What influences a health risk assessment?

Toxicology Letters 167 (2006) 201–204

What influences a health risk assessment? Christina Rud´en ∗ Division of Philosophy, Royal Institute of Technology, Teknikringen 78B, SE-100 44 Stockholm, Sweden Received 6 July 2006; received in revised form 19 September 2006; accepted 20 September 2006 Available online 28 September 2006

Abstract In this paper it is claimed that the health risk assessment process is influenced by (at least) four general factors, namely: the regulatory framework, the quality and availability of scientific data, general risk assessment principles, and case-by-case assumptions. Furthermore, the scientific basis of risk assessment relies on three overall types of methods for data generation: standardized animal experiments, epidemiology, and non-standardized mechanism data. In this paper, the use of the different types of data for risk assessment purposes are analyzed in the light of the factors claimed to influence the risk assessment process. It is concluded that the availability of pre-defined criteria for the interpretation and evaluation of data for regulatory health risk assessment purposes need to be further developed. Especially with the implementation of the new European chemicals legislation REACH. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Risk assessment; Regulatory toxicology; REACH; Toxicity testing; Epidemiology; Mechanism data

1. The regulatory risk assessment process The risk assessment process is influenced by (at least) four general factors: the regulatory framework, the availability and quality of scientific data, general risk assessment principles, and case-by-case assumptions (Fig. 1). The regulatory framework of the risk assessment process can be determined by legislation, or by some other institution, such as an organization performing risk assessments. The regulatory framework determines the scope of the process. This includes for instance what substances are included/prioritized, data requirements, risk assessment requirements and procedures, the range of possible risk management decisions, and how the processes are structured, for instance in terms of how experts are appointed, who is responsible for what, and how and by whom formal decisions are taken.

∗

Tel.: +46 8 790 9587; fax: +46 8 790 9517. E-mail address: [email protected].

The regulatory framework, thus, determines the (minimum) data availability. The data requirements differ widely for different chemical groups/legislations. Pharmaceuticals and pesticides are examples of substance groups for which extensive testing is required. For existing industrial chemicals on the other hand, no test data are currently effectively required. (In the proposed new European chemicals legislation new and existing chemicals will be treated in single system.) General risk assessment principles can be defined as the principles that the risk assessment process is based upon. These principles are general in the sense that they remain the same independent of which chemical or exposure that is under scrutiny. (It should however be noted that general principles can potentially differ between different regulatory frameworks and they may also change with increased knowledge.) The general principles can furthermore range from detailed criteria, such as the predefined criteria for classification of substances according to their inherent properties as specified in the European council directive 67/548, to very general and fundamen-

0378-4274/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.toxlet.2006.09.008

202

C. Rud´en / Toxicology Letters 167 (2006) 201–204

Fig. 1. Factors that influence the risk assessment process.

tal principles such as the principles stating that animal data should generally be considered relevant for human risk assessment (European Commission, 2003). Besides the general principles, the risk assessment process is also dependent on case-by-case assumptions based on expert judgment. Case-by-case assumptions are assumptions that are adopted for the purpose of assessing a specific chemical substance; they may thus differ from substance to substance. According to the European Commission Technical Guidance Document for risk assessment (TGD) “the risk assessment process relies heavily on expert judgment”, and examples of issues to be determined by expert judgment are: the relevance of test data obtained with non-standardized methods, the interpretation of conflicting data, data quality evaluation, assessment of carcinogenicity and mutagenicity, and whether an effect is causally connected to exposure or not (European Commission, 2003). Several ambitious measures have been taken to make a systematic account of the regulatory process (European Commission, 2000a; National Research Council, 1994). And attempts have even been made to harmonize risk assessment procedures as in the IPCS harmonization project (www.who.int/ipcs; Sonich-Mullin et al., 2001). However, the combined effect of scientific uncertainties and a significant reliance on individual experts has made it difficult to achieve a risk assessment process that is fully consistent and systematic in all aspects. It is therefore essential to continuously scrutinize and evaluate this process, and to provide systematic feedback from its practical performance. This requires careful systematic studies of the actual workings of the system (Hansson and Rud´en, 2006), including investigations of the potential for introducing bias due to conflict of interest pertaining to individual experts (for a discussion see, e.g. Maurissen et al., 2005). 2. The scientific basis of risk assessment Health risk assessment is performed by comparing the levels of human exposures to the exposure/dose levels at which no adverse effects have been identified, either in humans or in experimental animal models. Risk

assessment thus relies on scientific data on exposures and effects. Effect data are usually obtained from three main sources: from standardized experiments, often animal models (as required in some legislation), from studies of exposed humans (epidemiology data), and from nonstandardized experiments (e.g. toxicological research data). 3. Standardized animal experiments As mentioned above, a generally accepted principle in toxicological risk assessment is that adverse effects seen in animal studies indicate that the chemical under study will cause a similar effect in humans. Typically standardized animal experiments are used for the purpose of hazard identification, and adverse effects of chemicals have in many cases been first identified in animal models. The design and procedures of standardized experiments are laid down in detailed guidelines, such as the OECD Testing Guidelines and the principles of Good Laboratory Practices (GLP). Animal experiments performed in accordance with the OECD Test Guidelines and GLP are readily accepted in the risk assessment process (European Commission, 2003), as well as for classification and labeling (Council Directive 67/5481 ). In this system, the use of standardized data for classification and risk assessment is also supported by the detailed criteria for the interpretation of data specified in the classification and labelling directive. In combination, the test guidelines, the principles of Good Laboratory Practices, the Technical Guidance Document, and the classification and labelling criteria provides a well developed and generally accepted set of evaluation and interpretation criteria for standardized animal data. 4. Epidemiology data If data from exposed humans are available, no species extrapolation is necessary and therefore, high quality epidemiological data are usually assigned significant weight in the risk assessment process. The quality and strength of the epidemiological evidence for specific health effects depends on the overall design of the study, and on the resulting power of the study to detect an effect. According to the European Commission Technical Guidance Document, relevant epidemiological study designs may include cohort, case-control and correlation studies. Cluster investigations and case reports, on 1

The classification and labelling system is currently undergoing a global harmonization procedure.

C. Rud´en / Toxicology Letters 167 (2006) 201–204

the other hand, can serve as “supporting information in specific cases” (European Commission, 2003). In epidemiology human beings are studied in their daily life. The design of epidemiological studies is, therefore, dependent on the prerequisites available, i.e. on the identification and enrolment of individual human beings. Therefore, detailed guidelines for study design are inherently difficult to specify. According to the European Commission Technical Guidance Document, criteria for assessing the adequacy of epidemiology studies include: “. . . the proper selection and characterisation of the exposed and control groups, adequate characterisation of exposure, sufficient length of follow-up for disease occurrence, valid ascertainment of effect, proper consideration of bias and confounding factors, and a reasonable statistical power to detect an effect” (European Commission, 2003). However, these are not criteria in the usual sense of the word; rather they can be seen as some guidance for evaluating fundamental aspects of study design. For epidemiological data, the criteria for data quality evaluation and interpretation are, thus, less detailed than for the standardized animal experiments. 5. Non-standardized, mechanism data In addition to animal data and epidemiology, also other types of data can be used in risk assessment; typically data to clarify the mechanism of toxicity (or mode of action) in the experimental species, with the purpose to refine the species extrapolation. Standardized animal experiments produce little mechanistic information. Instead the use of non-standardized methods is often needed to obtain this kind of information. These methods include data on: metabolism (including studies in cell cultures from different species), absorption (which may also differ between exposure routes and species) and various aspects of toxicity (e.g. tests for cytotoxicity in different types of cells, macromolecule binding studies, tests using embryo culture systems, sperm motility tests etc.) (European Commission, 2003). The use of mechanism data has been established, e.g. the criteria for classification of carcinogenic substances according to the European union directive, which states that a substance should not be classified as a carcinogen if “the mechanism of experimental tumor formation is clearly identified, with good evidence that this process cannot be extrapolated to man” (Council Directive 67/548/EEC, par. 4.2.1.2). How much and what kind of data that are needed to consider a mechanism “clearly

203

identified”, and what is meant by “good evidence” however, is not further specified in this directive. According to the European Commission Technical Guidance Document, data obtained from nonstandardized methods should be evaluated on a case-by-case basis using expert judgment (European Commission, 2003). Some general guidance is however, given as to what aspects that should be taken into consideration. That is for animal data: if a complete test report is available, including information on the purity/impurities and origin of the test substance, the species tested, dosing procedures, and completeness of the report. Furthermore, if there are other studies or calculations available on the substance, it should be considered whether the data under consideration are consistent with them. There are also guidance on the evaluation of in vitro data, listing a number of factors that should be taken into account: • The range of exposure levels used, taking account of the toxicity of the substance towards the bacteria/cells, its solubility and, as appropriate, its effect on the pH and osmolality of the culture medium. • Whether, for volatile substances, precautions have been taken to ensure the maintenance of effective concentrations of the substance in the test system. • Whether, when necessary, an appropriate exogenous metabolism mix (e.g. S9 from induced rat liver or from hamster liver) has been used. • Whether appropriate negative and positive controls were included as integral parts of the tests. • Whether an adequate number of replicates (within the tests and of the tests) were used (European Commission, 2003). Again, these are not criteria in the usual sense of the word; rather they can be seen as some guidance to fundamental aspects of study design. The availability of detailed evaluation and interpretation criteria are, thus, limited also for non-standardized (mechanism) data. 6. Discussion and conclusions Toxicity data can be obtained from different experimental models, representing different levels of biological organization. (Indeed data relevant to a risk assessment can also be generated without including a biological component, e.g. quantitative structure activity relationships (QSARs) or data on physico-chemical characteristics, but such methods are not the focus of this paper.) The use and interpretation of animal data obtained from standardized tests are usually not controversial in risk assess-

204

C. Rud´en / Toxicology Letters 167 (2006) 201–204

ment. One, not too uncommon, exception being cases when different tests report conflicting results (Rud´en, 2001a, 2001b, 2003, 2004, 2006). The regulatory acceptance of animal data obtained from standardized models rests on e.g. an extensive experience from this type of testing resulting in the availability of detailed guidelines and criteria for data interpretation and evaluation. The use of epidemiology data is more diverse, not seldom leading to diverging interpretations and evaluations by different risk assessors (Rud´en, 2001a, 2001b, 2003, 2004, 2006). This is due to the nature of the epidemiological method being a non-experimental approach which, in general, leads to reduced statistical powers to detect effects due to the large number of confounding factors. Previous studies for the use of mechanism data in risk assessment indicates that the use of mechanism data in risk assessment is diverse. Different risk assessors may chose different data to serve as a basis for their assessment, they may draw diverging conclusions from the same data, and they may also evaluate the quality of different data differently (Rud´en, 2002, 2003, 2006). Furthermore, the use of mechanism data is in some cases non-transparent in the sense that it is not easy for the reader of the risk assessment to follow the argumentation from data reference to conclusion, due to lack of explicit referral to particular data (Rud´en, 2002, 2006). This is due to the lack of generally accepted criteria for generation, use, and interpretation of mechanism data. For instance, criteria that in a meaningful way addresses the question of how much, and what type of mechanism data are sufficient for overruling the results obtained in a long-term in vivo standardized experiment. The current lack of stringency and heavy reliance on case-by-case assumptions and expert judgement in health risk assessment should be seen in the light of the forthcoming European legislation for industrial chemicals – the REACH system – in which the responsibility for risk assessment to an increasing extent will be put on the producers of chemicals rather than on public agencies.

References European Commission, 2000a. First report on the harmonisation of risk assessment Procedures, Part 1, 26–27 October 2000. European Commission, 2003. Technical guidance document in support of commission directive 93/67/EEC on risk assessment for new notified substances and commission regulation (EC) No. 1488/94 on risk assessment for exiting substances. Available on-line at http://www.ecb.jrc.it/. Council Directive (EEC) 67/548 of 27 June 1967 on the approximation of laws, regulations and administrative provisions relating to the classification, packaging and labeling of dangerous substances. Official Journal 196, 1–5. Hansson, S.O., Rud´en, C., 2006. Evaluating the risk decision process. Toxicology 218 (2–3), 100–111. Maurissen, J.P., Gilbert, S.G., Sander, M., Beauchamp, T.L., Johnson, S., Schwetz, B.A., Goozner, M., Barrow, C.S., 2005. Workshop proceedings: managing conflict of interest in science. A little consensus and a lot of controversy. Toxicol. Sci. 87 (1), 11–14. National Research Council, 1994. Science and judgement in risk assessment. In: Committee on risk assessment of hazardous air pollutants, Commission on life sciences. National Academy Press, NRC Washington, DC. Rud´en, C., 2001a. Interpretations of primary carcinogenicity data in 29 trichloroethylene risk assessments. Toxicology 169 (3), 209–225. Rud´en, C., 2001b. The use and evaluation of primary data in 29 trichloroethylene carcinogen risk assessments. Regulat. Toxicol. Pharmacol. 34 (1), 3–16. Rud´en, C., 2002. The use of mechanistic data and the handling of scientific uncertainty in carcinogen risk assessment—the trichloroethylene example. Regulat. Toxicol. Pharmacol. 35 (1), 80–94. Rud´en, C., 2003. Science and transscience in carcinogen risk assessment—the European regulatory process for trichloroethylene. J. Toxicol. Environ. Health B: Crit. Rev. 6 (3), 257–277. Rud´en, C., 2004. Acrylamide and cancer risk—expert risk assessments and the public debate. Food Chem. Toxicol. 42, 335–349. Rud´en, C., 2006. Science and policy in risk assessments of chlorinated ethenes. In: Mehlman, M.A., Soffritti, M., Landrigan, P., Bingham, E., Belpoggi, F. (Eds.), Annals of the New York Academy of Sciences, vol. 1076, Living in a Chemical World Framing the Future in Light of The Past. Sonich-Mullin, C., Fielder, C.R., Wiltse, J., Baetcke, K., Dempsey, J., Fenner-Crisp, P., Grant, D., Hartley, M., Knaap, K.A., Kroese, D., Mangelsdorf, I., Meek, E., Rice, J.M., Younes, M., 2001. IPCS conceptual framework for evaluating a mode of action for chemical carcinogenesis. Regulat. Toxicol. Pharmacol. 34, 146–152.