Author + information
- Received March 29, 2018
- Revision received April 19, 2018
- Accepted April 20, 2018
- Published online July 2, 2018.
- C. Michael Gibson, MS, MDa,b,∗ (, )@CMichaelGibson@BrighamWomens,
- Bennett Levitan, MD, PhDc,
- William J. Gibson, MD, PhDa,b,
- Megan K. Yee, MPHa,
- Sabina A. Murphy, MPHb,
- Zhong Yuan, MD, PhDc,
- Anjan K. Chakrabarti, MDa,
- Michael Lee, PhDd and
- Eugene Braunwald, MDb
- aCardiovascular Division, Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts
- bTIMI Study Group, Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts
- cJanssen Research and Development, Titusville, New Jersey
- dJanssen Research and Development, Raritan, New Jersey
- ↵∗Address for correspondence:
Dr. C. Michael Gibson, Beth Israel Deaconess Medical Center, Cardiovascular Division, 930 Commonwealth Avenue, Boston, Massachusetts 02215.
Background Net clinical outcome analyses of acute coronary syndrome (ACS) mingle fatal or irreversible events with survivable or reversible events that vary significantly in clinical impact.
Objectives A comparison of efficacy and safety limited to fatal or irreversible ischemic and adverse or seriously harmful events is one way to assess net clinical outcome and risk-benefit overall, given the fact that these events have a similar clinical impact.
Methods In the ATLAS ACS 2-TIMI 51 (Anti-Xa Therapy to Lower Cardiovascular Events in Addition to Standard Therapy in Subjects with Acute Coronary Syndrome-Thrombolysis In Myocardial Infarction) trial of rivaroxaban in the secondary prevention of events among patients with ACS treated with aspirin plus clopidogrel or ticlopidine (clopidogrel/ticlopidine) or aspirin alone, fatal and irreversible efficacy events including nonbleeding cardiovascular death, myocardial infarction, and ischemic stroke were compared to fatal or irreversible safety events, including fatal and intracranial bleeding.
Results Rivaroxaban, 2.5 mg orally twice per day, in patients treated with aspirin and clopidogrel/ticlopidine was associated with 115 (95% confidence interval [CI]: 18 to 212) fewer fatal or irreversible ischemic events (663 for placebo vs. 548 for therapy) and 10 (95% CI: −11 to 32) additional fatal or irreversible seriously harmful events (33 vs. 23 for placebo) per 10,000 patient-years of exposure. Taken together, there would be 105 (95% CI: 6 to 204) fatal or irreversible events prevented per 10,000 patient-years of exposure to rivaroxaban compared with placebo, with 11 (10 of 115) fatal or irreversible ischemic events prevented for each fatal or irreversible seriously harmful event caused. If only nonbleeding cardiovascular death is included as a fatal or irreversible event, then 95 events would be prevented per 10,000 patient-years of exposure in the group taking 2.5 mg orally twice per day.
Conclusions Both fatal or irreversible ischemia and bleeding are clinically significant events that can be compared to assess the net clinical outcomes associated with therapy. Rivaroxaban therapy at an oral dose of 2.5 mg twice daily in patients treated with aspirin and clopidogrel is associated with a net reduction in fatal or irreversible events. (Anti-Xa Therapy to Lower Cardiovascular Events in Addition to Standard Therapy in Subjects with Acute Coronary Syndrome-Thrombolysis In Myocardial Infarction [ATLAS ACS 2-TIMI 51]; NCT00809965)
Despite the intuitive notion that, to be useful clinically, the benefit of a pharmacotherapy must outweigh its risk, there is currently little regulatory guidance regarding the evaluation of net clinical outcomes. To date, in acute coronary syndrome (ACS) trials, risk-benefit and net clinical outcome analyses have often mingled fatal or irreversible events with survivable or reversible events that vary considerably in their clinical impact. For instance, one endpoint that has been used in cardiovascular trials of antithrombins and antiplatelets is net adverse clinical events (1,2). This endpoint is a composite that includes cardiovascular (CV) death, myocardial infarction (MI), target vessel revascularization, stroke, and major bleeding, each weighted equally and typically computed using a time-to-first event analysis. Although this can provide the practicing clinician with some sense of the overall risk-benefit profile of a therapy, this type of endpoint has limitations and can be misleading. First, all types of events are weighted equally; for example, reversible nonintracranial hemorrhage (non-ICH), nonfatal bleeds that can be managed with supportive care, are weighted equally with death and disabling stroke. Second, stroke can be either hemorrhagic or ischemic, and the relative contributions of hemorrhagic or ischemic stroke may not be appropriately assigned to risk-versus-benefit categories in many analyses.
Unger et al. (3) and Beasley et al. (4) recently suggested that cardiovascular risk-benefit analyses should focus on events that are clinically meaningful and are of similar clinical significance such as fatal or irreversible events. A comparison of efficacy and safety that is initially limited to fatal or irreversible ischemic and harmful events as proposed by Unger et al. (3) would be one way to assess the net clinical outcome of efficacy and safety, given that the events have a similar clinical impact. The goal of the present analysis was to apply such an approach to a comparison of the risk-benefit of adding low-dose rivaroxaban, a novel factor Xa inhibitor, with that of placebo in ACS patients receiving standard-of-care therapy.
ATLAS ACS 2-TIMI 51 (Anti-Xa Therapy to Lower Cardiovascular Events in Addition to Standard Therapy in Subjects with Acute Coronary Syndrome-Thrombolysis In Myocardial Infarction), as described previously (5,6), was a randomized, double-blind, placebo-controlled, event-driven trial, which enrolled a broad population of patients with ACS, including those with ST-segment elevation MI (STEMI), non–STEMI (NSTEMI), or unstable angina. Subjects were enrolled within 7 days of hospital admission for ACS, after stabilization with either revascularization or revascularization and use of parenteral therapies. Major exclusion criteria included gastrointestinal bleeding within 12 months, previous ICH, previous stroke, or transient ischemic attack among patients taking both aspirin and clopidogrel or ticlopidine (clopidogrel/ticlopidine), who had a creatinine clearance <30 ml/min, a platelet count <90,000 per cc, or hemoglobin concentration <10 g /dl. Subjects were randomized 1:1:1 to either rivaroxaban, 2.5 mg orally twice per day, or 5 mg orally twice per day, or placebo, in addition to standard of care, which included low-dose aspirin (75 to 100 mg/day). Use of clopidogrel/ticlopidine was recommended but was initiated at the discretion of the enrolling physician, and the enrollment and analysis were stratified by co-administration of clopidogrel/ticlopidine. All study procedures were approved by the institutional review board, and study participants provided written informed consent.
Definitions of the ATLAS ACS 2-TIMI 51 trial’s endpoints have been described previously (5). Net clinical outcome, a secondary endpoint of the trial, was measured as a composite of CV death, MI, ischemic stroke, and non-coronary artery bypass graft (non-CABG) Thrombolysis In Myocardial Infarction (TIMI) major bleeding, each weighted equally and typically computed using a time-to-first event analysis. All safety and efficacy endpoints were adjudicated by members of a clinical events committee who were blinded to study group assignment. Fatal and irreversible efficacy endpoints included CV death (not due to bleeding), MI, and ischemic stroke. These events were considered separately and as a composite endpoint. In the construction of the composite endpoint, only 1, the first benefit (CV death, MI, or stroke prevented) or seriously harmful event, was counted. As a sensitivity analysis, nonfatal MIs excluding peripercutaneous intervention events were counted instead of all MIs.
Fatal and intracranial bleeding constituted the fatal or irreversible harmful events. As a sensitivity analysis and to show a complete risk-benefit profile with all types of bleeding, TIMI life-threatening bleeding and TIMI major bleeding were also assessed.
All risk-benefit analyses were performed using a modified intention-to-treat approach, which included all randomized subjects and the endpoint events that occurred after randomization and no later than the completion of the treatment phase of the study (i.e., the global treatment end date, including 72.4% of study population), 30 days after early permanent discontinuation of the study drug (including 27.6% of study population), or 30 days after randomization for subjects who did not receive a study drug (1.1% of study population) (5,6). The intention-to-treat mode of analysis was performed as a sensitivity analysis. Given the mortality advantage and the reduced risk of fatal bleeding of 2.5 mg orally twice per day compared to 5 mg orally twice per day of rivaroxaban (5), the primary analysis focused on the net clinical outcome associated with 2.5 mg orally twice per day. Sensitivity analyses also were performed using the pooled data from the 2.5 mg orally twice daily and 5 mg orally twice daily groups.
In the ATLAS ACS 2-TIMI 51 trial, clopidogrel/ticlopidine exposure was independent of study drug exposure, and randomization was stratified on the basis of planned use of clopidogrel/ticlopidine. Stratum 1 was defined as those subjects whom the enrolling physician chose to treat with aspirin alone and stratum 2 as those subjects whom the enrolling physician chose to treat with aspirin plus clopidogrel/ticlopidine. However, the net clinical outcomes used for this analysis included only patients in stratum 2 so that bleeding outcomes were not underestimated. Results were examined according to major subgroups for general consistency of treatment effect. Additionally, because risk-benefit may vary with duration of exposure to study drug over time, differences in Kaplan-Meier rates for benefit (ischemic) and harm (bleeding) events over time were calculated. Results are presented according to 10,000 patient-years of exposure with 95% confidence intervals.
Results of the conventional net clinical outcome analysis are shown in Table 1. The 1.7% absolute difference in cardiovascular mortality advantage for 2.5-mg twice daily dosage of rivaroxaban compared with placebo is offset by a 1.3% increase in non-CABG TIMI major bleeding events, most of which were reversible and nonfatal bleeds associated with a drop in hemoglobin and/or blood transfusion.
However, in an analysis assessing events that had a similar clinical impact, rivaroxaban at an oral dose of 2.5 mg twice daily was associated with 115 (95% confidence interval [CI]: 18 to 212) fewer fatal or irreversible ischemic events (663 CV deaths not due to bleeding, MIs or ischemic strokes for placebo vs. 548 events) per 10,000 patient-years of exposure. Results are shown with risk differences rather than hazard ratios to clearly show events caused versus those prevented and include results from the pooled analysis as well (2.5 mg and 5 mg orally twice per day) (Central Illustration, Figure 1⇓⇓).
On the other hand, rivaroxaban at an oral dose of 2.5 mg twice daily was associated with 10 (95% CI: −11 to 32) excessive, fatal, or irreversible seriously harmful events (33 fatal bleeds or ICHs for a 2.5-mg dose of rivaroxaban vs. 23 events for placebo) per 10,000 patient-years of exposure (Central Illustration). Considered together, there would be 105 (95% CI: 6 to 204) fatal or irreversible events prevented per 10,000 patient-years of exposure to 2.5 mg of rivaroxaban taken orally twice a day compared with placebo. An alternate interpretation of the data is that there would be 11 (10 of 115) fatal or irreversible ischemic events prevented for each fatal or irreversible harmful event caused.
In a sensitivity analysis, if periprocedural MIs were excluded, 115 fatal or irreversible ischemic events would still be prevented in the group taking rivaroxaban, 2.5 mg twice daily. If all MIs were excluded from the analysis of fatal or irreversible ischemic events and only nonbleeding CV death or ischemic strokes were included, then 90 fatal or irreversible events would be prevented in the group taking rivaroxaban, 2.5 mg twice daily. If only nonbleeding CV death were included as a fatal or irreversible event, then 95 events would be prevented per 10,000 patient-years of exposure in the group taking rivaroxaban, 2.5 mg twice daily. In all cases, the fatal or irreversible events prevented are 9 to 11 times the fatal or irreversible seriously harmful events caused.
Figures 2A and 2B show additional sensitivity analyses of safety or seriously harmful events. Even if nonfatal, non-ICH TIMI life-threatening, and TIMI major bleeding were counted as fatal or irreversible events, there remains a numerical net clinical outcome advantage in favor of rivaroxaban, 2.5 mg taken orally twice per day (Figures 2A and 2B). In a sensitivity analysis of the 5-mg oral, twice daily dosage of rivaroxaban, the 5-mg dose was associated with the prevention of 108 fatal or irreversible ischemic events, while causing 27 fatal bleeds or ICHs for a net benefit of 81 fatal or irreversible events prevented per 10,000 patient-years of exposure. In addition, a sensitivity analysis using intention-to-treat mode was performed, and results were similar (data not shown). Among patients without TIMI non-CABG major bleeding, both of the rivaroxaban dosage groups demonstrated a total mortality (hazard ratio [HR] for 2.5-mg twice a day was 0.58; 95% CI: 0.43 to 0.79; HR for 5 mg twice a day was 0.75; 95% CI: 0.56 to 0.99) and CV mortality (HR for 2.5 mg twice a day was 0.57; 95% CI: 0.42 to 0.78; and HR for 5 mg twice a day was 0.77; 95% CI: 0.58 to 1.02) benefit compared to placebo.
Time-to-event analysis demonstrated that the risk of fatal or irreversible harm remained low and constant over time, whereas reduction in fatal or irreversible ischemic events expanded (Figure 3). By 720 days, a net of 142 fatal or irreversible events (165 fatal or irreversible ischemic events were prevented, and 23 fatal or irreversible seriously harmful events were caused) would have been prevented by 2.5-mg oral doses twice per day of rivaroxaban. Additional time-to-event sensitivity analyses demonstrated similar results, even when TIMI major bleeding was included as a fatal or irreversible event (Online Figures 1 to 3).
The present study used a methodology that separated events into those fatal or irreversible ischemic events that were prevented and compared them with the harmful fatal or irreversible events caused by a pharmacotherapy. In this form of analysis, only fatal or irreversible events were included so that benefit and seriously harmful events of similar clinical impact were compared. This is particularly important when the endpoints and analyses do not include measurements of subjective clinical impact such as utility measurements or preference weights. This approach also uses risk differences rather than relative measurements such as hazard ratios, so the number of events prevented and caused are clearly distinguished.
There is currently limited regulatory guidance regarding the appropriate method to evaluate risk-benefit or measurements of net clinical outcome quantitatively. Although risk-benefit evaluations are important to compare the tradeoff between efficacy and safety measures, such comparisons are not always straightforward because the endpoints often vary in their clinical importance from patients’ and physicians’ perspectives. Although some investigators have advocated that fatal and irreversible harmful outcomes may be more important than nonfatal extracranial hemorrhage (3,4), others have expressed alternate views (7–9). In a recent editorial, Verheugt et al. (10) highlighted the importance of evaluating each component of composite outcomes and using novel methodologies, such as the “win ratio,” to evaluate net clinical benefit (11). Our risk-benefit assessment of rivaroxaban therapy focused not only on clinically important fatal and irreversible harmful outcomes but also presented results of sensitivity analyses that included events of varying clinical impact. For example, in the primary analysis, MI was included as an irreversible ischemic event. Although the myonecrosis of MI may be irreversible from a pathologic perspective, the clinical relevance of MI, particularly periprocedural MI, has been debated (12–15). In a sensitivity analysis, even when periprocedural MIs were excluded from the present analysis, a net clinical outcome benefit persisted in favor of rivaroxaban therapy. Likewise, even when all MIs were excluded from the analysis, a net clinical outcome benefit persisted in favor of rivaroxaban. When both MI and ischemic stroke were excluded from the analysis and only nonbleeding cardiovascular death was included, the results favored rivaroxaban. Likewise, in sensitivity analyses that incorporated reversible, nonfatal forms of TIMI major or TIMI life-threatening bleeding, the results favored rivaroxaban numerically.
It is also worth noting that our risk-benefit assessment focused on absolute risk differences rather than hazard ratios, as the comparison of the latter form is not always straightforward. For instance, even with similar clinical importance, a 20% relative reduction (e.g., HR: 0.8) in efficacy endpoints may not be equivalent to a 20% relative increase in safety endpoints (e.g., HR: 1.2), because the background rates for efficacy and safety may differ.
In the primary analysis, the benefit to a typical patient exposed for the typical (median) duration was calculated. This primary method of analysis does not evaluate changes in net clinical outcome over time. Indeed, the net benefit increased as the duration of exposure to rivaroxaban increased and that by 2 years, 142 fatal or irreversible events would be prevented per 10,000 patient-years of exposure.
These data are drawn from a randomized trial, and the results can be applied only to patients who met the eligibility criteria of the randomized trial. Patients were not given either prasugrel or ticagrelor in this study, and the results reported here apply only to patients treated with either clopidogrel or ticlopidine. This analysis does not incorporate patient and physician perspectives regarding the relative benefit or harm of death, MI, or stroke. Bleeding data and analyses are limited by the absence of the long-term impact that transfusion might have had on these patients. Additionally, although far less important clinically, TIMI minor or less significant bleeding events were not included, although minor bleeding events that might have affected adjunctive treatment might have had an impact on clinical endpoints.
These are typically time-to-first-event analyses; patients may have more than 1 bleeding and/or ischemic event, and the methods applied in this study ignore the contribution of later events and account only for the event which occurred first. For example, a patient with only a major bleed would not be distinguished from a patient with a major bleed who subsequently died from a stroke.
Transfusion and surgery were not included as irreversible events in this analysis. Of note, any death, stroke, ICH, or MI that resulted from any bleeding event during the timeframe of the trial was included in the net outcome analysis, regardless of the fact that the event was immediately associated with bleeding.
Fatal or irreversible ischemia or bleeding are both clinically significant events that can be compared to assess the net clinical outcome associated with a therapy. Rivaroxaban at an oral dose of 2.5 mg twice daily is associated with a net reduction in fatal or irreversible events.
COMPETENCY IN MEDICAL KNOWLEDGE: In patients with ACSs, addition of rivaroxaban, 2.5 mg orally twice daily, to dual antiplatelet therapy with aspirin and clopidogrel was associated with a net reduction in fatal or irreversible events compared to dual antiplatelet therapy alone.
TRANSLATIONAL OUTLOOK: Further studies are needed to assess the net clinical of other antithrombotic drug combinations in patients with ACS.
ATLAS ACS 2-TIMI 51 trial was a collaboration of the TIMI Study Group, the associated investigators from the executive and steering committees, and the sponsors Johnson & Johnson and Bayer Healthcare. The study design was approved by national and institutional ethics committees. Trial safety was evaluated by ongoing reviews of unblinded data by an independent data safety monitoring committee. Statistical analyses were performed and validated by the TIMI Study Group and the PERFUSE (Percutaneous-Pharmacologic Endoluminal Revascularization for Unstable Syndromes Evaluation) Study Group, using an independent copy of the complete raw database. Members of the TIMI Study Group bear responsibility for accuracy, completeness of data reported, and drafting of the first and last drafts of the manuscript.
ATLAS ACS 2-TIMI 51 was supported by Johnson & Johnson and Bayer Healthcare. Dr. Gibson has received funding through his institution from Johnson & Johnson and Bayer Healthcare; and has received grants and honoraria from Bayer, Johnson & Johnson, and Portola. Dr. Levitan is an employee of Janssen Research and Development; and holds stock in Janssen, Baxter International, Inc., Pharmaceutical Holdrs Trust, and Zimmer Holdings, Inc. Drs. Yuan and Lee are employees of Janssen Research and Development. Dr. Braunwald, Ms. Murphy, and the TIMI Study Group have received research grants from Johnson & Johnson and Daiichi-Sankyo. Dr. Braunwald has received speakers honoraria from Daiichi-Sankyo and Bayer. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- acute coronary syndrome
- intracranial hemorrhage
- myocardial infarction
- Thrombolysis In Myocardial Infarction
- Received March 29, 2018.
- Revision received April 19, 2018.
- Accepted April 20, 2018.
- 2018 American College of Cardiology Foundation
- Gibson C.M.,
- Mega J.L.,
- Burton P.,
- et al.
- Eikelboom J.W.,
- Mehta S.R.,
- Anand S.S.,
- Xie C.,
- Fox K.A.,
- Yusuf S.
- Harris B.M.,
- Nageh T.,
- Marsden J.T.,
- Thomas M.R.,
- Sherwood R.A.
- Januzzi J.L.,
- Lewandrowski K.,
- MacGillivray T.E.,
- et al.