Opportunities and pitfalls when measuring harm reduction through pharmacovigilance activities

Background/Introduction: Pharmacovigilance (PV) activities aim to reduce harm by better use of medicines and promote public health. Assessing the impact of PV actions on public health at population level is an area currently under-investigated [1, 2]. Determining and measuring the right outcomes can be challenging and complicated due to lack of publically available data. Measuring impact should not only include an assessment of the intended but also take into account unintended effects and other simultaneous events such as changes in clinical practice, or secular trends in health outcomes. Objective/Aim: To investigate the feasibility of measuring impact of pharmacovigilance activities based on publically available data in the Netherlands. Methods: Five different identified safety risks related to use of medicines were assessed. 1. thrombotic risk due to (off-label) use of Diane-35 (cyproterone/ ethinyloestradiol) 2. pergolide and bromocriptine and the risk of cardiac valvulopathy 3. proton pump inhibitors and the risk of hypomagnesaemia 4. rosiglitazone withdrawal due to cardiovascular effects 5. valproate and the risk of congenital abnormalities and developmental disorders Assessment was done using the following approach: a. Based on the information on the safety signal and following pharmacovigilance actions a definition of impact on health was formulated for each signal; b. based on this pre-defined definition of impact of the specific PV activities, data on change of health were searched in publically available data sources; c. If not available, data were searched needed to estimate the impact on health: data on change in use of the medicine or other safety measurements concerning the use of the medicine and the AR or RR of the safety issue. For b. and c. also other factors that could have influenced the health outcomes were taken into account. The feasibility of pre-defining a definition of impact, and the public availability of the needed data were evaluated. Results: Determining the definition of the impact was feasible. However, measuring or estimating the impact was hampered by a lack of publically available data both on outcomes and drug use. For some safety signals (2, 4) it was challenging to determine a time frame. For only one signal (5) data on change of health were publically available, although only in a small cohort. For signal 3 no data on drug use were available, and for signal 4 data on the risk were doubtful, so an estimation of the impact was not possible. For two signals (1, 2) a crude estimation of the impact could be made with a varying degree of assumptions. Conclusion: Lack of data hampers assessment of the impact of PV activities, but with assumptions a crude estimation of the impact on health can be possible.