Vorapaxar and Clopidogrel After Non-ST ACS
Vorapaxar and Clopidogrel After Non-ST ACS
In this analysis, we compared the efficacy and the safety of vorapaxar for NSTE ACS patients with and without concomitant clopidogrel use. The use of clopidogrel was common in TRACER, and treatment was continued chronically, with a decrease observed at 1 year. The subgroup analyses using baseline clopidogrel use suggested a better efficacy and safety profile of vorapaxar among patients who were not on clopidogrel. However, baseline use of clopidogrel only modestly reflects clopidogrel treatment during the study. We used the MSM method to account for ongoing clopidogrel use during follow-up and to adjust for time-dependent factors that may affect clinical outcomes and the decision to start or discontinue clopidogrel treatment. When we evaluated clopidogrel use over time, there was no evidence of an interaction between vorapaxar and clopidogrel on efficacy and safety outcomes. One caveat of these analyses is that when all patients are on clopidogrel (ie, after stent placement), it is not possible to assess for an interaction; therefore, our results are not applicable to the first months after an ACS among patients who were treated with a coronary stent.
The TRACER trial was designed to test the hypothesis that vorapaxar, in addition to standard of care, would reduce recurrent ischemic events, with the expectation that most patients would be treated with a combination of aspirin and clopidogrel. In fact, >90% of patients received clopidogrel during the index ACS hospitalization. After discharge, there were differences in clopidogrel use and treatment duration according to whether a stent was received and the type of stent, but most stented patients went on to receive clopidogrel for ≥1 year.
The main results of TRACER indicated an overall modest effect of vorapaxar, with a nonsignificant reduction of the primary end point and a nominally significant reduction of the key secondary end point. In terms of safety, there was a significant increase in major bleeding with vorapaxar. In other trials with triple therapy, similar bleeding results were observed. Therefore, we wanted to assess whether there was an interaction between use of clopidogrel and vorapaxar. The subgroup analysis of clopidogrel use before randomization suggested a larger estimated effect size in patients who were not on clopidogrel compared with patients who were on clopidogrel, with a trend for statistical interaction. These results were paralleled by trends possibly suggesting less of an incremental bleeding increase in patients who were not on clopidogrel at randomization. Nonetheless, as we have shown, the use of medication before randomization only modestly reflects subsequent use. In particular, nearly 50% of patients who were not on clopidogrel at baseline started clopidogrel, most of them during the index hospitalization. This makes the results using baseline clopidogrel difficult to interpret. It may be that other factors related to the use of clopidogrel at baseline (baseline characteristics or treatment strategies) may explain the association, or it could be a spurious finding. Because patterns of clopidogrel use may be complex (with patients starting, stopping, and restarting), to properly address the interaction with clopidogrel, one has to (1) account for whether a patient was on clopidogrel and whether clopidogrel use was continuous for the duration of the follow-up period, and (2) determine whether events occurred while the patient was on clopidogrel. The association between chronic antiplatelet use and outcomes may be confounded by several factors, including baseline and time-dependent variables (such as procedures and bleeding events).
Using the MSM method, we evaluated associations among exposure to vorapaxar, clopidogrel use, and outcomes with analyses assessing the entire period in which the patient was followed up and accounting for known potential baseline and time-dependent confounders. Applying this method, we observed no evidence of interaction between clopidogrel use and efficacy of vorapaxar, and point estimates for vorapaxar effect were consistent with and without concomitant clopidogrel. Similarly, there was increased bleeding with vorapaxar regardless of concomitant clopidogrel. It is important to highlight that, with the interaction, one only assesses whether the effect of vorapaxar was modified by clopidogrel use and not absolute risk of bleeding when vorapaxar is used with or without clopidogrel. It is possible that absolute bleeding risk deriving from a combination of 3 agents may become clinically unacceptable. Furthermore, we reiterate that these results do not apply to the early periods: 6 months after placement of a DES; 1 month after placement of a BMS; and the in-hospital period for the remaining patients, as the interaction cannot be assessed when all or nearly all patients are on clopidogrel. Although these cutoffs are arbitrary, sensitivity analyses were performed after more extended duration of clopidogrel treatment, and results did not change. In the TRA 2P-TIMI 50 trial, comparing vorapaxar versus placebo in patients treated for secondary prevention of atherosclerotic disease, patients were stratified according to the intent to administer a thienopyridine. Because patients were randomized ≥2 weeks after an acute MI, the intention was therefore captured after definitive treatment strategies were established. In this context and similar to our result, the trial found no interaction between use of thienopyridine and vorapaxar efficacy and safety.
Overall, the phase III experience with vorapaxar suggests that vorapaxar and P2Y12 inhibition do not have a clinically relevant interaction, and thus, they likely act independently. It remains uncertain in patients with ACS whether PAR-1 antagonism alone would provide an advantage compared with P2Y12 inhibition alone or whether use of PAR-1 antagonism could be beneficial without aspirin, with or without P2Y12 inhibition. Those are important questions that give opportunity to further study PAR-1 antagonists to improve treatment of ACS patients.
This is an exploratory analysis attempting to address a complex issue of interaction between time-varying chronic antiplatelet therapies and, as such, has several limitations. First, it is based on postrandomization variables, which may be affected by numerous confounders. We have relied on a complex statistical method, which we believe adequately accounts for known baseline and time-dependent confounders. Although we believe the MSM method allowed us to address those complexities in a more comprehensive way than conventionally used statistical models, the findings of this analysis rely on the method used. This method depends upon several technical assumptions that cannot be directly verified. We also cannot exclude the possibility of unmeasured or unknown confounders that influenced treatment decisions and outcomes. As already stated, during periods in which nearly all patients were on or were expected to be on clopidogrel, the interaction cannot be tested, as there is no sufficient variation in the use of clopidogrel. Therefore, our analysis is unable to assess whether an interaction may exist during the early phase after an ACS. The analysis periods used for patients with a DES are arbitrary but also reflect uncertainties and current practice for duration of dual antiplatelet therapy.
In spite of these limitations and the hypothesis-generating nature of our study, we think that there is value in this type of analysis if performed with adequate methods. Whether there is an interaction between concomitant antiplatelet therapies is often a key question in trials testing these agents and may potentially apply to different treatments. We believe that investigating the interaction based on baseline clopidogrel use may be potentially misleading, as it does not reflect subsequent use. Although randomized clinical trials remain the approach to provide definitive data, it is unrealistic to assume that adequately powered trials with factorial designs examining combination therapies will always be available. Therefore, analyses using adequate statistical methods may provide helpful insights and may also help guide the design of future studies.
It would also be desirable to establish consensus within the clinical, statistical, and medical journal community about the recommended or acceptable statistical methods to address issues of chronic concomitant therapies. That may also help plan and design these important secondary analyses during the early stages of trials as well as ensuring that key data are collected, and that quality is ensured during the conduct of the trial.
Discussion
In this analysis, we compared the efficacy and the safety of vorapaxar for NSTE ACS patients with and without concomitant clopidogrel use. The use of clopidogrel was common in TRACER, and treatment was continued chronically, with a decrease observed at 1 year. The subgroup analyses using baseline clopidogrel use suggested a better efficacy and safety profile of vorapaxar among patients who were not on clopidogrel. However, baseline use of clopidogrel only modestly reflects clopidogrel treatment during the study. We used the MSM method to account for ongoing clopidogrel use during follow-up and to adjust for time-dependent factors that may affect clinical outcomes and the decision to start or discontinue clopidogrel treatment. When we evaluated clopidogrel use over time, there was no evidence of an interaction between vorapaxar and clopidogrel on efficacy and safety outcomes. One caveat of these analyses is that when all patients are on clopidogrel (ie, after stent placement), it is not possible to assess for an interaction; therefore, our results are not applicable to the first months after an ACS among patients who were treated with a coronary stent.
The TRACER trial was designed to test the hypothesis that vorapaxar, in addition to standard of care, would reduce recurrent ischemic events, with the expectation that most patients would be treated with a combination of aspirin and clopidogrel. In fact, >90% of patients received clopidogrel during the index ACS hospitalization. After discharge, there were differences in clopidogrel use and treatment duration according to whether a stent was received and the type of stent, but most stented patients went on to receive clopidogrel for ≥1 year.
The main results of TRACER indicated an overall modest effect of vorapaxar, with a nonsignificant reduction of the primary end point and a nominally significant reduction of the key secondary end point. In terms of safety, there was a significant increase in major bleeding with vorapaxar. In other trials with triple therapy, similar bleeding results were observed. Therefore, we wanted to assess whether there was an interaction between use of clopidogrel and vorapaxar. The subgroup analysis of clopidogrel use before randomization suggested a larger estimated effect size in patients who were not on clopidogrel compared with patients who were on clopidogrel, with a trend for statistical interaction. These results were paralleled by trends possibly suggesting less of an incremental bleeding increase in patients who were not on clopidogrel at randomization. Nonetheless, as we have shown, the use of medication before randomization only modestly reflects subsequent use. In particular, nearly 50% of patients who were not on clopidogrel at baseline started clopidogrel, most of them during the index hospitalization. This makes the results using baseline clopidogrel difficult to interpret. It may be that other factors related to the use of clopidogrel at baseline (baseline characteristics or treatment strategies) may explain the association, or it could be a spurious finding. Because patterns of clopidogrel use may be complex (with patients starting, stopping, and restarting), to properly address the interaction with clopidogrel, one has to (1) account for whether a patient was on clopidogrel and whether clopidogrel use was continuous for the duration of the follow-up period, and (2) determine whether events occurred while the patient was on clopidogrel. The association between chronic antiplatelet use and outcomes may be confounded by several factors, including baseline and time-dependent variables (such as procedures and bleeding events).
Using the MSM method, we evaluated associations among exposure to vorapaxar, clopidogrel use, and outcomes with analyses assessing the entire period in which the patient was followed up and accounting for known potential baseline and time-dependent confounders. Applying this method, we observed no evidence of interaction between clopidogrel use and efficacy of vorapaxar, and point estimates for vorapaxar effect were consistent with and without concomitant clopidogrel. Similarly, there was increased bleeding with vorapaxar regardless of concomitant clopidogrel. It is important to highlight that, with the interaction, one only assesses whether the effect of vorapaxar was modified by clopidogrel use and not absolute risk of bleeding when vorapaxar is used with or without clopidogrel. It is possible that absolute bleeding risk deriving from a combination of 3 agents may become clinically unacceptable. Furthermore, we reiterate that these results do not apply to the early periods: 6 months after placement of a DES; 1 month after placement of a BMS; and the in-hospital period for the remaining patients, as the interaction cannot be assessed when all or nearly all patients are on clopidogrel. Although these cutoffs are arbitrary, sensitivity analyses were performed after more extended duration of clopidogrel treatment, and results did not change. In the TRA 2P-TIMI 50 trial, comparing vorapaxar versus placebo in patients treated for secondary prevention of atherosclerotic disease, patients were stratified according to the intent to administer a thienopyridine. Because patients were randomized ≥2 weeks after an acute MI, the intention was therefore captured after definitive treatment strategies were established. In this context and similar to our result, the trial found no interaction between use of thienopyridine and vorapaxar efficacy and safety.
Overall, the phase III experience with vorapaxar suggests that vorapaxar and P2Y12 inhibition do not have a clinically relevant interaction, and thus, they likely act independently. It remains uncertain in patients with ACS whether PAR-1 antagonism alone would provide an advantage compared with P2Y12 inhibition alone or whether use of PAR-1 antagonism could be beneficial without aspirin, with or without P2Y12 inhibition. Those are important questions that give opportunity to further study PAR-1 antagonists to improve treatment of ACS patients.
This is an exploratory analysis attempting to address a complex issue of interaction between time-varying chronic antiplatelet therapies and, as such, has several limitations. First, it is based on postrandomization variables, which may be affected by numerous confounders. We have relied on a complex statistical method, which we believe adequately accounts for known baseline and time-dependent confounders. Although we believe the MSM method allowed us to address those complexities in a more comprehensive way than conventionally used statistical models, the findings of this analysis rely on the method used. This method depends upon several technical assumptions that cannot be directly verified. We also cannot exclude the possibility of unmeasured or unknown confounders that influenced treatment decisions and outcomes. As already stated, during periods in which nearly all patients were on or were expected to be on clopidogrel, the interaction cannot be tested, as there is no sufficient variation in the use of clopidogrel. Therefore, our analysis is unable to assess whether an interaction may exist during the early phase after an ACS. The analysis periods used for patients with a DES are arbitrary but also reflect uncertainties and current practice for duration of dual antiplatelet therapy.
In spite of these limitations and the hypothesis-generating nature of our study, we think that there is value in this type of analysis if performed with adequate methods. Whether there is an interaction between concomitant antiplatelet therapies is often a key question in trials testing these agents and may potentially apply to different treatments. We believe that investigating the interaction based on baseline clopidogrel use may be potentially misleading, as it does not reflect subsequent use. Although randomized clinical trials remain the approach to provide definitive data, it is unrealistic to assume that adequately powered trials with factorial designs examining combination therapies will always be available. Therefore, analyses using adequate statistical methods may provide helpful insights and may also help guide the design of future studies.
It would also be desirable to establish consensus within the clinical, statistical, and medical journal community about the recommended or acceptable statistical methods to address issues of chronic concomitant therapies. That may also help plan and design these important secondary analyses during the early stages of trials as well as ensuring that key data are collected, and that quality is ensured during the conduct of the trial.
Source...