Statistical aspects of the design of biological tests for the detection of low genotoxic activity

When an environmental factor is tested for mutagenicity or ability to induce related stochastic effects, a negative result of the test is unable, without additional information, to demonstrate non‐mutagenicity, etc. The confidence interval of the accepted no‐effect might include a response that corresponds to a non‐acceptable risk in exposed populations. Tests in microorganisms for genetic toxicity due to non‐thermal effects of microwaves and radio‐waves, mostly giving negative results, are discussed from these points of view. As a basis for experimental studies (A nderstam et al. 1983, Hereditas 98 : 11–32) methods are indicated to estimate the size of a test that would render it able to detect or exclude, with sufficient statistical power, non‐acceptable risks due to occupational exposure. This calculation has to take into consideration the ratio of absorbed doses in exposed populations and in experiments, and a translation factor from response in the test system to risk in man. The latter factor was estimated by referring the response in test systems and the risk for cancer or heritable damage in man, to dose‐equivalents of γ‐radiation. Statistical methods for treatment of experimental data, especially for the purpose of minimizing errors, are discussed.