The analyses conducted by Ten Berge et al. (1986) examined the relationship between concentration and exposure time for predicting mortality response. OEHHA used the chemical-specific regression coefficients (n) reported by Ten Berge et al. (1986), when available, to adjust concentrations administered to humans or animals in the study selected for a duration different than one-hour. However, because these regression coefficients were derived based on mortality data, they are not applicable to other endpoints. . . .

OEHHA has apparently chosen to use a default approach, rather than consider mechanistic data or in the absence of such data, a conservative approach designed to mitigate potential overestimation of exposure concentrations. If so, that should be stated openly rather than attempting an ad hoc scientific justification based on Ten Berge et al. (1986).

Response: The comment expresses disagreement with the use of a default procedure for estimating 1-hour concentrations that cause adverse effects. It is correct that the Ten Berge et al. (1986) paper used mortality data to derive exponents of "n" for the equation Cⁿ x T = constant. OEHHA uses the principle of the Ten Berge et al. paper without necessarily using specific values contained in the paper. OEHHA agrees with the comment that mortality data do not necessarily reflect the dose-duration response relationships for other endpoints. When specific data on the endpoints of concern existed to derive the exponential term for this relationship, we have used the data and derived "n" (see chlorine, ammonia, and phosgene as examples). The generic example given in the comment for using "mechanistic data" to derive the exponential term when response data do not exist describes differences between systemically and locally acting irritants and the usefulness of considering different mechanisms when deriving "n". While the example given contains no specific data or citations, it does point out the importance of considering biological mechanisms. The comment advocates the exchange of default values for the assumption that mechanistic information is an adequate substitute for actual exposure data. This comment overestimates the quantity and utility of "mechanistic" data for such acute exposures. Furthermore, no specific recommendations are made in the comment on how such information can be used consistently and quantitatively in estimating the exponential term.

OEHHA has derived acute exposure levels by methods that we believe are scientifically sound and based on existing data. Because data gaps exist, health-protective parameters are used. If data should become available that suggest changes to this methodology or to specific values, we will consider such data. Mechanistic data may be of use in estimating exponential terms for the modified Haber’s Law. However, use of such data can only produce estimates, which should be validated by actual exposure data. Until methodology exists for the incorporation of such data, we believe that the use of the health-protective values are justified.

Comment: There is no factual basis for the statement that, ‘For acute toxicity data, the log-probit model usually provides a good fit and is generally used.’

Response: The log-normal model is among the most widespread models used for toxicity testing and has traditionally been used extensively for determination of acute lethality and other dichotomous responses (Finney et al., 1971; Rees and Hattis, 1994). Furthermore, the log-normal distribution aspect of the model is biologically plausible and accounts for some degree of inter-individual variability (Rees and Hattis, 1994).

Comment: OEHHA’s decision to define the BMD as a probability of a 1% response will always result in more stringent standards than the use of the NOAEL approach, resulting in an overly conservative approach. In a study by Allen et al. (1994) the experimentally derived NOAEL was compared to the BMD. Based on the work of Allen et al. (1994), of BMDs defined as a 10%, 5%, or 1% probability of a response based on quantal data, the 10% BMD level on average was closest to the NOAEL derived from the experimental data. A BMD level for quantal data based on a 1% probability of a response was on an average of 29 times lower than the NOAEL, while at the 10% probability level, the average BMD level was 3 times lower than the NOAEL. Therefore, it appears that the proposed guidance using the BMD approach will in general result in lower levels than the NOAEL approach and will be highly conservative. An estimate of exposure producing a 1% probability of response may be model dependent; however if a higher response level, such as 10%, was used, the BMD would be less model dependent. Therefore, if the BMD approach is to be used for the development of acute exposure levels, OEHHA’s BMD should be defined as a probability of a 10% response, rather than a 1% response.

Response: Although our analysis shows that there is not much difference between the BC₀₁ and the BC₀₅, in order to be consistent with USEPA, OEHHA has changed the proposed benchmark from 1% to 5%.

It is not the goal of the benchmark dose approach to emulate results from the NOAEL approach, and the continued reliance on such comparisons is inherently flawed. Furthermore, the data upon which the comment relies is an analysis of developmental data sets by Allen et al. (1994). In their analysis, Allen et al. show that BC₀₁ is 29-fold below the level of the NOAEL, on average. The BC₀₅ values were also below the NOAEL, on average, by about 6-fold. However, the analysis is extremely limited due to the examination of only developmental toxicity studies. In our analyses, the BC₀₁ is less than 2-fold below NOAELs taken from acute studies, and only about 1.3-fold below the BC₀₅, on average, for acute responses other than developmental defects. The dose-response slope for developmental toxicity studies may not be representative for all other endpoints. It is therefore premature to conclude that the BC₀₁ values are overly conservative based on the analysis by Allen et al. (1994).

Comment: OEHHA recommends that when using the BMD approach uncertainty factors of 10 be applied to the BMD for uncertainty involved in animal to human extrapolation and human intraspecies variability. However, because the BMD approach takes into account intraindividual variability, which is what determines the shape of the dose response, OEHHA recommends adjustments resulting in factors of less than 10. This recommendation has merit, but it should not be a rigid default recommendation, like the modifying factor of 0.3 recommended by OEHHA. The biological aspects of the response should regulate the magnitude of the uncertainty factors and/or modifying factors used, rather than setting a default value.

Response: OEHHA agrees that if "biological aspects of the response" can be specifically quantified, they should be incorporated into uncertainty factors and the existing value should be replaced. However, this needs to be determined according to the individual chemical. The comment presents no data or examples that pertain to this process.

Instead of a modifying factor of 0.3, the REL development methodology now specifies an uncertainty factor of 3 to extrapolate from a LOAEL to a NOAEL for mild sensory irritation.

Comment: The preferred approach for the derivation of acute toxicity levels based on studies of longer duration is a method similar to the method reported by Guth and associates (1991) in Appendix D of OEHHA’s Technical Document.

Response: The methodology proposed by Guth and associates in 1991 and suggested in the comment has potential applications that may make it the preferred one under some circumstances. It allows the information from a large number of smaller studies reporting NOAEL or LOAEL data to be combined, therefore strengthening the conclusions reached. However, because it has not yet been approved for use by USEPA and because it is useful only for chemicals on which there are a large number of studies, it is not a recommended approach for the development of reference exposure levels.

References
Finney DJ. Probit analysis. Cambridge, England: Cambridge University Press; 1971.

Rees DC, Hattis D. Developing quantitative strategies for animal to human extrapolation. In: Hayes AW, editor. Principles and methods of toxicology. 3rd ed. New York: Raven Press; 1994.

Chemical-Specific Comments
Up | Down

Comment: For the Level I concentration, OEHHA used a regression coefficient, n, of 4.6 reported in Ten Berge et al. (1986). However, this regression coefficient (n) is based on a dose-response relationship for mortality. The use of any value based on mortality data is not usable for extrapolation to other endpoints, especially for nonsystemic endpoints, such as irritation.