engleski

Optimization of the number of experimental animals using new computing techniques

Modern research involving animal testing are under constant pressure to reduce the number of animals that will be used. In addition to the usual reasons such as cost reduction, ethical reasons are becoming more prominent. Therefore, the requirements for reducing the number of animals in the experiments are growing. Besides of importance of minimizing the number of animals used, there is also a need for obtaining accurate and reliable results. Therefore, one of the crucial tasks of experimental design which includes use of animals is optimization of the number of organisms that will be used in the experiment. Classical statistical methods can fairly accurate determine the sample size required for appropriate decisions making regarding acceptance or rejection of the hypothesis. However, due to the nature of endpoints that are measured during the experiment, this number is often relatively high. Modern statistical techniques combined with modern computers and calculation techniques, are enabling a substantial reduction of animals in experiments. Although in such approaches additional requirements are set for researchers in terms of the use of existing biological knowledge and understanding of statistical methods, they allow much better assay performance, in economical and ethical sense. Resampling methods in combination with the use of Bayesian statistics and parallel computing form the basis of optimal design of experiments using animals. In this paper application of various statistical methods of designing and determining sample sizes for individual endpoints measured in laboratory mice, rats, carps and earthworms are used. The results showed that the use of these techniques for experimental designs the number of animals can be significantly reduced with the same parameters of statistical inference.

sample size; design of experiments; resampling; statistical techniques; Bayes

nije evidentirano