Pablo Lamelas –
One of the top surprises from ESC congress 2019 was ISAR-REACT 5 trial, which concluded that in acute coronary syndromes Prasugrel was superior to Ticagrelor for efficacy without paying the price of excess bleeding. Ticagrelor is the most popular antiplatelet accompanying aspirin in acute coronary syndromes: should we change practice?
ISAR-REACT 5 trial summary
Was an investigator-initiated, multicenter, randomized, open-label trial, that randomly assigned 4,018 patients with acute coronary syndromes (with or without ST-segment elevation) and invasive evaluation planned to receive ticagrelor or prasugrel. The primary end point was death, myocardial infarction, or stroke at 1 year, while major bleeding major safety outcome. The primary-end point event occurred in 9.3% in ticagrelor and 6.9% in the prasugrel group (HR 1.36; 95% CI 1.09-1.70; p=0.006). The individual components of the primary end point in the ticagrelor group and the prasugrel group were: death, 4.5% vs 3.7%; myocardial infarction 4.8% vs 3.0%; and stroke, 1.1% vs 1.0%. Definite or probable stent thrombosis was 1.3% in ticagrelor and 1.0% in prasugrel. Major BARC bleeding was 5.4% in ticagrelor and 4.8% in prasugrel group (HR 1.12; 95% CI 0.83-1.51; p=0.46). Authors concluded that in patients with acute coronary syndromes, the incidence of death, myocardial infarction, or stroke was significantly lower among those who received prasugrel than among those who received ticagrelor, and the incidence of major bleeding was not significantly different between the two groups.
Two-tailed test hypothesis
This study is a great example why studies should consider two-tailed hypothesis, especially in trials with an active comparator. Authors powered the study to show superiority of Ticagrelor when actually ended being inferior for efficacy. If authors specified a single-tailed superiority analysis (2.5% type I error risk) study power (sample size) calculations would have been the same but, strictly talking, cannot claim that prasugrel was superior at the end of the study.
But do we care about this when applying evidence-based medicine? Not much in my opinion. Evidence-based medicine means to integrate the best available evidence to guide clinical decisions. The actual data of this evidence does not come from a single study, but a critically appraised summary of the body of the evidence (all the evidence available about a specific topic). Then, when summarizing the evidence we don’t really care if the primary study used a single-tailed, two-tailed, 5%, 2.5% type I error, but we care about the point estimate and its precision (95% CI). Then, if this was (hypothetically talking) the only study comparing these two drugs directly in a randomized trial, Ticagrelor would look harmful vs Prasugrel, regardless of primary study plan.
Primary composite outcome
All outcomes included in the primary composite are patient-important with similar importance to patients, move in the similar direction but in different magnitudes (mainly driven by myocardial infarction, a mirroring trend in death but “identical” numbers in stroke).
Should stroke be on the primary outcome of this specific trial? Is dual antiplatelet therapy the standard treatment after acute coronary syndromes because they prevent stroke? What is the rationale that Ticagrelor or Prasugrel will be superior to the other in strokes when there is no indication for preventing that outcome of either drug?
Adding outcomes in which the interventions are not supposed to have a difference (unnecessary pooling) reduces study power and does not help in interpretation of the composite outcome result. For instance, in the complete vs culprit-only revascularization after STEMI, the COMPARE-ACUTE trial did include stroke in the primary outcome when the hypothesis was that complete revascularization is better: whats the rationale that doing more coronary angioplasty will end up in lower strokes? If there is any difference, will be on the other direction: more PCI, more risk of stroke, and mixing outcomes expected to go in different directions do not help when interpreting it.
Sometimes I wonder about a “clinical trial inertia” that makes us including stroke as “MACE” for every trial rather than designing the trial based on what is supposed to work on.
Is important to highlight that this was not just a head-to-head comparison of Ticagrelor vs Prasugrel, since interventions were delivered differently. Ticagrelor was delivered as PLATO, allowing pre-treatment in NSTEACS. Prasugrel was delivered as TRITON, which loading dose was given once anatomy was known in NSTEACS. Then, the observed differences can be partially explained by the way the drug was delivered, rather than just the drug effect.
Against a significant impact of timing for loading dose in the results, 46% of the population were STEMI, in which Ticagrelor and Prasugrel loading doses were administered as soon as possible after randomization. STEMI patients also belong to the highest risk subgroup, accounting of more than half of both ischemic and bleeding events.
Lower ischemic events and no higher bleeding with Prasugrel
In the antithrombotic literature we should always be skeptic when we see a signal in lower thrombotic events and no increase in bleeding. Should we be concerned about this trial then?
Well, we know from other scenarios that this scenario is possible (gain something without paying the price of the other side of the rope). For instance, in patients receiving oral anticoagulation with anti-vitamin K, having a high INR (ie, >5) is associated with exponential risk of bleeding, while no much gain in reduction of ischemic events vs normal INR (2 to 3). Novel oral anticoagulants like Dabigatran 150 mg BID demonstrated to be superior to warfarin in preventing ischemic stroke and similar major bleeding to anti-vitamin K.
On the other hand, this trial was powered for ischemic events, which happened with more frequency than major bleeding events. Then, precision of major bleeding estimates was not great (95% CI 0.83-1.51). Does this exclude a clinically significant bleed increase by Prasugrel? What is a clinically significant difference in bleeding? Tough questions, and depends on many factors, among them: Severity and frequency of major bleeding for patient prognosis and quality of life, also depends if Prasugrel is really superior to Ticagrelor and its magnitude, costs, etc.
In my opinion, we may probably say about bleeding is: this study was unable to detect statistically significant differences major bleeding events (cannot rule out smaller, clinically significant or not given to underpower for smaller differences) in favor or against Prasugrel, while large (clinically important) differences (in favor or against) are unlikely.
A big lesson learned
Indirectness and network meta-analysis
Ticagrelor in PLATO suggested superiority to Clopidogrel, including cardiovascular mortality. Prasugrel in TRITON had positive results vs Clopidogrel (same comparator), but not as impressive as PLATO, including signal of harm if prior stroke or low body weight. Then, applying common sense with some degree of indirectness, Ticagrelor became the antiplatelet of choice in ACS, also supported by network meta-analysis suggesting that Ticagrelor was the “most likely winner” in post-ACS patients.
Now we have a large, high-quality randomized trial showing exactly the opposite. One may hypothesize if different aspirin doses in PLATO, or pre-treatment not allowed in TRITON for NSTEACS and other methodolical aspects may have had a role in this paradoxical result. However, even a long list of limitations is unlikely to revert from benefit (the expected relative effect of Ticagrelor over Prasugrel) to harm (observed) in a high-quality randomized trial. If adherence (secondary to adverse effects specific of Ticagrelor, costs and two daily doses) explain part of the observed difference, then that goes even more in favor of Prasugrel, since real life adherence much lower than the observed in a controlled RCT.
Regardless, ISAR-REACT 5 meets the highest quality evidence describing the direct relative efficacy and safety of both interventions. Then, until new similar randomized trials come to light, we should prioritize this trial rather than network meta-analysis and ¨logic¨ which are limited with indirectness of the comparison in question.