Drug-drug interactions in atrial fibrillation patients receiving direct oral anticoagulants

Polypharmacy is common in patients with atrial fibrillation (AF), making these patients vulnerable to the occurrence of potential drug-drug interactions (DDIs). We assessed the risk of ischemic stroke and major bleeding in the context of concomitant treatment with potential DDIs in patients with AF prescribed direct oral anticoagulants (DOACs). Using the common data model (CDM) based on an electronic health record (EHR) database, we included new users of DOACs from among patients treated for AF between January 2014 and December 2017 (n = 1938). The median age was 72 years, and 61.8% of the patients were males, with 28.2% of the patients having a CHA2DS2-VASc score in category 0–1, 49.4% in category 2–3, and 22.4% in category ≥ 4. The CHA2DS2-VASc score was significantly associated with ischemic stroke occurrence and hospitalization for major bleeding. Multiple logistic regression analysis showed that increased risk of ischemic stroke and hospitalization for major bleeding was associated with the number of DDIs regardless of comorbidities: ≥ 2 DDIs was associated with ischemic stroke (OR = 18.68; 95% CI, 6.22–55.27, P < 0.001) and hospitalization for major bleeding (OR = 5.01; 95% CI, 1.11–16.62, P < 0.001). DDIs can cause reduced antithrombotic efficacy or increased risk of bleeding in AF patients prescribed DOACs.


Methods
Data sources. Electronic health records (EHRs) data from Seoul National University Bundang Hospital were transformed into OMOP-CDM version 5.2. Diagnoses were coded according to the 6th Korean Classification of Disease modified classification systems from the International Classification of Disease-10 (ICD-10). Drug names were mapped to the Anatomical Therapeutic Chemical Classification System. This study was performed in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Seoul National University Bundang Hospital. The need for informed consent from each patient was waived because the authors did not have access to identifiable information (IRB No: X-19040/535-901). Study population. We included patients with diagnostic codes for AF from January 2013 to December 2017 and who received DOAC (rivaroxaban, apixaban, dabigatran, or edoxaban) treatment for 7 days or longer from initial diagnosis (n = 3681). We used several exclusion criteria to maximize data accuracy. First, we excluded AF patients previously prescribed DOAC between January 2013 and December 2013 to analyze only new cases (n = 1188). Second, patients < 20 years of age diagnosed with AF and patients with valvular AF were excluded (n = 2). Third, patients with an alternative indication for DOAC treatment and prophylaxis, including deep vein thrombosis, pulmonary embolism, and joint replacement surgery, were excluded (n = 25). Fourth, patients with end-stage renal disease were excluded (n = 0). Lastly, patients with ischemic stroke (IS), intracranial hemorrhage (ICH), or gastrointestinal (GI) bleeding in the 6 months prior to cohort entry were excluded in the analysis of primary prevention (n = 528). Finally, a total of 1938 patients (diagnosed between January 2014 and December 2017) were selected for this study. The detailed patient enrollment flow is described in Supplementary Figure 1. DDIs. Forty-five concurrent medications that may have a potential DDI with DOACs were selected as follows: (1) drugs affecting platelet function such as antiplatelet agents, non-steroidal anti-inflammatory drugs (NSAIDs), and serotonergic agents such as selective serotonin reuptake inhibitors (SSRIs), and serotonin and norepinephrine reuptake inhibitors (SNRIs); (2) P-glycoprotein inhibitors or CYP3A4 inhibitors such as amiodarone, clarithromycin, cobicistat, fluconazole, itraconazole, and voriconazole; and (3) P-glycoprotein inducers or CYP3A4 inducers such as carbamazepine, phenobartibal, phenytoin, and rifampin [22][23][24][25] (Supplementary  Table 1).

Study outcomes.
We identified four clinical outcomes as follows: IS, ICH, hospitalization for GI bleeding, and hospitalization for major bleeding. To assess the outcomes, we followed up the patients for 1 year. Detailed definitions of the clinical outcomes are described in Supplementary Table 2.
DOACs administered within 30 days before the clinical outcomes were examined to assess clinical outcomes associated with DOACs. If DOACs were not administered within 30 days before the events, the event was not counted in our analysis. To identify potential drug interactions causing an increased risk of DOAC-related bleeding or a reduced antithrombotic efficacy, we examined the use of DDI drugs 30 days prior to the events.

Comorbidities.
Comorbidities were included in the model as CHA 2 DS 2 -VASc scores by assigning 1 point for age between 65 and 74 years, female sex, hypertension, diabetes, congestive heart failure, or vascular disease, and adding 2 points for age 75 years or older, history of stroke, transient ischemic attack, or systemic thromboembolism 26 . CHA 2 DS 2 -VASc scores were divided into three categories: 0-1, 2-3, and ≥ 4. Statistical methods. Patient characteristics were summarized using descriptive statistics. The median and interquartile range (IQR) was reported for continuous variables, and categorical variables were expressed as frequencies (percentage). Multiple logistic regression analysis was performed using a forced entry method to examine the associations of the CHA 2 DS 2 -VASc score and DDIs with the risk for poor clinical outcomes. For each independent variable, the odds ratio (OR) and 95% confidence interval (CI) were determined. All tests were 2-tailed, with P < 0.05 considered significant. All statistical analysis was performed using R Statistical Software version 3.6.3 (Vienna, Austria; http:// www.R-proje ct. org/).

Results
Baseline characteristics. Between 2014 and 2017, a total of 1,938 patients with AF who were newly administered DOACs were included in the study. The clinical baseline characteristics of the study are shown in Table 1. The median age was 72 years, and 61.8% of the patients were males. Among the DOACs, rivaroxaban was the most used (29.4%), followed by apixaban (22.3%) and dabigatran (15.2%). The proportion of subjects in each CHA 2 DS 2 -VASc score category was as follows: 28.2% in category 0-1, 49.4% in category 2-3, and 22.4% in category ≥ 4.
Ischemic stroke. IS events associated with DOACs occurred in 29 patients (1.5%) during the observation period ( Table 2). Although not statistically significant, the median age was higher in the IS group than in the group without IS (76 vs. 71 years, P = 0.062). The CHA 2 DS 2 -VASc score was significantly associated with IS occurrence; as the score increased, so did the risk of IS. In addition, IS events were more common in patients who simultaneously took drugs with potential DDIs and DOACs, with risk increasing alongside the number of DDIs: 1 (OR = 6.22; 95% CI, 2.65-15.67, P < 0.0001) and ≥ 2 (OR = 12.22; 95% CI, 4.21-34.72, P < 0.001). The use of P-glycoprotein inducers or CYP3A4 inducers was not observed in patients with IS (Supplementary Table 3). Most of the potential drugs used in patients with IS were identified as pharmacodynamic drugs: antiplatelet agents such as aspirin (n = 15) and clopoidogrel (n = 5) (Supplementary Table 3 Table 3). The median age was significantly higher in the hospitalization for major bleeding group than in the group without hospitalization for major bleeding (81 vs. 71 years, P < 0.001). When analyzed against the CHA 2 DS 2 -VASc score, events of hospitalization for major bleeding were significantly more likely patients with 4 points or more compared to patients with 0-1 points (OR = 11.51; 95% CI, 2.15-212.75, P = 0.021). There was no association between DOAC type and hospitalization for major bleeding events. Hospitalization for major bleeding events had a statistically significant relationship with potential DDIs (OR = 2.81; 95% CI, 1.20-6.60, P = 0.016), and the risk tended to increase as the number of DDIs increased. Most of the potential drugs used in patients with hospitalization for major bleeding were antiplatelet agents (n = 4) or NSAIDs (n = 3) (Supplementary Table 4). The relationship between ICH and hospitalization for GI bleeding and DDI is described in Supplementary Tables 5 and 6, respectively. GI bleeding showed a statistically significant relationship with DDI, whereas ICH did not.

Multiple logistic regression analysis for clinical outcomes. Multiple logistic regression analysis
showed that increased risk of IS and hospitalization for major bleeding was associated with the number of DDIs regardless of comorbidities: ≥ 2 DDIs was associated with IS (OR = 18.68; 95% CI, 6.22-55.27, P < 0.001) and hospitalization for major bleeding (OR = 5.01; 95% CI, 1.11-16.62, P < 0.001) ( Table 4).

Discussion
Management of oral anticoagulant drug interaction is essential to ensure safe and effective use. Warfarin has over 200 identified drug interactions that must be considered before use 27 . Wang et al. conducted meta-analysis based on low-to moderate-strength evidence supporting interaction between warfarin and a small group of medications leading to bleeding risk or thromboembolic outcomes 28 . Although DOACs have comparable efficacy and enhanced safety compared to warfarin, the appropriateness and accuracy of prescribing medications are important to prevent increased risk of bleeding or reduced antithrombotic efficacy. In the current study, we found that potential DDIs were associated with a substantially high risk for both ischemic stroke and hospitalization for major bleeding regardless of comorbidities. Drug interactions have been previously associated with decreased potency of DOACs 25 . Based on pharmacokinetic data and published case reports, there is a significant decrease in DOAC drug concentration and an increased risk of adverse thrombotic events in patients receiving concomitant P-glycoprotein inducers or CYP3A4 inducers [29][30][31][32] . DDIs were associated with a significantly higher risk for IS, in particular for DDIs with ≥ 2  27). Most of the DDIs related to IS were pharmacodynamic drugs such as antiplatelet agents or NSAIDs and were not related to P-glycoprotein inducers or CYP3A4 inducers. The ARISTOTLE trial reported that participants on aspirin were at higher risk for ischemic events, with higher CHADS 2 scores, than were participants not receiving aspirin 33 . In the current study, concomitant use of antiplatelet agents was observed in patients with IS, which could explain the higher CHA 2 DS 2 -VASc scores of those on antiplatelet agents. Previous studies demonstrated a thrombotic risk associated with NSAIDs 34-36 . Kent et al. demonstrated that the use of NSAIDs was associated with an increased risk for ischemic stroke as well as major bleeding 37 . It is difficult to assess the clinical relevance of DOAC drug interactions because the available data are frequently limited to pharmacokinetic studies in a small number of healthy volunteers or retrospective case-control or cohort studies. Romoli et al. showed that switching between DOACs is frequent, occurring in up to 11% of patients prescribed with DOAC for AF. In this study, one in 4 patients had to switch between DOACs within 30 days before ischemic stroke or bleeding events. Due to the anonymization of CDM data, the cause of switching between DOACs was not known, but other studies reported that minor bleeding and non-CV adverse events had been reported as one of the most common causes to justify switching between DOACs 38,39 . IS was more common in those who changed DOACs compared to those who took rivaroxaban alone, with an OR of 2.28. Considering that continuous adherence to DOACs is essential to maintain stroke risk reduction 40,41 , the inability to take continuous doses could result in IS events.
In the current study, among those hospitalized for major bleeding, GI bleeding showed a statistically significant relationship with DDIs, but ICH did not. Our group reported that the rate of exposure to DDIs in bleeding events was about 57%, and NSAIDs and antiplatelet agents were the most common drugs for DDIs 17 . Although the data is not presented in this study, NSAIDs and antiplatelet agents caused potential DDIs as their use alongside DOACs was concomitant when GI bleeding occurred. The combination of anticoagulation medication and NSAIDs or aspirin increased the risk of GI bleeding 42,43 . The mechanism for the increased GI bleeding associated with NSAIDs relies on the effect of NSAIDs on platelet aggregation and gastric mucosa 44,45 . In addition, considering the potential renal effect of NSAIDs, it is possible that the exposure increased in the concomitant with DOACs, which are mainly excreted by the kidney 46 .
Our study has several limitations. First, due to the nature of the CDM data, we could not review the medical records of any individual patient such as lifestyle risk factors like cigarette smoking and alcohol consumption or change of DOACs due to serious side effects. Second, according to the recent domestic claim data, less than 10% of AF patients using DOAC had CHA 2 DS 2 -VASc scores of 0-1 17,47 , but the current CDM study showed that proportion to be close to 28%. The reason for the higher proportion of patients with CHA2DS2-VASc scores of 0-1 compared to other studies is that the diagnosis of comorbidities such as hypertension and diabetes are likely to be missed. To correct this error, even if there was no diagnosis, CHA2DS2-VASc scores reflected whether a patient was taking medication for hypertension or diabetes. Third, the CDM database did not have data on adherence to prescribed medications and information on other hospital medications. In Korea, there is a Korean Health Insurance Review and Assessment Service database to check the history of drug administration while visiting a hospital for the last 90 days. In the future, the limitations of the CDM database can be overcome by reflecting this information. Fourth, as with all non-experimental studies, associations can be drawn, but causality can only be inferred. The number of DDIs can either decrease or enhance the effect of the involved drugs, increasing the risk of side-effects, as well as indicate the presence of multimorbidity. Further studies are required to better understand DDIs leading to adverse events in patients with polypharmacy. Lastly, the potential DDIs of this study may contain inaccuracies or contradict other interaction databases. In light of these limitations, it is essential to regularly reassess the literature for optimal clinical decision-making.
In conclusion, physicians prescribing DOACs for AF should be aware that DDIs can cause significantly increased risk for both IS and hospitalization for major bleeding regardless of comorbidities.