Guidelines for the statistical evaluation of SCE.

Abstract

When planning studies by the sister chromatid exchange (SCE) test, it is necessary to calculate the size of the test groups, taking into consideration the variance of the test result and the statistical distribution of SCE frequencies. This paper deals with these problems. Recommendations are given for the preparation of slides under standardized conditions, for the subsequent counting of SCEs/cell in a random sample of 30 cells from each slide, and for the condensation of the information contained in the sample of 30 counts into a single statistic, that may be treated as a normally distributed variable. The adequacy of this transformation is shown for the data from 170 different subjects. Of these, 165 (58 nonsmokers and 107 cigarette smokers) had mean values of SCE/cell ranging from 6.5 to 13.5, while the remaining 5 subjects were on intermittent treatment with cytostatics every fourth wk, and exhibited a mean value of SCE/cell in the range 15-23. The variance associated with the recommended statistic has been decomposed into 4 variance components: variance within slides, variance between slides prepared from the same blood sample, variance within subjects, and variance between subjects. Based on a total of 680 SCE analyses in 218 persons, estimates of these variance components are given and used to calculate the necessary number of samples for the detection of a prescribed difference in SCEs/cell for selected values of Type I and Type II errors.