Polypharmacy and the Efficacy and Safety of Rivaroxaban Versus Warfarin in the Prevention of Stroke in Patients With Nonvalvular Atrial FibrillationCLINICAL PERSPECTIVE
Background—Patients with atrial fibrillation (AF) often take multiple medications.
Methods and Results—We examined characteristics and compared adjusted outcomes between rivaroxaban and warfarin according to number of concomitant baseline medications and the presence of combined cytochrome P450 3A4 and P-glycoprotein inhibitors in the Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF) study. At baseline, 5101 patients (36%) were on 0 to 4 medications, 7298 (51%) were on 5 to 9, and 1865 (13%) were on ≥10. Although polypharmacy was not associated with higher risk of stroke or non–central nervous system embolism (adjusted hazard ratio, 1.02 for ≥10 versus 0–4 medications; 95% confidence interval, 0.76–1.38), it was associated with higher risks of the combined end point of stroke, non–central nervous system embolism, vascular death, or myocardial infarction (adjusted hazard ratio, 1.41 for ≥10 versus 0–4 medications; 95% confidence interval, 1.18–1.68) and nonmajor clinically relevant or major bleeding (adjusted hazard ratio, 1.47 for ≥10 versus 0–4 medications; 95% confidence interval, 1.31–1.65). There was no significant difference in primary efficacy (adjusted interaction P=0.99) or safety outcomes (adjusted interaction P=0.87) between treatment groups by number of medications. Patients treated with 0 to 4 medications had lower rates of major bleeding with rivaroxaban (adjusted hazard ratio, 0.71; 95% confidence interval, 0.52–0.95; interaction P=0.0074). There was no evidence of differential outcomes in those treated with ≥1 combined cytochrome P450 3A4 and P-glycoprotein inhibitors.
Conclusions—In a population of patients with atrial fibrillation, two thirds were on ≥5 medications. Increasing medication use was associated with higher risk of bleeding but not stroke. Rivaroxaban was tolerated across complex patients on multiple medications.
Multiple medications are common in the treatment of adult patients. This is particularly true in elderly patients with chronic medical conditions such as atrial fibrillation (AF). The use of multiple medications in a single patient, called polypharmacy, is becoming more prevalent in contemporary, guideline-driven practice.1,2 Patients with AF often have comorbidities that are associated not only with an increased risk of stroke but also with the need for multiple cardiovascular and noncardiovascular medications. Although an individual medication may be demonstrated to have a positive impact on outcomes when tested against placebo, multiple medications and eventual polypharmacy entail risk. These risks include drug-drug interactions, an increased risk of adverse events, decreased adherence, lower quality of life, and delirium.3–5
Clinical Perspective on p 360
Vitamin K antagonists, although efficacious, have multiple drug-drug interactions, often require frequent dose adjustments, and have a narrow therapeutic window. Drug-drug interactions in warfarin-treated patients on multiple medications are common and are associated with increased bleeding risks.6 Rivaroxaban is a once-daily, oral, direct factor Xa inhibitor that has been shown to be noninferior to warfarin for the prevention of stroke in nonvalvular AF with a lower risk of intracranial hemorrhage. Although rivaroxaban has fewer drug-drug interactions, it is affected by strong cytochrome P450 3A4 (CYP3A4) inducers and inhibitors and by medications with P-glycoprotein (P-gp) inhibition or induction. We aimed to examine the prevalence of polypharmacy and the impact of concomitant medications on ischemic and hemorrhagic events. Furthermore, to better understand the comparative benefits of warfarin and rivaroxaban in patients on multiple medications, we investigated the efficacy and safety of rivaroxaban versus warfarin according to concomitant medications and the presence of combined CYP3A4 and P-gp inhibitors.
The rationale and design of the Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF) have previously been described.7 In brief, ROCKET AF was an international, multicenter, double-blind, double-dummy, randomized, noninferiority trial that compared rivaroxaban 20 mg once daily (or 15 mg daily in patients with creatinine clearance 30–49 mL/min) with dose-adjusted warfarin (target international normalized ratio, 2.0–3.0) in patients with nonvalvular AF at moderate to high risk of stroke. Moderate to high stroke risk was defined by CHADS2 (clinical heart failure, hypertension, age ≥75 years, diabetes mellitus [1 point each], and previous stroke or transient ischemic attack [2 points]) risk score ≥2. Enrollment of patients with only 2 risk factors was capped at 10% for each geographic region. Key exclusion criteria included use of a strong CYP3A4 inhibitor or inducer, prosthetic heart valves, hemodynamically significant mitral stenosis, and creatinine clearance <30 mL/min. Ninety-three patients from 1 site were excluded from efficacy analyses because of violations of Good Clinical Practice guidelines. The study was approved by ethical/institutional review boards at all participating sites, and all patients provided written informed consent.
Definitions and End Points
Patients were classified according to the number of baseline concomitant medications (in addition to randomized study medication) that were individually recorded on the case report form. Although there is no consensus definition of polypharmacy, for the purpose of this analysis, we considered 3 clinically relevant categories of concomitant medications and polypharmacy (0–4, 5–9, and ≥10).8 We also conducted an additional analysis using the number of concomitant medications as a continuous variable. The primary efficacy end point in the ROCKET AF trial was the occurrence of stroke (ischemic or hemorrhagic) or non–central nervous system (CNS) embolism. Secondary efficacy outcomes included all-cause mortality; myocardial infarction (MI); the composite of stroke, non-CNS embolism, or vascular death; and the composite of stroke, non-CNS embolism, MI, or vascular death. The intention-to-treat population was used for all efficacy analyses. The primary safety end point was the occurrence of International Society on Thrombosis and Haemostasis major or nonmajor clinically relevant (NMCR) bleeding.9 Secondary safety end points included major bleeding, fatal bleeding, intracranial hemorrhage, and NMCR bleeding. Safety analyses were conducted among randomized patients who received at least 1 dose of study drug. All study end points were adjudicated by an independent clinical events committee blinded to treatment assignment. Beyond the adjudicated end points, we assessed study drug discontinuation (warfarin versus rivaroxaban) as an additional safety end point.
Baseline categorical data and early study drug discontinuation rates are provided as counts (percentages), and differences were tested among polypharmacy groups with the Pearson χ2 test. Continuous variables are summarized as medians with 25th and 75th percentiles, and the differences were analyzed with the Wilcoxon rank sum test. Outcomes are presented as events per 100 patient-years, and cumulative event rates were illustrated graphically with plots indicating the number of patients at risk for each event. Cox proportional hazards models were used to assess the associations of the number of concomitant medications with risk of outcomes. As described above, the number of concomitant medications was analyzed according to clinically relevant groups (0–4, 5–9, and ≥10 concomitant medications). We also assessed the interaction between randomized treatment and the number of concomitant medications for each efficacy and safety outcome to determine whether treatment risk relationships seen in the full ROCKET AF cohort are consistent across levels of concomitant medication use.
As a result of concerns about the impact of combined CYP3A4 and P-gp inhibitors on the pharmacokinetics of rivaroxaban, additional analyses were performed to compare treatment outcomes in those patients receiving ≥1 combined inhibitors with those subjects who did not receive any combined inhibitor. Combined inhibitors included amiodarone, diltiazem, verapamil, quinidine, ranolazine, felodipine, erythromycin, or azithromycin. Dronedarone was not taken by any subjects at study entry.
All Cox models included randomized treatment and covariates found to be predictive of outcomes in the full ROCKET AF cohort. Models for efficacy end points included age, sex, body mass index, region, diabetes mellitus, previous stroke/transient ischemic attack, vascular disease (MI, peripheral artery disease, carotid occlusive disease), congestive heart failure, hypertension, chronic obstructive pulmonary disease, paroxysmal AF, diastolic blood pressure, creatinine clearance (calculated with the Cockcroft-Gault equation), heart rate, and alcohol use. Models for safety end points included age, sex, region, previous stroke/transient ischemic attack, anemia, previous gastrointestinal bleed, chronic obstructive pulmonary disease, diastolic blood pressure, creatinine clearance, platelets, albumin, and previous aspirin, vitamin K antagonist, or thienopyridine use. Rates of missing data were generally quite low; missing values of covariates were imputed with the use of the median for continuous variables and the most common value for categorical variables. Risk relationships are presented as hazard ratios (HRs) with 95% confidence intervals (CIs).
The study was supported by Johnson & Johnson Pharmaceutical Research & Development (Raritan, NJ) and Bayer HealthCare AG (Leverkusen, Germany). The Duke Clinical Research Institute (Durham, NC) coordinated the trial, managed the database, and performed the primary analyses independently of the sponsors. All analyses were performed with the SAS version 9.2 statistical software (SAS Institute, Cary, NC).
Among 14 264 subjects with nonvalvular AF, 98% (n=13 965) were taking at least 1 concomitant medication at baseline (Figure 1). The median number of concomitant medications at baseline was 5 (25th and 75th percentiles, 4 and 8). At the last follow-up, 94% (n=13 461) were taking at least 1 concomitant medication (median, 6; 25th and 75th percentiles, 4 and 8). In the entire cohort, 36% were taking 0 to 4 concomitant medications at baseline, 51% were taking 5 to 9 medications, and 13% were taking ≥10 medications. As shown in Table 1, a higher number of concomitant medications was much more prevalent in North America. Patients taking more medications were older, were more often diabetic, more often had a diagnosis of chronic obstructive pulmonary disease, and were more likely to have taken previous vitamin K antagonists, but they were less likely to have had a previous stroke or transient ischemic attack (Table 1). The number of concomitant medications was also associated with increasing use of ≥1 combined P-gp and weak or moderate CYP3A4 inhibitors. Of note, the type of AF and the presence of heart failure did not correlate with the number of concomitant medications.
The occurrence of the primary end point (stroke or non-CNS embolism) was not significantly different according to the number of concomitant medications (Figure 2 and Table 2). In contrast, major and NMCR bleeding (Figure 3) and all-cause mortality increased with the number of concomitant medications. Results of multivariable Cox proportional hazards modeling for outcomes are shown in Table 3. After adjustment for covariates, patients receiving ≥10 medications experienced increased risks of the combined end point of stroke, non-CNS embolism, vascular death, or MI compared with those receiving 0 to 4 medications (HR, 1.41; 95% CI, 1.18–1.68; P=0.0009). The risk of MI was higher in those receiving ≥10 compared with those receiving 0 to 4 medications (HR, 2.28; 95% CI, 1.53–3.41; P=0.0002). Similarly, patients receiving ≥10 medications experienced increased risks of all-cause mortality compared with those receiving 0 to 4 medications (HR, 1.44; 95% CI, 1.18–1.74; P=0.0005).
There was evidence of increasing early study drug discontinuation with more concomitant medication. The discontinuation rates were 31.8%, 35.4%, and 42.4% in those taking 0 to 4, 5 to 9, and ≥10 medications, respectively (P<0.0001 for difference across groups). Early study drug discontinuation was similar in each treatment arm across concomitant medication groups (Table I in the online-only Data Supplement).
The frequency of the primary safety end point (major or NMCR bleeding) increased according to the number of concomitant medications: 11.64 versus 14.79 versus 23.42 events per 100 patient-years in those receiving 0 to 4, 5 to 9, and ≥10 medications, respectively. As shown in Table 2, similar relationships were observed for major bleeding, intracranial hemorrhage, and NMCR bleeding. In adjusted analyses, the risk of major or NMCR bleeding was higher in those taking ≥10 medications compared with those taking 0 to 4 medications (HR, 1.47; 95% CI, 1.31–1.65; P<0.0001 across groups; Table 3).
When we conducted a sensitivity analysis omitting topical medications, ophthalmological medications, and herbal supplements, the results for both the efficacy and safety end points were consistent with the overall analysis (Table II in the online-only Data Supplement). We also examined the number of concomitant medications as a continuous variable. As shown in Table III in the online-only Data Supplement, the relationships were similar to the results of the main categorical analysis. It is notable that each additional concomitant medication was associated with a 3% increase in the risk of all-cause mortality.
Outcomes According to Treatment Assignment
In general, the efficacy end points were numerically higher in the warfarin arm compared with the rivaroxaban arm across all concomitant medication groups (Figure 4 and Table IV in the online-only Data Supplement). The occurrence of major and NMCR bleeding was similar between the rivaroxaban and warfarin arms, whereas intracranial hemorrhage was less frequent in the rivaroxaban arm across all 3 groups of concomitant medications. Interaction tests failed to identify any heterogeneity in the associations between outcomes and treatment assignments except for major bleeding (P=0.0074). The risk of major bleeding was reduced with rivaroxaban compared with warfarin among patients taking 0 to 4 medications (HR, 0.71; 95% CI, 0.52–0.95); this reduced risk was not evident in patients taking more medications (HR, 1.23; 95% CI, 1.01–1.49 in those taking 5 to 9 medications; HR, 1.17; 95% CI, 0.87–1.56 in those taking ≥10 medications).
Impact of Combined CYP3A4 and P-gp Inhibitors
Because of concerns about the impact of combined CYP3A4 and P-gp inhibitors on the pharmacokinetics of rivaroxaban, additional analyses were performed to compare treatment outcomes in those patients receiving ≥1 combined inhibitors with those subjects who did not receive any combined inhibitor. As shown in Figure 5 and Table V in the online-only Data Supplement, there was no evidence of an interaction between treatment assignment (rivaroxaban versus warfarin) and combined inhibitor status for any of the outcomes, including all-cause mortality, intracranial hemorrhage, major or NMCR bleeding, and stroke or non-CNS embolism.
We also conducted a supplemental analysis in which we analyzed event rates in patients treated with ≥2 combined inhibitors (Table VI in the online-only Data Supplement). Mortality and the composite end point of stroke, non-CNS embolism, vascular death, or MI were numerically lower in the patients treated with rivaroxaban. However, major bleeding and NMCR bleeding were numerically lower in those treated with warfarin.
In this analysis of polypharmacy and outcomes in more than 14 000 patients with nonvalvular AF, there are 4 main findings. First, two thirds of patients were taking ≥5 concomitant medications in addition to their oral anticoagulant. Second, after adjustment for prognostically important covariates, the frequencies of bleeding events and mortality (but not thromboembolism) were associated with the number of concomitant medications. Third, there was no evidence of significant heterogeneity in outcomes according to the number of concomitant medications in patients treated with rivaroxaban versus warfarin, except for International Society on Thrombosis and Haemostasis major bleeding. In these subjects, rivaroxaban was associated with a lower risk of major bleeding in patients taking 0 to 4 medications. Finally, and perhaps most important, despite concerns for increased risks in patients taking combined CYP3A4 and P-gp inhibitors, there was no evidence of increased bleeding or all-cause mortality in those treated with ≥1 combined inhibitors with rivaroxaban compared with warfarin. Although the sample size was limited, use of ≥2 combined inhibitors was associated with higher bleeding rates in those treated with rivaroxaban compared with warfarin, however, the cardiovascular composite end point and all-cause mortality were lower with rivaroxaban.
Clinicians prescribe medications to patients largely on the basis of efficacy data from placebo-controlled, randomized, controlled trials. However, patients receiving these medications are often on a large number of concomitant medications that could affect the risk-to-benefit profile of the prescribed medication, especially in patients taking large numbers of concomitant medications. Polypharmacy is becoming increasingly common in modern clinical practice and is associated with an increased risk of adverse events, decreased adherence, a lower quality of life, and delirium.3–5 These risks are not trivial; the Centers for Disease Control and Prevention estimate that drug-related deaths now exceed fatalities from motor vehicle accidents.10 National data suggest that 7% of all emergency department visits that result in hospitalization are attributable to adverse drug events.11 Despite the importance of medication safety, there are few investigations into the risks of concomitant medications in specific disorders. Although the lifetime risk of AF is ≈25%,12 the risks of AF are particularly pronounced in the elderly and those with comorbid cardiovascular illness. Such patients are more likely to be taking concomitant medications. In this analysis, we found that 13% of patients with nonvalvular AF were taking ≥10 medications, and 64% were taking ≥5 medications. This rate of polypharmacy in this moderate- to high-risk AF cohort is increased 2- to 5-fold compared with that observed in primary care populations (2%–5%).13,14
Although several analyses have examined patient safety and polypharmacy, fewer studies have analyzed the impact of polypharmacy on specific treatment effects or end points. We found that an increasing number of concomitant medications was independently associated with bleeding events and all-cause mortality after adjustment. Surprisingly, the number of concomitant medications was not associated with an increased risk of stroke or non-CNS embolism. This finding may be attributable to low power, particularly because the composite outcome (stroke or non-CNS embolism, vascular death, or MI) was associated with increasing number of concomitant medications. An earlier investigation in Scotland found that polypharmacy (defined as ≥11 medications) was more common in those with a diagnosis of stroke (12.6% versus 1.5%).15 The failure to identify an association between polypharmacy and stroke in ROCKET AF could also be attributable to the selection criteria used in the ROCKET AF trial. As a result of the inclusion criteria, patients with previous stroke (and greatest risk of future stroke) tended to be younger and were on fewer medications.
There has been significant interest in the use of bleeding scores to help identify patients at risk for bleeding while on oral anticoagulant therapy. It is interesting to note that the number of medications appears to mirror bleeding risk. In our cohort, patients on ≥10 medications experienced a 2-fold increase in major and NMCR bleeding. The presence of polypharmacy should prompt and remind clinicians to minimize bleeding risk, including the discontinuation of concomitant antiplatelet therapy if and when appropriate.16
In terms of the relative treatment effects in patients treated with rivaroxaban and warfarin according to the number of concomitant medications, we saw little evidence of heterogeneity. The lone exception was in the occurrence of major bleeding. Patients on 0 to 4 concomitant medications had a lower risk of major bleeding with rivaroxaban. Given the small sample sizes in these subgroups and the multiple tests performed, it is possible that this finding represents a false-positive association (type I error).
Strong CYP3A4 inhibitors (eg, ketoconazole, clarithromycin, and ritonavir) and inducers (eg, phenytoin, carbamazepine, and rifampin) were excluded from ROCKET AF. However, combined mild to moderate CYP3A4 inhibitors and P-gp inhibitors were permitted in the trial. Furthermore, it is important to note that many medications used to treat AF, including amiodarone, verapamil, and diltiazem (among others), are combined CYP3A4 and P-gp inhibitors.
At present, the package insert for rivaroxaban cautions that the use of rivaroxaban with combined P-gp and moderate CYP3A4 inhibitors should be avoided because of the potential for increases in drug absorption, drug concentrations, and increased bleeding risks. Despite this warning, we saw no evidence of differential outcomes between rivaroxaban and warfarin according to combined inhibitor status across all 3 concomitant medication groups. This should be reassuring for clinicians who care for patients on rivaroxaban and combined inhibitors. In a smaller supplemental subgroup analysis, there was evidence of more bleeding events in patients on ≥2 combined inhibitors with rivaroxaban compared with warfarin. However, major adverse cardiac events and mortality were lower in the rivaroxaban-treated patients. These data raise the hypothesis that treatment of patients with rivaroxaban and multiple combined inhibitors (eg, amiodarone with verapamil) may lead to increased bleeding but lower cardiovascular events compared with treatment with warfarin. Again, given the limitations of the sample size, larger studies in more heterogeneous populations are needed.
There are several limitations that should be kept in mind in the consideration of these data. This study was a retrospective, exploratory analysis. There is the potential for selection bias because patients with previous stroke did not need to have additional risk factors for stroke to be included in ROCKET AF. Despite extensive adjustment, residual confounding may have influenced the observed association. Although we adjusted for a significant number of comorbidities, we cannot exclude the possibility that the observed associations were attributable to differences among patients who received more or fewer medications in ways that are not accounted for in our models. Given the large number of statistical tests, in particular the interaction tests, we cannot exclude the possibility of type I error. Finally, the data are limited to baseline prescription burden. However, patients with polypharmacy that is driven by chronic medical conditions do not often have a dramatic reduction in the number of medications they are taking.
This is the first study examining concomitant medications and polypharmacy in patients with AF. In a population of patients with nonvalvular AF at moderate to high risk of stroke, polypharmacy was not infrequent. Increasing medication use was associated with higher risk of bleeding but not stroke risk. There was no evidence of differential outcomes in those treated with rivaroxaban compared with warfarin according to the number of concomitant medications, except for a lower risk of major bleeding in those treated with rivaroxaban and <5 medications. There was no evidence of heterogeneity of treatment outcomes according to the use of combined CYP3A4 and P-gp inhibitors unless patients were taking ≥2 combined inhibitors. In summary, rivaroxaban was tolerated across complex patients on multiple medications.
Sources of Funding
ROCKET AF was supported by Johnson & Johnson Pharmaceutical Research & Development (Raritan, NJ) and Bayer HealthCare AG (Leverkusen, Germany).
Dr Piccini receives grants for clinical research from ARCA biopharma, Boston Scientific, Gilead, Janssen Pharmaceuticals, ResMed, and St. Jude Medical and serves as a consultant to GlaxoSmithKline, Johnson & Johnson, Laguna Pharmaceuticals, Medtronic, and Spectranetics. Dr Becker serves as a consultant/advisory board member for Janssen Research & Development, Portola, Cook, and Boehringer Ingelheim. Dr Breithardt has served as a consultant to Bayer HealthCare, Johnson & Johnson, Boehringer Ingelheim, Sanofi-Aventis, MSD, and 3M. Dr Berkowitz is an employee of Bayer HealthCare Pharmaceuticals. Dr Halperin has received consulting fees from Bayer AG HealthCare, Boehringer Ingelheim, Daiichi Sankyo, Johnson & Johnson, Ortho-McNeil-Janssen Pharmaceuticals, Pfizer, and Sanofi-Aventis. Dr Hankey has received consulting fees from Bayer and Sanofi. Dr Mahaffey’s full disclosures before August 1, 2013, are available at www.dcri.org and after August 1, 2013, at https://med.stanford.edu/profiles/47970?tab=research-and-scholarship. Dr Nessel is an employee of Janssen Research and Development. Dr Singer has received research funding from Johnson & Johnson, Bristol-Myers Squibb, Boehringer Ingelheim, and Medtronic, as well as consulting fees from Boehringer Ingelheim, Bristol-Myers Squibb, CVS Health, Johnson & Johnson, Merck, Pfizer, and St. Jude Medical. Dr Fox has received research funding from Bayer, Janssen, and AstraZeneca; honoraria from Bayer, AstraZeneca, GlaxoSmithKline, Janssen, and Sanofi; and consulting fees from Bayer, Lilly, AstraZeneca, and Sanofi. Dr Patel has received research funding from Johnson & Johnson and AstraZeneca and served on advisory boards for Bayer, Janssen, AstraZeneca, and Genzyme. The other authors report no conflicts.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA.115.018544/-/DC1.
- Received July 16, 2015.
- Accepted December 11, 2015.
- © 2015 American Heart Association, Inc.
- Schulman S,
- Kearon C
- 10.↵Centers for Disease Control and Prevention. Quickstats: death rates for three selected causes of injury—National Vital Statistics System, United States, 1979–2012. November 21, 2014. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6346a19.htm. Accessed July 10, 2015.
- Lloyd-Jones DM,
- Wang TJ,
- Leip EP,
- Larson MG,
- Levy D,
- Vasan RS,
- D’Agostino RB,
- Massaro JM,
- Beiser A,
- Wolf PA,
- Benjamin EJ.
- Gallacher KI,
- Batty GD,
- McLean G,
- Mercer SW,
- Guthrie B,
- May CR,
- Langhorne P,
- Mair FS.
- Steinberg BA,
- Kim S,
- Piccini JP,
- Fonarow GC,
- Lopes RD,
- Thomas L,
- Ezekowitz MD,
- Ansell J,
- Kowey P,
- Singer DE,
- Gersh B,
- Mahaffey KW,
- Hylek E,
- Go AS,
- Chang P,
- Peterson ED
Patients with atrial fibrillation often take multiple medications. In the Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF) study population, we examined patient characteristics and compared adjusted outcomes between rivaroxaban and warfarin according to number of concomitant baseline medications and the presence of combined cytochrome P450 3A4 and P-glycoprotein inhibitors. Among ROCKET AF patients (nonvalvular atrial fibrillation with moderate to high risk of stroke), polypharmacy was not infrequent. Two thirds of the patients were on ≥5 medications. Increasing medication use was associated with higher risk of bleeding but not stroke risk. There was no evidence of differential outcomes in those treated with rivaroxaban versus warfarin according to the number of concomitant medications, except for lower risk of major bleeding in those treated with rivaroxaban and <5 medications. There was no evidence of heterogeneity of treatment outcomes according to the use of combined cytochrome P450 3A4 and P-glycoprotein inhibitors unless patients were taking ≥2 combined inhibitors. In summary, rivaroxaban was tolerated across complex patients on multiple medications.