All Times EDT
Keywords: Survival curve; Kaplan Meier estimate; mixture model; prediction; interim analysis; hazard function
When you read this title, you may wonder what aspects we need to discuss about this. Since Kaplan Meier (KM) procedure was introduced in 1958, it has been the commonly used for displaying and analysis of survival data in medical research. KM method is nonparametric and very flexible in terms of fitting any arbitrary shaped survival curves. However, limitations for KM estimates include some reduced efficiency, unstable estimation when the risk set size is small, and the lack of capability in practical applications • to adjust for covariates especially for continuous covariates, • to predict future events and determine interim analysis time for trial and resource planning, • to evaluate costs and effectiveness for a medical product, health economic models rely on data from trials to project the risk of events (eg, death) over time beyond the span of the available data from clinical trials • to explain the biologically plausibility for categorization of continuous data, which is often criticized in medical research because of the loss of power, residual confounding, and leading to considerable bias by using of the data-driven “optimal” cut-points. • to explain reasonable well for the relative risk and relative survival of interest
With these practical objectives in mind, the KM procedure lacks of applicability and additional ways of characterization of a survival cure is warranted. In this roundtable discussion, the roundtable discussion team member will share his/her company experience and view on characterization of a survival curve for different objectives. I will lead the discussion and share our experience using a fully parametric mixture model to fit the survival data, which is as flexible as KM for the observed data but avoid the aforementioned limitations of KM procedure and have the nice features beyond the trial time, such as predicting future events, survival probability, and hazard function.