Author + information
- Received February 21, 2018
- Revision received April 3, 2018
- Accepted April 17, 2018
- Published online July 9, 2018.
- Anthony P. Kent, MDa,
- Martina Brueckmann, MDb,c,
- Mandy Fraessdorf, PhDb,
- Stuart J. Connolly, MDd,
- Salim Yusuf, MD, DPhild,
- John W. Eikelboom, MDd,
- Jonas Oldgren, MD, PhDe,
- Paul A. Reilly, PhDf,
- Lars Wallentin, MD, PhDe and
- Michael D. Ezekowitz, MD, PhDg,∗ (, )@TJUHospital@JeffersonUniv@LIMR
- aDepartment of Internal Medicine, Bridgeport Hospital, Yale New Haven Health, Bridgeport, Connecticut
- bBoehringer Ingelheim GmbH & Co KG, Ingelheim, Germany
- cUniversity of Heidelberg, Mannheim, Germany
- dMcMaster University and Hamilton Health Sciences, Population Health Research Institute, Hamilton, Ontario, Canada
- eUppsala Clinical Research Center and Department of Medical Sciences, Uppsala University, Uppsala, Sweden
- fClinical Development, Cardiology, Boehringer Ingelheim Pharmaceuticals, Ridgefield, Connecticut
- gSidney Kimmel Medical College at Thomas Jefferson University and Lankenau and Bryn Mawr Hospitals, Philadelphia, Pennsylvania
- ↵∗Address for correspondence:
Dr. Michael D. Ezekowitz, Department of Cardiology, Bryn Mawr Hospital and Lankenau Medical Center, 830 Lancaster Road, Suite 301, Bryn Mawr, Pennsylvania 19010.
Background Nonsteroidal anti-inflammatory drugs (NSAIDs) are commonly used medications that can potentially increase the risk of bleeding and thrombosis.
Objectives This study quantified the effect of NSAIDs in the RE-LY (Randomized Evaluation of Long Term Anticoagulant Therapy) trial.
Methods This was a post hoc analysis of NSAIDs in the RE-LY study, which compared dabigatran etexilate (DE) 150 and 110 mg twice daily (b.i.d.) with warfarin in patients with atrial fibrillation. Treatment-independent, multivariate-adjusted Cox regression analysis assessed clinical outcomes by comparing NSAID use with no NSAID use. Interaction analysis was obtained from treatment-dependent Cox regression modeling. Time-varying covariate analysis for NSAID use was applied to the Cox model.
Results Among 18,113 patients in the RE-LY study, 2,279 patients used NSAIDs at least once during the trial. Major bleeding was significantly elevated with NSAID use (hazard ratio [HR]: 1.68; 95% confidence interval [CI]: 1.40 to 2.02; p < 0.0001). NSAID use did not significantly alter the risk of major bleeding for DE 150 or 110 mg b.i.d. relative to warfarin (pinteraction = 0.63 and 0.93, respectively). Gastrointestinal major bleeding was significantly elevated with NSAID use (HR: 1.81; 95% CI: 1.35 to 2.43; p < 0.0001). The rate of stroke or systemic embolism (stroke/SE) with NSAID use was significantly elevated (HR: 1.50; 95% CI: 1.12 to 2.01; p = 0.007). The use of NSAIDs did not significantly alter the relative efficacy on stroke/SE for DE 150 or 110 mg b.i.d. relative to warfarin (pinteraction = 0.59 and 0.54, respectively). Myocardial infarction rates were similar with NSAID use compared with no NSAID use (HR: 1.22; 95% CI: 0.77 to 1.93; p = 0.40). Patients were more frequently hospitalized if they used an NSAID (HR: 1.64; 95% CI: 1.51 to 1.77; p < 0.0001).
Conclusions The use of NSAIDs was associated with increased risk of major bleeding, stroke/SE, and hospitalization. The safety and efficacy of DE 150 and 110 mg b.i.d. relative to warfarin were not altered. (Randomized Evaluation of Long Term Anticoagulant Therapy [RE-LY]; NCT00262600)
The class of medications known as nonsteroidal anti-inflammatory drugs (NSAIDs) consists of a variety of agents with a common pathway of inhibiting the cyclooxygenase (COX) enzyme and reducing the synthesis of prostaglandins, thromboxane, and prostacyclin. The COX enzyme is subdivided into 2 isoenzymes: COX-1 and COX-2 (1). Examples of nonselective NSAIDs include ibuprofen, naproxen, meloxicam, diclofenac, and ketorolac, whereas celecoxib is an example of a selective NSAID (selective COX-2 inhibitor) (2–4). Aspirin irreversibly inhibits COX, and preferentially, COX-1 at low doses, as well as both isoenzymes at higher doses. Clinically, NSAIDs are associated with an increased risk of bleeding and thrombosis (5–8), adverse effects on renal function (9), and worsening heart failure (10).
NSAIDs are commonly used and easily accessible medications for patients, and are frequently self-used in the community in conflict with labeling information (11). For example, the use of NSAIDs for analgesia is prevalent among patients initiating dabigatran (12). Both the transient or prolonged combination of NSAIDs with direct oral anticoagulation (OAC) poses a serious risk of bleeding (13). Transient or prolonged use of NSAIDs occurs for a variety of reasons, including arthritis (osteoarthritis and rheumatoid arthritis). Importantly, the burden of arthritis is increasing, particularly among patients with heart disease (14), making the overlap use of NSAIDs and OAC more probable.
In the context of atrial fibrillation, concomitant administration of NSAIDs with vitamin K antagonists poses an elevated risk of thromboembolism and bleeding (7,15). Direct OACs, including the direct thrombin inhibitor dabigatran and factor X inhibitors rivaroxaban, apixaban, and edoxaban, are understudied with regard to NSAID use, with current reports based on venous thromboembolism prevention and treatment (16–23). The effect of NSAIDs on the safety and efficacy of dabigatran etexilate (DE), specifically in the context of patients with atrial fibrillation in the randomized, controlled RE-LY (Randomized Evaluation of Long Term Anticoagulant Therapy) trial, has not been previously reported.
The objective of this post hoc analysis of the RE-LY trial was to assess the effects of the combined use of NSAIDs with OAC therapy, specifically dabigatran or warfarin, in the context of patients with atrial fibrillation to quantify bleeding and thrombotic risks, respectively.
The RE-LY trial was a prospective evaluation of DE (150 or 110 mg twice daily [b.i.d.]) in comparison to dose-adjusted warfarin (mean time in therapeutic range: 64%) among 18,113 patients with atrial fibrillation. The rationale and design of the RE-LY trial, as well as its primary outcomes were previously described (16,24).
This post hoc analysis of the RE-LY trial compared the group of patients who used nonselective NSAIDs at least once during the RE-LY trial (n = 2,279) with patients who never used NSAIDs during the trial (n = 15,834). Concomitant NSAID use was recorded at study randomization and during routine patient follow-up. The baseline characteristics and outcomes of patients were described according to NSAID use versus no NSAID use, and treatment-independent (OAC therapy) or with respect to treatment arm. The analysis of NSAID use excluded aspirin and selective COX-2 inhibitors. The concomitant use of antiplatelet therapy with DE in patients with atrial fibrillation was previously described (25). Clinical outcomes from the RE-LY trial were adjudicated for verification of the outcome event. The definition of major bleeding was published previously (24).
Cox regression models were used to assess time to relevant outcome events, by providing hazard ratios (HRs), 95% confidence intervals (CIs), and p values for treatment-independent outcomes (Table 1). In the Cox model, patients were censored after their first event. Multivariate adjustment analysis was applied to all clinical outcomes data to correct for confounding variables. HRs were based on multivariate-adjusted analysis, which adjusted for 17 total covariates as described in Table 1. Time-varying covariate analysis was applied to the Cox model to account for the variability in NSAID use over time, its duration of use, and its use before an outcome event (Table 2). Interaction analysis was performed to assess for any interaction of NSAIDs with dabigatran relative to warfarin with respect to clinical outcomes (Table 3). Interaction p values were generated based on the relative differences in the HRs for dabigatran compared with warfarin in the context of NSAID use versus no use according to clinical outcome.
Proportion of NSAID use
Of the 18,113 total patients randomized in the RE-LY trial to either dabigatran or warfarin, 12.6% of patients (n = 2,279) used NSAIDs at least once during the study period. There were a similar proportion of patients who used NSAIDs per OAC treatment group (dabigatran 110 mg b.i.d., 13.0%; dabigatran 150 mg b.i.d., 11.9%; and warfarin, 12.8%). Most patients never used an NSAID during the study period (87.4%; n = 15,834).
The baseline characteristics of patients at randomization to treatment group (dabigatran 110 mg b.i.d., dabigatran 150 mg b.i.d., or warfarin) in the RE-LY study are described according to NSAID use versus no NSAID use (Table 4).
At randomization into RE-LY, the mean CHA2DS2VASc score was the same for patients who took NSAIDs (3.6) compared with patients who did not take NSAIDs (3.6). Renal function was similar between the NSAID use and never used groups. There was a smaller proportion of patients with a history of myocardial infarction (13.1% vs. 17.1%; p < 0.0001) and a smaller proportion with a history of heart failure (30.1% vs. 32.3%; p = 0.0349) among those who took NSAIDs compared with patients who did not take NSAIDs. The proportion of patients with valvular heart disease was greater among patients who used NSAIDs (24.3%) compared with patients who did not use NSAIDs (21.4%) (p < 0.0001).
Aspirin, clopidogrel, or dipyridamole were used in slightly greater proportion among patients who took NSAIDs at least once during the RE-LY (44.9%) trial than among patients who never took NSAIDs (41.2%) (p = 0.0007). The NSAID use group had a moderately higher proportion of patients who took proton-pump inhibitors compared with patients who did not take NSAIDs (16.5% vs. 13.8%; p = 0.0008). Multivariate-adjusted analysis was performed for all outcomes to adjust for differences in baseline characteristics.
The annualized rate of major bleeding when NSAIDs were used at least once in combination with OAC therapy was significantly elevated compared with patients who did not use NSAIDs (Cox model: HR: 1.39; p < 0.0001; time-varying covariate analysis: HR: 1.68; p < 0.0001) (Tables 1 and 2, Figure 1). The rates of major bleeding for each OAC treatment arm, dabigatran 110 mg b.i.d., dabigatran 150 mg b.i.d., and warfarin, were elevated with NSAID use (Table 5, Figure 2). Major bleeding rates with dabigatran 110 and 150 mg b.i.d. were not altered relative to warfarin with NSAID use (pinteraction = 0.93 and 0.63, respectively) (Table 3).
Gastrointestinal (GI) major bleeding occurred significantly more often in patients who took NSAIDs at least once with OAC therapy than in patients who did not take NSAIDs (Cox model: HR: 1.48; p = 0.0018; time-varying covariate analysis: HR: 1.81; p < 0.0001) (Tables 1 and 2, Figure 3). GI bleeding also occurred significantly more frequently in the NSAID group despite a higher proportion of patients on a proton-pump inhibitor in the unadjusted analysis. GI major bleeding was elevated across all OAC treatment arms with NSAID use (Table 5, Figure 4). NSAID use did not alter the rate of GI major bleeding with dabigatran 110 and 150 mg b.i.d. relative to warfarin (Table 3).
In the basic Cox model, without the time-varying covariate analysis, there was a trend toward significance for stroke or systemic embolism (stroke/SE) with NSAID use (HR: 1.27; p = 0.0553), and ischemic stroke and hemorrhagic stroke did not show a significant difference with NSAID use (Table 1). In the time-varying covariate analysis of the Cox model, stroke/SE was significantly elevated among patients taking NSAIDs in combination with OAC therapy (2.0%/year vs. 1.4%/year; HR: 1.50; 95% CI: 1.12 to 2.01; p = 0.007) (Table 2, Figure 5). Notably, ischemic stroke was significantly elevated with NSAID use (HR: 1.55; p = 0.0102). Hemorrhagic stroke rates were similar with NSAID use or nonuse (HR: 1.08; p = 0.8631). Rates of stroke/SE were elevated with dabigatran 110 mg b.i.d., dabigatran 150 mg b.i.d., or warfarin combined with NSAID use (Table 5, Figure 6). In the interaction analysis, the efficacy of dabigatran 110 and 150 mg b.i.d. in preventing stroke/SE relative to warfarin was not significantly altered by the use of NSAIDs (Table 3).
Myocardial infarction rates were low and occurred at a similar rate (0.70%/year) among patients who took an NSAID at least once during the study period in combination with OAC therapy compared with patients who did not take an NSAID (0.67%/year), which was consistent with the time-varying covariate analysis (Tables 1 and 2). With respect to the OAC treatment arm (dabigatran 110 or 150 mg b.i.d., or warfarin), the rates of myocardial infarction were similar with or without NSAID use (Table 5). The use of NSAIDs did not interact with dabigatran 110 or 150 mg b.i.d. to alter the rates of myocardial infarction relative to warfarin (Table 3).
Patients were more frequently hospitalized if they used an NSAID with OAC therapy compared with patients who did not take an NSAID (p < 0.0001) (Tables 1 and 2, Figure 7). Hospitalizations were more frequently associated with NSAID use in combination with each treatment arm, dabigatran 110 mg b.i.d., dabigatran 150 mg b.i.d., and warfarin, compared with patients who did not take an NSAID with the respective anticoagulant (Table 5, Figure 8).
All-cause mortality among patients who used NSAIDs with dabigatran or warfarin was not significantly different compared with patients who did not use NSAIDs (Tables 1 and 2). Mortality with NSAID use and warfarin (3.3%/year) was notably lower compared with warfarin use without NSAID use (4.3%/year) (Table 5).
The present analysis was the first to evaluate NSAID use with DE in the context of atrial fibrillation. The rates of major bleeding, GI major bleeding, stroke/SE, ischemic stroke, and hospitalization when NSAID was used in combination with OAC therapy (dabigatran 110 or 150 mg b.i.d., or warfarin) were significantly elevated compared with patients who did not use NSAID. The rates of myocardial infarction with NSAID use in combination with OAC therapy were similar compared with not taking NSAIDs (Central Illustration). The use of NSAIDs did not significantly alter the relative efficacy or safety for DE 110 mg b.i.d. or DE 150 mg b.i.d., respectively, relative to warfarin.
The mechanism for the observed bleeding associated with NSAIDs was likely a result of its antiplatelet effects and a reduction in gastric mucosal protection. Non-GI bleeding represented 60% and GI bleeding represented 40% of major bleeding events among the NSAID group. NSAIDs are known to reduce the glomerular filtration rate (9), and DE is excreted approximately 80% via glomerular filtration. Despite the potential renal effects of NSAIDs, the interaction analysis did not show that NSAIDs had any direct interaction with DE to alter its bleeding profile relative to warfarin.
Previous studies demonstrated a thrombotic risk associated with NSAIDs (5–8). The present analysis demonstrated a significantly elevated stroke/SE rate, specifically, ischemic stroke, with NSAID use. The time-varying covariate analysis of the Cox model accounted for the duration of NSAID use and its use before an outcome event compared with the basic Cox model. The present analysis did not include celecoxib or other selective COX-2 inhibitors due to the small sample size (∼160 per treatment arm), and the cardiovascular risk associated with COX-2 inhibitors does not necessarily make them a safe alternative to nonselective NSAIDs (26,27).
Hospitalization was common (∼20%/year) among a patient population with atrial fibrillation on OAC without NSAID use, and was significantly elevated with NSAID use (∼29%/year). In addition to atrial fibrillation, there was a prominent comorbid burden among patients, which was highlighted by the mean CHA2DS2-VASc score of 3.6 for both the NSAID and no NSAID use groups.
Seeking safer alternatives to NSAIDs for patients with atrial fibrillation who receive OAC therapy is recommended from this study. Atrial fibrillation is an epidemic with an estimated burden of 2.7 to 6.1 million people in the United States (28), and >30 million people worldwide (29). Notably, >2 million people in the United States are addicted to prescription opioids (30). Safer alternatives beyond NSAIDs and opioids are available (31). Updated guidelines and better therapeutic interventions (nonpharmacological and pharmacological) are needed to address analgesia in patients with heart disease because current approaches with NSAIDs and opioids pose a risk to patients (32).
The present analysis was consistent with a previous study of NSAID use in combination with warfarin in patients with atrial fibrillation that demonstrated an elevated rate of bleeding and stroke with NSAID use (7,15). In previous studies, the use of NSAIDs with dabigatran 150 mg b.i.d. in the context of venous thromboembolism did not significantly effect the rate of bleeding (20,21), which was in contrast to the present analysis, and potentially due to the different populations studied.
The present analysis adds to the body of literature that have investigated the safety of NSAIDs and reinforces the evidence that NSAIDs are not innocuous medications (26,33–39). Thus, there is a need for careful clinical review of NSAID use in patients with atrial fibrillation who are receiving OAC therapy. Further analysis is required to investigate the effects of NSAIDs when used concomitantly with direct OACs in patients with atrial fibrillation.
The study was a post hoc analysis. It was limited by the lack of data regarding the specific type of nonselective NSAIDs, the dosage of NSAIDs, and the reasons for NSAIDs use, which were not captured in the standardized case report form used during the RE-LY trial follow-up visits. Patients were allowed to stop and start NSAIDs periodically. The basic Cox model evaluated NSAID use versus nonuse, irrespective of the duration and timing of NSAID use. Baseline characteristic data did not capture the prevalence of osteoarthritis, rheumatoid arthritis or other inflammatory conditions that might have partially explained NSAID use and might have added to the investigators’ understanding of the patient population that used NSAIDs during RE-LY. Treatment-dependent outcomes for NSAID use versus no use were not analyzed with the Cox model due to small sample sizes per OAC treatment arm among patients using NSAIDs.
The use of NSAIDs in combination with OAC therapy (DE 110 or 150 mg b.i.d., or warfarin) in patients with atrial fibrillation was associated with an increased risk of major bleeding, GI major bleeding, stroke/SE, ischemic stroke, and hospitalization, compared with patients who did not use NSAIDs. No significant difference in myocardial infarction or mortality was observed with NSAID use. The safety and efficacy of DE 110 and 150 mg b.i.d. relative to warfarin was not significantly altered by the use of NSAIDs.
COMPETENCY IN MEDICAL KNOWLEDGE: In anticoagulated patients with atrial fibrillation, concurrent use of NSAIDs was associated with a higher risk of major bleeding, thromboembolism, and hospitalization, but the safety and efficacy of dabigatran 150 and 110 mg b.i.d. relative to warfarin were not altered.
TRANSLATIONAL OUTLOOK: The effects on clinical outcomes of anti-inflammatory drugs, and selective COX-2 inhibitors in particular, should be evaluated in patients managed with target-specific OACs, including apixaban, rivaroxaban, and edoxaban.
The RE-LY clinical trial was funded by Boehringer Ingelheim GmbH & Co KG, Ingelheim, Germany. The post hoc analysis did not have a funding source. Dr. Connolly has received grants and personal fees from Boehringer Ingelheim, Bristol-Myers Squibb, Sanofi, and Bayer; has received personal fees from Portola; has received grants from Boston Scientific; and has received an institutional research grant from Boehringer Ingelheim. Dr. Yusuf has received grants, honoraria, and travel reimbursements for speaking engagements from Boehringer Ingelheim, Bayer, and Bristol-Myers Squibb; and has received a research grant from Boehringer Ingelheim. Dr. Eikelboom has received consulting fees and/or honoraria and grant and/or in-kind support from AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, Eli Lilly, GlaxoSmithKline, Pfizer, Janssen, and Sanofi. Dr. Oldgren has been a consultant (including steering and data monitoring committees) and has received lecture fees from Bayer, Bristol-Myers Squibb/Pfizer, Boehringer Ingelheim, Daichii-Sankyo, and Sanofi; and has been a member of the advisory board for Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Pfizer, and Sanofi. Dr. Wallentin has received institutional research grants from AstraZeneca, Bristol-Myers Squibb/Pfizer, Merck & Co., Roche, GlaxoSmithKline, and Boehringer Ingelheim; and has received consulting fees from Abbott. Dr. Ezekowitz has received consulting fees from Boehringer Ingelheim, Pfizer, Bristol-Myers Squibb, Portola, Daiichi-Sankyo, and Armetheon; and has received grant support from Boehringer Ingelheim and Pfizer. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- twice daily
- confidence interval
- dabigatran etexilate
- hazard ratio
- nonsteroidal anti-inflammatory drug
- oral anticoagulation
- systemic embolism
- Received February 21, 2018.
- Revision received April 3, 2018.
- Accepted April 17, 2018.
- Rouzer C.A.,
- Marnett L.J.
- Solomon D.H.,
- Husni M.E.,
- Libby P.A.,
- et al.
- Chang C.H.,
- Shau W.Y.,
- Kuo C.W.,
- Chen S.T.,
- Lai M.S.
- Arfè A.,
- Scotti L.,
- Varas-Lorenzo C.,
- et al.
- Ilomäki J.,
- Helin-Salmivaara A.,
- Huupponen R.,
- Rikala M.,
- Kirkpatrick C.M.,
- Korhonen M.J.
- Davidson B.L.,
- Verheijen S.,
- Lensing A.W.,
- et al.
- U.S. Department of Health and Human Services
- Lip G.Y.
- Friedman R.J.,
- Kurth A.,
- Clemens A.,
- Noack H.,
- Eriksson B.I.,
- Caprini J.A.
- ↵Schulman S, Eriksson H, Goldhaber SZ, et al. Influence of concomitant NSAID or ASA on the efficacy and safety of dabigatran versus warfarin for the treatment of acute venous thromboembolism: a pooled analysis from RE-Cover and RE-Cover II. Poster presented at: 55th Annual Meeting of the American Society of Hematology. New Orleans, LA; December 7−10, 2013.
- Dans A.L.,
- Connolly S.J.,
- Wallentin L.,
- et al.
- Gislason G.H.,
- Torp-Pedersen C.
- Brophy J.M.,
- Lévesque L.E.,
- Zhang B.
- Mozaffarian D.,
- Benjamin E.J.,
- Go A.S.,
- et al.
- Chugh S.S.,
- Havmoeller R.,
- Narayanan K.,
- et al.
- ↵Prescription drug use and misuse in the United States: results from the 2015 National Survey on Drug Use and Health. NSDUH Data Review. Available at: https://www.samhsa.gov/data/sites/default/files/NSDUH-FFR2-2015/NSDUH-FFR2-2015.htm#topofpage. Accessed June 5, 2018.
- Murthy V.H.
- Antman E.M.,
- Bennett J.S.,
- Daugherty A.,
- Furberg C.,
- Roberts H.,
- Taubert K.A.
- Gislason G.H.,
- Jacobsen S.,
- Rasmussen J.N.,
- et al.
- Schjerning Olsen A.M.,
- Fosbøl E.L.,
- Lindhardsen J.,
- et al.
- Felson D.T.
- Grosser T.,
- Ricciotti E.,
- FitzGerald G.A.
- Coxib and Traditional NSAID Trialists' (CNT) Collaboration,
- Bhala N.,
- Emberson J.,
- Merhi A.,
- et al.
- Bally M.,
- Dendukuri N.,
- Rich B.,
- et al.