Analyzing Carcinogenicity Assays without Cause of Death Information
- PDF / 710,967 Bytes
- 19 Pages / 504 x 720 pts Page_size
- 58 Downloads / 183 Views
Drug Informution Journal, Vol. 31, pp. 489-507, 1997 Printed in the USA. All rights reserved.
ANALYZING CARCINOGENICITY ASSAYS WITHOUT CAUSE OF DEATH INFORMATION DIRKHECKBOLDEBUCK AND GEORGNEUHAUS University of Hamburg, Institut fur Mathematische Stochastik, Hamburg, Germany
G ~ N T EHEIMANN R Schering AG Berlin, Corporate Biometry, Berlin, Germany
A widely used approach to analyzing carcinogenicity assays was established by Pet0 et al. (I), In addition to assessing time to death and the tumor types present at death for each animal, one needs information on the cause of death for each animal to apply this kind of analysis. The “cause of death” data appear to be very unreliable, and they tend to vary heavily between different pathologists. In other studies, they are not assessed at all, and one tends to analyze these data as if they had “time to death from tumor” instead of “time to death with tumor” information available. In this papel; a model (see Groeneboom 121) to analyze “time to death with tumor” data is presented. An overview of the two available sample tests will be given, and the results of a simulation study will be presented. Key Words: Interval censoring; Two-sample tests; Rank tests AMS 1980 subject classification: 62 E 20, 62 G 10
INTRODUCTION A WIDELY USED approach to analyzing data from carcinogenicity assays was established by Pet0 et al. (1). In addition to assessing time to death and the tumor types present at death for each animal, one needs information on the cause of death for each animal to apply this kind of analysis. The “cause of death” data appear to be very unreliable, and they tend to vary heavily between different pathologists. In other studies, the cause of death is not assessed at all. The data are analyzed using the classical type I1 censored data model (Kaplan Meier estimator log rank test), although it does not really apply. Lagakos and Louis (3) discuss these issues and give recommendations on how to proceed, using the logrank and Hoel Walburg (4)test. This paper proposes tests that may be applied instead of the Hoel Walburg test and that are more powerful. In the last few years, a new censored data model has attracted some attention in the
Presented at the DIA Workshop “Statistical Methodology in Non-Clinical & Toxicological Studies,” March 25-27, 1996, Bruges, Belgium. Reprint address: Giinter Heimann, Schering AG, Corporate Biometry, 13342 Berlin, Germany.
489 Downloaded from dij.sagepub.com at Bobst Library, New York University on May 17, 2015
490
Dirk Heck Boldebuck. Genrg Neuhuus. and Gunter Heimann
literature. This model is more appropriate for the analysis of carcinogenicity assays if “cause of death” information is not recorded, or if one wishes to neglect it. Schabe ( 5 ) presented a method to estimate the distribution function of tumor onset times. The information necessary to apply this method is “time to death” and whether a tumor (or a tumor of a specific type) is present or not. Schabe’s ( 5 ) proposal is based upon kernel density estimators. For the same
Data Loading...
