Use of direct oral anticoagulants in patients with atrial fibrillation in Scotland: Applying a coherent framework to drug utilisation studies

Abstract Purpose To report the use of direct oral anticoagulants (DOACs) for stroke prevention in patients with atrial fibrillation in Scotland and advocate the standardisation of drug utilisation research methods. Methods Retrospective cohort study using linked administrative data. Patients included those with a diagnosis of atrial fibrillation (confirmed in hospital) who received a first prescription for a DOAC (dabigatran, rivaroxaban, or apixaban) from September 2011 to June 2014. Drug utilisation measures included discontinuation, persistence, and adherence. Results A total of 5398 patients (mean CHA2DS2‐VASc score 2.98 [SD 1.71], 89.7% with ≥5 concomitant medicines) were treated with DOACs for a median of 228 days (interquartile range 105‐425). Of 35.6% who discontinued DOAC treatment, 11.0% switched to warfarin, and 48.3% reinitiated DOACs. Persistence after 12 and 18 months was 75.9% and 69.8%, respectively. Differences between individual DOACs were observed: Discontinuation rates ranged from 20.4% (apixaban) to 60.6% (dabigatran) and 12 months persistence from 60.1% (dabigatran) to 85.5% (apixaban). Adherence to treatment with all DOACs was good: Overall DOAC median medication refill adherence was 102.9% (interquartile range 88.9%‐115.5%), and 82.3% of patients had a medication refill adherence > 80%. Conclusions In Scotland, adherence to DOAC treatment was good, and switching from DOAC to warfarin was low. However, discontinuation and persistence rates were variable—although treatment interruptions were often temporary. To decrease the inconsistencies in drug utilisation methods and facilitate meaningful study comparison, the use of a coherent framework—using a combination of discontinuation, persistence, and adherence—and the standardisation of measurements is advocated.

instructions. 1 Nonadherence to drugs is widespread [1][2][3] and has been linked to increases in morbidity, premature death, and health care expenditure; 1,4,5 especially nonadherence to drugs with complex pharmacological profiles such as warfarin and other vitamin K antagonists (VKAs) is known to negatively affect treatment outcomes. [6][7][8][9] Warfarin is used for multiple cardiovascular conditions, and high discontinuation rates of warfarin treatment have been reported in clinical trials and observational studies; [10][11][12] poor adherence has been ascribed to a variety of issues, from the occurrence of bleeding events to the inconvenience of treatment. 6,[13][14][15] As warfarin is widely used long-term in patients with atrial fibrillation (AF)-a common condition causing irregular heartbeat and, as such, a major independent risk factor for stroke 16,17 -efforts to replace warfarin have resulted in 4 new direct oral anticoagulants (DOACs) being introduced since 2008: dabigatran, a direct thrombin inhibitor, and rivaroxaban, apixaban, and edoxaban, direct factor Xa inhibitors.
Treatment with DOACs follows easier dosing schemes and theoretically requires no monitoring, primarily because their therapeutic windows are wider and interactions with other drugs/foods are shown to be less than with warfarin. 18,19 Treatment with DOACs is deemed as effective and safe as with warfarin; 20-23 the effect of nonadherence on bleeding risk and stroke incidence among AF patients has been addressed but as yet not intensely studied. 24,25 Recently, concerns have been raised about the potential impact of no monitoring and the presence of multimorbidity and polypharmacy on DOAC adherence. 24,[26][27][28][29] Knowledge about patients' adherence to DOACs in real life is however still limited, and studies conducted thus far differ in scale, methodology, and focus. 25,[30][31][32][33][34][35][36][37][38][39][40][41] Results emerging from DUR studies are frequently used to compare uptake and use of drugs over time and between regions. Additionally, information on treatment adherence can be used to improve the accuracy of drug exposure estimates when investigating treatment outcomes in clinical practice-as compared to clinical trials results. 42 Nevertheless, although a comprehensive framework of drug adherence and a common terminology for measuring adherence have been proposed by the European Society for Patient Adherence, Compliance and Persistence, 43 DUR studies still make use of a variety of definitions and measurements, with methods of calculation also differing. [44][45][46] The objective of this study is therefore 2-fold: first, to report on the use of DOACs for stroke prevention in patients with AF identified in Scottish secondary care; and second, to advocate the standardisation of DUR by applying an evolving methods approach, based on a sound theoretical framework and well-documented measurements of adherence.

| METHODS
In Scotland (population approximating 5.3 million 47 ), all residents are covered by the publicly funded National Health Service (NHS), with most clinical services including prescriptions provided free of charge at the point of care. 48 Prescription details have been derived from the Prescribing Information System (PIS)-an NHS Scotland national database created primarily for reimbursement purposes. 49 The PIS holds data on the prescribed and dispensed items, patients, and prescribers, including a range of drug-specific information based on the British National Formulary. 49 to all residents registered with a general practitioner in Scotland. 54 The study population comprises patients with a diagnosis of AF, con- Drug utilisation can be divided into 3 distinct parts: initiation, implementation, and discontinuation. 43,55 While initiation and discontinuation indicate the start and end of a therapy, the process of implementation illustrates whether medication was taken as prescribed. To give a valid representation of patients' drug taking behaviour and enable analysis of drug exposure, duration and intensity of treatment should be taken into account. Hence, this study includes measures of both discontinuation/persistence and adherence. Discontinuation has been calculated using the refill-gap method with censoring of patients after the first discontinuation event, 56 defined as a gap of more than 28 days without supply following the assumed end of a prescription, based on summary statistics of the data and comparable to previous studies. 30,35 In addition, a second discontinuation rate-allowing for reinitiation of treatment during the study period-has been introduced (cessation rate). In line with the literature, persistence at prespecified points in time (6, 12, and 18 months after treatment initiation) has been assessed using the anniversary method to account for intermediary KEY POINTS • Adherence to DOAC treatment was good, and switching from DOAC to warfarin was low. However, discontinuation and persistence rates were variable.
• Standardisation of drug utilisation studies using the ESPACOM framework, combining measurements of discontinuation, persistence, and adherence, is strongly advocated.    Table 3 for details.

| DISCUSSION
This is the first study in Scotland using linked data from PIS and SMR01 to analyse discontinuation, persistence, and adherence to DOAC treatment and one of a small number of studies analysing routinely collected data at a national level. 25 ing that the method of calculation where patients are censored after the occurrence of a first discontinuation event likely leads to an overestimation of discontinuation. This lower discontinuation rate is also more in agreement with the persistence rates found in this study, calculated using the anniversary method that is insensitive to periods of treatment interruptions. 56 The percentage of patients switching from a DOAC to VKA treatment was low in this study (11.0%) compared to patients in Denmark (51.2%) and Japan (54%), but higher than in England (6.0%), and similar to the 1-year follow-up EORP-AF pilot registry (11.8%); 35,36,38,59 the reasons for these diverse findings are unclear, but might be rooted either in differences in clinical guidelines and physicians' preferences, or in timing of studies and availability of DOACs.
Although many studies report on DOACs as a group rather than separately by individual drugs, our results confirm previous findings indicating sizable differences between individual drugs: Dabigatran frequently exhibits discontinuation rates higher than rivaroxaban and apixaban. 32,38 Consequently, 12 months persistence has reportedly Discontinuation of DOAC treatment has often been attributed to changes in underlying disease severity, including restoration of sinus rhythm, worsening kidney function, and side effects including bleeding events or, particularly in case of dabigatran, gastrointestinal disturbances. 31,38,41 Because of the data available for this study, specific reasons for treatment discontinuations among the study population remain unknown.
Adherence to medication was high for all drugs and did not considerably decline over time. Treatment gaps were rare, patients generally having enough medication to cover the treatment period; median medication refill adherence, and compliance rate during the first 6 months of treatment, indicated oversupply rather than undersupply of drugs, although these findings might be due to the timing of prescriptions.
High median adherence to DOAC treatment has been shown before:  To calculate adherence, only patients who received at least 2 prescriptions have been included. Discrepancies between the total number of patients receiving 2 prescriptions for any DOAC (n = 4555) and the sum of patients receiving at least 2 prescriptions for dabigatran, rivaroxaban, or apixaban (n = 4507) are due to patients switching drugs after only one prescription for an initial drug. a Includes all patients with at least 2 prescriptions during the study period.
b Includes only patients with sufficient follow-up time to cover the respective prescription period.
limit the impact of variations in dosing schedules on adherence. Finally, as prescription records do not cover secondary care, in-patient periods are not included; this might have impacted adherence and persistence, as hospital days could have appeared to be treatment interruptions.
Sensitivity analysis of the lengths of admissible gaps and the additional measurement of treatment cessation have been used to account for the potential effect of in-patient episodes on discontinuation and persistence.
This study has nevertheless several strengths: Access to health care is universal, and electronic health records in Scotland cover the entire population. Because of the presence of a unique patient identifier, records can easily and reliably be linked; a large variety of variables is therefore available, including those essential for calculating adherence to medication. Furthermore, PIS and SMR01 have previously been used for research, and validity and accuracy of the data has been established. 49,52

| CONCLUSION
In Scotland, adherence to DOAC treatment was high, and switching from DOAC to warfarin was low. However, discontinuation and persistence rates were variable-although treatment interruptions were often temporary. The effects of nonadherence, including treatment interruptions, on the safety and effectiveness of DOACs need to be investigated further; more research is needed to analyse whether treatment with DOACs does indeed result in better disease outcomes as compared to warfarin.
To decrease the inconsistencies in drug utilisation methodology impacting the comparability of results across studies, the use of a coherent framework-using a combination of discontinuation, persistence, and adherence-and the standardisation of measurements is strongly advocated.

ETHICS STATEMENT
The authors state that no ethical approval was needed.