raising questions of significant human toxicity without significant additional data. In general, it would not be advisable to solely use Category B or C effects to specify the seriousness of the adverse effects expected in humans.

The level of concern appropriate for different adverse findings increases when effects have been documented in well-designed and well-conducted animal studies, when the observed effects increase in severity or incidence with increasing dose, and/or when the observed effects are otherwise clearly related to the substance. An ideal study would be appropriately controlled, define the composition of the test material, administer the test material in measured quantities by the oral route, and use standardized and validated methods to measure toxicity accompanied by appropriate statistical analysis, interpretation, and reporting (see Box 5-1). While many studies will not meet this ideal, they will nonetheless provide useful information and should be used if they suggest possible risk to human health.³ The strength of the evidence for toxicity is substantially increased if the effects were observed in more than one animal species, and even more so if supported by additional experimental data (e.g., in vitro data) or human data.

Under current law, FDA has the burden of providing evidence that one or more uses of a dietary supplement poses some identifiable significant or unreasonable risk to human health. Issues confronting FDA in regulating dietary supplements are not exactly analogous to those that arise in the premarketing approvals of other substances such as food additives, for which protocols for using animal data to establish safety have been developed. The traditional use of animal toxicity data to establish acceptable exposures has imbedded within it an element of caution—animal toxicity findings are used without significant question regarding the predictive power of specific findings for humans, and uncertainty factors are used to ensure safety. Use of protocols for setting safe levels (as opposed to evaluating risk) are outlined in Box 5-2. In assembling evidence regarding risks to health,