Article Text

Original research
Comparison of medication adherence to different oral anticoagulants: population-based cohort study
  1. Arnar B Ingason1,2,
  2. Jóhann P Hreinsson3,
  3. Sigrún H Lund4,
  4. Arnar S Ágústsson1,2,
  5. Edward Rumba1,2,
  6. Daníel A Pálsson1,2,
  7. Indriði E Reynisson5,
  8. Brynja R Guðmundsdóttir6,
  9. Páll T Önundarson1,6,
  10. Einar S Björnsson1,2
  1. 1Faculty of Medicine, University of Iceland, Reykjavik, Iceland
  2. 2Department of Gastroenterology and Hepatology, Landspitali University Hospital, Reykjavik, Iceland
  3. 3Department of Molecular and Clinical Medicine, Sahlgrenska Academy, Goteborg, Sweden
  4. 4DeCODE Genetics, Reykjavik, Iceland
  5. 5Fjordur Health Care Center, Hafnarfjordur, Iceland
  6. 6Department of Laboratory Hematology, Landspitali University Hospital, Reykjavik, Iceland
  1. Correspondence to Dr Arnar B Ingason; arnarbingason{at}gmail.com

Abstract

Objective Previous observational studies have yielded conflicting results on whether medication adherence differs between patients receiving warfarin and direct oral anticoagulants (DOACs). Importantly, no study has adequately accounted for warfarin dosing being continuously modified based on INR values while dosing of DOACs is fixed. We aimed to compare non-adherence between new users of apixaban, dabigatran, rivaroxaban and warfarin in a population-based cohort.

Methods New users of apixaban, dabigatran, rivaroxaban and warfarin from 2014 to 2019 living in the Icelandic capital area were included. Non-adherence was defined as proportion of days covered below 80%. Inverse probability weighting was used to yield balanced study groups and non-adherence was compared using logistic regression. Factors associated with non-adherence were estimated using multivariable logistic regression.

Results Overall, 1266 patients received apixaban, 247 dabigatran, 1566 rivaroxaban and 768 warfarin. The proportion of patients with non-adherence ranged from 10.5% to 16.7%. Dabigatran was associated with significantly higher odds of non-adherence compared with apixaban (OR 1.57, 95% CI 1.21 to 2.04, p<0.001), rivaroxaban (OR 1.45, 95% CI 1.12 to 1.89, p=0.005) and warfarin (OR 1.63, 95% CI 1.23 to 2.15, p<0.001). The odds of non-adherence were similar for apixaban, rivaroxaban and warfarin. Apart from the type of oral anticoagulants (OACs) used, female sex, hypertension, history of cerebrovascular accident and concomitant statin use were all independently associated with lower odds of non-adherence.

Conclusion Dabigatran was associated with higher odds of non-adherence compared with other OACs. Non-adherence was similar between apixaban, rivaroxaban and warfarin users. Female sex and higher comorbidity were associated with better medication adherence.

  • anticoagulation
  • quality in health care
  • risk management
  • clinical pharmacology

Data availability statement

Data are available upon reasonable request. Technical appendix and statistical code are available from ABI (abingason@gmail.com) upon request. Study dataset is available with license from the National Bioethics Committee (vsn@vsn.is).

http://creativecommons.org/licenses/by-nc/4.0/

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

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • The study was population-based and used robust propensity score-weighting model to account for indication bias.

  • The study accounted for all dose adjustments for both warfarin and direct oral anticoagulants during the study period, greatly increasing the accuracy of non-adherence calculations.

  • As data on dose adjustments for warfarin were only available from Landspitali University Hospital (the only university hospital in the capital area), the primary analysis was restricted to patients living in the capital area.

  • The propensity score-weighting model did not account for important socioeconomic factors that may affect risk of non-adherence, such as smoking, alcohol consumption, education level or household income.

Introduction

Based on the outcomes of large clinical trials, direct oral anticoagulants (DOACs) are currently recommended as first-line treatment for patients with atrial fibrillation and venous thromboembolism (VTE).1–4 However, concerns have been raised that the efficacy of DOACs may be lower in clinical practice than reported in clinical trials,5 possibly due to lower medication adherence. Warfarin treatment needs to be controlled by regular INR measurements, while no such monitoring is deemed necessary for DOACs. However, the regular monitoring of warfarin treatment may not only secure optimal therapeutic dosing, but may also serve as a safety marker ensuring that the drug is used correctly and that patients are adherent to their dosing regimen.

Results from observational studies comparing adherence between warfarin and DOACs have been ambiguous.6–9 Importantly, previous studies have failed to account for dosing adjustments between prescriptions for warfarin.6–9 This is important as warfarin dosing is continuously being modified according to INR values, while DOACs are prescribed at fixed doses. While patients are usually started on a standard dose when initiating warfarin treatment, the final maintenance dose can vary by as much as 40-fold.10 Therefore, results from previous studies may be unreliable. Indeed, a previous study from Sweden comparing the adherence and persistence of oral anticoagulants (OACs), excluded patients receiving warfarin when comparing adherence as ‘the proportion of days covered could not be calculated for warfarin treatment due to the highly variable dosage regimens’.11

This study aimed to compare non-adherence between warfarin and DOAC users in a population-based cohort where all dose adjustments were taken into account.

Methods

Study population

The Icelandic OAC database has been previously described in detail.12 Briefly, data were collected on all patients receiving their first drug prescription for OAC from 1 March 2014 to 28 February 2019 using the Icelandic Medicine Registry, which contains a centralised record of all outpatient drug prescriptions in the country. The personal identification numbers of these patients were subsequently linked to the electronic medical record system of Landspitali University Hospital, and all four regional hospitals in Iceland (Akureyri, Akranes, Ísafjörður and Neskaupsstaður hospitals).

To yield an OAC naïve cohort, patients were excluded if they had filled an OAC prescription in the preceding 12 months before their eligibility in the study. Additionally, as the catchment area of the Landspitali Anticoagulation Management Center is limited to the capital area, we excluded patients with residence outside the capital area. Furthermore, patients were excluded if they had a mechanical heart valve, mitral stenosis, end-stage renal disease or treatment indication other than atrial fibrillation, VTE or ischaemic stroke (figure 1). Finally, patients with a follow-up of less than 30 days or who only had a single OAC prescription were excluded, as well as patients receiving warfarin who had missing data on dose adjustments.

Figure 1

Flowchart for selection of study cohort. AF, atrial fibrillation; OAC, oral anticoagulant; VTE, venous thromboembolism.

Follow-up

Patients were followed from the filling of their first OAC prescription until 28 February 2019 or earlier if treatment was discontinued or switched to another OAC, if the patient had a major bleeding or thromboembolic event, or if death occurred. Adherence was calculated from the date of first prescription to the date of the last prescription before follow-up stopped. As the last OAC prescription before the end of follow-up was excluded from calculations, the calculated adherence was unaffected by abrupt treatment cessations.

Study outcomes

The primary study outcome was medication non-adherence, defined as proportion of days covered (PDC) below 80%. This cut-off has been shown to be optimal in stratifying adherent and non-adherent patients,7 13 and has been widely used in previous studies.7–9 13 Secondary outcomes were rates of any thromboembolic event or major bleeding, and factors associated with non-adherence.

Calculation of medication adherence

In Iceland, 75% of all warfarin treatment is regulated by the Icelandic Anticoagulation Management Center, which has a database containing information on every dose adjustment between visits for these patients. Using these data, we calculated the mean daily dose of warfarin for each patient (in mg). The PDC was subsequently calculated by dividing the total amount of warfarin dispensed (in mg) by the predicted amount needed during the study period (in mg).

As opposed to warfarin, the dosing of DOACs is fixed. Therefore, PDC for DOACs was calculated as the number of tablets dispensed during the study period divided by the total amount needed during that period. This accounted for patients being switched from standard to reduced dosing during the study period (or less commonly vice versa). Patients receiving rivaroxaban due to VTE were estimated to have received rivaroxaban 15 mg two times a day for 3 weeks, followed by 20 mg once a day as per the medication’s monograph. Similarly, patients receiving apixaban due to VTE were estimated to have received dosing of 10 mg two times a day for 1 week, followed by 5 mg two times a day. All other doses were estimated to be once daily for rivaroxaban, and two times a day for apixaban and dabigatran.

Data acquisition

Information on thromboembolic and bleeding events during the study period, as well as prior bleeding and thromboembolic events, comorbidities and indication for treatment were gathered using relevant ICD-10 codes as previously described (online supplemental table 1).12 Additionally, events were identified by searching the Icelandic death registry and by manually reviewing results from computed tomographies of the head and pulmonary arteries and all endoscopic procedures. To estimate the comorbidity burden of patients, Charlson comorbidity index and CHA2DS2-VASc scores were calculated for each patient using previously verified ICD-10 codes.14 15

Information on concomitant drug use was obtained from the Icelandic Medicine Registry (online supplemental table 2). Concomitant drug use was defined as filling a relevant drug prescription within 6 months of the start of a patient’s follow-up.

Statistical analysis

Inverse probability weighting (IPW) was used to achieve balanced study groups. It assigns weights to patients based on propensity scores, therefore yielding a balanced pseudopopulation that includes the whole study population. The following variables were used in the model: age, sex, treatment indication, all variables in the Charlson comorbidity index (except for AIDS, which was too sporadic), bleeding or coagulation disorders, hypertension, history of VTE or gastrointestinal bleeding requiring hospital admission, and concomitant use of antihistamines, antihypertensives, antiplatelets, corticosteroids, non-steroidal anti-inflammatory drugs, proton pump inhibitors, selective serotonin receptor inhibitors and statins. Standardised mean difference was used to evaluate balance between study groups after weighting, with values below 0.1 being considered ideal and values below 0.2 being considered acceptable.12 16 The model yielded acceptable balance for all variables, except treatment indication and the highly correlated variable history of VTE. Therefore, a sensitivity analysis including patients with atrial fibrillation only was performed. Additionally, to account for potential differences in medication adherence based on area of residence, a second analysis was performed which included patients receiving DOACs and living in all regions of Iceland. This analysis used the same statistical methods as described above except area of residence was included as a variable in the IPW model as well. Finally, as the average length of follow-up varied considerably between individual OACs, a sensitivity analysis was performed with follow-up limited to the first 18 months of treatment.

Non-adherence was compared using propensity score-weighted logistic regression that accounted for length of follow-up. Univariate analysis was performed to identify factors associated with non-adherence. Categorical variables were compared using the χ2 test and continuous variables using analysis of variance. This analysis compared 32 variables. Therefore, to account for multiple testing, a p value of less than 0.0016 was considered significant. Multivariable analysis was performed including variables from the univariate analysis with significant association with non-adherence. For this analysis, the linearity assumption of continuous variables was assessed by visualising the logit of non-adherence in quantiles of OAC users and the goodness of fit was assessed using the Hosmer-Lemeshow test.

Statistical analysis was performed in R, V.4.2.1 (R Foundation for Statistical Computing), using RStudio, V.2022.07.1. All statistical tests were two-tailed. Apart from the univariate analysis, a p value of less than 0.05 was considered significant.

Patient and public involvement

There was no active patient involvement in this study.

Results

Study population

In total, 14 611 patients received OAC during the study period. Of those, 6967 patients were excluded as they had filled an OAC prescription during the preceding 12 months, 2761 patients as they had permanent residence outside the capital area and 782 patients were excluded as they had a follow-up of less than 30 days or only filled a single OAC prescription during the follow-up period. Additionally, 254 patients were excluded for other reasons as listed in figure 1. The final study population consisted of 3847 patients. Thereof, 1266 patients had received apixaban, 247 dabigatran, 1566 rivaroxaban and 768 warfarin. The mean follow-up period was 1.4 years for patients receiving apixaban, 2.1 years for dabigatran, 2.0 years for rivaroxaban and 1.3 years for warfarin. Baseline characteristics of the study population are provided in table 1.

Table 1

Baseline characteristics of study population

Comparison of medication non-adherence between users of individual OACs

Overall, the majority of patients had near-perfect adherence, with a median PDC of 100% for apixaban users (IQR 94.3%–100%), 99.7% for dabigatran users (IQR 90.8%–100%), 100% for rivaroxaban users (IQR 95.2%–100%) and 97.0% for warfarin users (IQR 85.4%–100%). The distribution of PDC for the overall population is provided in figure 2 and online supplemental figure 1. When patients were stratified by length of follow-up using 6 month intervals, the proportion of patients with non-adherence was around 15.2%–18.9% for patients with less than 18 months of follow-up, but 7.1%–8.1% for patients with longer than 18 months of follow-up (online supplemental figure 2).

Figure 2

Bar graph demonstrating the distribution of proportion of days covered for the overall study population. Non-adherence was defined as proportion of days covered below 80%.

Before propensity score-weighting, 16.7% of warfarin users (95% CI 14.0% to 19.3%), 16.2% of dabigatran users (95% CI 11.6% to 20.8%), 12.4% of rivaroxaban users (95% CI 10.8% to 14.0%) and 10.5% of apixaban users (95% CI 8.8% to 12.2%) were non-adherent (figure 3A). After adjusting for propensity score-weighting and length of follow-up, dabigatran was associated with significantly higher odds of non-adherence compared with apixaban (15.5% vs 11.9%, OR 1.57, 95% CI 1.21 to 2.04, p<0.001), rivaroxaban (15.5% vs 11.3%, OR 1.45, 95% CI 1.12 to 1.89, p=0.005) and warfarin (15.5% vs 11.1%, OR 1.63, 95% CI 1.23 to 2.15, p<0.001). The odds of non-adherence were similar between apixaban, rivaroxaban and warfarin users (figure 3B). The results were similar when the analysis was limited to 18 months of follow-up (online supplemental figure 3).

Figure 3

Bar graphs comparing the proportion of patients with non-adherence for the primary analysis (A) before and (B) after inverse probability weighting. Non-adherence was defined as proportion of days covered below 80%. Values are presented as means ± 95% confidence intervals.

Patient characteristics associated with non-adherence

Patients who were found to be non-adherent were younger than adherent patients, had a lower CHA2DS2-VASc score and were more commonly males (table 2). Non-adherent patients were also less likely to have hypertension, history of cerebrovascular accident, or to have received concomitant statin treatment.

Table 2

Univariate analysis comparing factors associated with non-adherence

After multivariable logistic regression, dabigatran usage and male gender were both associated with higher odds of non-adherence. Meanwhile, hypertension, history of cerebrovascular accident and concomitant use of statins were all independently associated with lower odds of non-adherence (table 3).

Table 3

Logistic regression estimating factors associated with non-adherence

Comparison of outcomes between adherent and non-adherent patients

Rates of thromboembolism (1.9 events per 100 person-years (events/100-py) vs 2.1 events/100-py, HR 0.86, 95% CI 0.31 to 2.37) and major bleeding (2.8 events/100-py vs 2.3 events/100-py, HR 0.97, 95% CI 0.55 to 1.73) were similar between adherent and non-adherent patients.

Sensitivity analysis

A sensitivity analysis including patients with atrial fibrillation only was performed. This analysis included 1104 patients receiving apixaban, 223 receiving dabigatran, 1307 receiving rivaroxaban and 229 receiving warfarin. The mean follow-up period was 1.5 years for apixaban, 2.2 years for dabigatran, 2.2 years for rivaroxaban and 1.6 years for warfarin. IPW yielded an acceptable balance between all study groups (online supplemental table 3).

The crude proportion of patients with non-adherence was 9.9% for apixaban, 13.4% for rivaroxaban, 13.5% for warfarin and 16.6% for dabigatran (figure 4A). After propensity score-weighting, dabigatran was associated with higher odds of non-adherence compared with apixaban (16.2% vs 10.5%, OR 2.05, 95% CI 1.51 to 2.81, p<0.001), rivaroxaban (16.2% vs 12.5%, OR 1.34, 95% CI 1.00 to 1.80, p=0.05) and warfarin (16.2% vs 7.6%, OR 2.75, 95% CI 1.63 to 3.37, p<0.001). Additionally, rivaroxaban was associated with higher odds of non-adherence compared with apixaban (OR 1.53, 95% CI 1.11 to 2.12, p=0.009) and warfarin (OR 2.05, 95% CI 1.41 to 2.01, p<0.001) (figure 4B).

Figure 4

Bar graphs comparing the proportion of patients with non-adherence for patients with atrial fibrillation only (A) before and (B) after inverse probability weighting. Non-adherence was defined as proportion of days covered below 80%. Values are presented as means ± 95% confidence intervals.

Similar to the primary analysis, dabigatran use, younger age and male sex were all independently associated with higher odds of non-adherence, while a history of cerebrovascular accident and concomitant statin use were associated with lower odds of non-adherence. Similarly, rates of major bleeding (3.0 events/100-py vs 2.1 events/100-py, HR 0.83, 95% CI 0.43 to 1.58) and thromboembolism (1.5 events/100-py vs 1.6 events/100-py, HR 0.97, 95% CI 0.35 to 2.68) were similar between adherent and non-adherent patients.

A second analysis including patients receiving DOACs and living anywhere in the country was performed to account for potential differences due to area of residence. Comparison of the baseline characteristics of this population is provided in online supplemental table 4. Similar to the primary analysis, dabigatran was associated with higher odds of non-adherence compared with both apixaban (16.3% vs 11.5%, OR 1.81, 95% CI 1.47 to 2.23, p<0.001) and rivaroxaban (16.3% vs 11.6%, OR 1.53, 95% CI 1.25 to 1.88, p<0.001), while no differences were noted between apixaban and rivaroxaban. The odds of non-adherence were similar between patients who lived in the capital area and those who did not (OR 1.05, 95% CI 0.87 to 1.27).

Discussion

In this population-based study, medication adherence for OACs was found to be good overall, with a median PDC around 100% for all medications. After accounting for patients’ baseline characteristics, non-adherence was significantly more common among dabigatran users compared with patients receiving other OACs, while the odds of non-adherence were similar between apixaban, rivaroxaban and warfarin.

The higher odds of non-adherence for patients receiving dabigatran reported in the current study are consistent with several previous studies.7 9 11 17–22 This lower adherence might be due to a combination of two factors. First, it is administered two times a day as opposed to warfarin and rivaroxaban which are given as a single daily dose. Previous studies on chronic cardiovascular medications have demonstrated that once a day dosing is associated with greater adherence compared with two times a day dosing.23 However, apixaban is also administered two times a day but had similar odds of non-adherence as rivaroxaban and warfarin in the primary analysis. Additionally, apixaban was associated with lower odds of non-adherence than rivaroxaban when only patients with atrial fibrillation were analysed. Second, dabigatran’s lower adherence has been speculated to be due to its frequent gastrointestinal side effects.12 24 Supporting this, 12% of patients receiving dabigatran in the RE-LY trial reported dyspepsia and three times as many patients discontinued treatment due to gastrointestinal upset compared with warfarin.25

Previous studies comparing adherence between warfarin and DOACs have yielded conflicting results.6–9 This might, at least partly, be due to the inability of these studies to account for dosage adjustments being performed for warfarin according to patient’s INR values. For example, while patients are usually put on a standard dose when initiating warfarin treatment, the final maintenance dose can vary by as much as 40-fold.10 This can easily lead to false estimations of non-adherence if dose adjustments between drug prescriptions are not taken into account. In the current study, all dose adjustments during warfarin treatment were taken into account, greatly increasing the accuracy of the data. Similarly, the study accounted for patients on DOACs switching from high-dose treatment to reduced doses or, less commonly, vice versa.

Non-adherence in the current study was found to be 10.5%–16.7%. This is comparable to previous studies from Scandinavia that have reported non-adherence ranging from 4.3% to 23.2%,11 26 but considerably lower than the reported non-adherence for OACs in American studies which has ranged from 24.6% to 61.5%.7 9 22 27–30 This difference may be due to the different study settings. Most studies from the USA have gathered data on drug prescriptions from either Veteran Health Administration pharmacies27 or insurance claims9 22 28–30 which may be more prone to missing prescriptions compared with the centralised nationwide prescription database used in the current study. Additionally, the observed difference may, at least partly, be explained by increased social disparity in the American population, as patients with low socioeconomic status may not afford to fill their drug prescriptions on time.

Apart from the type of OAC received, female gender, hypertension, history of cerebrovascular accident and concomitant statin use were all associated with lower odds of non-adherence. This suggests that patients with higher comorbidity burden are at reduced risk of non-adherence, which is reassuring as those patients are likely at higher risk of complications following inadequate drug consumption.

In the current study, thromboembolism and major bleeding rates were not significantly different between adherent and non-adherent patients after accounting for baseline characteristics. In comparison, a previous study demonstrated that among patients with a CHA2DS2-VASc score of 2 or higher, poor adherence was associated with higher rates of the composite outcome of ischaemic stroke, systemic embolism and all-cause mortality.9 Similarly, these patients had lower rates of major bleeding compared with adherent patients. Another study demonstrated that poor adherence to dabigatran was associated with higher likelihood of all-cause mortality and stroke, but not myocardial infarction or major bleeding.27 The reason for the differences between previous studies and the current one is likely due to the fact that our study was based on an ‘on treatment’ analysis, where all thromboembolic and major bleeding events were manually verified and events excluded if the patients had not been receiving OACs in the preceding 2 days. Previous studies have not manually verified outcome events and are therefore more representative of an ‘indication-to-treat’ analysis.9 27

The current study has several strengths. It is population-based using previously verified databases, such as the Icelandic Medicine Registry, which includes data on all outpatient drug prescriptions in the country. The study also has minimal missing data. Additionally, to our best knowledge, it is the first study to account for all dose adjustments for warfarin and DOACs during the study period when estimating non-adherence.

The study also has several limitations. First, as data on dose adjustments for warfarin were only available from Landspitali (the only university hospital in the capital area), the primary analysis was restricted to patients living in the capital area. This is important as treatment adherence in rural areas may differ from the catchment area of university hospitals. However, a sensitivity analysis of patients receiving DOACs and living in all regions of Iceland demonstrated no difference in non-adherence between patients who lived in the capital area and those who did not. Second, although a robust propensity score-weighting model was used to account for indication bias, data were missing on socioeconomic factors such as smoking, alcohol consumption, education level and household income. These variables may be associated with medication non-adherence. In fact, lower household income and decreased socioeconomic status have both been associated with lower self-reported overall medication adherence.31 Importantly, patients with low socioeconomic status may also be more likely to receive warfarin. However, even though the cost of warfarin is lower than for DOACs in Iceland, DOACs are reimbursed, which likely substantially reduces the magnitude of this bias compared with many other populations. Third, adherence was calculated based on the assumption that all filled OAC prescriptions were consumed. How closely this resembles the true medication adherence of patients is unknown. Fourth, our study did not account for hospital admissions. In Iceland, patients are provided medications during their hospital stay, a lengthy stay may therefore lead to inaccurate calculation of non-adherence. However, the current study did censor the follow-up period to the first major bleeding or thromboembolic event, which likely attenuates this risk. Fifth, analysis of the data is complicated by the fact that while warfarin and DOACs were both considered first-line treatment at the start of the study period, during the 2016 update of the European Society of Cardiology (ESC) guidelines, DOACs were recommended over warfarin for patients with atrial fibrillation.32 Meanwhile, warfarin and DOACs were both recommended as first-line treatment for pulmonary embolism by the ESC guidelines throughout the study period.4 33 Therefore, proportionally higher proportion of patients were started on warfarin during the first half of the study period and patients receiving warfarin were more often being treated for VTE than patients receiving DOACs. Having said that, the IPW model accounted for treatment indication, and a subanalysis including only patients with atrial fibrillation demonstrated similar results as the primary analysis. Therefore, the results are unlikely to be explained by in-group differences in treatment indication. Sixth, the length of follow-up varied from 1.3 years for patients receiving warfarin to 2.1 years for patients receiving dabigatran. This is important as non-adherence may be affected by treatment length. In fact, we found that for patients with follow-up less than 18 months, non-adherence ranged from 15.2% to 18.9%, compared with 7.1%–8.1% for patients with more than 18 months of follow-up. However, a sensitivity analysis limited to the first 18 months of follow-up yielded similar results as the primary analysis. Additionally, length of follow-up was accounted for in the logistic regression when comparing non-adherence between groups. Therefore, the results are unlikely to be explained by differences in length of follow-up. Finally, the current study did not include data on time in therapeutic range (TTR) for warfarin patients. This is important as TTR is routinely used to assess the quality of warfarin management. However, previous studies from Iceland using the same population have demonstrated TTR of 77%–79%.34 35

In conclusion, after accounting for baseline patient characteristics, dabigatran was associated with higher odds of non-adherence compared with apixaban, rivaroxaban and warfarin. The odds of non-adherence were similar for patients receiving warfarin, apixaban and rivaroxaban. Apart from the type of OAC received, female sex, hypertension, history of cerebrovascular accident and concomitant statin use were all associated with lower odds of non-adherence.

Data availability statement

Data are available upon reasonable request. Technical appendix and statistical code are available from ABI (abingason@gmail.com) upon request. Study dataset is available with license from the National Bioethics Committee (vsn@vsn.is).

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by the Icelandic Bioethics Committee (VSN-18-111-V1). The need for informed consent was waived by the Icelandic Bioethics Committee as all data was anonymised and no active patient participation was needed.

References

Supplementary materials

Footnotes

  • Twitter @abingason

  • Contributors ABI, ESB, JPH, SHL and PTO designed the study. ABI and ESB were involved in obtaining funding for the study. ABI, ASA, ER, DAP, IER and BRG performed the data acquisition. ABI performed the statistical analysis and visualised the data with help from JPH and SHL. ABI wrote the first draft of the manuscript. All other coauthors critically reviewed the manuscript and approved the final manuscript before submission. ABI and ESB act as co-guarantors for the study.

  • Funding This study was funded by the Icelandic Research Fund (207113-051 to ABI) and the Landspitali University Hospital Research Fund (A-2020-009 to ESB). The funding sources had no role in the design, reporting or publishing of the study.

  • Competing interests PTO and BRG together with the Landspitali University Hospital and the University of Iceland hold a patent for the Fiix-prothrombin measurement. All other authors declare no competing interests.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.