A Practical Approach, Second Edition=Ronald D. Ho.pdf

354 DEVELOPMENTAL REPRODUCTIVE TOXICOLOGY: A PRACTICAL APPROACH, SECOND EDITIONdo not respond completely independently. 41,56 The propensity for fetuses of a given litter to exhibitsimilar responses to toxic insult, thereby artificially inflating the apparent group response, has beendesignated the “litter effect.” Mathematically, using the litter as the experimental unit to accountfor this influence requires a determination of the percentage of embryos or fetuses within eachlitter that are affected. A grand mean is then calculated from the individual litter means. Using thisapproach, the variance among litters (standard deviation and standard error) can also be calculated.While the litter proportion calculation is generally applied to embryo and fetal survival data,many investigators fail to use this approach to analyze fetal malformation data. These investigatorssimply determine a percentage of litters with at least one malformed fetus, failing to apply correctlitter-based statistics. With this approach, the number of malformed fetuses within each litter is nottaken into account (e.g., a litter with 1 of 12 fetuses malformed is given the same weighting as alitter with 6 of 12 fetuses malformed). Therefore, variance among litters cannot be determined.Initially determining the percentage of malformed fetuses within a litter, followed by calculatinga litter grand mean, is the most appropriate way to analyze fetal malformation data.Table 9.13 presents five comparative examples of calculations using the litter as the experimentalunit. These examples include the following endpoints: numbers of resorptions (prenatal death),malformations, and fetal weight. For each example, an explanation of the derivation of the pertinentendpoint is provided. These examples contrast the incorrect and correct litter-based statistics andclearly demonstrate the different means of computation. Surprisingly, even though considerableresearch has been conducted and literature published on this topic in the 1960s and 1970s, fewlaboratories, and even fewer commercial software programs, properly calculate these values.Depending on a particular data set, the correct and incorrect means may differ little, but variance(sum-of-mean-square-derived) may not be obtained at all using the improper calculations. Whenthe incorrect statistic is used, the N (number of fetuses) is exaggerated, increasing false positiveresults, and in the case of fetal weight, “within-litter variance components” are not determined orincorporated. The latter omission often fails to identify the within-litter response dimension (i.e.,increase in number of affected fetuses per litter), which becomes lost in the overall among-littervalue. Because of the incorrect statistics, litters with large or small numbers of fetuses will bedisproportionately weighted in the analysis. In addition, the analysis will not be able to assessclustering of effect in limited numbers of litters. The precise calculations, which obey the litterunit, were first applied in the late 1970s at the NCTR. 56,57 Use of the appropriate method ofcalculation can make significant differences in interpretations. This becomes of paramount importanceto postnatal data evaluations because litter influences persist well beyond weaning. 37IV. DOSE RANGE–CHARACTERIZATION STUDIESVS. SCREENING STUDIESDose range–characterization studies provide information to design a definitive study (one used forhazard identification in risk assessment). Under most circumstances, dose range–characterizationstudies do not eliminate a test article from development, although unexpected results can lead tothat decision. Conversely, toxicologic screening studies may be used expressly for selecting productcandidates. The following sections discuss the differences between dose range–characterization andscreening studies, and where the two types of studies may converge.A. Dose Range–Characterization StudiesDose range–characterization studies (also referred to as dose range–finding, preliminary, pilot, ordose-finding studies) are an essential component of a valid research program. These studies provideinvestigators with information necessary to properly select dose levels for definitive developmentaland reproductive toxicity studies. They represent a different level of interpretation and are pivotal© 2006 by Taylor & Francis Group, LLC