Requirements for a Credible Extrapolation Model Derived from Health Effects in Rats Exposed to Particulate Air Pollution: A Way to Minimize the Risks of Human Risk Assessment?

Abstract

For some years, several regulatory agencies have attempted to develop rather simple mathematical and biostatistical models to yield quantitative assessments of human risk derived from data of experimental rat exposures to typical particulate air pollutants. In our view, these models have failed to adequately consider some important biological aspects. The assessments have to bridge the gap between the low levels of human exposures and the intentionally high levels used in rat exposure studies. This is controversial even without the interspecies problem. Four crucial questions may be raised that have not been addressed adequately in past efforts: (1) What constitutes the effective dose of biopersistent particulate matter in the lung? (2) Is the effective dose in humans mechanistically the same as in rats? (3) Are the slope and shape of dose–response curves of rats applicable to humans? (4) Are model extrapolations from high to low exposures and from rats to humans valid? As the mechanistic similarity will remain an open question for some time, so will the lack of suitable data for chronic retention in human lungs continue to impede validation of today's rat-to-human extrapolations. For the other problems, however, improvements are possible and necessary. Invariably, published procedures treat particles as chemically active agents whether they are soluble or not. Furthermore, much emphasis is placed on the phenomenon of lung overload. However, the evaluation of diesel exhaust exposure studies shows that the retention of particles in the environmental range of ambient exposure concentrations up to 150 μg/m³ does not cause an impairment of alveolar macrophages. Focusing on insoluble particles, this article proposes that particle mass accumulated in the lung cannot per se be the relevant effective dose. Such a dose must be related to a target tissue or cell population and is represented by an integral of the total particulate surface and its residence time in the target area. Based on this effective relative dose, it can be shown by using data of lifetime exposures of rats that even crude dose–response relationships for lung tumor incidences show a strong nonlinearity. No-threshold probit analyses of rat data indicate that ubiquitous exposure concentrations for rats at diesel soot levels somewhere between 40 and 85 μg/m³ are virtually safe from lung cancer incidences. This is a very conservative assessment, because in reality the suspicious, nonchemical, nongenotoxic, rat-specific carcinogenicity can be seen only under heavy lung overload and probably involves a no-effect threshold. Another major drawback of published risk assessment procedures is the exclusive use of inappropriate lung retention models. Most frequently, postexposure models have been applied to chronic exposures. Almost all models are models of data, which, according to a definition by DiStefano and Landaw [Am. J. Physiol. (Regul. Integrat. Comp. Physiol. 15) 2456: R651–R664 (1984)], are only acceptable for interpolations (e.g., the radiological applications of the ICRP model). However, physiologically based models of systems that discriminate between various compartments in the pulmonary region give plausible disposition patterns of the retained particles and thus more credible extrapolations.