Statistical design, analysis, and inference issues in studies using sister chromatid exchange.

While underlying biological mechanisms responsible for sister chromatid exchange (SCE) formation are not fully understood, scientists worldwide are increasingly using SCEs in the evaluation of excess risk from exposure to chemical and biological agents. SCEs are being used as endpoint measures of cell damage in many types of experimental and nonexperimental investigations. The former includes both simple and complex randomized experiments using both animals exposed in vivo and cells exposed in vitro as experimental units. The latter includes the important, yet potentially misleading, human case-control studies in which a group of humans who are or have been exposed to some agent are compared with a selected nonexposed group on SCE frequency. As more is learned about those factors which result in SCE variations, the assay techniques and study protocols can be adjusted to enhance study sensitivity and to minimize potential bias. Although research concerning sample sizes and statistical analysis methods has been conducted, more is needed. Search for other sources of intersubject variations in SCEs should continue so that such sources can be controlled in future studies, particularly the human exposure types. A number of experimental designs are presented and contrasted with their nonexperimental counterparts. Statistical methods are summarized and sample size options are given for the human and animal exposure studies and for the studies in which cells are exposed in vitro.