Tid: 7 maj 2012 kl 15.15-16.00.Seminarierummet 3721, Institutionen för Matematik, KTH, Lindstedts väg 25, plan 7. Karta!
Föredragshållare: Matteo Bottai, Professor and Head, Unit of Biostatistics, Institute of Environmental Medicine, Karolinska Institutet
Titel: Analysis of percentiles of survival time with censored quantile regression
Abstract Suppose a researcher in a hospital wishes to study the effectiveness of a new treatment for a terminal disease and plans to follow up a group of patients on that treatment. The major research interest lies in the number of days that the patients survive. Some patients may survive over the entire study period or be lost to follow up (e.g. a patient survived at least one month before he moved away and we lost contact). These provide only partial information and are often called "censored" observations.
Survival time is often described by Kaplan-Meier curves. These can estimate percentiles of survival time like median survival, the time by which half of the patients would die. Other percentiles can be estimated as well like, for example, the 90th percentile of survival, the time by which 90% of the patients would die. Percentiles of survival may also be used to compare survival in different groups of patients (e.g. patients on treatment versus patients on placebo).
The seminar gives an introduction to censored quantile regression with focus on Laplace regression, a general method for estimating percentiles of survival (e.g. median survival, 90th percentile of survival) that extends the Kaplan-Meier method to multivariable analyses in which multiple risk factors can be evaluated jointly and potential confounding adjusted for. The talk is meant for general audience with minimal familiarity with regression methods. Practical examples and results from analyses of simulated data will describe advantages and limitations of Laplace regression over other traditional approaches such as Cox proportional-hazard regression. Some mathematical details are given for completeness but are not essential for understanding the rest of the presentation.
|Sidansvarig: Filip Lindskog