|
33 ELR 10071 | Environmental Law Reporter | copyright © 2003 | All rights reserved Genomics and Toxic Substances: Part I—ToxicogenomicsGary E. MarchantThe author is Associate Professor and Executive Director, Center for the Study of Law, Science, and Technology, Arizona State University College of Law. J.D. (1990); M.P.P. (1990); Ph.D. (Genetics) (1986). Portions of this Article were presented at a National Institutes of Environmental Health Sciences conference on Toxicogenomics in December 2001, and at presentations in 2002 to the Woodrow Wilson International Center for Scholars, the Environmental Law Institute, and faculty colloquia at Seton Hall Law School and Arizona State University College of Law. The author appreciates the many helpful comments and questions from the participants at those events, including some particularly valuable suggestions from Andrew Askland and Michael Saks. Advances in genomics, the study of the structure and function of our genetic make-up, are fundamentally transforming toxicology, the science of how toxic substances affect our bodies. These changes will inevitably spill over into the legal regimes that frequently rely on toxicological data, including toxic torts and environmental regulation.1 Genomic data, and the techniques with which they are generated, have the potential to make toxic torts and environmental regulation more effective, efficient, and fair, but at the same time will present many new doctrinal, evidentiary, and ethical challenges. Two key applications of genomic data for toxic torts and environmental regulation are2: (i) the study of the expression of genes in cells or tissues in response to exposure to a toxicant, known as toxicogenomics; and (ii) the identification of genetic variations affecting susceptibility to toxic agents, sometimes referred to as toxicogenetics.3 This Article will address the application of toxicogenomics to toxic torts and environmental regulation; a subsequent companion article will address toxicogenetic applications. After first describing the scientific background of toxicogenomics, this Article explores some potential uses of toxicogenomic data in regulation and litigation involving toxic substances. Toxicogenomics is defined "as the study of the relationship between the structure and activity of the genome (the cellular complement of genes) and the adverse biological effects of exogenous agents."4 A major focus of toxicogenomics is to characterize changes in gene expression in cells or tissues after exposure to toxic substances.5 Such exposure invariably results, either directly or indirectly, in characteristic changes in gene expression6 These gene expression changes [33 ELR 10072] may sometimes be the cause or in other cases the consequence of the early stages of a toxic response.7 Gene expression changes can be analyzed by collecting and characterizing messenger ribonucleic acid (mRNA) using a deoxyribonucleic acid (DNA) microarray. A DNA microarray (sometimes also referred to as a gene chip or DNA chip) consists of a set of many different singlestranded genetic sequences fixed to a substrate, such as a glass slide or membrane, in a defined pattern. The genetic markers can consist of short (500 to 2,000 base pair) DNA sequences that are complementary to, and thus bind to, genes of potential interest. Fifty thousand or more of these specific DNA sequences can be spotted (or printed) onto the fixed substrate in precise locations or spots on a grid, with each spot on the array containing several million identical copies of a DNA segment from a specific gene.8 Alternatively, shorter synthesized DNA sequences, called oligonucleotides, can be constructed directly onto the substrate using a process called photolithography.9 At least 20 companies have commercialized DNA microarray products,10 and at least 1 manufacturer has already released a commercial gene chip containing the entire human genome.11 DNA microarrays can be used to identify and characterize the response of a cell or tissue to an external stimulus or perturbation, such as exposure to a toxic substance.12 The complement of mRNA culled from the cytoplasm of a cell provides a snapshot of the genes that are being expressed in the cell at that time.13 The mRNA from treated or control cells can be collected and then copied to form DNA sequences known as complementary DNA (cDNA) that can be "tagged" with a flourescent marker.14 The cDNA sample is then added to the microarray, and cDNA sequences will bind (or hybridize) to sites on the microarray that contain DNA with a matching sequence. A laser scans the microarray and generates flourescent spots at the locations where the cDNA binds, and the intensity of the signal at each spot will be proportional to the abundance of matching mRNA in the original sample isolated from the treated or control cells.15 Sophisticated computer programs are available to "read" the microarrays and produce digital graphic readouts of the genes being expressed in the cells of interest. DNA microarrays thus permit the almost instantaneous and simultaneous genome-wide detection of the expression of thousand of genes, even if the function of some of the genes is unknown.16 This genome-wide or "global" gene expression analysis is a major advancement over previous methodologies which only permitted analysis of the expression of one or two individual genes at a time.17 Toxicity usually involves the induction (up regulation) and repression (down regulation) of [33 ELR 10073] many different genes,18 and thus only the genome-wide as-says made possible by DNA microarrays can evaluate the entire cascade of gene responses to toxic exposure.19 Exposure to chemicals that cause toxicity through a mechanism involving DNA alkylation, for example, results in changes in the expression of over 2,000 genes.20 Other chemicals induce changes in the expression of a more modest but still substantial number of different genes.21 The use of DNA microarrays to study global gene expression provides "a tool of unprecedented power for use in toxicology studies."22 Gene expression changes measured by microarrays have the potential to provide a more sensitive, characteristic, and earlier indicator of a toxic response than typical toxicological endpoints such as morphological changes, carcinogenicity, or reproductive toxicity.23 Microarray data promise greater specificity because while "there are a limited number of cellular, organ, and organismal manifestations of chemically-induced toxicity, the possible number of gene expression patterns for encoding those manifestations is enormous."24 Many different toxic agents may be capable fo causing the same toxicological endpoint, e.g., a liver tumor, in many cases by different mechanisms, whereas each chemical will produce a unique gene expression profile, thus providing a higher resolution tool with much greater specificity than simply monitoring the toxicological endpoint.25 Microarrays also permit evaluation of all toxicological endpoints in a single assay, whereas traditional toxicological methods generally require separate studies for carcinogenicity, mutagenicity, reproductive toxicity, teratogenicity, immunotoxicity, neurotoxicity, and endocrine disruption.26 Yet another advantage of studying gene expression changes to assess toxicity is that such alterations can occur almost immediately following exposure, whereas the clinical manifestation of toxicity may take days, months, or even years to develop.27 Because these toxicological endpoints are the end result of earlier molecular events that can be monitored by microarrays, it is possible to screen for toxicity much more quickly and earlier using microarrays than with traditional toxicological methods.28 Moreover, because they represent an earlier step in a toxic response, gene expression changes will be detectable in a larger percentage of the exposed animal or human population than will ultimately go on to develop clinical disease, thereby providing a more statistically robust measure of effect. For these reasons, gene expression changes assayed using DNA microarrays have the potential to provide both an earlier and more sensitive biomarker of a toxic response. Of particular interest is the rapidly growing body of evidence demonstrating that specific chemicals or classes of chemicals with similar toxicological properties produce a characteristic gene expression "fingerprint" or signature profile. Initial "proof-of-principle" experiments have successfully identified the identity or toxicological mechanism of chemicals based on their gene expression profiles.29 The [33 ELR 10074] potential applications of these findings are significant. The observation that groups of chemicals with a common toxicological mechanism produce a characteristic pattern of gene expression changes means that it may be possible to predict the toxicological nature and mechanism of an untested chemical in a quick and inexpensive gene expression assay.30 The finding that it is possible to discern exposure to an individual chemical based on unique gene expression changes suggests that it may be possible to use microarrays to measure exposure or toxic responses to specific chemicals in individuals or populations.31 Finally, the capability to identify the genes that are affected by a particular chemical may be useful for discovering the toxicological mechanism of a toxicant, which would be very helpful in characterizing the risks associated with that toxic agent.32 Toxicogenomics not only has tremendous potential, but this potential is unfolding at an unprecedented rate. As described by one set of reviewers, "unlike other new approaches or methods in toxicology that have been adopted slowly," toxicogenomic methods "are being evaluated and adopted rapidly by all sectors of industry, academia, and regulatory agencies at an unprecedented rate."33 Others have referred to the rapid emergence of toxicogenomics as "the microarray stampede."34 In a mere five years, toxicogenomics has transformed the scientific study of toxic substances, as microarray analysis has become one of the most common and informative methodological approaches of toxicology.35 The toxicogenomic revolution will have many important potential applications for both toxic torts and environmental regulation. Some of these applications are discussed below. Toxicogenomic Applications for Toxic Torts Plaintiffs often face significant obstacles in toxic tort cases in satisfying their burden of proof to demonstrate exposure and causation.36 Gene expression data may help plaintiffs to demonstrate exposure and causation when they exist, and conversely to help defendants demonstrate the absence of exposure or causation when they are lacking. One of the first hurdles a plaintiff must surmount in a toxic tort lawsuit is to prove that he or she was exposed sufficiently to a toxic agent associated with the defendant's product or conduct.37 In many toxic tort cases, direct evidence of the fact or quantity of exposure is limited or lacking altogether.38 Examples include plaintiffs who allege they were injured by chemicals that had leached into their groundwater, pesticides that were sprayed in their homes or workplaces, or pollutants emitted into the air nearby. In these types of cases, courts have frequently dismissed plaintiffs' cases for failure to meet their burden of proof to demonstrate (or quantify) exposure.39 Gene expression data using microarrays may assist plaintiffs in demonstrating exposure, or in other cases to support a defendant's argument that there was not sufficient exposure.40 Gene expression assays of the plaintiffs' blood or skin cells may demonstrate the presence (or absence) of gene expression "fingerprints" that are characteristic of the toxic substance to which the plaintiff was allegedly exposed. Such an assay might even be capable of quantifying the level and duration of plaintiff's exposure. A number of uncertainties and questions would need to be addressed to establish the reliability of this exposure evidence. For example, how well characterized and validated is the gene expression "fingerprint"? In other words, how certain can we be that a particular gene expression pattern is indeed representative of a particular toxic exposure?41 Can other potential sources of exposure to that same toxic substance (or other substances that cause similar responses) be excluded? Are the gene expression changes in the easily as-sayed [33 ELR 10075] tissues such as blood or skin cells representative of the changes in the less accessible target tissue in which the plaintiff's disease is more likely to have occurred?42 What is the quantitative relationship between the level of exposure and the magnitude of gene expression changes? Over what range of exposure is this relationship valid? Do interindividual differences in susceptibility (genetic or nongenetic) affect gene expression patterns in different individuals?43 How does gene expression vary with single, acute exposures versus long-term chronic exposures? A critical set of issues relates to the timing of gene expression changes. Specifically, what is the time course of the gene expression changes following toxic exposure, and are these changes transient or longer term?44 Very little existing data address the issue of the duration of gene expression changes in cells, and how those changes either progress or diminish over time.45 The limited data that are available suggest that gene expression changes may only provide a valid measure of exposure within a few days of exposure.46 If that is the case, then gene expression data will have little or no utility for latent diseases unless the gene expression data is collected close to the time of exposure, rather than when the disease manifests itself many years later.47 On the other hand, some exposures to a toxic substance may produce permanent changes in gene expression, perhaps as a result of gene mutations, gene amplifications, or changes in DNA methylation patterns, all of which control gene expression.48 To the extent that these types of changes in gene expression can be verified, they may produce a much more durable marker of exposure. The legal significance of the temporal component with respect to the use of genetic biomarkers to quantify exposure is demonstrated in an analogous context by the litigation brought by residents near the Three Mile Island (TMI) nuclear reactor, who claimed that a radioactive plume from the 1979 reactor accident caused their cancer. The plaintiffs lacked direct measurements or adequate modeling evidence to prove that they received sufficient radiation exposure to cause their cancers, which the court described as the "critical issue" in the case.49 They attempted to overcome this problem by introducing evidence that they had an increased frequency of a particular type of chromosome aberration (dicentric chromosomes)50 in their blood cells (lymphocytes), which they claimed provided a quantitative biomarker of radiation exposure.51 The U.S. Court of Appeals for the Third Circuit held that such applications of genetic markers "is an accepted method, not simply for determining if the subject of the analysis was irradiated, but also for estimating radiation dose to the individual."52 The court found, however, that while "radiation dose estimation based on dicentric enumeration is a valid and reliable scientific methodology," the "validity and reliability decrease as the time gap between the alleged irradiation and the dicentric count increases."53 The court concluded that dicentric chromosomes could only provide an accurate indicator of dose within one or two years of exposure, and thus plaintiff's reliance on dicentric chromosome levels assayed 15 years after the TMI accident were no longer a reliable measure of exposure.54 The lesson from the TMI litigation is that litigants who seek to rely on gene expression changes to quantify exposure will need to evaluate the gene expression changes in the plaintiffs' cells as soon as possible after exposure. In addition, it will be necessary to lay a proper foundation for using such evidence to quantify exposure by providing data on how such gene expression changes change over time from exposure. Notwithstanding these limitations, the use of DNA microarrays to monitor gene expression changes within a plaintiff's cells has the potential to provide a very [33 ELR 10076] specific and informative quantitative molecular dosimeter that can be used by plaintiffs to demonstrate and quantify their exposure to toxic substances or by defendants to demonstrate the lack of sufficient exposure.55 Plaintiffs who are able to adequately quantify their exposure must next prove that this exposure caused their injuries. Causation analysis typically involves two steps.56 First, the plaintiff must prove general causation, which requires a demonstration that the toxic agent produced by the defendant is capable of causing the health effect incurred by the plaintiff.57 Most courts have required relevant evidence that is specific to both the toxic substance and the health effect in question for demonstrating general causation. Thus, courts have generally excluded evidence that the same chemical can cause diseases perhaps related to but different than the condition afflicting the plaintiff, such as tumors in other types of tissues in the case of a plaintiff with cancer.58 Plaintiffs have likewise often been precluded from relying on evidence showing that chemicals related to the one which the individual plaintiff has been exposed can cause the specific health effect for which the plaintiff has been diagnosed.59 The challenge facing plaintiffs in proving general causation is thus much more daunting than, for example, the challenge facing a regulatory agency attempting to regulate the same substance. The regulatory agency need only show that the chemical might cause any adverse health effect in some people. In contrast, a toxic tort plaintiff has the burden of proof to show that the chemical did cause a specific adverse effect, i.e., that from which plaintiff suffers, in a particular person. "Toxic ignorance,"60 or the lack of adequate testing data for many potential toxic substances, thus severely limits a plaintiff's ability to introduce the required data on a specific chemical-health effect relationship, given that data evaluating many such relationships will often be nonexistent.61 Toxicogenomic data may provide some opening for plaintiffs to proceed when no data are available for the specific toxicant-health effect combination relevant to their case. For example, consider a case in which plaintiffs have been exposed to agent A and had developed kidney cancer, but there is no toxicological data directly linking agent A and kidney cancer. Plaintiffs may nevertheless be able to rely on data showing that agent A stimulates gene expression changes that are similar to those induced by agent B, which has been found to cause kidney cancer. Alternatively, plaintiffs may be able to introduce evidence that agent A causes liver cancer by a mode of action that involves a characteristic gene expression profile, and that agent A has also been observed to cause a similar gene expression change in kidneys in animal studies, even though such studies have not detected a statistically significant increase in kidney tumors.62 These types of toxicogenomic data may provide a molecular link between agent A and kidney cancer even in the absence of data directly showing such a relationship. A Texas case, Austin v. Kerr-McGee Refining Corp.,63 can be used to further illustrate the potential role of toxicogenomics to demonstrate general causation. In that case, the spouse of a deceased worker who had worked with mineral spirits claimed that benzene in the mineral spirits caused her deceased husband's chronic myelogenous leukemia (CML). The plaintiff's causation expert relied primarily on studies showing that benzene caused acute myelogenous leukemia (AML), and argued that these and other data show that benzene causes all types of leukemia, including CML.64 [33 ELR 10077] The court held that the plaintiff's expert must demonstrate "that all types of leukemia are related or interchangeable," which the expert attempted to demonstrate by alleging that benzene causes a common genetic mutation in bone marrow cells that can result in all types of leukemia.65 The court concluded that the expert failed to adequately support his argument that all types of leukemia derive from a common genetic mutation.66 Gene expression data might have provided the missing link required by the court to treat AML and CML as related. If microarray data were available showing that benzene causes the same changes in gene expression in patients who ultimately developed either AML or CML, a persuasive case could be made that the two types of leukemia shared a common mechanism. Alternatively, DNA microarrays could be used to compare the DNA mutations in patients with AML and CML, rather than their gene expression profiles.67 Microarray evidence that AML and CML tumors contain similar mutations would also be probative of a common mechanism. If microarray data can provide such a molecular link between AML and CML, the available epidemiology studies showing that benzene increased the incidence of AML would therefore arguably be relevant to whether benzene can cause CML. Alternatively, if benzene exposure results in different patterns of gene expression changes in AML and CML patients, the defendant's case would be strengthened, because such a finding would indicate that benzene does not cause a similar response in patients who develop AML and CML. In that situation, the question of whether benzene indeed does cause CML would require direct evidence of such a causal relationship, and could not be based on the AML findings. The second step of the causation inquiry is specific causation. While general causation refers to the question of whether an agent can cause the disease from which the plaintiff suffers, specific causation asks whether the agent did in fact cause the disease in that specific individual.68 Proving specific causation is perhaps the most formidable challenge facing a plaintiff, because human or animal studies of toxic risks evaluate the overall rate of disease in an exposed group versus a control population, but generally have no way of discerning which individuals in the exposed group developed the disease from the toxic exposure as opposed to background factors. As one court succinctly stated it, "science cannot tell us what caused a particular plaintiff's injury."69 Plaintiffs generally try to overcome this specific causation hurdle using one of two approaches.70 The statistical approach attempts to show that exposure to defendant's agent more than doubles the background risk of the health effect at issue, i.e., relative risk >2.0, thus making it statistically more likely than not that the plaintiff's health effect was caused by defendant's product.71 This approach has two shortcomings. First, relatively few toxic agents double background risk, especially for health effects that are relatively common in the general population.72 Second, some commentators have argued that even if a plaintiff demonstrates a doubling of background risk, such purely statistical evidence is insufficient to establish specific causation in the absence of "particularized" data relating to the specific plaintiff, although few courts have adopted this suggestion.73 The second approach for establishing specific causation is through differential diagnosis, in which a physician rules out other known possible causes of the health effect based on the case history and other clinical evidence.74 The courts have been inconsistent on whether and under what circumstances they will allow differential diagnosis evidence to be introduced to prove specific causation.75 Some courts have held that expert testimony based on differential diagnosis [33 ELR 10078] can be used to establish specific causation,76 but other courts have been more skeptical of this approach.77 Given the limitations of the two existing approaches for establishing specific causation, there is likely to be significant potential and interest in the use of DNA microarrays for proving that a particular agent did or did not cause the disease process in an individual plaintiff. At least two potential applications of microarrays are relevant to specific causation. First, a plaintiff could assay for gene expression changes in his or her cells that are characteristic of the specific toxic agent associated with the defendant. Here, the types of gene expression changes that would be most relevant are not those of the initial cellular response to exposure to the toxic agent, but rather the subsequent gene expression changes that are typical of the developing disease process. Preliminary studies have demonstrated that different toxic compounds produce a "unique expression profile."78 A second potential use of DNA microarrays would be to assay not for changes in gene expression, but rather for changes in the DNA sequence that represent chemical-specific mutations in genes relevant to the disease process.79 For example, the p53 tumor suppressor gene is mutated in over 50% of human tumors, and several important human carcinogens appear to induce chemical-specific "mutational fingerprints" at precise sites in the p53 gene.80 Microarrays could be used to detect these specific mutations in a plaintiff with cancer, and used to establish specific causation by showing that the p53 mutation in their tumor was characteristic of the specific agent produced by the defendant.81 Conversely, the absence of such biomarkers would support the argument of defendants that there was no specific causation. By allowing scientists to "peer" inside cells and look for chemical-specific genetic markers of disease processes, whether they be changes in gene expression or mutational spectra, toxicogenomics offers to provide the first direct evidence of specific causation. The lack of such direct evidence explains in large part why toxic tort cases have generally been much more controversial and difficult than traditional personal injury cases involving traumatic injury, such as automobile accident cases, where the issue of "specific causation" is obvious.82 If toxicogenomic data can provide reliable direct evidence of specific causation in toxic tort cases, there may no longer be a need to evaluate general causation, as direct evidence that a particular substance did or did not in fact cause a given plaintiff's illness moots the issue of whether the substance is capable of causing such disease.83 Toxicogenomic data thus offers the potential of an unprecedented advance in directly demonstrating causation or the lack thereof in toxic tort cases. In recent years, plaintiffs exposed to hazardous substances frequently seek recovery for their latent risks that have not yet manifested into clinical disease. Such claims usually seek damages for the increased risk of future disease as well as recovery for the present fear associated with the increased risk. To prevent a flood of latent risk claims, yet at the same time providing the possibility of recovery for the most compelling claims, courts have imposed stringent threshold requirements for such claims.84 For example, most courts require proof of a "present injury" for increased risk and fear of disease claims,85 as well as a demonstration (and often quantification) of a sufficient quantum of increased risk.86 Most plaintiffs exposed to hazardous substances are unable to meet these threshold requirements, at least with the types of scientific evidence presently available.87 Gene expression data may assist plaintiffs in appropriate cases to demonstrate both an existing injury and a sufficient increase in risk to trigger recovery. By providing a highly sensitive and specific assay of toxicological response at the molecular level, microarrays may demonstrate subcellular effects that may qualify as a "present physical injury" in at least those jurisdictions that permit asymptomatic conditions to satisfy the present injury requirement.88 Other jurisdictions require symptomatic disease to satisfy the present injury requirement, motivated in large part by the difficulty in objectively proving alleged subcellular injuries.89 Even in those jurisdictions, a plaintiff that can objectively show gene expression changes that have been validated as a reliable marker of developing toxicological injury may be able to assert a credible argument for relaxing the legal insistence on symptoms to establish present injury. Likewise, gene expression data may provide objective quantitative evidence of increased risk, which, if of adequate magnitude, could satisfy the other pre-condition for recovery for latent risks which is that the plaintiff demonstrate a sufficiently enhanced risk. The issue of whether and when to allow recovery for latent risks has been described as the most difficult problem confronting toxic torts.90 Most jurisdictions are still grappling with this issue, often seeking to balance the competing policy considerations for and against recovery for latent risks by imposing restrictive threshold requirements that until now have excluded most latent risk claims. Toxicogenomics offers a tool of unprecedented power for satisfying the evidentiary requirements for latent risk claims, and will likely make recovery for latent risk both more and less problematic. It will be less problematic to the extent that toxicogenomic data provide some objective, scientific evidence of future risk that can better inform cases that are today litigated with almost complete ignorance of an individual's actual future risk. On the other hand, gene expression assays may bring to fruition the fears that latent risk claims could flood the courts with an almost unlimited number of new, asymptomatic litigants.91 Legal decisionmakers (both courts and legislatures) are likely to be confronted with the difficult question of whether to allow recovery for a much larger number of latent risk claims that meet the existing threshold requirements using toxicogenomic data.92 Courts in at least 17 states and the District of Columbia have recognized claims for medical monitoring, in which exposed at-risk plaintiffs can recover for future periodic medical tests intended to detect the onset of latent diseases resulting from exposure to toxic substances.93 The precise formulation of the standard for awarding medical monitoring costs varies somewhat between different States, but most courts have required that plaintiffs must suffer from an increased risk of contracting a serious latent disease as a proximate result of defendant's negligent act, that this increased risk makes periodic diagnostic medical examinations reasonably necessary, and that monitoring and diagnostic methods exist that make early detection and treatment of the disease both possible and beneficial.94 Unlike increased risk and fear of disease claims, medical monitoring [33 ELR 10080] claims in most jurisdictions do not require proof of present injury.95 Plaintiffs may soon seek medical monitoring expenses to conduct gene expression assays on exposed individuals at increased risk of developing future disease. Such assays, if properly validated, may provide a much more reliable assessment of pre-clinical disease progression in such individuals, in many cases perhaps leading to timely medical intervention. By providing a more sensitive and specific diagnosis of the disease process before it manifests into clinical symptoms, gene expression assays have the potential to greatly expand the number of valid medical monitoring claims as well as to prevent or mitigate many new disease cases. On the other hand, microarrays may have the potential to produce too much information and too many putative plaintiffs. Every American has certainly been exposed to toxic substances in some form or level, whether it be from living near a hazardous waste site or polluting facility, exposure to pesticides, inhaling second-hand smoke, or using products with hazardous constituents.96 Microarrays may for the first time provide the scientific capability to directly monitor molecular changes in the exposed population. As mentioned above, most courts do not require plaintiffs to demonstrate a significant quantum of increased risk as a threshold to recover medical monitoring damages.97 To the extent that gene expression changes resulting from toxic exposures meet the criteria for medical monitoring tests, every person could conceivably be entitled to medical monitoring damages.98 Hence, the scientific expansion of the capability to test for disease development may force the legal contraction of the right to recover for the costs of such testing. Toxicogenomic Applications for Environmental Regulation The revolutionary impact of toxicogenomics for the science of toxicology will translate into equally fundamental changes in regulatory risk assessment and decisionmaking. Some of these transitions have already begun. Several potential applications of toxicogenomic data to environmental regulation are discussed below. Several major uncertainties limit the confidence in and utility of risk assessment for informing regulatory decisions.99 These uncertainties include extrapolating from animal results to humans, extrapolating from high-dose experimental results to more typical low-dose human exposures, understanding the mechanism of action of a toxicant and its implications for risk assessment, determining the shape of the dose-response curve, and estimating the exposure levels for actual human populations.100 Gene expression data may help to overcome many of these limitations.101 In the summer of 2002, the U.S. Environmental Protection Agency (EPA) issued an Interim Policy on Genomics which stated that "EPA believes that genomics will have an enormous impact on our ability to assess the risk from exposure to stressors and ultimately to improve our risk assessments."102 EPA's Interim Policy states that genomics can be used "to explore the possible link between exposure, mechanism(s) of action, and adverse effects," and may also be useful to EPA "in setting priorities, in ranking of chemicals for further testing, and in supporting possible regulatory actions."103 While the Interim Policy states that genomic data may be considered in current regulatory decisions, it cautions that for at least the time being, such data alone are "insufficient as a basis for decisions," and "EPA will consider genomics information on a case-by-case basis."104 There are several potential ways toxicogenomic data can improve risk assessment. First, gene expression data, by providing a characteristic "fingerprint" of different toxicological mechanisms,105 can be used to characterize the mechanism or mode of action of a toxicant.106 Regulatory [33 ELR 10081] agencies such as EPA have recently focused on mode of action as a central factor in risk assessment, because this information is critical for estimating the shape of the dose-response curve, extrapolating of results from animals to humans, and deciding whether or not the agent is likely to exhibit a threshold below which there is no significant toxicity.107 Gene expression assays "provide a new and powerful way of determining the mode of action," in that "association of a given toxic endpoint (e.g. carcinogenicity, genotoxicity, hapatoxicity) with a particular pattern of gene/protein expression … may provide a 'fingerprint' that is characteristic of a specific mechanism of induction of that toxicity."108 Indeed, DNA microarray analyses have already been used to identify unique mechanistic pathways through which the body reacts to certain classes of toxic exposures.109 Second, gene expression data will be useful in extrapolating results obtained in animal and epidemiology studies that typically involve high-dose levels to lower doses more relevant for the general human population.110 Until now, lowdose effects have generally been refractory to empirical analysis, and risk assessors have had to rely on models to extrapolate results from high- to low-dose levels.111 For example, the risks of low-dose ionizing radiation have long been a subject of controversy, but direct testing of the dose-response relationship at low doses has been beyond the reach of existing toxicological methods. Direct detection of low-dose effects from ionizing radiation and other agents by assessing gene expression changes will provide much needed information to better characterize and quantify risk levels at low doses.112 A finding that gene expression changes characteristic of the carcinogenic response at high doses are also observed in low-dose groups, even though those low-dose animals may not develop tumors, may indicate that low-dose exposures present a carcinogenic risk in large populations. Alternatively, the absence of any characteristic gene expression response in low-dose animals may suggest that the carcinogenic response only occurs at high doses.113 Third, comparing gene expression changes in rodent and human cells after a similar exposure may provide information on the relevance of rodent tumor responses for human health risk.114 Most toxicology studies are necessarily conducted in animal species, and the extrapolation of animal results to humans raises an additional important element of uncertainty in risk assessment. While most chemicals that cause cancer in mice or rats are also carcinogenic in humans, there are now a number of examples where an agent causes toxicity in rodents but not humans, or humans but not rodents.115 By providing a quick and inexpensive test of whether a chemical is causing a similar response in rodents and humans, gene expression assays can help detect and prevent what would otherwise be false positives for chemicals that cause toxicity in rodents but not humans, and false negatives for chemicals that cause toxicity in humans but not rodents.116 Gene expression profiling can thus provide a "bridging biomarker" that can connect toxicological responses in animals and humans.117 Fourth, gene expression data will also be useful in quantifying human exposure, a key input to risk assessment. Some experts consider that the limited information available on exposure is perhaps the most serious problem afflicting risk assessment.118 Without reliable data on human exposure, it is not possible to estimate accurately the relationship between dose and response that underlies risk assessment estimates. By characterizing gene expression patterns in exposed persons, microarrays have the potential to provide more precise quantitative estimates of exposure to specific toxic substances in contemporaneous and prospective human studies.119 Fifth, gene expression profiling may be particularly useful for evaluating the toxicity of chemical mixtures,120 which are the most typical human exposure scenarios, but which are hard to evaluate using traditional toxicological methods.121 The combined effects of exposure to several different toxic substances present in a mixture may not be additive, but may instead be greater, e.g., synergistic, or lesser, e.g., antagonistic, than expected from simply adding the predicted effects of the individual compounds.122 Because DNA microarrays permit the simultaneous monitoring of all gene expression changes within a cell in a single experiment, they "are particularly suitable to evaluate any kind of combinational effect resulting from combined exposure to toxicants."123 The National Institute of Environmental Health Sciences has made the study of mixtures its "top priority," in significant part because the availability of microarrays will for the first time make the toxicological assessment of mixtures feasible.124 Finally, gene expression assays may also provide a more sensitive methodology for examining other risk assessment issues such as the differential sensitivity of children versus adults to specific environmental exposures. For example, if gene expression changes show that a pesticide induces a greater relative response in neonatal rodents than in adult rodents, there would be good reason to suspect that human children might be more susceptible to a toxic response than adults. Conversely, if there are no differences in gene expression between neonatal and adult rodents after exposure to a particular pesticide, there would be less concern about differential susceptibility of human children, and perhaps grounds for not applying the default additional tenfold uncertainty factor for children under the Food Quality Protection Act (FQPA).125 These potential applications of gene expression data may help reduce many of the most important uncertainties in risk assessment, although by no means eliminating such uncertainties altogether. The major existing uncertainties in risk assessment and the resulting controversies they have spawned have produced significant delays in implementing risk-based regulation, and have resulted in a shift away from risk-based standard setting in recent years to other approaches such as setting standards based on the best available technology.126 Perhaps these frustrations with risk assessments were best expressed by one prominent senator's declaration during the 1990 Clean Air Act Amendments that he "would be glad to declare risk assessment dead."127 Toxicogenomics has the potential to help restore confidence in risk assessment by reducing many of the most important uncertainties, and may thereby open the door to a renewed emphasis on risk-based standards, since after all it is an acceptable level of risk and not best technology per se that provides the most direct measure of what environmental regulation seeks to protect, i.e., human and environmental health. High Throughput Toxicity Screening of Chemicals The majority of chemicals in commercial use in the United States have not been comprehensively tested for human toxicity and carcinogenicity potential.128 Other than pharmaceuticals and pesticides, there is no legal duty imposed on manufacturers to pre-market test their products for toxicity. EPA and the chemical industry have begun to address this data gap for chemical risk assessment with the High-Production Volume (HPV) chemical testing initiative.129 However, given that there are now some 80,000 chemicals in commerce,130 it is not feasible to conduct traditional toxicological testing for all or even most chemicals in commerce with existing test methods. For example, the "gold standard" assay for carcinogenicity is the chronic rodent bioassay, in which rats or mice are exposed to a potentially carcinogenic substance over a two-year [33 ELR 10083] period, followed by a comprehensive pathological examination.131 The largest chemical testing program in the United States, conducted by the National Toxicology Program of the National Institute of Environmental Health Sciences (NIEHS), has recently completed its 500th chronic rodent bioassay for evaluating carcinogenicity after 30 years of testing.132 A chronic rodent bioassay can take five or more years to complete from start to finish, and costs $ 2 to $ 6 million per chemical.133 Given these time and cost requirements, it is simply not feasible to conduct chronic bioassays for all or even most chemicals in commerce. Moreover, even for those relatively few chemicals that are evaluated in chronic bioassays, additional tests may be needed to assess other toxicological endpoints such as reproductive toxicity, developmental toxicity, immunotoxicity, neurotoxicity, endocrine disruption, and other possible effects. The Director of the NIEHS recently testified to the U.S. Congress that a large number of commercial products require additional testing, but "we can never satisfy this testing requirement using traditional technologies."134 There is a pressing need for rapid, inexpensive, and reliable assays that can be used to screen a large number of chemicals for toxicity. Genotoxicity assays and structureactivity relationship (SAR) analyses are currently used to screen many chemicals relatively quickly and cheaply, but these assays are limited in their utility and predictiveness.135 Alternative testing models, such as transgenic mice, are being developed to provide less expensive and more rapid assays, but these models have not yet been fully validated, and moreover still involve a considerable expenditure of time and resources.136 Gene expression assays have tremendous potential for providing a rapid, inexpensive and high throughput screening of chemicals for a wide range of genotoxic and nongenotoxic responses.137 The Director of the NIEHS predicts that by using DNA microarrays the time it takes to test potential carcinogens will be reduced from several years to "a few days," while the costs will be reduced from millions of dollars to conduct a chronic bioassay "to less than $ 500 dollars" to test each chemical using DNA microarrays.138 Microarrays can be used to interrogate the gene expression of cells either in tissue culture or in living mice or rats that have been treated with chemical candidates, with the resulting gene expression profiles used to classify those chemicals to specific toxicological categories and to characterize their likely risks.139 It may be possible to successfully and accurately make such classifications within 24 hours of an initial exposure, long before any physical manifestation of toxicity.140 In addition to providing a cheaper and quicker toxicity screen than existing test methods, microarrays offer several other important advantages for toxicity screening. Because microarrays simultaneously monitor all changes in the expression of all genes within a cell, all toxicological end-points can be evaluated in a single microarray assay, whereas today separate tests are currently needed to evaluate carcinogenicity, genotoxicity, developmental toxicity, reproductive toxicity, immunotoxicity, neurotoxicity, and endocrine disruption.141 Microarrays are also more sensitive than current methods, because they can detect immediate changes within every exposed cell or organism, whereas existing methods can only detect observable toxicity that develops many weeks, months, or even years after exposure, and in only some cells or organisms within the study population. Initially, gene expression assays will need to be conducted in association with traditional toxicity testing until a sufficiently robust and validated data set has been accumulated to reliably correlate specific gene expression profiles with particular toxicological mechanisms and endpoints.142 Used in conjunction with traditional toxicology tests, gene expression data have the potential to improve the sensitivity and interpretability of the standard tests.143 Once such a database has been established, gene expression assays might replace some or all of the current toxicological screening and testing assays, or at least to narrow and select the specific assays that are indicated by the observed gene expression pattern.144 One possible initial regulatory application of this gene expression technology, which would also help to build the [33 ELR 10084] necessary database to validate gene expression data, would be to require companies submitting premanufacturing notices (PMNs) for new chemical substances under § 5 of the Toxic Substances Control Act (TSCA)145 to include the results of a gene expression assay in their submission. Currently, companies are not required to generate any new data to support PMNs; they are only required to submit relevant data already in their possession.146 Requiring a manufacturer of a new substance to conduct and submit a gene expression assay would not be unduly burdensome, and would begin to build an experiential database of chemical-specific gene expression data that would then be available to EPA. Such a database would only be useful to the extent that the submitted data were roughly consistent in the genes and methods used, and so some form of standardization of microarray platforms and methods would be required before such a program could be implemented. In addition to potential regulatory applications in the screening of new chemicals under statutes such as TSCA, screening of chemicals using DNA microarrays have a number of other potential regulatory applications. For example, the chemicals included on the toxic release inventory (TRI) list of reportable substances might be based at least in part on the results of DNA microarray analyses.147 Similarly, the identification of listed hazardous wastes or hazardous wastes based on the characteristic of toxicity could be based on a quick and inexpensive microarray assay evaluating whether the waste induces a gene expression profile that is characteristic of a known toxicity mechanism.148 Under the FQPA, EPA must combine all pesticides that share the same mechanism of toxicity in a single cumulative risk assessment.149 Toxicogenomic data may indicate which pesticides should be grouped together for such assessments.150 Gene expression data may also be useful for prioritizing contaminated sites. Indeed, in the foreseeable future EPA might want to consider adding gene expression assays to its hazard ranking scheme for establishing the national priorities list (NPL) under Superfund.151 Gene expression data may also be useful in assessing risks and selecting appropriate cleanup options at individual waste sites. Many abandoned waste disposal sites contain large quantities of soils and sediments with moderate or low levels of contamination which present uncertain risks but very large cleanup costs.152 In most cases, the primary potential hazard is to local ecosystems and species rather than to human health. Assessment of these ecological risks using standard toxicological tests can be very expensive and highly uncertain due to factors such as incomplete information on the speciation of metals and other toxic substances present in the material, a lack of data on the bioavailability and interaction of contaminants, and the limited ability to test the contaminants in the most sensitive species that may be affected by such contamination.153 In the face of such uncertainties, EPA generally applies conservative default assumptions that will often overestimate risks and result in unnecessarily stringent and costly remediation.154 In other cases, however, the risk assessment may fail to recognize synergistic interactions between different contaminants in a mixture, resulting in underestimation of risk and inadequate health and ecological protection.155 Differential expression of stress-response and other genes known to be involved in toxicity response could be evaluated and used to rank and prioritize contaminated soils and sediments for cleanup. High throughput, automated DNA microarray systems are being developed for the direct, rapid, and affordable assessments of soil and sediment toxicity based on changes in gene expression levels.156 In one initial test of such a system, which have been described as "genosensors," the microarray system was capable of assaying 672 environmental samples for effects on the expression of 64 different stress response genes in a two-hour period.157 This represents a major advancement in the cost and speed of analyzing environmental samples.158 More importantly, [33 ELR 10085] this technology also offers a more accurate prediction of risk than conventional toxicology methods, because it directly assesses the toxicity of contaminated soils or sediments in a manner that eliminates critical uncertainties relating to contaminant bioavailability, speciation of metals, and interactive effect of contaminant mixtures.159 The major limitation of the genosensor is in selecting which genes to include in the microarray to provide the most accurate and comprehensive assay for toxicity that encompasses a variety of potentially affected species, but notwithstanding this issue rapid progress is being made in developing this technology and overcoming the remaining limitations.160 Finally, in addition to these potential screening applications of microarrays by regulatory agencies, industry will be able to use the technology to screen potential future products for toxicity. By providing an earlier and more specific biomarker of toxicity, DNA microarrays have the potential to create significant cost and time savings for product development by screening out potentially harmful products early in the developmental cycle.161 EPA traditionally uses reference doses (RfDs) or reference concentrations (RfCs) that are listed in the Agency's Integrated Risk Information Systems (IRIS) in making regulatory decisions for noncarcinogenic chemicals.162 An RfD or RfC is defined as "an estimate (with uncertainty spanning perhaps an order of magnitude)" of an ongoing exposure "to the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects during a lifetime."163 EPA attempts to set risk-based regulatory standards for noncarcinogens at a level of exposure that does not exceed the applicable RfD or RfC. In effect then, the RfC or RfD is used as a "safe" threshold value for noncarcinogenic chemicals, recognizing that there will always be some uncertainty about any such value. RfDs and RfCs are calculated by applying a series of uncertainty factors to the no observed adverse effect level (NOAEL),164 or in the absence of a NOAEL, the lowest observed adverse effect level (LOAEL).165 In the past, the terms no observed effect level (NOEL) and lowest observed effect level (LOEL) were used in many regulatory programs, but these terms were modified to LOAEL and NOAEL to require an effect to be "adverse" before it has regulatory significance.166 If studies show that a chemical induces gene expression changes at levels below the existing NOAEL or LOAEL, should these changes be considered "adverse" and used to establish a lower NOAEL or LOAEL, which in turn will mean a more stringent RfC or RfD? Industry groups are understandably concerned that gene expression changes that are not adequately validated as true markers of a toxic response will be used to establish more stringent reference values.167 Many changes in gene expression in response to chemical exposures will merely be an adaptive response of the cell or organism to exposures that help the organism maintain homeostasis, or represent a temporary aberration that would not normally progress to toxicity.168 Such adaptive responses will not be indicative of a true toxicological response. It will often be difficult to distinguish whether a particular response is adverse or adaptive.169 There is generally no bright line that can be used to differentiate adverse and adaptive responses, and often there is a continuum in which the same general category of response may be adverse in some circumstances and adaptive in others.170 Another complication is that a similar exposure may produce a response that is adverse in some individuals [33 ELR 10086] but not others based on various factors affecting susceptibility, including genetics, health, weight, other toxic exposures, age, gender, or nutritional status.171 While some attempts have been made to define when an effect is "adverse,"172 there is no single accepted definition.173 In a recent draft report on the process for determining RfDs and RfCs, EPA stated that "professional judgment is required to decide, on the basis of a thorough review of all available data and studies, whether any observed effect is adverse and how the results fit with what is known about the underlying mode of action."174 The draft report indicates that "biological significance" will be the critical factor in determining whether an effect is adverse, and proceeds to define biological significance as "the determination that the observed effect (a biochemical change, a functional impairment, or a pathological lesion) is likely to impair the performance or reduce the ability of an individual to function or to respond to additional challenge from the agent. Biological significance is also attributed to effects that are consistent with steps in a known mode of action."175 Under this limited available guidance, the determination of whether a particular gene expression change is "adverse" will require expert judgment on a case-by-case basis.176 Gene expression changes per se are unlikely to "impair the performance" of an individual, although they conceivably may be indicative in some cases of a reduced capability to accommodate additional exposures to the same or a similar toxic agent. The more relevant inquiry in most cases will be whether a specific gene expression change is consistent with a "known mode of action." This will require validated data showing that the particular gene expression change is a consistent biomarker for a known toxicological response. If EPA does determine that changes in gene expression is an "adverse effect" in a particular case, the next question is whether EPA should adjust the traditional uncertainty factors it uses to calculate the RfC or RfD based on this new critical adverse effect. The standard set of uncertainty factors that EPA applies to calculate an RfD or RfC do not take into account the severity of the adverse effect that defines the LOAEL.177 EPA has occasionally applied on a case-by-case basis a reduced overall uncertainty factor when the relevant "adverse effect" is of low severity, such as minor irritation lesions in the nasal cavity after inhalation of a chemical,178 but there is no general requirement for such an adjustment in the IRIS methodology. Should EPA reduce the uncertainty factors applied to calculate an RfD or RfC based on gene expression effects to compensate for the low severity of this "adverse" effect? There is a case to be made that the availability of a more sensitive test to detect a toxic response provided by microarrays should not necessarily result in more stringent standards, but rather should be used to only provide a more precise and certain assay for characterizing whether a toxicant does indeed cause a toxic response in humans. Under this view, EPA should compensate for the more sensitive test by reducing the otherwise applicable uncertainty factors to compensate for the low severity of the critical adverse effect. Again, this is a new issue for which there is currently no applicable guidance. In many cases, environmental risks are not discovered until they manifest in human disease or death. Microarray assays may provide an early warning of potentially dangerous exposures before adverse health effects occur by providing "a rapid means of assessing the bioavailability and potential toxicity of complex mixtures of chemicals released into the air and into groundwater."179 Pre-symptomatic detection of hazardous exposures would permit early intervention to monitor and treat affected persons in a more timely and effective manner, as well as to minimize further exposure to those and other persons. Such a surveillance program could [33 ELR 10087] be applied to individuals living or working near a polluting facility or hazardous waste site,180 or it could be applied to a cohort of individuals exposed to a potentially hazardous substance such as consumers using a particular household product that is suspected of toxicity.181 It could also be used to monitor citizens living near the site of an environmental accident such as TMI or Bhopal. A recent study demonstrated the potential of microarrays to provide real-time surveillance of potentially exposed individuals or populations. The study involved exposing peripheral blood lymphocytes from several different human donors to ionizing radiation while the cells grew in tissue culture.182 The study found that 48 genes were significantly up-regulated and 7 genes significantly down-regulated after radiation exposure.183 These changes in gene expression were reproducible, peaked at 24 hours after exposure, and still remained significantly above background levels 72 hours after exposure.184 Importantly, the quantitative response was very similar in cells from different donors, which greatly enhances the practical utility of such genes as potential markers of exposure across an exposed population, without having to account for individual differences in background levels of the marker.185 These findings suggest that it "may be possible to establish normal ranges for expression levels of these genes to distinguish irradiated individuals, whose expression levels of these genes would fall outside the normal range."186 Ongoing or targeted screening of an exposed population using microarrays could detect such abnormal gene expression profiles in individuals, facilitating both individualized intervention to assist those at-risk people and also population-wide risk assessment and risk management measures.187 Microarray technology may also be used to monitor pollutant effects on nonhuman organisms, such as aquatic species.188 For example, one recent study demonstrated that an estrogenic compound produced a characteristic pattern of changes in the expression of estrogen responsive genes in sheepshead minnows.189 The minnow can thus be used as a living sensor for the presence and effect of endocrine disruptive chemicals in coastal habitats.190 Another study showed that characteristic changes in gene expression in tadpoles can be detected from laboratory exposure to the herbicide acetochlor prior to development of overt morphological changes brought about by the endocrine disruption effect of the herbicide.191 This information could be used to provide a more sensitive and earlier indicator of potential toxicity in sensitive species, as well as to help understand the nature and mechanism of the toxicological response in particular species. Tester animal or plant species could be placed intentionally near a hazardous facility or environmental accident to monitor for changes in their gene expression, which could serve as a "sentinel" for nearby residents or natural species.192 Gene expression assays have several advantages over traditional toxicological endpoints for real-time surveillance of potentially at-risk populations. Microarrays have the potential to provide real-time, on-site estimates of both exposure and risk.193 In particular, it is possible, using high throughput gene expression screening to quantify the potential exposures of a large number of people in a quick and minimally intrusive manner.194 Another important advantage of microarrays for real-time surveillance is that microarrays provide a more sensitive and earlier indication of a potential risk than traditional methods.195 Of course, one of the inevitable consequences of using a more sensitive assay is that the results will require careful interpretation and the exercise of judgment by both regulators and companies to avoid false alarms while recognizing truly significant early toxicological responses. The potential for real-time surveillance provided by microarrays may trigger or raise questions about some existing regulatory requirements for product safety surveillance by manufacturers. For example, § 8(e) of TSCA requires the manufacturer of a substance or mixture to report to EPA information received "which reasonably supports [33 ELR 10088] the conclusion that such substance or mixture presents a substantial risk of injury to human health or the environment."196 If a manufacturer obtains data showing that one of its chemical products induces gene expression changes in animal studies, but there is no other indication of toxicity, will the company be required to report that data under § 8(e)? Is the requirement to report stronger if the gene expression changes are characteristic of a known mechanism of toxicity? EPA's guide for § 8(e) reporting indicates that reporting is generally only required for "serious" toxic effects, and thus effects of less certain significance such as organ weight change or in vitro genotoxicity test results need not be reported unless other evidence or factors show that the observed effect is indeed predictive of the potential for a more serious effect.197 While gene expression changes of unknown toxicological significance should clearly not be reportable under those criteria, changes that are signatures of known toxicological mechanisms present a closer question. On the one hand, the gene expression changes are in and of themselves not a "serious" toxicological effect, but because such changes have the potential to provide unprecedented acumen in predicting toxicity, there is at least an argument that § 8(e) reporting incorporates such a nonserious but informative effect.198 Similarly, § 6(a)(2) of the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) requires pesticide registrants to report "factual information regarding unreasonable adverse effects" associated with their products.199 Again, agency guidance on this reporting requirement is ambiguous on whether gene expression data alone can ever represent information on "unreasonable adverse effects." EPA's implementing regulations require that a pesticide registrant report any "information relevant to the assessment of risks or benefits" of a registered pesticide,"200 which is further defined to include "the results of a study of the toxicity of a pesticide to humans or other non-target domestic organisms if they show an adverse effect" not previously reported to EPA.201 Thus, the § 6(a)(2) reporting requirement for gene expression results will depend on whether such findings represent an "adverse effect," which as discussed above is not clear under existing definitions.202 Product manufacturers are thus likely to face significant uncertainties about the reporting of gene expression changes under both TSCA and FIFRA in the absence of more specific guidance provided by EPA. Setting Environmental Standards Several environmental statutes require risk-based standards that protect the public health against "adverse effects" with an adequate "margin of safety."203 For example, § 109 of the Clean Air Act requires EPA to establish national ambient air quality standards that protect the public health "with an adequate margin of safety."204 The legislative history of the statute indicates Congress' intent that EPA protect the public from "adverse effects" resulting from exposure to air pollutants,205 and one of the key issues EPA addresses in setting such standards is determining whether a particular response is an "adverse effect." In the most recent revision to the ozone air quality standard, for instance, EPA concluded that "transient and reversible" effects on lungs from ozone exposure were not "adverse effects."206 However, regulatory and judicial precedent hold that an "adverse effect" need not have clinical symptoms. When EPA first promulgated its ambient air quality standard for lead in 1978, the "adverse effect" on which the Agency based the standard was elevated erythrocytein protoporphyrin (EP) levels, a "subclinical" molecular change in the cell that may indicate an impairment of heme synthesis.207 EPA itself acknowledged that initial elevation of EP levels from exposure to lead "may not be a disease state or be seen as a clinically detectable decline in performance," but found that EP elevation becomes progressively more significant as an indicator of physiological response as lead exposure increases.208 EPA concluded that, "as with other subclinical manifestations of impaired function, it is a prudent public health practice to exercise corrective action prior to the appearance of clinical symptoms."209 EPA did not find that any amount of EP elevation that could be detected would qualify as an "adverse effect," but only elevations above a certain quantum that EPA concluded indicated that the EP elevation "has progressed to the extent that it should be regarded as an adverse health effect."210 In the case Lead Industries Ass'n v. EPA,211 the U.S. Court of Appeals for the D.C. Circuit upheld EPA's determination to base its standard on EP elevation, holding that EPA need not show that an effect caused by exposure to an air pollutant was "clearly harmful" to health. Instead, it was sufficient, according to the court, that the chemical changes relied on by EPA indicated that "lead has begun to affect one of the basic biological functions of the body."212 In the same lead rulemaking, EPA also considered whether it should base its standard on a different subclinical effect, the inhibition of the enzyme [delta]-aminolevulinic acid dehyratase ([delta]-ALAD) in red blood cells and other tissues.213 This enzyme catalyzes the formation of one of the components involved in the cellular synthesis of heme, and lead inhibits this enzyme at a concentration significantly below that which produces any other physiological or molecular effect from lead exposure, including EP elevation. EPA concluded that this subclinical effect was not "adverse," and thus should not be used as the basis for the standard, "because of the absence of evidence that there is an impairment of heme synthesis" at levels below which other adverse effects, i.e., EP elevation, occur.214 The evidence before EPA suggested that while lead reduced the activity of [delta]-ALAD at low-exposure levels, this reduced enzymatic activity did not correspond to any changes in the rate of heme synthesis, and thus resulted in no "functional impairment."215 If data showed that a criteria pollutant induced gene expression changes that were characteristic of a known toxicological profile at levels below the existing national ambient air quality standards (NAAQS), would this finding constitute an adequate "adverse effect" to require tightening of the NAAQS?216 The precedent established in the Lead Industries case suggests that an effect need not be clinically detectable or "clearly harmful" to be considered adverse, and that an effect can be adverse if it occurs solely at the molecular level. It therefore appears that at least some gene expression changes may be considered "adverse." On the other hand, EPA's own precedent, established in its decision that [delta]-ALAD alterations were not an adverse effect for purposes of setting the lead air quality standard, suggests that a gene expression change should only be considered adverse if it results in "functional impairment." Once again, there is likely to be considerable uncertainty and disagreement about whether gene expression changes are an "adverse effect," a reflection of the unique nature of toxicogenomic data. Gene expression data provides information much earlier in the disease process than toxicologists have traditionally used to identify toxicity, long before any symptoms appear. At the same time, gene expression changes may have powerful predictive value, providing more sensitivity, specificity, and information than much more obvious toxicological responses. It remains to be seen how this new category of data will impact regulatory standard setting given that the existing criteria and precedents for such standards were established in the pre-genomic era. Toxicogenomics: Caveats and Limitations Gene expression profiling using DNA microarrays, by providing a window to peer within the cell to observe the earliest molecular responses to toxic exposures, offers a tool of unprecedented power for understanding and predicting the body's response to toxic exposures.217 As such, this new technology will have numerous potential applications in toxic torts and environmental regulation, no doubt including many uses in addition to those discussed above. Some of these applications will be available in the very near future, and in some cases are available now, at least in "proof of concept" form, while other applications are further into the future, albeit still likely within the next decade. Scientists predict that "it is almost certain" that the widespread use of DNA microarrays will become routine and inexpensive in the near future.218 Nevertheless, many obstacles and uncertainties remain to be resolved before toxicogenomic data can be given widespread practical and legal effect outside of the research laboratory.219 In the words of one toxicological expert, microarrays [33 ELR 10090] are "going to revolutionize science. But the technology is in its infancy, so there are going to be some growing pains."220 To begin with, the toxicological significance of gene expression changes must be validated,221 which is not an easy undertaking given the rapid pace at which microarray technology is still evolving.222 Validation will involve evaluating the robustness and reproducibility of toxicogenomic assays between or across different laboratories, species, individuals, tissues, development stages, exposure levels, and exposure durations,223 all of which could potentially affect gene expression patterns.224 "Normal" gene expression is a dynamic and ever-changing condition.225 The smallest perturbations in the microenvironment of the cell, such as the position of a tissue culture dish in an incubator or the time of day in which the assay is performed, can cause perturbations in gene expression.226 Such perturbations may at least partly explain the significant interlaboratory variability in results that has been reported, as well as the observation that the same laboratory can produce significantly different results in experiments repeated just a few weeks apart.227 Given that each microarray use can involve assays of tens of thousands of different genes, "hundreds of false positives are almost guaranteed" from the widespread use of microarrays given the potential for background fluctuations in gene expression.228 A major challenge to the successful use of microarrays will therefore be the capability to distinguish background "noise" caused by fluctuations in background conditions from true, biologically significant cellular responses to toxic exposures. Data sets of background levels of gene expression in unexposed persons, and how these levels vary between individuals and in response to differences in location, health status, nutritional intake, lifestyle, time of day, and other potential modifiers are needed.229 These background data sets can then be compared to gene expression patterns in exposed persons to determine if observed changes are "real" or artifacts of other external or internal factors. One possible measure to minimize the number of false positives and signal noise from changes in the expression of genes with unknown significance is to monitor only a limited set of genes with established associations with toxic responses, while excluding genes that are highly sensitive to external perturbations.230 Of course, limiting the number of genes monitored runs the risk of missing relevant information, and may also lead to disagreement between laboratories about which genes should be included. Microarray analyses need not only distinguish "real" changes in gene expression from background fluctuations, but also need to discriminate between two types of changes in gene expression in response to toxic exposures. Some changes will be nothing more than the adaptive response of cells to external stimuli having no toxicological significance [33 ELR 10091] or increased risk, whereas other changes will truly represent the early stages of disease progression.231 One important factor to be considered is the extent to which gene expression changes have been correlated with "classical" observed effects of toxicity.232 Some experts have suggested that toxicogenomic data should only be used as a "hypothesis-generating tool" in the initial years, whereby it would be understood that microarray results may point to interesting possibilities that need to be confirmed by additional testing, but would not alone be used to establish any final conclusions regarding toxicity.233 Another inherent limitation of DNA microarrays is that they only measure changes in the expression of genes into mRNAs, but not the subsequent synthesis of proteins from that RNA. Proteins are the functional and structural units of the cell that are likely to be the specific molecular target for many toxic substances, and otherwise involved directly in the toxic response in other cases.234 While gene expression is a critical determinant of protein synthesis, it is not the only factor, as protein levels may also be affected by factors such as RNA and protein stability and turnover.235 The RNA:protein abundance ratio can vary over a range of at least tenfold for different RNAs within the cell.236 The related fields of proteomics and metabonomics attempt to measure changes in cellular proteins and metabolites, respectively, but to date it has been much more difficult to accurately measure these parameters than gene expression.237 In sum, because gene expression is one step removed from the biological interaction of a toxic chemical with the critical components of a cell, there are therefore limitations on how useful it can be in delineating toxic responses and mechanisms.238 There will ultimately be a need to standardize DNA microarrays, although premature standardization may carry its own risks by freezing a rapidly developing technology before it matures.239 Many different microarray formats (or "platforms"), methodologies, and content have been developed and utilized by commercial companies and individual laboratories, making difficult interlaboratory comparison and reconciliation of results.240 For example, different microarray platforms often contain different gene sets, guaranteeing variability in results. In addition, laboratories currently use different data analysis systems to analyze the massive quantity of data produced by microarrays, which is responsible for some of the interlaboratory variability in results.241 While these various limitations and challenges associated with microarrays are significant and in many cases formidable, there is no question that toxicogenomic data from microarrays are already contributing to our understanding of toxic substances, and will play an increasingly important role in evaluating toxicity in the future.242 It is critical to the successful deployment of this technology that scientific and regulatory bodies develop standards and guidelines for the appropriate use of microarray data. To that end, the NIEHS established the National Center for Toxicogenomics (NCT) in 2000 "to promote the evolution and coordinated use of gene expression technologies and to apply them to the toxicological effects in humans."243 This undertaking will likely include an effort to move toward the establishment of standardized procedures and data quality standards.244 The efforts of the NCT as well as regulatory agencies such as EPA must go beyond the use of toxicogenomic data in research applications, but must also anticipate and develop guidelines for the regulatory use of toxicogenomic data, which may also be indirectly relevant to toxic tort applications.245 EPA has recently taken the first steps in providing such regulatory guidance by issuing its Interim Policy on Genomics in the summer of 2002.246 This document conveys considerable enthusiasm for the potential usefulness of genomic data in risk assessment and regulatory decisions.247 At the same time, the EPA guidance expresses appropriate caution with respect to some of the limitations and uncertainties discussed above. For example, the Interim Policy states that while gene expression data may provide "valuable insights" for predicting toxicity of environmental stressors, it cautions that "the relationship between changes in gene expression and adverse effects are unclear at this time and may likely be difficult to elucidate."248 EPA states that it "expects that genomics data may be received, as supporting information for various assessment and regulatory purposes," and that "genomics data may be considered in decision-making at this time" although "these data alone are insufficient as a basis for decisions."249 The Interim Policy therefore necessarily walks a fine line. On one hand, EPA is encouraging the development and submission of genomics data by affirming that it will indeed consider such data in making decisions. On the other hand, EPA recognizes the existing limitations and uncertainties involved with toxicogenomic data, and provides an important backstop against inappropriate use of genomic data by taking the position that it will not make decisions at this time based solely on genomic data. In other words, gene expression changes must be connected to traditional toxicological effects before they can be used for regulatory purposes at the present time. Another important aspect of the Interim Policy is that it expressly provides that the agency will continue to monitor and participate in the development of this technology, and update its guidance accordingly.250 The use of the term "interim" policy conveys its transitional status. With such a rapidly developing technology as microarrays, it is critical that the policies of agencies such as EPA stay current with the evolving science, which requires that policies remain flexible and current. If there is one area where the Interim Policy falls short, it is the lack of guidance for industry on the relevance of microarray data for various reporting requirements such as the TSCA § 8(e) requirement to report "substantial risk" information or the FIFRA § 6(a)(2) requirement to report information on "unreasonable adverse effects" associated with pesticides.251 While the Interim Policy indicates that regulatory decisions will not need be based exclusively on gene expression data, it does not address the issue of whether gene expression data must be submitted under TSCA and FIFRA reporting requirements. Microarray data are unique relative to other toxicological information in that changes in gene expression in and of themselves are not a health problem, yet on the other hand are highly prognostic of the development of toxicity. Most of the existing reporting guidance and precedent on defining an "adverse effect" do not contemplate or "fit" such a powerful predictive technology, but rather assume that observed toxicological effects must be "serious" or "significant" before they can be considered predictive of toxicity. Without specific guidance on the implications of gene expression changes for reporting requirements, regulated parties and agency staff will face significant uncertainties with respect to whether some gene expression changes will trigger reporting requirements.252 While EPA may understandably be reluctant to "lock in" any specific policies at such an early stage of a rapidly developing technology, the concept of "fair notice"253 requires that the Agency provide advance notification if it intends to use toxicogenomic data in an enforcement context, such as penalizing a company for failure to submit "substantial risk" or "adverse effect" notifications based on gene expression changes. EPA might be able to provide more guidance for these and other questions, without unduly tying its hands with rigid regulatory requirements, by producing more detailed "points to consider" or "best practices" guidelines that provide specific guidance on how microarray data should be developed and used for regulatory purposes.254 An alternative approach for the interim might be for EPA to expressly provide a "safe harbor" for toxicogenomic data, in which the Agency encourages the development and submission of such data, but commits not to take enforcement action for the failure to submit gene expression data or otherwise use the data for enforcement purposes until the methodology has been adequately validated.255 Such an approach would be consistent with the recently enacted ICCVAM256 Authorization Act of 2000, which requires that "any new or revised acute or chronic test method" should be "determined to be valid for its proposed use" prior to being required or encouraged by a federal agency.257 In contrast to the role of EPA in the regulatory context, there is no expert agency that can provide guidance and [33 ELR 10093] oversee the appropriate introduction of toxicogenomic data in the toxic torts context. Although there are no cases reported to date involving the use of gene expression data, there has been increasing interest by practitioners in this subject area,258 and it is only a matter of time before trial lawyers and their experts seek to introduce toxicogenomic data for the many potential applications in tort cases. The high-stakes and one-shot dynamics of tort litigation will provide strong incentives for the use of all potentially helpful evidence, even if some uses may be perceived as premature or inappropriate by the scientific and regulatory communities.259 Lay judges and juries are unlikely to be in a position to screen the reliability and validity of toxicogenomic data on their own. Clear and carefully developed codes of practice or guidance documents produced by influential scientific and regulatory bodies are likely to be the best option for defending against inappropriate or premature use of toxicogenomic data in civil litigation and other nonregulatory contexts.260 For many years, environmental regulators and toxic tort fact finders have had to make their decisions about the risks of toxic substances under conditions of paralyzing uncertainty. Toxicogenomics offers a tool of unprecedented power to look within the black box of the cell and directly observe the earliest stages of the toxicological response with information that is both highly specific and sensitive. While the potential applications and benefits of toxicogenomics for both environmental regulation and toxic torts are immense, the use of this technology in such contexts is not without limitations and the need for caution. In particular, toxicogenomics has the potential to produce too much information—in that it has the potential to identify too many chemicals and products that are interacting with biological systems, and too many people that are experiencing gene expression changes as a result of exposures to environmental agents—to permit practical decisions and priority setting based on gene expression changes alone. Careful judgment and rigorous validation will be needed to discriminate those gene expression changes that warrant public health concern from those with no public health significance that merely reflect innocuous adaptive responses and normal fluctuations within dynamic cells. One thing that seems clear, however, is that environmental regulation and toxic tort litigation will both look very different 10 years from now, in large part due to the revolutionary capabilities and information provided by toxicogenomics. 1. See, e.g., P. Trinia Simmons & Christopher J. Portier, Toxicogenomics: The New Frontier in Risk Analysis, 23 CARCINOGENESIS 903, 903 (2002) ("the complete sequence of the human genome will cause a fundamental paradigm shift in the science of risk assessment"); Wendy Yap & David Rejeski, Environmental Policy in the Age of Genetics, ISSUES IN SCI. & TECH., Fall 1998, at 33. 2. Many other types of genetic data may be useful in environmental regulation and toxic torts, including deoxyribonucleic acid (DNA) adducts, chromosomal aberrations, DNA breakage studies, reporter gene assays, and mutational spectra associated with specific chemicals. See generally Stefano Bonassi & William W. Au, Biomarkers in Molecular Epidemiology Studies for Health Risk Prediction, 511 MUTATION RES. 73 (2002). These applications are outside the scope of this Article. 3. Emile F. Nuwaysir et al., Microarrays and Toxicology: The Advent of Toxicogenetics, 24 MOLECULAR CARCINOGENESIS 153, 158 (1999); Richard J. Albertini, Developing Sustainable Studies on Environmental Health, 510 MUTATION RES. 317, 323 (2001). Toxicogenetics therefore involves the toxicological implications of single genes, whereas the focus of toxicogenomics is on the entire genome. A "genome" refers to the complete set of genes contained within a cell. This same distinction between focusing on one or a few susceptibility genes (toxicogenetics) versus studying the expression of the whole genome (toxicogenomics) in response to exposure to toxic substances also applies in the related field of genomic approaches to pharmaceuticals. See Allen D. Roses, Pharmacogenetics, 10 HUM. MOLECULAR GENETICS 2261, 2261 (2001) ("Pharmacognetics is defined as the study of variability in drug responses attributed to hereditary factors in different populations. Pharmacogenomics is the determination and analysis of the genome (DNA) and its products (RNA and proteins) as they relate to drug response."). 4. Marilyn J. Aardema & James T. MacGregor, Toxicology and Genetic Toxicology in the New Era of "Toxicogenomics": Impact of "-Omics" Technologies, 499 MUTATION RES. 13, 15 (2002). 5. A small proportion of the genes in any cell are "turned on" or "expressed" in a given cell at any one time. Although different types of cells, e.g., skin, blood, nerve cells, within a body contain identical genetic information, they have very different functional and structural properties primarily because they have different subsets of genes that are expressed. One estimate is that approximately 25% of all genes are active (turned on) in a given cell type, and that on average about 5% of genes that are active in one cell type are different from the genes turned on in another cell type. Toby G. Rossman, Cloning Genes Whose Levels of Expression Are Altered by Metals: Implications for Human Health Research, 38 AM. J. IND. MED. 335, 335 (2000). In addition to these differences in gene expression between cell types, the gene expression in any one cell type varies over time in response to external stimuli. A gene is expressed by a process called transcription, in which a replicate of the functional DNA sequence of the gene is created (known as messenger RNA (mRNA)), which then moves from the cell nucleus into the cell cytoplasm to produce a protein, the primary functional and structural units of the cell. In addition to characterizing these gene expression changes, "toxicogenomics" also generally encompasses other types of data including profiling the proteins (proteomics) or metabolites (metabonomics) in a cell or tissue. See Aardema & MacGregor, supra note 4, at 14. 6. See Nuwaysir et al., supra note 3, at 153 ("Almost without exception, gene expression is altered during toxicity, as either a direct or indirect result of a toxicant exposure."); Spencer Farr & Robert T. Dunn, Concise Review: Gene Expression Applied to Toxicology, 50 TOXICOLOGICAL SCI. 1, 1 (1999) ("The fundamental assumption of toxicogenetics is that there are no toxicologically relevant outcomes in vitro or in vivo, with the possible exception of rapid necrosis, that do not require differential gene expression."); Russell S. Thomas et al., Identification of Toxicologically Predictive Gene Sets Using cDNA Microarrays, 60 MOLECULAR PHARMACOLOGY 1189, 1189-90 (2001) ("Toxicity is commonly manifested as inflammation, proliferation, apoptosis, necrosis, and/or cellular differentiation. All of these toxic endpoints are intimately linked to specific alterations in gene expression."); Albertini, supra note 3, at 321 ("It has long been known that cells almost always respond to noxious stimuli by altering gene expression."). 7. See Christine Debouck & Peter N. Goodfellow, DNA Microarrays in Drug Discovery and Development, 21 (Suppl.) NATURE GENETICS 48, 49 (1999). 8. See Hisham K. Hamadeh & Cynthia A. Afshari, Gene Chips and Functional Genomics, 88 AM. SCI. 508, 510-11 (2000); Nuwaysir et al., supra note 3, at 153-54; Stephen H. Friend & Roland B. Stoughton, The Magic of Microarrays, SCI. AM., Feb. 2002, at 44, 46-47. 9. Nuwaysir et al., supra note 3, at 154; Timothy J. Aitman, DNA Microarrays in Medical Practice, 323 BRIT. MED. J. 611, 612 (2001). These types of microarrays are particularly useful for detecting small changes in DNA sequences, such as point mutations in a gene, because it is possible to synthesize many closely related sequences that differ by only a single base pair. Microarrays can generally be used both for characterizing gene expression and identifying variations in a DNA sequence of genes independent of their gene expression. In the former application, the mRNA is collected from the cell and hybridized to the microarray as described infra, while in the latter the cell's DNA is itself collected and hybridized to the microarray. 10. See Friend & Stoughton, supra note 8, at 46. 11. Affymetrix, News Release: Affymetrix Launches First Commercial Human DNA Array to Use Draft of Human Genome, Jan. 21, 2002, at http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=AFFX&script=410&layout=-6&item_id=248283 (last visited Aug. 29, 2002). 12. There are other important medical and public health applications of DNA microarrays in addition to toxicogenomics, including differentiating similar appearing tumors with respect to prognosis and treatment based on gene expression patterns, and rapidly identifying potentially dangerous microorganisms in the context of food safety, biowarfare defense, and other applications. See, e.g., Edward K. Lobenhofer et al., Progress in the Application of DNA Microarrays, 109 ENVTL. HEALTH PERSP. 887 (2001); Charles M. Perou et al., Molecular Portraits of Human Breast Tumours, 406 NATURE 747 (2000); Friend & Stoughton, supra note 8, at 44-53; Aitman, supra note 9. 13. See Hamadeh & Afshari, supra note 8, at 509. 14. Jennifer Medlin, Array of Hope for Gene Technology, 109 ENVTL. HEALTH PERSP. A34, A35-36 (2001). 15. Id. at A36; Hisham K. Hamadeh et al., An Overview of Toxicogenomics, 4 CURR. ISSUES MOLECULAR BIOL. 45, 46 (2002). One frequently used technique is to compare gene expression in two samples, e.g., control versus treatment cells, by making copies of the mRNA isolated from the two samples using modified nucleotides carrying flourescent tags. For one sample, a tag called Cy3 which fluoresces green could be used to synthesize the cDNA from the control, i.e., untreated, cells while Cy5 which fluoresces red is used to produce the cDNA from the treated cells. The fluorescently labeled sequences from the control and treated cells are then mixed together and hybridized to the DNA microarray and scanned by a laser to produce flourescent patterns. If a gene is expressed in equal amounts in both the control and treatment samples, then the spot on the microarray corresponding to that gene will fluoresce yellow. If a gene was expressed more in the treated cells than the control cells, i.e., "up-regulated," the spot will be red. If the gene is expressed at relatively lower levels in the treated cells, i.e., "down-regulated," the spot will be green. The intensity of the signal will provide a quantitative estimate of the extent by which a particular gene is up- or down-regulated in the treated cells. Id.; see also Hisham K. Hamadeh et al., Discovery in Toxicology: Mediation by Gene Expression Array Technology, 15 J. BIOCHEM. MOLECULAR TOXICOLOGY 231, 232 (2001). 16. See W.D. Pennie & I. Kimber, Toxicogenomics; Transcript Profiling and Potential Application to Chemical Allergy, 16 TOXICOLOGY IN VITRO 319, 320 (2002); Hamadeh & Afshari, supra note 8, at 509. 17. See Hamadeh & Afshari, supra note 8, at 509 ("Traditional assays measure RNA transcripts from one gene at a time over a three-day period. Gene chips can measure transcripts from thousands of genes in a single afternoon."). As one scientific review of this technology commented, "a global analysis of gene expression has the potential to provide a more comprehensive view of toxicity than has been possible previously, since toxicity generally involves change not only in a single or few genes but rather is a cascade of gene interactions." Aardema & MacGregor, supra note 4, at 14. See also Hisham K. Hamadeh et al., Gene Expression Analysis Reveals Chemical-Specific Profiles, 67 TOXICOLOGICAL SCI. 219, 219 (2002). Gene expression signatures for identifying particular classes of toxicants cannot be defined using classical methods where genes are investigated individually for potential association to chemical exposure. This is because the most highly characterized chemical-responsive genes, such as genes encoding proteins or enzymes that regulate metabolism, tend to be frequently modulated by many compounds, and therefore do not provide a solid footing for providing specificity for distinguishing multiple classes. 18. For example, most of the differential gene expression in cultured liver cells exposed to ethanol involve down-regulation, whereas similar treatment with carbon tetrachloride results in both up-regulation and down-regulation of different genes. H.M. Harries et al., The Use of Genomics Technology to Investigate Gene Expression Changes in Cultured Human Liver Cells, 15 TOXICOLOGY IN VITRO 399, 401-02 (2001). 19. Hamadeh et al., supra note 17, at 219 ("DNA microarrays enable the study of levels of expression of thousands of genes at the mRNA level. The concerted expression pattern across those genes constitutes the expression profile of a compound at a certain dose and time."); Aardema & McGregor, supra note 4, at 14; 20. See William E. Bishop et al., The Genomic Revolution: What Does It Mean for Risk Assessment?, 21 RISK ANALYSIS 983, 986 (2001). 21. For example, 24-hour treatment of cultured liver cells with ethanol produces changes in the expression of approximately 85 different genes. Harries et al., supra note 18, at 402. Dioxin appears to change the expression of approximately 300 genes by a factor of 2 or more. Felix W. Frueh et al., Use of cDNA Microarrays to Analyze Dioxin-Induced Changes in Human Liver Gene Expression, 122 TOXICOLOGY LETTER 189, 198-99 (2001). 22. Nuwaysir et al., supra note 3, at 153. See also Hamadeh et al., supra note 15, at 231 (DNA microarrays "provide a revolutionary platform to perform genome-wide gene expression analyses through comparison of virtually any two biological samples"); Pennie & Kimber, supra note 16, at 319 (microarrays "represent nothing short of a revolution in our ability to characterize simultaneously an unprecedented number of biological endpoints"). 23. See Harries et al., supra note 18, at 399 ("Changes in gene expression often provide a far more sensitive, characteristic and measurable endpoint than the toxicity itself."); Hamadeh et al., supra note 15, at 231; Nuwaysir et al., supra note 3, at 154-55. 24. See Farr & Dunn, supra note 6, at 2. There are generally less than 100 toxicological outcomes evaluated in human or animal studies, whereas hypothetically there are an estimated 10<70"> different gene expression patterns per cell, although the number of toxicologically relevant gene expression patterns actually observed is obviously much smaller than the possible number of patterns. Id. 25. Charles P. Rodi et al., Revolution Through Genomics in Investigative and Discovery Toxicology, 27 TOXICOLOGICAL PATHOLOGY 107, 109 (1999). 26. Pennie & Kimber, supra note 16, at 319 (microarrays can "characterize simultaneously an unprecedented number of biological endpoints"). 27. See Farr & Dunn, supra note 6, at 1. 28. See Rodi et al., supra note 25, at 107. 29. E.g., Thomas et al., supra note 6, at 1193 (a microarray analysis using only 12 diagnostic genes was able to correctly classify the toxicological class of 24 toxicants with 100% predictive accuracy); Michael E. Burczynski et al., Toxicogenomics-Based Discrimination of Toxic Mechanisms in HepG2 Human Heatoma Cells, 58 TOXICOLOGICAL SCI. 399 (2000) (gene expression patterns used to successfully discriminate compounds based on toxic mechanism); Jeffrey F. Waring et al., Microarray Analysis of Hepatotoxins in Vitro Reveals a Correlation Between Gene Expression Profiles and Mechanisms of Toxicity, 120 TOXICOLOGY LETTER 359 (2001) (microarray analysis of the effects of 15 known liver toxins on cultured cells showed that each compound produced a unique signature, and compounds with similar toxic mechanisms had the same distinct patterns ("clusters") of gene expression); Matthew Bartosiewicz et al., Applications of Gene Arrays in Environmental Toxicology: Fingerprints of Gene Regulation Associated With Calcium Chloride, Benzoapyrene, and Trichloroethylene, 109 ENVTL. HEALTH PERSP. 71, 73-74 (2001) (three important environmental contaminants belonging to different chemical classes produced unique patterns of gene expression in mice); Hamadeh et al., supra note 17, at 219, 228-29 (structurally unrelated compounds from the same general class of toxicants produce similar, but distinguishable, gene expression profiles); Hisham K. Hamadeh et al., Prediction of Compound Signature Using High-Density Gene Expression Profiling, 67 TOXICOLOGICAL SCI. 232 (2002) (microarray analysis of gene expression profiles from liver samples of chemical-treated rats was able to correctly predict the toxicological mechanism of 22 of 23 blinded chemicals); Christopher M.L.S. Bouton et al., Microarray Analysis of Differential Gene Expression in Lead-Exposed Astrocytes, 176 TOXICOLOGY & APPLIED PHARMACOLOGY 34, 44-45 (2001) (microarrays used to distinguish cells exposed to lead from nontoxicant exposed cells); Steven J. Bulera et al., RNA Expression in the Early Characterization of Hepatotoxicants in Wistar Rats by High-Density DNA Microarrays, 33 HEPATOLOGY 1239 (2001) (microarray analysis of mRNA able to successfully distinguish six different liver toxicants). See also Raymond W. Tennant, The National Center for Toxicogenomics: Using New Technologies to Inform Mechanistic Toxicology, 110 ENVTL. HEALTH PERSP. A8, A8 (2002); Nuwaysir et al., supra note 3, at 154-56. 30. See Hamadeh et al., supra note 17, at 228-29; supra notes 139-40 and accompanying text. 31. See supra notes 118-19 and accompanying text. 32. See supra notes 105-15 and accompanying text. 33. Aardema & MacGregor, supra note 4, at 15, 22. See also David Stipp, Gene Chip Breakthrough, FORTUNE, Mar. 31, 1997, at 56 ("Microprocessors have reshaped our economy, spawned vast fortunes and changed the way we live. Gene chips could be even bigger."). 34. Elizabeth Pennisi, Recharged Field's Rallying Cry: Gene Chips for All Organisms, 297 SCIENCE 1985, 1986 (2002). 35. One measure of the rapid development of toxicogenomics is the number of published scientific studies containing the term "microarray" indexed in the National Library of Medicine's "PubMed" database, at http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed). Only 2 published studies mentioned microarrays in 1997, but this quickly increased to 21 studies in 1998, 83 in 1999, 285 in 2000, 805 in 2001, and as of August 30, 2002, 874 in the first eight months of 2002. 36. See, e.g., Troyen A. Brennan, Causal Chains and Statistical Links: the Role of Scientific Uncertainty in Hazardous-Substance Litigation, 73 CORNELL L. REV. 469, 469 (1988) (proving causation is the "paramount obstacle to just resolution of tort claims based on injury from toxic substances"); Steve Gold, Causation in Toxic Torts: Burdens of Proof, Standards of Persuasion, and Statistical Evidence, 96 YALE L.J. 376, 376 (1986) ("Proving the cause of injuries that remain latent for years, are associated with diverse risk factors, and occur at background levels even without any apparent cause, is the 'central problem' for toxic tort plaintiffs.") (footnotes omitted). 37. See, e.g., Allen v. Pennsylvania Eng'g Corp., 102 F.3d 194, 199 (5th Cir. 1996) ("Scientific knowledge of the harmful level of exposure to a chemical, plus knowledge that the plaintiff was exposed to such quantities, are minimal facts necessary to sustain the plaintiffs' burden in a toxic tort case."). 38. See generally Susan R. Poulter, Science and Toxic Torts: Is There a Rational Solution to the Problem of Causation?, 7 HIGH TECH. L.J. 189, 236-41 (1992). 39. E.g., Wright v. Williamette Indus., Inc., 91 F.3d 1105, 1107 (8th Cir. 1996) ("at a minimum, we think that there must be evidence from which the fact finder can conclude that the plaintiff was exposed to levels of that agent that are known to cause the kind of harm that the plaintiff claims to have suffered"); Mitchell v. Gencorp Inc., 165 F.3d 778, 781 (10th Cir. 1999) ("guesses, even if educated, are insufficient to prove the level of exposure in a toxic tort case"); Allen, 102 F.3d at 197 ("Scientific knowledge of the harmful level of exposure to a chemical, plus knowledge that the plaintiff was exposed to such quantities, are minimal facts necessary to sustain the plaintiffs' burden in a toxic tort case."). But see Donaldson v. Central Ill. Pub. Serv. Co., 767 N.E.2d 314, 332 (Ill. 2002) (plaintiffs are "not required to show the exact amount of exposure" in cases involving environmental toxics because such cases "do not afford litigants the opportunity to specify with such certainty the exact level of and dose of exposure"). 40. See Gary E. Marchant, Toxicogenomics and Toxic Torts, 20 TRENDS IN BIOTECH. 329, 330 (2002). 41. This inquiry will involve evaluating both the specificity and sensitivity of the gene expression assay. Specificity refers to the capability of the assay to identify only exposures that are indeed associated with the toxic agent of interest, i.e., avoiding false positives, while sensitivity refers to the capability of the assay to detect all exposures to the relevant toxic agent, i.e., avoiding false negatives. See generally Ralph R. Cook, The Importance of Test Validity and Predictive Values to Screening Programs, 41 JURIMETRICS 111, 112 (2000). 42. See Tennant, supra note 29, at A9 ("it is important to determine whether or not serum/blood cells can be used as an alternative to specific target organ tissue"). The NIEHS is currently testing the hypothesis that blood cells can serve as a surrogate for tissue-specific chemical effects. Id. See also John C. Rockett et al., DNA Arrays to Monitor Gene Expression in Rat Blood and Uterus Following 17B-Estradiol Exposure; Biomonitoring Environmental Effects Using Surrogate Tissues, 69 TOXICOLOGICAL SCI. 49 (2002) (pilot study finding that peripheral blood lymphocytes can serve as adequate surrogate for uterus for monitoring gene expression changes in response to endocrine disrupting chemicals). 43. If gene expression samples are available for an individual before and after exposure, then comparison of those two results might provide a valid measure of exposure regardless of any unique genetic characteristics of the individual affecting gene expression levels. In the absence of pre-exposure data for an individual, however, the effect of individual susceptibility would likely have to be considered to quantify exposure if such susceptibilities affect the relationship between the quantum of exposure and the magnitude of the gene expression changes. However, one study examining gene expression changes after exposure to ionizing radiation found only slight variation in gene expression levels in peripheral blood lymphocytes from different donors, suggesting that interindividual genetic variability had a relatively minor effect on gene expression changes for at least this response. Sally A. Amundson et al., Identification of Potential mRNA Biomarkers in Peripheral Blood Lymphocytes for Human Exposure to Ionizing Radiation, 154 RADIATION RES. 342, 343, 346 (2000). 44. See Bonassi & Au, supra note 2, at 76 ("For many types of biomarkers the most important consideration is the stability of the biomarkers with respect to time after exposure."); Carol J. Henry et al., Use of Genomics in Toxicology and Epidemiology: Findings and Recommendations of a Workshop, 110 ENVTL. HEALTH PERSP. 1047, 1049 (2002) ("An additional challenge [of toxicogenomic methods] is to examine gene expression at the time period that is relevant to the health outcome of interest."). 45. Most toxicogenomic data produced to date only examines the short-term effect of toxic exposure to cells isolated in tissue culture or rodents exposed for 24 to 72 hours. But see Hamadeh et al., supra note 17, at 225, 227 (finding quantitative and qualitative differences in gene expression between one-time exposure to toxicant and after two weeks of daily exposure to same toxicant). 46. One study evaluated the time course of gene expression changes in human peripheral blood lymphocytes grown in tissue culture and exposed to a single dose of ionizing radiation. Amundson et al., supra note 43, at 342-46. The maximal response for most of the marker genes occurred 24 hours after irradiation, and then gradually declined, but remained significantly above background levels at 72 hours. Id. at 344. 47. See Henry et al., supra note 44, at 1049 ("Recently examined gene expression products may not be relevant to disease that have long latency periods."). 48. See Rossman, supra note 5, at 336 (suggesting that some changes in gene expression may be permanent). 49. In re TMI Litig., 193 F.3d 613, 622 (3d Cir. 1999), cert. denied, 120 S.Ct. 2238 (2000). 50. Dicentric chromosomes contain two rather than the normal one centromere. A centromere is the structural center of a chromosome at which the left and right chromosome arms are joined. Dicentric chromosomes are formed when chromosomes are broken by an agent such as radiation and then rejoin in an aberrant manner. Id. at 688 n.24. 51. Id. at 688. 52. Id. at 690. 53. Id. at 692. 54. Id. 55. The potential of genomic data to verify exposure to specific products is provided by the recent actions of a prominent personal injury law firm specializing in foodborne illness. The firm has posted on the Internet at http://www.fsis-pfge.org the genetic fingerprints of pathogenic E.Coli 0157:H7 strains associated with recalls of ground beef. Individuals who contract food poisoning from E. Coli 0157:H7 can check if the genetic fingerprints of the strain that has infected them matches one of the genetic profiles from the contaminated meet. Given the variability between different bacterial strains, such a match would provide strong evidence of causation in a personal injury case. Allison Beers, Marler Clark Posting E. Coli Genetic Fingerprints From Recalls, FOOD CHEM. NEWS, June 3, 2002, at 1. 56. See, e.g., Sterling v. Velsicol Chem. Corp., 855 F.2d 1188, 1200, 19 ELR 20404, 20408 (6th Cir. 1988); Raynor v. Merrell Pharmaceuticals, Inc., 104 F.3d 1371, 1376 (D.C. Cir. 1997); Merrell Dow Pharmaceuticals, Inc. v. Havner, 953 S.W.2d 706, 714 (Tex. 1997). 57. Havner, 953 S.W.2d at 714 ("General causation is whether a substance is capable of causing a particular injury or injury in the general population …."). 58. See, e.g., General Elec. Co. v. Joiner, 522 U.S. 136, 144, 28 ELR 20227, 20228 (1997) (study finding that polychlorinated biphenyls (PCBs) cause alveologenic adenomas at high concentrations in mice cannot be used to show that PCBs caused plaintiff's small-cell carcinoma); Allen v. Pennsylvania Eng'g Corp., 102 F.3d 194, 197 (5th Cir. 1996) ("Evidence has been found that suggests a connection between [ethylene oxide] exposure and human lymphatic and hematopoietic cancers, but this is not probative on the causation of brain cancer."); Christophersen v. Allied-Signal Corp., 939 F.2d 1106, 1115-16 (5th Cir. 1991) (evidence that defendant's chemicals may cause small-cell carcinoma of the lung inadmissible to show that same chemicals may cause small-cell carcinoma of the colon). See generally Poulter, supra note 38, at 227 ("it is not uncommon for plaintiffs' experts to assert that evidence that a substance causes any cancer is evidence that it can and has caused other cancers. Although substances that are discovered to cause one type of cancer may cause other types of cancer as well, that possibility does not permit a prediction of what those other cancers, if any, are likely to be.") (footnotes omitted). But see Donaldson v. Central Ill. Pub. Serv. Co., 767 N.E.2d 314, 327 (Ill. 2002) (Illinois permits extrapolation between similar but not identical cause and effect relationships in the "limited instances" where science unable to directly establish cause of disease). 59. E.g., Glastetter v. Novartis Pharmaceuticals Corp., 252 F.3d 986, 990 (8th Cir. 2001) ("Even minor deviations in molecular structure can radically change a particular substance's properties and propensities."); Amorgianos v. National Rd. Passenger Corp., 137 F. Supp. 2d 147, 190 (E.D.N.Y. 2001) (too great of an "analytical gap" between plaintiff's short-term exposure to xylene and studies relied on by plaintiff's experts involving longer term exposure to other solvents); Lynch v. Merrell-National Labs., 830 F.2d 1190 (1st Cir. 1987) (rejecting expert's reliance on toxicological studies with "analogous chemicals" to show causation). 60. ENVIRONMENTAL DEFENSE FUND, TOXIC IGNORANCE (1997). 61. See Richard J. Pierce Jr., Causation in Government Regulation and Toxic Torts, 76 WASH. U. L.Q. 1307, 1308 (1998) ("There are thousands of regulated substances that rarely, if ever, could be the subject of a successful tort action. The available evidence is sufficient to support a finding that they probably cause nontrivial injuries of some types, but it is insufficient to support a finding that they probably caused any particular injury."). 62. For example, in Christophersen v. Allied-Signal Corp., 939 F.2d 1106 (5th Cir. 1991) the plaintiff argued that evidence showing defendant's chemicals can cause small-cell carcinoma of the lung may be relevant to show chemicals can also cause small-cell carcinoma of the colon based "on the nature of the biochemical reaction that results in the development of small cell carcinoma." Id. at 1116 n.10. The court ruled that the plaintiff's expert had failed to adequately substantiate this connection, but microarray data may be able to show that a particular type of cancer, e.g., small-cell carcinoma, in two different tissues is (or is not) caused by the same molecular events, and thus perhaps the same chemical agent. 63. 25 S.W.2d 280 (Tex. Ct. App. 2000). 64. Id. at 288. 65. Id. 66. Id. at 290-91. 67. See supra note 9. 68. See Merrell Dow Pharmaceuticals, Inc. v. Havner, 953 S.W.2d 706, 714 (Tex. 1997) ("specific causation is whether a substance caused a particular individual's injury"). 69. Id. at 715. See also In re Agent Orange Prod. Liab. Litig., 597 F. Supp. 740, 834 (E.D.N.Y. 1984) ("it may be impossible to pinpoint which particular person's cancer would have occurred naturally and which would not have occurred but for exposure to the substance"); Gold, supra note 36, at 379 ("Cancers and mutations provide no physical evidence of the inducing agent, so direct observation of individual plaintiffs provides little or no evidence of causation in many instances."). 70. For a few diseases, such as mesothelioma caused by asbestos or clear cell adenocarcinoma caused by the drug DES, almost every case of disease is caused by only one known cause, in which case the specific causation inquiry becomes elementary. Such examples of "signature" diseases are rare. See Daniel A. Farber, Toxic Causation, 71 MINN. L. REV. 1219, 1251-52 (1987). 71. E.g., Daubert v. Merrell Dow Pharmaceuticals, Inc., 43 F.3d 1311, 1321, 25 ELR 20856, 20860 (9th Cir. 1995); Hall v. Baxter Health-care Corp., 947 F. Supp. 1387, 1403-04 (D. Or. 1996); Allison v. McGhan Med. Corp., 184 F.3d 1300, 1315 (11th Cir. 1999). Other courts have rejected the requirement to show a doubling of relative risk. See generally Russellyn S. Carruth & Bernard D. Goldstein, Relative Risk Greater Than Two in Proof of Causation in Toxic Tort Litigation, 41 JURIMETRICS 195 (2001). 72. See Frederica P. Perera, Environment and Cancer: Who Are Susceptible?, 278 SCIENCE 1068, 1072 (1997) ("In epidemiology, it has been difficult to detect relative risks of 1.5 or even 2.0."); Gary Taubes, Epidemiology Faces Its Limits, 269 SCIENCE 164, 165 (1995) (noting only a handful of carcinogenic agents have produced relative risk greater than two in epidemiology studies). 73. In re Agent Orange Prod. Liab., 597 F. Supp. at 835 (describing (but not adopting) the requirement for particularistic evidence in addition to a relative risk of at least two as the "strong" version of the preponderance rule); Michael Dore, A Commentary on the Use of Epidemiological Evidence in Demonstrating Cause-in-Fact, 7 HARV. ENVTL. L. REV. 429, 434 (1983) (epidemiological evidence alone cannot establish causation). 74. See Glastetter v. Novartis Pharmaceuticals Corp., 252 F.3d 986, 989 (8th Cir. 2001): In performing a differential diagnosis, a physician begins by "ruling in" all scientifically plausible causes of the plaintiff's injury. The physician then "rules out" the least plausible causes of injury until the most likely cause remains. The final result of a differential diagnosis is the expert's conclusion that a defendant's product caused (or did not cause) the plaintiff's injury. 75. See Joseph Sanders & Julie Machal-Fulks, The Admissibility of Differential Diagnosis Testimony to Prove Causation in Toxic Tort Cases: The Interplay of Adjective and Substantive Law, 64 LAW & CONTEMP, PROBS, 107, 1201-29 (2001) (reviewing case law); Margaret A. Berger, The Supreme Court's Trilogy on the Admissibility of Expert Evidence, in FEDERAL JUDICIAL CENTER, REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 9, 34 (2d. ed. 2000) ("Judges disagree on whether a physician relying on the methodology of clinical medicine can provide adequate proof of causation in a toxic tort action."). 76. E.g., Westberry v. Gislaved Gummi AB, 178 F.3d 257, 262-66 (4th Cir. 1999); Turner v. Iowa Fire Equip. Co., 229 F.3d 1202, 1208 (8th Cir. 2000); McCullock v. H.B. Fuller, 61 F.3d 1038, 1043-44 (2d Cir. 1995). 77. E.g., Daubert v. Merrell Dow Pharmaceuticals, Inc., 43 F.3d 1311, 1319, 25 ELR 20856, 20859 (9th Cir. 1995); Glastetter, 252 F.3d at 989-92; Meister v. Medical Eng'g Corp., 267 F.3d 1123, 1129, 1131 (D.C. Cir. 2001). 78. See, e.g., Waring et al., supra note 29, at 367. See also generally supra note 29 and accompanying text. 79. See supra note 9. 80. S. Perwez Hussain et al., Tumor Suppressor Genes: At The Crossroads of Molecular Carcinogenesis, Molecular Epidemiology, and Human Risk Assessment, 34 LUNG CANCER S7 (2001); Ian C. Semenza & Lisa H. Weasel, Molecular Epidemiology in Environmental Health: The Potential of Tumor Suppressor Gene p53 as a Biomarker, 105 (Suppl. 1) ENVTL. HEALTH PERSP. 155, 155-56 (1997). 81. See Gary Marchant, Genetics and Toxic Torts, 31 SETON HALL L. REV. 949, 971-72 (2001). 82. See Daubert, 43 F.3d at 1320 n.13, 25 ELR at 20860 n.13 ("unfairness is inevitable when our tools for detecting causation are imperfect and we must rely on probabilities rather than more direct proof"); In re Agent Orange Prod. Liab. Litig., 597 F.Supp. 740, 836 (E.D.N.Y. 1984): There would appear to be little harm in retaining the requirement for "particularistic" evidence of causation in sporadic accident cases since such evidence is almost always available in such litigation. In mass exposure cases, however, where the chance that there would be particularistic evidence is in most cases quite small …. 83. Similarly, there is no general causation requirement in most traumatic injury cases because the general propensity of the technology or action involved is beyond dispute, and the only contested issue is whether it did cause the injury in the specific case. See AMERICAN LAW INSTITUTE, RESTATEMENT OF THE LAW TORTS: LIABILITY FOR PHYSICAL HARM (BASIC PRINCIPLES), TENTATIVE DRAFT No. 2 § 28, at 102 (2002) ("In cases involving traumatic injuries, such as a broken bone following an automobile accident, the absence of other causal sets and better understanding of the causal mechanisms involved moots the necessity for independent proof of general causation beyond the 'specific causation' evidence in the case."). 84. See, e.g., Metro-North Commuter R.R. Co. v. Buckley, 521 U.S. 424, 433 (1997) (courts generally deny recovery for latent risks because of policy considerations including: (i) preventing defendants from being subjected to "unlimited and unpredictable liability"; (ii) protecting courts from having to sift through meritorious versus frivolous claims; and (iii) avoiding a "flood of comparatively unimportant claims"). 85. E.g., Adams v. Johns-Manville Sales Corp., 783 F.2d 589, 591-93 (5th Cir. 1986); Anderson v. W.R. Grace & Co., 628 F. Supp. 1219, 1226-27, 16 ELR 20577, 20579-80 (D. Mass. 1986). 86. E.g., Ayers v. Township of Jackson, 525 A.2d 287, 308, 17 ELR 20858, 20862 (N.J. 1987) (rejecting claim for unquantified enhanced risk of disease because of speculative nature of unquantified risk); Abuan v. General Elec. Co., 3 F.3d 329, 334 (9th Cir. 1993) (recovery for increased risk only where plaintiff shows that toxic exposure will more likely than not result in disease); Gideon v. Johns-Manville Sales Corp., 761 F.2d 1137-38 (5th Cir. 1985) (increased risk of cancer must be more likely than not to occur for claim to be recognized). 87. See, e.g., Brafford v. Susquchanna Corp., 586 F. Supp. 14, 18 (D. Col. 1984) ("the inability to precisely quantify the extent of present damage to the chromosomes is a function of medical technology's inability to make such a measure"); Andrew R. Klein, A Model for Enhanced Risk Recovery in Tort, 56 WASH. & LEE L. REV. 1173, 1179 (1999) (threshold requirements imposed by courts create "a nearly insurmountable barrier for enhanced risk plaintiffs"). 88. For example, in Anderson, 628 F. Supp. at 1226-27, 16 ELR at 20579-80, the court required plaintiffs seeking to recover for emotional distress associated with an increased risk from toxic exposure to demonstrate a present injury "manifested by objective symptomatology," and held that subcellular injuries could meet this standard provided they were "objectively evidenced." See also Brafford, 586 F. Supp. at 17-18 (denying summary judgment against plaintiff who relied on an inference that he must have incurred subcellular chromosomal damage from radiation exposure); Bryson v. Pillsbury Co., 573 N.W.2d 718, 720-21 (Minn. Ct. App. 1998) (asymptomatic, subcellular injury may constitute a legally recognized present injury). 89. See, e.g., Dodge v. Cotter Corp., 203 F.3d 1190, 1202 (10th Cir. 2000) (requiring evidence of "a chronic objective condition caused by their increased risk of developing cancer" to permit recovery for emotional distress damages); In re Hawaii Fed. Asbestos Cases, 734 F. Supp. 1563, 1567 (D. Haw. 1990) (requiring "an objectively verifiable functional impairment"); Schweitzer v. Consolidated Rail Corp., 758 F.2d 936, 942 (3d Cir. 1985) (subclincial injury insufficient for recovery). 90. Geoffrey C. Hazard, The Futures Problem, 148 U. PA. L. REV. 1901, 1901 (2000) ("Perhaps the most difficult problem in addressing mass torts is that of future claimants."); Richard W. Wright, Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts, 73 IOWA L. REV. 1001, 1067 (1988) (liability for risk exposure is "the most problematic area of current tort practice"). 91. For example, approximately one-third of all humans will contract cancer at some point in their lives, and every person incurs numerous mutations throughout their lives that would eventually progress to cancer if the person lived indefinitely. ROBERT A. WEINBERG, ONE RENEGADE CELL: HOW CANCER BEGINS 156 (1998) ("given enough time cancer will strike every human body"); Donald T. Ramsey, The Trigger of Coverage for Cancer: When Does Genetic Mutation Become "Bodily Injury, Sickness, or Disease?," 41 SANTA CLARA L. REV. 293, 298 (2001) ("even people who die from some other cause, before they can develop cancer, carry many thousands of cells bearing mutations to key genes throughout most of their existence"). Except for those caused by random copying errors, these mutations will have been caused by some exogenous or endogenous agent. New genetic technologies that make possible the detection of these mutations, and perhaps their cause, would therefore create the dilemma that every person could be classified as injured if such mutations are determined to be a "present injury." See id. at 329 ("according to the view that brands the mere initiation of a DNA mutation in a cell as injury, everyone is injured all the time."). See also Andrew R. Klein, Fear of Disease and the Puzzle of Futures Cases in Tort, 35 U.C. DAVIS L. REV. 965, 966 n.2 (2002) (collecting statistics on large percentages of population who have been exposed to various toxic agents); Arvin Maskin et al., Medical Monitoring: A Viable Remedy for Deserving Plaintiffs or Tort Law's Most Expensive Consolation Prize? 27 WM. MITCHELL L. REV. 521, 528 (2000) (listing toxic exposures which most Americans have experienced). 92. The current controversy over latent risk claims relating to asbestos exposure gives a flavor of the difficult issues to be faced by the proliferation of latent risk claims. See, e.g., James A. Henderson Jr. & Aaron D. Twerski, Asbestos Litigation Gone Mad: Exposure-Based Recovery for Increased Risk, Mental Distress, and Medical Monitoring, 53 S.C. L. REV. 815 (2002) (discussing how the "massive, never-ending que of claimants" litigating latent risk claims for asbestos exposure has "become a tragic chapter in American jurisprudence" and "will remain so unless courts put an end to the madness"). Even some plaintiffs' counsel are now advocating restricting claims for unimpaired plaintiffs who have been exposed to asbestos because their latent risk claims are consuming a disproportionate share of the available funds from bankrupt defendants. Mark P. Goodman et al., Plaintiffs' Bar Now Opposes Unimpaired Asbestos Suits, NAT'L L.J., Apr. 1, 2002, at B14, B15; Alex Berenson, A Surge in Asbestos Suits, Many by Healthy Plaintiffs, N.Y. TIMES, Apr. 10, 2002, at A1, C4. 93. See Badillo v. American Brands, Inc., 16 P.3d 435, 438-39 (Nev. 2001) (surveying case law). 94. See, e.g., In re Paoli R.R. Yard PCB Litig., 916 F.2d 829, 852, 21 ELR 20184, 20196 (3d Cir. 1990), cert. denied, 499 U.S. 961 (1991); Hansen v. Mountain Fuel Supply Co., 858 P.2d 970, 979 (Utah 1993); In re Asbestos Cases, 265 F.3d 861, 866 (9th Cir. 2001). 95. See Maskin et al., supra note 91, at 532-33 (most courts have required plaintiffs to demonstrate exposures that have increased their risks, but impose no requirement with respect to the quantification or magnitude of such increased risk). For example, in the landmark medical monitoring case of Ayers v. Jackson Township, 525 A.2d 287, 312 (N.J. 1987), the New Jersey Supreme Court held that plaintiffs could recover medical monitoring costs, even though the lower court had found that plaintiffs' experts could not exclude the possibility that plaintiffs' increased risk was "so microscopically small as to be meaningless." See also Badillo v. American Brands, Inc., 16 P.3d 435, 441 (Nev. 2001) (surveying the requirements of different states). 96. See supra note 91. 97. See supra note 95. 98. See, e.g., Metro-North Commuter R.R. Co. v. Buckley, 521 U.S. 424, 442 (1997) ("tens of millions of individuals may have suffered exposure to substances that might justify some form of substance-exposure-related medical monitoring"). The potential for a flood of new claims would be particularly pronounced in those jurisdictions that recognize medical monitoring as a separate cause of action that can be brought in the absence of any other injury or claim. 99. See, e.g., Donald T. Hornstein, Reclaiming Environmental Law: A Normative Critique of Comparative Risk Analysis, 92 COLUM. L. REV. 562 (1992); Mark Eliot Shere, The Myth of Meaningful Risk Assessment, 19 HARV. ENVTL. L. REV. 409 (1995). 100. NATIONAL RESEARCH COUNCIL, RISK ASSESSMENT IN THE FEDERAL GOVERNMENT: MANAGING THE PROCESS 29-33 (1983) (listing over 50 uncertainties in risk assessment) 101. See generally Bishop et al., supra note 20, at 986-987; Henry et al., supra note 44, at 10478 (increased understanding from toxicogenomic data could reduce need to apply default uncertainty factors in risk assessment). 102. U.S. EPA, SCIENCE POLICY COUNCIL, INTERIM POLICY ON GENOMICS (2000) at 1, available at http://epa.gov/osp/spc/genomics. pdf. See also Pat Phibbs, EPA Scientist Expects Genomic Information to Improve Analyses of Chemical Effects, Daily Env't Rep. (BNA), Nov. 8, 2001, at A2 (reporting on presentation by high-ranking EPA scientist claiming that EPA has already evaluated or is developing gene expression data from exposure to dioxins, formaldehyde, disinfection byproducts, phthalates, and other chemicals); John C. Rockett & David J. Dix, Application of DNA Arrays to Toxicology, 107 ENVTL. HEALTH PERSPECT. 681, 681 (1999) (two EPA scientists write that "[EPA] is interested in applying DNA array technology to ongoing toxicologic studies."). 103. INTERIM POLICY, supra note 102, at 2. 104. Id. 105. See supra note 29 and accompanying text. 106. See Aardema & MacGregor, supra note 4, at 16-17; INTERIM POLICY, supra note 102, at 3 (genomics will "likely provide a better understanding of the mechanism or mode of action of a stressor and thus assist in predictive toxicology, in the screening of stressors, and in the design of monitoring activities and exposure studies"). While "mode of action" and "mechanism" are often used interchangeably, the former involves a more generalized level of knowledge than the latter. "Mode of action" refers to the critical events caused by a particular substance that leads to toxicity, whereas "mechanism" refers to the precise molecular changes involved in the toxicity response. See Robert J. Golden et al., Chloroform Mode of Action: Implications for Cancer Risk Assessment, 26 REG. TOXICOLOGY & PHARMACOLOGY 142, 143 (1997) ("The mode of action for a carcinogenic substance refers to the primary obligatory step(s) in the carcinogenic process (e.g., DNA reactivity resulting in mutations), whereas the mechanism of action refers to the myriad of primary and secondary effects, interactions, and biochemical alterations that can occur in conjunction with chemical carcinogenesis.") (citations omitted). 107. U.S. EPA, Proposed Guidelines for Carcinogen Risk Assessment, 61 Fed. Reg. 17960, 17980-81 (Apr. 23, 1996); Lewis L. Smith, Key Challenges for Toxicologists in the 21st Century, 22 TRENDS IN PHARMACOL. SCI. 281, 282 (2001) ("It is increasingly recognized that an understanding of the mechanism of toxicity of a chemical in experimental animals, together with a knowledge of the biochemistry and physiology of humans, provides a more reliable basis to predict whether a chemical is likely to prove harmful."). 108. Aardema & MacGregor, supra note 4, at 17. 109. See, e.g., Scott A. Jelinsky & Leona D. Samson, Global Response of Saccharomyces cerevisiae to an Alkylating Agent, 96 PROC. NAT'L ACAD. SCI. 1486, 1490 (1999) (gene expression profiling of yeast cells following exposure to alkylating agent reveals role of protein degradation in responding to toxic damage); Hamadeh et al., supra note 17, at 227 (microarray data provides new insights on molecular mechanism of liver toxicity caused by phenobarbital); Ahmet Zeytun et al., Analysis of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin-Induced Gene Expression Profile in Vivo Using Pathway-Specific cDNA Arrays, 178 TOXICOLOGY 241 (2002) (microarray analysis provided useful information on mechanism of toxicity of dioxin based on the nature of genes up- and down-regulated following exposure). 110. See Aardema & MacGregor, supra note 4, at 18. 111. The Director of the National Institute of Environmental Health Sciences, the nation's preeminent center of environmental toxicological research, concedes that "we have no idea what kinds of risks are posed by low-dose exposures … because testing to this point has, out of necessity, focused on higher exposure levels." Kenneth Olden et al., A Bold New Direction for Environmental Health Research, 91 AM. J. PUB. HEALTH 1964, 1965 (2001). 112. The Department of Energy's Low-Dose Radiation Program is thus utilizing gene expression data using microarrays to study empirically the shape of the dose-response curve for ionizing radiation at low levels of exposure. See http://www.er.doe.gov/production/ober/lowdose.html. 113. See Farr & Dunn, supra note 6, at 1 ("measurement of gene expression may allow us to identify threshold concentrations below which there is little health risk") 114. See Aardema & MacGregor, supra note 4, at 19-20. 115. See THE PRESIDENTIAL/CONGRESSIONAL COMMISSION ON RISK ASSESSMENT AND RISK MANAGEMENT, 2 RISK ASSESSMENT AND RISK MANAGEMENT IN REGULATORY DECISION-MAKING, FINAL REPORT 64 (1997) (while the results of most animal studies were found to be relevant to humans, "some chemicals elicit tumors in rodents only through mechanisms or at doses that have been clearly demonstrated to be very different from mechanisms and exposures in humans"); E. Dybing et al., Hazard Characterisation of Chemicals in Food and Diet: Dose Response, Mechanisms, and Extrapolation Issues, 40 FOOD & CHEM. TOXICOLOGY 237, 259 (2002) (listing examples). For example, saccharin causes bladder tumors in rats by a species-specific mechanism that does not apply to humans, and based on these findings saccharin has recently been delisted as a human carcinogen. See NATIONAL TOXICOLOGY PROGRAM, 9th REPORT ON CARCINOGENS B3-B4 (2000) (explaining decision to delist saccharin as reasonable anticipated to be a carcinogen). 116. For example, one recent study compared gene expression changes in the livers of rodents with cancer resulting from exposure to arsenic with gene expression patterns in human liver cells removed by biopsy from humans exposed to arsenic. The finding that the gene expression changes in the rodent and human tissues were largely in agreement indicated that arsenic was producing a similar toxicological response in humans and rodents. Tong Lu et al., Application of cDNA Microarray to the Study of Arsenic-Induced Liver Diseases in the Population of Guizhou, China, 59 TOXICOLOGICAL SCI. 185, 190 (2001). 117. See Aardema & MacGregor, supra note 4, at 19. This use of gene expression data to extrapolate from animals to humans, while providing additional data for a more informed risk assessment judgment, will nevertheless create its own set of uncertainties and issues. For example, how much weight should be given to the lack of a gene expression response in human cells for extrapolating a positive response in animals to humans? If a particular chemical produces both tumors and gene expression changes in mice, but no gene expression changes (and no indications of carcinogenicity) in rats or human cell cultures, would this provide regulators sufficient confidence to conclude that the chemical is likely not a human carcinogen? Conversely, if another chemical does not increase tumors in a chronic rodent study, but does induce a significant alteration in gene expression that is characteristic of a carcinogenic response for other compounds, should regulators disregard the absence of tumors in the rodent study and nevertheless classify the chemical as a possible human carcinogen based solely on the gene expression changes? 118. See Olden et al., supra note 111, at 1966. 119. See Nuwaysir et al., supra note 3, at 157. 120. See Aardema & MacGregor, supra note 4, at 20-21; V.J. Feron & J.P. Groten, Toxicological Evaluation of Chemical Mixtures, 40 FOOD & CHEM. TOXICOLOGY 825, 834 (2002); INTERIM POLICY, supra note 102, at 3 ("Genomic analysis … holds promise to evaluate the cumulative impacts resulting from the interplay of factors such as genetic diversity, health status, and life stage in responding to exposures(s) to multiple stressors."). 121. See Feron & Groten, supra note 120, at 825 ("Mixtures are tough for everybody.") (quoting Jonathan Sarnet). 122. Id. at 826. 123. Id. at 834. 124. See Olden et al., supra note 111, at 1966. 125. 21 U.S.C. § 346a(b)(2)(C); Cheryl Hogue, Toxicogenomics Uses Genes as a Toxic Screen, CHEM. ENG. NEWS, Mar. 19, 2001, at 33, 34 ("microarrays using DNA from babies and youngsters may shed light on whether they are more sensitive or more resilient to pesticide toxicity than adults are"). 126. See Wendy E. Wagner, The Triumph of Technology-Based Standards, 2000 U. ILL. L. REV. 83 (2000). For example, the 1990 Clean Air Act Amendments replace the previous risk-based approach for regulating hazardous air pollutants with primarily a technology-based approach based on "maximum available control technology" or "MACT." 42 U.S.C. § 7412(d), ELR STAT. CAA § 112(d); see Arnold W. Reitze Jr., Control of Hazardous Air Pollution, in AIR POLLUTION CONTROL LAW: COMPLIANCE AND ENFORCEMENT 123 (Envtl. L. Inst. 2001). 127. 136 CONG. REC. S16895, S16932 (Oct. 27, 1990) (statement of Sen. Durenberger). 128. See Bernard D. Goldstein & Mary Sue Henifin, Reference Guide on Toxicology, in FEDERAL JUDICIAL CENTER, REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 401, 412 (2d ed. 2000) ("less than 1% of the 60,000-75,000 chemicals in commerce have been subjected to a full safety assessment, and there are significant toxicological data on only 10%-20%"); NATIONAL RESEARCH COUNCIL, TOXICITY TESTING: STRATEGIES TO DETERMINE NEEDS AND PRIORITIES (1984); ENVIRONMENTAL DEFENSE FUND, supra note 60. 129. Under the HPV Challenge program, chemical manufacturers have volunteered to collect and make public by 2005 basic toxicity and environmental fate data on chemicals produced or imported into the United States in volumes exceeding one million tons annually. See http://www.epa.gov/opptintr/chemrtk/volchall.htm. 130. NATIONAL TOXICOLOGY PROGRAM, CURRENT DIRECTIONS AND EVOLVING STRATEGIES 2 (2002) ("More than 80,000 chemicals are registered for use in commerce in the United States, and an estimated 2,000 new ones are introduced annually for use in everyday items such as foods, personal care products, prescription drugs, household cleaners, and lawn care products."). 131. See Elaine M. Faustman & Gilbert S. Omenn, Risk Assessment, in CASARETT & DOULL'S TOXICOLOGY: THE BASIC SCIENCE OF POISONS 83, 88 (Curtis D. Klansmen ed., 6th ed. 2001); Ernest E. McConnell, Historical Review of the Rodent Bioassay and Future Directions, 21 REG. TOXICOLOGY & PHARMACOLOGY 38 (1995). 132. Press Release, National Institute of Environmental Health Sciences, NTP Completes 500th Two-Year Rodent Study and Report; series is the Gold Standard of Animal Toxicology (Release # 01-03, Jan. 25, 2001). 133. NIH Environmental Health Prevention Research, Prepared Testimony of Kenneth Olden, Director of the National Institute of Environmental Health Sciences, before the Subcomm. on Public Health, Senate Comm. on Health, Education, Labor, and Pensions, at 6 (Mar. 6, 2002). 134. Id. See also Olden et al., supra note 111, at 1965 ("Without new, high-throughput technologies, … we will not be able to assess the toxicity of the thousands of chemicals on which there are inadequate toxicity data."). 135. See Joseph Sanders, From Science to Evidence: The Testimony of Causation in the Bendectin Cases, 46 STAN. L. REV. 1, 19 (1993) ("Molecules with minor structural differences can produce very different biological effects."); Goldstein & Henefin, supra note 39, at 421 (reliability of SAR "has a number of limitations"); R. Julian Preston & George R. Hoffman, Genetic Toxicology, in CASARETT & DOULL'S TOXICOLOGY: THE BASIC SCIENCE OF POISONS 321 (Curtis D. Klaassen ed., 6th ed. 2001); James D. McKinney et al., The Practice of Structure Activity Relationships (SAR) in Toxicology, 56 TOXICOLOGICAL SCI. 8 (2000). 136. See Raymond W. Tennant, Evaluation and Validation Issues in the Development of Transgenic Mouse Carcinogenicity Bioassays, 106 (Suppl. 2) ENVTL. HEALTH PERSP. 473 (1998). 137. See Aardema & MacGregor, supra note 4, at 17-18. 138. Press Release, National Institute of Environmental Health Sciences, National Center for Toxicogenomics to Study Genetic Basis of Disease Caused by Environmental Pollution (Dec. 7, 2000), available at http://www.niehs.nih.gov/nct/pr07de00.htm. 139. See supra note 29; Thomas, supra note 6, at 1194 (initial promising results using microarrays "open[] the door to a new era of toxicological testing where relatively short and inexpensive studies using transcript expression as an endpoint allow the prioritization of untested chemicals based upon their classification"). Potential toxicity mechanisms that may induce characteristic gene expression "finger-prints" include DNA alkylation, inflammation, oxidative stress, peroxisome proliferators, estrogenic action, and many others. Id. at 1189-90. 140. See, e.g., Hamadeh et al., supra note 17, at 225. 141. See Bartosiewicz et al., supra note 29, at 73. 142. See Aardema & MacGregor, supra note 4, at 18 ("it will be necessary to characterize multiple classes of agents with well-defined mechanisms of action before expression profiles for new biomarkers can be used reliably in regulatory decision-making."). 143. See Nuwaysir et al., supra note 3, at 157. 144. See Aardema & MacGregor, supra note 4, at 18. 145. 15 U.S.C. § 2604, ELR STAT. TSCA § 5. See BIOTECHNOLOGY DESKBOOK 43 (Envtl. L. Inst. 2001). 146. 15 U.S.C. § 2604(d), ELR STAT. TSCA § 5(d). See TSCA DESKBOOK (Envtl. L. Inst. 1999); Carolyne R. Hathaway et al., A Practitioner's Guide to the Toxic Substances Control Act: Part I, 24 ELR 10207 (May 1994). 147. Section 313 of the Emergency Planning and Community Right-To-Know Act of 1996 establishes a list of chemicals for which certain facilities are required to report annual releases into the environment. 42 U.S.C. § 11023, ELR STAT. EPCRA § 313. EPA can add chemicals to this list that are "known to cause or can reasonably be anticipated to cause" major types of toxicity "based on generally accepted scientific principles or laboratory tests." Id. § 11023(d)(2), ELR STAT. TSCA § 313(d)(2). EPA has established a toxicity screening methodology for determining the addition or deletion of chemicals from this TRI list. See U.S. EPA, Addition of Certain Chemicals; Toxic Chemical Release Reporting; Community Right-to-Know, 59 Fed. Reg. 61432, 61432-33 (Nov. 30, 1994). Gene expression assays may be a quick and inexpensive addition to this screening procedure. 148. There are two ways that EPA can designate hazardous wastes—either by listing wastes by regulation based on their toxicity, or by finding that they exhibit one or more hazardous "characteristics," including the "toxicity" characteristic. See generally U.S. EPA, RCRA ORIENTATION MANUAL, III-3 to III-22 (EPA/530-R-00-006, June 2000). EPA conducts a screening toxicity assessment for adding listed wastes, and tests for the toxicity characteristic using an assay called the Toxicity Characteristic Leaching Procedure (TCLP). This assay, which simply measures the leaching potential of wastes, has been strongly criticized. See, e.g., David Montgomery Moore, The Toxicity Characteristic Rule for Hazardous Waste Determination: Has EPA Satisfied Congress' Mandate?, 7 TUL. ENVTL. L.J. 467 (1994). Microarray analysis of gene expression changes in test cells or organisms treated with a candidate waste have the potential to substantially improve the scientific determination of both listed wastes and wastes exhibiting the toxicity characteristic. 149. 21 U.S.C. § 346a(b)(2)(D)(v). 150. See Hogue, supra note 38, at 34. 151. EPA has developed a "Hazard Ranking System" (HRS) to determine which hazardous waste sites should be placed on the NPL, which is EPA's list of sites that are priorities from long-term evaluation and remediation under the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA). U.S. EPA, Hazard Ranking System; Final Rule, 55 Fed. Reg. 51532 (Dec. 14, 1990). See also Terry C. Clarke, A Practitioner's View of the National Priorities List, 2 ENVTL. LAW. 57 (1995). The HRS uses a scoring system that attempts to assess the relative potential of a site to pose a threat to human health or the environment. Gene expression data from nearby species could be factored into this scoring system. 152. Herbert L. Frederickson et al., Towards Environmental Toxicogenomics—Development of a Flow-Through, High-Density DNA Hybridization Array and Its Application to Ecotoxicity Assessment, 274 SCI. TOTAL ENV'T 137, 138 (2001). 153. Id. at 139. 154. Id. 155. Id. 156. Id. 157. Id. at 141. 158. Id. at 147. 159. Id. Because the genosensor tests soil and sediment samples directly, unlike conventional toxicology methods which test individual constituents separately, many of the uncertainties associated with conventional test methods do not apply to genosensor technology. 160. Id.; NATIONAL RESEARCH COUNCIL, BIOAVAILABILITY OF CONTAMINANTS IN SOILS AND SEDIMENTS: PROCESSES, TOOLS, AND APPLICATIONS 243 (2002) ("Using microarray techniques, it is possible to develop a sensitive and inclusive snapshot of the response of cells, tissues, and organisms to a contaminant without the time requirements, labor, or subjectivity of more traditional analyses. Validating these techniques, and increasing their practicality for specifically assessing contaminant bioavailability from soils and sediments, should occur in the near future."). 161. See William D. Pennie et al., Application of Genomics to the Definition of the Molecular Basis for Toxicity, 120 TOXICOLOGY LETTER 353, 354 (2001) (gene expression profiling "could highlight potential toxicity earlier in a new compound's development"); European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC), ECETOC White Paper on Genomics, Transcript Profiling, Proteomics and Metabonomics (GTPM) (Mar. 2001). 162. The IRIS database is available online at http://www.epa.gov/iris/index.html. RfCs are used when the route of exposure for a chemical is inhalation, while RfDs are used for oral routes of exposure (e.g., via drinking water). 163. U.S. EPA, INTEGRATED RISK INFORMATION SYSTEM, GLOSSARY OF IRIS TERMS (rev. 1999), available at http://www.epa.gov/iris/gloss8.htm [hereinafter U.S. EPA IRIS GLOSSARY]. 164. EPA defines a NOAEL as the "highest exposure level at which there are no statistically or biologically significant increases in the frequency or severity of adverse effect between the exposed population and its appropriate control; some effects may be produced at this level, but they are not considered adverse, nor precursors to adverse effects." Id. 165. A LOAEL is defined by EPA as the "lowest exposure level at which there are statistically or biologically significant increases in the frequency or severity of adverse effects between the exposed population and its appropriate control group." Id. 166. Richard W. Lewis et al., Recognition of Adverse and Nonadverse Effects in Toxicity Studies, 30 TOXICOLOGICAL PATHOLOGY 66, 67 (2002). 167. ECETOC, supra note 161, at 1 ("There is … the real danger that indiscriminate application of these technologies will lead to the generation of misleading data. Furthermore, the current (relative) lack of reference data could easily lead to mis- or over-interpretation and subsequently to undue concern by regulatory agencies."); Pennie & Kimber, supra note 16, at 321 ("Where the results are likely to influence the derivation of No-Observed-Adverse-Effect Levels (NOAELs) the relevance to the in vivo situation must be established by correlating the observed changes with 'classical' adverse effects."). 168. See Dybing et al., supra note 115, at 254 (adaptive responses are "stress reactions to environmental influences whereby the organism tries to maintain homeostasis. Enzyme induction, changes in hormone levels, and indicators of slightly altered cellular functions are examples of such adaptive responses. In many instances, these responses do not lead to clinically significant altered structure or function, namely adverse reactions."). 169. Id. at 254-55 ("it will be a great challenge to clarify whether such [gene expression] changes simply represent non-adverse alterations of physiological function or if they predict impending development of more serious irreversible injury, should exposure to the chemical continue"). 170. For example, the induction of increased enzyme activity, a frequent response to exposure to foreign substances, "may in some situations be present as an adaptive response without any biological significance; sometimes it may be beneficial in that it leads to more rapid metabolism and elimination of potentially toxic compounds; or it may be a truly adverse response in that it may lead to increases in reactive intermediates and thus potentiate toxic effects." Id. at 254. 171. Id. at 255. 172. Lewis et al., supra note 166, at 68-74 (proposing multi-factor frame-work for defining adverse effects); Dybing et al., supra note 115, at 254 (the National Academy of Sciences has defined an adverse effect as a change in morphology, growth, development, or life span, an impairment of the organism's capacity to compensate for additional stress, or an increased susceptibility to additional toxic exposures). EPA defines an "adverse effect" as "[a] biochemical change, functional impairment, or pathological lesion that affects the performance pf the whole organism, or reduces the organism's ability to respond to an additional environmental challenge." U.S. EPA IRIS GLOSSARY, supra note 163. This definition obviously leaves much room for subjectivity. 173. Lewis et al., supra note 166, at 67. 174. U.S. EPA, RISK ASSESSMENT FORUM, A REVIEW OF THE REFERENCE DOSE AND REFERENCE CONCENTRATION PROCESSES, EXTERNAL REVIEW DRAFT 4-9 (May 2002), available at http://www.epa.gov/ncea/raf/pdfs/RfDRfC/rfdrfcextrevdrft.pdf. 175. Id. 176. See Lewis et al., supra note 166, at 72 ("often the distinction between adverse and nonadverse effects is not clearly defined and interpretation needs scientific judgment on a case-by-case basis"); William D. Pennie et al., The Principles and Practice of Toxicogenomics: Applications and Opportunities, 54 TOXICOLOGICAL SCI. 277, 282 (2000): It must be recognized that the interaction of xenobiotics with biological systems will in many instances result in some changes in gene expression, even under circumstances where such interactions are benign with respect to adverse effects. The challenge again is to ensure that sound judgment and the appropriate toxicological skills and experience are brought to bear on the data generated, so that toxicologically relevant changes in gene expression are distinguished from those that are of no concern. 177. See George V. Alexeeff et al., Characterization of the LOAEL-to-NOAEL Uncertainty Factor for Mild Adverse Effects From Acute Inhalation Exposures, 36 REG. TOXICOLOGY & PHARMACOLOGY 96 (2002); A.G. Renwick, The Use of an Additional Safety or Uncertainty Factor for Nature of Toxicity in the Estimation of Acceptable Daily Intake and Tolerable Daily Intake Values, 22 REG. TOXICOLOGY & PHARMACOLOGY 250 (1995); John D. Graham, Historical Perspective on Risk Assessment in the Federal Government, 102 TOXICOLOGY 29, 33 (1995): The severity of adverse effects caused by toxic agents can vary from mild cases of reversible skin irritation to death. Historically, the severity of the adverse health effect caused by an agent has not played an explicit role in the choice of safety factors, although some extremely mild biological responses have on occasion not been considered "adverse." 178. See Alexeef et al., supra note 177, at 103 ("The value of 3 is often used when the adverse effect at the LOAEL is considered mild in severity."). For example, EPA reduced the standard uncertainty factors in calculating the RfC for acrylic acid "because the effect is considered mild." U.S. EPA, INTEGRATED RISK INFORMATION SYSTEM, ACRYLIC ACID (CASRN 79-10-7), available at http://www.epa.gov/iris/subst/0002.htm (last visited Sept. 9, 2002). In Chemical Manufacturers Association v. EPA, 28 F.3d 1259, 24 ELR 21210 (D.C. 1994), the D.C. Circuit rejected EPA's reliance on the RfC for methylene diphenyl diisocyanate in promulgating a regulation because EPA failed to give any weight to the seriousness of the health effect upon which the RfC was based. 179. Bartosiewicz et al., supra note 29, at 71. 180. Philip M. Iannaccone, Toxicogenomics: "The Call of the Wild Chip," 109 ENVTL. HEALTH PERSP. A8, A10 (2001) (using microarrays, "it may be possible to screen biological samples obtained from workers at Superfund sites for the adverse effects of exposure to compounds present in the site"). Some studies have reported an increased frequency of chromosomal anomalies in residents living near hazardous waste sites, E.g., M. Vrijheid et al., Chromosomal Congenital Anomalies and Residence Near Hazardous Waste Landfill Sites, 359 LANCET 320 (2002). Gene expression assays have the potential to provide a monitoring assay of such residents that is both more sensitive (by producing a positive response in a higher proportion of exposed individuals) and informative (by providing more specific information on the class of toxicant causing the response and the risk associated with that response) than currently available methods. 181. See Bishop et al., supra note 20, at 986. 182. Amundson et al., supra note 14, at 342. 183. Id. at 343. 184. Id. at 344. 185. Id. at 343. If different people with varying genetic and environmental backgrounds do indeed respond in a similar way to a particular exposure, it is not necessary to obtain unexposed control samples from each individual in order to quantify exposures. Obtaining preexposure background gene expression samples from an entire at-risk population would be very difficult after an accident or other exposure scenario had occurred. Id. 186. Id. at 346. 187. Another recent study found that humans exposed to arsenic exhibited gene expression changes in their liver cells removed by biopsy that closely resembled the gene expression changes in rodent livers that developed carcinomas from arsenic exposure. Lu et al., supra note 116, at 190. Such assays using microarray systems could therefore be used to identify individuals at increased risk of cancer from exposure to toxicants such as arsenic. 188. See Nuwaysir et al., supra note 3, at 157 (bioassays based on standard ecotoxicity model systems could be improved by the addition of microarray analysis). 189. Patrick Larkin et al., Array Technology as a Tool to Monitor Exposure of Fish to Xenoestrogens, 54 MARINE ENVTL. RES. 395 (2002). 190. Id. 191. Doug Crump et al., Exposure to the Herbicide Acetochlor Alters Thyroid Hormone-Dependent Gene Expression and Metamorphosis in Xenopus Laevis, 110 ENVTL. HEALTH PERSP. 1199 (2002). 192. See Bartosiewicz et al., supra note 29, at 71. 193. See Iannaccone, supra note 180, at A10. 194. Other currently available surveillance assays, such as monitoring for chromosomal changes or electron spin resonance of dental enamel (for detecting radiation exposure), are not capable of large-scale population monitoring. See Amundson et al., supra note 14, at 346. In addition, these methods may be more intrusive, e.g., the electron spin resonance method requires extraction of a tooth. Id. 195. See Farr & Dunn, supra note 6, at 1 ("the detection of altered gene expression can serve as an early warning for subsequent deleterious outcomes"); Tennant, supra note 29, at A9 ("If array data can be 'phenotypically anchored' to conventional indices of toxicity … it will be possible to search for evidence of injury prior to its clinical or pathological manifestation. This approach could lead to development of early biomarkers of toxic injury …."). See also supra notes 137-61 and accompanying text. 196. 15 U.S.C. § 2607(e), ELR STAT. TSCA § 8(e). 197. EPA's Reporting Guide states the two factors to be considered in determining whether an effect indicates "substantial risk" and thus should be reported are: "1) the seriousness of the adverse effect, and 2) the fact or probability of the effect's occurrence." U.S. EPA, TSCA SECTION 8(E) REPORTING GUIDE 3 (1991), available at http://www.epa.gov/oppt/tsca8e/doc/rguide91.pdf. The guide then discusses some specific examples to illustrate these factors. For example, it states that "serious in vivo genotoxicological effects (e.g., gene or chromosomal mutations) are reportable in and of themselves," whereas "a positive in vitro genotoxicity test, when considered alone, is usually insufficient to cause reporting under Section 8(e)." Id. at 28-29. The guide continues that a positive in vitro test would, however, normally suggest the need for additional studies, which if also positive, may trigger a reporting requirement. Id. at 29. Another relevant example discussed is organ weight change, "which in and of itself, may not reflect a serious or prolonged incapacitation," but the reportability of such effects will depend on other factors such as "the biological significance of the change." Id. at 45. 198. EPA's § 8(e) guidance requires reporting of "any pattern of effects or evidence which reasonably supports the conclusion that the chemical substance or mixture can produce cancer, mutation, birth defects or toxic effects resulting in death, or serious or prolonged incapacitation." U.S. EPA, Toxic Substances Control Act: Notification of Substantial Risk Under Section 8(e), 43 Fed. Reg. 11110, 11112 (Mar. 16, 1978). 199. 7 U.S.C. § 136d(a)(2). 200. 40 C.F.R. § 159.158(a). 201. Id. § 159.165(a). 202. See supra notes 166-215 and accompanying text. 203. See, e.g., 42 U.S.C. § 7409(b)(1), ELR STAT. CAA § 109(b)(1) (requiring EPA to set national ambient air quality standards that protect the public health with an adequate margin of safety); 42 U.S.C. § 300g-1(b)(4)(A), ELR STAT. SDWA § 1412(b)(4)(A) (requiring EPA to set maximum contaminant level goals (MCLGs) under the Safe Drinking Water Act at "the level at which no known or anticipated adverse on the health of persons occur and which allows an adequate margin of safety"). 204. 42 U.S.C. § 7409(b)(1), ELR STAT. CAA § 109(b)(1). 205. E.g., S. REP. 1196, 91st Cong. 86 (1970) (air quality standards must ensure "an absence of adverse effect on the health of a statistically related sample of persons in sensitive groups …."). 206. U.S. EPA, National Ambient Air Quality Standards for Ozone, Final Rule, 62 Fed. Reg. 38856, 38868 (July 18, 1997); U.S. EPA, National Ambient Air Quality Standards for Ozone—Final Decision, 58 Fed. Reg. 13008, 13011 (Mar. 9, 1993). 207. U.S. EPA, National Primary and Secondary Ambient Air Quality Standards: for Lead; Final Rulemaking, 43 Fed. Reg. 46246, 46247 (Oct. 5, 1978). Protoporphyrin combines with iron in the red blood cells (erythrocytes) to form heme, a critical component of hemoglobin, which transports oxygen in the blood. Lead blocks the formation of heme, resulting in elevated levels of protoporphyrin in erythrocytes that would otherwise have been incorporated into heme. Id. at 46253. 208. Id. at 46247. 209. Id. 210. Id. at 46251. See also id. at 46252 ("EPA is making a distinction between the blood lead level that is the threshold for detection of the biological effect, impaired heme synthesis, and the blood lead level at which this effect has progressed to an extent that it is regarded as adverse to health."). 211. 647 F.2d 1130, 10 ELR 20643 (D.C. Cir.), cert. denied, 449 U.S. 1042 (1980). 212. Id. at 1139, 10 ELR at 20646. 213. 43 Fed. Reg. at 46252. 214. Id. 215. U.S. EPA, Lead; Proposed National Ambient Air Quality Standard, 42 Fed. Reg. 63076, 63078 (Dec. 14, 1977). 216. The Chairman of EPA's Clean Air Act Scientific Advisory Committee (CASAC) testified to Congress in 1997 that as our ability to detect subtle responses of lung tissue to O3 improves, and as we gain more refined information from especially susceptible humans exposed in the environment and in laboratories, it is becoming increasingly apparent that there will probably be no clear threshold for detecting responses at exposure concentrations within the concentration range amenable to regulatory control. It should be understood that our growing ability to detect and study responses to exposures to O3 and other pollutants down to, and perhaps including, background (uncontrollable) concentrations is necessarily accompanied by an increasing demand on our judgment of … when cellular responses should be considered sufficiently adverse to warrant regulatory action …. U.S. EPA's Proposed Clean Air Act Regulations, Prepared Statement of Joe L. Mauderly, Testimony Before the Subcomms. on Health & Env't and Oversight & Investigations, House Comm. on Commerce (Apr. 10, 1997), available at 1997 WL 10571479. 217. See supra note 22 and accompanying text. See also Tennant, supra note 29, at A10: Given the vast numbers and diversity of drug, chemicals, and environmental stressors, the diversity of species in which they act, the time and dose factors that are critical to the induction of beneficial and adverse effects, and the diversity of phenotypic consequences of ex posures, it is only through the development of a rich knowledge base [of microarray data] and its availability to all of the scientific community that toxicology and environmental health can rapidly advance. 218. Aardema & MacGregor, supra note 4, at 23; Hamadeh & Afshari, supra note 8, at 515 ("Microarray technology will undoubtedly have a profound impact on many avenues of biological and biomedical research, including toxicology …."). 219. See generally Hadley C. King & Animesh A. Sinha, Gene Expression Profile Analysis by DNA Microarrays; Promise and Pitfalls, 286 J. AM. MED. ASS'N 2280 (2001); Rockett & Dix, supra note 102, at 684. 220. Jonathan Knight, When the Chips Are Down, 410 NATURE 860, 860 (2001) (quoting toxicologist Timothy Zacharewski). 221. See Stefano Bonassi et al., Validation of Biomarkers as Early Predictors of Disease, 480 MUTATION RES. 349 (2001); Bonassi & Au, supra note 2, at 79. 222. Bernard A. Schwetz, Toxicology at the Food and Drug Administration: New Century, New Challenges, 20 INT'L J. TOXICOLOGY 3, 6 (2001): New chips can be developed faster than chips can be validated. Validation is a significant underpinning of the quality of toxicology data today. If microarray technology moves faster than the validation technology, how will we use these data for the development of new products and the development of data to assess safety and efficacy? 223. Criteria for validation of new toxicological test methods were defined in a report issued by the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM), established in 1997 by the National Institute of Environmental Health Sciences and subsequently established as a permanent committee of representatives of 15 federal agencies by the ICCVAM Authorization Act of 2000, Pub. L. No. 106-545 (2000). These validation criteria include, inter alia, requirements that (i) the relationship of the test method's endpoint(s) to the biological effect of interest must be described, (ii) a detailed protocol of the test method must be available, (iii) the extent of within-test variability, and the reproducibility of the test within and between laboratories, must have been demonstrated, and (iv) the test method's performance must have been demonstrated using representative reference chemicals or test agents. ICCVAM, VALIDATION AND REGULATORY ACCEPTANCE OF TOXICOLOGICAL TEST METHODS 21-22 (NIH Pub. # 97-3981, Mar. 1997), available at http://iccvam.niehs.nih.gov/docs/guidelines/validate.pdf. 224. See Hamadeh et al., supra note 15, at 239; King & Sinha, supra note 219, at 2281 ("The ethnicity, sex, age, and genetic background of a patient are likely to affect the gene expression profiles of many tissues to varying extents."); Mark R. Fielden & Tim R. Zacharewski, Challenges and Limitations of Gene Expression Profiling in Mechanistic and Predictive Toxicology, 60 TOXICOLOGICAL SCI. 6, 8 (2001) ("transcriptional responses may differ between one target cell and another, from cell culture to in vivo conditions, or from rodent models to humans"). 225. See Timothy R. Hughes et al., Functional Discovery via a Compendium of Expression Profiles, 102 CELL 109 (2000) (comparison of gene expression profiles in genetically identical strains of yeast grown under identical conditions finds significant fluctuations in expression of some genes, which apparently represent a form of "biological noise"); King & Sinha, supra note 219, at 2281; Rockett & Dix, supra note 102, at 684. Most laboratories require at least a two-fold increase or decrease in gene expression before a change in gene expression will be considered significant. Id. (detection of 1.5 to 2-fold changes in gene expression reported, but expressing concern about the capability to achieve such detection levels generally); King & Sinha, supra note 219, at 2284 (arguing that a twofold change in expression may not always indicate a meaningful effect given the large natural variability in gene expression); Hamadeh & Afshari, supra note 8, at 513 (microarrays can "detect changes in the expression level of a gene of about 1.5 times"). 226. Eric S. Lander, Array of Hope, 21 (Suppl.) NATURE GENETICS 3, 3 (1999); Hogue, supra note 38, at 33. 227. See Stephen A. Bustin & Sina Dorundi, The Value of Microarray Techniques for Quantitative Gene Profiling in Molecular Diagnostics, 8 TRENDS MOLECULAR MED. 269, 269 (2002). 228. Kathleen McGowan, As Pipelines Wither, Pharma and FDA Explore Whether Microarrays Are Ready for Primetime, Genome-Web.com, Apr. 15, 2002, available at http://www.genomeweb.com/articles/view-article.asp?Article=200241594143. 229. See Marjorie F. Oleksiak et al., Variation in Gene Expression Within and Among Natural Populations, 32 NATURE GENETICS 261, 263 (2002) (finding substantial variations in gene expression within natural population of fish species); Nuwaysir et al., supra note 3, at 157 (diet, health, and lifestyle factors can affect gene expression patterns); King & Sinha, supra note 219, at 2281-82 (same). See also Henry et al., supra note 44, at 1049 ("Little information is available on the prevalence of mutations and gene expression patterns across various population groups, lifestyles, and health conditions."). 230. See Zeytun et al., supra note 109, at 256 ("the important finding of this study was that exclusion of genes that are highly sensitive to slight variations in the experimental condition leads to better correlation between gene expression profiles and the mode of action of the toxicant"). See also Fielden & Zacharewski, supra note 65, at 8 ("gene expression profiles cannot be used as an explanation or predictor of toxicity unless correlated with an adverse effect"); Pat Phibbs, Testing: Using Fewer Genes to Classify Chemicals Said to Be More Effective, Limit Liability, 25 Chem. Reg. Reptr. (BNA), Dec. 10, 2001, at 1753 (reporting findings that a microarray containing only the 12 most important genes for toxicity was able to accurately classify chemicals and may "also limit liability by reducing the number of 'smoking guns' in a company's closet"). 231. See Pennie et al., supra note 161, at 356 ("The technology does not … distinguish causative events from adaptive response (or even system noise) ….") (emphasis in original); Henry et al., supra note 4, at 1049 (necessary to determine "whether observed changes in gene/protein expression are causative, coincidental, or adaptive responses to a chemical"); Smith, supra note 107, at 283: This technology is not without its dangers, at least during the phase where toxicologists are beginning to understand appropriate application of toxicogenomic and proteomic data. It is already known that the response of organs, or indeed of individual cell types, to exposure to individual toxicants can lead to an alteration in many genes in a simultaneous and, as yet, undefined manner. However, many of these changes are almost certainly adaptive and unrelated to the mechanism of toxicity and/or human safety. As this technology develops, scientists will be able to describe altered gene expression provoked by chemicals long before they are able to offer valid interpretations of their meaning. the potential for inadvertently raising concerns over the effect of chemicals in experimental animals (and hence humans), or even the intentional misrepresentation of the results to suggest that chemicals are "playing" with our genes, is enormous. See also supra notes 169-71 and accompanying text. 232. See Bonassi & Au, supra note 2, at 82; Pennie & Kimber, supra note 16, at 321; Henry et al., supra note 44, at 1047 ("there is a critical need to establish relationships between gene expression data and toxicological changes, enabling an integration of 'omics' information with known toxicological measures"). 233. Hogue, supra note 38, at 33. 234. See Fielden & Zacharewski, supra note 65, at 7. 235. Id.; Iannaccone, supra note 180, at A10. 236. See Bustin & Dorundi, supra note 227, at 271. 237. See, e.g., Carol Ezzell, Proteins Rule, SCI. AM., April 2002, at 41; Daniel C. Liebler, Proteomic Approaches to Characterize Protein Modifications: New Tools to Study the Effects of Environmental Exposures, 110 ENVTL. HEALTH PERSP. 3 (2002). 238. See Fielden & Zacharewski, supra note 65, at 7 ("our ability to define the mechanism of action of a compound using gene expression profiling technologies will be highly limited in resolution"); lannaccone, supra note 180, at A10. 239. Henry et al., supra note 44, at 1047 ("given the speed with which the field is evolving, standardization of research platforms or methods does not appear to be appropriate at this time"). 240. See Aardema & MacGregor, supra note 4, at 22; King & Sinha, supra note 219, at 2284. 241. See Bustin & Dorundi, supra note 227, at 269-70. 242. See Hamadeh & Afshari, supra note 8, at 515 ("Microarrays are certainly a giant leap into the future of performing a quality biological research that holds the promise to aid in discovery of better chemicals, diagnostics and pharmaceutical compounds and ultimately, to improve the quality of life of future generations."); Pennie & Kimber, supra note 16, at 319. 243. Tennant, supra note 136, at A8. 244. Id. 245. See Aardema & MacGregor, supra note 4, at 18 ("It is critical that toxicologists in industry, regulatory agencies, and academic institutions develop a consensus, based on rigorous experimental data, about the reliability and interpretation of endpoints such as global gene expression patterns prior to use in regulatory and industrial settings."); Scott W. Burchiel et al., Analysis of Genetic and Epigenetic Mechanisms of Toxicity: Potential Roles of Toxicogenomics and Proteomics in Toxicology, 59 TOXICOLOGICAL SCI. 193, 194 (2001) ("Whereas scientists pursuing mechanistic data may be interested in identifying a broad range of genes that are affected by drugs, chemicals, and complex mixtures, the regulatory communities will be interested in only those genes that are indicative of a critical health effect."). 246. U.S. EPA, supra note 102. 247. E.g., id. at 1 ("EPA believes that genomics will have an enormous impact on our ability to assess the risk from exposure to stressors and ultimately to improve our risk assessments."). 248. Id. at 2. 249. Id. 250. Id. at 3-4: As EPA gains experience in applying genomics information and refines its understanding of the use of such information, it will develop guidance to explain how genomics data can be better utilized in informing decisionmaking and related ethical, legal, and societal implications. EPA is working with other federal, state, and tribal organizations, as well as with academic, international, and industry groups to facilitate scientifically sound applications of genomics data. In addition, EPA will continue to build partnerships and communicate with all interested stakeholders as an essential component of the Agency's future activities in genomics. 251. See supra notes 196-202 and accompanying text. 252. See Henry et al., supra note 44, at 1047 (initially, "'omics' findings will likely be misinterpreted, because no guidelines currently exist for correlating quantitative or qualitative changes in gene/protein/metabolite expression with the potential for adverse effects"). 253. See General Elec. Co. v. EPA, 53 F.3d 1324, 1329, 25 ELR 20982, 20984 (D.C. Cir. 1995) (an "agency must always provide 'fair notice' of its regulatory interpretations to the regulated public"). 254. See, e.g., Henry et al., supra note 44, at 1047 (recommending adoption of recommendations for best practices for microarray systems). 255. See David Rejeski, Exploring the Genomics Frontier, RISK POLICY REP., Aug. 20, 2002, at 24, 25. 256. ICCVAM refers to the Interagency Coordination Committee on the Validation of Alternative Methods, an interagency coordinating committee established by the National Institute of Environmental Health Sciences to improve the toxicological test methods used by federal agencies. The ICCVAM Act of 2000 made this committee permanent. See supra note 223. 257. 42 U.S.C. § 2581-4(c). 258. A number of recent conference workshops and newsletter articles for trial lawyers have addressed the potential applications of toxicogenomics for toxic tort and product liability litigation. 259. See Gary E. Marchant, Genetic Susceptibility and Biomarkers in Toxic Injury Litigation, 41 JURIMETRICS 67, 106 (2000). 260. See Henry et al., supra note 44, at 1049 ("Because of the excitement surrounding the discovery of a new technology, some people may use information based on 'omics' data … before its effectiveness is proven. A certain amount of guidance for tests and results is needed to ensure that misrepresentation and misinterpretation do not occur."). 33 ELR 10071 | Environmental Law Reporter | copyright © 2003 | All rights reserved |