The use of benchmark dose methodology with acute inhalation lethality data.

Benchmark dose methodology has been proposed as a refinement to the no observed adverse effect level (NOAEL) methods currently used for health risk assessments. We compared log-normal probit and quantal Weibull benchmark concentration (BMC) estimates using 1, 5, and 10% response incidences with inhalation toxicity NOAELs and LOAELs from 120 acute lethality data sets. These studies yielded relatively steep dose-response slopes, which in turn influenced the suitability of selecting response incidences. The mean magnitude of difference between the 95% lower confidence limits (LCLs) for 1, 5, or 10% BMCs and corresponding NOAELs was less than twofold using the probit model and less than fourfold using the Weibull model. BMC estimates at the 10% response exceeded the observed LOAEL in some cases. Maximum likelihood estimates for doses with 1, 5, or 10% responses frequently exceeded LOAELs. The probit model repeatedly gave a better fit for the data compared with the Weibull model, resulting in improved goodness of fit tests and reduced 95% confidence intervals. The 95% LCL appears to be necessary at the 1, 5, or 10% response levels in order to safely estimate a concentration below that resulting in a LOAEL. Copyright 1999 Academic Press.