance of surviving from the beginning of the study
Median survival time: time until 50% of the population survive
Do not extrapolate beyond the defined time frame, and only specific hard endpoints should be used
Fact check
If 95% confidence intervals overlap by more than 25%, the difference is not statistically significant — misleading — Overlapping confidence intervals do not determine significance; formal testing (log-rank/Cox for survival data, or a CI for the difference) is required. The >25% overlap rule is not a standard statistical criterion — source