Article Text
PDF
Abstract
Objectives To describe opportunities and challenges experienced from the four pharmacoepidemiological database studies included in the rivaroxaban post authorisation safety study (PASS) programme and propose ways to maximise the value of population-based observational research when addressing regulatory requirements.
Design PASS programme of rivaroxaban carried out as part of the regulatory postapproval commitment to the European Medicines Agency.
Setting Clinical practice in Germany, the Netherlands, Sweden and the UK (electronic health records)—undertaken by pharmacoepidemiology research teams using country-specific databases with different coding structures.
Participants 355 152 patients prescribed rivaroxaban and 338 199 patients prescribed vitamin K antagonists.
Results Two major challenges that were encountered throughout the lengthy PASS programme were related to: (1) finalising country-tailored study designs before the extent of rivaroxaban uptake was known, and (2) new research questions that arose during the programme (eg, those relating to an evolving prescribing landscape).
Recommendations We advocate the following strategies to help address these major challenges (should they arise in any future PASS): conducting studies based on a common data model that enable the same analytical tools to be applied when using different databases; maintaining early, clear, continuous communication with the regulator (including discussing the potential benefit of studying drug use as a precursor to planning a safety study); consideration of adaptive designs whenever uncertainty exists and following an initial period of data collection; and setting milestones for the review of study objectives.
- EPIDEMIOLOGIC STUDIES
- Cardiology
- Vascular medicine
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Introduction
Rivaroxaban is a factor Xa inhibitor with several cardiovascular indications.1 Between 2011 and 2020, Bayer and partners conducted a large rivaroxaban postauthorisation safety study (PASS) programme comprising eight observational studies as part of the regulatory postapproval commitment to the European Medicines Agency (EMA). Hereafter, we describe opportunities and challenges experienced from the four pharmacoepidemiological database studies included in the programme and propose ways to maximise the value of population-based observational research when addressing regulatory requirements.
A detailed description of the rivaroxaban PASS programme (designed to address the approved cardiovascular indications) has been published previously.2 Briefly, of the eight observational studies, four were population-based database studies aimed at evaluating rivaroxaban prescribing, safety and effectiveness in routine primary and/or secondary care versus the existing standard of care (vitamin K antagonists (VKAs)) over a 9-year period from approval—learnings from which are the focus of this paper. All four database studies were based on secondary use of patient-level data from electronic healthcare sources well established for pharmacoepidemiological research, with rivaroxaban use captured from prescriptions issued by primary care physicians or pharmacy dispensations.2 We also conducted four other studies: one using Modified Prescription-Event Monitoring methodology in primary care, two using Specialist Cohort Event Monitoring methodology in secondary care and a study of the effectiveness of risk minimisation activities2 3; these other studies are not discussed hereafter. All studies were conducted by experienced independent research organisations (all members of the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance)2 in collaboration with Bayer, with data collected from four European countries (Germany, the Netherlands, Sweden and the UK).
The PASS programme was initially designed to support three cardiovascular indications in adults that were authorised in Europe from 2008: prevention of venous thromboembolism (VTE) in adults undergoing elective total hip/knee replacement surgery; treatment of VTE and prevention of its recurrence; and prevention of stroke and systemic embolism in patients with non-valvular atrial fibrillation (SPAF indication). Three of the four database studies (from the UK, Germany and the Netherlands) were designed with respect to those indications. In 2013, the EMA approved rivaroxaban for concomitant use with acetylsalicylic acid with/without clopidogrel/ticlodipine for the prevention of atherothrombotic events in adults with elevated cardiac biomarkers after an acute coronary syndrome (ACS). This approval included the condition to collect further information in the postauthorisation stage to monitor rivaroxaban use and address safety when used for secondary prevention of ACS in real-world clinical practice. In particular, reassurance was sought regarding whether the distribution of risk factors among patients prescribed rivaroxaban in clinical practice was consistent with the ATLAS ACS2 TIMI 51 trial population.4 To meet this request, Bayer expanded the programme with the development of the fourth database study (from Sweden). Additionally, the protocols for the three original database studies were modified to also capture rivaroxaban use in patients with ACS, and were upgraded from PASS category 3 (ie, legally required to investigate safety of the authorised drug as part of the pharmacovigilance plan) to category 1 (imposed as a condition of the marketing authorisation, which is included in the risk management plan (RMP)). The programme was extended in duration with the requirement for annual progress reports and additional interim reports; in total, 11 interval or cumulative interim/final reports since 2015 were provided. The four pharmacoepidemiological database studies, each covering all four indications, included a total of 355 152 patients receiving rivaroxaban and 338 199 patients receiving VKAs. The programme completed in autumn 2020, with EMA’s assessment and opinion adopted in September 2022 and endorsed by the European Commission in December 2022.
Patient and public involvement
There was no patient or public involvement.
Opportunities demonstrated from the PASS programme
This unique PASS programme exemplifies an approach whereby the prescription of a medication and its safety and effectiveness can be evaluated in a single initiative, covering all indications to be assessed, and by using well established and validated population-based European databases already familiar to researchers, industry and regulators. By using a cohort study design in the database studies, we were able to evaluate real-world patterns of rivaroxaban use including episodes of use between treatment interruptions, and by performing nested case-control analyses we were able to evaluate its safety and effectiveness, including by duration and recency of use. Additionally, although the coding structures of the country-specific databases differed, study investigator teams harmonised the design, objectives, exposures, outcomes and available look-back periods of the four studies through clear communication and transparency. However, while this approach was implemented successfully, we recognise that it was labour intensive and more operationally efficient processes are now available for managing these scenarios. The past decade has seen major developments in harmonising data from country-specific electronic health record (EHR)/claims databases with different coding systems into a common data model (CDM), enabling the same analytical tools to be applied. One such initiative, increasingly being adopted in pharmacoepidemiological research, is the Observational Health Data Sciences and Informatics (OHDSI) community, which provides access to a global network of administrative claims and EHR data sources standardised to OHDSI’s Observational Medical Outcomes Partnership CDM 5.5 Providing that no detailed information is lost through the application of a CDM, we advocate consideration of such an approach in future multicountry umbrella programmes that encompass several pharmacoepidemiological studies, as in the Data Analysis and Real World Interrogation Network European Union (DARWIN EU) project.6
Challenges encountered during the PASS programme
Two major challenges that were encountered throughout the lengthy PASS programme were related to: (1) finalising country-tailored study designs before the extent of rivaroxaban uptake was known, and (2) new research questions that arose during the programme. We outline these challenges below and propose ways to manage them for future PASS initiatives (see also tables 1 and 2).
Examples/wider context of the two main challenges encountered during the PASS programme
Summary of opportunities recognised and challenges encountered during the rivaroxaban PASS programme, and recommendations to help effectively address these should they arise in future
Finalising study designs before product uptake is known
In contrast to randomised controlled trials or epidemiological studies of established medications, the level of future uptake of newly approved medications cannot necessarily be accurately predicted at product launch. Also, the sample size and type of drug exposure can vary between countries due to treatment guidelines and reimbursement practices introduced at different times. Consequently, upfront planning of prospective pharmacoepidemiological safety studies with undefined follow-up duration—potentially of many years—is challenging. Rivaroxaban uptake for the ACS indication was expected to be low in the initial months after EMA approval yet to increase over time. However, uptake remained low throughout follow-up. This limited the drawing of any robust conclusions (after data analysis) about risk factor distribution between these patients in clinical practice versus those in the pivotal ATLAS ACS2 TIMI 51 trial population. In hindsight, it would have been operationally more efficient, and in line with pharmacoepidemiological thinking, to have considered a stepwise approach starting with a drug utilisation study to inform the regulator about the feasibility of a safety study for this indication. This could have involved formulating potential scenarios, based on different assumptions of projected sample sizes and timelines, to inform when a sufficient sample size would be obtained for a safety study.7 Such an approach could potentially be harnessed for any drug indication. It could also inform about appropriate timings for final data analysis, which could be earlier for some indications—as transpired with the SPAF indication due to high uptake and availability of a large dataset soon after the study began—thereby enabling earlier regulatory review and dissemination of results. Like a clinical trial scenario, we also advocate systematically including a statement in the study protocols that (1) premature study cessation will be considered whenever clinically justified, and (2) interim analyses will be conducted where the study objectives will be reviewed (ie, further objectives added where deemed necessary).
Research questions/challenges that emerged throughout the programme
We received several relevant scientific questions following programme completion that related to safety. Although some of these did not relate to the original study protocol, we were able to provide satisfactory responses following appropriate post-hoc analyses based on additional data released. Furthermore, we were also able to adequately respond to requests relating to data gaps specified in the RMP concerning specific patient subgroups. Another challenge encountered during the programme was that direct oral anticoagulants overtook VKAs as the standard of care for the indications studied. Consequently, patients treated with a VKA during the end of the study period were probably substantially different from those treated in the beginning,8 making safety comparisons between rivaroxaban and VKAs challenging.
Key lessons learned
We propose three key strategies that could help avoid/effectively manage these challenges in future PASS initiatives to facilitate operationally efficient programmes that generate timely results and enable robust conclusions to be drawn from the analyses. First, we advocate maintaining early, clear, continuous and open communication with the regulator, as encouraged by the EMA through the seeking of their scientific advice. Second, for programmes intended to span several years, we propose setting milestones for study review following an initial period of data collection. This could potentially be linked to the formal interim review period where key discussions between the regulator and marketing authorisation holder (MAH) are undertaken—preferably during meetings to accompany the assessment reports. Crucially, this would be an opportune time for the MAH and regulator to discuss any additional study questions that arose after PASS initiation, and for feedback on programme direction. A proactive approach involving dialogue about feasibility evaluations, and appropriate changes to statistical analysis plans, could then be undertaken accordingly, in turn helping investigators with resource and operational planning. The third is to, in consultation with the regulator, allow adaptation of study protocols to enable the inclusion of additional, or changes in, study objectives as well as modifications of study designs to reflect evolving prescription behaviours, newly emerging treatments and guidelines changes. This would ensure that the outcomes of the PASS are relevant for regulators and clinical practice.
Ethics approval
Not applicable.
Acknowledgments
We thank Susan Bromley, EpiMed Communications, Abingdon, UK, for medical writing assistance, funded by Bayer AG and in accordance with Good Publication Practice. We thank Irene Bezemer, former employee of PHARMO Institute for Drug Outcomes Research, Utrecht, Netherlands, for contributing to the harmonisation of the studies. We also thank Saad Shakir, Drug Safety Research Unit, Southampton, UK for conducting event monitoring studies as an integral part of the overall PASS programme. Finally, we thank Mari-Ann Wallander, former employee at Bayer AG, Berlin for creating the first version of the epidemiology PASS programme in collaboration with the investigators.
Footnotes
Contributors GB developed the original manuscript draft. All authors commented on subsequent drafts and approved the final version for journal submission.
Funding This manuscript relates to experiences and lessons learned from a post-authorisation safety study programme funded by Bayer AG.
Competing interests GB was an employee of Bayer AB, Sweden, at the time the PASS programme was conducted and has received consultancy fees from Bayer AG, Germany, thereafter as an external consultant. YB, CT, KS-W, MS-G, PV, TD, GN and PA are current employees of Bayer AG, Germany. MH was an employee of Bayer AG, Germany, at the time the PASS programme was conducted. AR and LAGR work for Spanish Centre for Pharmacoepidemiological Research, Spain, which received research funding for the study carried out within the rivaroxaban PASS programme. TS is an employee of Leibniz Institute for Prevention Research and Epidemiology, Germany, which received research funding for the study carried out within the rivaroxaban PASS programme. AV was an employee at BIPS at the time when the study was conducted. RH, ES and KMAS work for PHARMO Institute for Drug Outcomes Research, Netherlands, which received research funding for the study carried out within the rivaroxaban PASS programme. LF works for Friberg Research AB, Sweden, which received research funding for the study carried out within the rivaroxaban PASS programme.
Provenance and peer review Not commissioned; externally peer reviewed.