ranged from 2.7-56% of nominal concentrations (Torkelson et al. 1961; Rusch et al. 1986; Bowden 2005). Studies identifying end points other than mortality were few. No data were available to evaluate the potential for boron trifluoride to cause developmental or reproductive toxicity or carcinogenicity in animals. Boron trifluoride was not mutagenic in several stains of Salmonella typhimurium.
AEGL-1 values are based on a no-effect level for irritation. A group of 10 rats exposed for 4 h to measured concentrations of boron trifluoride at 25 mg/m3 had no abnormal findings, whereas rats exposed to the next higher concentration of 74 mg/m3 had histopathologic changes in the larynx and tracheal bifurcation indicative of irritation (Bowden 2005). The concentration of 25 mg/m3 was selected as the point of departure for calculating AEGL-1 values. The irritant effects seen at 74 mg/m3 are more severe than the threshold effects for the AEGL-1values. A total uncertainty factor of 10 was applied. An interspecies uncertainty factor of 3 was applied because irritation is a direct contact effect and is not expected to vary greatly among species. An intraspecies uncertainty factor of 3 was applied because the mechanism of irritation is not expected to vary greatly in subpopulations. The same AEGL value was applied to all AEGL durations because the point of departure is a no-effect level for mild irritation.
Relevant data for deriving AEGL-2 values were not available. Therefore, the AEGL-3 values were divided by 3 to obtain reasonable estimates of the AEGL-2 values. Dividing AEGL-3 values by 3 is supported by the steep dose-response curve (Rusch et al. 1986).
The derivation of AEGL-3 values was based on the threshold for lethality. Rusch et al. (1986) calculated a 4-h LC50 (lethal concentration 50% lethality) of 1,210 mg/m3 (exposures were to liquid aerosols of boron trifluoride dihydrate; concentrations reported are based on boron trifluoride). Using individual mortality data, a 4-h BMCL05 (benchmark concentration, 95% lower confidence limit with 5% response) was calculated by a log-probit analysis using EPA Benchmark Dose Software version 1.4.1c  (EPA 2012). The resulting 4-h BMCL05 of 554 mg/m3 was used to derive the AEGL-3 values. An interspecies uncertainty factor of 3 was applied because boron trifluoride is a corrosive irritant, and the mechanism of action is not expected to vary greatly among species. An intraspecies uncertainty factor of 3 was chosen, because the mechanism of irritation is not expected to vary greatly among subpopulations. An intraspecies uncertainty factor of 3 is also supported by the steep dose-response curve for lethality (3/10 rats died at 1,010 mg/m3, while 9/10 rats died at 1,540 mg/m3), which indicates little variability in the response within a population. The Rusch et al. (1986) study is supported by the Kasparov and Kiriĭ (1972) study that reported a 4-h LC50 of 1,180 mg/m3 in rats. Time scaling was performed using the concentration-time relationship equation Cn × t = k, where C = concentration, t = time, k is a constant, and n generally ranges from 0.8 to 3.5 (ten Berge et al. 1986). An empirical value for n could not be determined because of inadequate data, so the default value of n = 1 was used for extrapolating from shorter to longer exposure periods and a value of n = 3 was used to extrapolate from