Epidemiological studies of cancer have well-recognized limitations, including: (i) lack of sensitivity; (ii) difficulty in discriminating between several plausible risk factors ; (iii) the inadequacy of past exposure data ; (iv) inability to evaluate the impact of recent exposures; and (v) uncertainties in interpreting negative studies or inverse relationships. Only the first and last points will be discussed below.

Low-level exposures and negative studies

Much of the information suggesting a lack of response to suspected carcinogenic exposures at low doses is dependent on the evaluation of negative epidemiological studies. A negative study is one of adequate statistical power in which the relative risk for a suspected factor is not significantly different from unity at the 5% level of statistical significance. The interpretation of apparently negative data is complicated by the impossibility of proving a true negative – no matter how large a study, there remains some chance of missing a carcinogenic effect. These statistical limitations are also inherent even in large scale animal bioassays, where lack of carcinogenic activity is unprovable. Study of low-level potential carcinogenic hazards in the general environment, such as air, water or soil pollution from waste sites or industry, is fraught with many uncertainties (Grisham, 1986). It is not always possible to define the exposed population precisely in order to establish an accurate baseline for the presumed health effect.