Hence, within one experiment, only one investigator should be assigned the tasks of recovering the S-9 fraction. Other sources of variation may be attributed to differences in the analyst, temperatures in the lab and in the spectrofluorometer, timing of thawing and preparation of reaction solutions, and reagent quality. Munkittrick et al. (1993) pointed out large differences among laboratories in reported extinction coefficients of standard resorufin solutions, GSK126 reflecting differences among batches of standard, the instruments used to measure extinction coefficient, and the procedures of each laboratory. To assess the occurrence and extent of variation, we maintain control

sheets showing the variations among assays in activity of standard S-9 fractions, prepared Protein Tyrosine Kinase inhibitor from control rainbow trout or trout exposed to ß-naphthoflavone (BNF), a model CYP1A inducer. These ‘lab standards’ were prepared by mixing the S-9 fractions from numerous control and BNF-exposed fish, dividing the mixed S-9s into small aliquots and storing them frozen at −80 °C. One

of each is analyzed with each experimental set of samples over a 6–18 month period to demonstrate that the analytical method works on each occasion, and to identify occasions when the method generated data that might be higher or lower than normal. After a new batch of S-9s has been prepared and stored, the control chart is prepared from the first five samples of positive and negative control samples analyzed. The chart consists of the 95% confidence limits about the geometric mean EROD activity of the positive and negative controls, and of the induction (positive divided by negative). Subsequent samples are plotted on the same chart, and most of the new values should fall within the 95% confidence limits, and any random or systematic change in Liothyronine Sodium expected activity can be identified quickly. As Fig. 1 demonstrates for one batch of positive and negative control S-9s tested over 16 months, that EROD activities of induced and control

fish, and induction (the ratio of induced to control activities) varied considerably among assays. Because some of this variation could be due to poor mixing of the original S-9 fractions from individual fish, we also analyzed five control and five BNF S-9 standards on one occasion. The coefficient of variations for the positive and negative controls, and for induction based on arithmetic means were 31%, 19%, and 39%, respectively, much lower than the ‘among assay’ variations of 140%, 39%, and 104%, respectively from the first five samples tested in Fig. 1. Therefore, the scatter observed in Fig. 1 was due to ‘among assay’ variance rather than ‘within assay’ variance, and reflected differences in procedures or assay conditions, even though the analyses were done by the same person. The data also suggest an increasing trend in positive and negative control results, but not in induction.