|
Abstract:
|
In safety assessment, we should control the error rate of wrongfully rejecting a true safety signal and be less stringent to that of wrongfully accepting a false safety signal. Current safety analyses have emphasized detection of significant safety signals. It is obvious that statistically nonsignificant results do not necessarily mean there is no safety concern. It is important to emphasize that unless we can show no harm with certain confidence, a drug cannot be declared safe---even if there is no statistically significant signal. For this purpose, we may consider establishing three types of safety testing hypotheses: showing protective effect, demonstrating no worse than placebo, and detecting risk signals. We would like to discuss if it is feasible to test the hypotheses, how to perform the tests, and what would be the impact on drug labeling and safety management.
|
