Document n9aawLx7K9M55xBq0Y2JaJLVa

FINAL REPORT Submitted by Sara M. Debanne, Ph.D. November 30, 1987 ftp- 3//? <k$-'s&H Data identified as describing the mortality responses to four doses (geometric progression of weights in g) of each of 16 PVC sample compositions tested using the New York Combustion Toxicity test was received from Dr. Hinderer. The data for each sample were analyzed by the Debanne-Haller method of deriving statistically preferred LD50's with their 95% fiducial limits (Debanne-Haller, 1985). Table 1 summarizes the results of this analysis by presenting LD50's, their fiducial limits, and the results of statistical tests addressing the presence of a dose-response relationship and goodness of model fit for each of the arc sine and probit transformations of mortality data. The preferred LD50 was chosen as that estimate derived from the model yielding shortest fiducial limits where a significant dose-response relationship was indicated. In situations where the chi-square goodness-of-fit statistic showed significant departure from the model, a heterogeneity factor was utilized in the calculations. There were three types of results indicated in Table 1; namely, (1) those situations resulting in solid fiducial intervals, (2) those resulting in vague fiducial intervals due to arithmetically forcing a calculation of fiducial limits even when no significant dose-response relationship was detected, and (3) those situations which did not permit even arithmetical calculation of fiducial limits. 12 substances yielded solid fiducial limits, substances B-3 and PVC-2 yielded only vague fiducial intervals and substances P-1 and-PVC-4 had no preferred model. tootozzz To address the issue of whether the 16 PVC sample compositions indicated disjoint toxicity classes, it is noted that restricting consideration to solid fiducial intervals (i.e., those of the first type) yields a set of overlapping intervals over the range 5.3383 to 22.1027 grams. If the vague fiducial intervals are also considered, the fiducial intervals overlap moreso, as the vague intervals were extremely broad. It may also be added that although the fiducial intervals of certain substances were disjoint from those of others (viz., A-2, A-9, P-2, and P-9 had distinct intervals from that of P-6), they all overlapped with the solid preferred fiducial intervals for each of the substances A-1, B-2, and B-6. In summary, using the fiducial interval approach of Debanne-Haller to determine toxicity classes leads to the conclusion that the 12 substances with solid fiducial intervals are of the same classification since the technique gives them mutually overlapping fiducial intervals. REFERENCES Debanne, S.M. and Haller, H.S. (1985) "Evaluation of statistical methodologies for estimation of median effective dose". TAP 79, 274-282. eooeozzT! TABLE 1 - SUMMARY STATISTICS PROBIT 'n LD50 LL UL 2 tX A1 14.9127 2.8030 21.9629 * NS A2 7.7562 -- -- NS NS A9 8.6372 6.1723 10.8433 * NS PI 5.6875 -- -- NS NS ?2 8.3361 -- -- NS NS ?6 14.3104 11.8347 18.2839 * NS ?9 8.6302 -- -- NS NS 31 9.0282 7.0790 11.3575 NS NS 32 10.3645 7.4059 13.0118 * NS 33 13.8880 -- -- NS NS 34 11.5887 -- -- NS NS 36 12.4005 2.3320 18.2677 * NS ^VCl 16.6730 13.2814 23.3368 * NS 'j^2 13.8874 -- -- NS NS 7C3 13.7136 11.3743 16.7396 * NS 7C4 9.9645 NS NS LD50 ARC SINE LL UL PREFERRED 2 tX 15.0142 6.4191 19.9544 * NS arc sine 7.6884 7.3801 8.0815 X NS arc sine 8.7495 6.4753 10.5744 X NS arc sine 5.6453 -- -- NS NS -- 8.5533 7.0484 9.6906 * NS arc sine 14.1656 12.2081 18.7657 NS probit 8.8512 7.6226 10.0073 * NS arc sine 9.1289 7.5182 11.0834 X NS arc sine 10.4993 7.7698 12.6898 X NS arc sine 13.8509 10.3609 34.9137 NS NS arc sine 11.2989 9.9935 13.1217 * NS arc sine 12.4857 5.3383 16.5913 * NS arc sine 16.5875 13.8733 22.1027 X NS arc sine 13.8498 10.3581 34.8446 NS NS arc sine 13.6070 11.7378 16.5215 X NS arc sine 10.0002 __ NS NS __ 22203004 ^=4 Signature Date '\/aou /o? / *7 cf sooeotzz