Duration of antiplatelet therapy after DES implantation: can we trust non-inferiority open-label trials?

This editorial refers to ‘Three, six, or twelve months of dual antiplatelet therapy after DES implantation in patients with or without acute coronary syndromes: an individual patient data pairwise and network meta-analysis of six randomized trials and 11 473 patients’†, by T. Palmerini et al., on page 1034. The optimal duration of dual-antiplatelet therapy (DAPT) following stenting remains a topic of major clinical importance and a source of ongoing debate, given the inherent trade-offs of continued protection against thrombotic events which must be balanced against the prolonged exposure to a higher bleeding risk. Many clinical trials have explored the issue, some comparing the standard 12-month DAPT duration with shorter treatment, others investigating longer treatment durations. The results of the trials have been mixed and have generated some confusion among clinicians. The confusion is related to the fact that trials evaluating shorter DAPT duration than 12 months (most of which were relatively small, open-label non-inferiority trials) generally claimed non-inferiority in the prevention of ischaemic and thrombotic events when compared with 12 months DAPT, whereas two large double-blind superiority trials studying DAPT durations longer than 12 months after stenting or myocardial infarction (the DAPT and PEGASUS trials1 , 2) found a reduction in ischaemic events with prolonged DAPT, albeit at the expense of increased bleeding. In this issue of the journal, Palmerini et al. report the results of an individual patient-level meta-analysis of randomized trials investigating the safety and efficacy of shortening DAPT to < 1 year after drug-eluting stent (DES) implantation.3 They pooled data from six of the seven trials which tested DAPT durations <1 year after DES placement, representing 11 473 patients, and found that while there was no overall difference in rates of myocardial infarction (MI) or definite/probable stent thrombosis when comparing short-term (3 or 6 months) DAPT with 1-year DAPT, there were important additional findings: there was a borderline interaction between DAPT duration and clinical presentation; patients with acute coronary syndrome (ACS) appeared to derive potential harm from shorter DAPT (3 or 6 months) whereas ischaemic event rates were similar regardless of DAPT duration in stable patients. In addition, among ACS patients, harm appeared potentially greater in patients treated with only 3 months DAPT. Conversely, as expected, shorter DAPT durations resulted in lower bleeding rates, irrespective of clinical presentation. These important observations are consistent with the findings of the prior meta-analysis of Udell et al. indicating the benefit of prolonged DAPT after MI4 as well as with the observation in the DAPT trial that the benefits of prolonged DAPT after DES implantation were greater in patients presenting with MI than in those undergoing elective stenting.5 Meta-analyses have long been viewed as providing one of the highest levels of evidence. However, more recently, in light of the increasing number of meta-analyses and systematic reviews, sometimes even more frequent than individual trials performed, concerns have been voiced regarding their quality.6 The present meta-analysis used solid methods and, importantly, relied on individual patient-level data. As such, it provides a robust contribution to the field, and its authors must be congratulated. An important limitation of the present analysis, which the authors have properly acknowledged but which must be emphasized, is that, while 41.5% of the total population was labelled as presenting with ACS, two-thirds of that group had biomarker-negative unstable angina. Therefore, MI was the index clinical presentation of only 14% of the total population. Unstable angina is, in the context of modern contemporary care using high-sensitivity troponin, a poorly defined entity with a low risk of subsequent ischaemic events7 and in those with this condition intensive or protracted antiplatelet therapy may be less beneficial than in patients with recent or prior MI.8 Therefore, the benefits of prolonged DAPT (or conversely the harms of shortened DAPT) may have been underestimated by studying such a low-risk ACS population. The authors adequately acknowledge the limitations of their data set and of the analysis, and appropriately conclude that ‘1-year mandatory DAPT would be a prudent minimum in high risk ACS patients’. However, a meta-analysis can only be as good as the quality of the trial data entered and, in that respect, one has to raise several concerns regarding the methods and designs of many of the DAPT duration trials. First, many of these trials were underpowered, increasing the risk of failure to distinguish relevant differences between treatment strategies adequately. In that respect, it is critical to remember that ‘lack of evidence of a difference does not establish evidence of lack of difference’. The trials had small planned sample size, but also sometimes enrolled fewer than the number of planned patients and were prematurely terminated for slow enrolment. Moreover, in some trials, the number of primary outcome events in both arms was very low, and the event rates were lower than expected, suggesting selection of very low-risk elective patients that may not be representative of routine clinical practice or at least that the trial findings may not be applicable to the usual case mix of many catheterization laboratories. It is noteworthy that while the majority used a control arm ‘anchored’ at 12 months (considered as the clinical standard of care), some used a different duration as control. Secondly, many of the shorter DAPT duration trials were non-inferiority trials and used lenient non-inferiority margins, sometimes based on absolute percentage increases which has been previously pointed out to risk accepting as non-inferior a treatment resulting in an substantial increase in the risk of an ischaemic event. This observation illustrates the point made by Pocock et al.9 that ‘when the noninferiority margin is a difference in percentages, it becomes easier (perhaps too easy) to achieve noninferiority if the overall event rate is lower than expected’ and underscores the value of selecting a relative rather than an absolute difference when deciding on a non-inferiority margin, since this is supposed to be the highest ‘acceptable’ difference between strategies to make them clinically interchangeable. (A) Non-inferiority margins, and actual vs. expected event rates in the reference arm of non-inferiority trials exploring short dual-antiplatelet therrapy (DAPT) duration following stenting. The observed event rates were generally much lower than the expected event rates, and the non-inferiority margins (generally based on absolute percentage increases) were high compared with the observed event rates. (B) Numbers of patients lost to follow-up, numbers of patients with crossover, and total number of primary outcome events in open-label randomized trials exploring short DAPT duration following stenting. The numbers of patients either lost to follow-up or with treatment crossover (which do not occur at random in open label trials) often vastly exceed the total number of primary outcome events. An important value of clinical trials, even when these are small or underpowered, is the ability to pool the data from all trials into meta-analyses, providing sufficient power to tease out robustly the benefits and harms of each strategy. Underpowered trials provide imprecise but ‘true’ estimates of treatment effect, and pooling several small trials can improve the precision in the assessment of treatment effect estimates. However, for open-label non-inferiority trials, there is a critical caveat: if there are high crossover rates or substantial numbers of patient lost to follow-up, this will result in biased estimates of treatment effect. Therefore, it is important to interpret conservatively meta-analyses of small randomized trials when they have these shortcomings, which are quite common: many of these trials enrolled fewer patients than planned, and had optimistic assumptions of event rates. Most of the non-inferiority trials had wide non-inferiority margins (often based on absolute percentage increases from the already optimistic assumptions of event rates) which vastly exceed what would be called ‘clinically interchangeable’ (Figure 1). These limitations, however, do not hamper pooling of the data into meta-analyses. More concerning are the frequent potential sources of bias such as large numbers of patients with cross-over in open-label trials, or high numbers of patients lost to follow-up (which can exceed the number of patients with outcome events) (Figure 1). The bias potentially introduced by these limitations is not reduced by pooling of the data and, regardless of how large the data set is, can produce misleading results. These limitations are important to keep in mind when reading or interpreting meta-analyses, even if these are as carefully conducted as the present one. Given these considerations, how should clinicians interpret the observations of Palmerini et al.? A conservative interpretation is consistent with clinical common sense,12 , 13 ESC guidelines,14 and with the cautious conclusions from the authors themselves: in the highest risk patients, particularly post-MI patients, 12 months DAPT is a prudent minimum. Conversely, in stable patients undergoing elective stenting, the risk of ischaemic event is low and there may be limited benefit to prolonged DAPT. In these patients, tools such as the DAPT risk score,15 which assists clinicians in evaluating the risk/benefit ratio of prolonged DAPT, may be useful to guide clinical decisions. Obviously, the best way to resolve the issue of whether a shorter DAPT duration is acceptable in patients undergoing elective stenting would be to conduct a very large, well-powered, double-blind randomized trial with a narrow non-inferiority margin in that patient population. Conflict of interest: P.G.S. reports grants from Merck, Servier, and Sanofi, personal fees from Amarin, AstraZeneca, Bayer, Boehringer-Ingelheim, Bristol-Myers-Squibb, CSL-Behring, Daiichi-Sankyo, GlaxoSmithKline, Janssen, Lilly, Merck Novartis, Pfizer, Regeneron, Sanofi, Servier, and The Medicines Company, outside the submitted work; and was involved in the leadership of two of the trials mentioned in the discussion (DAPT and OPTIDUAL). T.S. reports grants from Astrazeneca, Daiichi-Sankyo, Eli-Lilly, GSK, MSD, Novartis, and Pfizer, and personal fees from Board membership, consultancy, or lectures from AstraZeneca, Astellas, Bayer, BMS, Lilly, MSD, and Sanofi, outside the submitted work.