Innovations in Post-Marketing Safety Research

Safety surveillance is important during the entire life cycle of a drug. Pre-marketing trials have been shown to be ineffective in establishing the full safety profile of the drug, mainly due to their relatively small sample size and characteristics of the patients, which are usually younger and healthier than the average patient using the drug in general practice. Hence, the safety profile of a drug will be incomplete at market entry. Regulations have been implemented to ensure drug safety and market authorization holders are at present required to provide updated safety information as evidence accumulates. During the post-marketing phase additional safety information is mainly gathered by signal detection, observational studies, and meta-analysis of trials. The thesis is divided into three subtopics: perspectives and concepts in post-marketing safety research, utilizing data from randomized controlled trials (RCTs) in safety research and finally future steps in risk monitoring. First we focused on available safety information and what we can learn by taking a different approach to this data. We discuss comparability of data from RCTs of same class drugs, where one has been withdrawn from the market due to safety concern and the second one still remains on the market. Further, we discuss learning during the post-marketing phase. We aimed to investigate whether same class drugs had similar adverse drug reaction profiles and whether learning occurs between the drug pair and at which time point. Finally, we categorize different types of off-label use and discuss the potential shift in risk associated with moving out of the licensed domain. We hypothesize that, only when the off-label population is more vulnerable to pharmacologically expected (Type A) events than the domain will we expect higher risk. Type B (unpredictable) adverse drug reactions occur randomly, also in the off-label population. The second topic focuses on utilizing data from RCTs to estimate drug safety in situations where potential safety issues rise during the post-marketing phase. We used data from the ADVANCE trial to first assess whether the intensity of glucose control influences cancer risk in patients with type II diabetes and second to discuss methodological challenges of post-hoc analysis of RCTs that explore drug associated cancer outcomes. We found that intensity of treatment did not influence cancer risk in these patients. For the last topic we propose a new strategy that could aid safety evaluation during the post-marketing phase. We structured a Bayesian meta-analysis model that combines information from phase III and phase IV and thereby, combines both trial and observational data. For this we used the example of rosiglitazone which was withdrawn from the European market due to cardiovascular risk. We conclude with a discussion on benefits of new strategies for pharmacovigilance such as the Bayesian meta-analysis which is currently not a part of routine pharmacovigilance activities and the importance of seeking the most efficient ways to use available information. We identified several areas where data may be used more efficiently; these include information about same class drugs and dosages, trial databases for post-hoc analysis and limiting uncertainties regarding off-label use.