ED01-02

The exploration of the role of chemical carcinogens in the etiology and progression of human cancers has progressed enormously with the rapid discovery of the underlying biological mechanisms of cancer. Since it is becoming more apparent that each individual’s cancer is unique, the application of our mechanistic understanding of the disease is opening a new venue for individual based biomarker development and application to prevention and therapy. A biomarker may be defined as a chemical, physical or biological agent in accessible body matrices, an in vivo response to an exposure or set of exposures, or a genotype or phenotype indicative of susceptibility to disease, all measurable in body fluids, cells or tissues. The promises of biomarkers are many fold, though, as yet, largely unrealized. Biomarkers have the potential to improve assessments of ambient environmental exposures; improve methods in risk estimation and classification of at-risk individuals, communities, and populations; define the mechanisms of exposure-disease linkages and underlying susceptibility factors; define the interactions of multiple agents and exposures on disease outcomes; and, ultimately, improve and expedite methods for assessing the effect on disease outcomes of exposure remediation and preventive interventions (Hulka, 1990; Hulka, 1991; Schulte & Perera, 1993; Muñoz & Gange, 1998; Wogan et al, 2004). Of great interest is the translation of the myriad of -omics based platforms to the field of biomarker based strategies as they impact cancer prevention and control.
 It is readily evident that population studies into etiology that use disease as the end point are of necessity large, lengthy and costly. While such investigations will always remain the ultimate standard for establishing linkage, they are unwieldy and inefficient for the timely translation of our accelerating understanding of the molecular basis of cancer towards preventive strategies. Thus, inclusion of biomarkers, despite some intrinsic limitations, into the paradigm of prevention and therapy are of central importance for the advancement of the field. To accomplish this end, validation strategies must be developed, refined and implemented.
 The mere existence of a biomarker does not mean that it will be useful to the field. At present the opportunity for the biased use of biomarkers likely outweighs prospects for informed use. This concern arises from the simple fact that few of the biomarkers currently applied in either population, preclinical or clinical settings have undergone anything approaching rigorous validation. Indeed, paradigms for the validation of biomarkers, themselves, are still evolving (Freedman et al., 1992; Schulte and Perera, 1993; Groopman and Kensler, 1999; Kensler et al, 2001; Franco et al, 2004). Recognizing that considerable effort will be required for the validation of current and future biomarkers of potential use in environmental studies, this discussion seeks to highlight the types of approaches that are being used in the development and application of such biomarkers that, in turn, reflect different components of the multistage, multifactorial process of carcinogenesis. Of particular importance is the recognition of the concept that the utility of biomarkers in population and prevention studies are not dichotomous (i.e., good, bad), but rather continuous, with some markers being more informative than others depending upon how they are used.
 One of the cornerstones for development and validation of biomarkers can be built from a major set of tenets set out in epidemiology for defining causality. In 1965, Sir Austin Bradford Hill (Bradford Hill, 1965) set forth in his presidential address to the section of occupational medicine a systematic view to facilitate an analysis of the role of an environmental exposure to human disease. This view was outlined in the form of nine categories that have since become known as the Bradford Hill Criteria. These criteria now should be re-visited for the validation of biomarkers in the cancer etiology and disease linkage paradigm.
 In the original Bradford Hill address there were several comments that looked to the future role for advances in understanding the mechanisms of disease and how they would impact upon the issues of causality. Reflecting upon the classic observation studies of John Snow, Bradford Hill commented that the effect of the water source was so strong upon the outbreak of cholera that the knowledge of the specific etiologic factor in the disease was not needed. Indeed it would be another 30 years before Koch characterized the Vibrio cholera as the specific organism leading to the disease. However, Bradford Hill recognized that the integration of experimental and mechanistic findings into this model will pose future challenges and opportunities.
 The Bradford Hill Criteria were developed at a time when causality of disease was frequently measured in terms of exposure directly leading to disease. Thus, the questions being asked did not often incorporate mechanism based strategies. This is the extension to the validation paradigm that biomarkers bring to the field.
 Strength of association was the first criteria listed and the discussion involved the examination of both the absolute difference in the disease outcome and the fold change of an incident disease linked to an exposure. Bradford Hill also raised the issue that a slight association may in fact have a profound impact on disease risk. The strength of the association is generally measured in terms of either relative risk or odds ratio which are both group or population based statistical analyses. Consistency is the next criteria and addresses whether the study had been repeated in different people/populations, places, circumstances and times. A strong consistency argument would be enhanced if different study designs, e.g. prospective and retrospective analyses, were done with the same conclusions. Specificity is next major criteria. Often there will not be a one-to-one correspondence between an exposure and disease and such an occurrence is likely to be a fairly rare event. Nonetheless there should be specificity in the magnitude of the association. It would appear that biomarkers have a major role to play in sorting through the complexities associated with specificity. The issue of temporality involves the recognition that a disease outcome must be proceeded by an exposure event. This type of analysis works well with a cross-sectional study design. However in longitudinal studies using biomarkers one can envision a more complex pattern of events when the outcome is driven by a multistage process. For example, an exposure that leads to an enzyme induction which is only relevant if there is a follow-up exposure could violate strict temporality rules. Biological gradient is addressed under the framework that a dose response relationship should exist for the exposure/biomarker/disease investigation. Plausibility is the next criteria. Unfortunately, what is biologically plausible depends upon the biological knowledge base of the day of the investigation. For single agents this is easier since animal or experimental models can address plausibility. This problem is much more difficult in gene-gene, gene-environment, vector-environment situations, thus the plausibility should be examined by testable experimental hypotheses. Coherence is invoked by the use experimental data to buttress human observations. Similarly, experiment involves the simple concept that the lower an exposure is the lower disease should occur. The final criteria was analogy and this implies that we can use data or biomarkers in a generic sense. Therefore, a commonly accepted phenomenon in one area can be extrapolated to another.
 In conclusion, not all biomarkers are suitable for all purposes. Some will be helpful in understanding etiology, or selecting study cohorts, others will find use in assessing participant compliance, and others key to determining agent efficacy in prevention and therapeutic trials.
 #
 REFERENCES
 Bradford Hill, A. (1965) The Environment and Disease: Association or Causation? Proceedings of the Royal Society of Medicine. Pp.296-300.
 Franco, E.L., Correa,P., Santella,R.M., Wu,X., Goodman,S.N., and Petersen,G.M. (2004) Role and limitations of epidemiology in establishing a causal association. Seminars in Cancer Biology, 14, 413-426.
 Freedman, L.S., Graubard, B.I. & Schatzkin, A. (1992) Statistical validation of intermediate end points for chronic diseases. Stat. Med. 11, 167-178.
 Groopman, J.D. & Kensler, T.W. (1999) The light at the end of the tunnel for chemical-specific biomarkers: daylight or headlight: Carcinogenesis, 20, 1-11.
 Hulka, B.S. (1991) Epidemiological studies using biological markers: issues for epidemiologists. Cancer Epidemiol., Biomarkers & Prev., 1, 13-19.
 Hulka, B.S., Wilcoxky, T.C. & Griffith, J.D. (1990) Biological Markers in Epidemiology. New York: Oxford University Press.
 Kensler, T.W., Davidson, N.E., Groopman, J.D. and Munoz, A. (2001) Biomarkers and Surrogacy: Relevance to Chemoprevention. In: Biomarkers in Cancer Chemoprevention. Miller, A.B., Bartsch, H., Boffetta, P., Dragsted, L., and Vainio, H., eds. IARC Scientific Publications, No. 154, 27-47.
 Muñoz, A. & Gange, S.J. (1998) Methodological issues for biomarkers and intermediate outcomes in cohort studies. Epidemiol. Rev., 20, 29-42.
 Schulte, P.A. & Perera, F.P. (1993) Molecular Epidemiology: Principles and Practices. San Diego, Academic Press.
 Wogan,G.N., Hecht,S.S., Felton,J.S., Conney,A.H., and Loeb,L.A. (2004) Environmental and chemical carcinogenesis. Seminars in Cancer Biology, 14, 473-486.

Sixth AACR International Conference on Frontiers in Cancer Prevention Research-- Dec 5-8, 2007; Philadelphia, PA