Page:NIOSH DM DFM respirator evaluation draft.pdf/64

rates ranged from about 9 to 14 users per 100 users in these three studies. That is, after necessary consideration of statistical sampling error for these three studies, the best one can conclude with 95% confidence is that the true failure rates for an APF of 10 was as high as 9 to 14 user failures per 100 users in these studies. These failure-rate results are in sharp contrast to Hyatt's requirement (and the expectations of most respirator purchasers and users) that no user failures will occur after OSHA-approved fit-test screening.^[1]
  Based on the preceding discussion and the results given in Table O for the nine studies, NIOSH concludes that: (1) (2) (3) (4) The failure rates in Table O were obtained under ideal conditions and it is highly likely that actual failure rates in typical American workplaces are sub- stantially higher. The WPFs reported in eight of the nine studies had measurement biases and most likely were substantially overestimated because: (A) A NIOSH-type deep probe was not used and failure to use this type of probe can erroneously overestimate WPFs by up to 100% and (B) Lung retention was not corrected for and failure to perform this correction can erroneously overestimate WPFs by up to 25%. (C) Filter-holder wall deposition was not corrected for and failure to perform this correction can erroneously overestimate WPFs by 300 to 500%. Because the individual WPFs reported in eight of the nine studies had mea- surement biases, both the computed point estimates for the control failure rates and the associated upper confidence limits are biased (i.e., incorrect). That is, the values reported in Table O erroneously underestimate the point estimates and confidence limits because of the inherent measurement biases in the WPF data values. In at least three studies of face-seal leakage for non-powered, air-purifying

halfmask, the actual control-failure rates could have been as high as 9 to 14