A variety of assumptions are required to conduct screening level risk assessments and assign risk scores. These assumptions can be broadly grouped into those that address data gaps and those that define the environmental fate, exposure and risk models used to calculate Toxic Equivalency Potentials (TEPs). In both situations, assumptions are made by balancing the need to assign risk scores to the largest possible number of chemicals being released into the environment with the need to develop a scientifically-defensible ranking system.
As a result of data gaps or modeling problems, not all chemicals with reported environmental releases possess the information required to weight their mass release by toxicity and exposure potential. It is also not possible to derive risk scores for several important categories of chemicals (e.g., dioxin compounds), because these categories include specific chemicals with very different toxicity or physical-chemical characteristics. While these chemicals and categories do not appear in ranked lists of cancer and noncancer risks, they should not be assumed to be safe.
ASSUMPTIONS MADE TO ADDRESS DATA GAPS
The method used to develop TEPs requires three types of environmental data for specific chemicals: physical-chemical and environmental process data for fate and exposure modeling, and risk assessment value data for risk estimation. Since chemicals that are not assigned TEPs are not included in the ranking system (and may therefore be interpreted in the policy process as less hazardous or lower priorities), it is imperative to make use of available data as much as possible.
A variety of references were consulted to obtain required chemical parameters, using a hierarchy of sources based on data quality and electronic accessibility. However, the requirement of scientific credibility precluded filling outstanding data gaps on specific chemicals with generic default values. Use of generic default values for missing data could introduce substantial uncertainties and potentially undermine rankings based on chemical-specific data.
Major assumptions made to address specific data gaps are detailed below:
PHYSICAL-CHEMICAL DATA GAPS
Some physical-chemical parameters can be obtained for chemicals lacking data using predictive methods or default assumptions. Diffusion coefficients for a number of chemicals were estimated using structure-activity relationship methods. If no published half-lives were available for a chemical from reliable sources, the CalTOX default of zero degradation was utilized (effectively assigning such chemicals an infinite half-life). TEPs were developed for a number of inorganic metals by setting their vapor pressure to zero.
LANDSCAPE PARAMETER DATA GAPS
The risk scoring system utilizes CalTOX to generate TEPs that are applied to national TRI releases, using an area-weighted average U.S. landscape data set. Uncertainty and sensitivity analyses have shown that CalTOX calculations are much less sensitive to landscape properties than to chemical properties.
RISK ASSESSMENT VALUE DATA GAPS
For most chemicals, risk assessment values are only available for one of the three exposure routes CalTOX considers (ingestion, inhalation, and dermal contact). As a general rule, any available risk assessment value is applied to other routes of exposure lacking toxicity data, unless there is a clear toxicological rationale against making this assumption. This assumption of cross-route applicability is made to avoid treating chemicals as if they pose no health risk at all if exposures occur via a route that lacks data.
ASSUMPTIONS MADE IN THE DESIGN OF RISK ASSESSMENT MODELS
A series of assumptions are required in the design of the environmental fate, total exposure and risk assessment models utilized in CalTOX. Value judgments implicit in model design can have significant impacts on the results of TEP estimates.
MODELING ENVIRONMENTAL FATE
The CalTOX multimedia fate and exposure model is well suited to assess the fate of non-polar organic chemicals, but special care is required to assess the fate of some inorganic substances (such as metals), some highly reactive substances (such as acids), and volatile metals (such as mercury). Predicting the environmental fate and exposure potential associated with release of these substances to different media compartments requires consideration of special environmental processes (such as the neutralization of acids). TEPs were not generated for chemicals which require substantial modifications to the model in order to accurately capture special environmental processes that determine exposure potential.
The CalTOX model was run assuming that the region being modeled (U.S.) constitutes a closed control volume (meaning that chemicals released here are not transferred out of the country via transport processes like wind or runoff). This model choice has the effect of increasing the importance of a chemical's environmental persistence in determining its risk score.
CalTOX was also run to calculate steady-state concentrations of chemicals in each compartment, and not to evaluate changes in compartment concentrations over time. This effectively assumes that there is no significant time lapse between release and exposure, and that exposures occur to the ultimate steady-state concentration. This model choice provides a good indicator of long-term, average population exposures, but may underestimate the exposure of populations near the release site.
MODELING TOTAL HUMAN EXPOSURE
Human exposures to chemical concentrations resulting from an environmental release were modeled using factors that are representative of an average member of the population, and do not address the risks posed by atypical exposure patterns in some populations (such as subsistence fishers).
MODELING HEALTH RISKS
CalTOX incorporates conventional approaches to characterizing cancer and noncancer risks, which are based on a number of important assumptions. Some of these assumptions (results in animal tests are predictive of human toxicity, cancer risks are linear at low doses) are health-protective and may result in overestimating risks, while others (there are thresholds for noncancer effects, humans do not vary greatly in susceptibility to toxicants) may result in underestimating risks.
NORMALIZING TO A REFERENCE COMPOUND
This risk scoring method does not take into account qualitative differences in the types of cancer or noncancer health effects that chemicals may cause, or in the weight of evidence supporting the identification of a chemical as a specific type of hazard. For example, benzene is a known human carcinogen that causes leukemia. Expressing the potential health risk of other carcinogens in benzene-equivalents does not indicate they are known human carcinogens, or that they will cause leukemia; it indicates only the relative magnitude of added cancer risks associated with a one pound release.