Apixaban Plus Mono vs Dual Antiplatelet Therapy in ACS
The trial design and main results of APPRAISE-2 have been published. In brief, APPRAISE-2 was a double-blind, placebo-controlled randomized clinical trial of patients who experienced ACS within the previous 7 days, with symptoms of myocardial ischemia at rest, elevated cardiac biomarkers, or dynamic ST-segment changes on the electrocardiogram, and who received aspirin or aspirin plus a P2Y12 receptor inhibitor. Patients were also required to have ≥2 of the following risk factors: age ≥65 years, diabetes, myocardial infarction (MI) within the previous 5 years, cerebrovascular disease, peripheral artery disease, clinical heart failure or a left ventricular ejection fraction (LVEF) <40% associated with the index event, creatinine clearance <60 ml/min, or no revascularization after the index event.
Patients were randomized to apixaban (5 mg twice daily or 2.5 mg twice daily for a creatinine clearance <40 ml/min) versus placebo. Other treatment decisions, including the use, choice, and duration of antiplatelet therapy, were left to the discretion of treating providers. The trial was stopped early at the recommendation of the independent data monitoring committee due to increased bleeding events with apixaban without any offsetting reduction in ischemic events. A total of 7,392 patients at 858 sites from 39 countries were randomized.
The median duration of follow-up was 241 days (25th, 75th percentiles: 131, 352). The primary efficacy endpoint for APPRAISE-2 and for this analysis was a composite endpoint of cardiovascular death, MI, and ischemic stroke. Secondary efficacy outcomes included cardiovascular death and all-cause mortality. Thrombolysis In Myocardial Infarction (TIMI) major bleeding was the primary safety endpoint, with TIMI major or minor bleeding assessed as a secondary safety outcome. All outcomes were adjudicated by a clinical events committee blinded to treatment assignment.
For the first study objective to describe antiplatelet therapy use in the overall trial population, we defined mono antiplatelet therapy as the use of any antiplatelet agent alone (aspirin, clopidogrel, prasugrel, or ticlopidine). Dual antiplatelet therapy was defined as use of aspirin plus at least 1 thienopyridine (clopidogrel, prasugrel, or ticlopidine). Due to the small numbers of patients on non-aspirin, single antiplatelet regimens, we examined specific regimens of aspirin alone and aspirin plus clopidogrel for our main analysis population and for modeling. Daily records for antiplatelet therapy use were summarized to weekly data. Patients were considered to be on a particular regimen if they received this regimen for >50% of the week. Dosage information regarding aspirin or P2Y12 receptor inhibitors was not available.
Mono (any antiplatelet agent alone) versus dual (aspirin plus any thienopyridine) antiplatelet therapy use was described over time among the overall APPRAISE-2 trial population. The proportion of patients on mono versus dual antiplatelet therapy of all available patients was calculated for each day since randomization. Due to a small number of patients (n = 23) not on aspirin or a P2Y12 receptor inhibitor at randomization, the proportions of mono and dual antiplatelet therapy use do not equal the sum of 1. Next, we assessed the proportion of patients on mono versus dual antiplatelet therapy before randomization, at randomization, and during the course of the trial. We calculated the number, direction, and timing of switch episodes, as well as the clinical rationale for switching between mono and dual antiplatelet therapy.
The main analysis population for our study included patients taking aspirin alone or aspirin plus clopidogrel. We examined patients according to their antiplatelet therapy regimens and according to whether they received treatment with PCI. Baseline categorical variables were presented as proportions and compared using the chi-square test; continuous variables were presented as medians with 25th and 75th percentiles and compared using the Wilcoxon rank-sum test.
Outcomes were presented as the number of events and rates per 100 patient-years. The impact of aspirin alone versus aspirin plus clopidogrel use at baseline on time-to-event for outcomes was characterized using the unadjusted hazard ratio (HR) and corresponding 95% confidence interval (CI) from a Cox proportional hazards model with a binary indicator for mono antiplatelet therapy. Next, we adjusted the Cox model for the following baseline variables that were associated with either the propensity to use aspirin or aspirin plus clopidogrel and the outcomes of interest: age; sex; body mass index; history of stroke; LVEF <40%; diabetes; hypertension; MI; peripheral artery disease; congestive heart failure; renal insufficiency; history of PCI; and history of coronary artery bypass graft surgery. To assess the impact of apixaban versus placebo according to baseline antiplatelet therapy, treatment effects within each baseline antiplatelet therapy group were characterized by adjusted HRs and associated 95% CIs of the binary treatment indicator for apixaban versus placebo (simple model), including an interaction term for assigned treatment (apixaban or placebo) by antiplatelet regimen (aspirin alone or aspirin plus clopidogrel). Although apixaban versus placebo was randomized, we adjusted for the same covariates associated with either the propensity to use aspirin or aspirin plus clopidogrel and the outcomes of interest, because this randomization was not performed within the antiplatelet therapy groups, nor was the use of aspirin or aspirin plus clopidogrel groups randomized. Adjustment was performed to ensure results were due to differences in assigned treatment and not to differences in case-mix between antiplatelet groups. Interactions were tested at a significance level of 0.05.
We next examined the influence of antiplatelet therapy on outcomes based on the actual antiplatelet regimen taken by patients during follow-up, instead of categorizing patients based on antiplatelet regimen at time of randomization, as was done for the simple model analyses. We achieved this by fitting a marginal structural model (MSM) involving a Cox model and propensity score methodology with antiplatelet therapy status as a time-dependent covariate. Rather than only considering baseline antiplatelet status, this approach allowed group membership to change during follow-up, and patients contributed information to the antiplatelet group depending on whether they were taking aspirin or aspirin plus clopidogrel at a particular time. This strategy allowed for estimation of a more accurate and clinically relevant effect of antiplatelet therapy on outcomes. In the MSM approach, the propensity for use of aspirin versus aspirin plus clopidogrel was modeled as a function of baseline and post-baseline characteristics, and inverse probability weights were used for covariate balancing. Through an iterative process, patients were re-weighted weekly during follow-up to balance the case-mix and to adjust for baseline covariates and post-baseline, time-dependent confounders related to both outcome and switching between aspirin and aspirin plus clopidogrel. In addition to the previously listed baseline covariates, the post-randomization, time-dependent confounders for which we adjusted for in the MSM included MI, angina, PCI, PCI with stent implantation, major bleeding, any bleeding, use of study drug, and use of dual antiplatelet therapy in the previous week. Similar to the main simple Cox model, adjusted MSMs were first used to assess the association of aspirin versus aspirin plus clopidogrel with outcomes and then to examine the impact of assigned treatment according to antiplatelet therapy and the interactions therein.
Finally, we repeated simple and MSM analyses according to treatment with PCI, because these patients might be expected to have differential risks of ischemic and bleeding outcomes. Simple model analyses could not be performed in the PCI subgroup due to insufficient patient numbers. The APPRAISE-2 trial was funded by Bristol-Myers Squibb (Princeton, New Jersey) and Pfizer, Inc. (New York, New York), and the sponsors participated in the trial design and data collection. Analyses presented in this analysis were designed by the authors and performed at the Duke Clinical Research Institute using SAS (version 9.2, SAS Institute, Cary, North Carolina).
Methods
Data Source and Patient Population
The trial design and main results of APPRAISE-2 have been published. In brief, APPRAISE-2 was a double-blind, placebo-controlled randomized clinical trial of patients who experienced ACS within the previous 7 days, with symptoms of myocardial ischemia at rest, elevated cardiac biomarkers, or dynamic ST-segment changes on the electrocardiogram, and who received aspirin or aspirin plus a P2Y12 receptor inhibitor. Patients were also required to have ≥2 of the following risk factors: age ≥65 years, diabetes, myocardial infarction (MI) within the previous 5 years, cerebrovascular disease, peripheral artery disease, clinical heart failure or a left ventricular ejection fraction (LVEF) <40% associated with the index event, creatinine clearance <60 ml/min, or no revascularization after the index event.
Patients were randomized to apixaban (5 mg twice daily or 2.5 mg twice daily for a creatinine clearance <40 ml/min) versus placebo. Other treatment decisions, including the use, choice, and duration of antiplatelet therapy, were left to the discretion of treating providers. The trial was stopped early at the recommendation of the independent data monitoring committee due to increased bleeding events with apixaban without any offsetting reduction in ischemic events. A total of 7,392 patients at 858 sites from 39 countries were randomized.
Outcomes and Definitions
The median duration of follow-up was 241 days (25th, 75th percentiles: 131, 352). The primary efficacy endpoint for APPRAISE-2 and for this analysis was a composite endpoint of cardiovascular death, MI, and ischemic stroke. Secondary efficacy outcomes included cardiovascular death and all-cause mortality. Thrombolysis In Myocardial Infarction (TIMI) major bleeding was the primary safety endpoint, with TIMI major or minor bleeding assessed as a secondary safety outcome. All outcomes were adjudicated by a clinical events committee blinded to treatment assignment.
For the first study objective to describe antiplatelet therapy use in the overall trial population, we defined mono antiplatelet therapy as the use of any antiplatelet agent alone (aspirin, clopidogrel, prasugrel, or ticlopidine). Dual antiplatelet therapy was defined as use of aspirin plus at least 1 thienopyridine (clopidogrel, prasugrel, or ticlopidine). Due to the small numbers of patients on non-aspirin, single antiplatelet regimens, we examined specific regimens of aspirin alone and aspirin plus clopidogrel for our main analysis population and for modeling. Daily records for antiplatelet therapy use were summarized to weekly data. Patients were considered to be on a particular regimen if they received this regimen for >50% of the week. Dosage information regarding aspirin or P2Y12 receptor inhibitors was not available.
Statistical Analysis
Mono (any antiplatelet agent alone) versus dual (aspirin plus any thienopyridine) antiplatelet therapy use was described over time among the overall APPRAISE-2 trial population. The proportion of patients on mono versus dual antiplatelet therapy of all available patients was calculated for each day since randomization. Due to a small number of patients (n = 23) not on aspirin or a P2Y12 receptor inhibitor at randomization, the proportions of mono and dual antiplatelet therapy use do not equal the sum of 1. Next, we assessed the proportion of patients on mono versus dual antiplatelet therapy before randomization, at randomization, and during the course of the trial. We calculated the number, direction, and timing of switch episodes, as well as the clinical rationale for switching between mono and dual antiplatelet therapy.
The main analysis population for our study included patients taking aspirin alone or aspirin plus clopidogrel. We examined patients according to their antiplatelet therapy regimens and according to whether they received treatment with PCI. Baseline categorical variables were presented as proportions and compared using the chi-square test; continuous variables were presented as medians with 25th and 75th percentiles and compared using the Wilcoxon rank-sum test.
Outcomes were presented as the number of events and rates per 100 patient-years. The impact of aspirin alone versus aspirin plus clopidogrel use at baseline on time-to-event for outcomes was characterized using the unadjusted hazard ratio (HR) and corresponding 95% confidence interval (CI) from a Cox proportional hazards model with a binary indicator for mono antiplatelet therapy. Next, we adjusted the Cox model for the following baseline variables that were associated with either the propensity to use aspirin or aspirin plus clopidogrel and the outcomes of interest: age; sex; body mass index; history of stroke; LVEF <40%; diabetes; hypertension; MI; peripheral artery disease; congestive heart failure; renal insufficiency; history of PCI; and history of coronary artery bypass graft surgery. To assess the impact of apixaban versus placebo according to baseline antiplatelet therapy, treatment effects within each baseline antiplatelet therapy group were characterized by adjusted HRs and associated 95% CIs of the binary treatment indicator for apixaban versus placebo (simple model), including an interaction term for assigned treatment (apixaban or placebo) by antiplatelet regimen (aspirin alone or aspirin plus clopidogrel). Although apixaban versus placebo was randomized, we adjusted for the same covariates associated with either the propensity to use aspirin or aspirin plus clopidogrel and the outcomes of interest, because this randomization was not performed within the antiplatelet therapy groups, nor was the use of aspirin or aspirin plus clopidogrel groups randomized. Adjustment was performed to ensure results were due to differences in assigned treatment and not to differences in case-mix between antiplatelet groups. Interactions were tested at a significance level of 0.05.
We next examined the influence of antiplatelet therapy on outcomes based on the actual antiplatelet regimen taken by patients during follow-up, instead of categorizing patients based on antiplatelet regimen at time of randomization, as was done for the simple model analyses. We achieved this by fitting a marginal structural model (MSM) involving a Cox model and propensity score methodology with antiplatelet therapy status as a time-dependent covariate. Rather than only considering baseline antiplatelet status, this approach allowed group membership to change during follow-up, and patients contributed information to the antiplatelet group depending on whether they were taking aspirin or aspirin plus clopidogrel at a particular time. This strategy allowed for estimation of a more accurate and clinically relevant effect of antiplatelet therapy on outcomes. In the MSM approach, the propensity for use of aspirin versus aspirin plus clopidogrel was modeled as a function of baseline and post-baseline characteristics, and inverse probability weights were used for covariate balancing. Through an iterative process, patients were re-weighted weekly during follow-up to balance the case-mix and to adjust for baseline covariates and post-baseline, time-dependent confounders related to both outcome and switching between aspirin and aspirin plus clopidogrel. In addition to the previously listed baseline covariates, the post-randomization, time-dependent confounders for which we adjusted for in the MSM included MI, angina, PCI, PCI with stent implantation, major bleeding, any bleeding, use of study drug, and use of dual antiplatelet therapy in the previous week. Similar to the main simple Cox model, adjusted MSMs were first used to assess the association of aspirin versus aspirin plus clopidogrel with outcomes and then to examine the impact of assigned treatment according to antiplatelet therapy and the interactions therein.
Finally, we repeated simple and MSM analyses according to treatment with PCI, because these patients might be expected to have differential risks of ischemic and bleeding outcomes. Simple model analyses could not be performed in the PCI subgroup due to insufficient patient numbers. The APPRAISE-2 trial was funded by Bristol-Myers Squibb (Princeton, New Jersey) and Pfizer, Inc. (New York, New York), and the sponsors participated in the trial design and data collection. Analyses presented in this analysis were designed by the authors and performed at the Duke Clinical Research Institute using SAS (version 9.2, SAS Institute, Cary, North Carolina).