Historically, toxicity testing and evaluation schemes have always implicitly included thresholds to prioritize concern or delineate the toxicity testing needed to reach a safety decision. In most cases, such thresholds for testing have been qualitatively based on a general knowledge of a chemical or class of chemicals. Moreover, decisions regarding testing recommendations typically weigh the two main factors of the safety assessment paradigm; the likely consumer exposure to a particular chemical or class of chemicals and the likely inherent toxicity of those chemicals. A larger risk, due either to an expectation of high consumer exposure or greater concern for potential toxicity will ordinarily correspond with a need for greater levels of toxicity testing data to demonstrate the safety of the compound under its intended conditions of use. Hence, greater amounts of toxicity data are generally needed to demonstrate the safety of bioactive compounds such as drugs, herbicides, and other biocides, and to support the safety of compounds with relatively high consumer exposure such as food ingredients. Less data are ordinarily necessary to demonstrate safety for constituents of food contact materials. As a general principle, starting reactants and reaction byproducts of food contact materials may be expected to be more bioactive and less stable, and therefore more toxic, than the end product. Fortunately, these reactive species tend to be present at relatively low residual levels in end product food contact materials.
The distinction between traditional thresholds used to set specific testing recommendations and the newer approaches of FDA's threshold of regulation and the recently proposed thresholds of toxicological concern is that the latter approaches are based on a more quantitative evaluation of the toxicity of industrial chemicals. Frawley (1967) first proposed a probabilistic analysis of toxicity data to support the establishment of a 300 mg/p/d dietary threshold for toxicity testing for the regulation of food contact materials. Frawley's proposal was based on his review of classical toxic endpoints in a limited number of chronic toxicity studies. In the mid-1980s, the FDA began examining the available scientific literature on toxicity testing, to develop a more quantitative basis for testing recommendations for food contact materials. The advent of larger collections of toxicity data such as the Registry of Toxic Effects of Chemical Substances (NIOSH, 2005), the carcinogenic potency database (CPDB) (Gold and Zeiger, 1997; EPA, 2005b), and others made it possible to perform a more quantitative analysis of the potential risks from consumer exposure to the world of chemicals.
Beginning in the 1980s, Flamm et al. (1987) and Rulis (1989) documented FDA's exploration of the use of large databases of toxicity data to address very low exposures to components of food contact materials more efficiently. Flamm et al. (1987) performed a probabilistic analysis of carcinogenic potency data in an attempt to discern a dietary level below which no specific toxicity testing data should be considered prerequisite to judge the safety of a compound used in a food contact material.
Continuing the treatment by Flamm et al. (1987), Rulis (1989) proposed a range of possible threshold of regulation levels for components of food contact materials, extending consideration to the practical limitations of analytical chemistry and regulatory review. Later, Machuga, et al. (1992) reasoned that such practical considerations support a level on the order of 3 mg/p/d. Rulis's (1989) proposed level of 0.15 mg/p/d would eliminate virtually all risk from potential carcinogens as well as other toxins. FDA established a minimum testing recommendation level of 1.5 mg/p/d in its policy on the threshold of regulation (FDA, 1995). This higher threshold was based in part on practical considerations including the ability of analytical chemistry to detect migrants at migration levels related to risk-based dietary thresholds. Although the higher limit of consumer exposure implies a higher potential risk, in practical terms the ability to identify compounds of concern without testing is substantially the same for all thresholds considered by FDA.
A critical assumption in FDA's development of a threshold of regulation is that carcinogenicity is the most sensitive toxicological endpoint at very low (mg/kg) dietary concentrations. FDA expects that protection against this critical endpoint will also protect against the less severe non-neoplastic toxic effects of food contact materials. Epidemiological data adequate to perform risk assessments for carcinogenicity are very rarely available for food contact materials or their constituents. Therefore, risk assessments for carcinogenic potential are usually based on extrapolations from the results of animal bioassays (relatively high doses) to levels of dietary exposure that may be of negligible or acceptable risk. Historically, upper-bound lifetime cancer risks estimated to be less than one in one million (10-6) have been considered to represent negligible risk. Upper-bound risk estimates in the range of one in a hundred thousand (10-5) have been accepted in specific cases, as the weight-of-evidence data mitigated concern. This extrapolation to negligible carcinogenic risk is tantamount to the application of a safety factor 2-3 orders of magnitude larger than those typically applied to non-neoplastic toxicological endpoints. The work by Munro et al. (1996), Cheeseman et al. (1999) and Kroes et al. (2000, 2004) provide a strong basis of support for the assumption that carcinogenicity is the toxic effect of most concern at the lowest dietary concentrations.
Cheeseman et al. (1999) examined the Gold CPDB underlying FDA's threshold of regulation to determine if other, more precise, thresholds could be established based on additional information on the structure or toxicity of food contact materials. They also conducted a more thorough examination of the range of concern levels for specific toxic endpoints other than carcinogenicity. Likewise, attempts by Kroes et al. (2000, 2004) to define a toxicological threshold of concern have also included a more careful consideration of the chemicals in the database which underlie FDA's threshold of regulation and the structure-based toxicological thresholds first proposed by Munro (Munro et al., 1996, 1999; Munro and Kroes, 1998). In addition, both Cheeseman et al. (1999) and Kroes et al. (2004) show the value of considering separately the structure activity relationships of the most potent compounds for different toxic endpoints. Both of these treatments demonstrate that toxicity thresholds may be safely established for most compounds or broad classes of compounds without toxicity data specific to the compound of interest. However, implementation of those thresholds should include special considerations for evidence that might identify more potent subsets of compounds.
FDA's threshold of 1.5 mg/p/d is an example of the threshold of toxicological concern approach. Below this threshold level of consumer exposure, FDA does not consider specific testing necessary to identify compounds with significant carcinogenic potential. This threshold of regulation is not meant to be a level below which no chemicals could be harmful. FDA's threshold of regulation analysis identifies many compounds of potential concern below 1.5 mg/p/d but also concludes that compounds of potential concern may be identified, by SAR analysis, without specific testing data and any necessary testing can be requested. Therefore, this threshold does not represent the level below which data will never be necessary to demonstrate safety. For example, a concern raised by preexisting test data or by structural analysis may be adequately addressable only by additional toxicity testing. It is essential that regulators and industry appropriately apply these caveats to any threshold of toxicological concern that is derived from a probabilistic analysis.
However, when raising such concerns as a regulator or risk manager, it is important to understand the data and assumptions underlying any threshold and the meaning that these data and assumptions give to individual decisions. Only a fraction of the untested chemicals would show evidence of carcinogenicity if actually tested (for obvious practical reasons, testing has been focused on compounds of greater concern). For example, only about 10 percent of known carcinogenic compounds would result in a potential lifetime risk greater than one in one million at the lower proposed threshold of 0.15 mg/p/d. Thus, a decision regarding the safety of an untested substance at this level can be based on an understanding of the likelihood that the substance is as potent as one of the top 10 percent of the most potent carcinogens known. Future testing might verify that this subset represents the top 1 percent of compounds of potential concern in the world of chemicals. Because the carcinogenic potency of each of these chemicals is inherently related to its chemical structure and resulting physical/chemical properties, structural analysis can be an effective approach to determining the safety at a reasonable degree of certainty for low consumer exposures (Cheeseman, 2005; Bailey et al, 2005).
As consumer exposures increase, regulators become progressively more concerned about the less potent compounds. As the group of compounds of concern grows to include members of lesser and lesser potency, the scope of chemical structures becomes more diverse and the decision-making process based on distinct structural clues becomes less certain. Although approaches such as the use of SAR analysis software bridge this gap by providing a systematic review tool, at some point consumer exposure and the commensurate risk rises to a level at which specific testing data become necessary to clarify the real likely risk. Whenever such levels of consumer exposure are defined, they can be viewed as thresholds of toxicological concern, above which additional steps are necessary to mitigate risk. As potential risk increases, these thresholds of toxicological concern must trigger requirements for additional testing, implementation of a higher level program of structural analysis, or procedures to prevent unsafe consumer exposure.
Just as a threshold level can be set to address the toxicological concern for carcinogenic risk, other threshold levels for toxicity testing may be established in order to address the likelihood of other toxicological risks. Kroes proposed multiple thresholds of toxicological concern based on Munro's structural classification of compounds tested for a diverse range of toxicological endpoints (Kroes et al., 2000, 2004; Munro et al., 1996). Munro used the structural classification scheme of Cramer et al. (1978) to segregate a representative data set of compounds into three structural classes. Munro then performed a probabilistic analysis of the range of acceptable daily intakes for each of these groups and proposed a threshold for each group based on the lower 95 percent confidence level for the acceptable daily intakes. As an aside, the threshold of regulation concept is separate from the concept of toxicity thresholds. The thresholds of toxicological concern approach proposed by Munro does not investigate mechanisms of compensation and repair, but rather is a method for prioritization and an efficient approach to risk-safety determination.
Kroes et al. (2000) further analyzed the need to specifically consider thresholds for different toxicological endpoints including neurotoxicity, developmental toxicity, immunotoxicity, and developmental neurotoxicity. They also considered the need for a separate threshold for teratogens, allergens, and endocrine-active compounds. Consideration of these specialized endpoints resulted in only a slight change in the originally proposed thresholds of toxicological concern; that being the addition of a consideration of organophosphate compounds as an especially potent subset of neurotoxins.
Collectively, the work by Munro et al. (1996) and Kroes et al. (2000, 2004) proposed several thresholds of toxicological concern based on toxicological and structural classifications. Table 7.2 summarizes these thresholds, which are cumulative as dietary concentrations increase. These thresholds are based on the analysis of compounds grouped using the so-called Cramer decision tree and on the structural analysis of compounds testing positive for specific toxic endpoints. Maintenance of such a system depends upon maintenance of the decision tree and the ability to continue to discern relationships between toxicity effects and structural information and to then translate those relationships into a decision tree.
Table 7.2 Thresholds of toxicological concern (from Kroes et al. 2004)
Criteria for safety at threshold
Verification that compound is not a polyhalogenated dibenzodioxin, dibenzofuran; or biphenyl, aflatoxin-like, azoxy-, or N-nitroso-compound.
There is no reason to believe the compound is genotoxic. The compound is not an organophosphate. Compound is not a member of Cramer structural class three (Cramer et al., 1978)
Compound is not a member of Cramer structural class two (Cramer et al., 1978).
Was this article helpful?
I lost over 60 pounds and 4+ inches off my waist without pills, strenuous exercise, or any of the things that the diet experts tell you to do...and I did it in less than 4 months! If you have the desire, and can read through my e-book , then this is for you! I could have easily made it a lot more difficult, with stacks of information that people will never read, but why?