No free lunches: balancing bleeding and efficacy with ticagrelor.

The development of new oral antagonists of the platelet P2Y12 receptor has generated both anticipation and anxiety among practising physicians. In a field in which clopidogrel was the only choice available for the better part of a decade, prasugrel and ticagrelor represent new treatment options in many parts of the world. Weighing efficacy, safety, and cost to determine appropriate use of these three agents will represent a great challenge over the next few years for physicians, as well as individuals and organizations involved in formulary decisions. To make appropriate decisions, both for the individual patient and for the larger healthcare system, in-depth understanding of safety data is absolutely critical. Thus, the comprehensive report from Becker et al. describing bleeding complications in the PLATO trial is particularly timely and important. The primary bleeding definition in PLATO combined noncoronary artery bypass graft (CABG) and CABG-related events, which differs notably from other contemporary antiplatelet trials that have used non-CABG bleeding as the primary safety endpoint. While we do not dispute the importance of CABG bleeding, we believe a focus on total bleeding in the PLATO trial may provide a misleading assessment of the safety of ticagrelor relative to clopidogrel. A strong argument can be made that the CABG bleeding definition in PLATO was too soft: any transfusion of ≥ 4 units of red blood cells was defined as a major bleed, regardless of clinical consequences. In essence, usual operating room procedures related to transfusion of blood products met the definition of a major bleed. The result was that a huge proportion of patients who underwent CABG met the study definition of a major bleed, with rates . 80% in each treatment arm. By combining soft CABG ‘bleeds’ with more rigorously characterized non-CABG bleeding events, important differences in bleeding were diluted by the high number of CABG-related transfusions. Indeed, although only 10% of the study population underwent CABG, more than two-thirds of the major bleeding events in the PLATO trial were CABG-related. Therefore, non-CABG bleeds, which would be less influenced by transfusion practice and more likely to represent events of clinical importance to the patient, should remain the primary focus for safety comparisons between the two agents. Multiple key points emerge from this comprehensive analysis of the bleeding profile of ticagrelor. (i) Study drug discontinuation due to non-procedural bleeding occurred .2-fold more frequently with ticagrelor than with clopidogrel. (ii) Although no differences were evident regarding fatal bleeding, rates of non-CABG major bleeding were higher with ticagrelor than clopidogrel using both the PLATO definition [hazard ratio (HR) 1.19, 95% confidence interval (CI) 1.02–1.38] and the TIMI definition (HR 1.25, 95% CI 1.03–1.53). (iii) The composite of major + minor non-CABG bleeding, which is a highly relevant outcome since ‘minor’ bleeds often cause considerable suffering for patients and increase length of stay and costs for the healthcare system, occurred in 8.7% and 7.0% in the two groups (HR 1.27, 95% CI 1.14–1.42). (iv) The excess bleeding appeared to be restricted to spontaneous bleeding events (HR 1.31, 95% CI 1.08–1.60), with no differences reported previously in multiple parameters of procedural bleeding related to CABG and percutaneous coronary intervention (PCI). With prasugrel, although PCI-related bleeding rates were also similar to those with clopidogrel, CABG-related bleeding was increased; moreover, rates of fatal bleeding were slightly increased with prasugrel, which was not seen in these analyses of ticagrelor. (v) The excess spontaneous bleeding events were largely due to gastrointestinal (GI) bleeding and epistaxis, with a small excess of intracranial bleeds also noted (0.34% vs. 0.19%) (Figure 1). (vi) Although the non-CABG bleeding curves appeared to diverge slightly after 30 days of treatment, an excess of major and minor bleeding was still evident within 30 days for patients receiving ticagrelor. (vii) Excess bleeding with ticagrelor was not restricted to individuals predicted to be at high bleeding risk— no treatment interactions were observed in subgroups defined by clinical characteristic or using the CRUSADE bleeding model. Thesefindingshave interesting implications for interpretationof the mortality reduction seen with ticagrelor in PLATO. In the primary manuscript, the authors speculated that ‘the improved survival rate