CardioExchange Archive

PLATO is one of the most impressive trials in recent years, demonstrating substantial benefits for ticagrelor over clopidogrel in a wide population of ACS patients. However, patients enrolled in the US showed no benefit from ticagrelor, and experts have been unable to agree on a cause. Possible factors could include much higher doses of aspirin used in the US versus the rest of the world, differences in baseline characteristics (the US population was heavier and had more diabetes), and different patterns of invasive procedures. It is also impossible to exclude the play of chance. At the recent FDA advisory panel which recommended US approval of ticagrelor, one FDA official summed up the dilemma: “this drug doesn’t do what it’s supposed to do in the country in which it’s being asked to get approved.”

CardioExchange asked several panel members and other experts to respond to 3 questions about the panel. Our series starts with the chairman of the ticagrelor panel, Sanjay Kaul, who voted for approval but urged caution in the drug’s usage and recommended that the FDA require the company to perform additional trials in the US. Tomorrow we will post answers from Mori Krantz, the one panel member who voted against ticagrelor approval. Other installments are from Jonathan Halperin, another panel member, and Robert Harrington, who presented PLATO data to the panel on behalf of the sponsor.

1. What is your interpretation of the finding in the subgroup of US patients?

The reasons for US vs. OUS (outside US) disparity in efficacy outcomes are not readily discernible from the trial data. Potential explanations include the play of chance; aspirin dose interaction with ticagrelor treatment effect; and the “process of care” differences.

Much has been made of the interaction term that was statistically significant: P value of 0.045 for the US vs. overall cohort, or P value of 0.01 for the US vs the OUS cohort. In my opinion, from a statistical perspective, the US vs. OUS is not a proper subgroup comparison because:

1. It was not prespecified.

2. There is no biologically plausible reason for such an interaction.

3. The qualitative nature of the interaction (difference in direction with evidence of benefit in OUS but harm in US) makes it unlikely to be reliable or replicable. And,

4. It was not adjusted for multiple comparisons (when adjusted for 31 prespecified comparisons, the P values are no longer significant).

Thus, the estimate of treatment effect in the overall cohort is the most reliable estimate of the treatment effect in any subgroup, including the US. Finally, the bleeding data did not track with the efficacy data, i.e., there were no differences in the hazard for bleeding between the US (risk ratio of 1.09 for total or 1.24 for non-procedural major or minor bleeding) and the OUS cohort (risk ratio of 1.10 for total or 1.38 for non-procedural major or minor bleeding).

The validity of ASA dose-efficacy outcome interaction would have been strengthened if the pharmacokinetic (PK), pharmacodynamic (PD) or bleeding data were congruent. There were no pharmacokinetic or pharmacodynamic interactions that would predict an undesired effect at high aspirin doses. Similarly, no difference in bleeding was apparent between high-dose aspirin (risk ratio of 1.06 for total major or minor bleeding) vs. low-dose aspirin (risk ratio of 1.03 for total major or minor bleeding). Unless the interaction is mediated by a non-platelet effect, one would have expected reduced bleeding with high-dose aspirin as a result of reduced antiplatelet efficacy of ticagrelor.

Thus, if the US vs. OUS difference is real, it is likely related to differences in the “process of care” issues, and the aspirin dose might simply be a marker of these differences. However, this is at present a conjecture, albeit a plausible one, that clearly warrants confirmation in prospective assessment.

2. How should this result influence the guidelines?

There is no doubt that the guideline committees will be keen (if not under pressure) to incorporate the findings from PLATO in their latest recommendations. However, given the cloud of doubt (justified or not) whether the drug does really work within the US practice environment, I would urge caution. In my opinion, guidelines are unlikely to be integrated into medical practice if the clinicians feel that the evidentiary support for the recommendations is not generalizable to their practice environment. Regardless of the lack of a statistically or biologically plausible explanation for the US vs. OUS treatment disparity, it is hard to conceive that this will not have any impact on the clinical use of this drug. Accordingly, it might be worthwhile to wait until we have a better idea regarding its efficacy and safety in the US practice environment.

I think it would be in the best interest of the sponsor to sort out this issue ASAP, i.e., conduct another study in ACS patients with at least 50% patients enrolled from within the US. A potential study design would be a 2×2 factorial design (ticagrelor vs. clopidogrel as the first factorial, and low-dose ASA vs. high-dose ASA as the second factorial) to sort out both the US vs. OUS and the low-dose vs. high-dose ASA issues. The trial can be designed to rule out a clinically unacceptable harm with ticagrelor (for example 30% harm or an equivalent hazard ratio of 1.3). Such a trial can be conducted with a sample size of less than 5000. Piggybacking on top of the chronic stable angina trial, as proposed by the sponsor, is unlikely to address potentially relevant issues that might best explain the disparities, such as difference in “process of care” for ACS. Some will justifiably question whether such a trial would be feasible post-approval. In fact, some in the FDA have already deemed it “awkward” and even “unethical,” given the cardiovascular mortality advantage with ticagrelor, while others in the FDA and half of the advisory panel members expressed support for this trial. I think the trial can be designed without major ethical concerns, but I am also aware of the challenges, especially if the drug ends up getting a “broad” claim by the FDA.

3. If approved now, whom would you treat with this drug… in the United States? and why?

Recommended use of ticagrelor can be limited only to the ACS setting alone at the present time.

1. Patients with moderate- to high-risk NSTEMI ACS who are managed invasively with a PCI exhibit the most optimal benefit-risk profile in favor of ticagrelor, followed closely by the same patients managed medically.

2. Although, patients with STEMI ACS managed invasively also benefit overall, there does not appear to be any treatment benefit in the first 30 days, a finding which is in contrast with the prasugrel benefit observed in TRITON. This finding tempers my enthusiasm for this drug as the treatment of choice during primary PCI, where early benefit is desirable. There were too few STEMI cases managed medically in PLATO (less than 2.5% of the overall cohort) to draw clinically meaningful inferences.

3. In contrast, patients with unstable angina alone or low-risk (TIMI risk score, 0-2) or troponin-negative NSTEMI ACS managed medically have the least favorable benefit-risk profile.

With respect to the US cohort, there were very few primary endpoint events in patients with unstable angina (17/148 events) or STEMI (24/148 events) as index events. In the NSTEMI subgroup (107/148 events), compared with those managed with PCI (where the outcomes favored clopidogrel, HR of 1.46), those managed medically did not appear to be harmed with ticagrelor (HR of 1.01).

It is interesting to note that while one half of the total number of primary endpoint events in PLATO occurred within the first 30 days, only one third of the ticagrelor benefit was evident during this time. The lack of early treatment effect is inconsistent with other P2Y12 antagonists such as prasugrel or clopidogrel. Moreover, the drug shows a delayed mortality benefit, which no other P2Y12 antagonist has shown. These observations raise questions about the mechanism behind the beneficial effects of ticagrelor and suggest the tantalizing possibility that they may be mediated, in part, via a non-platelet effect.

The rapid “onset” and “offset” of ticagrelor’s antiplatelet effect observed in the laboratory setting would make it desirable for use in patients with NSTEMI ACS with unknown coronary anatomy or in whom urgent CABG or any other surgery is not deferrable. Unfortunately, this is not borne out by outcome data in PLATO as no early or short-term advantage in clinical efficacy or safety was apparent. In fact, there was increased bleeding requiring treatment cessation at least 5 days prior to CABG or surgery, which is not much different from the 7 days required for clopidogrel. This then raises questions regarding the predictive value of PK and PD (platelet function tests) evaluation in informing treatment decisions. In other words, do these tests have clinical utility?

Although the reported adverse events are clinically manageable, I would, nevertheless, urge caution (but not avoid ticagrelor) in patients with COPD, severe chronic kidney disease, advanced bradyarrhythmias (without pacemaker support), and history of severe gout or hepatic dysfunction.

I would avoid ticagrelor in patients at high– risk for bleeding, especially intracranial hemorrhage, such as those with history of stroke or TIA, even though these variables were not predictive of intracranial hemorrhage in PLATO. The numerical imbalance in fatal intracranial hemorrhage (11 vs. 2) is a concern, even though there was a counterbalancing imbalance in fatal extracranial bleeding (9 vs. 21). Some have raised concerns regarding noncompliance leading to increased thrombotic risk (related to rapid “offset”). As mentioned above, I am not convinced if there is more to this other than laboratory phenomenology (unless, of course the bleeding effect is dissociated from the antithrombotic effect)!