The Effect of Dabigatran Plasma Concentrations and Patient Characteristics on the Frequency of Ischemic Stroke and Major Bleeding in Atrial Fibrillation Patients

The RE-LY Trial (Randomized Evaluation of Long-term Anticoagulation Therapy)

Paul A. Reilly, PHD; Thorsten Lehr, PHD; Sebastian Haertter, PHD; Stuart J. Connolly, MD; Salim Yusuf, MD, DPHIL; John W. Eikelboom, MB BS; Michael D. Ezekowitz, MD, PHD; Gerhard Nehmiz, PHD; Susan Wang, PHD; Lars Wallentin, MD, PHD


J Am Coll Cardiol. 2014;63(4):321-328. 

In This Article


The design and results of the RE-LY study have been previously published.[1–3] Briefly, the primary objective of RE-LY was to establish the noninferiority of 2 doses of DE compared with warfarin for stroke prevention in patients with AF and 1 additional risk factor for stroke. This trial randomized 18,113 AF patients to 1 of 2 blinded doses of DE, DE 110 or DE 150, or to dose-adjusted warfarin titrated to an international normalized ratio of 2 to 3. Median follow-up was 2.0 years. The study, including the PK sampling, was approved by all appropriate national regulatory authorities and ethics committees. All patients provided written informed consent before study entry.

All patients with a valid blood sample and all ischemic stroke/systemic embolic events (SEE) or bleeding events that occurred on-treatment in these patients were included in the analysis, regardless of when the event occurred in relation to the sampling. Patients who were off-treatment at the time of sampling or the time of event were not included in the analysis. All primary and secondary outcome events were blindly and doubly adjudicated. The primary RE-LY study outcome was stroke or systemic embolism. Stroke type was subdivided into ischemic, hemorrhagic, and unknown. Definitions of RE-LY endpoints are described elsewhere.[1,3] CHADS2 score is a simple validated risk score that assigns 1 point for a history of congestive heart failure, hypertension, age 75 years or older, and diabetes mellitus and 2 points for a history of stroke or transient ischemic attack. CHA2DS2-VASc[8] and HAS-BLED[9] are other methods to assess risk for stroke or bleeding in AF patients.

PK and Statistical Methods

Peak and trough samples at steady state were collected for determination of drug concentration, activated partial thromboplastin time, and ecarin clotting time at 1-month post-randomization in all DE subjects who gave consent to participate, regardless of the time of any ischemic or bleeding event that occurred. Samples were collected from patients in all geographic regions. The frequencies generally approximated the number of recruited subjects. For trough concentrations, only samples collected within 10 to 16 h after the previous DE dose were considered. Similarly, for post-dose samples, only samples collected within 1 to 3 h after dosing were considered. Approximately 12% of samples were excluded from evaluation because of questionable records in blood sampling date/time or in administration date/time. Additional samples were taken at 3, 6, and 12 months from 2,143 subjects who participated in a PK substudy. For analyses reported here, the data were merged with the data from the substudy and analyzed together. The first trough and post-dose samples fulfilling the time-window rule were used from these subjects with multiple blood samples. The population PK[4] and pharmacodynamic data (activated partial thromboplastin time and ecarin clotting time) will be reported elsewhere.

Plasma concentrations of nonconjugated (free) dabigatran and of total dabigatran after alkaline cleavage of conjugates were determined by a validated high-performance liquid chromatography tandem mass spectrometry method at AAIPharma Deutschland GmbH & Co. KG, Neu-Ulm, Germany.[10] Total dabigatran concentration was determined in all PK plasma samples, whereas nonconjugated dabigatran was determined in a subset (n = 1,085) of samples in order to assess the prevalence of dabigatran acylglucuronides in this population.

Logistic regression of events (ischemic stroke/SEE and major and minor bleeds) and associated log-transformed trough plasma concentrations was performed with and without covariates. If patients had more than 1 observation, the sample closest to the planned time after dose (2 h and 12 h) was used, regardless of the occurrence of an event. The best-fit model was defined based on the Akaike Information Criterion. The c-statistic for predictive value based on the receiver-operating characteristic curve was also calculated. The impact of the covariates age, sex, body mass index, creatinine clearance (CrCl), coronary artery disease (CAD), diabetes mellitus, prior stroke or transient ischemic attack, hypertension, heart failure, CHADS2 score, concomitant aspirin (ASA) use, and concomitant clopidogrel use were investigated. Only covariates with p < 0.2 were retained in the model if the Akaike Information Criterion was improved. CrCl at baseline was calculated using the Cockroft-Gault equation.

The performance of the final logistic regression models for major bleeding and ischemic stroke/SEE was tested by predicting the event rates in the DE patients without plasma concentration measurements (n = 3,584). For these patients, the validation dataset, the trough plasma concentrations, was predicted using the population PK model from RE-LY,[4] and the individual stroke and bleed rates were calculated based on the final logistic regression models. Patients were sorted according to their calculated risk and divided into ascending deciles. For each decile, the mean predicted risks for an event and for no event were calculated and compared with observed rates. The predictive performance was evaluated by the Hosmer-Lemeshow test (Online Table 1, Online Fig. 1). Values of p > 0.05 indicate no significant difference between the predicted and observed event rates.

Figure 1.

Probability of Clinical Outcomes Versus Dabigatran Plasma Concentrations
Major bleeding event (left) and ischemic stroke/systemic embolic event (right) versus trough dabigatran plasma concentration in atrial fibrillation patients by age (65, 75, and 85 years). Covariates include sex, prior stroke, and diabetes. Conc. = concentration.

Post-dose concentrations provided no advantages compared with trough concentrations, so only the latter are reported in the logistic regression analyses. Cox regression spline analyses were performed to explore the covariate–concentration–response relationship without the restrictions of a covariate linear effect. Event rates for bleeding and thromboembolic events are for the entire duration of the trial (median duration 2.0 years). All statistical analyses were performed with SAS version 9.1 and higher (SAS Institute, Cary, North Carolina).