The Neoliberal Regulatory State,Industry Interests,and the Ideological Penetration of Scientific Knowledge

Introduction

Since the 1970s, a core element of science and technology studies (STS) has been the sociology of scientific knowledge (SSK). Notwithstanding underlying debates about relativism, constructivism, and realism, scholars agree on the fundamental claim of SSK that social, political, and economic factors can shape the very content of scientific knowledge  (Abraham 1995, 1-35; Barnes 1974; Bloor 1984; Gieryn 1982; Millstone 1978; Chubin and Restivo 1983; van Zwanenberg and Millstone 2000; Webster 1991). In particular, MacKenzie (1981) found that, in late-nineteenth-/early–twentieth-century Britain, the content of statistical knowledge was influenced by (eugenics) ideology and the interests and goals of the professional middle class, rather than inexorably determined by mathematical logic and/or empirical discoveries.

This article follows that SSK tradition, together with STS’s long-standing concerns about governance of technological risk and the newly emerging field of “STS and neoliberal science” (Bal and Halfman 1998; Bloomfield and Doolin 2011; Braun and Kropp 2010; Lave, Mirowski, and Randalls 2010; van Zwanenberg and Millstone 2005; Siegel Watkins 2011). Consistent with Lave, Mirowski, and Randalls (2010, 659), who “urge STS scholars to undertake detailed exploration of exactly how the external political–economic forces of neoliberalism are transforming technoscience,” we trace the links between neoliberal ideology, the interests of the pharmaceutical industry and drug regulatory agencies, and the content of toxicological knowledge about pharmaceutical carcinogenicity in the last thirty years in the United States and Europe. Specifically, we are concerned with what counts as knowledge when identifying and categorizing new pharmaceuticals as carcinogenic risks to humans for purposes of regulatory screening. The STS scholars, Gillespie, Eva, and Johnston (1979), Brickman, Jasanoff, and Ilgen (1985), Abraham (1993), and Abraham and Davis (2007a) also investigated carcinogenic risk assessment of chemicals/pharmaceuticals, but their emphasis was on international comparisons of different countries’ regulatory assessments of case-study chemicals, and was limited to before the 1990s. Contradistinctively, and like Winickoff and Bushey’s (2010), research on emerging global food regulatory standards, our main focus is more recent and on cross-national technoregulatory standards applied to carcinogenicity testing for all pharmaceuticals. Although Groenewegen (2002) examined how the direction of Dutch toxicology altered according to industry demands, his research was about environmental/industrial chemicals, rather than pharmaceuticals and made no attempt to link toxicological knowledge to the regulatory state. Similarly, Frickel (2004) and Shostak (2005, 2007) focused on the “geneticization” of American environmental toxicology, though Shostak (2007) analyzed the “translation” of such geneticization on to regulatory agencies’ agendas in the United States.

In the pharmaceutical sector, we argue that neoliberalism as a political movement and ideology has redefined the regulatory state to have much greater convergence of interests and goals with the drug industry than previously, particularly regarding acceleration and cost reduction of drug development and regulatory review. Consequently, drug regulatory agencies have accepted and supported the pharmaceutical industry’s agenda of streamlining new drug testing, including carcinogenicity evaluation requirements. Although pharmaceuticals are the focus of this article, neoliberal enhancement of industry influence over carcinogenicity evaluation since the 1980s has been reported in other fields, such as tobacco and environmental chemicals (Ong and Glantz 2000; Huff and Tomatis 2002). For example, the tobacco and chemical industries have successfully influenced that evaluation process at the World Health Organization’s expert body, known as the International Agency for Research on Cancer (IARC)—apparently without any condemnation from Member State regulatory authorities (Cook and Bero 2006). In the United States, neoliberal conservatives in the Supreme Court and Congress curbed the Food and Drug Administration’s (FDA’s) ambitions to regulate the tobacco industry, while chemical manufacturers supported by the George W. Bush Administration effectively halted the Environmental Protection Agency’s integrated carcinogenic risk assessment reports on the grounds that they were too “alarming” to the public (Hilts 2003, 292-94; Furlow 2011).

Unlike many reports of neoliberal and business influence on other regulatory sectors, we provide an in-depth analysis of the mechanisms of such influence, and how they relate to the techno-scientific knowledge claims in the pharmaceutical sector. The pharmaceutical industry has been permitted to set the agenda about how shorter term and cheaper alternative carcinogenicity testing systems are investigated for validity. In particular, we contend that, with the tacit approval of the neoliberal regulatory state, the commercial interests of the pharmaceutical industry framed the process and interpretation of validating these new test systems, thereby influencing what counts as knowledge about the carcinogenic status of new pharmaceuticals. Such framing was not determined by the technoscientific logic of toxicology or a lack of prevailing professional standards of validation, but rather by the sociopolitical goals of the controlling institutions. Indeed, a different validation process could have been conducted had the priority been to develop carcinogenicity testing in the interests of public health protection, especially because validation as a professional standard among scientists was well advanced when the validation process was undertaken.

To investigate the development and validation of these alternative short-term carcinogenicity tests involving genetically engineered rodents, we employed documentary and interview research methods. Fieldwork in Europe and the United States spanned three years from the mid-2000s. The documentary research included an extensive review of the technical literature regarding pharmaceutical toxicology since 1950 and the Web sites of all relevant institutions/organizations. For example, PubMed (1965-2010) was searched electronically using keywords, “carcinogenicity testing,” “risk assessment,”  “transgenic animals,” and “genetically altered mouse models”; and Scrip (1998-2010)—the twice-weekly pharmaceutical trade newsletter—was searched manually. E-mail notifications of latest developments took the literature review beyond books and articles to ongoing organizational practices via consultation documents, letters, notices of meetings, newsletters, and press releases.

Interviewees were mainly “informants” who had either specialist knowledge of, or some involvement with carcinogenicity testing/regulation of pharmaceuticals. Fifty-three interviews were conducted with informants from the pharmaceutical industry, drug regulatory bodies and their advisory committees, academic and government sciences institutions, public health advocacy groups, and patient organizations. Lasting 1-2 h, interviews were semistructured to allow probing and impromptu follow-up inquiries, as well as preplanned core questions. All were tape-recorded and transcribed, except for three when contemporaneous field notes were employed. The data were coded and analyzed independently by two researchers.

Neoliberal Ideology and New Regulatory Goals

Before new pharmaceuticals are permitted on to the market, drug firms must test them for safety according to government regulations. After pharmaceutical companies submit their test results to the appropriate regulatory agencies, government scientists review the results before government decides whether to permit the drugs on to the market. Those basic principles of pharmaceutical regulation remain, but the United Kingdom led the world in the neoliberal ideological reconstruction of them.

The commercial interests of the pharmaceutical trade are as old as capitalism itself (Abraham 1995, 36-86; Braithwaite and Drahos 2000, 360-98; Liebenau 1981). Concern that industry might have temporarily captured government regulatory agencies is as old as the legislation that created them (Abraham 1995, 36-86; Mitnick 1980). Nonetheless, in the decades before neoliberalism there was a governmental and legislative expectation that the basic goal, and indeed raison d’etre of pharmaceutical regulation was to protect public health over and above the commercial interests of pharmaceutical firms, albeit too frequently with lacklustre conviction. However, the neoliberal shift, which began in, and has persisted since, the 1980s changed that expectation into enrolment of the state in the service of industry’s ever-expanding appetite for increased market access and profits (Lave, Mirowski, and Randalls 2010).

The UK election of Prime Minister Thatcher in 1979 and her New Right Conservative Party in three subsequent elections marked the beginning of the neoliberal shift. The Thatcher government was sympathetic to the pharmaceutical industry’s claims that state regulation was insufficiently responsive to the needs of business and innovation because it did not approve new drugs on to the market fast enough (Abraham and Lewis 2000, 60-65). Consequently, a reform of the British civil service, including the Department of Health’s state-funded pharmaceutical regulation, was instigated. This created the UK Medicines Control Agency (MCA), which came to be entirely funded by fees paid by pharmaceutical firms for the regulatory work it conducted in exchange for a more “efficient service” by which was meant “lighter touch” regulation and faster drug approvals (Anon. 1988a, 1989a). On arrival, the new director of the MCA, who came from the pharmaceutical industry, stated that his goal was to reduce “processing times” for new drugs by 24 percent in the first year, even though he did not know the scientific quality (regarding safety or efficacy) of the new drug applications for that year (Anon. 1989b). With applause and large salary bonuses for the MCA’s executives from government Ministers, the goals of the UK drug regulatory agency were reset to emphasize its “dual responsibilities” to industry and patients (MCA 1991). Between 1989 and 1998, new drug processing times fell by more than two-thirds, from 154 working days to just 44—results for which the pharmaceutical firms paid handsomely, making the MCA one of the richest regulatory agencies in Europe (Abraham and Lewis 2000, 65-66).

Similarly, in Sweden, the Government created the Medical Products Agency (MPA) in 1990 in response to industry complaints that drug approval times were too long. This removed pharmaceutical regulation from the state’s National Board of Health and Welfare. Like the MCA, the MPA was a semiautonomous drug regulatory agency formed to encourage consultation with industry and accelerate drug approval times in exchange for high fees from industry, which entirely funded the Swedish agency (Anon. 1990). As in the United Kingdom, between 1989 and 1993 approval times fell sharply, by more than half (Anon. 1994a).

In Germany too, the election of the neoliberal Christian Democrats in the 1980s made the Government more responsive to pharmaceutical industry complaints about regulation. In 1987, pharmaceutical firms brought about thirty court cases against the German drug regulatory agency, the Bundesgesundheitsamt (BGA), for “failing to act” on drug applications (Anon. 1989c). The BGA argued that approval rates were not always a high priority because of the poor quality of new drugs, noting that in 1988 only two of sixty-two new drugs approved were of outstanding therapeutic significance (Anon. 1989d). Despite the BGA’s protestations, after continuing criticism, it was disbanded in 1994 and replaced with the Bundesinstitut fur Arzneimittel und Medizinprodukte (BfArM), another semiautonomous regulatory agency, instructed by the German Ministry of Health that it “could not continue to be so conservative [as the BGA] in its approach to approval of new products” (Anon. 1994b). By 1995, drug approval times in Germany were cut in half and regulators had established extensive mechanisms of consultation with pharmaceutical companies to meet industry needs (Zahn 1995).

Thus, these neoliberal changes occurred in three countries with significant pharmaceutical industries and drug regulatory agencies in Europe. The shaping of the European Union (EU) supranational drug regulatory system, which was becoming established during the early and mid 1990s, was strongly influenced by these neoliberal developments taking place in many of its most important member states for the pharmaceutical sector (Abraham and Lewis 2000). In 1995, the supranational European Medicines Evaluation Agency (EMEA) and concomitant regulatory systems were put in place. The EMEA was to be largely funded by fees from pharmaceutical companies—initially 28 percent in 1995, but rising to about 70 percent by 1999 and thereafter (Abraham and Davis 2007b, 397). Moreover, supranational EU regulators were required to make approval decisions within strict timelines consistent with the reduction in regulatory review times that had been introduced in Germany, Sweden, and the United Kingdom (Abraham and Lewis 2000). The supranational EU drug regulatory system, therefore, reflected the neoliberal framework that had come to dominate its most influential member states in the pharmaceutical sector.

Meanwhile, in the United States, in 1981 probusiness Reagan became President and the Republicans gained control of the Senate against a background of reported deterioration in the international competitive position of US industries (NAS 1983). Believing that America’s economic and industrial decline resulted from excessive government interference with the private sector, the New Right within the Reagan Administration and Congress committed to radical deregulatory agendas (Hilts 2003, 210-11). Reagan’s health policy adviser, formerly president of the US Pharmaceutical Manufacturer’s Association, invited the pharmaceutical industry to propose FDA reforms (Anon. 1981a). Congress convened a Commission to address “regulatory overkill at the FDA” (Anon. 1981b, 10). The Commission recommended reforms designed to promote more rapid approval of new drugs, which the FDA management implemented by narrowing the scope of the FDA’s regulation of industry data and allowing closer and more frequent contact between industry and FDA officials (Anon. 1982a; FR 1985, 1987). Some FDA scientists expressed concern before a Senate Committee that such management pressure to expedite approvals might “compromise the scientific integrity of [new drug] reviews” (Anon. 1981c, 1982b). Meanwhile, Reagan’s Secretary of State for Health praised the “new spirit” of FDA-industry cooperation (Anon. 1986).

Under Reagan and Bush (senior), the President’s Task Force on Regulatory Relief and the White House Council of Competitiveness under Vice President Quayle exerted continual pressure on the FDA throughout the 1980s and early 1990s to remove regulatory barriers to industry products by “streamlining” the drug approval process (Anon. 1991, 1988b). However, because the FDA had been subject to severe budgetary cuts during the Reagan and Bush (senior) Administrations, it could not reduce its approval times sufficiently to meet industry demands (Hilts 2003, 255). Neither the Bush (senior) Administration nor Congress were willing to increase its budget, so the pharmaceutical industry was permitted to partly fund the agency via fees under the 1992 Prescription Drug User Fees Act (PDUFA; Anon. 1992). Conditions of PDUFA included that the industry fees could only be used to fund/accelerate the FDA’s drug review process, and that the fee payments would only be reauthorized after every five years if the pharmaceutical industry and Congress could agree on specific performance goals for the FDA in each period (USGAO 2002, 7). These terms further accentuated the formation of a neoliberal regulatory state, by allowing the industry de facto influence over the FDA’s priorities and goals. Between 1993 and 2003, the extent of funding for the FDA from industry fees rose from 25 to 50 percent, while the agency cut its drug approval times by half in the same period (Abraham and Davis 2007b).

It was in this context that the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) was formed in the early 1990s. It comprises the three pharmaceutical industry associations and government regulatory agencies of the EU, United States, and Japan. The ICH sought to harmonize, streamline, and reduce pharmaceutical testing requirements as means of accelerating both drug development and regulatory review times across the three regions, which made up about 85 percent of the global market (Abraham 2009). This was in the commercial interests of industry because it reduced drug development costs and could accelerate product access to markets, but it would not be in the interests of patients and public health if it compromised safety checks and increased risks of drug toxicity/injury. Entirely consistent with the neoliberal drug regulatory states that had emerged in Europe and the United States, regulators from the three agencies willingly participated in the ICH project of reducing drug testing, which was to be enormously influential on global technoregulatory standards for pharmaceuticals. Furthermore, throughout the 1990s, the regulatory agencies permitted the industry to set the agenda of the ICH, whose secretariat was the International Federation of Pharmaceutical Manufacturing Associations (Abraham and Reed 2001). Gripped by neoliberal ideology, the ICH project could be construed by managers of regulatory agencies as in the interests of their institutions because with fewer/shorter tests, there would be less data to review and a concomitant reduction in their workload to meet ever-shrinking drug-approval-time goals. A significant target for streamlining at the ICH was the expensive and time-consuming process of carcinogenicity testing. Thus, the pharmaceutical sector is highly congruent with Lave et al.’s (2010, 660) more general observation that, “while varying across national contexts,” “over the past 30 years,” there has been a “broad global movement towards neo-liberalism.”

New Ways of Testing Pharmaceuticals for Carcinogenicity

It is well established that exposure to some chemicals, including pharmaceuticals, may cause cancer in people. Hence, all newly discovered pharmaceutical entities could pose some carcinogenic risk. In order to assess that risk, since the 1960s, government drug regulatory agencies have required scientists in pharmaceutical firms (or their contract research organizations) to test their new drugs for carcinogenicity before deciding whether to approve them on to the market (WHO 1969).

When testing new drugs for carcinogenicity, scientists have had to rely on the extrapolation of results from nonhuman tests because most carcinogenic risks accelerate over the life span—seventy to ninety years for humans—far too long for clinical trials (Schou 1992, 210). Two types of carcinogenicity test have been developed: short-term in vitro mutagenicity tests, and life span in vivo studies in rodents (Hayashi 1994, 291). The former involves adding the test drug to disembodied mammalian/human cells or to microorganisms in glass dishes (in vitro) to see if the chemical alters/damages DNA, causing mutations associated with carcinogenicity (King 1996, 93-94). Pharmaceuticals found to damage DNA are known as genotoxic carcinogens, However, many carcinogens cause cancer without damaging DNA as a primary biological activity (Purchase 1992). They are known as nongenotoxic carcinogens and are not detected by in vitro mutagenicity tests (Ashby and Tennant 1991).

To screen for nongenotoxic carcinogens, tests in whole live animals (in vivo) have been deployed. These animal models of human carcinogenesis have involved feeding rodents the test drug over (most of) their life span, usually between eighteen and twenty-four months. At the end of the study, the incidence and nature of the tumors found among the rodents given the test drug are compared with those in a “control group,” which do not receive the drug. Due to the problem of extrapolating findings to another species, and to humans in particular, the WHO (1969) recommended that these “life span” carcinogenicity testing should be conducted in at least two species (typically mice and rats) before government regulatory review, on the grounds that, if a drug caused cancer in more than one animal species (a trans-species carcinogen), then it would pose a greater carcinogenic risk to humans. By the late 1970s, the regulatory agencies in North America, Western Europe, and Japan had all adopted this view (Abraham 1998). Exceptionally, such rodent life span testing could be deferred to postmarketing of new drugs intended to treat life-threatening conditions to avoid delaying therapy for desperately ill patients (D’Arcy and Harron 1996, 258).

During the 1990s, the regulatory requirement of two life span studies in two rodent species began to be challenged within the ICH agenda. By 1998, the ICH had decided that two life span carcinogenicity tests (one in rats and one in mice) should no longer be a necessary regulatory requirement to screen for nongenotoxic carcinogens. As an alternative, the ICH proposed that only one life span carcinogenicity study (typically in rats) was necessary and that the second life span study (typically in mice) could be replaced with a much cheaper, smaller scale, and short-term study in genetically engineered rodents (typically mice), which would last approximately six months, involve about a third of the number of animals, and incur only a fifth of the cost of a life span study (Lumley and Van Cauteren 1997).

That these types of alternative tests were available as prominent alternatives was due to the “molecularization” of toxicology in the aftermath of the geneticization of cancer causation, together with the extent to which some government scientists, especially at the U.S. National Institute of Environmental Health Sciences (NIEHS) and the FDA, promoted and “translated” genetically engineered animal models for use in risk assessment and regulatory decision making (Dalpe et al. 2003; Frickel 2004; Shostak 2005, 2007). The FDA was able to push this new conceptualization of carcinogenicity testing through to acceptance at ICH because the agency is responsible for governing the largest pharmaceutical market—nearly half the world market—which all major drug companies seek to access. That institutional power within the global political economy of pharmaceuticals meant that the drug industry at ICH paid attention to the regulatory concepts preferred by the FDA, as they had done previously regarding the bioequivalence concept in the United States (Carpenter and Tobbell 2011). Such was the influence of the ICH that the proposal to replace one life span carcinogenicity study with a short term in vivo test involving genetically engineered rodents was transposed into pharmaceutical regulation around the world, especially in the EU, Japan, and the United States (EMEA 1997; Lubiniecki 1997; Pettit 2001).

The short-term in vivo tests proposed as alternatives to the life span carcinogenicity studies involved genetic manipulation to produce rodents, into which genes were introduced that were associated with tumor development (known as oncogenes), or in which genes thought to suppress tumor development, known as tumor-suppressor genes, were removed—“knocked out” (Tennant 1996). The technoscientific rationale for these short-term in vivo studies was that the early stages of tumor development, known as “initiation,” could be built into genetically engineered rodents so that carcinogenic effects could be detected in the whole live animal much sooner than in “normal” rodents because only the later stages of carcinogenesis would need to occur. On this logic, if a new drug were carcinogenic, then it should be detected fairly quickly by these short-term tests because the initiated animals should develop more cancer tumors more rapidly than the control animals (Schou 1992).

In practice, most genetically engineered rodent test models were mouse models, and the general expectation was that, under this new regulatory regime, the life span study would be in rats implying that the “second-species” short-term in vivo test would need to be in mice, albeit genetically engineered. Three types of genetically engineered mouse tests were outlined at the ICH, namely, those involving transgenic mice with the oncogene, v-Ha-ras, introduced (the tgAC model), “knock-out” mice with the tumor-suppressor gene, p53, removed (the p53 model), and transgenic mice with the oncogene, c-Ha-ras, introduced (the rasH2 model). While such mouse models had certainly been developed by 1998, they had not been validated. That is to say, their capability to correctly identify carcinogens and noncarcinogens had not been systematically assessed. Despite this, following the ICH, from 1998, regulatory agencies permitted pharmaceutical firms to conduct life span carcinogenicity studies in just one rodent species so long as they also employed one of these new short-term in vivo tests appropriately, even though it was not known whether those new tests could adequately screen for nongenotoxic carcinogens. Insofar as that new testing regime took on a life of its own, the FDA’s institutional power could be regarded as having been partly converted into what Carpenter (2010, 394) calls “conceptual power.” However, it was neither unchallenged, nor even secure, conceptual power because of lack of validation.

For example, the UK government expert advisory Committee on Carcinogenicity (UKCoC) concluded that this regulatory change “lacked appropriate scientific rigour to justify its use as a working document in the provision of information to support regulatory decisions” (Department of Health 1997, 113). One UK regulator asserted that “it was bad science to include an unvalidated assay as an alternative,” ¹ while a senior academic scientist in the field considered validation to be “absolutely essential” because “one should never introduce into regulation a completely novel procedure of such a major type without validating it—both to show the intrinsic worth of the method and to make sure there are enough laboratories around the world with experience of the procedure.” ² Even some ICH experts acknowledged that “conclusions cannot be drawn [about the utility of the new short-term tests] until the results of validation studies are obtained” (MacDonald 1998, 272).

The Meaning and Increasing Significance of Validation as a Professional Scientific Standard

After these changes to the regulations for carcinogenicity testing, “validation” studies were conducted during the 2000s, primarily in the United States with some input from European scientists. Before considering those “validation” studies, it is necessary to examine the prevailing technoscientific and regulatory standards regarding validation processes that were available in the United States and Europe. This puts in context the significance attached to validation as both a principle and a process in regulatory science.

In the United States, efforts to develop professional scientific standards for validation of regulatory science have taken many forms. For instance, one well-documented idea is the creation of the Health Effects Institute (HEI) in 1980, which became responsible for sponsoring research on air quality standards and the toxicology of air pollutants. The underlying principle, which guided the formation of the HEI, was that the validity of regulatory science could be achieved by removing the direct funding decisions for toxicological research from government and industry. Rather, the HEI, an early “public–private partnership” received half of its funding from the government’s EPA and half from the automobile industry (Jasanoff 1990, 209-16). Hence, the HEI provided a functional separation between the key stakeholders and research funding decisions affecting the interests involved—at least that was the claim of the Institute’s promoters, which received some support from an investigation of the validity of HEI’s research by the US General Accounting Office (GAO) in 1986. However, the GAO also found that the research had low utility for regulatory intervention, suggesting that the quest for neutrality between industry and government was not the most effective way to inform public-health protective regulation that might conflict directly with industry interests.

While the scientific community has grappled with the principles of validity pertaining to toxicological tests for decades, in the United States, prior to 1997, there was no formalized procedure for validating new test methods in regulatory agencies. The lessons of the HEI experience, especially in terms of limiting industry control of regulatory science research, were never formalized at the EPA or other federal agencies. The FDA’s Division for Toxicological Research (now known as the National Center for Toxicological Research) determined the validity of new test methods by “scientific consensus” involving “review by scientists having particular expertise in a particular field” (ICCVAM 1997, Appendix B, 66-75). This informal approach came under scrutiny after the US Congress passed the National Institutes of Health Revitalisation Act of 1993, which directed the NIEHS to “establish criteria for the validation and regulatory acceptance of alternative testing methods, and [to] recommend a process through which scientifically validated alternative methods [could] be accepted for regulatory use” (ICCVAM 1997, preface). Between 1994 and 2000, the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM) became established to take on this role for NIEHS with responsibilities across many federal regulatory agencies (ICCVAM 2003). Such institutionalization of scientific validation culminated in the ICCVAM Authorization Act of 2000, which stated that “each federal agency shall ensure that any new or revised acute or chronic toxicity test method, including animal test methods and alternatives, is determined to be valid for its proposed use prior to requiring, recommending, or encouraging the application of such test method” (Public Law 106-545, 4C).

In 1998, the US National Toxicology Program’s Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM) was formed to provide operational and scientific support for ICCVAM, whose efforts were further strengthened by the creation of the Scientific Advisory Committee on Alternative Toxicological Methods (SACATM) in 2004. These bodies review evidence and make recommendations, but do not themselves conduct validation studies (Schechtman 2002). Formally, each of the federal regulatory agencies, such as the FDA, remain autonomous in deciding whether they wish to accept and/or implement ICCVAM’s recommendations about new tests’ “potential usefulness, limitations and applicability to federal testing requirements” (Stokes et al. 2002, 27). Throughout the 2000s, the FDA’s position was that it would not necessarily withhold approval of a new pharmaceutical on to the market solely because some test methods have not been formally validated (SACATM 2004, section IX, discussion). Nonetheless, the scientific community’s growing commitment to the principle of validation in carcinogenicity testing is illustrated by the American Chemistry Council’s assertions in 2002 that “the US National Toxicology Program, through ICCVAM, is obliged to evaluate the transgenic mouse-model test methods in a formal validation program if it intends to routinely evaluate data from such assays as part of its carcinogen screening program” (Brozena and Becker 2002).

Meanwhile in Europe, the supranational European Commission established the European Centre for the Validation of Alternative Methods (ECVAM) in 1991 (Directive 86/609/EEC). Although ECVAM is very similar to ICCVAM, it differs in its greater emphasis to coordinate, manage, promote, and even develop/research alternative/new test methods, including those relating to carcinogenicity. ECVAM has looser connections with government regulatory agencies than ICCVAM, working instead with broad networks of laboratories, including the pharmaceutical industry (Balls 2002; van Zeller and Combes 1999). It has been more forthcoming than ICCVAM in stipulating criteria for validation processes, while less concerned about persuading regulators of the legislative imperative to integrate the scientific principle of validation into regulatory policy. Specifically, by the early 2000s, ECVAM had established criteria for validation studies, including “clarity of defined goals,” “quality of overall design,” “independence of management,” “independence of selection of test materials,” “independence of data collection and analysis,” “number and properties of test materials,” and “quality and interpretation of results” (Worth and Balls 2002, 18).

The Process of Validation

It was widely acknowledged, therefore, that the new short-term in vivo carcinogenicity tests using genetically engineered mice needed to be validated in order to have confidence that they could screen for carcinogens, especially nongenotoxic carcinogens. The lessons from the HEI in the United States combined with heightened professional concerns in both the EU and United States about validation of regulatory science might have suggested the need to separate industry influence from that confidence-building process. However, the dominance of industry interests, rather than public health protection, permitted by the neoliberal regulatory state, was to shape the validation process and deeply influence the construction of the technoscience involved, so that the new regulatory science and associated potential conceptual power could be harnessed to maximum commercial gain. Our investigation reveals that the choices about who should conduct the validation, and how the data were collected, analyzed, and interpreted were consistent with the pharmaceutical industry’s agenda, but might well have been different if the primary driver of validation had been the protection of public health.

Framing the Science as Screening for Risks to Industry

As Murphy (2001) found in tobacco regulation debates, such “affiliation bias” in expertise framed the content of risk assessments. The large involvement of industry scientists and many of the proponents of the alternative short-term in vivo mouse studies at the ICH had consequences for the “technical” approach and design of the ACT program’ in several key respects. For example, the basic methodology chosen was retrospective, rather than prospective. That is to say, the approach was to validate the genetically engineered mouse models using chemicals already found to be carcinogens or noncarcinogens in humans and/or life span rodent studies. However Tennant (1997, 240), one of the preeminent world experts in chemical carcinogenicity testing and transgenics suggested that:

A more objective approach [than the retrospective methodology of the ACT programme] is to assay chemicals that are currently undergoing long-term [lifespan] bioassays and to evaluate the results of both types of assays [lifespan and short-term in vivo studies] concurrently with the stated objective for the transgenic models of detecting trans-species carcinogens but not responding to non-carcinogens.

One senior US government scientist in the field explained why the ACT program rejected Tennant’s “more objective” prospective approach in favor of a retrospective validation methodology:

The consensus amongst particularly the pharmaceutical industry scientists was that they … primarily… felt obligated to take some pharmaceutical compounds that have had a lot of human exposure, that they knew by the personal experience and by epidemiology—should [be] negative, but often had turned out to be positive in some, if not all long-term [lifespan rodent] studies. Because the fear was that if the short-term [in vivo] studies would be particularly sensitive, and might show up to be positive, and if that had happened then they would have lost all confidence in trying to do a short-term [in vivo] study. ¹⁰

In other words, the primary concern that informed the methodological design was that the alternative short-term in vivo tests with genetically engineered mice might produce results that would propel scientists to designate even more pharmaceuticals as presenting carcinogenic risk than the life span mouse studies had done. Indeed, according to many scientists from academia, industry and government, including some closely involved with the ICH process, “the ILSI [ACT] study was an attempt by the industry to convince themselves that these assays [short-term in vivo tests] were not overly sensitive,” that is, they would not produce more results that industry scientists believed were false positives by “mis-identifying” (as industry saw it) noncarcinogens as carcinogens. ¹¹ The industry’s attitude to the ACT program was paraphrased by a former senior FDA scientist, who was involved in both the ICH and the ACT program, as: “let’s make sure these [short-term in vivo] assays don't start popping up findings that are worse than the two-year mouse and two year rat together.” ⁸ Clearly, this characterization meant “worse” for the commercial interests of the pharmaceutical industry, as was made explicit by another expert scientist, who elaborated on industry’s concerns as being that the new alternative short-term in vivo tests in genetically engineered mice might “throw off a lot of false positives, and cause a lot of problems in regulatory decisions [about future pharmaceutical products].” ⁷

A second consequence of industry ascendancy within the ACT program for methodological design was in the selection of compounds to be tested by the new short-term mouse models. The pervasive industry concern to check that the short-term in vivo tests did not produce more false positives than the rodent life span tests resulted in most of the compounds selected for the ACT program being noncarcinogens. As one senior scientist observed, this explained why the selection of drugs was “heavily weighted toward those which would be presumed to have no [carcinogenic] effect.” ⁷ In fact, of the twenty-one compounds used, only six were human carcinogens, while fifteen were noncarcinogens in humans. Yet, from the perspective of screening for carcinogens in order to protect public health, the priority is to have test systems that will be able to detect human carcinogens, so that regulatory agencies can then consider whether that risk is sufficient to prohibit human exposure to the drug, including the denial of marketing approval. Thus, the ACT program gave over twice as much attention to checking the validity of these short-term in vivo tests according to industry interests (not too many false positives regarding noncarcinogens) compared with the interests of public health protection (not too many false negatives regarding human carcinogens).

The commercial interests of pharmaceutical companies also affected the selection of drugs used because firms did not want any products with market value to be part of the ACT program in case the new mouse models indicated that they were carcinogens. According to one US regulator, it would have been scientifically interesting to investigate some pharmaceutical products that had been permitted on to the market despite previous positive carcinogenicity results in rodent lifespan studies, but added that that would “probably have alienated more people [meaning industry] than was necessary.” ¹² Consequently, in the methodological design of the ACT program, one of the inclusion criteria for the compounds selected was that they should be “non-proprietary” (Robinson and MacDonald 2001, 18). A former FDA regulator at the ICH, who had subsequently moved to a US pharmaceutical firm surmised that “if the regulatory authorities had had their freewill about which compounds would be in, there probably would have been a few different ones,” but significantly added that “the FDA went along with it—understanding that this was really [to let industry] become comfortable [with the new mouse models].” ⁸ The compounds finally selected for the ACT program are listed in Table 1. Even at an ACT workshop, many participants suggested that, from a scientific perspective, “a different suite of test chemicals from those used might have been more informative” (Pettit 2001, 194). More specifically, these findings imply that the ACT program could scarcely be said to meet the ECVAM standard for validation studies regarding “independence of selection of test materials.”

OPEN IN VIEWER

Table 1.

Compounds Assessed in the ACT Study

Class	Compound
(1) Evidence of tumorigenic activity in humans:
Genotoxic human carcinogen	Cyclophosphamide, melphalan, phenacetin
Immunosuppressant human carcinogen	Cyclosporin A
Hormonal human carcinogen	Diethylstilbestrol, estradiol
(2) Nongenotoxic rodent carcinogens putatively noncarcinogenic in humans:
Rodent carcinogens/putative human noncarcinogens (human data)	Phenobarbital, clofibrate, reserpine, dieldrin, methapyrilene
Rodent carcinogens/putative human noncarcinogens (by mechanisms)	Haloperidol, chlorpromazine, chloroform, metaproterenol, WY-14643 DEHP, sulfamethoxazole
(3) Noncarcinogens	Ampicillin, d-mannitol, sulfisoxazole

Interpretation of the Results of Validation

Although the ACT program involved only twenty-one compounds, some were tested in more than one short-term in vivo mouse model. Across the six known human carcinogens, involving thirty-two tests, the mouse models correctly identified these carcinogens in only seventeen (53 percent), produced nine false negatives (28 percent), and six (19 percent) of the results were equivocal. Hence, in nearly half of the cases, these animal models failed to identify human carcinogens, which could hardly inspire confidence in their capability to screen for the human carcinogenicity of pharmaceuticals in drug regulation. By contrast, the mouse models correctly identified 82 percent of compounds that were both human noncarcinogens and rodent carcinogens, and 100 percent of the compounds that were noncarcinogens in both humans and rodents (Eastin et al. 2001; Storer et al. 2001; Usui et al. 2001; van Kreijl et al. 2001).

After completion of the ACT program, the results were discussed by a panel of its expert scientists chaired by Professor Samuel Cohen, an American University Professor of Pathology and Oncology with extensive expertise in toxicology, a trustee of HESI, and a former advisor on the ILSI ACT Steering Committee. ⁴ Downer (2007) has noted that tests of reliability can contain irreducible ambiguities requiring judgments with real consequences. Our analysis of interpretations of the ACT program results shows how such judgments are structured by the power and goals of the social interests at stake.

On the whole, the pharmaceutical industry and regulators involved with the program found the results reassuring and became more comfortable about the prospect of using the short-term in vivo mouse models. ^8,15 One regulator, who had criticized the ICH for recommending use of these short-term mouse models in 1998 before validation, was similarly reassured:

we’ve actually looked at all these models, decided which ones we think are good enough for regulatory acceptance, and put forward how we think these should be applied, and general recommendations for study design. So the guidelines are there now. You couldn’t produce the guidelines before the studies were validated because they weren’t validated. ¹⁶

Apparently a method had been found to manage new drugs through the carcinogenic-risk-assessment part of the regulatory process that would lessen workload for regulators and result in faster and cheaper drug development for industry with no greater risk than before of denial of marketing approval by regulators because of carcinogen identification. With greater confidence that the new short-term in vivo genetically engineered mouse models did not produce “excessive” false positives, their introduction became much better aligned than before with the commercial interests of industry and the political interests of regulators, who, under neoliberalism, had redefined the mission of regulation as the rapid processing of drug development.

Goodman (2001), an academic scientist closely involved with ILSI, illustrated how these institutional interests could shape the very content of  scientific knowledge about carcinogen identification and definition. He considered two aspects of the short-term mouse models to offer a “level of comfort,” namely, that these models could be used to increase “confidence in a negative result [with a compound in a life span rat study]” or provide “the ability to place a positive result [in a life span rodent study] into proper context” (Goodman 2001, 174). In other words, there was reason for the pharmaceutical industry (and regulators sympathetic to industry) to be comfortable with the new short-term mouse models so long as they confirmed negative results in rats or made overall negative interpretations more likely by prompting reconsideration of positive findings in rats. Notably, the converse scientific possibilities, namely, the confirmation of positive results or the ability to put negative findings in “proper context,” which were contrary to industry interests, but might increase public health protection, were not mentioned within a “level of comfort.” As Winickoff and Bushey (2010) report regarding the internationalization of new standards for food risk regulation, judgments about the adequacy of frameworks for risk assessment embody values about public health and the priority given to its protection relative to industry interests.

Yet, as a close examination of the technoscience underpinning the ACT program reveals, there was comfort not because of high levels of confidence in scientific validity, but rather because the new short-term mouse models appeared to satisfy the interests of the dominant stakeholders—the pharmaceutical industry and the government regulatory agencies. For, as many scientists, including some within the ACT program, acknowledged, the evidence that these new short-term in vivo mouse tests offered any scientific improvement over, or could even adequately replace, the life span studies in mice as means to screen for potential nongenotoxic human carcinogens in the regulatory protection of public health was, at best, scant, especially given a mere 53 percent detection rate. Indeed, scientists in the ACT program Panel could not agree about how to interpret some unexpected positive carcinogenicity findings from the short-term mouse models. They disagreed as to whether an unexpected positive results in the model was the evidence of irrelevant oversensitivity (false positives) or additional information indicating that the test compound might truly be a carcinogen (Pettit 2001, 193).

Although the panellists “generally agreed that these new models might substitute for the two-year [life span] mouse bioassay,” even some of the ACT program scientists significantly qualified this point by specifying that the new short-term mouse models might have a “particular utility for identifying genotoxic compounds,” but “there was not a consensus that any individual model presents any particular mechanistic advantage over the traditional mouse strains for evaluating chemicals operating through non-genotoxic modes of action” (Pettit 2001, 191). Other scientists involved in the ACT program also alluded to the limitations of the alternative mouse models regarding screening for nongenotoxic pharmaceutical human carcinogens because the organ specificity of some of the models “[did] not appear to correlate well with potential carcinogenicity at specific sites in humans” which “will significantly limit the insight that can be developed with respect to mechanism of action for these compounds, particularly for chemicals that are not genotoxic” (Cohen, Robinson, and MacDonald 2001, 18). Particularly sceptical about the reliability of extrapolation from the mouse models to nongenotoxic human carcinogens, Cohen, Robinson, and MacDonald (2001, 188) noted that although the specific genes inserted or deleted in these mice “may have increased relevance to human carcinogenesis, [they did] not give a clear indication of mechanistic-relevant human carcinogenesis.” This is because, as one industry scientist acknowledged, “experimental evidence shows that [non-genotoxic] cancer is often secondary to some other biological effect” not reducible to genetic make-up. ¹⁷

Yet from the perspective of patient and public health protection, if the new mouse models were to replace life span studies in mice or all rodents, then the crucial issue was whether the new mouse models could be relied upon to detect nongenotoxic carcinogens because it was those carcinogens for which the life span studies were used to screen. As several scientists, including the UKCoC explained, the new short-term mouse models “best detected genotoxic carcinogens,” but that “in most cases” such carcinogens would be “detected in a standard battery of genotoxicity tests,” which have been used alongside life span studies for decades (UKCoC 2003, section 2i; Goodman 2001, 175). As one industry toxicologist outside the ACT program reflected:

With the benefit of hindsight, I don’t think the outcome of that [ACT programme] was worth the effort … because we are actually pretty close to where we were before it was done. It hasn’t helped me make decisions about what we are going to do with our drugs in development. … There are already studies for detecting genotoxicity but we haven't solved the problem of non-genotoxic carcinogens. ¹⁷

Despite these major reservations, Cohen, Robinson, and MacDonald (2001, 189) contended that the new short-term mouse models might provide “useful information in the overall evaluation of potential carcinogenic hazard to humans” and “provide the next step in our ultimate goal of being able to accurately predict human carcinogenesis for a given chemical and to do so without relying on long-term [lifespan] bioassays in any species.” The contradiction inherent in the ACT program experts’ optimism about how these new mouse tests could match their commercial and political interests in avoiding life span tests is encapsulated in their comment that “although they [the ACT program experts] consider the [new genetically engineered mouse] models scientifically valid, the current data set is not sufficient to validate the models” (Pettit 2001, 195). Such optimism infected some government regulatory agencies. For example, the EU’s supranational drug regulatory Committee for Proprietary Medicinal Products (2002), now known as the Committee for Human Medicinal Products, considered that these new genetically engineered mice “could be used for regulatory purposes as an alternative to the long-term mouse study in conjunction with a long-term [life span] rat study.”

Other expert scientists and toxicology bodies outside the ACT program were often much more forthright in their criticisms. The UKCoC (2003) noted “that these mice identify some genotoxic compounds” (section 12), but advised that “further development and validation of short-term in-vivo models to evaluate non-genotoxic carcinogenesis and tumor promoters may be valuable” (section 2iii). Overall, the Committee’s view was that there was not enough data from the ACT program “to really judge those questions” about nongenotoxic carcinogenicity.¹⁸ Focusing specifically on whether these new short-term genetically engineered mouse models could replace life span mouse studies, the Committee concluded that “none of the models” were suitable in that respect. Indeed, one regulator suggested that most scientists outside the ICH-ILSI-HESI-ACT nexus were not convinced by the new mouse models as carcinogenicity tests:

Overall—I think the majority of scientists were obviously quite disappointed with the end results—because we really would have liked to have a good model, and obviously we haven’t got [that]. ¹⁶

Conclusions and Policy Implications

Frickel (2004) and Shostak (2005, 2007) have shown how shifts in American toxicology of environmental chemicals toward toxicogenomics involved the “activism” of scientists promoting their genetically engineered animal models as testing systems relevant for risk assessment in the regulatory sphere, including the FDA. Regarding international pharmaceuticals regulation, that, together with the institutional power of the FDA in the global pharmaceutical political economy, helps to explain why the genetically engineered mouse models were available for “translation” into carcinogenicity testing. However, as Raman and Tutton (2010) note, such “molecularization” is only one dimension of biopolitics and scientific-knowledge formation. It does not explain why drug regulators were looking for alternative tests to translate in the first place, or why the reception, management, and interpretation of those alternatives took the form they did in the international pharmaceutical sector. To explain those things, one needs to understand why the government regulatory agencies collaborated with the industry’s agenda of streamlining drug testing, and carcinogenicity testing, in particular. One also needs to understand why the government regulatory agencies essentially gave the pharmaceutical industry a “free hand” in designing, managing, and interpreting the process of determining whether the alternative tests were valid for regulatory purposes of carcinogen identification and screening.

We have shown that the precise ways in which these phenomena occurred resulted from a neoliberal ideology, which radically altered the goals of drug regulatory agencies and their relationships with the pharmaceutical industry, culminating in regulator–industry interactions of the kind displayed in the ICH project and subsequently the ACT program. It was for these reasons that, on the watch of government regulators, attempts to determine the validity of short-term mouse models were shaped primarily by the commercial goals of the industry, rather than the interests of public health or logical imperatives internal to toxicological science. Indeed, the professional scientific standards regarding validation had never been clearer in their stipulations, but had no compelling effects on the ACT program, whose interest-driven practices clashed with those scientific principles. As Lave, Mirowski, and Randalls (2010, 660) have observed in other domains, “neoliberalism had shifted methods, organization and content of technoscience.”

Moreover, the interpretation of the short-term mouse models as valid replacements for the life span tests in mice in order to identify and screen for nongenotoxic human carcinogens is highly questionable, even by the standards of some of the ACT program scientists themselves. Yet, such validity is what is needed if the short-term mouse models are to maintain regulatory levels of public health protection. Nonetheless, the short-term mouse models have been introduced into pharmaceutical regulations, thus contributing to the “scientific knowledge” about whether or not new pharmaceuticals are carcinogens. The industry has achieved its objective of having the option to jettison life span carcinogenicity studies in mice in favor of cheaper and shorter mouse tests, which the industry and its allies in regulatory agencies may present as valid because the ACT program has shown that the short-term mouse models are not “overly sensitive” to noncarcinogens. Such validation indicated that the short-term mouse models pose small risks to the commercial interests of pharmaceutical firms, but provided little reassurance that they would not permit patients to be exposed to greater risks than before from undetected carcinogens that find their way on to clinical trials or the market.

This raises the question of whether the drug regulatory agencies should be characterized as “captured” by industry regarding “technoscientific” standards for carcinogenic risk assessment. Abraham (1995), who studied pharmaceutical regulation in the United Kingdom and the United States from the late nineteenth century to the mid-1980s, found that there had been capture in both countries in that period, though significantly more in the United Kingdom than the United States. However, overall, he concluded that UK and US drug regulation should be characterized as “corporate bias,” rather than capture because “industry interests played a key role in the establishment and evolution of regulation, and the state, under certain conditions, defined its own interests independently of the industry” (Abraham 1995, 85). Capture theory, therefore, tends to glorify the beginnings of state regulation as adversarial toward industry during “cycles” of regulatory zeal, while simultaneously exaggerating the powerlessness of regulatory agencies in “cycles” of “capture” by industry (Abraham 2007).

As we have noted, discussions about regulatory capture long predate neoliberalism. It is possible for the former to operate in the absence of the latter, so clearly they are distinct mechanisms of political influence and change. Neoliberalism advocates that the state should be minimal and subject to the tests of the market, whereas regulatory capture may thrive under conditions of antimarket protectionism supported by a large state. Nevertheless, neoliberalism can accentuate opportunities for capture if the minimalization and marketization of the state are undertaken largely to meet business interests. In this case, it may be said that the regulatory agencies permitted the commercial interests of industry to capture the agenda for pharmaceutical carcinogenicity testing in ways that did not prioritize the interests of public health. They permitted the industry to do that because the neoliberal political context in which they found themselves did indeed seek to make regulatory agencies more responsive to industry interests. However, the regulators have not relinquished their authority regarding these matters, only their leadership and political will to intervene in the interests of public health. In that respect, we prefer to characterize these developments in carcinogenicity testing standards since the 1980s as a part of what Abraham and Lewis (2000) call “neoliberal corporate bias,” rather than capture, and a neoliberal corporate bias striking at the heart of what counts as  scientific knowledge in ways hitherto unexamined.