Weibull Failure-Time Mixture Models for Evaluating Efficacy in the Presence of a Biomarker
- PDF / 8,203,470 Bytes
- 9 Pages / 612 x 792 pts (letter) Page_size
- 72 Downloads / 154 Views
Weibull Failure-Time Mixture Models for Evaluating Efficacy in the Presence of a Biomarker markY is a m a t h that +ds the differentialflcacy of a particulmthempy based on markY status. Standard statistical methods compare tmtments, not validate prediction. We pmpase a faih-time mixture mod$ that adueves both abjectives. We also explain how to evaluate f i q in bionrmlEa submatim. We use the mawimum l i W i h d method to csti-
A +dive
Kallappa M. Koti, PLD US Food and DNg Administrntion. Silver Spring, Maryland
Kry Words Biomarker development; Survival data; Gordonk m o d 4 Extreme value distribution; Subroutine NLPTR; Likelihood rntio tat; Wald t a t Corrrspo~dricrAddrrrr Kallappa M. Koti. m c e of Biostatistics, Food and Drug Administration. 10903 New Hampshire Avenue, Silver Spring, MD 20993 (email: [email protected]).
INTRODUCTlO N A biological marker (biomarker) is a characteris-
tic that is objectively measured and evaluated as an indicator of normal biologic processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention (1).To mention a few, CD4 count, prostate-specific antigen, arrhythmias, and blood glucose are biomarkers having some degree of clinical utility (2). Biomarkers are the foundation of evidence-based medicine, and it is imperative that biomarker development be accelerated along with therapeutics (3). The number of marker-based clinical trials is on the rise. Investigatorsare trying to find out if developing markers is useful to guide therapy or predict disease progression. However, some in the scientific community are wary of marker-based clinical trials. There is a counterargument. Drugs have multiple mechanisms of action; the biologic markers typically focus on just one of them and, therefore, biologic markers are imperfect measures in predicting drug benefit or harm (4). See also Wang (S), and Simon and Wang (6),for an in-depth discussionon biomarkers as classifiers in pharmacogenomics clinical trials. Consequently, it is necessary to validate the biomarker-originated benefit of an intervention using valid data analyses. In this article, we propose a data analysis method that compares treatments and validates prediction. We consider a clinical trial that is designed to compare two treatments in the presence of a
mate the mixture model. We explain the computational as@ of the model and discuss the underlying statistical infmce. We discuss sample size determination. We illustmte the methoddogywitha cmputesgendeddata set. Theproposed mixtun modd is simple and capable of d n g investigators seeking to design m a h ~ based dinical trials in their analyses.
single marker. Biomarkers can take on continuous or binary values. We consider the latter case. We name the biomarker under investigation as B. In this note we presume that the presence of biomarker B is indeed a possible major guiding factor in determining efficacy of an experimental therapy. In what follows, we use B+ to mean that a subject has the marker B, and B- to mean that he or she does not have
Data Loading...
