Abstract:
|
Two important classes of biomarkers in oncology are prognostic and predictive biomarkers. Sample size planning is an essential step in designing a biomarker study. Many statistical methods for sample size estimation have been developed for therapeutic trials and can be readily adopted when designing biomarker studies. These methods rely on assumptions about design parameters and are suitable for prospective studies before any data are collected. However, biomarker studies are frequently conducted using specimens collected from previous treatment trials. In these situations, power estimation is more relevant as the sample size available is restricted by the size of the treatment trial. Furthermore, at the time of the biomarker study, the number of events would have been observed. In this talk, I will present simple methods for estimating power in prognostic and predictive biomarker studies which utilize the information that is available in the treatment trial. I will show that power is robust to the assumed hazard rates so long as the event numbers are consistent with those observed. Real-life examples in cancer research will be drawn to illustrate the use of the proposed methods.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.