Commentary on “Toxicity Testing in the 21st Century: Implications for Human Health Risk Assessment” by Krewski et al.

The main goal of Toxicity Testing in the 21st Century: A Vision and a Strategy(1) is to provide a practical means of evaluating the heath risks of chemicals. The vision is in stark contrast to the current situation, where the focus on apical testing in vivo is too expensive and too slow to keep up with ongoing influx of new chemicals into commerce. Key aspects of the vision are (a) in vitro evaluation of perturbations of toxicity pathways rather than in vivo evaluation of apical effects and (b) evaluation of dose response over relevant concentration ranges, as opposed to the current practice of focusing on doses that cause adverse effects in laboratory animals. While the NAS report lacks specifics on how the vision should be implemented, and even a precise definition of a toxicity pathway, its concern for a practical and dose-relevant approach to risk assessment is laudable. The central concern of risk assessment is to ensure that the public health is not adversely affected by exposure to chemicals and other environmental agents. It follows that risks should be predicted as accurately as possible, since accurate prediction helps to ensure that (a) the associated exposure standards actually do protect the public health and (b) the economic consequences of compliance with the standards are commensurate with the risk. In today's world we have little assurance that predicted risks are accurate, and while it is possible to argue that current practice is health protective, it is unlikely that exposure standards and their economic consequences are commensurate with actual risk. Since remediation is expensive,(2) assurance that cost is aligned with risk is necessary if we are to avoid the situation where inaccurate risk assessment results in misallocation of economic resources. Strikingly absent from Krewski et al.(3) and from Toxicity Testing in the 21st Century: A Vision and a Strategy(1) is any clear-cut statement of concern for the accuracy of predicted risks. This raises the possibility that the vision could be implemented in a manner that achieves high throughput, and that focuses on relevant doses, but that at the end of the day does not lead to more accurate risk assessment. Concern for protection of the public health, and for unnecessary economic consequences, would remain. This result is likely if the vision is implemented with a less-than-adequate understanding of the quantitative relationships between acute evaluations of in vitro perturbations of toxicity pathways and development of apical effects in humans, both acute and chronic. What factors should be considered to avoid such an unfortunate outcome? The vision clearly articulates the importance of an understanding of exposure and of the relationship between exposure and tissue dose. However, the focus on in vitro tests suggests that estimates of dose at the level of individual cells will be needed, not the average dose to all the cells in a tissue. We know very well that concentration gradients across tissues can be steep, as shown by Kimbell et al.(4) for the respiratory tract, and that important dose-dependent transitions occur in the activation of toxicity pathways.(5) Similar observations have been made for other tissues, including the liver(6) and kidney.(7) Implementation of the vision will thus require significant refinement in the current state of the art for both pharmacokinetic data collection and for dosimetry modeling. Today, average tissue doses are typically measured and dosimetry models consist of homogeneous, well-stirred compartments. The vision, however, will require some combination of “microdosimetry” data at the level of the individual cells and a robust ability to computationally predict such data. In the future, pharmacokinetic models that incorporate spatial information on tissue architecture may provide this predictive capability at the cellular level or, at a minimum, at some greater level of resolution than is typically available today. It should be noted that this need is already starting to be addressed in projects whose long-range goals are the development of “virtual tissues,” which are in silico descriptions of tissues across multiple levels of biological organization.(8) Beyond the issue of dosimetry, it is important to ask how accurately, in the quantitative sense, we can expect in vitro assays to toxicity pathways to recapitulate apical toxicity in vivo. A central concern here is that the quantitative tissue dose-apical toxicity relationship in vivo reflects adaptive responses designed to maintain homeostasis(9, 10) and that these adaptive responses may depend on higher levels of biological organization that will not be captured in in vitro assays. For example, the dose response for apical toxicities that are sensitive to steroid hormones will be affected by feedback signaling across the multi-organ endocrine system. It is difficult to envision an in vitro test, or even a suite of in vitro tests, that would adequately recapitulate such system-wide behavior. Here again, virtual tissues may prove be a key enabling technology, since they will in theory be able to integrate cellular-level data to higher levels of biological organization. Perhaps with time it will be possible to construct systems of linked virtual tissues capable of predicting dose response in vivo and of providing mechanistic insight into discrepancies between dose responses measured in vitro and in vivo. Such a capability is only a theoretical possibility today, but this is not inconsistent with the long-range nature of the vision and the recognition that substantial funding over a 10–20 year timeframe, or perhaps even longer, will be needed for its implementation.(1) In summary, issues that will need to be considered if the NAS/NRC vision is to be successfully implemented have been discussed. If we are to strive for risk assessment that is both high throughput and accurate, the issues of microdosimetry, adaptive responses, and homeostasis will need to be addressed. It is hoped that this commentary will be viewed not as a negative criticism of the vision but, rather, as encouragement intended to help focus the vision along a pathway leading to its successful implementation. This work was reviewed by the U.S. EPA and approved for publication but does not necessarily reflect EPA policy.