response data, the point of departure has generally been defined as the no-observed-adverse-effect level (NOAEL), lowest-observed-adverse-effect level (LOAEL), or a modeled dose corresponding to an incremental effect (e.g., the lower 95% limit of the dose or concentration corresponding to a 10% increase in response [LED10 or LEC10]). Use of NOAEL and LOAEL has been criticized because of their dependence on features of the experimental design of the study from which they are derived (e.g., spacing of the dose groups) and their lack of consideration of statistical error or the shape of the dose-response curve (Crump 1984). The NOAEL and LOAEL provide only a single summary statistic and are of limited use in describing the quantitative dose-response relationship. Continuous dose-response models are thus preferred (Faustman and Bartell 1997). However, not all dose-response data sets are suitable for estimating parameters in continuous dose-response models. At least three dose groups are required for continuous dose-response modeling, whereas studies with as few as one or two dose groups can sometimes be used to identify a NOAEL or LOAEL. In its draft risk assessment for trichloroethylene, EPA (2001b) compared a variety of NOAEL, LOAEL, and LED or LEC values calculated from different doseresponse data sets and converted to human-equivalent doses by various approaches. Points of departure for inhalation dosing and for oral dosing were then selected from among the lower NOAEL, LOAEL, and LED10 or LEC10 values for each route of exposure.
The committee found that determining points of departure for noncancer end points in EPA’s draft risk assessment for trichloroethylene was generally consistent with common practice and the dose-response evidence available at the time of the assessment. However, several points should be addressed in the future. First, the criteria used to determine when toxicologic or epidemiologic data are suitable for continuous dose-response modeling should be specified. Second, the rationale for choosing a 10% response level should be provided, and presenting results for several other response levels should be considered. The ability to quantify specific response levels depends on the study design, which often differs in epidemiologic and toxicologic studies. Third, the dose-response model(s) used to estimate LEDs should also be presented. Fourth, the methods used to derive human-equivalent doses from animal data should be described. It is important that the summary statistic used for the conversion (e.g., area under the curve or peak values) be provided and be readily apparent (not placed in footnotes or separate documents). Given the variety of approaches available to derive human-equivalent doses, the results using the different approaches should be presented in tables that allow them to be easily identified and compared. This suggests that multiple dose metrics should be considered for each data set to help inform the selection of the appropriate adjustment methods.