Document LgeoOK2EqEVD303g5kmGp0NbQ

Federal Register / .Vol- 51, No. 119 / Friday, June 20, 1986 / Rules and Regulations 22687 sampling strategies be designed to yield ' and on the horizontal axis.we havethe.mean 'TOO fields" stopping rule of the set A samples with fiber densities of at least 80 fibers per 100 fields when using the NIOSH 7400 method. Intralaboratory Variability number of fibers counted in that sample:....... Since coming to the United States I have plotted the results on the same axis which were published by (Busch ct al.| .... If we superimpose those; you can see they lie in very much the same kind of area. (Tr. 6/20, rules (i.e.. for samples with a fiber density of less than 1 fiber per field, 100 fields must be counted, and for samples wjth a fiber.density of more than 5 fibers per field, 20 fields must be counted). For NIOSH's statistical analysis of the pp. 6-7) samples with fiber densities between 1 Johns-Manville analytical data, and the Dr. Ogden's estimate of 0.2 ns an and 5 fibers per field, the NIOSH "A" resulting estimates of the precision of the P&CAM 239 method, were criticized by rulemaking participants, who believed NIOSH's estimates to be loo low (Ex. 01-16, Tab D; 91-10, Tab E, 232B, 233-B, 328; Tr. 7/6, p. 66J. In summarizing the record evidence on intralaboratory variability, which was . average CV representing inlralaboratory .variability is a consequence of his use of a minimum fiber density of SO fibers per 100 fields, in contrast with NIOSH's recommendation in the 7400 method that a minimum fiber density of 80 fibers per 100 field be used, which yields an estimated CV of 0.15. Therefore, OSHA rules require that enough fields be counted to yield 100 fibers. According to Busch et al. (Ex. 84-62, Appendix C. p. 75). calculating the overall CV for samples counted using the "100 fibers" stopping rule rather than the "100 fields" stopping rule cannot be done unless additional statistical techniques are one of the three sources of variability included in NIOSH's overall estimate of precision for the method, the A1A finds that Dr. Ogden's results actually confirm NIOSH's estimates of the inlralaboratory coefficient of variation developed, although "indications are . that the TpO fibers' stopping rule (wouldj yields a CVT similar to that for concluded that ". . . NIOSH for asbestos samples analyzed by phase the TOO fields' stopping rule when 100 should . . . recognize a more reasonable CV value in the range of 0.2 contrast microscopy. The AIA also relied on the analysis fibers are counted" (Ex. 84-62. p. 75). Both because of the uncertainty to 0:3--a range which accords with the performed by Dr.Crockett (Ex. 312-A, surrounding the calculation of a CV for remainder of the evidence in the record Tab PJ of the Johns-Manville data to samples counted using the 100 fibers on the best achievable total refote NIOSH's estimate of the precision stopping rule, as reported by Busch (Ex. intralaboratory error" (Ex. 128. p. A-15). of the P&CAM 239 method (Ex. 328. pp. 84-62. p. 75). and because the Johns- In arriving at this conclusion, the AIA A-12 to A-15). The AIA explains Dr. Manville data were appropriately : relied on the following evidence: Crockett's analysis as follows: excluded from the NIOSH analysis. Testimony from Dr. Ogden staling the! hu had achieved intralnboralory CV's of approximately 0.2 in British Inboratories. Dr. Crockett identified and plotted . . mure than forty data points that were excluded from the NIOSH analysis. ... In OSHA disagrees with Dr. Crockett's contention that NIOSH's estimated CV is unreliable. Analysis of the Johns-Manville data by the very important range of 60 to 100 fibers The final study cited by the ALA to Dr. Patrick Crockett, who projected a CV of counted, only six or seven of the eighteen support its contention that the oyer 0.31 for a total fiber count of 100. data points [in this range) were included and inlralaboratory CV associated with the The study of the P&CAM 239 method by Chase and Rhodes (Ex. 86-002). who reported an inlralaboratory CV of 0.38 for a total Tiber count of 100. - . OSHA has analyzed the evidence presented by the ALA and finds that the included points represented dramatically lower CV estimates than the excluded * points . . When Dr. Crockett applied a close reproduction of the NIOSH statistical method to the entire Johns-Manville data set. his projected CV for a total-fiber count of 100 P&CAM .239 method is higher than that estimated by NIOSH is the round robin study by Chase and Rhodes (Ex. 86-002), which reported a random inlralaboratory CV of 0.36 for a total fiber count of 100. According to the AIA: these date do not necessarily refute NIOSH's estimates of the CV for samples analyzed by the P&CAM 239 or NIOSH'7400 methods; In fact, the evidence of Dr. Ogden cited by the AIA . closely parallels the results obtained by. Busch et al.(Ex. 84-62, Appendix C). w'as over 0.31, about three times as high as NIOSH's result based on incomplete data. In any event, the published NIOSH method does hot represent the original Johns-manville data base, but instead reflects only a subset of that database with much of the high variability data deleted. (Ex. 128. pp. A-13 to A-1.4) This study included virtually all relevant error sources and of the studies in the record, is the most representative of everyday . American experience with P&CAM 239 and commonly encountered workplace samples, and should be accorded significant weight by OSHA. (Ex. 328, p. A-li) Dr.. Ogden examined .intralaboratory variation among technicians analyzing ' 66 asbestos-samples in British Health and Safety Executive'(HSE) laboratories (Ex..B4-447). In this testimony about this work. Dr. Ogden stated, that his. investigation, as,well as. those from . . other laboratories, resulted la estimated intralaboratory CV's similar to the estimates obtained in the NIOSH study: When questioned as to why NIOSH eliminated some of the data points in the analysis, Ken Busch of NIOSH replied that these data were excluded because they were outside the fiber density range permitted by the "A" rules: I'm absolutely certain that thcre,was.no' intent to eliminate counts which would be the cause of high variability. The elimination of counts . . . which were based on large ALA did acknowledge that the study may overstate the magnitude of the "best achievable inlralaboratory CV" because of the absence of quality control programs in some of theparticipatingTaboratories, and because. some of the samples analyzed were difficult to count (Ex. 320. p. A-ll, Footnote 17). The lack of quality control programs, as evidenced by the . There is a lot of evidence from different numbers of fibers was simply because this participation in the NIOSH PAT laboratories' that repeated evaluation of the . procedure of counting more than 100 fibers program of only 19 of the 46 laboratories same asbestos-loaded membrane Tiller by the same counter, or by different counters closely linked within a laboratory, can give a coefficient of variation of between 10 and 20 percent. .... Figure l'|from Ex. 84-447| demonstrates' results in our laboratory of a detailed study of one year's quality.assurance results. did not correspond to the standard procedure . for the NIOSH method. (Tr. 6/21,p. 192) In addition, the statistical model developed by Busch et al. (Ex. 84-62. Appendix C) was developed to estimate (he relationship between CV and number of fibers counted. As such, the included in the study, was one reason suggested by OSHA in the November proposal-(48 FR 51136).for the Chase and Rhodes study's, higher reported CV. OSHA also believes that the design of the Chase and Rhodes study is deficient. The authors collected and analyzed a On the vertical axis we haVe.the coefficient model can only appropriately be applied total of 1.774 full-shirt samples, of which of variation of the repealed determination. to samples that were counted using the 541 were submitted for a second GLEASON-000935