Stroke prevention in atrial fibrillation changes after dabigatran availability in China: The GLORIA‐AF registry

Abstract Background Until the approval of dabigatran etexilate, treatment choices for stroke prevention in patients with atrial fibrillation (AF) were vitamin K antagonists (VKAs) or antiplatelet drugs. This analysis explored whether availability of non‐vitamin K antagonist oral anticoagulants post‐dabigatran approval was associated with changing treatment patterns in China. Methods Global Registry on Long‐Term Oral Antithrombotic Treatment in Patients with Atrial Fibrillation (GLORIA‐AF) collected data on antithrombotic therapy choices for patients with newly diagnosed nonvalvular AF at risk for stroke. In China, enrollment in phase 1 (before dabigatran approval) and phase 2 (after dabigatran approval) occurred from 2011 to 2013 and 2013 to 2014, respectively. Analyses were restricted to sites within China that contributed patients to both phases. The weighted average of the site‐specific results was estimated for standardization. Sensitivity analyses used multiple regression. Results Thirteen sites participated in both phase 1 (419 patients) and phase 2 (276 patients), 76.1% and 16.0% were known to be at high risk for stroke (CHA2DS2‐VASc ≥2) and bleeding (HAS‐BLED ≥3); 55.5% were male. In phase 1, 16.7%, 61.6%, and 21.7% of patients were prescribed oral anticoagulants (OACs), antiplatelet agents, and no treatment, respectively. Respective proportions were 26.4%, 40.6%, and 33.0% in phase 2. The absolute increase in the site‐standardized proportion of patients prescribed OACs after dabigatran availability was 9.9% (95% confidence interval [CI]: 3.7%‐16.0%). There was a standardized 17.3% (95% CI: −24.3% to −10.4%) absolute decrease in antiplatelet agent use. Conclusions There was an increase in OAC and decrease in antiplatelet agent prescription since dabigatran availability in China. However, a large proportion of AF patients at risk for stroke remained untreated.


| INTRODUC TI ON
Oral anticoagulant (OAC) therapy prevents ischemic stroke and reduces mortality in atrial fibrillation (AF) patients. 1 treated with a vitamin K antagonist (VKA) and 61% with antiplatelet agents without VKAs, while 25% received no antithrombotic therapy. 5 Suggested explanations are a perceived higher baseline risk of anticoagulant-induced intracerebral hemorrhage among Chinese patients, 6,7 a belief there is a low risk of ischemic stroke among Chinese AF patients 8,9 and the potential for herb-drug interactions due to the high prevalence of herbal consumption. 10 Warfarin interacts with many drugs and food ingredients, which may pose significant challenges in administration and monitoring. 9 In addition, it has been stated that many patients in China lack access to good laboratory control for VKA therapy and therefore cannot be prescribed VKAs. 11,12 Non-vitamin K antagonist oral anticoagulants (NOACs) have been suggested to be preferentially indicated over warfarin for Asian patients. 11 14,15 and, following the introduction of NOACs, more educational programs have been conducted. 3 Given the obstacles to VKA use in China, the introduction of dabigatran (along with the contemporaneous expert consensus and educational programs) can be expected to have increased the proportion of patients prescribed OACs for stroke prevention in AF. The objective of this analysis was to explore whether availability of NOACs was associated with changing treatment patterns in China by comparing the proportion of treated patients between pre-and post-dabigatran approval periods.

| Study design, participants, and setting
The design of the Global Registry on Long-Term Oral Antithrombotic Treatment in Patients with Atrial Fibrillation (GLORIA-AF) program has been reported previously. 16

| Variables
Baseline characteristics and antithrombotic treatment strategy were collected using electronic case report forms at the time of the enrollment (baseline) visit. The main outcome for this analysis was prescription of an OAC, excluding patients prescribed multiple OACs, but including patients prescribed an OAC in association with an antiplatelet drug. Prescription of an antiplatelet drug alone was a secondary outcome. Antithrombotic treatment strategy was defined based on antithrombotic treatments prescribed for long-term use at the time of the baseline visit, or any antithrombotic treatments that the patient was already using at the time of the baseline visit and would continue using long term.
Baseline patient characteristics included demographic and clinical data, AF characteristics, medical history, and selected concomitant treatments. Labile international normalized ratio was not collected at baseline; a modified hypertension, abnormal renal function, abnormal liver function, stroke (prior), bleeding (prior), labile international normalized ratio, elderly (age >65 years), prior alcohol or drug usage history, medication usage predisposing to bleeding (antiplatelet agents, nonsteroidal anti-inflammatory drugs) (HAS-BLED) score not considering labile international normalized ratio was used.
Potential confounders in the association between the phase of the study and antithrombotic prescriptions considered were as follows: risk scores for stroke (CHA 2 DS 2 -VASc) and bleeding (modified HAS-BLED), age, sex, body mass index, AF characteristics (categorization as asymptomatic, minimally symptomatic, or symptomatic; and type as permanent, persistent, or paroxysmal AF), history of thromboembolic disease (stroke), cardiovascular history (myocardial infarction, coronary artery disease, congestive heart failure/left ventricular dysfunction, hypertension), history of bleeding events, other medical history (diabetes mellitus, chronic gastrointestinal disease), enrollment site, and month of enrollment.

| Statistical methods
The objective was to assess the changes in antithrombotic treatment patterns after dabigatran availability, which involved comparing patterns during phases 1 and 2 of the GLORIA-AF registry program.
To avoid bias due to different sites participating in the study phases, analyses were restricted to sites from mainland China (the main country enrolling patients in phase 1) enrolling patients in both phases 1 and 2.
As site participation may have differed between phases 1 and 2 of the study, and sites may have differed in their antithrombotic prescription patterns, the main analysis involved within-site comparisons with standardization across sites. 17 Standardization involved taking a weighted average of the site-specific results, weighting by the total number of patients enrolled by each site during both phases.
Patient characteristics were described according to phase. We conducted a sensitivity analysis using log-binomial regression 18 to account for phase, site, and characteristics imbalanced between phases 1 and 2. Log-binomial regression provides probability (in this case, antithrombotic treatment prescription) and probability ratios (in this case, the ratio between phase 2 and phase 1).
Imbalanced characteristics were defined as any variable for which the absolute value of the standardized difference 19 between phase 1 and phase 2 was higher than 10%. The standardized difference compared the difference in means in units of the pooled standard deviation.
For multiple regression, missing values were treated as an absence of the considered risk factor; standardized differences were also assessed after single imputation for missing values.
To avoid confusion between the use of "standardized" for imbalance diagnosis (standardized difference) and of "standardization" as our main analysis approach controlling for site effects, "site-standardized" will be used in the remainder of the article to describe the latter.
The proportion of OAC use during baseline was also described according to the month of baseline visit for both phases 1 and 2 to assess underlying time trends. In addition, a log-binomial model estimating the association between time and OAC use was fitted to the subset of patients enrolled in phase 1, adjusting for site and patient characteristics found to be imbalanced between study phases. A potential phase 2 trend was not explored by multivariable modeling, since the focus was on assessing the extent to which the change in OAC use observed in phase 2 could be explained by a prescription trend before dabigatran availability.
Inclusion of both phase and time as predictors in a multiple regression was not reasonable because of high multicollinearity between time and phase.

| Patient characteristics
There were signs of imbalance (standardized difference >10%) in some patient characteristics between phase 1 (pre-dabigatran approval) and phase 2: category of AF (65% of patients in phase 1 and 44% in phase 2 were symptomatic; standardized difference: +43%), known high-bleeding risk (modified-not considering labile international normalized ratio-HAS-BLED ≥3), body mass index, known prior bleed, congestive heart failure, gastrointestinal disease, and prior myocardial infarction (Table 1).    By definition, no NOAC was available in phase 1 (before availability of dabigatran) and all OAC use corresponded to VKA use. In phase 2, eight patients were prescribed NOACs, constituting 2% of all phase 2 patients, in addition to 24% prescribed VKAs.
Differences between phase 2 and phase 1 of the study may be related to a temporal trend that predates dabigatran availability. In Abbreviations: AF, atrial fibrillation; BMI, body mass index; HAS-BLED, hypertension, abnormal renal function, abnormal liver function, stroke (prior), bleeding (prior), labile international normalized ratio, elderly (age >65 y), prior alcohol or drug usage history, medication usage predisposing to bleeding (antiplatelet agents, nonsteroidal anti-inflammatory drugs); SD, standard deviation. a Patients with missing HAS-BLED (n = 13) were considered at low risk.
b Congestive heart failure/left ventricular dysfunction.

| D ISCUSS I ON
The principal study findings are as follows: (a) OAC use increased in  In line with other registries, this analysis from GLORIA-AF indicates that OAC use increased in China after dabigatran approval.
Nonetheless, although all patients in the GLORIA-AF registry had a CHA 2 DS 2 -VASc score of ≥1, and 76% in this analysis had a CHA 2 DS 2 -VASc score ≥2, many patients were treated with antiplatelet drugs alone (40.6% after dabigatran approval) or were not treated (33.0%). In this analysis, the increase in OAC use after dabigatran approval (+10%) was only partially explained by NOAC prescriptions alone (+2%). It is possible that, in a setting where NOACs were not covered by medical insurance, economic factors may have had a greater influence on what was prescribed.
Nonetheless, NOAC approval may have had an indirect role, triggering guideline updates and increasing the awareness of optimal management of stroke prevention in AF.

| Limitations
Some limitations should be considered. The number of patients available for the analyses was modest, limiting our ability to adjust for confounders and identify time trends. As phase 2 of the study started when dabigatran was approved in each respective country, the enrollment period in phase 2 was relatively short (10 months

| CON CLUS ION
The patterns observed in patients enrolled in GLORIA-AF in mainland China are consistent with an increase in the proportion of patients prescribed OACs since dabigatran became available. In parallel, the rate of antiplatelet prescriptions showed a decrease following dabigatran availability and is in line with published guidelines. Despite the increase in oral anticoagulation use since dabigatran availability, a large proportion of patients remain untreated, or treated with antiplatelet only.

ACK N OWLED G M ENTS
Editorial support was provided by Sarah J Petit of Parexel, supported by Boehringer Ingelheim. The study was funded by Boehringer Ingelheim. The authors thank the patients who participated in this trial, and their families, as well as the investigators, study coordinators, study teams, and nurses.  Note: Intercept and the 12 coefficients associated with the 13 sites are not shown.

CO N FLI C T O F I NTE R E S T
Abbreviations: AF, atrial fibrillation; BMI, body mass index; CI, confidence interval; HAS-BLED, hypertension, abnormal renal function, abnormal liver function, stroke (prior), bleeding (prior), labile international normalized ratio, elderly (age >65 y), prior alcohol or drug usage history, medication usage predisposing to bleeding (antiplatelet agents, nonsteroidal anti-inflammatory drugs). a Patients with missing prior bleed (n = 1) or HAS-BLED (n = 13) were considered at low risk.