Assessment of the impact of pharmacovigilance actions at the population level is an area currently under-investigated but with increasing importance for regulators. Impact research identifies the net impact of a regulatory intervention by measuring both the intended outcomes and the unintended consequences of a regulatory intervention, such as stopping a useful medication or switching to alternatives. A detailed guidance on the methodological conduct of impact studies is provided in Annex 2 of this Guide, together with a comprehensive reference list.
Although it uses existing datasources and methods, the area of impact research has some distinctive characteristics that are worth discussing.
To measure the impact of pharmacovigilance activities, process or outcome indicators can be used depending on the type of intervention, target population, drug or disease characteristics. Determining and measuring the right outcomes can be challenging. It may be further complicated by unavailability of data and may therefore require use of surrogate outcomes. Data sources for the analysis include both primary data and secondary use of data, the latter being used more frequently as they reflect routine clinical practice (real world population). However, secondary use of data that is originally collected for other purposes as such presents limitations, especially in terms of missing relevant data.
If the date or time period of the intervention is known, a before/after time series is a design frequently used allowing to analyse changes of trends in incidence or prevalence of an outcome before and after the intervention occurred. Changes may be affected by simultaneously occurring interventions or events and the use of comparator groups that did not receive the intervention may facilitate the interpretation of any associations found.
The analytical methods will depend on the study design and type of data collection. Interrupted time series (ITS) regression is a strong analytical tool for before/after time series, especially if autocorrelation and adjusting for seasonality are taken into account, and the time point (or period) of the intervention is known. For adequate power, sufficient time points before and after the intervention are required. Joinpoint regression models calculating time points of trend line changes offer an alternative if the date of the intervention is unknown.
Specific analytical approaches are needed to measure unintended effects of pharmacovigilance activities which may not be expected at the design stage, for example switching to alternative medicines following product withdrawal or restriction, and determine the net attributable impact on patient outcomes.
Future challenges include the identification of long-term consequences of regulatory actions and the definition of thresholds for successful risk minimisation activities.