The Pharmacovigilance Symposium at the 38^th Annual Meeting of the Japanese Society of Toxicology (JSOT) marked another interesting exchange between scientists and physicians working in all aspects of safety science to better understand, predict and ultimately manage the safety risks to patients associated with pharmaceutical products. I would like here to offer some observations on the presentations contributed by our excellent speakers as well as highlight some ongoing international collaborations that aim to improve our work in safety science and a heuristic approach to interpreting the results of nonclinical safety studies which I have found useful as a clinical scientist.

　The clinical development and eventual approval for use of pharmaceuticals respectively candidate pharmaceuticals will require an evaluation of the safety in non-clinical in vitro and in vivo animal studies as well as in humans (volunteers and patients). The investigations into human safety are initially guided by non-clinical data and will then broaden to focus on evidence from humans. In this process human data may confirm or not what has been observed in vitro or in animals. In this context, it is worthwhile to enhance understanding and communication between the non-clinical and clinical safety experts to direct human safety investigations into the right area and to create understanding about the quite often limited predictive capabilities of the non-clinical studies. Further, it is of utmost importance to ensure a thorough understanding about some of the areas of safety assessment, which are not easily amenable for either side to thoroughly assess. A few of these areas shall be mentioned here:

・Non-clinical data are of limited value especially in the area of involvement of the immune system. The human immune system is distinctively different from that of animals. Further, the individual history of immunizations and challenges to the immune system of humans creates an immense background of variability in the human population with varying individual susceptibilities. Hence, data in animals or in vitro surrogate data from human matrices (blood, tissues) may often only serve for hypothesis generation but cannot replace directed studies into human safety in vivo.

・Similarly, everything that is associated with assessment and monitoring of brain function and signaling (pain, headache, mood disturbances, suicidality, etc.) is rather human specific and cannot be mimicked adequately with animal investigations. Hence, these areas will normally be exclusively evaluated based on human safety information.

・Conversely, teratogenicity, mutagenicity and carcinogenicity are areas of safety deemed of major importance for human use of medicines but which are only extremely difficult to be assessed in humans. Hence, safety assessment is often based on animal data a priori. But these data, if indicating risk for humans, may require an extensive, very long and costly human biomonitoring program to give an answer about human safety.

　It is apparent from the short elaboration above that the methods used by clinical safety (observational, monitoring) and non-clinical safety (experimental, directed, dose-response) are often quite different and require different sets of scientific methods and skills. Hence, a few of the different principles of both disciplines shall be mentioned below.

　Toxicology testing is predicated on the use of animal models for the prediction of adverse effects of chemicals in humans. The assumption that experimental animals are appropriate and relevant models of human toxicity responses is difficult to fully test for many chemicals. By contrast, pharmaceuticals typically have an extensive body of toxicology data from both animals and humans that could be used to test this assumption. These data are largely unpublished and confidential, residing in the databases and files of pharmaceutical companies.