Predictors of Direct Oral Anticoagulants Utilization for Thromboembolism Prevention in Atrial Fibrillation.

PURPOSE
Several factors have been associated with the prescription of direct oral anticoagulants (DOAC) over warfarin such as younger age, fewer concomitant medications, and lower CHADS2 or bleeding scores. The primary objective of this study was to identify predictors of DOAC choice compared with warfarin for patients who are starting a new oral anticoagulant (OAC) for atrial fibrillation (AF). The secondary objective was to describe the proportion of DOAC prescriptions in new users of OAC for AF.


METHODS
A retrospective cross-sectional study was conducted in a teaching hospital in Canada. Medical records of adult patients hospitalized in any medical units between October 1st, 2011 and October 1st, 2014, who were newly prescribed an OAC for non valvular AF were systematically reviewed. Baseline characteristics of warfarin and DOAC users were compared and a multivariate logistic regression analysis was completed to identify predictors of DOAC use. Variables included in the multiple regression analysis were: age, hypertension, diabetes, history of stroke or transient ischemic attack, coronary artery disease, peripheral arterial disease, CHADS2 score of 2 or more, creatinine clearance 30mL/min or more, polypharmacy, concomitant use of ASA or clopidogrel, and prescription by a neurologist.


RESULTS
Among OAC users (144 patients on DOAC and 295 patients on warfarin), older age (odds ratio [OR] 0.97; 95%CI 0.95-0.98), peripheral arterial disease (OR: O.41;95%CI: 0.21-0.82), polypharmacy (OR: 0.30;95%CI:0.10-0.89), and concomitant use of clopidogrel (OR: 0.19;95%CI:0.07-0.56) decreased the probability of DOAC use. Prescription by a neurologist (OR: 2.77;95%CI:1.34-5.76) and an estimated creatinine clearance of at least 30mL/min (OR: 3.53;95%CI:1.18-10.57) increased the likelihood of DOAC prescription.


CONCLUSION
To the best of our knowledge, this is the first observational study finding that concomitant use of clopidogrel reduced the likelihood of DOAC utilization while prescription by a neurologist increased the probability of receiving a DOAC over warfarin in patients with AF.This article is open to POST-PUBLICATION REVIEW. Registered readers (see "For Readers") may comment by clicking on ABSTRACT on the issue's contents page.


INTRODUCTION
Oral anticoagulants (OAC) are the cornerstone of therapy to prevent thromboembolic events associated with atrial fibrillation (AF). 1 Direct oral anticoagulants (DOAC) such as apixaban, dabigatran, edoxaban and rivaroxaban have been shown to be non inferior 2,3 or superior 4,5 in terms of efficacy when compared to warfarin in randomized clinical trials. All were associated with a lower risk of intracranial haemorrhage [2][3][4][5] , but a greater risk of gastrointestinal bleeding was found with the higher doses of dabigatran 2 and rivaroxaban 3 . Several real-world studies conducted in North America have described patterns and predictors of DOAC use in AF. [6][7][8][9][10][11][12] DOACs were prescribed in proportions of 33.9% 10 and 42.2% 11 among new users of OAC for AF. Factors associated with the prescription of a DOAC over warfarin were younger age 9-12 , male gender 10,11 , prior ischemic stroke 10 , lower scores for the risk of stroke 8,9,11,12 or bleeding 8,11,12 , fewer comorbidities 8 or comedications 8,11 , a higher _________________________________________ Corresponding Author: Marie-France Beauchesne, Faculté de pharmacie, Université de Montréal, C.P. 6128, succ Centre-Ville, Montréal, Qc, Canada; Email: mariefrance.beauchesne@umontreal.ca calculated creatinine clearance 9 , having a private medical insurance 9 , and prescription by a cardiologist 10 . Most studies focused on dabigatran, only a few assessed predictors in new users of OAC 10,11 , and several were based on data retrieved from administrative databases only 7,8,10,11 . The main objective of this study was to assess clinical predictors associated with the use of DOACs over warfarin among new users of OAC for AF in a hospital based population.

Design
A retrospective, cross-sectional study was conducted in a single teaching hospital in Canada, the Centre Hospitalier Universitaire de Sherbrooke (CHUS). The CHUS offers specialized and over-specialized care and services, including cardiology, neurosurgery, oncology and neonatology.The scientific and ethics committees of the CHUS's research center approved the conduct of this study.

Objectives
The primary objective of this study was to identify predictors of DOAC choice compared with warfarin for patients who are starting a new anticoagulation therapy for AF. The secondary objective was to describe the proportion of DOAC prescriptions in new users of OAC for AF.

Inclusion/Exclusion criteria
We included subjects who were hospitalized in all medical units between October 1 st , 2011 and October 1 st , 2014, who were at least 18 years old, who had a documented diagnosis of non valvular AF, who had a complete hospital discharge prescription documented in their medical chart, and an OAC prescribed during hospital admission and at discharge. Subjects were excluded if they were prescribed an OAC prior to hospital admission, moderate to severe mitral stenosis, heart valve repair or replacement, congenital heart disease, hypercoagulability, transient AF (i.e. lasting less than 48 hours), another indication for anticoagulation (e.g. venous thromboembolism), uncontrolled hyperthyroidism, illicit drug use (amphetamines, cocaine, crack), admission for drug intoxication (eg. tricyclic antidepressant) or perioperative AF.

Data collection and outcomes
A medical archivist identified charts of hospitalized subjects who had a documented diagnosis of AF (using ICD-10 codes I48X, I48.XN-001 and I48.XN-002), and apixaban, dabigatran, rivaroxaban or warfarin (edoxaban was not available during the study period) on the pharmacologic profile's field. We selected the first eligible hospitalization between October 1 st , 2011 and October 1 st , 2014.
Data were collected from individual electronic medical records (EMRs). More than one evaluator completed data collection. Therefore, all data collectors used the same form developed for the purpose of this study and it was pretested on ten medical charts for standardization.The information collected corresponds to the one available any time during admission, but when several values were available, the one that was closest to OAC prescription was documented. Demographic data collected included age, gender, weight, and body mass index. Medical history data included: heart failure, hypertension, diabetes, stroke or transient ischemic attack, coronary artery disease (CAD), peripheral artery disease (PAD), dementia, history of fall, and history of major bleeding. 13 The CHADS 2 score was calculated based on the information retrieved from the discharge summary. Laboratory data included serum creatinine (to estimate creatinine clearance using the Cockcroft-Gault formula) and liver enzymes (i.e., aspartate transaminase [AST] and alanine transaminase [ALT], reporting the proportion of subjects who had values higher than twice the upper limit of normal). Number of drugs at discharge, concomitant use of ASA and other antiplatelets (i.e., clopidogrel, prasugrel, ticagrelor) were documented. Polypharmacy was defined as the prescription of at least five chronic medications. We also collected the medical specialty of the OAC's prescriber.
Predictors of DOAC prescription, compared to warfarin, were assessed among incident users of OACs. Incident users were subjects who were not prescribed an OAC at the time of admission (i.e. OAC was initiated during hospitalization). By selecting only incident users, we avoided the potential bias associated with different length of oral anticoagulant use in prevalent users. Variables related to demographics, medical history, laboratory data, and concomitant medications described previously were considered as potential predictors of DOAC use.

STATISTICAL ANALYSIS
Baseline characteristics are described using proportions for categorical variables and means with standard deviations for continuous variables. To identify predictors of DOAC prescription over warfarin, univariate logistic regressions were completed to estimate the crude odds ratio (OR) for each potential predictor (i.e. demographic, medical history, and laboratory data along with concomitant medications described above). Variables that reached statistical significance (α = 0.05) in the univariate analysis were included in a multiple logistic regression analysis (with forward selection of the variables). Variables included in the multiple regression analysis were: age (continuous variable), hypertension, diabetes, history of stroke or TIA, CAD, PAD, CHADS 2 score of 2 or more, creatinine clearance 30mL/min or more, polypharmacy, concomitant use of ASA or clopidogrel, and prescription by a neurologist. The log-linearity of the continuous variables was tested with the 4-Tidwell transformation before integration in the multivariate logistic regression model. 14 Based on the log-linear analysis, the CHADS 2 score, creatinine clearance and number of medications (i.e., polypharmacy) were included as dichotomous variables. Adjusted ORs were estimated for the variables that were included in the final regression model. Data analyses were conducted using SPSS version 20.

Screening
A total of 3849 medical charts were screened and 2958 were excluded because they were from prevalent users of OAC (2401 on warfarin, 41 on apixaban, 259 on dabigatran, and 257 on rivaroxaban). From the 891 incident users of OAC, 349 on warfarin and 103 on a DOAC (total of 452) were excluded (see figure 1). The main reason for exclusion was because of the unavailability or incomplete discharge prescription. A total of 439 incidents users of OAC were included in this study; 295 on warfarin (67.2%) and 144 (32.8%) on a DOAC (29 apixaban, 38 dabigatran and 77 rivaroxaban).

Baseline characteristics
Characteristics of incident users of OACs are shown in Table 1. About half of the subjects were females, mean age was of 79,6 years old in the warfarin group and 73,9 in DOAC users. Mean CHADS2 scores were of 2.63 and 2.44 in the warfarin and DOAC groups, respectively. The most common comorbidity was hypertension. The proportions of subjects: who were 75 years of age or older, with hypertension, with diabetes, with CAD, with PAD, on ASA, on clopidogrel, with polypharmacy, were lower in the DOAC group. The mean CHADS 2 score and the mean number of drugs were lower for subjects on DOAC. The percentage of subjects with a stroke or TIA history and the mean estimated creatinine clearance were higher in the DOAC group. Overall, cardiologists mostly prescribed OAC. Furthermore, neurologists were more likely to prescribe a DOAC over warfarin.

DISCUSSION
This study describes predictors of DOAC utilization in new users of OAC. DOACs were prescribed in 32.8% of incident users. This proportion is similar to the one described by AbuDagga et al., which reported that 33.9% of OAC naïve subjects were prescribed dabigatran (OAC prescription claims from October 2010 to October 2011 in the Integrated Research Database, USA). 10 It is lower than the proportion reported by Desai et al., which was of 42.2%, but this study included outpatients covered by a health care benefits company (Aetna, USA) with OAC prescriptions between 2010 and 2013. 11 Unsurprisingly, DOACs were more likely prescribed in younger patients, such as reported by others. [9][10][11] We also found that subjects with fewer concomitant medications (based on the variable "polypharmacy") and no PAD (which could be an indicator of less comorbidities) were more likely prescribed a DOAC, similar to others. 8,11   Interestingly, subjects on clopidogrel were more likely prescribed warfarin rather than a DOAC. This could be explained by limited evidence on the safety of the DOAC-clopidogrel association. Furthermore, at the beginning of the study period, the 2010 Canadian Cardiovascular Society Atrial Fibrillation Guidelines recommended warfarin over dabigatran in patients with AF and coronary artery disease requiring an OAC and antiplatelet therapy 15 , while more recent guidelines favors DOACs based on limited evidence. 16 This result suggests that clinicians are cautious about the use of DOACs in subjects on antiplatelet therapy. Moreover, subjects on DOACs had better renal function than the ones on warfarin, which was also reported by Steinberg et al. 9 This result is reassuring considering that DOACs are contraindicated when kidney function is severely reduced. This advocates the appropriate use of DOACs in our population of subjects with AF.
A new finding was that neurologists were more likely to prescribe a DOAC. This results could be explained by the fact that DOACs are associated with a lower risk of intracranial haemorrhage, and because they are effective in stroke prevention. [2][3][4][5] Therefore, quality improvement projects on DOACs should not only target cardiologists who regularly prescribe OAC for AF, but also neurologists.
Our study has several limitations. First, data were collected retrospectively and relevant information could have been missed (eg. the presence of a comorbidity if it was not documented in the EMR). Secondly, a high number of patients were excluded, mostly because of incomplete discharge prescriptions. Thirdly, this study was conducted in a single teaching hospital. Finally, there are reimbursement restrictions for DOAC prescriptions in subjects covered by the public medication insurance plan in the province of Quebec, which limits the generalization of the results.
Our study supports previous findings, but it differs from others by the inclusion of subjects on rivaroxaban and apixaban, and collection of laboratory data (i.e., serum creatinine, AST and ALT) to take into account renal and liver impairments, which are associated with higher risks of bleeding in users of OAC. To the best of our knowledge, this is the first study finding that prescription by a neurologist increased the likelihood while concomitant use of clopidogrel reduced the probability of DOAC prescription over warfarin. Results from this study add to the currently limited research on DOAC predictors of use among OAC naive patients. It allowed us to identify subjects who are more likely to be prescribed a DOAC, and thus target interventions to promote the proper use of these drugs.