Statistical issues with respect to workplace protection factors for respirators

A workplace protection factor (WPF) for a respirator wearer is the measured concentration of a contaminant outside the respirator divided by the simultaneous concentration of that contaminant inside the respirator. The Occupational Safety and Health Administration (OSHA) proposed an assigned protection factor (APF) of 10 to negative-pressure, half-facepiece, air-purifying respirators (HFAPR), based on the criterion that the 5th percentile of WPF for HFAPR be larger than the APF. This class of half-facepiece respirators includes both filtering facepiece and elastomeric half-mask respirators. Nicas and Neuhaus developed a statistical model for log-normally distributed WPF that separated between-wearer and within-wearer variation. Using results from applying this model to seven studies of HFAPR, they proposed an APF of 5 for this class of respirator, based on the criterion that the 5th percentile of the 5th percentile of individual worker WPF distributions be larger than the APF. In this article, two reasons are suggested for these differing proposals: (1) the Nicas and Neuhaus criterion is inherently more conservative than that of OSHA, and (2) substantially different databases were used to evaluate the two criteria. The Nicas and Neuhaus model is expanded to allow for differences in WPF distributions for different types of HFAPR and different contaminants and for separately estimating within-wearer variation and variation due to measurement error. Appropriate statistical methods are illustrated for implementing these models when some of inside-the-respirator measurements are nondetects. Results from applying this expanded model to two new WPF studies suggest that an APF of 10 would be sufficiently protective in these studies using the OSHA criterion but more marginally so using the Nicas and Neuhaus criterion.