“…While substitution of an analytical LOQ or a fraction thereof has been common practice, random effects, nonparametric, and censoring regression models have also been used. (29,57,58) When most or all C i samples in a study are below the quantification limit, as is often the case in studies of high-performing respirators, (17,20,23,24) meaningful WPF distributions cannot be calculated. In these cases, reporting WPFs as values “greater than” a minimum value calculated using the quantification limit as the C i mass (for example) may be most appropriate: This process avoids bias and error introduced by assumptions regarding the C i distribution or the expected level of performance (APF) for the device.…”