A Kaplan-Meier (KM) curve is a graphical tool used in survival analysis to estimate the probability of an event not occurring over time. This kind of analysis is common in medical research, where the event of interest might be death or disease recurrence. The curve provides a visual estimate of the proportion of a group surviving a specified outcome as the study period progresses.
Deconstructing the Visual Elements
The graph is defined by two axes. The horizontal axis (X-axis) represents time, measured from the start of the observation. This time can be measured in units such as days, months, or years, depending on the event being studied.
The vertical axis (Y-axis) represents the estimated probability of survival, ranging from 1.0 (100%) to 0.0 (0%). The line plotted on the graph is a step-function that only drops vertically when one or more events occur.
The curve starts at 1.0 (100% survival probability) at time zero, as every participant is event-free initially. The probability calculation is updated only when an event happens, causing the vertical drops. The horizontal lines between the vertical drops represent intervals where the estimated survival probability remains unchanged.
Interpreting the Survival Line
The overall shape of the survival line reflects the rate at which the event is occurring within the studied group. A survival curve that drops steeply indicates a high event rate, suggesting a less favorable outcome or a shorter time until the event happens. Conversely, a flatter curve signifies a lower event rate, pointing to a better prognosis or a longer event-free period.
The horizontal segments represent periods where no participants experienced the event. When an event occurs, the line drops vertically. The magnitude of this drop is proportional to the number of events relative to the number of participants still under observation. A curve with many small steps suggests a large study population, while large, infrequent drops may indicate a smaller sample size.
The median survival time is the point where the survival probability crosses the 0.5 (50%) mark on the Y-axis. To find this value, trace a horizontal line from 0.5 to the survival curve, then drop a vertical line to the X-axis to read the corresponding time. This time represents the point at which half of the study participants have experienced the event and half have not.
Understanding Censoring Marks
Censoring accounts for incomplete time-to-event data collected during a study. On a KM curve, censored individuals are typically marked with a small tick mark or a vertical line placed directly on the survival line.
The most common form is right censoring, representing participants who did not experience the event by the study’s end or were lost to follow-up. These individuals are included in the survival probability calculation up to the time they were last observed, but their data does not cause a vertical drop. The presence of a censored mark indicates the person was removed from the “at-risk” population for subsequent probability calculations.
A high number of censoring marks, especially toward later time points, suggests that survival estimates become less precise and reliable. As the number of individuals remaining decreases due to events or censoring, the probability estimate relies on fewer data points. Interpreting the tail end of a KM curve requires caution if the number of participants still at risk is not provided.
Comparing Multiple Curves
KM curves are frequently used to compare the outcomes of two or more distinct groups, such as treatment versus control. The graph displays multiple step-function lines, each representing a different group’s survival experience. The group whose curve lies higher on the graph has the better outcome, indicating a greater probability of not experiencing the event.
The degree of separation illustrates the magnitude of the difference in outcomes. A large vertical gap suggests a substantial difference in the event-free probability. While visual separation provides an immediate impression, a statistical test is needed to determine if this difference is truly meaningful.
The log-rank test is the standard statistical method used to formally compare two or more KM curves over the observation period. This test generates a p-value, which measures the evidence against the hypothesis that there is no difference in survival between the groups. A sufficiently small p-value confirms that the observed separation is statistically significant.