Kaplan Meier Estimate

The Kaplan Meier estimate is a nonparametric statistic used in survival analysis to estimate the probability of surviving (or remaining event free) beyond successive time points. It is displayed as a step function curve that decreases over time, with each step corresponding to an observed event, and accounts for the fact that some participants leave the study before the event occurs (censoring).

Survival analysis is fundamental to clinical research, epidemiology, and health economics. Many outcomes of interest, including death, disease recurrence, treatment failure, and hospital readmission, are defined as the time until an event occurs. The Kaplan Meier method provides a visual and numerical summary of survival experience that is intuitive, widely understood, and reported in most clinical trials with survival endpoints. It forms the basis for comparing survival between treatment groups (using the log rank test), for estimating median survival times, and for presenting evidence to regulatory agencies and HTA bodies.

The Kaplan Meier estimator calculates the survival probability at each time point where an event (such as death or recurrence) is observed. At each event time, the conditional probability of surviving past that time is calculated as the number of subjects who survived beyond that time divided by the number at risk just before that time. The cumulative survival probability is the product of all the conditional survival probabilities up to and including that time point.

Censoring is the defining feature of survival data. A censored observation occurs when the event has not been observed for a participant, either because the study ended before the event occurred, the participant was lost to follow up, or the participant experienced a competing event. The Kaplan Meier method handles censoring by removing censored individuals from the denominator (the number at risk) at the time of censoring, without counting them as events. This assumes that censoring is noninformative; that is, censored individuals have the same future event probability as those who remain in the study.

The Kaplan Meier curve is a step function that starts at a survival probability of 1.0 (or 100%) at time zero and steps down at each event time. The size of each step depends on the number of events and the number at risk. The curve flattens when few events occur and drops steeply when events cluster. Median survival is read from the curve as the time at which the survival probability crosses 0.50. Confidence intervals for the survival function are typically calculated using Greenwood's formula.

Comparison of Kaplan Meier curves between groups is performed using the log rank test, which tests the null hypothesis that the survival distributions are identical. The hazard ratio, estimated from a Cox proportional hazards model, provides a measure of the relative rate (hazard) of the event in one group compared to another.

Use the Kaplan Meier method whenever your outcome is the time from a defined starting point to an event, and when some observations are censored. It suits many types of time to event outcomes, including overall survival, progression free survival, time to treatment failure, and time to disease onset. It is preferred when the goal is descriptive (estimating and visualizing the survival function) rather than explanatory (modeling the effect of covariates). For multivariable analyses, the Cox proportional hazards model extends survival analysis to include covariates.

A common mistake is ignoring the assumption of noninformative censoring. If participants who are censored (for example, because they drop out) have a different prognosis from those who remain in the study, the Kaplan Meier estimate is biased. Another error is reading the Kaplan Meier curve at a time point where the number at risk is very small, producing an unreliable estimate with wide confidence intervals. Some researchers fail to report the number at risk at regular time points below the curve, making it impossible for readers to assess the precision of the estimate. Finally, using the Kaplan Meier method in the presence of competing risks (events that prevent the event of interest from occurring, such as death from another cause in a cancer recurrence study) can overestimate the cumulative incidence of the event of interest; the competing risks framework (using cumulative incidence functions) is more appropriate in such settings.

A phase III trial randomizes 600 patients with stage III colon cancer to adjuvant chemotherapy with a new regimen versus standard of care. The primary endpoint is overall survival, and patients are followed for up to 5 years. The Kaplan Meier curves show a 3 year survival probability of 78% (95% CI: 73% to 83%) in the new regimen arm and 71% (95% CI: 66% to 76%) in the standard arm. The log rank test yields p = 0.03, and the hazard ratio from a Cox model is 0.74 (95% CI: 0.57 to 0.97), indicating a 26% lower rate (hazard) of death at any given time with the new regimen. The number at risk at 12, 24, and 36 months is displayed beneath the curve.