TESTS FOR BIOEQUIVALENCE OF CONTROL MEDIA AND TEST MEDIA IN STUDIES OF TOXICITY

Statistical tests of the classical (null) hypothesis - that there is n o difference in effects of control media and test media - are commonly used to make statistical inferences toward the no-observed-adverse-ef fect concentration. However, failing to reject this hypothesis is not considered as scientific proof the hypothesis is true. An effect may e xist, but high variation due to inadequate replication, variation in e xperimental units, or imprecise measurement techniques may yield data for which the hypothesis is not rejected. An experiment may also be to o precise, yielding effects that are statistically significant but not biologically important. We propose the use of tests of bioequivalence of control media and test media to alleviate these unsatisfactory cha racteristics of tests of the classical hypotheses for regulatory decis ions. We review and illustrate the test for bioequivalence using acute and chronic toxicity data. We also define a procedure for determining the level of effect at which there will be high power to refute the h ypothesis that there is a lack of bioequivalence if in fact the biolog ical response in the control media is identical to the response in the test media.