Comparison of Power of Trend Tests for Toxicological Application Based on Dose–Response Profiles Enumerated Using a Nonparametric Algorithm
Tests for dose–response trend may be used to support conclusions on the toxicity of a substance in a range of doses evaluated in a comparative experiment. Criteria are proposed for evaluating and comparing alternative tests, and are applied in a comparison of certain tests applicable with continuous responses, namely, a multiplicity-adjusted pairwise comparison procedure, a trend test based on a single contrast, and four multiple-contrast trend tests. Methods considered treat dose as ordinal information. A particular focus was comparison of the simple test of a single contrast, with coefficients proportional to dose rank, with the multiple-contrast methods available with modern computing. To explore the role of the dose–response curve shape, regular dose–response shapes (or profiles), defined by the mean response for each dose level, using a simple grid-based algorithm were generated. Power for individual profiles may be used to illustrate the effect of curve shape on the power of alternative tests. To facilitate selection of tests, we report average power over profiles, averaging at different levels of granularity, first separately for each of several curve shape categories, and then over all profiles, weighting shape categories equally. Power results for the six tests evaluated at least provide clear distinctions between some procedures that are more-preferred, versus less-preferred. Our analyses expand the comparisons of tests available and contribute a methodology for choice of statistical methods for toxicological data analysis. Environ Toxicol Chem 2022;41:1754–1764. Published 2022. This article is a U.S. Government work and is in the public domain in the USA.