AM 095

Cardiovascular outcomes, bleeding risk, and achieved blood pressure in patients on long-term anticoagulation with the thrombin antagonist dabigatran or warfarin: data from the RE-LY trial

Michael Bohm, Martina Brueckmann, John W. Eikelboom, Michael Ezekowitz, Mandy Fraßdorf, Ziad Hijazi, Stefan H. Hohnloser, Felix Mahfoud, Roland E. Schmieder, Helmut Schumacher, Lars Wallentin, and Salim Yusuf
1 Klinik fu¨r Innere Medizin III, Saarland University Medical Center, Kirrberger Str. 1, 66421 Homburg, Germany;
2 Medicine CardioMetabolism & Respiratory, Boehringer Ingelheim International GmbH, Binger Strasse 173, 55216 Ingelheim am Rhein, Germany;
3 Faculty of Medicine Mannheim, Heidelberg University, Grabengasse 1, 69117 Heidelberg, Germany;
4 Population Health Research Institute, McMaster University and Hamilton Health Sciences, Hamilton Health Sciences – King West, PO Box 2000, Hamilton, Ontario L8N 3Z5, Canada;
5 Sidney Kimmel Medical College at Jefferson University, 1025 Walnut Street, Philadelphia, PA 19107, USA;
6 Lankenau Medical Center, 100 East Lancaster Avenue, Wynnewood, PA 19096, USA;
7 CardioMetabolism & Respiratory Medicine, Boehringer Ingelheim Pharma GmbH & Co. KG, Binger Straße 173, 55216 Ingelheim am Rhein, Germany;
8 Department of Medical Sciences, UCR Uppsala Clinical Research Center, Uppsala Science Park, Hubben, Dag Hammarskjo¨ lds va¨g 38, SE-751 85 Uppsala, Sweden;
9 Department of Cardiology, UCR Uppsala Clinical Research Center, Uppsala Science Park, Hubben, Dag Hammarskjo¨ lds va¨g 38, SE-751 85 Uppsala, Sweden;
10 Department of Cardiology, Goethe-Universita¨t Frankfurt am Main, 60629 Frankfurt/Main, Germany;
11 Department of Nephrology and Hypertension, University Hospital Erlangen, Friedrich-Alexander University, Schloßplatz 4, 91054 Erlangen, Germany; and
12 Statistical Consultant, 55216 Ingelheim, Germany

Abstract
Aims
A J-shaped association of cardiovascular events to achieved systolic (SBP) and diastolic (DBP) blood pressure was shown in high-risk patients. This association on oral anticoagulation is unknown. This analysis from RELY assessed the risks of death, stroke or systemic emboli, and bleeding according to mean achieved SBP and DBP in atrial fibril-lation on oral anticoagulation.
Methods and results

RE-LY patients were followed for 2 years and recruited between 22 December 2005 until 15 December 2007. 18.113 patients were randomized in 951 centres in 54 countries and 18,107 patients with complete blood pressure (BP) data were analysed with a median follow-up of 2.0 years and a complete follow-up in 99.9%. The association between achieved mean SBP and DBP on all-cause death, stroke and systemic embolic events (SSE), major, and any bleeding were explored. On treatment, SBP >140 mmHg and <120 mmHg was associated with all-cause death compared with SBP 120–130 mmHg (reference). For SSE, risk was unchanged at SBP <110 mmHg but increased at 140–160 mmHg (adjusted hazard ratio (HR) 1.81; 1.40–2.33) and SBP >_160 mmHg (HR 3.35; 2.09–5.36). Major bleeding events were also increased at <110 mmHg and at 110 to <120 mmHg. Interestingly, there was no increased risk of major bleeding at SBP >130 mmHg. Similar patterns were observed for DBP with an increased risk at <70 mmHg (HR 1.55; 1.35–1.78) and >90 mmHg (HR 1.88; 1.43–2.46) for all-cause death compared to 70 to <80 mmHg (reference). Risk for any bleeding was increased at low DBP <70 mmHg (HR 1.46; 1.37–1.56) at DBP 80 to <90 mmHg (HR 1.13; 1.06–1.31) without increased risk at higher achieved DBP. Dabigatran 150 mg twice daily showed an advantage in all patients for all-cause death and SSE and there was an advantage for 110 mg dabigatran twice daily for major bleeding and any bleeding irrespective of SBP or DBP achieved. Similar resultswere obtained for baseline BP, time-updated BP, and BP as time-varying covariate. Conclusion Low achieved SBP associates with increased risk of death, SSE, and bleeding in patients with atrial fibrillation on oral anticoagulation. Major bleeding events did not occur at higher BP. Low BP might identify high-risk patients not only for death but also for high bleeding risks. Introduction Recent guidelines of the European Society of Cardiology/ European Society of Hypertension1 and the American Heart Association2 recommend systolic blood pressure (SBP) targets of 120–129 mmHg and diastolic blood pressure (DBP) targets of 70–79 mmHg, if treatment is well tolerated, for almost all (<120 mmHg) and DBP (<70 mmHg) have been associated with increased risk, thus providing a J-shape curve in the blood pres- sure (BP)-risk association in patients with established cardiovascu- lar disease or at high cardiovascular risk.3,4 Hypertension predisposes to atrial fibrillation5,6 even at mildly elevated BP,7 and atrial fibrillation increases stroke risk.8 It is known that anticoagu- lation with Vitamin K-antagonists (VKA) reduces stroke risk9 but thrombin antagonists like dabigatran10 or factor Xa-antagonists like apixaban,11 rivaroxaban,12 or edoxaban13 provide at least similar or even better stroke protection and lower bleeding risk compared to VKA. However, it is unknown whether in patients on anticoagulation the J-shape BP-risk association is altered. Furthermore, the BP-risk association for bleeding is not known, al- though the ESC guidelines provide a Class IIa-B recommendation for target BP <140–90 mmHg before starting oral anticoagula- tion.1 The RE-LY trial (Randomized Evaluation of Long-term anti- coagulation therapY) database provides a unique opportunity to directly assess cardiovascular events and bleeding rates associated with achieved SBP or DBP. RE-LY randomized patients with atrial fibrillation at increased risk of stroke to two fixed blinded doses of dabigatran or open-label warfarin.10,14 We report the associations of mean achieved, in-trial SBP and DBP with the key outcomes all- cause death, stroke or systemic embolism (SSE), and any or major bleeding events in patients on long-term anticoagulation treat- ment. Mean in-trial BP might give the most precise integration of BP load over time. However, physicians might want the informa- tion what the risk might be before starting treatment or the BP predicting risk just before an event. Finally, change in BP over time might bias results (‘immortal time bias’). Therefore, this analysis was extended to risk associations with baseline BP, time-updated BP (i.e. BP before an event or last BP available) and time-varying BP as time-varying covariate. The key objective was to study the risk association over a broad SBP and DBP spectrum with death, SSE and bleeding in high-risk cardiovascular patients on oral anticoagulation. Patients were recruited from 951 centres in 44 countries. After the screening of 20 377 patients, 18 113 patients were randomized. In six patients no baseline or post-baseline BP was available, resulting in a final number of 18 107 patients for the analysis of mean achieved BP (for all- cause death). The study flow and treatment allocation are given in Figure 1. All primary and secondary outcomes (with the exception of any bleeding events) were assessed by a blinded central committee according to standard criteria. An average of 8.3 (SD 2.6) BP measurements was available measured during an average of 21.0 months (SD 9.6). Methods Study design and population In this analysis, we report outcome data in patients with documented non-valvular atrial fibrillation and an age of at least 75 years or 65– 74 years plus diabetes mellitus, hypertension or coronary artery disease, or at any age with a stroke risk factor, such as heart failure or prior stroke or TIA, according to mean achieved on treatment SBP and DBP ranges. Mean achieved BP were the mean of baseline and following post-baseline BP values until an event occurred. The achieved BP was the observed BP on given treatments but not a targeted BP during the trial, since BP adjustment was managed by the investigators with the advice to treat to guideline-recommended targets. Post-event BP values were not used to exclude an effect of primary outcome and bleeding events on BP. Further categorization of BP was done by using baseline BP values, i.e. BP values prior to any study mediation and by using time-updated BP values, i.e. last BP measurement until event of interest (excl. measurement at day of event) or last BP measurement available. The design, treatment algorithm, and the results of the RE-LY study have been published previously.10,14 In brief, patients with documented atrial fibrillation on electrocardiography performed at or within 6 months before screening with the criteria of or either previous stroke or transient ischaemic attack, left ventricular ejection fraction of _75 years or 65–74 years plus diabetes, hypertension, and coronary artery disease were assigned to either 110 mg or to 150 mg dabigatran twice daily in a blinded fashion. Warfarin was applied in an un-blinded fashion in tablets of 1, 3, and 5 mg and was adjusted locally to international normalized ratio (INR of 2.0–3.0) with monthly measured INR. Study medication was given on top of standard treatment used by the treating physicians according to best clinical practice. Blood pressure and other measurements Follow-up visits were scheduled 14 days after randomization and every 3 months thereafter in the first year and then every 4 months until the study end. At each visit brachial SBP was measured in a sitting position after resting for three minutes using an automatic validated device (Omron model HEM 757, Omron, Kyoto, Japan) and the measurement was attended by the study nurses or investigators. Only patients with available baseline or post-randomization BP data were included in the analysis. Patients randomized to warfarin underwent INR testing at least once every four weeks. All patients had monthly laboratory evaluations of hepatic function during the first year using a central laboratory. Patients were recruited from 951 centres in 44 countries. After the screening of 20 377 patients, 18 113 patients were randomized. In six patients no baseline or post-baseline BP was available, resulting in a final number of 18 107 patients for the analysis of mean achieved BP (for allcause death). The study flow and treatment allocation are given in Figure 1. All primary and secondary outcomes (with the exception of any bleeding events) were assessed by a blinded central committee according to standard criteria. An average of 8.3 (SD 2.6) BP measurements was available measured during an average of 21.0 months (SD 9.6). Outcomes The primary composite outcome of RE-LY was stroke or SE. The primary safety outcome was major haemorrhage. Secondary outcomes were stroke, systemic embolism, or death among others. Stroke was defined as a sudden onset of neurological deficit consistent with a territory of a major cerebral artery, ischaemic, haemorrhagic, or unspecified. Systemic embolism was defined as an acute vascular occurrence of any extremity or organ documented by means of imaging surgery or autopsy. Major bleeding was defined as a reduction of the haemoglobin level of at least 20 g/L, transfusion of at least two units of blood or symptomatic bleeding in a critical area or organ. Any other bleeding events of lower severity were added to major bleeding events and defined as any bleeding. Intracerebral bleedings were too rare to be studied according to BP groups. Statistical analyses Randomized groups were combined for this analysis on SBP and DBP. Patients were divided into groups on the basis on mean achieved seated clinical BP according to the following cut-offs: SBP <110 mmHg, 110 to <120 mmHg, 120 to <130 mmHg, 130–<140 mmHg, 140 to <160 mmHg and >_ 160 mmHg. For DBP, cut-offs were <_70 mmHg, 70 to <80 mmHg, 80 to _90 mmHg. Patient baseline data were tested for differences using analysis of variance (ANOVA) for continuous data and the v2 test for categorical data. Yearly event rates and Kaplan–Meier curves for all-cause death, stroke or SE, major bleeding, and any bleeding are presented treatment-independently as well as by treatment for SBP and DBP categories. Hazard ratios (HRs) and corresponding confidence intervals are obtained from Cox regression analyses, adjusting for risk predictors as detailed in the figure legends. Finally, cubic splines for the adjusted HRs using mean achieved SBP as a continuous variable were generated (using four knots expressed as quantiles of 0.05, 0.35, 0.65, and 0.95, respectively). In addition, to support results obtained from meanachieved BP analyses further sensitivity analyses by dividing patients into groups by baseline BP and time-updated BP were conducted. Finally, a time-varying covariate analysis was applied to the Cox model to account for the variability in BP use over time by using BP as time-varying covariate. All analyses were done using SAS version 9.4 (SAS Institute, NC, USA) 0.95, respectively). Results Table 1 shows the demographic and clinical characteristics of the RELY population grouped by mean achieved SBP. Patients with higher mean achieved SBP were older, had a higher body mass index and were more likely women. Paroxysmal, persistent, or permanent atrial fibrillation had a slightly different distribution, whereas Aspirin was distributed similarly between the groups. Kaplan–Meier curves in Figure 2 show occurrence of all-cause death (Figure 2A) or SSE (Figure 2B) according to mean achieved SBP. Lowest risk occurred at SBP between 120 to 140 mmHg, while risk was visibly increased at <110 and >160 mmHg. These separations were less for SSE; however, lowest risk was preva- lent at 120–130 mmHg. A different picture was observed for major bleeding (Figure 2C) and any bleeding (Figure 2D). Risk was increased with an SBP <120 mmHg for any bleeding or major bleeding (Figure 2C and D) with no increased risk for any bleeding or major bleeding events at higher achieved BP. Figure 3 shows Forest plots demonstrating relative hazards. The guideline-recommended target SBP of 120 to <130 mmHg is given as reference.2 All-cause death risk was increased at <110 mmHg (HR 2.78; 2.28–3.38) and 110–120 mmHg (HR 1.48; 1.25–1.75) and at >_160 mmHg (HR 1.79; 1.24–2.60) for mean achieved SBP (red). The results were similar for baseline (blue), time-updated (yellow), and time-varying SBP (green), but mean-achieved SBP (red) showed stronger risk increases at low and high SBP. Similar patterns, with, however, wider distributions were observed for SSE but with no increase of risk at lower mean-achieved SBP (Supplementary material online, Figure S1A). Major bleeding (Figure 3C) and any bleeding (Supplementary material online, Figure S1B) were increased at mean-achieved SBP _140 mmHg only any bleeding (Supplementary material online, Figure S1B) but not major bleeding was increased. Mean-achieved SBP was associated with a higher risk at high SBP for SSE and high and low SBP for any bleeding than baseline, time-updated and time-varying SBP. Analysis taking mean-achieved SBP as continuous variable and using cubic spline regression revealed that the relationship between outcomes and bleeding events were non-linear. Figure 4 shows HRs for all-cause death (Figure 4A), SSE (Figure 4B), major bleeding (Figure 4C), and any bleeding (Figure 4D). Risk was increased, in particular, for all-cause death (Figure 4A), at low SBP, and for SSE at high BP (Figure 4B). Bleeding events were increased at low SBP, while at higher SBP no increased risk for major (Figure 4C) but increased rates for any bleeding events were observed. The numbers for intracerebral bleeding were low (i.e. 6/915 patients at <110 mmHg, 32/3418 patients at 140 to <160 mmHg, and 2/346 patients at >_160 mmHg) preventing a detailed analysis by BP with the other events. Bleeding rates, in particular, at low SBP <110 mmHg, were higher than stroke risk. However, only ca. 5% of RELY patients were at such low SBP. Kaplan–Meier curves showed that mean-achieved DBP <70 mmHg was associated with higher rates of all-cause death (Figure 5A) but not for SSE (Figure 5B). Bleeding events were also higher in the low DBP group (<70 mmHg) for major bleeding (Figure 5C) and any bleeding (Figure 5D). Major bleeding events were not increased at DBP >_90 mmHg. Results are summarized in the Forest plots in Figure 6 and Supplementary material online, Figure S2 with the guideline recommended DBP of 70 to <80 mmHg as refer- ence. All-cause death was increased at mean-achieved DBP <70 mmHg (HR 1.55; 1.36–1.77) and 80 to <90 mmHg (HR 1.19; -1.03 to 1.37) or >_90 mmHg (HR 1.88; 1.43–2.46) vs. reference. There were no relevant differences between baseline (blue), time-updated (yellow), time-varying (green) DBP, but mean-achieved (red) DBP was associated with higher risk at high DBP for all-cause death (Figure 6A).
High DBP >_80 mmHg was associated with higher rates for any (Supplementary material online, Figure S2B) but not for major bleeding (Figure 6B). Furthermore, we performed an exploratory analysis combining SBP and DBP. There was a significant interaction with SBP and DBP for all-cause death (P< 0.0001) and for any bleed (P< 0.0001). Figure 7 summarizes the treatment-dependent HR according to achieved SBP. There was no significant interaction between dabigatran 110 mg bid vs. warfarin and 150 mg bid dabigatran vs. warfarin for all-cause death (Figure 7A) and SSE (Figure 7B) according to achieved SBP. An interaction was seen for major bleeding (Figure 7C, P= 0.0141) but not for any bleeding (Figure 7D). In general, the results are consistent with the overall reduction of SSE and all-cause death with 150 mg dabigatran compared to warfarin [all-cause death: HR 0.85; 0.75–0.97; SSE: 0.62; 0.49–0.77, Figure 7A and B, right]. For 110 mg dabigatran, major bleeding (HR 0.80; 0.70–0.93) and any bleeding (HR 0.79; 0.74–0.84) were lower for dabigatran 110 mg compared to warfarin with no relevant interaction between the different SBP groups (Figure 7C and D left). Similar results were observed for DBP with lower all-cause death and SSE (Supplementary material online, Figure S3A and B) of 150 mg dabigatran vs. warfarin. Lower rates of major bleeding (Supplementary material online, Figure S3C), (HR 0.80; 0.70–0.93), and any bleeding (HR0.79; 0.74–0.84) were observed for dabigatran 110 mg vs. warfarin (Supplementary material online, Figure S3C and D). Patients with low SBP (<110 mmHg) also had creased rates of non-cardiovascular death (HR 2.22; 1.68–2.92, Supplementary material online, Figure S4). There were no significant differences by regions. Adjusted hazards of Asia compared to other regions did not differ significantly (P = 0.142 for interaction for all-cause death). Across the levels of SBP the quality of anticoagulation as judged from time in therapeutic range (TTR, at INR 2–3) was similar (warfarin treated TTR <110 mmHg 62.8 ± 19.8%, 110 to <120 mmHg 64.7 ± 18.5%, 120 to <130 mmHg 63.8 ± 19.8%, 130 to <140 mmHg 65.3 ± 19.6%, 140 to <160 mmHg 64.1 ± 20.9%, >_160 mmHg 64.2 ± 23.1%, all 64.4 ± 19.8%).

Discussion
This study shows that in a population of patients with atrial fibrillation at high stroke risk and a hypertension prevalence of 78.9% with most patients taking antihypertensive drugs, U- and J-shape curves were observed for all-cause death over a broad spectrum of achieved SBP and DBP. For SSE, risk was increased at high but not low SBP/DBP. Interestingly, the risk for any bleeding or major bleeding was higher in patients with low SBP and DBP compared with higher BP levels, whereas no risk increase was observed for major bleeding. As in the overall population, the treatment effect resulting in more pronounced reduction of all-cause death and SSE for 150 mg dabigatran vs. warfarin and less bleeding for 110 mg dabigatran was not significantly altered by achieved SBP or DBP. Mean-achieved low or high SBP and DBP was more closely associated to outcomes compared to baseline, time-updated or time-varying BP values. In recent studies in high-risk patients after myocardial infarction3 or after myocardial infarction, stroke or diabetes with proven vascular disease,4 an increased risk has been observed at lower mean-achieved SBP and DBP resulting in a J-shape curve for the BPrisk association. In the RE-LY study, patients with atrial fibrillation at higher stroke risk and cardiovascular comorbidities reflected by prior myocardial infarction (17.6%), heart failure (32%), diabetes (23%), and hypertension (79%) were randomized.10 The risk association for all-cause death or SSE to SBP or DBP was also comparable to previous studies. In RE-LY, however, risk of death was more increased at lower SBP levels. This could reflect the higher prevalence of heart failure (32% overall)10 with a particularly high prevalence of heart failure in patients with SBP failure, a particular high risk was observed at low SBP and DBP,15 which might be due to reduced coronary perfusion, reverse caus- ation or use of hypotensive drugs.16 While in previous studies3,4 only a minority of patients received long-term anticoagulation, an effect of the administered anticoagulation in RE-LY in patients at lower BP could have contributed to higher mortality rates because bleeding rates were increased at low BP. Nevertheless, one might speculate that bleeding rates might contribute to the J-shape curve in hyperten- sion and heart failure in general, which has not been addressed in previous studies. This study also extends previous findings by show- ing that not only mean-achieved BP but also baseline, time-updated, and time-varying BP values shows similar results. However, mean- achieved SBP and DBP showed closer associations to outcomes than the other parameters, because it might give a good integration of intermediately occurring morbidities affecting BP and outcomes.
There are no prospective data on the effect of oral anticoagulation on the BP-risk association for all-cause death and SSE, and in particu- lar data on bleeding rates in patients with atrial fibrillation are lacking. This analysis has shown that, counterintuitively, low BP values (<120 mmHg SBP or <70 mmHg DBP) were associated with higher rates of any or major bleeding. Several risk factors have been identi- fied in RE-LY to be significant confounders to cardiovascular events and bleeding events.10 For these confounders, appropriate adjustments were done. In particular, the rate of aspirin prescription was comparable across the different SBP or DBP groups. These findings suggest that patients with low BP are possibly frailer and are, therefore, more likely suffer from bleeding events. This is supported by the increase in non-cardiovascular death at low SBP (Supplementary material online, Figure S5). Furthermore, low DBP was observed to be associated with any bleeding and major bleeding as well as all-cause death in particular at high SBP. This strengthens the finding that vascular vulnerability reflected by low DBP and high pulse pressure (at high SBP) is associated with complications. The recommendation of the 2018 ESC/ESH guidelines to target BP; however, no on treatment BP data were pro- vided.20 In a retrospective analysis of the Japanese J-RHYTHM Registry, increased bleeding rates at the highest BP quartile were observed, showing that BP could be relevant.21 In a Korean registry on 298 374 patients with AF and with 62% hypertensives among them, without anticoagulation, major cardiovascular event risk was higher in hypertensives as non-hypertensives at low BP.22 Therefore, it would have been interesting to explore the role of duration of hypertension, which has been shown to continuously increase the risk of stroke for 7 years.23 However, duration of hypertension was not captured in RELY. Other contemporary studies should evaluate risk BP associations to confirm these findings. Recent hypertension guidelines1,2 recommend strict treatment goals in high cardiovascular risk patients of 120–129 mmHg SBP and 70–79 mmHg DBP whenever tolerated. In agreement with the guideline recommendations, the rates of all-cause death and SSE were lowest at these guidelines-recommended SBP and DBP tar- gets. In line, a recent secondary analysis from the CLARIFY regis- try showed that patients titrated from 130–139 to 120– 129 mmHg had the lowest risk.24 However, major bleeding rates were not increased at high SBP and even higher at low achieved on-treatment SBP and DBP. When stricter BP targets are recommended by guidelines,3,4 more patients might achieve lower BP values of available on the BP-risk association for SSE, bleeding and all-cause death in patients on long-term oral anticoagulation. Therefore, this analysis fills the gap, which might be relevant for future guideline recommendations in this growing group of patients. The data are summarized in the Take home figure. Conclusions Our study indicates that there are J-shape curves in patients on oral anticoagulation for the association of AM 095 and DBP to all- cause death, and any bleeding. Interestingly and partly counterin- tuitively, low achieved BP rather than high achieved BP was associ- ated with higher rates of major bleeding and allows to identify patients on high risk in general. Regardless of whether or to which extent inverse causation is involved, in patients on anticoagulation at low BP, the higher bleeding rates in addition to the higher death rates suggest caution, identifies individuals at high risk and sup- ports lower boundaries of BP recommendations despite general stricter BP control in contemporary guidelines, although causal relationships are not proven.