Potent antiplatelets in comparison to clopidogrel and each other

Olier et al present a comparative analysis of all 89 067 primary PCI procedures in England and Wales between 2007 and 2014 where the antiplatelet regimen was known. The mortality outcomes are extracted from a well-validated national database while bleeding data are less rigorous as it is self-reported by the PCI institutions. Crude mortality rates at 30 days and 1 year were significantly lower in patients taking either prasugrel or ticagrelor compared with clopidogrel. As there are marked baseline differences between the arms, statistical adjustment is necessary. Logistic regression modelling suggested that while prasugrel was associated with a significantly lower chance of death at 30 days and 1 year compared with clopidogrel, there was no difference in major adverse cardiovascular events  (MACE) rates. Somewhat surprisingly, ticagrelor did not demonstrate a significant reduction in mortality compared with clopidogrel at either 30 days or 1 year. Furthermore, ticagrelor had a higher rate of in-hospital MACE compared with clopidogrel. Interestingly, when ticagrelor was compared with prasugrel, a higher rate death rate was observed at both time points, as a well as higher in-hospital MACE rate.

From this immense data set, we can infer that the newer potent agents are superior to clopidogrel. The data also imply that prasugrel offers advantages over ticagrelor, and that the benefits of ticagrelor are smaller than observed in the pivotal PLATO study.

Novel and challenging: this study in relation to prior studies

The data are novel and challenging. How do we reconcile the findings in light of prior outcome studies in which ticagrelor demonstrated significant mortality and MACE benefit over clopidogrel?4 Similarly, why is prasugrel effective in this data set when no significant benefit was observed in patients with STEMI undergoing primary PCI in the TRITON-TIMI-38 study?5 More philosophically, were primary angioplasty centres right to make a widespread switch to either prasugrel or ticagrelor in patients with STEMI based only on subgroup analyses? Should we change from ticagrelor to prasugrel now? How do these real-world data compare with other real-world data sets?

First, we must applaud the authors and the thousands of physicians and allied health professionals who complete the BCIS data set. The sheer size of the data overwhelms the prior pivotal studies and may dwarf criticisms of its retrospective observational nature. While it lacks randomisation and blinding, a larger study of invasively managed patients with STEMI is unlikely and the data are directly applicable to the UK population. However, there are limitations.

Observational databases may have inaccuracies and here, the data are collected at the height of emergency. Not all patient factors are known at the time of STEMI and could impact baseline characteristics. For example, the rates of ‘renal disease’ are sub-1% in each arm and this is likely an underestimation created by defining ‘renal disease’ as those needing dialysis. Frailty and body mass are also not readily captured but it is known these predict death, MACE and bleeding risk. The marked baseline differences between the study arms imply a systematic bias in prescribing.

Patients on clopidogrel were more likely to be older with higher risk features suggesting confounding by indication. Although at greatest risk of death during STEMI, the presence of multiple comorbidities and higher bleeding risks mean clinicians are pushed away from using prasugrel or ticagrelor. Such patients are not typically recruited into randomised studies or specifically excluded by trial criteria. Their inclusion in real-world data is undoubtedly helpful but can artificially push the data in favour of the newer antiplatelets.

Prasugrel prescribing is constrained to avoid patients at highest risk of bleeding in the TRITON study (avoid in patients over 75 years, those with prior stroke or transient ischaemic attack, those with prior bleeding and those in whom surgical revascularisation may be preferable). In each of these high-risk cases, the patient will have received clopidogrel, potentially increased the event rate in the clopidogrel arm while removing a high-risk event from the prasugrel arm. Although this bias in selecting antiplatelet is appropriate clinical behaviour, such confounding by indication introduces a bias that cannot always be corrected for.

Ticagrelor as a twice a day medication may have higher event rates in real-world populations as drug adherence is lower than in motivated and closely monitored trial populations. Moderate discontinuation rates of ticagrelor secondary to dyspnoea are recognised in trial patients and higher rates in the real world may remove its apparent benefit. The drug can only be useful if patients are taking it.

Further issues revolve around the time period being assessed. Not only are there marked regional variations seen in antiplatelet use suggesting policy and economic differences between centres, but the more potent antiplatelets are used towards the end of the time period when many other clinical behaviours will have also changed. Since primary PCI has altered significantly over time, the study naturally is a comparison of practice and approach to STEMI across time. The radial access revolution means the current UK rates are 80%–85% while this study has 47% in the clopidogrel arm a reflection of the fact that the majority of clopidogrel patients are from 2011 and earlier. Similarly, bare metal stent rates are highest in the clopidogrel arm.

Finally, self-reporting of bleeding and inpatient events is a significant limitation and caution is needed in interpretation. The bleeding rates are low (although inpatient only) and notably less than the pivotal studies and may reflect under-reporting.

Clopidogrel, prasugrel or ticagrelor?

It remains a challenge to know when it is right to change clinical practice in response to new randomised data. The data by Olier et al should give pause to those still using clopidogrel as their default drug in STEMI. The move from prasugrel to ticagrelor is less clear-cut and the data presented here are provocative. Should centres using ticagrelor move back to prasugrel for all STEMI that meet its prescribing restrictions? This choice is brought into sharp focus in the knowledge that prasugrel will soon be off-patent.

This question is not fully solved by the present data set, which can only generate hypotheses. Future studies are awaited, such as the ISAR-REACT 5 study (‘Prospective, Randomized Trial of Ticagrelor Versus Prasugrel in Patients With Acute Coronary Syndrome’) which will randomise 4000 patients in Europe to ticagrelor or prasugrel but it will include all ACS rather than STEMI alone, and is still recruiting.

Finally, it may be instructive to consider the Swedish experience of ticagrelor which was compared with clopidogrel in 45 073 patients with ACS within the SwedeHeart Registry.8 In that well-defined cohort with noted high data accuracy, the risk of death, myocardial infarction or stroke was halved by ticagrelor at 24 months without an increase in bleeding-related admissions. However, STEMI only accounted for a third of patients and prasugrel data were not reported. As such, Olier’s work remains the largest data set in this domain. Perhaps analysis of the 2014–2018 cohort will add clarity.

Going forward, the BCIS data set represents a fantastic resource for future analyses and growth in its data quality will support real-world analyses. The Olier paper demonstrates the real-world data can elucidate novel findings that complement randomised studies and push us towards better patient care.