Conclusions

CONCLUSIONS

Although there are limitations, non-clinical data is predictive of some human toxicities seen in clinical use. Experience shows that the degree of prediction is not as good for idiosyncratic toxicities as for those associated with the pharmacology or metabolism of the drug, as would be expected. Pharmacovigilance data will therefore always be required to detect rare and idiosyncratic human toxicities. Thorough and objective review of non-clinical data does, however, detect toxicities that are linked more directly to the actions of the drug and which have the potential to affect humans. For the foreseeable future, therefore, non-clinical data will continuously be used to iden-tify potential human toxicities, to identify safe starting doses and dose regimens and to develop the appropri-ate safety monitoring procedures for the clinical trial protocol. These aspects should be addressed in the risk–benefit analysis for the trial.

Once humans have been exposed in clinical trials, the data generated should be considered carefully with the non-clinical data until a picture of the human toxi-cities has been developed. Experience shows that not all toxicities are predictable based on the non-clinical and early clinical trial data; however, literature suggests that the rate of prediction is approximately one half to two thirds. Increasing knowledge of the mechanisms of toxicity and of species differences, together with the judicious use of in vitro metabolism and recep-tor binding methodologies, is allowing better species selection. This, together with the increasing availability of non-clinical disease models gives hope that predic-tivity will increase, or at least not decrease, provided that the data are carefully and objectively assessed.