Two-stage Testing of Safety: A Statistical View

Abstract

Sample sizes given in regulatory guidelines are not based on statistical reasoning. However, from an ethical, scientific, and regulatory point of view, a mutagenicity experiment must have a reasonable chance of supporting the decision as to whether a result is negative or positive. Consequently, the sample size should be based on type I and type II errors, the underlying variability, and the specific size of a treatment effect. A two-stage adaptive interim analysis is presented, which permits an adaptive choice of sample size after an interim analysis of the data from the first stage. Because the sample size of the first stage is considered to be a minimum requirement, this stage can also be regarded as a pilot study.

Dose–Response and Thresholds in Mutagenicity Studies: A Statistical Testing Approach

“…Sample size has a marked influence on decisions relating to statistical tests (see, for example, Hauschke & Hothorn [23]). Given the same "true" dose-response relationship with a threshold, the estimated maximum threshold dose increases with increasing sample size.…”

Biostatistical Assessment of Mutagenicity Studies: A Stepwise Confidence Procedure

“…Some simultaneous inferences remit multiplicity adjustments by invoking the partition principle proposed by Finner and Strassburger [6]: where the parameter space is partitioned into many disjoint subsets and only one of these nonempty disjoint subsets contains the true parameter of interest, so that the FWER will be properly controlled. In literature, mutagenicity dataset has been assessed according to the proof of safety by utilizing the concept of the maximum safe dose (Hothorn and Hauschke [7], by numerous authors, among them Hauschke and Hothorn [8], Hauschke et al [9], Hothorn and Bretz [10]). As a result, this article discusses statistical aspects in terms of design and analysis using stepwise confidence set-based procedure for identification of maximum safe dose: that is, the highest experiment dose with no biological relevant increase in safety effect in comparison with negative control (Hothorn amd Hauschke [9]).…”

Introduction

“…This resulted in the award of contract number 17159-2000-11F1ED ISP DE to Ludwig Hothorn of the Bioinformatics Unit, University of Hannover, Germany, and, in turn, to the six papers which are presented together in this supplement to ATLA (17)(18)(19)(20)(21)(22). Statistical aspects of the evaluation of validation studies are discussed: for qualitative outcomes, the use of sensitivity and specificity, and for quantitative outcomes, the use of the receiver operating characteristics technique and new approaches for categorical outcomes, are recommended (17).…”

“…The direct inclusion of relevance into a test statistic, in terms of the difference between a positive and the negative control, is demonstrated for in vitro muta-genicity assays (19). The critical problem of sample size determination is proposed for proof of safety by a two-stage approach, starting with the common guideline-recommended numbers (20). Various testing procedures for monotone and non-monotone dose-response relationships, including threshold shapes, are proposed by means of contrast tests (21,22).…”