Does additional monitoring status increase the reporting of adverse drug reactions? An interrupted time series analysis of EudraVigilance data

Abstract Purpose To evaluate the impact of including a medicine in the list of medicinal products subject to additional monitoring (AM) on the reporting of adverse drug reactions (ADRs) in the european economic area (EEA). Methods Interrupted time series using the monthly number of EEA ADR reports in EudraVigilance during 12 months before and after the addition to AM list. The main outcome was the change (%) in reporting of ADRs with step change as the a priori impact model. Further time series analysis was performed using Joinpoint Regression. Results The analysis included 11 active substances. No significant immediate (step change) increase of reporting was identified for any product at time of addition to AM list. We identified a significant gradual increase of ADR reporting after addition to AM list (slope change) for two out of five new products—boceprevir (10% per month, 95% confidence interval (CI) 3%–18%) and denosumab‐Xgeva (13% per month, 95% CI 4%–22%). No change was identified for Prolia, another denosumab‐containing product not subject to AM. No significant increase was identified for any product included in the AM list due to the requirement to conduct a PASS. Conversely, a gradual decrease in reporting was identified for natalizumab (−5% per month; 95% CI −10% to −1%), rivaroxaban (−5%; −8 to −3%), and varenicline (−16%; −21 to −10%). The results were corroborated by the Joinpoint analyses, which yielded similar results. Conclusions We identified limited evidence that reporting of ADRs increased modestly and gradually for some new products and not for products with PASS requirement.


| INTRODUCTION
Pharmacovigilance is the science and activities relating to the detection, assessment, understanding and prevention of adverse drug reac- The impact of additional monitoring as a policy intervention on the reporting of ADRs is unknown. Measuring the impact of regulatory decisions is important for all pharmacovigilance activities in order to improve existing processes. 5,6 Consequently, we undertook this study using EudraVigilance data to investigate whether the inclusion of medicines in the AM list increases the reporting of ADRs for those medicines in Europe.
The study was part of a data gathering project on the experience with additional monitoring, together with a survey of patients' and healthcare professionals' attitudes and behaviours towards reporting ADRs. 1 To account for changes in the number of patients exposed to the medicines over time, we calculated the size of at-risk population using medicine consumption/exposure data estimates obtained from Periodic Safety Update Reports 11 (PSUR) held by EMA for centrallyauthorised products. Exposure data are normally reported as the total person-years exposed during the interval covered by the PSUR which can range from 6 months to several years. We therefore divided the  • The real impact of additional monitoring on reporting of ADRs is currently unknown.

| Analysis
• Using an interrupted time series analysis of EudraVigilance ADR reporting data, we identified limited evidence that reporting of ADRs increased modestly and gradually for some new products. In contrast, reporting of ADRs did not increase (or even decreased) for products subject to AM due to the requirement to conduct a PASS.
• Further work is required to determine the effectiveness of AM as a policy intervention.
defined time point. 18 We applied segmented regression analysis to the interrupted time series data using Poisson regression and modelling the count of events per month, whilst offsetting the changes in at-risk population. To account for over-dispersion of the data we corrected the standard error by applying a scale parameter based on Pearson chi-square statistic divided by the residual degrees of freedom and we adjusted for seasonality using Fourier terms with two pairs of sine/cosine functions. 18,19 We postulated an immediate effect As the final number of joinpoints is established on the basis of a statistical criterion and their position is not fixed it does not require that an intervention date is pre-specified unlike interrupted time series regression. 21 We compared the results of both statistical analyses.

| RESULTS
We identified 82 eligible products corresponding to 79 substances from the AM list with at least 12 months of baseline data and excluded 68 substances with low ADR reporting. The final analysis therefore contained 11 substances, five of which were included in the AM list due to new substance status (boceprevir, telaprevir, vemurafenib, fingolimod and denosumab) and six that were included because of an imposed PASS (imatinib, lenalidomide, natalizumab, rivaroxaban, valproic acid and varenicline), as detailed in Flowchart 1 and Table 1. Analyses were performed at substance level except for denosumab, for which we identified two products with different indications (Xgeva and Prolia) and discrepant AM status. Xgeva is indicated for the prevention of skeletal related events in adults with advanced malignancies involving the bone and for the treatment of giant cell tumour of bone and is subject to AM. 22 Prolia is indicated for the treatment of osteoporosis and for bone loss associated with hormone ablation in men with prostate cancer and is not subject to AM. 23 The analysis of these products was performed separately at product level with Prolia used as a control for Xgeva.
As presented in Table 2 and Figures 1 and 2 When the two denosumab-containing products were compared, the reporting for Xgeva, a product that was subject to AM, increased by 13% (95% CI 4%-22%) per month, but we detected no significant changes in reporting of ADRs for Prolia (0%, 95% CI −6% to 7%), a product that was not subject to AM.
The results were broadly comparably with the results of the Joinpoint analyses, which are presented in Figures 3 and 4, with the exception of natalizumab, for which no change was identified in the Joinpoint analysis.

| DISCUSSION
In our study we identified that ADR reporting increased after addition to the AM list for two out of five new products, in the order of 10%- explain the limited effect we found on increasing the reporting of ADRs.
Due to methodological limitations, the maximum number of substances that we were able to include in our study (11 of 79 substances) was limited by the availability of their baseline data or low reporting, which restricts the overall generalisability of the findings. A large number of new medicines were excluded simply because there was no comparative data available before their inclusion in the AM list. The small sample of substances that we were able to study is also limited in terms of chemical/biological and pharmacological classes and may therefore not be universally generalisable to all medicines subject to AM. A possible solution to overcome this limitation would be to stagger future implementation of such policy interventions in various regions over time, which would serve as comparators in a step-wedge type of approach, but this may not be feasible for EU wide interventions. Alternatively, where possible, a controlled ITS design can be used, as illustrated with the denosumab example.
Another approach that could be employed would be to study the substances which we were able to study as well as the low monthly ADR counts we were unable to stratify the analyses by the reason of addition to AM list, by country, or by reporter type (patients or healthcare professionals) as initially intended. Therefore although differences in ADR reporting may exist between different reporter types, as reported in our separate study on reporter attitudes and behaviours, 1 we could not examine this effect in this study. Seven of the studied substances were included in the first version of the list in April 2013 and the remaining four between 2013 and 2015. Therefore, changes in these reporting sub-categories and changes over time cannot be excluded.
The application of ITS methods to medicine use data, other than being limited by the lack of baseline data for new medicines, also requires reliable estimates of exposure data due to the often rapid changes in the size of the population at risk, as opposed to population studies. Medicine use can increase rapidly after authorisation/reimbursement and inclusion in clinical guidelines or decrease due to safety concerns or replacement with a more effective, safer or more convenient competitor. The estimates of exposure that we have relied on are often based on approximations from sales data using an expected dose and treatment duration 11  We did not collect details about the lifecycle of individual medicines included in our study, and therefore couldn't investigate these factors in the current study.
Indeed, half of the substances included in our study were included in the AM list due to the obligation to conduct a PASS. The imposition of a PASS often follows the emergence of serious ADRs, concerns about medicine safety or the evaluation of benefit-risk balance by the PRAC (e.g. via a referral procedure), with consequent media attention. Such a setting is prone to confounding due to other regulatory actions and media attention possibly influencing ADR reporting and therefore the results are difficult to interpret in terms of causality. This may be one reason why we observed a fall in ADR reporting for several products added to the AM list due to an imposed PASS. However, should this observation be confirmed in future research, it may serve to inform any future discussions on the legislation governing additional monitoring.

| CONCLUSION
In summary, we identified limited evidence that reporting of ADRs increased modestly and gradually for some new products and did not increase for products subject to AM due to the requirement to conduct a PASS. The small number of medicines that we were able to include in our study, together with its ecological design, makes the causality of this observation difficult to establish. We suggest that, in the future it would be worthwhile to pre-specify the methods for the evaluation of policy interventions such as the introduction of additional monitoring, to help overcome the shortcomings of a retrospective evaluation. Given the limitations in our results, we would welcome suggestions from the research and regulatory communities on complementary methods that might be applied to study the impact of AM.

ETHICAL STATEMENT
The study did not involve human subjects and relied on routine data.