Authors P. van den Brandt, L. Voorrips, I. Hertz-Picciotto, D. Shuker, H. Boeing, G. Speijers, C. Guittard, J. Kleiner, M. Knowles, A. Wolk, A. Goldbohm
Abstract Epidemiologic studies directly contribute data on risk (or benefit) in humans as the investigated species, and in the full food intake range normally encountered by humans. This paper starts with introducing the epidemiologic approach, followed by a discussion of perceived differences between toxicological and epidemiologic risk assessment. Areas of contribution of epidemiology to the risk assessment process are identified, and ideas for tailoring epidemiologic studies to the risk assessment procedures are suggested, dealing with data collection, analyses and reporting of both existing and new epidemiologic studies. The dietary habits and subsequent disease occurrence of over three million people are currently under observation worldwide in cohort studies, offering great potential for use in risk assessment. The use of biomarkers and data on genetic susceptibility are discussed. The paper describes a scheme to classify epidemiologic studies for use in risk assessment, and deals with combining evidence from multiple studies. Using a matrix approach, the potential contribution to each of the steps in the risk assessment process is evaluated for categories of food substances.
The contribution to risk assessment of specific food substances depends on the quality of the exposure information. Strengths and weaknesses are summarized. It is concluded that epidemiology can contribute significantly to hazard identification, hazard characterisation and exposure assessment. Research needs Research needs address two areas of application: those that are needed to facilitate the use and conduct of epidemiologic studies for the purpose of risk assessment and those needed to integrate epidemiology with toxicology into the regular risk assessment process. They are ordered by priority:
Develop decision trees for dietary risk assessment, in which toxicological and epidemiologic research each have their role. Such decision trees may be incorporated in regulations for food safety evaluations.
Work towards a common terminology and approach (i.e. risk models, weight of evidence).
Development of short-term effect biomarkers that are highly predictive of risk of disease and that can be applied in human studies on a relatively large scale at low cost, without reducing response rates. The need for such established markers is great both with regard to cancer as well as for many other conditions.
Expansion of food composition tables with data on various chemicals in food (additives, contaminants, etc)
Further development and testing of framework/guidelines to select epidemiologic studies suitable for risk assessment for the various categories of chemicals in food (additives, contaminants, nutrients, foods etc)
Investigate the opportunities to incorporate more probabilistic approaches into the risk characterisation process. The probabilistic approach, which relies on probability distributions for many of the ingredients of risk characterisation such as exposure, the shape of the dose-response curve, toxicokinetic and toxicodynamic parameters, human variability and precision of the data, is already more familiar to epidemiologists than toxicologists, in particular in Europe. The probabilistic approach should provide a more rational and realistic framework for risk characterisation and would facilitate the integration of toxicology and epidemiology into the framework.(see also Kroes et al., (2002) this issue)
Study interactions regarding disease risk between nutrients and non-nutritive compounds taking into account human patterns of consumption.
Development and validation of methods to combine risks and benefits. Since exposures can be protective for particular disease(s) while enhancing risk of other diseases, further development and use of composite public health measures will provide insight to optimal dose ranges of exposures.
Increased use of genetic susceptibility measures on a population basis and studying interaction between genetic susceptibility and dietary exposures regarding disease risk to refine risk assessment for subgroups of the population; may possibly reduce uncertainty factors that are currently used to account for human variability.
Study the bioavailability of various compounds in normal food matrix and combine these data with exposure information to better define dose.
Further development of PLM using modern food assessment techniques (EAN codes etc) and investigating the possibilities for record linkage to routinely collected data in registries.
Refinement of methods to calculate lifetime risk from epidemiologic studies conducted in cross-section of population in various exposure groups covering a relatively short part of life.
Establishment of the value of applying non-parametric methods to model empirically the association between exposure and disease in epidemiologic studies (i.e. spline regression).
Establish an acceptable and standardised set of (already existing) biomarkers for organ function in humans (i.e. liver, and kidneys).
Download the full document here (593 kb - PDF)