Statistical aspects of the determination of mortality rates in bioassays
article
Likelihood based confidence sets for LC50 are shown to be applicable, even in cases where point estimates lose their value due to a low number of concentrations of a test-compound showing partial mortality. This is particularly useful in the classification of chemicals by their LC50's, when concentration-response relations are steep with respect to the class sizes. Measures for experimental efficiency in choosing concentrations of compounds to be tested are derived on the basis of the minimization of the variation of the relevant parameters, such as the LC50; this may be helpful in designing toxicity experiments. The maximum likelihood estimation of toxicity parameters is generalized to include the case where living test animals are removed in a chronic toxicity test, in order to analyse their tissue for instance. These animals contribute additively to the relevant quantities in such an estimation with respect to the non-removed animals, i.e. the gradient of the ln likelihood function and the information matrix. The maximum likelihood estimation of the concentration of toxic compound from survival data is described if the toxicity parameters are known. This may be applied in monitoring the toxicity reduction of a compound that is hard to analyse chemically and undergoes degradation. Finally, the estimation of mortality parameters is described in reproducing populations of test animals, if the longevity is reduced, due to a one-compartment accumulation of a toxic compound. It is shown that the estimates for the elimination rate and the LC50 can be highly correlated, which complicates the analysis of mortality patterns in such populations. The application of the theory is illustrated by several examples.
TNO Identifier
229323
ISSN
00431354
Source
Water Research, 17(7), pp. 749-759.
Pages
749-759
Files
To receive the publication files, please send an e-mail request to TNO Repository.