Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment (2007)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

11 Ethical, Legal, and Social Issues Toxicogenomic research and its applications will raise many ethical, legal, and social issues. Because toxicogenomics involves the collection and analy- sis of personal genetic and phenotypic information from large numbers of indi- viduals, it raises more significant ethical, legal and social issues than does, for example, release of reference genome sequences. Although the issues often overlap, they are discussed below in the following five categories: research is- sues, ethical and social issues, regulatory implications, litigation, and communi- cation and education. The early parts of the chapter cover general concepts and the later parts apply the concepts to toxicogenomics. The following discussion includes examples relating to the efficacy of pharmaceuticals, but similar issues may also apply to genomic-related investigations of the toxicity of pharmaceuti- cal and environmental chemicals. RESEARCH ISSUES Research on or using toxicogenomics raises three categories of ethical is- sues. First, toxicogenomic research not involving human subjects may raise is- sues common to all biomedical research, including research integrity, conflicts of interest, and commercial relations and disclosures. Second, toxicogenomic research involving human subjects raises generally applicable human subject issues, such as the Common Rule and the Health Insurance Portability and Ac- countability Act of 1996 (HIPAA) Privacy Rule, equitable selection of subjects, informed consent, privacy and confidentiality, and special protections for vul- nerable populations. Third, toxicogenomic research with human subjects in- volves distinct issues related to the collection and use of biorepositories in re- search, special issues of informed consent in genetics research, community 172 

Ethical, Legal, and Social Issues 173 consultation when doing genetics research in discrete subpopulations, and the duty to notify sample donors and their relatives of research findings. Research Priorities One myth about scientific research is that science is value neutral—that new scientific understandings await discovery, that these discoveries have no independent moral significance, and that they take on moral significance only when individuals and groups of individuals assign weight to scientific findings. One flaw in this argument is that there are seemingly limitless areas for scien- tific inquiry, yet there are finite numbers of scientists and limited resources to pursue research. Therefore, scientists and society must set priorities for research, and those priorities are a function of societal values. Even though scientists of- ten make adventitious discoveries, they generally discover what they are looking for, and what they look for are the things that science and society value discov- ering. Pharmacogenomics provides a good illustration. Pharmaceutical research- ers may discover numerous polymorphisms in the gene for a particular drug target. Before spending tens or hundreds of millions of dollars on studying a particular polymorphism, any revenue-conscious biotech or pharmaceutical company will try to ascertain the population frequency of these drug targets. Consequently, initial drug development does not focus on genetic variations found most often among individuals in developing countries, where most people do not have the money to pay for basic medicines, let alone expensive new products that treat persons with particular genotypes. Even in developed coun- tries, without some external support from government or private sources, devel- oping drugs directed at rare genotypes is not cost-effective. Another way to set research priorities is to focus on the nature of the harm to be prevented. It has been asserted that a disproportionate share of health re- sources (research and treatment) are directed at specific diseases simply because of the effectiveness of advocacy groups working on behalf of affected individu- als (Greenberg 2001). Similarly, environmental policy may be influenced by considering certain environmental risks to be of greater societal concern than others. For example, it has been argued that U.S. environmental policy is flawed because we spend millions of dollars on a relatively few Superfund sites in need of remediation but insufficient sums on air and water pollution, which is a more widespread threat to public health (Breyer 1995). Undoubtedly, the targets of toxicogenomic research will be influenced by a myriad of social factors, and priority setting will be influenced by economic and political concerns. The Federal Rule for the Protection of Human Subjects The federal Rule for Protection of Human Subjects (45 C.F.R. Part 46, Subpart A) is usually referred to as the Common Rule, because most federal 

174 Applications of Toxicogenomic Technologies agency sponsors of research in addition to the Department of Health and Human Services (DHHS) have adopted it (DHHS 1997). The Common Rule describes basic procedures and principles for conducting federally sponsored research on human subjects. It applies to researchers who receive federal financial support from a signatory federal agency, research conducted in contemplation of a sub- mission to the Food and Drug Administration (FDA) for approval, and research conducted by an institution that has signed a “multiple project assurance” with DHHS promising to comply with the Common Rule in all research, regardless of the funding source. As a result, the Common Rule applies to much, but not all, research involving human subjects in the United States. The Common Rule is designed to safeguard the welfare of human subjects of research, and therefore an important preliminary matter is to define “re- search.” The Common Rule defines it as “a systematic investigation, including research development, testing, and evaluation designed to develop or contribute to generalizable knowledge” (45 C.F.R. § 102[d]). Research also includes the development of repositories for future research (DHHS 1997). The Common Rule does not apply to research conducted on specimens or health records that are not individually identifiable. According to a guidance document issued by the Office for Human Research Protections (OHRP), private information and specimens are not “individually identifiable when they cannot be linked to specific individuals by the investigator(s) either directly or through coding systems” (DHHS 2004). The guidance further provides that research involving only coded private information or specimens does not involve human subjects if the investigator cannot “readily ascertain” the identity of the individ- ual because the key was destroyed before the research began, the key holder has agreed not to release the key to investigators under any circumstances, there are Institutional Review Board (IRB)-approved written policies prohibiting release of the key until the individuals are deceased, or other legal requirements prohibit releasing the key to the investigators until the individuals are deceased. Neither the OHRP guidance nor other authoritative policies address all the major issues surrounding specimen collection and use, including identifiability (McGuire and Gibbs 2006) and the potential for group harm (Foster and Sharp 2002). It is also possible that definitions of “readily identifiable” will need to change with the incorporation of large amounts of polymorphism and similar data in databases. Health Insurance Portability and Accountability Act HIPAA (42 U.S.C. §§ 300gg-300gg-2) contains a section called “Admin- istrative Simplification” (Title II, subtitle F, §§ 261-264), which commits the United States to more efficient health claim processing through standard elec- tronic transactions. To protect the confidentiality of protected health informa- tion, DHHS was directed to promulgate Standards for Privacy of Individually Identifiable Health Information (Privacy Rule) (45 C.F.R. Parts 160, 164). The Privacy Rule applies to three classes of covered entities: health care providers, 

Ethical, Legal, and Social Issues 175 health plans, and health clearinghouses. The covered entities under the Privacy Rule and the Common Rule overlap to a great extent (for example, both cover large academic medical center institutions), but they are not of equal scope. The Privacy Rule and the Common Rule also differ on other key issues, including reviews preparatory to research, research involving health records of deceased individuals, and revocation of consents and authorizations (Rothstein 2005a). Biorepositories Biorepositories are repositories of human biologic materials collected for research. According to the most recent and widely cited estimate, more than 300 million specimens are stored in the United States, and the number grows by 20 million per year (Eiseman and Haga 1999). Because the ability to access large numbers of well-characterized and annotated samples is an essential part of new genomic research strategies (see Chapter 10), biorepositories are increasingly important (Andrews 2005; Clayton 2005; Deschênes and Sallée 2005; Janger 2005; Knoppers 2005; Majumder 2005; Malinowski 2005; Palmer 2005; Roth- stein 2005b). Although biorepositories raise numerous issues involving research ethics and social policy, three issues are especially important to toxicogenomics, pharmacogenomics, and related research. First, biorepositories collect samples and data for unspecified, unknown, future research and thereby differ from traditional biomedical research, which generally contemplates a single, discrete project for which a single process of informed consent is used. Thus, the question that arises is whether it is permissi- ble under the Common Rule for researchers to obtain informed consent for fu- ture uses of specimens collected by a biorepository. Under the Common Rule, a research subject may consent to future, unspecified research. Nevertheless, IRBs generally do not approve “blanket consent” for all possible research on the grounds that consent about unknown future uses by definition is not “informed” (Strouse 2005). Using “tiered” or “layered” consent, research subjects are pre- sented with a menu of possible future research categories, such as cancer re- search, AIDS research, and genetic research. They can then indicate for which types of research it is permissible to use their sample (NBAC 1999). Because of the growth of biorepositories and new methods of informed consent, researchers need to monitor the understanding of research subjects and the effectiveness of the informed consent process to determine whether changes or improvements in the informed consent process are needed. Under the HIPAA Privacy Rule, however, each new research project re- quires a new authorization. Thus, tiered consent does not satisfy the Privacy Rule. Although researchers can obtain a waiver of the authorization requirement (meaning that the researchers do not need to recontact each sample donor for each new research project) (45 C.F.R. §164.512[l][2][ii]), a separate waiver from an IRB or Privacy Board is needed for each proposed new use. Therefore, the inconsistency between the Common Rule and the Privacy Rule complicates 

176 Applications of Toxicogenomic Technologies compliance for researchers, especially when toxicogenomic research may in- volve large repositories of individually identifiable specimens that may be used for multiple research projects (Rothstein 2005a). A second key ethical issue about biorepositories concerns if or when re- searchers have a duty to recontact sample donors to notify them that research using their samples has identified health risk information of clinical or social significance to them or their relatives (Rothstein 2005b; Bookman et al. 2006; Renegar et al. 2006). If the samples are deidentified, it will be impossible to notify any individual of sample-specific research findings, but there still may be a duty to notify all sample donors of the overall conclusions of the study and the potential desirability of consulting a physician. The greatest challenge, however, involves individually identifiable samples. Because it is too late to develop a recontact policy after a discovery has been made, the most prudent course is for researchers to ask sample donors at the time of sample collection if they want to be recontacted in certain circumstances (for example, when there is a predictive test or therapy) and, if so, whether the contact should be through their physician or personally. A third set of issues on which there has been considerable debate in the bioethics literature, is who owns the property rights to tissue samples and whether researchers have a duty to undertake any type of “benefit sharing” with individual donors or groups of donors (Andrews 2005; Marchant 2005). For example, over and above informed consent, who ultimately controls the use, distribution, and disposal of donated tissue samples? In one of the first judicial decisions to address this issue, a federal district court recently held that the re- search institute, not the individual researcher or tissue donor, “owns” donated tissue samples (Washington University v. Catalona, 2006 U.S. Dist. LEXIS 22969 [E.D. Mo. 2006]). Some tissue donors, especially those who are members of identifiable groups such as Native Americans, are now insisting that they re- tain the property rights in their donated tissues as a precondition for providing samples for research (Marchant 2005). A related question concerns those instances in which research on biore- pository samples has led to commercially valuable discoveries: are sample do- nors entitled to some share of the proceeds (see Greenberg v. Miami Children’s Hospital Research Institute, Inc., 264 F. Supp. 2d 1064 [S.D. Fla. 2003]; Moore v. Regents of Univ. of Cal., 793 p.2d 479 (Cal. 1990)? There have been a few documented cases in which one individual’s sample has been extremely valuable in developing cell lines or other research products (Moore v. Regents of Univ. of Cal., 793 p.2d 479 [Cal. 1990]; Washington 1994). It is far more likely, how- ever, for analyses of numerous samples to be needed for development of any commercially viable finding. Whether it is ethical for individual research sub- jects to waive all their interests in the products of discovery remains unclear, although it is common for informed consent documents to so provide. Regard- less of the language in the informed consent document, the principle of benefit sharing would be satisfied if researchers declare in advance that they will set aside a small percentage of revenues derived from commercialization for dona- 

Ethical, Legal, and Social Issues 177 tion to a charitable organization that, for example, provides health care to indi- gent individuals with the condition that is the target of the research (HUGO 2000; Knoppers 2005). Intellectual Property The development, exploitation, and protection of intellectual property are important elements of contemporary research strategy. Many of the contentious intellectual property issues, such as publication, timing of filing, ownership, and licensing, arise “downstream,” after the discovery of a patentable invention. These issues are not unique to toxicogenomic and pharmacogenomic research. Nevertheless, because many toxicogenomic applications involve assays that produce large amounts of data, often involving large numbers of genes or pro- teins, obtaining intellectual property rights to all the materials used in an assay could be particularly burdensome and problematic for researchers and platform manufacturers in the field of toxicogenomics (NRC 2005b). An important, “upstream” issue, and one that needs to be addressed spe- cifically in the context of toxicogenomic and pharmacogenomic research, is whether to have an “open access” policy to raw samples and data and, if so, how open it should be and how it would work. Traditionally, both toxicologic and pharmaceutical research have been undertaken with closed access, with private companies maintaining proprietary control over their research until a patent has been filed or a company has decided to disclose information for some other rea- son. Increasingly, however, there is pressure to make preliminary data from government-funded (and even some privately funded) research more widely available to researchers in general. For example, the Human Genome Project (public), the Pharmacogenetics Research Network (public), and the SNP Con- sortium (private) have adopted the policy of making research findings available online promptly for other researchers to use. The Committee on Emerging Issues and Data on Environmental Contami- nants of the National Research Council held a Workshop on Intellectual Prop- erty in June 2006 at the National Academies. Among other issues, the workshop discussed the position of the National Institutes of Health (NIH) in support of immediate (or prompt) release of and free access to human genome sequence data. These policies, embodied in the so-called Bermuda Principles, have been endorsed by the International Human Genome Organization and other organiza- tions (HUGO 1997). They also have been applied more broadly, and they form the basis of the National Human Genome Research Institute’s policies of prompt or immediate disclosure of genomic information generated by NIH-supported research or collected in NIH-supported repositories. No standard policy for data release has been developed for toxicogenomic and pharmacogenomic research. In the absence of such a policy, the default po- sition is nondisclosure. Consequently, affirmative steps should be undertaken to 

178 Applications of Toxicogenomic Technologies ensure that toxicogenomic and pharmacogenomic data are promptly and freely made available to all researchers whenever possible. Conflicts of Interest Environmental, pharmaceutical, and occupational health research often is sponsored or undertaken by entities with an economic interest in the outcome of the study. Especially for sensitive research, such as genetic research involving vulnerable populations, concerns about the investigators’ actual or perceived conflicts of interest will be magnified. Transparency, informed consent, and external oversight are essential (Rothstein 2000a). Even with such precautions, conflicts may arise with regard to what actions to take on the basis of the study. ETHICAL AND SOCIAL ISSUES A number of ethical and social issues may apply to toxicogenomics. These issues include privacy and confidentiality, issues related to socially vulnerable populations, health insurance discrimination, employment discrimination, indi- vidual responsibility, issues related to race and ethnicity, and implementation. Privacy and Confidentiality Privacy has many dimensions. In the information sense, privacy is the right of an individual to prevent the disclosure of certain information to another individual or entity (Allen 1997). Increasingly, the bioethics literature also has recognized a negative right of informational privacy—that is, the right “not to know” certain information about oneself (Chadwick 2004). Toxicogenomics affects both of these aspects of privacy. Individuals may want to restrict the dis- closure of their toxicogenomic or pharmacogenomic information that was ob- tained in a research, clinical, workplace, or other setting. They also may not want to know of certain risks, especially when nothing can be done to prevent the risk of harm (for example, where there was past occupational or environ- mental exposure). Confidentiality refers to a situation in which information obtained or dis- closed within a confidential relationship (for example, physician-patient) gener- ally will not be redisclosed without the permission of the individual (Rothstein 1997). With regard to toxicogenomics, the most important relationship is the physician-patient relationship. Restrictions on redisclosure are an important ethical precept of health care professionals, and maintaining the confidentiality of health information is important in preventing intrinsic and consequential harm to individuals (Orentlicher 1997). Although toxicogenomic and pharmacoge- nomic information would be covered under the “genetic privacy” laws several states have enacted (NCSL 2005a), these laws afford limited and inconsistent protection. 

Ethical, Legal, and Social Issues 179 It should also be noted that DHHS is taking the lead in promoting wide- spread adoption of electronic health records and creation of a national infrastruc- ture to support a system of interoperable, longitudinal, comprehensive health records (NCVHS 2006). In such an environment, privacy and confidentiality protections are even more important because detailed health information can be disclosed to numerous sources instantaneously (Rothstein and Talbott 2006). Although it is important to safeguard the security of health records to pre- vent unauthorized disclosure of information, breaches of confidentiality through authorized disclosures may be an even greater problem. For example, in health care settings, not all employees (for example, billing clerks, meal service em- ployees, maintenance employees) need the same degree of access to patient health records. Thus, the degree of access needs to comport with the need to know about protected health information. “Role-based access restrictions” help to limit the scope of disclosure. In non-health care settings, individuals often are required to execute an au- thorization for disclosure of their medical records as a condition of employment or insurance. Development and application of “contextual access criteria” will help to limit these disclosures to the health information relevant to the purpose of the disclosure (for example, ability to perform a job) (Rothstein and Talbott 2006). Socially Vulnerable Populations Many genetic loci of actual or potential significance to toxicogenomics and pharmacogenomics have a differential allelic frequency in discrete subpopu- lations. Due to ancestral patterns of endogamy (marrying within a social group), migration, geographic isolation, and other elements of ancestral origins, some genetic traits have a higher frequency among subpopulations socially defined by race or ethnicity. Thus, there is great potential for stigma when an increased risk of a particular undesirable health condition is associated with a particular popu- lation group, especially when the group is a racial or ethnic minority in a soci- ety. These population groups are then said to be “vulnerable” with respect to a particular health condition. A socially vulnerable population also can be based on shared somatic mu- tations. For example, workers with an acquired mutation or biomarker based on a particular occupational exposure or residents of a certain area with toxic expo- sures who demonstrated the subclinical effects of a particular exposure may also be said to be socially vulnerable populations. These polymorphisms may not be phenotypically obvious and often do not correlate with race, ethnicity, and other traditional social categories. The notion of vulnerable populations based on toxicogenomics and phar- macogenomics raises numerous ethical concerns, including the ethical principles of justice, respect for persons, and beneficence. Because of the social risks to vulnerable populations from research, special efforts are needed to obtain infor- 

180 Applications of Toxicogenomic Technologies mation about community interests and concerns. Community engagement and consultation are essential elements of ethical research (Foster and Sharp 2002). In addition, if environmental or pharmaceutical exposures have already ad- versely affected a subpopulation or a subpopulation has been identified to be at greater risk, ethical considerations might require that vulnerable populations have fair access to health care for diagnostic and treatment measures; that steps be taken to prevent unfair discrimination in employment, insurance, and other opportunities; that there is adequate compensation for harms; that feasible reme- diation be undertaken; and that effective protections are in place for protecting privacy and confidentiality (Weijer and Miller 2004). Health Insurance Discrimination The leading concern among individuals at a genetically increased risk of illness is that they will be unable to obtain or retain their health insurance (Roth- stein and Hornung 2004). In the United States, individuals obtain health care coverage mostly through government-funded programs (for example, Medicare, Medicaid) or employer-sponsored group health insurance. A relatively small percentage (about 10%) of those with health coverage have individual health insurance policies. Although this group is the only one in which individual medical underwriting takes place, individuals who currently have group cover- age might be legitimately concerned that if they lost their job and their group health insurance, they would need to obtain individual health insurance. To address the concern about discrimination in health insurance, nearly every state has enacted a law prohibiting genetic discrimination in health insur- ance (NCSL 2005b). These laws have a fatal flaw: they prohibit discrimination only against individuals who are asymptomatic. If an individual later becomes affected by a condition, he or she is no longer protected by the law and is at risk of rate increases or even cancellation in accordance with general provisions of state insurance law (Rothstein 1998). Health insurance discrimination is very difficult to resolve. As long as a key element of our system of health care finance is individually underwritten, optional, health insurance, then it will contain less favorable access for individu- als at an increased risk of future illness. Meaningful reform would require major changes in health care financing. Employment Discrimination Employers have two concerns about employing individuals who are at a genetic risk of future illness. First, employee illness and disability of any etiol- ogy represents significant costs in terms of lost productivity, lost work time, high turnover, and increased health care costs. Genetic testing makes it possible to predict future risks and would make it more attractive for employers to ex- clude at-risk individuals from their workforces. To combat such discrimination, 

Ethical, Legal, and Social Issues 181 about two-thirds of the states have enacted laws prohibiting genetic discrimina- tion in employment (NCSL 2005c). These laws are seriously deficient because, although they make discrimination based on genetic information illegal, they do not prohibit employers from lawfully obtaining genetic information contained within the comprehensive health records that employers may require individuals to disclose as a condition of employment. Consequently, at-risk individuals, who might benefit from genetic testing, are often reluctant to do so because a future employer can lawfully obtain the results (Rothstein and Hornung 2004). Second, employers may be reluctant to hire or assign individuals to work where they are at an increased risk of occupational illness because of their ge- netics. In addition to the employers’ illness concerns listed above, occupational illness could result in additional costs based on workers’ compensation, compli- ance with the Occupational Safety and Health Act of 1970, personal injury liti- gation, and reduced employee morale. With regard to toxicogenomics in the workplace, is it lawful or ethical for employers to request or require workers to undergo genetic testing? See Box 11-1 for additional discussion on toxicoge- nomic research in the workplace. There are two ethical cross-currents in the laws regulating the workplace: paternalism and autonomy. Much workplace health and safety regulation may be characterized as paternalistic, including child labor laws, minimum wage laws, and the Occupational Safety and Health Act. A greater appreciation for worker autonomy in the workplace is reflected in the Supreme Court’s leading decision in International Union, UAW v. Johnson Controls, Inc. (1991). The Court held that the employer discriminated on the basis of sex, in violation of Title VII of the Civil Rights Act of 1964 (42 U.S.C. § 2000e), by excluding all fertile women from jobs with exposure to inorganic lead because of concerns that a woman could become pregnant and give birth to a child with deformities caused by maternal workplace exposures. According to the Court: “Congress made clear that the decision to become pregnant or to work while being either preg- nant or capable of becoming pregnant was reserved for each individual woman to make for herself” (499 U.S. at 206). A middle position between permitting employers to mandate genetic test- ing and prohibiting all employee genetic testing is that a worker who is currently capable of performing the job should have the option of learning whether he or she is at increased risk of occupational disease based on genetic factors. Op- tional genetic testing would be conducted by a physician or laboratory of the individual’s choice; the testing would be paid for by the employer; and the re- sults would be available only to the individual. The individual would then have the option of deciding whether to assume any increased risk of exposure (Roth- stein 2000b). Only if employment of the individual created a direct and immedi- ate threat of harm to the individual or the public would the employer be justified in excluding the individual based on future risk (Chevron U.S.A. Inc v. Echaza- bal 2002). See Box 11-2 for additional discussion on genetic screening in the workplace. 

182 Applications of Toxicogenomic Technologies BOX 11-1 Toxicogenomic Research on Workers Some employers that use toxic substances in the workplace have considered or have engaged in research on their workers to determine whether exposures are causing physiologic changes. Toxicogenomics increases the potential scope of research to include the possible effects of individual genetic variability. The Fed- eral Regulations for the Protection of Human Subjects, generally called the “Common Rule,” specifically provide: “When some or all of the subjects are likely to be vulnerable to coercion or undue influence, such as children, prisoners, preg- nant women, mentally disabled persons, or economically or educationally disad- vantaged persons, additional safeguards [should be] included in the study to pro- tect the rights and welfare of these subjects” (45 C.F.R. § 46.111[b]). Although research conducted by private employers is unlikely to be subject to federal regulations, the underlying principles of ethical conduct of research suggest that special steps are undertaken to prevent coercion and to safeguard the rights and welfare of workers. It has been suggested that the following guidelines, intended to supplement the Common Rule, should apply to medical research in- volving workers. These guidelines appear to have equal force when applied in the context of toxicogenomic research. 1. If possible, the research should be performed by a party other than the employer, such as a university, medical institution, nongovernmental organization, or government agency. 2. Employers and employees (including union representatives) should be involved from the beginning in developing all aspects of the study, including study design, recruitment practices, criteria for inclusion and exclusion, informed con- sent process, confidentiality rules, and dissemination of findings. 3. The sponsor of the research should be indicated to potential participants, and investigators should disclose any financial conflicts of interest in the research. 4. Individuals with supervisory authority over potential research partici- pants should not be involved in the recruitment process, and lists of research par- ticipants should not be shared with supervisors. 5. No inducements should be offered for participation in the research. 6. The informed consent process should make it clear to potential partici- pants that there will be no adverse employment consequences for declining to participate or withdrawing from the research; potential participants also should be informed whether treatment or compensation for injuries will be provided. 7. Research should be conducted, and results should be disclosed, in the least identifiable form consistent with sound scientific methodology. 8. If the investigators believe the findings will be of sufficient scientific va- lidity and clinical utility to warrant offering participants the opportunity to obtain their individual results, the participants should be advised of all the potential risks of disclosure, including any psychologic, social, or economic risks. 9. Reasonable steps should be taken to ensure the confidentiality of partici- pant-specific information generated by the research, including storing the informa- tion in secure locations separate from other employment records, destroying bio- logic samples that are no longer needed, destroying linking codes when they are no 

Ethical, Legal, and Social Issues 183 longer needed, and applying for a certificate of confidentiality from the Depart- ment of Health and Human Services when appropriate. 10. Investigators should take special precautions at all stages of the study if the research has the potential to adversely affect groups of individuals on the basis of race, ethnicity, gender, age, or similar factors. Source: Rothstein 2000a. Individual Responsibility To what extent should individuals have a responsibility to learn if they are at increased risk from certain exposures, to avoid those exposures, and to deal with the consequences of illness caused by those exposures? To what extent does society have a duty to respect the autonomy of individuals to decline to learn of possible increased risks or to accept those risks? To what extent is soci- ety obligated to provide special protections for such individuals and health care if they become sick? Several of these policy choices have been settled in the workplace setting for some time. Under workers’ compensation law, an employer “takes the em- ployee as is,” and therefore it is no defense that the individual was at an in- creased risk of illness (Perry v. Workers’ Comp. App. Bd. 1997). Under the Oc- cupational Safety and Health Act, health standards, at least in theory, are designed to protect all workers (29 U.S.C. § 655 [c][5]). Under HIPAA, em- ployer-sponsored group health plans may not charge employees different rates or have different levels of coverage based on their health status, including ge- netic predisposition (42 U.S.C. § 300gg-1[a] and [b]). Nevertheless, even with regard to employment, many issues remain to be resolved, including the role of individual variation in setting permissible exposure levels. It may be even more difficult to develop rules of general applicability be- yond the workplace setting. Both scientific evidence (for example, absolute risk, relative risk, severity, latency, treatability) and context-specific social concep- tions of reasonable conduct in the face of increased risk are likely to affect the development of public policy. For example, if members of a low-income family with a genetic predisposition to illness and with few options of places to live “voluntarily” agree to accept the increased risk of living in the vicinity of toxic emissions, it is doubtful that society would shift the moral blame from the pol- luter to the individuals harmed by the pollutants. On the other hand, if toxicoge- nomic testing indicated that a particular individual would be much more likely to suffer serious harm from exposure to tobacco smoke and the individual started smoking or made no effort to stop, it is likely that at least some of the moral blame would shift to the cigarette smoker (Christiani et al. 2001). 

184 Applications of Toxicogenomic Technologies BOX 11-2 Genetic Screening for Susceptibility to Chronic Beryllium Disease Beryllium is a lightweight, stable, and strong metal that is ideal for use in the aerospace, nuclear, electronics, and other industries. Workplace exposure to beryllium can result in chronic beryllium disease (CBD), a granulomatous lung disease occurring in 2% to 16% of exposed workers (Fontenot and Maier 2005). Researchers have identified an increased risk of CBD among workers with the allelic substitution of glutamate for lysine at position 69 in the HLA-DPB1 gene on chromosome 6. This marker, known as Glu69 P, has a population prevalence of 40% (McCanlies et al. 2003). The Brush Wellman Company, the leading commercial beryllium- processing company, adopted a pilot program of genetic screening for the Glu69 P marker in job applicants and employees before exposure to beryllium. The pro- gram featured the following elements: testing was voluntary, informed consent was obtained before testing, testing was performed at the employer’s expense by an independent laboratory under contract with the company, results were provided directly to the individuals, management received no individual test results, and counseling was available to workers upon request. The decision to accept em- ployment rested with the worker. Although the testing program for beryllium susceptibility incorporated the essential elements of voluntariness, confidentiality, and autonomy, the program had limitations that may be instructive in the development of similar screening programs (Holtzman 1996). First, the test’s positive predictive value was only about 11%. Second, the presence of the Glu69 marker is not the only risk factor for development of CBD and the absence of the marker may give workers a false sense of protection. Third, in any screening program, care must be taken to ensure that the individual’s decision to undergo testing is truly informed and voluntary and that confidentiality of participation and results is maintained. Fourth, screen- ing out workers from exposure, even indirectly, is no substitute for industrial hy- giene controls to ensure the health of all workers. Race and Ethnicity Concerns about race and ethnicity are subsumed under the heading of “so- cially vulnerable populations,” discussed above, but they are raised in distinctive ways by pharmacogenomics. All humans are genetically 99.9% alike. There is more genetic diversity within certain socially defined groups (for example, race, ethnicity) than between groups (Ossorio and Duster 2005). The individual and group differences, however, are measurable and may be significant. Among the traits for which subpopulation variance has been observed are those dealing with pharmacokinetics and pharmacodynamics (Rothstein and Epps 2001a). Although pharmacogenomics holds great promise for increasing drug effi- cacy and decreasing adverse drug events, there is a risk of “racializing” medi- cine. It is facile to rely on race as a surrogate for genotype, but the social cost of 

Ethical, Legal, and Social Issues 185 doing so may be high. “By heedlessly equating race with genetic variation and genetic variation with genotype-based medications, we risk developing an over- simplified view of race-specific medications and a misleading view of the scien- tific significance of race” (Rothstein 2003, p. 330). The controversy surrounding the FDA’s approval of the drug BiDil in 2005 for self-identified African Ameri- cans demonstrates the currency of the issue (Saul 2005). See Box 11-3. Implementation Issues Traditionally, pharmaceutical companies have searched for new block- buster drugs to provide effective treatments for common conditions, such as hypertension, hypercholesterolemia, and musculoskeletal pain. A single product is designed to be safe and effective for substantially all members of the adult population. With pharmacogenomics, however, the potential market is seg- mented by genotype and many different companies may develop products for each affected population. Although virtually every biotech and pharmaceutical company is using genomic technologies in researching new products, whether “personalized medicine” will prove to be a viable business model remains to be seen (Reeder and Dickson 2003; Royal Society 2005; Ginsburg and Angrist 2006). Once medications that target people with particular genotypes are devel- oped, it is not clear how widely they will be adopted. Targeted medications may eliminate the costly trial and error of current prescribing and dosing, but the substantial research and development costs almost certainly will result in more expensive medications. Whether private and public payers will include these therapies on their formularies will depend on several factors, including the con- dition being treated, the relative improvement in safety and efficacy over stan- dard therapies, and cost (La Caze 2005). Patient demand produced by direct-to- consumer advertising (Ausness 2002) and provider concerns about potential liability (Palmer 2003) undoubtedly will increase adoption of the new products. New technologies will also expand the responsibilities of physicians, nurses, pharmacists, and other health care providers to use genetic information in prescribing, dispensing, and administering medications (Rothstein and Epps 2001b). Professionals not only will need to integrate genetic information into their practices, they also will need to become competent in genetic counseling (Clayton 2003). In addition, even though genetic testing for medication purposes is more limited in scope than genetic tests for assessment or prediction of health status, the test results will still contain sensitive genetic information. For exam- ple, identifying someone as being more difficult or expensive to treat could lead to discrimination in employment or insurance. Consequently, pharmacogenom- ics could increase the demand for new protections for the confidentiality of health information. 

186 Applications of Toxicogenomic Technologies BOX 11-3 BiDil and Race-Linked Pharmaceuticals BiDil is a combination of two drugs, hydralazine and isosorbide (nitroglyc- erin), that have been available in generic form for decades. Researchers postulated that this combination of drugs would benefit individuals with end-stage cardiovas- cular disease. In 1997, the FDA rejected the drug (based on a lack of efficacy) after a trial in a mixed-race group, although a subgroup of African Americans appeared to show benefit. According to Nitromed, Inc. (which now has rights to BiDil), African Americans are more likely than white Americans to have decreased amounts of nitric oxide, and BiDil is a nitric oxide enhancer. This was the biologic hypothesis for an improved outcome in this subpopulation. Nitromed joined with the Associa- tion of Black Cardiologists to sponsor the African American Heart Failure Trial, which involved 1,050 patients at 170 sites. On July 19, 2004, the trial was halted because of the significant success of patients enrolled in the treatment arm of the trial. Subjects treated with BiDil demonstrated a 43% improvement in survival and a 33% reduction in first hospitalization. A patent was issued for BiDil on August 31, 2004. In June 2005, based on the assumption that BiDil has a differential efficacy, the FDA approved Nitromed’s new drug application for use only by self-identified African Americans. Even if there is a polymorphism with some role in this proc- ess, it is scientifically implausible either that the genotype would be found exclu- sively in this population group or that members of the group would have only this polymorphism. The BiDil approval is very controversial because race is being used as a crude surrogate for an assumed but unknown genotype. Genotyping ef- forts in different populations (for example, African, European, and Asian) have demonstrated that most human genetic variation is shared across human popula- tions, with different allele frequencies between populations. Moreover, most func- tional human genetic variation is not population specific (Hinds et al. 2005 and references therein). Thus, there is no scientific basis for using race as an indicator of genetic variability or as a surrogate for a specific genotype. According to one expert, “Race as a proxy for genotype should never count as a justification accept- able to regulatory authorities” (Ossorio 2004, p. 424). In contemplating more widespread application of race in pharmacogenomics, substantial scientific, ethi- cal, social, and legal issues need to be resolved (Kahn 2005). REGULATORY IMPLICATIONS Toxicogenomic data have numerous potential applications in environ- mental, pharmaceutical, and occupational health regulation (Rothstein 2000b; Baram 2001; Marchant 2003a,b; Salerno and Lesko 2004; Grodsky 2005; Galla- gher et al. 2006). Regulatory agencies such as the Environmental Protection Agency (EPA) and the FDA have already produced guidance documents on the incorporation of genomic information into their regulatory programs (see Box 

Ethical, Legal, and Social Issues 187 11-4). Some examples of potential regulatory applications of toxicogenomics and their implications are discussed below. Defining New Adverse Effects Many regulatory requirements are based on a finding of “adverse effect.” For example, national ambient air-quality standards set under the Clean Air Act are set at a level that protects against adverse effects in susceptible populations. Some subclinical effects (for example, changes in erythrocyte protoporphyrin concentration in blood) have been found to be adverse effects under this statu- tory provision and have required more stringent standards to prevent such ef- fects from occurring in exposed populations (Lead Industry Association v. EPA 1980). Furthermore, many EPA regulations for noncarcinogenic substances are based on the reference concentration (RfC) or reference dose (RfD). Some gene expression or other changes measured with toxicogenomic technologies may constitute new adverse effects under these programs and thus lower the RfDs and RfCs, resulting in more stringent regulations. Manufacturers of pesticides and toxic chemicals are required to notify the EPA of new scientific findings about adverse effects associated with their products. In pharmaceutical regula- tion, the detection of a biomarker that suggests an adverse effect may likewise trigger additional regulatory scrutiny or restrictions. Therefore, a critical issue in the regulatory application of toxicogenomics will be determining whether and when a change constitutes an adverse effect. Many changes in gene expression, protein levels, and metabolite profiles will be adaptive responses to a stimulus that are not representative or predictive of a toxic response. Other toxicogenomic changes may be strongly indicative of a toxic response. Therefore, it will be important to distinguish true biomarkers of toxicity from reversible or adaptive responses and to do so in a way that is transparent, predictable, and consistent for the affected entities (Gallagher et al. 2006). At least initially, phenotypic anchoring of toxicogenomic changes to well-established toxicologic end points will likely be necessary to identify toxi- cologically significant markers. Regulatory Decisions Based on Screening Assays The availability of relatively inexpensive and quick toxicogenomic assays that can be used for hazard characterization of otherwise untested materials of- fers a number of potential regulatory opportunities (Gallagher et al. 2006). Full toxicologic characterization of the approximately 80,000 chemicals in com- merce using a battery of traditional toxicology tests (for example, chronic rodent bioassay) is not economically or technologically feasible in the foreseeable fu- ture. Screening with a toxicogenomic assay, such as a gene expression analysis that classifies agents based on transcript profiles, might be useful to quickly 

188 Applications of Toxicogenomic Technologies BOX 11-4 Regulatory Agency Guidance Documents EPA: The EPA adopted its “Interim Guidance Policy” in June 2002 (EPA 2002). This Interim Policy states that the EPA expects to receive genomic data for various risk assessment and regulatory purposes and expresses the agency’s belief “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.” Specifi- cally, the EPA identifies potential applications of genomic data to include better understanding of the relationship between exposure, mechanisms of action, and adverse effects in risk assessment as well as in setting agency priorities, ranking chemicals for testing programs, and supporting possible regulatory actions. The EPA expressly states that it will accept and consider genomic data in making regu- latory decisions “on a case-by-case basis.” However, although “genomics data may be considered in decision-making at this time, these data alone are insufficient as a basis for decisions.” In other words, whereas the EPA will consider available genomic data as part of a weight-of-evidence risk assessment, it will not base regu- latory decisions solely on genomic data at this time. Two years later, the EPA elaborated on the potential applications of genom- ics for risk assessment and regulation purposes in a white paper issued in Decem- ber 2004 entitled “Potential Implications of Genomics for Regulatory and Risk Assessment Applications at EPA” (EPA 2004b). This 50+-page document identi- fied four major areas where the EPA could use genomic data: (1) prioritizing con- taminants and contaminated sites, (2) monitoring, (3) reporting provisions, and (4) assessing risk. The document describes potential, and in some cases actual, cur- rent applications of genomic data in these four areas. FDA: The FDA issued a “Guidance for Industry: Pharmacogenomic Data Submissions” in March 2005 (FDA 2005a). This guidance seeks to encourage the collection and use of pharmacogenomic data in drug development by providing the agency’s position on when and how pharmacogenomic data should be submitted to the FDA as part of the approval process for drugs or biologics, and how such pharmacogenomic data may be used in FDA regulatory decisions. The guidance sets forth “algorithms” describing submission requirements for investigational applications, unapproved marketing applications, and approved marketing applica- tions. Data on “known valid biomarkers” must be submitted to the FDA for each stage of approval, whereas probable biomarker data generally need to be submitted only at the final approval stage and not at the investigational approval stage for a drug or biologic. The guidance document recognizes that most pharmacogenomic data will be “exploratory” at this time and will not meet the definition of a valid known biomarker. The FDA guidance encourages the voluntary submission of such exploratory pharmacogenomic data and provides assurances that the agency will use such data only for learning purposes and not for regulatory purposes. The FDA has issued other draft guidance documents to support the development of pharmacogenomic data and products, such as its Drug-Diagnostic Co- Development Concept Paper (FDA 2005b) to encourage a drug and a related pharmacogenomic test to be developed and obtain regulatory approval together. 

Ethical, Legal, and Social Issues 189 screen chemicals for prioritizing substances for further investigation and possi- ble regulation (see Chapter 5). Another possibility is to require the manufacturer of a new chemical substance, or an existing substance that will be used in a new application, to include the results of a standardized toxicogenomic assay as part of a premanufacturing notice required under section 5 of the Toxic Substances Control Act (Marchant 2003a). Protection of Genetically Susceptible Individuals Many regulatory programs specifically require protection of susceptible individuals. For example, the Clean Air Act requires that national ambient air- quality standards be set at a level that protects the most susceptible subgroups within the population. Under this program, the EPA focuses its standard-setting activities on susceptible subgroups such as children with asthma. Recent studies indicate a significant genetic role in susceptibility to air pollution (Kleeberger 2003), which may lead to air-quality standards being based on the risks to ge- netically susceptible individuals (Marchant 2003b; Grodsky 2005). Regulations under other environmental statutes, such as pesticide regulations under the Food Quality Protection Act and drinking water standards under the Safe Drinking Water Act, may likewise focus on genetically susceptible individuals in the fu- ture, as might occupational exposure standards promulgated by the Occupational Safety and Health Administration. Likewise, pharmaceutical approvals may require considering and protecting individuals with genetic susceptibilities to a particular drug. The identification of genetic susceptibilities to chemicals, consumer prod- ucts, pharmaceuticals, and other materials raises a number of regulatory issues. One issue is the question of the feasibility of protecting genetically susceptible individuals. On the one hand, protecting the most susceptible individuals in so- ciety may be extremely costly, and perhaps even infeasible without major, for- midable changes in our industrial society. On the other hand, the concept of government regulators leaving the health of some individuals unprotected, who through no fault of their own are born with a susceptibility to a particular prod- uct or chemical, also seems politically and ethically infeasible (Cranor 1997). As more information on individual genetic susceptibility becomes available, regula- tors and society generally will confront difficult challenges in deciding whether and how to protect the most genetically vulnerable citizens in our midst (Grodsky 2005). A related set of issues will involve the extent to which we rely on societal regulation versus individual self-help in protecting genetically susceptible indi- viduals. All regulatory programs confront choices between societal regulation and targeting high-risk individuals (Rose 1994), but these choices will be heightened by the identification of genetic susceptibilities to specific products or substances. For example, should products be labeled to warn people with par- ticular genotypes to avoid using the product, just as diet sodas are labeled to 

190 Applications of Toxicogenomic Technologies warn people with the genetic disease phenylketonuria that the product contains phenylalanine? Should regulators approve for sale pharmaceuticals that are ef- fective only for individuals with compatible genotypes and are ineffective or even hazardous for other people? Under what conditions and with what types of warnings? Will workers with genetic susceptibilities to specific workplace haz- ards be precluded from working in those jobs, or will they be required to accept personal responsibility for any harms that result? In all these scenarios relying on self-help, how will susceptible individuals discover they carry a particular genetic susceptibility? How will the privacy and confidentiality issues associ- ated with such genetic knowledge be addressed? How well will people under- stand and adapt their behavior appropriately to information on their genetic sus- ceptibilities? Policymakers, scientists, regulators, and other interested parties will need to address these and other questions raised by new knowledge about genetic susceptibilities. Generic Regulatory Challenges Several generic issues will confront all regulatory programs that use toxi- cogenomic data. One issue will be ensuring that toxicogenomic data are ade- quately validated (see Chapter 9). Each agency has its own procedures and crite- ria for ensuring data quality. In deciding whether and when toxicogenomic data are ready for “prime time” application in formulating federal regulations, agen- cies must balance the risk of premature use of inadequately validated data versus the harm from unduly delaying the use of relevant data by overly cautious poli- cies. Agencies such as the EPA have been criticized for being too conservative in accepting new types of toxicologic data (NRC 1994). Another trade-off that agencies must face involves whether and when to standardize toxicogenomic platforms and assays (Gallagher et al. 2006). On the one hand, standardization is important in that it facilitates comparison between datasets (see Chapter 3) on different substances and ensures consistent treatment of different substances. On the other hand, standardization runs the risks of pre- maturely freezing technology before it has fully matured. A related issue is how regulatory agencies can encourage or require pri- vate parties to generate and submit toxicogenomic data. Many product manufac- turers are likely to be concerned that toxicogenomic markers may detect bio- logic effects from their products or emissions at lower doses that may lead to increased regulatory scrutiny. To address such disincentives, agencies have adopted different approaches to encourage data submission. The FDA has adopted an approach under which the voluntary submission of certain types of toxicogenomic data will be for informational purposes only to help regulators and regulated parties better understand toxicogenomic responses, with an assur- ance that the data will not be used for regulatory purposes (Salerno and Lesko 2004). The EPA has adopted an interim policy that it will not base regulatory 

Ethical, Legal, and Social Issues 191 decisions solely on genomic data, alleviating concerns that a new regulatory requirement may be imposed based solely on a toxicogenomic finding when no other indication of toxicity is present (EPA 2002). LITIGATION Toxic tort litigation involves lawsuits in which one or more individuals (the plaintiffs) who have been allegedly injured by exposure to a toxic agent sue the entity responsible for that exposure (the defendant) for compensation. Both plaintiffs and defendants are likely to seek to use toxicogenomic data for various purposes in future toxic tort litigation (Marchant 2002, 2006). Proving Exposure Many toxic tort claims involving exposures to environmental pollutants fail because the plaintiffs are unable to adequately demonstrate and quantify their exposure to a toxic agent. Individuals exposed to contaminated drinking water, hazardous chemicals in the workplace, or toxic releases from an envi- ronmental accident often lack access to objective environmental monitoring data that can be used to quantify their exposures. In such cases, courts often dismiss claims because the plaintiffs are unable to prove sufficient exposure with objec- tive data (Wright v. Willamette Indus 1996). Some courts have endorsed, in principle, the possibility of using genetic biomarkers (for example, chromoso- mal aberrations) rather than environmental monitoring data to demonstrate and even quantify exposure (in re TMI litigation, 193 F. 3d 613, 622-623 [3d Cir. 1999] cert. denied, 530 U.S. 1225 [2000]). Toxicogenomic data may be able to help prove or disprove exposure in appropriate cases. If a particular toxic agent creates a chemical-specific finger- print of DNA transcripts, protein changes, or metabolic alterations, those bio- markers potentially could be used to demonstrate and perhaps even quantify exposure. Alternatively, a defendant could argue that the lack of a chemical- specific toxicogenomic marker in an individual proves that there was not suffi- cient exposure. However, a number of technical obstacles and data requirements need to be addressed before toxicogenomic data can be used reliably in this manner (see Chapter 4). The chemical-specific attribution of the toxicogenomic change would need to be validated, as would the platform used to quantify the changes. The potential variability in expression between cell types and individu- als would also need to be addressed. Perhaps most significantly, the time course of toxicogenomic changes during and after exposure would need to be under- stood to correctly extrapolate an individual’s exposure history from after-the- fact measurements of toxicogenomic changes (Marchant 2002). 

192 Applications of Toxicogenomic Technologies Causation The other major evidentiary challenge that plaintiffs in toxic exposure cases must overcome is to prove causation. Plaintiffs bear the burden of proof to demonstrate both general causation and specific causation. General causation involves whether the hazardous agent to which the plaintiff was exposed is ca- pable of causing the adverse health effect the plaintiff has incurred. Specific causation concerns whether the exposure did in fact cause the health effect in that particular plaintiff. Courts generally consider these questions separately, and both inquiries frequently suffer from a lack of direct evidence, resulting in outcomes that are highly uncertain, speculative, and often unfair. Toxicoge- nomic data have potential applications for both general causation and specific causation. General causation determinations are impaired by “toxic ignorance” (NRC 1984; EDF 1997), in which valid scientific studies do not exist for many combi- nations of toxic substances and specific health end points. The lack of a valid published study evaluating whether the agent to which the plaintiff was exposed can cause the plaintiff’s health condition will generally bar tort recovery for failure to demonstrate general causation. Toxicogenomic assays that can reliably be used for hazard identification may provide a relatively inexpensive and quick test result that could be used to fill the gaps in general causation. For example, a gene expression assay that can identify a particular type of carcinogen might be used to classify a chemical as a carcinogen (and hence establish general causa- tion) in the absence of a traditional chronic rodent bioassay for carcinogenicity. Toxicogenomic data can also play a role in proving or disproving specific causation. Current toxicologic methods are generally incapable of determining whether a toxic agent caused an adverse health effect in a specific individual for all but so-called “signature diseases” that usually have a single cause (for exam- ple, mesothelioma and asbestos; diethylstilbestrol and adenocarcinoma) (Farber 1987). Lacking any direct evidence of specific causation, courts generally adju- dicate specific causation based on a differential diagnosis by an expert or based on epidemiologic evidence showing a relative risk greater than 2.0, which sug- gests that any individual plaintiff’s disease was more likely than not caused by the exposure under study (Carruth and Goldstein 2001). Such methods are very imprecise and prone to under- and overcompensation depending on the facts of the case. A toxicogenomic marker could provide direct evidence of specific cau- sation if an individual who develops a disease is shown to have the specific mo- lecular signature of toxicity attributable to a specific agent. Duty of Care Toxic tort litigation involves judgments about the duty of care by an actor who creates risks to those who may be injured by those risks. Toxicogenomic data could affect or shift those judgments about duty of care in a number of con- 

Ethical, Legal, and Social Issues 193 texts. For example, the finding that some members of the population may have a genetic susceptibility that makes them particularly sensitive to a product may impose new duties on the product manufacturer with regard to testing, labeling, and selling that product. Some individuals who alleged they had been harmed by the Lyme disease vaccine Lymerix brought lawsuits claiming that the vaccine manufacturer had a legal duty to warn vaccine users to obtain a genetic test for a polymorphism that allegedly affected the user’s propensity to develop serious side effects from the vaccine (Noble 2000). The litigation settled before a judg- ment was issued but likely contributed at least indirectly to the vaccine being removed from the market and was the first of what will likely be a new trend of plaintiffs claiming that a product manufacturer has additional duties to protect individuals genetically susceptible to their products. Alternatively, under the “idiosyncratic defense doctrine,” defendants may be able to argue in some cases that they have no duty to protect individuals with a rare genetic susceptibility to a product. Under this line of cases, courts have held that a product manufacturer can be reasonably expected to ensure the safety of “normal” members of the population and not individuals with an unusual susceptibility to the agent in question (Cavallo v. Star Enterprise 1996; Marchant 2000). The detection of toxicogenomic changes in exposed individuals using toxicogenomic technologies may also result in lawsuits seeking damages for an increased risk of disease or for medical monitoring. Historically, courts have been reluctant to award damages for an increased risk of disease that has not yet manifested in clinical symptoms, but courts in some states have recognized such a claim if the at-risk individual can sufficiently quantify a substantial increased risk of disease (Klein 1999). It is possible that toxicogenomic data could support such a claim (Marchant 2000). A related type of claim is for medical monitor- ing, now recognized by more than 20 states, which requires a defendant respon- sible for a risk-creating activity to pay for periodic medical testing of exposed, at-risk individuals (Garner et al. 2000). Toxicogenomic assays could be used to identify at-risk individuals who would be entitled to ongoing medical evalua- tions, or the assays could serve as a periodic medical test for people who in- curred a hazardous exposure. Damages Toxicogenomic data may also be relevant in the damages stage of toxic tort litigation. A defendant who is liable for a plaintiff’s injuries may seek to undertake genetic testing of the plaintiff to identify potential predispositions to disease that might have contributed to his or her disease or that might otherwise reduce the plaintiff’s life expectancy (Rothstein 1996). Either of these findings could reduce the damages the defendant would be ordered to pay to the plaintiff. Although the plaintiff’s genetic predispositions to disease may sometimes be pertinent (and may be used to benefit the plaintiff in some situations), there is a danger that defendants will undertake “fishing expeditions” in the plaintiff’s 

194 Applications of Toxicogenomic Technologies genome that may reveal sensitive and private personal information and poten- tially violate the plaintiff’s right not to know (Chadwick 2004). Legal, Policy, and Ethical Aspects of Toxic Tort Applications The many potential applications of toxicogenomic data in toxic tort litiga- tion raise a number of scientific, legal, policy, and ethical issues. One central concern is the potential for premature use of toxicogenomic data (Marchant 2000). Unlike regulatory agencies, which generally consider new types of data cautiously and deliberately, toxic tort litigants are unlikely to show similar re- straint. A toxic tort case is a one-time event often involving large stakes that, once filed, tends to move forward expeditiously toward a decision under a court- ordered schedule. Therefore, toxic tort litigants have every incentive to use any available data that may help them prevail, regardless of how well those data have been considered and validated by the scientific community. Premature use of toxicogenomic data should obviously be discouraged, but it is important to note that current scientific evidence on issues in litigation such as causation are often inadequate, and toxicogenomic data have enormous potential to make the resolution of toxic tort litigation more scientifically informed, consistent, and fair. The reliability of toxicogenomic data introduced in a court proceeding will be evaluated under the standards for admissibility of scientific evidence. In fed- eral courts and many state courts, the evidence will be evaluated under the crite- ria for scientific evidence announced by the U.S. Supreme Court in the 1993 Daubert decision (Daubert v. Merrell Dow 1993). Under that standard, the trial judge is to serve as a gatekeeper for scientific evidence to ensure that any such evidence presented to a jury is both relevant and reliable. The Supreme Court identified the following four nonexclusive criteria that courts can use to evaluate the reliability of scientific evidence: (1) whether the evidence can and has been empirically tested, (2) whether it has a known rate of error, (3) whether it has been peer reviewed and published, and (4) whether it is generally accepted within the relevant scientific field. These criteria are consistent with general criteria used to validate toxicologic tests and, if applied rigorously, should help ensure against the premature or inappropriate use of toxicogenomic data in toxic tort litigation. Specifically, a court might want to consider the following factors in decid- ing whether to admit toxicogenomic data in a toxic tort lawsuit. • Has the toxicogenomic response been shown to be associated with or predictive of a traditional toxicologic end point (for example, cancer, toxicity)? • Are there data showing that the observed toxicogenomic response is characteristic of exposure to the specific toxic agent at issue, with a similar time course of exposure as experienced by the plaintiff? 

Ethical, Legal, and Social Issues 195 • Have the data used or relied on for making the above determinations been published in peer-reviewed scientific journals? • Have one or more other laboratories replicated the same or similar re- sults under similar conditions? • Has the toxicogenomic platform used been shown to provide consistent results to the platform used in any other studies relied on? Ideally, all these questions would be answered affirmatively before toxi- cogenomic data were introduced into evidence. Given that it is not reasonable to impose greater barriers to the introduction of toxicogenomic data than other types of toxicologic evidence used in toxic tort litigation, data satisfying most of the above criteria would likely be sufficiently reliable to be admitted. The use of toxicogenomic data in toxic tort litigation raises a series of other issues. For instance, the capability of lay jurors to adequately understand and apply toxicogenomic data in their decision making will present challenges. The potential privacy and confidentiality issues raised by genetic testing of plaintiffs create another set of concerns. Protective orders that protect sensitive information used at trials from being publicly disclosed will be needed to help protect the confidentiality of personal genetic information. Product manufacturers may have concerns about future liabilities associ- ated with toxicogenomic biomarkers. For example, a manufacturer may sponsor a study that detects a biomarker of unknown toxicologic significance today but that years later may become established as a validated biomarker of toxicity. The use of such data to impose liability retroactively could raise fairness con- cerns and may deter manufacturers from conducting toxicogenomic studies. At the same time, shielding manufacturers from such liability could create the wrong incentives by deterring them from investigating the risks of their products and taking appropriate mitigation measures and may deprive injured product users from being compensated for their injuries. These countervailing factors demonstrate the complex and sensitive role that potential liability can exert on scientific research and applications of toxicogenomics. COMMUNICATION AND EDUCATION The increasing scientific capability of making individualized predictions of risk from toxic substances raises the important question of how this informa- tion will be communicated to individuals and the public. Such predictions, tai- lored for the individual, may be years away. But it is still important to commu- nicate with the public in advance about relevant scientific advancements to maintain public support for the science and to help educate members of the pub- lic for the day when many of them may have to make personal decisions based on their capacity to understand the customized predictions of the risk they face from one or more toxic hazards (NRC 2005c). A great deal of research is needed in this field. Among other things, we do not know how much information people 

196 Applications of Toxicogenomic Technologies want, in what form, their likely comprehension, their likely response, and the degree of variability in different socioeconomic and cultural groups. Neverthe- less, some basic risk-communication principles are instructive. Risk communication has been defined as “an interactive process of ex- change of information and opinion among individuals, groups, and institutions” (NRC 1989). Thus, the prescription for agencies, media, and others involved in risk communication is to abandon the traditional top-down, sender-based, “pub- lic education” model of risk communication (for example, launching a flurry of messages in an attempt to get the public to see things the way experts do). In- stead, they should favor an approach that emphasizes greater understanding of the emotional reactions, concerns, and motivations of a segmented public who can seek (or even avoid), process, and evaluate critically the risk information they encounter and who have varying desires and capacities to do so (Griffin et al. 1999). Indeed, “the degree to which individuals have the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions” has become the prime definition of “health liter- acy” and has been proposed as an essential component of health and risk com- munication programs (IOM 2004, p. 32). Thus, there are at least four key issues in risk communication: (1) suffi- ciency of information; (2) capacity of the individual or society to access, assess, and understand information; (3) emotional responses to risks; and (4) trust in scientific and mass media organizations that oversee communication channels (NRC 2005d). It is important to understand that effective risk communication requires multiple messages tailored to a particular audience. People interpret health risks in light of their everyday events and experiences, and the messages must be framed within a familiar context. Health disparities and environmental justice issues are of heightened concern in some communities. In all communities, a person’s emotional response to risk, such as worry, fear, anger, and hope, can determine risk perception as well as response to risk (e.g., Witte 1994; Griffin et al. 1999). The emotional response to risk is also a key factor in risk acceptabil- ity. Better understood and familiar risks may be more acceptable than dreaded, poorly understood risks of a lower magnitude. Complicating any discussion of risk communication is the public’s low level of scientific comprehension, low level of understanding probabilistic in- formation, and low level of understanding basic numerical concepts. Thus, risk communication, especially about predictions of individualized risk from toxic substances, must include upfront efforts designed to help the public improve their health and risk literacy. There are many dimensions to the translation of toxicogenomic and phar- macogenetic knowledge into health benefits. Research needs to focus on both individual-level and population-level educational and risk communication as- pects of translation into medical and public health practice (Kardia and Wang 2005). For example, critical research issues on an individualized level are exem- plified by research on genetic risk communication (Hopwood et al. 2003), in- 

Ethical, Legal, and Social Issues 197 formed consent processes (Geller et al. 1997), the decision-making process (Shiloh 1996), and provider knowledge and awareness of genetics (Suchard et al. 1999). These research arenas have direct practice implications for genetic education and genetic counseling (C. Wang et al. 2004), interventions incorpo- rating genetic information (Lerman et al. 1997), patient adherence to screening recommendations (Hadley et al. 2004), effectiveness of decision supports and aids (Green et al. 2004), and health care provider training (Burke et al. 2002). Researchers and practitioners in the field of health behavior and health education can play a pivotal role in integrating toxicogenomics into practice to improve the public’s health (e.g., Sorenson and Cheuvront 1993; C. Wang et al. 2005). Priority areas that are ripe for further exploration, understanding, and application include the following: (1) public and provider education about ge- netic information, (2) risk communication and interventions for behavior change, (3) sociologic sequelae of genetic testing, and (4) public health assur- ance and advocacy. An ecologic perspective should be considered when ad- dressing the educational and communication issues involved in applying toxico- genomics to decrease health risks. Many different types of stakeholders and practitioners should be considered, and, consequently, many levels of interven- tion and analysis should be pursued. CONCLUSIONS AND RECOMMENDATIONS Because this chapter presents a number of complex issues, the conclusion text precedes the corresponding, numbered recommendations. It is critical to ensure adequate protections on the privacy, confidentiality, and security of toxicogenomic information in health records. Safeguarding this information will further important individual and societal interests. It will also prevent individuals from being dissuaded from participating in research or un- dergoing the genetic testing that is the first step in individualized risk assessment and risk reduction. The potential consequences of disclosure of toxicogenomic information are greater with the growth of electronic health records. There is a lack of comprehensive legislation protecting the privacy, confi- dentiality, and security of health information, including genetic information. These protections are needed at all entities that generate, compile, store, trans- mit, or use health information, not just those affected by HIPAA. 1. Address the privacy, confidentiality, and security issues that affect the use of toxicogenomic data and the collection of data and samples needed for toxicogenomic research. 2. Role-based access restrictions should be used for the disclosure of health information in health care settings. Contextual access criteria should be developed and used for the disclosure of health information, pursuant to an au- thorization, beyond health care settings. 

198 Applications of Toxicogenomic Technologies Toxicogenomic research often uses large biorepositories and databases in anonymous, deidentified, linked, or identifiable forms as well as phenotypic data in health records. Inconsistencies between the Common Rule informed consent requirement and the HIPAA Privacy Rule authorization requirement burden and interfere with toxicogenomic research. 3. Consider approaches to harmonize standards for deidentification and informed consent and authorization under the Federal Rule for the Protection of Research Subjects (Common Rule) and the HIPAA Privacy Rule, to minimize unnecessary barriers to research while continuing to protect the privacy and wel- fare of human subjects. 4. DHHS should explore new approaches to facilitate large-scale biore- pository and database research while protecting the welfare and privacy of hu- man subjects. Subpopulation groups considered socially vulnerable based on race or eth- nicity, income, age, or other factors are at increased risk for discrimination, stigma, and other adverse treatment as a result of individualized toxicogenomic information. 5. In toxicogenomic research, especially involving or affecting socially vulnerable populations, special efforts should be made at community engage- ment and consultation about the nature, methods, and consequences of the re- search. 6. To minimize the risk of adverse impacts on socially vulnerable popula- tions from toxicogenomic research and implementation, access to adequate health care for diagnostic and treatment purposes will be critical and should be a priority for funding agencies and legislators. The decision to use toxicogenomic testing to learn about one’s individual risk should rest with the individual, including risk posed by the workplace set- ting. Employers have the primary responsibility, under the Occupational Safety and Health Act, to provide a safe and healthful workplace and, under the Ameri- cans with Disabilities Act, to provide nondiscriminatory employment opportuni- ties and reasonable accommodations for individuals with disabilities. 7. When toxicogenomic tests to provide individualized risk information have been validated, individuals should be able to learn of their particular risk from workplace exposures without the information being disclosed to their cur- rent or potential employer. Upon learning of their individualized risk informa- tion they should be able to decide for themselves whether to accept the risk of such employment. 8. Only in extraordinary circumstances, when employing an individual with a risk would create a direct and immediate threat to the individual, cowork- 

Ethical, Legal, and Social Issues 199 ers, or the public, would the employer be justified in excluding an individual from a particular employment based on increased risk. Toxicogenomic data have many promising applications in regulation and litigation. These data have the potential to help fill many of the scientific uncer- tainties and gaps regarding exposure, causation, dose response, and extrapola- tion that currently limit the toxicologic knowledge critical for making sound regulatory and liability decisions. 9. Although caution, scrutiny, and validation are required to protect against premature, inappropriate, and unethical use of toxicogenomic data in regulatory and litigation contexts, care should also be taken to ensure that a higher standard of proof is not imposed for toxicogenomic data relative to other types of toxicologic data used in regulation and litigation. 10. A regulatory agency or court should give appropriate weight to the fol- lowing factors in deciding whether to rely on toxicogenomic data:1 a. Has the toxicogenomic response been shown to be associated with or predictive of an adverse health outcome (for example, cancer, toxicity)? b. Has the specificity and sensitivity of the test been established to be within reasonable bounds? c. Are there data showing that the observed toxicogenomic re- sponse is characteristic of exposure to the specific toxic agent at issue, with a similar time course and level of exposure as experienced by the plaintiff? d. Have the data used or relied on for making the above determi- nations been published in peer-reviewed scientific journals? e. Have one or more other laboratories replicated the same or similar results under similar conditions? f. Has the toxicogenomic platform used been shown to provide results consistent with the platforms used in any other studies that were relied on? Risk communication is an essential component of translating toxicoge- nomic information into reduced health risks for the public. Currently, the gen- eral public, as well as health care practitioners, are ill-equipped to understand and use toxicogenomic information to alter adverse health outcomes. 11. Research is needed on how to communicate toxicogenomic risk infor- mation to the public using culturally and psychologically appropriate methods. 1 These factors could be operationalized through agency risk assessment guidelines in the regulatory context and through judicial precedent relying on authoritative sources, such as this report in the litigation context. 

200 Applications of Toxicogenomic Technologies 12. Educational initiatives are needed for vulnerable subgroups and the general public to raise awareness about toxicogenomic findings that can affect their health. 13. Educational initiatives need to be developed and implemented to pre- pare the medical and public health workforce to use toxicogenomic information. Finally, several areas of future research would address some of the issues raised in this chapter. These include the following: 14. Federal agencies should increase their support for research on the is- sues of ethical, legal, and social implications in applying toxicogenomic tech- nologies, including public attitudes toward individualized risk, social effects of personalized information on increased risks, and regulatory criteria for toxico- genomics. 15. The National Institute of Environmental Health Sciences (or other fed- eral agencies) should develop “points-to-consider” documents that identify and discuss the issues of ethical, legal, and social implications relevant to individual researchers, institutional review boards, research institutes, companies, and funding agencies participating in toxicogenomic research and applications.