Biomarkers as predictive tools in toxicity testing.

The Report and Recommendations of ECVAM Workshop 401,2

Reprinted with minor amendments from ATLA 28, 119-131.

Diane J. Benford,3 A. Bryan Hanley,4 Krys Bottrill,5 Sarah Oehlschlager,4 Michael Balls,6 Francesco Branca,7 Jean Jaques Castegnaro,8 Jaques Descotes,9 Karl Hemminiki,10 David Lindsay11 and Benoit Schilter12
3School of Biological Science, University of Surrey, Guildford, Surrey GU2 5XH, UK; 4Central Science Laboratory, Sand Hutton, York Y041 ILL, UK; 5FRAME, Russell and Burch House, 96-98 North Sherwood Street, Nottingham NG1 4EE, UK; 6ECVAM, Institute for Health & Consumer Protection, Joint Research Centre, European Commission, 21020 Ispra (VA), Italy; 7Unitadi Nutrizione Umana, Instituto Nazionale delta Nutrizione, Rome, Italy; 8Unit of Gene Environment Interactions, IARC, 150 Cours Albert Thomas, 69008 Lyon, France; 9Department of Pharmacology, Medical Toxicology and Medicine, INSERM U 98-X Faculté de Médecine Lyon RTH Laennec, 69372 Lyon Cedex 98, France; 10Department of Molecular Epidemiology, Centre of Nutritional Toxicology, Karolinska Institute, Novum, 141 57 Huddinge, Sweden; 11Euro Science Perspectives Ltd, 3 Patcham Grange, Brighton BN1 8UR, UK; 12Nestle Research Safety Centre, Verscheales Blanc, 1000 Lausanne 26, Switzerland

1ECVAM - The European Centre for the Validation of Alternative Methods. 2This document represents the agreed report of the participants as individual scientists.

Address for correspondence: Diane Benford, School of Biological Science, University of Surrey, Guildford, Surrey GU2 5XH, UK.

Address for reprints: ECVAM, TP 580, JRC Environment Institute, 21020 Ispra (VA), Italy

Preface:

This is the report of the fortieth of a series of workshops organised by the European Centre for the Validation of Alternative Methods (ECVAM). ECVAM's main goal, as defined in 1993 by its Scientific Advisory Committee, is to promote the scientific and regulatory acceptance of alternative methods which are of importance to the biosciences and which reduce, refine or replace the use of laboratory animals. One of the first priorities set by ECVAM was the implementation of procedures which would enable it to become well-informed about the state-of-the-art of non-animal test development and validation, and the potential for the possible incorporation of alternative tests into regulatory procedures. It was decided that this would be best achieved by the organization of ECVAM workshops on specific topics, at which small groups of invited experts would review the current status of various types of in vitro tests and their potential uses, and make recommendations about the best ways forward (1). In addition, other topics relevant to the Three Rs concept of alternatives to animal experiments have been considered in several ECVAM workshops.

This ECVAM workshop on Biomarkers as Predictive Tools in Toxicity Testing was held in Burnham Market (Norfolk, UK) on 5-8 October 1998, under the co-chairmanship of Diane Benford (University of Surrey, UK) and Bryan Hanley (Central Science Laboratory, UK). The participants, whose expertise included toxicology, biomarker research, the Three Rs and methods of validation, came from industry, universities and other research institutions. The aim of the workshop was to provide a forum for discussing the utility of Biomarkers as predictive tools for toxicity testing. A major focus of the workshop was to examine the possibilities afforded by the use of Biomarkers for the reduction, refinement or replacement of animal use in toxicity testing. This report summarises the outcome of the discussions and includes a number of conclusions and recommendations.

Introduction

The term biomarker has grown in popularity in recent years. With this growth has come a great increase in applications and a correspondingly increased diffuseness in the meaning of the term, to the extent that one of the major difficulties in the area of biomarker research is reconciling differing views on what constitutes an acceptable definition. This is reflected in the large number of definitions appearing in the literature, a selection of which are given below in chronological order. This is not designed to be either comprehensive or a critically evaluated collection, but it does represent the range of uses to which biomarkers are put and the justifications for the use of the word in different applications. They range from exposure measures, and biological indices which support a mechanistic postulate, to clinical markers with diagnostic implications.

  1. "Biological markers are indicators signalling events in biological systems or samples", and "Biological markers are measurements of body fluids, cells or tissues that indicate in biochemical or cellular terms the presence and magnitude of toxicants or of host responses." (2)
  2. Biomarkers are "cellular, biochemical, or molecular alterations which are measurable in biological media such as human tissues, cells or fluids." (3)
  3. "The term biomarker refers to the use made of a piece of information, rather than to a specific type of information. A biomarker is a change in a biological system that can be related to an exposure to, or effect from, a specific xenobiotic or type of toxic material." (4)
  4. "Biomarkers, broadly defined, are indicators of variation in cellular or biochemical components or processes, structures, or functions, that are measurable in biological systems or samples." (6)
  5. A biomarker is "an indicator that signals events in biological systems or samples, and it is generally taken to be any biochemical, genetic, or immunological indicator that can be measured in a biological specimen." (6)
  6. "The term biomarker has been used to describe measurements in the sequence of events leading from exposure... to disease. At each step, persons may differ in susceptibility, thus a biomarker may also refer to an indicator of susceptibility." (7)
  7. "A biomarker is a xenobiotically induced variation in cellular or biochemical components or processes, structures, or functions, that is measurable in a biological system or sample." (J.F. McCarthy, R.S. Halbrook & L.R. Shugart, 1991, Conceptual strategy for design, implementation and validation of a biomarker-based biomonitoring capability. Internal document of the Environmental Sciences Division of the US EPA, used as the discussion document for the NATO Advanced Research Workshop on Strategy for Biomarker Research and Application in the Assessment of Environmental Health).
  8. "... the Biomarkers are any of a series of biochemical or molecular responses to compounds that have entered an organism, reached sites of toxic action and are exerting an effect on the organism." (Proposal document for the NATO Advanced Research Workshop on Strategy for Biomarker Research and Application in the Assessment of Environmental Health, Texel, NL, May 1991)
  9. Biomarkers are "parameters that putatively represent some step along the causal pathway between exposure and effect." (8)
  10. Biomarkers are "markers of biologic activity that reflect not evidence of, but the potential for, neoplastic progression... Unlike tumour markers, which are biological indicators found in neoplasms, prevention biomarkers are specifically related to earlier stages of carcinogenesis. These intermediate endpoints can be defined as measurements of a particular biologic factor associated with the evolution of neoplasia and occurring with increased frequency in abnormal cells." (9)
  11. "The term biomarker is used in a broad sense to include almost any measurement reflecting an interaction between a biological system and an environmental agent, which may be chemical, physical or biological." (10)
  12. "A biomarker is a measurable event occurring in a biological system, such as the human body. In environmental epidemiology, a biomarker represents a sub-clinical and reversible change; it is not a diagnostic test, but an indicator that an early change has occurred that could later lead to clinical disease." (11)
  13. A biomarker is "a measurement made on body tissue, body fluid or excretion to give a quantitative indication of exposure to a chemical and which may give an estimate of the risks consequent on the exposure." (12)
  14. "Biomarkers are observable endpoints in a continuum of events leading from exposure to toxic agents to diseases that ultimately result from exposure." (13)
  15. "The term biomarker is used in a broad sense to describe parameters reflecting an interaction between a biological system and a potential hazard of chemical, biological and physical nature. The measured response may be functional physiological and biochemical at a cellular or molecular level." (14)
  16. "A biomarker is a parameter which can be evaluated quantitatively, semi-quantitatively or qualitatively, and which provides information on exposure to a xenobiotic, or on the actual or potential effects of that exposure in an individual or in a group." (15)

It is apparent from the breadth of these definitions that any use of the term "biomarker" must also include a definition of what is meant by it within the context of a specific discussion. This workshop highlighted the great difficulties associated with the imprecise use of this word. A realistic appraisal of the biomarker area is required in order to clarify between scientists and clinicians the different ways in which the term may be used.

Genesis/Taxonomy of Biomarkers

The term biomarker has a significant and lengthy history. The meaning for which it is used clearly depends upon the context and this is reflected most clearly in the parameters of the database which is used as the basis for any search. Since the topic of this workshop was the use of biomarkers as predictive tools in toxicity testing, the focus was on biological, medical and toxicological uses of the term. Therefore, an appropriate database set was likely to be found in Medline and Toxline, so these databases were searched for the term "biomarker", with the following results:

1966-1975: Only one record, which related to the measurement of porphyrin in soil samples (16).

1976-1985: 51 records, of which the majority were concerned with the use of tumour biomarkers to diagnose cancer, obtain a prognosis or monitor therapy. Other topics included ageing, monitoring wildlife rabies vaccination, rat pituitary enzymes, ecotoxicology, assessment of toxic exposure, and the molecular epidemiology of cancer.

1986-1990: 308 records, in which the major subjects were the use of tumour biomarkers for diagnosis, prognosis, monitoring of therapy and chemoprevention programmes, assessment of toxic effects, and ecotoxicology. Other topics included ageing, assessment of toxic exposure, use of biomarkers in risk assessment, assessment of susceptibility to toxic insult, monitoring wildlife rabies vaccination, and studies on other diseases.

1991-1995: 3579 records, which included the first references to validation (17, 18). The subjects covered included all those mentioned above, but with far less emphasis on ageing. There was increased use of biomarkers for the monitoring of chemoprevention programmes, the monitoring of specific groups of workers for exposure, and the assessment of susceptibility. There was a new subject for discussion, namely, the ethical aspects associated with the biomonitoring of individuals, especially with respect to their susceptibility to disease. Other new topics included the use of biomarkers in epidemiology and in the assessment of health risks within the general population, the use of biomarkers in studies on the mechanisms of carcinogenesis, and the identification of biomarkers of dietary intake. There was more theoretical consideration of methods to be used in the analysis of biomarker data, possible sources of bias, and the validation of biomarkers.

1996-November 1998: 3107 records covering topics very similar to those of the previous period.

This breakdown of the uses of the term shows a clear progression in the ideas associated with biomarkers. The earliest reference given above refers to the measurement of a biological compound (porphyrin) in a nonbiological matrix (shale), and as such, has no toxicological relevance.

Between 1976 and 1985, many of the studies in which the term biomarker was used were concerned with tumour biomarkers as diagnostic/prognostic tools. This use of the term remains current. Most of these diagnostic biomarkers are concerned with indicating current status and measuring the effectiveness of treatment. Such biomarkers are not strictly predictive of outcome, although they may contribute to an overall prognosis given by the physician.

A second use of the term biomarker in this period was in the detection of exposure to toxic agents. In some cases, biomarkers were measured in humans. But a major usage at this time was related to the use of biological systems as bioindicators of environmental pollutants. For the purposes of such studies, the most biologically relevant systems were those which were most sensitive in relation to the pollutants of interest. This type of study is not intended to be predictive of effects in other species, but is indicative of exposure that may represent a potential hazard to our species.

The final area of active research in this period was the development of biomarkers of ageing. These investigations have little overall relevance to classical toxicity testing, though they may contribute to the assessment of cumulative toxic insult in humans, if ageing is the result of a predominantly exogenous process.

In the period 1986-1990, the total number of references to biomarkers increased substantially, particularly in relation to toxicology. There was a gradual increase in studies on biomarkers of occupational exposure, which is likely to reflect a change in terminology from "biological monitoring" to "biomarkers", as well as an actual increase in research. The primary use of biomarkers of exposure is to establish the adequacy of control measures with respect to chemical exposure in the workplace, i.e. occupational exposure. They are not intended to be predictive of an adverse effect, but since occupational exposure limits are intended to protect the workforce from a known effect, infringement of the control measures clearly has the potential to cause harm. (Studies relating to smoking aimed to investigate specific causal agents and mechanisms of carcinogenicity.) A further area of development was that of biomarkers of effect. These are parameters that change in response to exposure and reflect a biological consequence to that exposure. For the first time, some studies which related biomarkers to risk assessment were carried out.

In the period 1991-1995, a number of definitions of biomarkers were published, five of which are listed above (11-15). A number of papers referred to biomarkers as a means of monitoring the effects of interventions. This usage is based on the hypothesis that a particular exposure is associated with a particular disease. Intervention methods may be used to modify the exposure under the assumption that this will reduce the risk of that disease. Biomarkers of exposure are used to verify that the intervention has the anticipated effect on the internal exposure; biomarkers of effect, if available, may give an earlier indication of whether the intervention is likely to result in reduced incidence of the disease. If such relationships are established, this helps to strengthen the evidence of causality. Because of their increasing availability and sensitivity, biomarkers are increasingly being used to explore the links between exposure and effect in order to establish causality. Because biomarkers can measure internal exposure in an individual, the importance of individual susceptibility and vulnerability become more clearly defined. Biomarkers that reflect these individual differences are referred to as biomarkers of susceptibility; they need not relate to specific exposures. In addition, some awareness of the ethical issues involved in biomarker research and monitoring populations began to emerge.

In the most recent period (1996-November 1998), the numbers of biomarker-related records in Medline and Toxline increased dramatically, although the topics being dealt with were broadly similar to those noted in the previous period.

In conclusion, work on biomarkers over the past 30 years or so has progressed, such that, currently, a range of definitions have been adopted, and this diversity is represented in Figure 1. However the most commonly used terms are biomarkers of exposure, biomarkers of effect and biomarkers of susceptibility.

Figure 1: The progression in the use of the term "biomarker"

For the context of this report, we will define a biomarker as "a parameter which can be measured in a biological sample, and which provides information on an exposure, or on the actual or potential effects of that exposure in an individual or in a group".

This report aims to consider the use of biomarkers as predictive tools in toxicity testing, so we must also consider what is meant by predictivity. The purpose of regulatory toxicity testing is to define the inherent hazardous properties of a substance. Previous ECVAM workshops have considered the value of in vitro test systems for the prediction of various hazardous properties, and important elements of the discussion include the required level of accuracy of predictivity in terms of false negatives and false positives, and how to achieve acceptance as replacement tests for regulatory toxicity testing. The point of reference is usually the results produced by in vivo animal toxicity studies; comparison with human data would be preferable, but sufficient human data of sufficiently high quality are rarely available. Clearly, the uses of biomarkers listed above do not include prediction of hazardous properties or of "toxicity". We therefore need to consider the broader aspects of toxicology (rather than toxicity testing) and the value of biomarkers as predictive tools that might also allow for refinement and reduction of animal use in toxicology studies. In particular, the ultimate aim of toxicology is to evaluate the potential risk to humans, where the risk is a function of exposure as well as the hazard.

Biomarkers of exposure could be predictive of risk, if sufficient information were available on the dose-response relationship. Biomarkers of effect cover a range of measurements which may be indicative of exposure to a particular agent, although the specificity is generally less than for biomarkers of exposure. They may also show correlations with possible clinical outcomes, but cannot be assumed to be predictive unless they are shown to reflect a significant step along the causal pathway between exposure and effect. The extreme example of a biomarker that reflects a current clinical condition, is actually a diagnostic marker and because it is rarely related to a chemical exposure, it does not meet the above definition of a biomarker. Biomarkers of susceptibility most notably include genetic susceptibility. The concept of susceptibility will not be dealt with in this report per se but it must be recognised that individual susceptibility and population susceptibility are important issues, which should be taken into account and may be appropriate for a future workshop, once the establishment of a coherent meaning to the use of biomarkers in predictive toxicology has been better defined.

It is important to distinguish between use of biomarkers in toxicological studies in experimental animals and in humans. Traditional toxicological endpoints are generally empirical observations of the consequences of exposure to specific agents. However, provided that a study is conducted under strictly controlled conditions, with appropriate statistical analysis, it is reasonable to assume that the observed effects were caused by the exposure under investigation. Exposure to the animal is well defined but, nevertheless there may be requirements for internal exposure to be verified by toxicokinetic measurements, such as blood levels of test material in guidelines for some types of toxicity test. These are not normally considered to be biomarker studies, although they clearly meet many of the definitions of a biomarker. Biomarkers of effect (although again not termed as such) are frequently included in a battery of clinical chemistry assays on blood and liver, and possibly other tissues, depending upon the nature of the effect under investigation. In contrast, human studies, even apparently well controlled volunteer studies, involve many other variables and inter-individual differences that mean effects cannot be so readily ascribed to the exposure under investigation. With investigations on "free-living" human populations, the potential for confounding factors increases enormously. It therefore becomes essential to establish the specificity of a "biomarker" for the exposure and/or effect of interest. For a biomarker of exposure, this relates to the measurement techniques and the toxicokinetics of the substance but, for a biomarker of effect, it must be based on understanding of the mechanism of action. This leads to two key questions.

  1. What are the current status and future prospects for the use of biomarker methodology in toxicity testing?
  2. What are the prospects for use of biomarkers leading to refinement, reduction and replacement of animal procedures in toxicology?

Answers to these questions form the basis of the rest of this report.

Biomarkers in Toxicity Testing

Toxicity tests are designed to identify the hazardous properties of a chemical substance, tested in an isolated form in laboratory animals. Because they are primarily intended for the testing of novel substances, the protocols have to be generic and cannot be related to specific mechanisms of action. There is little opportunity for biomarkers of exposure to be included in such studies. As noted above, there are some exceptions. One example is the requirement for plasma levels of compound in carcinogenicity studies of pharmaceutical agents, which are compared with data on human plasma levels at the therapeutic dose, in order to demonstrate that sufficiently high doses of drug have been tested. A second example is the mouse bone marrow micronucleus test, for which a negative result requires evidence that the compound is available to the bone marrow, either by specific chemical analysis (biomarker of exposure), or by changes in the ratio of normal to polychromatic erythrocytes. Because this reflects toxicity to the bone marrow, whereas the endpoint of the assay is chromosomal damage, this could be considered to be the use of a biomarker of (a secondary) effect as a biomarker of exposure.

In principle, biomarkers of effect could be valuable tools in toxicity testing protocols. Biomarkers that develop early in the course of repeat exposure, but that can be reliably correlated to the subsequent development of a lesion or an effect, would allow animal testing protocols to be refined and animal testing to be reduced. However, the mechanistic link to the lesion needs to be understood in order to define the uses and limitations of a particular biomarker. At the current time, there is insufficient understanding of mechanisms of toxicity to permit the identification of meaningful biomarkers for inclusion in routine toxicity testing protocols. For a very small number of compounds, some associations are known, such as liver glutathione content as an indicator of paracetamol hepatotoxicity (19), but these cannot be used in a general sense. Some of the "biomarkers" now used in clinical chemistry, such as serum levels of liver transaminases, are more appropriately termed "organ function tests", and arguably should be defined as diagnostic markers rather than as biomarkers. Nevertheless, they are valuable because they demonstrate an important link between in vivo and in vitro testing, i.e. that the same endpoints can be used.

Biomarkers have the potential either to be used as indicators of events at various stages in the progression of a toxic lesion, or to provide tools to test the relevance of proposed toxic mechanisms. The extent to which a biomarker is a predictive tool depends on its connection to the causal pathway leading from exposure to effect, and on the quantitative assessment of the consequences of changes in the biomarker in terms of toxicological endpoints. Our current understanding of mechanisms of toxicity is generally not sufficient to support development of biomarkers that could be early predictors of toxic effect. However, as our understanding increases, and with the enormous growth in the use of new techniques such as genomics and proteomics, which seek to determine the link between gene expression, gene function, protein synthesis and protein function and exposure to environmental and chemical agents, there is the potential to develop such predictive biomarkers in the future. The success of a genomics/proteomics based approach to biomarkers relies on the use of intelligent data manipulation, and on correlation between measurable changes in the functional properties of biological macromolecules and the properties of the cell, organ or organism.

In addition to traditional toxicity tests, predictive models are available for risk assessment, ranging from structure/activity models (i.e. QSAR) and knowledge based systems (for example, DEREK) to in vitro assays. The relevance, reliability and overall value of any predictive model system is crucially dependent on the quality of the database used in its development and the method used for assessing the reliability of various pieces of information. It is possible that any biomarker which has been shown to fit into an established mechanistic pathway could provide information which could be used by an established predictive model. Such predictive systems should also be capable of assessing the effects of substances which may reduce or limit toxicity (in dietary terms, the concept of "protective factors").

Clearly, the expression of a toxic effect is determined by the magnitude of exposure, i.e. the dose. There are a growing number of examples where the validity of extrapolating from high doses to low doses can be questioned, especially where this extrapolation is not based on detailed knowledge of the mechanisms of toxicity. (However, the concepts of minimum effective dose and biological differences between high and low dose scenarios must also be included. The dose level at which a biomarker is measured in response to a chemical challenge must be realistic in terms of "normal" human exposure. There is little purpose in developing effective biomarkers (in terms of their relevance to a toxic endpoint), if they are of little relevance to quantifiable measurements at human exposure levels, or cannot be measured at these levels. In addition, it is important to be able to compare conflicting data on the effects of compounds by working in similar test systems. For example, consumption of over-boiled coffee has been found to increase blood cholesterol (20) in a human study; however, the level of exposure was higher than would normally be encountered outside a testing regime. The same test material has been demonstrated to have a chemoprotective effect against genotoxicity (21) in a nonhuman test system by stimulating phase 2 enzymes at much lower (and probably more physiologically relevant) levels. At normal consumption levels, the benefits of the latter effect may outweigh the risks associated with the former; however, until directly comparable human studies are carried out, it is not possible to make a net toxicity/benefit prediction. The dose responsiveness of bioactive constituents, the way that a biomarker can reflect this, and the choice of appropriate test systems, are clearly keys to establishing the validity of biomarkers as predictive tools.

Exposure and Effect

The relationship between exposure and effect is crucial to any consideration of the risk associated with a given chemical. When considering experimental toxicology, it has been assumed that there is a linear progression from exposure to effect, with the likelihood of increasing severity of toxicity with increasing exposure. Therefore, it is possible to envisage a number of biomarkers indicative of the magnitude of exposure/response and/or the progression along the pathway (Figure 2).

Figure 2: The pathway from exposure to overt clinical effect

The control exerted over the process in the laboratory helps to assure of the link between exposure and effect. In contrast when considering the majority of human health problems, it is possible to identify a large number of potential exposures and a large number of diseases for a given population, but the links between them are poorly defined, if at all. If, on the basis of knowledge of mechanisms of toxicity, biomarkers can be developed that match those defined in Figure 2, there is the potential to:

  1. Unravel the links between exposure and effect, i.e. establish causality;
  2. use biomarkers of internal exposure and early biomarkers of effect to determine the efficacy of interventionary measures; and
  3. use biomarkers of effect as predictive tools in investigating agents that may exert a similar toxic effect.

Biomarkers of exposure can be defined strictly by chemical analysis. The simplest forms involve measurement either of the agent in question, or of a metabolite; however, they must be specific to the exposure of interest. For example, hippuric acid has been used as a urinary biomarker of exposure to toluene. While hippuric acid is a metabolite of toluene, and therefore could be expected to reflect exposure in a coherent way, it is non-specific, because it may also be derived from a number of food constituents (22). The value of urinary hippuric acid as a biomarker therefore depends upon the relative sources from dietary or occupational origins in a particular individual. In contrast, the presence of the urinary metabolite, equal, is completely specific for the ingestion of isoflavones. However, some individuals are unable to produce this metabolite, so its absence in urine may not be reflective of non-exposure (23). Therefore, knowledge of the toxicokinetics of the agent is required, in order to determine the most appropriate metabolite, tissue, and sampling times, and the relevance of the results obtained. In some instances, it may be possible to establish different biomarkers as measures of recent and cumulative exposure. An example of this can be seen in studies on lead. The half-life of lead in the blood is 36 days, and it is therefore most useful as a measure of recent exposure (24). Estimates of cumulative lead exposure can be made by measuring bone lead, where the half-life of lead in cortical bone is more than ten years, while that in the more metabolically active trabecular bone is less. However, although techniques that permit in vivo measurement of lead in bone are available, they are not suitable for routine monitoring purposes. A further complication is introduced by the fact that bone lead can be mobilised in various circumstances, such as pregnancy, and may therefore be a significant contributor to blood lead levels (24). Urinary lead levels may also be used as a marker of recent exposure, but are subject to considerable inter-individual variation.

Once considerations go beyond a direct measurement of chemical exposure, they begin to include an assessment of effect and lose specificity for the chemical in question. For example, other biomarkers that have been used in measuring lead exposure involve the monitoring of effects upon the haem synthesis pathway. This is based on the known haematotoxicity of lead, which is mediated via the inhibition of enzymes of haem synthesis, leading to the accumulation of the intermediates δ-aminolaevulinie acid (ALA), zinc protoporphyrin (ZPP) and coproporphorin (24). These biomarkers of effect are not specific to lead exposure, as they may be influenced by other haematotoxic agents or by nutritional or sub-clinical disease status. In addition, they are not connected to the disease endpoints now considered to be of most concern (because haematoxicity occurs at higher levels than are now likely to occur, whereas reproductive and neurobehavioural effects may occur at lower levels of exposure).

Provided there is sufficient evidence to assume that a given disease is associated with a specific exposure, it is not necessary for the mechanism of effect to be understood in order for biomarkers of exposure to be of value. Thus, it can be assumed that increasing exposure increases risk, and decreasing exposure decreases risk, even though these increases or decreases may not be quantifiable. However, in order to identify meaningful biomarkers of effect, as indicated in Figure 2, it is necessary to have an established, mechanistic causal pathway which links measurements of exposure, intermediate effects and the final outcome. At the far end of the chain of events are diagnostic markers. Just as exposure markers are tied to specific chemical exposures, diagnostic markers relate to specific clinical events. It is axiomatic that the further along the pathway from exposure to effect, the more difficult it is to define precisely the extent to which the exposure contributes to a specific clinical outcome, and vice versa. However, measurements of biological indices which are indicative of a causal pathway between exposure and outcome (i.e. biomarkers of effect) may be considered to have a degree of predictivity. Following the definition of the scope and limitations of such biomarkers, they could be incorporated into in vivo and in vitro toxicity studies.

In conclusion, biomarkers of exposure alone cannot be predictive of a toxic effect, but have established value in monitoring and controlling exposure to chemicals that are recognized or assumed hazards to health. Biomarkers of effect might be able to be used in investigations of causal links and into mechanisms of toxicity. Once causality has been established, a biomarker should fulfil the criteria discussed below, if it is to be used in a predictive sense, whether in experimental toxicology, or in human populations.

Validation Issues

The validation of biomarkers is not directly comparable to validation of in vitro tests, but must follow the basic principle of demonstrating reproducibility, reliability and fitness-for-purpose. The actual process will vary, depending on the type of biomarker and its intended use. It was beyond the scope of this workshop to consider this issue in detail, but some general comments can be made.

Validation of biomarkers of exposure should include consideration of:

  1. the toxicokinetics of the substance, i.e. the quantitative and time relationship of levels in body fluids and tissues when exposure is of short, medium or long duration;
  2. specificity, i.e. whether the substance or its metabolite may result from sources other than the exposure of interest;
  3. analytical quality control, i.e. the use of appropriate reference materials to establish the accuracy of results;
  4. stability in appropriate biological fluids; and
  5. potential confounding factors.

It is notable that a number of quality control schemes are in operation, at regional, national and international levels, for methods used in biological monitoring of occupational exposure (25). Such schemes could provide a model for other biomarker applications.

Validation of biomarkers of effect should include consideration of:

  1. relevance to the effect of interest;
  2. reliability of prediction of the associated endpoint;
  3. interlaboratory reproducibility;
  4. analytical quality control;
  5. stability in appropriate biological fluids; and
  6. potential confounding factors.

Future Considerations

Although biomarkers currently cannot replace or reduce the use of animals in toxicity testing, future technological developments show great promise for making this possible. In vitro techniques may play an important role in the development of biomarkers of effect, as the greater control which is possible in in vitro studies facilitates investigation of mechanisms of toxicity, and therefore the identification. A mechanistic understanding of the progression of a toxic effect is a necessary prerequisite for the use of biomarker approaches for the prediction of toxicity. It is possible to envisage a natural progression, in which a potential biomarker is identified during in vitro mechanistic studies, and its value tested in animal models, is then incorporated in in vivo testing protocols (refinement of animal testing), and is subsequently validated for use in in vitro toxicity tests (leading to reduction and replacement of animal testing). Once an acceptable level of use has been established for a new chemical, there is a potential for biomarkers to provide the continuity between initial toxicity testing and post-marketing surveillance, to ensure that the acceptable levels are adequate to protect the human population against adverse effects.

Biomarkers of exposure are not directly relevant to toxicity testing per se, but they may become relevant outside the laboratory environment in the determination of exposure and dose (molecular dosimetry) and in monitoring to check that acceptable exposure levels are not exceeded. They are also relevant to risk assessment and to investigating mechanistic causes of human diseases.

If biomarker techniques can be readily automated, they have the potential to facilitate the development of high throughput assays, which permit the testing of compounds which have not already been tested but to which populations are exposed. This is particularly relevant to dietary exposure, partly because of the vast number of chemical components of food, but also because the complex nature of whole foods makes delineation of the effects of individual components both difficult and misleading. For example, dietary oestrogens interact with each other and with the consumer in unpredictable ways. Therefore, markers which can combine the different active and protective components into an overall biological responsiveness (for example, increases in levels of sex hormone-binding protein or changes in the expression of specific genes) linked to clinical endpoints (for example, increased or decreased risk of developing particular diseases) are potentially of great value.

The importance of the establishment of a causal pathway was a recurring theme of the workshop. Currently, there are few examples of situations where this has been achieved. However, this reflects the lack of information on the mechanisms which underpin toxicity. At present, too little emphasis is placed on this issue in toxicological studies.

Conclusions and Recommendations

  1. The term "biomarkers" is employed very loosely. There are many differences of opinion about the definition, uses and potential value of biomarkers.
  2. Most controversy relates to the use of the term "predictive biomarker". A biomarker is, by definition, a reflection of the current status of the biological sample or system at the time of analysis. For a biomarker of exposure, that status could reflect either recent or longer term exposure, i.e. a past or present situation. Provided there is a known association with a disease, the biomarker of exposure can predict a change in risk of that disease, but it cannot predict a toxic effect. A biomarker of effect might reflect an early stage in the development of a disease, and therefore may be predictive of eventual disease. However, the development of such biomarkers is dependent upon understanding of the aetiology of the disease in question, and few examples are available as yet.
  3. The use of biomarkers as research tools to illuminate a postulated pathway or to measure imprecise endpoints is an important aspect of their role.
  4. Traditional toxicity testing is not predictive of the range of effects which may occur in a real-life exposure situation. This incomplete picture of the effects of exposure may be clarified by the development of biomarkers which have specific relevance to human disease. Such relevant biomarkers could ultimately be used to help establish the causal links between exposure and outcome and/or be indicative of a potential effect in humans. There is a need to establish the criteria by which biomarkers can be considered to be valid for such purposes. This should include an awareness of confounding factors, the establishment of a dose-response relationship in terms of the biomarker and exposure, and the further establishment of a response-outcome relationship.
  5. A biomarker can be identified as occurring along the exposure effect pathway, either as a result of empirical observation of an apparent correlation which helps to illustrate a mechanism, or from a mechanistically derived hypothesis.
  6. Biomarkers of exposure may have particular relevance in:
    1. post-toxicity testing of the consequences of long-term exposure in a real-life situation;
    2. monitoring of exposure; and
    3. investigating the factors which affect absorption and uptake.
  7. Biomarkers should be subject to quality control schemes similar to those already established for occupational exposure, including establishment of appropriate reference materials that could be made available to laboratories using a particular technique.
  8. At present, it is impossible to reduce, refine or replace animal experiments by applying Biomarkers in environmental bioassays for toxicological evaluation. The acquisition of knowledge, for example, as a result of the human genome project and the further development of genomic and proteomic technologies, should make this feasible in the future.

References

  1. ECVAM (1994). ECVAM News & Views. ATLA 22, 78-11.
  2. Nuclear Regulatory Commission (1987). Biological markers in environmental health research. Environmental Health Perspectives 74, 3-9.
  3. Hulka, B.S. & Wilcosky, T. (1988). Biological markers in epidemiologic reserch. Archives of Environmental Health 43, 83-89.
  4. Henderson, R.F., Bechtoid, W.E., Bond, J. & Sun, J.D. (1989). The use of biological markers in toxicology. Critical Reviews of Toxicology 20, 65-82.
  5. Nuclear Regulatory Commission (1989). Biologic Markers in Reproductive Toxicology, 40pp. Washington, DC. USA: National Academy Press.
  6. Schulte, P. & Mazzuckelli, L.F. (1991) Validation of biological markers for quantitative risk assessment. Environmental Health Perspectives 90, 239-246.
  7. Wiilett, W.C. (1991). The use of Biomarkers in nutritional epidemiology. In Biomarkers of Dietary Exposure (ed. F.J. Kok & P. van's Veer), pp. 9-14. London, UK: Smith-Gordon.
  8. Hattis, D. (1991). Use of biological markers and pharmacokinetics in human health risk assessment. Environmental Health Perspectives 90, 229-238.
  9. Greenwald, P., Within, K.M., Malone, W.F. Byar, D.P, Freedman, L.S. & Stern, H.R. (1992). The study of markers of biological effects in cancer prevention research trials. International Journal of Cancer 52, 189-196.
  10. WHO (1993). Biomarkers and Risk Assessment: Concepts and Principles (Environmental Health Criteria, No. 155). Geneva, Switzerland: WHO.
  11. Grandjean, P. (1995). Biomarkers in epidemiology. Clinical Chemistry 41, 1800-1803.
  12. Sherlock, J.C. (1995). Key issues in the use of Biomarkers for assessing risk from food chemicals. In Biomarkers in Food Chemical Risk Assessment (ed. H.M. Crews & A.B. Hanley), pp. 1-8. London, UK: Royal Society of Chemistry.
  13. Ward, J.B. & Henderson, R.E. (1996). Identification of needs in biomarker research. Environmental Health Perspectives 104, Suppl. 5, 895-900.
  14. van Poppel, G., Verhagen, H. & Heinzow, B. (1997). Biomarkers in epidemiological and toxicological nutrition research. In Food Chemical Risk Analysis (ed. D.R. Tennant), pp. 87-105. Blackie Academic and Professional Press.
  15. Bottrill, K. (1998). The use of Biomarkers as alternatives to current animal tests on food chemicals. ATLA 26, 421-480.
  16. Rho, J.H., Bauman, A.J., Boettger, H.G. & Yen, T.F. (1973). A search for porphyrin biomarkers in Nonesuch Shale and extraterrestrial samples. Space Life Science 4, 69-77.
  17. Vahrendorft, J. (1995). Design of studies for validation of Biomarkers of exposure and their effective use in environmental epidemiology. Toxicology 101, 89-92.
  18. Taioli, E., Kinney, P., Zhitkovich, A., Fulton, H., Voitkun, V., Cosma, G., Frenkel, K., Toniolo, P., Garte, S. & Costa, M. (1994). Application of reliability models to studies of biomarker validation. Environmental Health Perspectives 102, 306-309.
  19. Mitchell, J.R., Jollow, D.J., Potter, W.Z., Davis, D.C., Gillette, J.R. & Brodie, B.B. (1973). Acetaminophen-induced hepatic necrosis. IV. Protective role of glutathione. Journal of Pharmacology and Experimental Therapeutics 187, 185-194.
  20. van Tol, A, Urgert, R., de Jong-Daesar, R., van Gent, T., Scheek, L.M., de Roose, B. & Katan, M.B. (1997). The cholesterol-raising diterpenes from coffee beans increase serum lipid transfer protein activity levels in humans. Atherosclerosis 132, 251-254.
  21. Cavin, C., Holzhauser, D., Constable, A., Huggett, A.C. & Schilter, B. (1998). The coffee-specific diterpenes cafestol and kahweol protect against aflatoxin B1-induced genotoxicity through a dual mechanism. Carcinogenesis 19, 1369-1375.
  22. Lauwerys, R.R. & Hoet, P. (1993). Industrial Chemical Exposure: Guidelines for Biological Monitoring, 2nd edn, 318pp. Boca Raton, FL USA: Lewis Publishers.
  23. Setchell, K.D.R. & Cassidy, A. (1999). Dietary isoflavones: biological effects and relevance to human health. Journal of Nutrition 129, 758S-767S.
  24. International Programme on Chemical Safety (1995). Environmental Health Criteria, Vol. 165, Inorganic Lead, 300pp. Geneva, Switzerand: WHO.
  25. Lehnert, G., Schaller, K-H. & Angerer, J. (1999). Report of the Status of the External Quality Control Programme for Occupational Medical and Environmental-Medical Toxicological Analyses in Biological Materials in Germany. International Archives of Occupational and Environmental Health 72, 60-64.