|
LLNA Report Appendix K Page 1 of 5: NICEATM Assessment of Intra- and Inter- Laboratory Variability in the LLNA Previous Appendix 1 2 3 4 5 Next Page Return to the NICEATM-ICCVAM Document Search Page (July 11, 1998)
This assessment of the extent of intra- and inter- laboratory variability was based on the data provided in Table 2, page 12, of the LLNA Submission (Tab B). These are the only data located which are amenable to the type of analysis described in ASTM E691-92 A Standard Practice for Conducting an Interlaboratory Study to Determine the Precision of a Test Method. Two data sets were analyzed. The first one consisted of EC3 (dose calculated to induce a stimulation index of 3) data for DNCB tested twice in each of 5 laboratories. The second consisted of EC3 data for HCA tested six times in each of two laboratories. This analysis calculates h, the within laboratory consistency statistic, where h = d (the difference between each laboratory mean value and the mean for all laboratories)/the standard deviation of test averages, and k, the between laboratory consistency statistic, where k = the standard deviation for individual laboratories/ repeatability standard deviation. Once calculated, 95% confidence limits can be derived from a table provided in the ASTM Guideline. It should be appreciated that (i) the analysis is based on EC3 data, the calculation of which is not a part of the submitted protocol, and (ii) a corresponding analysis of guinea pig test data may not be feasible given the nature of the assay. 1. DNCB Data. The original data and calculations are provided in the attached table, the individual h and k values for each laboratory ate presented graphically in the accompanying figures. The 95% confidence limits for h and k were 1.74 and 2.11, respectively. None of the h and k values for the individual laboratories exceeded these confidence limits, indicating the lack of significant within and between laboratory variability. 2. HCA Data. The original data and calculations are provided in the attached table. The 95% confidence limits for h could not be calculated due to the fact that only two laboratories were involved; k was 1.52. The k values for the two laboratories did not exceed this confidence limit, indicating the lack of significant between laboratory variability. Previous Appendix 1 2 3 4 5 Next Page Return to the NICEATM-ICCVAM Document Search Page |
