2000), and particulate matter air pollution and premature mortality (Dominici et al., 2003).

There are no routine systems of cardiovascular morbidity ascertainment, although there are several large epidemiological cohort studies that include the endpoints of stroke or myocardial infarction, such as the Framingham study, the Nurses Health Study, the Women’s Health Initiative, the Physicians Health Study, and the Cardiovascular Health Study (Abbott et al., 1988; Ma et al., 1999; Psaty et al., 1999; Rimm et al., 1998; Rossouw et al., 2002). Mortality data are collected routinely at the local level and aggregated into county and state statistics with associated age, gender, and other sociodemographic information. By extracting information from national surveys such as NHANES, it is possible to impute values for functional status or quality of life, such as the Health Utility Index, or the SF36. These in turn can be integrated with mortality information to calculate quality-adjusted life years.


The ability to evaluate the unintended health effects of a genetically engineered organism that expresses a significantly different phenotype than its conventional counterpart is problematic (Kuiper et al., 2001). As recognized by other expert panels, current risk-assessment paradigms and drug-safety evaluation programs are inappropriate methods to apply to the determination of the potential for unintended adverse health effects of GE food (Atherton, 2002; FAO/WHO, 2000).

Unlike chemicals and drugs, a dose-response relationship cannot be established for food (i.e., it is not possible to dose an animal with 10, 1, and 0.1 times the volume of food). Foods also represent complex mixtures that must be tested as a whole to consider possible nonadditive interactions that can significantly impact toxicological outcome (Dybing et al., 2002). Consequently, even though these technologies may satisfy the hazard identification step of risk evaluation, there are no existing validated methods for dose-response characterization for a complex mixture such as food.

New approaches should be based on a risk-assessment strategy proposed by the National Research Council (NRC, 1983) and rely on “substantial equivalence” to illustrate distinctions that may exist between foods modified by genetic engineering compared with those modified through traditional (non-GE) methods. Further, such evaluations would be expected at all stages of product development, including gene discovery, selection, and advancement to commercialization, and followed by postmarketing studies to further assess both intended and unintended effects. Epidemiological studies may be helpful in the postmarketing phase, provided they are conducted with the rigor that contemporary methods allow. The more definition that can be brought to bear with respect to defining exposure, the greater the inferential potential of these observational population-based methods.

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement