Vorapaxar With or Without Clopidogrel After Non–ST-segment Elevation Acute Coronary Syndromes

Results From the Thrombin Receptor Antagonist for Clinical Event Reduction in Acute Coronary Syndrome Trial

Pierluigi Tricoci, MD, PhD; Yuliya Lokhnygina, PhD; Zhen Huang, MS; Frans Van de Werf, MD; Jan H. Cornel, MD; Edmond Chen, MD; Lars Wallentin, MD, PhD; Claes Held, MD, PhD; Philip E. Aylward, MD; David J. Moliterno, MD; Lisa K. Jennings, PhD; Harvey D. White, DSc; Paul W. Armstrong, MD; Robert A. Harrington, MD; John Strony, MD; Kenneth W. Mahaffey, MD

Disclosures

Am Heart J. 2014;168(6):869-877.

Abstract and Introduction

Abstract

Background Protease-activated receptor 1 antagonism with vorapaxar represents a novel strategy for platelet inhibition. In TRACER, vorapaxar was compared with placebo plus standard of care among 12,944 patients with non–ST-segment elevation acute coronary syndromes. We anticipated that most patients would have received clopidogrel as part of standard care. We investigated the modification of vorapaxar's effect associated with clopidogrel use over time.

Methods The marginal structural model method was used to estimate causal modification of vorapaxar effect by use of clopidogrel over time. The primary outcomes were the composite of cardiovascular death, myocardial infarction, or stroke and Global Use of Strategies to Open Occluded Coronary Arteries moderate or severe bleeding. The event accrual period excluded the time during which clopidogrel was clinically warranted.

Results Among 12,887 patients who received study medication, 11,117 (86.3%) received clopidogrel before randomization, of whom 38.5% stopped later in the trial (median time to stoppage 200 days with placebo; interquartile range [IQR] 14–367) (186 days with vorapaxar; IQR 17–366). In total, 1,770 (13.7%) patients were not on clopidogrel at randomization, of whom 47.8% started afterward (median time to start 2 days; IQR 2–4). During the period of event accrual, vorapaxar was associated with a 26% reduction in the composite of cardiovascular death, myocardial infarction, or stroke when used with clopidogrel (hazard ratio [HR] 0.74; 95% CI 0.60–0.91) and a 24% reduction when used without clopidogrel (HR 0.76; 95% CI 0.56–1.02) (interaction; P = .89). The hazard of Global Use of Strategies to Open Occluded Coronary Arteries bleeding with vorapaxar was not significantly different without clopidogrel (HR 1.33; 95% CI 0.81–2.20) or with clopidogrel (HR 1.09; 95% CI 0.76–1.56) (interaction; P = .53).

Conclusions We observed no interaction between vorapaxar and clopidogrel after non–ST-segment elevation acute coronary syndromes on efficacy or safety outcomes, supporting a complementary role of protease-activated receptor 1 and P2Y₁₂ antagonism.

Introduction

Vorapaxar is an oral, selective, protease-activated receptor 1 (PAR-1) antagonist. Thrombin-induced platelet activation is mediated by PARs and is considered to be the main thrombin receptor in humans.^[1] Vorapaxar administration results in potent inhibition of PAR-1–mediated platelet aggregation.^[1]

The TRACER trial was a large, international, phase III clinical trial that evaluated the efficacy and safety of vorapaxar compared with placebo in patients presenting with non–ST-segment elevation acute coronary syndromes (NSTE ACS) who were treated with standard of care.^[2] The TRACER trial did not meet its primary objective, with a nonstatistically significant 8% reduction in the composite of cardiovascular death, myocardial infarction (MI), stroke, coronary ischemia with hospitalization, or urgent coronary revascularization.^[2] A nominally significant 11% reduction in the key secondary end point of cardiovascular death, MI, or stroke was observed with vorapaxar and was driven by a reduction in MI. Vorapaxar significantly increased the risk of major bleeding. In the TRA 2P–TIMI 50 study, a large phase III trial of vorapaxar for secondary prevention of atherothrombosis, vorapaxar also significantly reduced cardiovascular death, MI, and stroke.^[3]

As part of the standard of care in TRACER, most patients were treated with dual antiplatelet therapy, with clopidogrel as the most commonly used P2Y₁₂ inhibitor. We designed this analysis to assess whether the observed effects of vorapaxar in TRACER were modified by concomitant exposure to clopidogrel. Assessing interaction between chronic antithrombotic therapies presents several analytical challenges, and thus, we have implemented the marginal structural model (MSM) method, aiming to control for time-dependent confounders.

1 2 3 4 5

Next Section

Abstract and Introduction
Methods
Results
Discussion
Conclusions
References
Appendix

Table I. Baseline demographics and inhospital procedure by use of thienopyridine
	No baseline clopidogrel	Baseline clopidogrel
	(n = 1770)	(n = 11,117)
Age, y	64 (58–72)	64 (58–71)
≥75	17.9	16.9
Female	35.0	27.1
White	84.2	85.7
Body weight, kilogram	81.6 (70.0–95.2)	80.0 (70.0–92.0)
<60 kg	8.7	7.5
Region
North America	48.1	23.1
Latin America	13.1	5.6
Western Europe	26.0	47.9
Eastern Europe	5.7	12.3
Asia	4.5	7.6
Australia/New Zealand	2.5	3.4
Hypertension	76.1	69.7
Hyperlipidemia	64.5	62.0
Diabetes mellitus	36.0	30.8
Prior stroke	4.2	4.3
Prior MI	27.4	29.6
PAD	7.5	7.2
CrCl <60 mL/min	15.4	13.3
Past tobacco use	33.4	32.2
Current tobacco use	25.9	27.5
Prior PCI	18.6	24.7
Prior CABG	11.8	12.0
Aspirin	90.9	97.5
≤100 mg	45.0	62.2
>100 mg	55.0	37.9
Intent to use GP IIb/IIIa inhibitor	30.1	19.6
Positive troponin or CK-MB	92.5	93.9
Inhospital procedures
Cardiac catheterization	84.8	88.5
PCI	30.8	61.7
Stenting	28.1	58.5
DES	54.8	56.9
BMS	48.5	47.5
CABG	22.7	8.3

Abbreviations: PAD, Peripheral artery disease; CrCl, creatinine clearance; GP, glycoprotein; CK-MB, creatine kinase–MB.
Data are presented as medians (IQRs) or percentages.

Table II. Concordance table indicating consistency of antiplatelet treatment
		Current week (patient/wk)
		Study drug	Clopidogrel	Study drug + clopidogrel	None	Total
Prior week (patient/wk)	Study drug	113,982	2	197	64	114,245
	Clopidogrel	2	170,740	98	1322	172,162
	Study drug + clopidogrel	1388	88	168,290	23	169,789
	None	113	191	12	109,240	109,556
	Total	115,485	171,021	168,597	110,649	565,752

In this table, the units are represented by each patient per each week of participation (ie, each patient is counted once for each week of participation). The table shows the concordance of treatment of each patient in each week compared with the week prior. "Diagonal cells" indicate that treatments have remained consistent with the week prior.

Appendix

Marginal Structural Cox Proportional Hazards Model for the Outcomes

Let X_vc(t) = 1 if a patient is on both vorapaxar and clopidogrel at time t and X_vc(t) = 0 otherwise; X_v(t) = 1 if a patient is on vorapaxar but not on clopidogrel at time t and X_v(t) = 0 otherwise; X_c(t) = 1 if a patient is on clopidogrel but not on vorapaxar at time t and X_c(t) = 0 otherwise. Let X(t) denote a patient's treatment at time t (vorapaxar/clopidogrel only) and X(t) denote treatment history through time t, and let V be the vector of baseline covariates. [An example of one possible X(t) for t = 365 days: treatment with vorapaxar only until day 85, then vorapaxar and clopidogrel until day 180, then vorapaxar only through day 365.] Then, for each X(t), we specify

where h_X(t |V) denotes the hazard of having an event at time t among patients with baseline covariates V in the original patient population, if all patients had, possibly contrary to the fact, followed treatment history X(t) through time t; h₀(t) denotes baseline hazard and bV is a term for the effect of baseline covariates.

The main question of interest, interaction between the study treatment and clopidogrel, was addressed by comparing (b_vc – b_c) and b_v.

Models for the weights

Let L(t) denote the history of time-dependent confounders through time t, including baseline covariates [L(0) = V]. In order to get a consistent estimate of treatment-related hazard ratios, we fit a time-dependent Cox model where at each time t the contribution of a patient i in the risk set is weighted by sw_i(t) * sw _i(t), where:

Here p_1i is the probability of a patient receiving his or her actual treatment at time t, given baseline covariates and treatment history through time (t-1); p_2i is the probability of a patient receiving his actual treatment at time t, given baseline covariates, time-dependent covariates history through time t and treatment history through time (t-1). Inclusion of p_1i in the weights is generally not necessary, but it helps to "stabilize" weights (ie, to avoid extremely large weights given to patients with rare treatment changes). Using "stabilized" weights typically leads to narrower confidence intervals for HRs and better actual coverage rates for the confidence intervals.

Models for the Censoring Weights

Weights sw _i(t) are used to account for censoring, where:

Here C(t) = 1 if a patient is censored by time t and C(t) = 0 otherwise; q_1i is the probability that a patient remains uncensored at time t, given he was not censored by time (t-1) and given baseline covariates and treatment history through time (t-1); p_2i is the probability that a patient remains uncensored at time t, given he was not censored by time (t-1) and given baseline covariates, time-dependent covariates history through time t and treatment history through time (t-1). To be mathematically rigorous, C(-1) and X(-1) are defined to be 0.

Baseline Covariates (V)

The following baseline covariates were used in the analyses: type of stents used during PCI, if PCI was performed (≥1 DES or BMSs only); age >70 years; sex; weight; heart rate; race; hypertension; diabetes mellitus; prior angina; prior MI; prior arterial vascular disease; prior heart failure; tobacco use; prior PCI; prior coronary artery bypass graft; ST segment elevation; symmetric T-wave inversions; Killip class III or IV; clopidogrel use at baseline; statin use at baseline; or glycoprotein IIbIIIa inhibitor use.

Time-dependent Confounders (L(T))

The following time-dependent confounders were used in the analyses: surgical or invasive procedure; non-index PCI; CABG; bleeding within the previous month defined by the investigator as GUSTO moderate or higher, or TIMI major, or TIMI requiring medical attention but not higher (i.e., not TIMI major); unstable angina in the previous week; and MI within the previous month.