ABSTRACT
To evaluate the possible beneficial or detrimental effect of a given agent, the ideal errorfree study design would be one that compares the outcome experiences of a single group of individuals under the opposite conditions of having and having not been exposed to the agent of interest. Such an unattainable comparison may be simulated in animal research. In cancer bioassays, for example, groups of rodents of the same strain, reared under identical conditions (environment, housing, and food and beverage) are randomly and blindly assigned to one or more exposed and unexposed subgroups and followed up life-long for the outcomes of interest. In experimental studies on humans, however, the assumptions of genetic homogeneity and habitat identity are violated. Furthermore, when ethical constraints prevent human experiments, the exposure-disease relationship may only be assessed by observational studies.
