Optimizing dual antiplatelet therapy duration after myocardial infarction: evidence-based, precision, or personalized medicine?

This editorial refers to ‘Personalising the decision for prolonged dual antiplatelet therapy: development, validation and potential impact of prognostic models for cardiovascular events and bleeding in myocardial infarction survivors’†, by L. Pasea et al., on page 1048. The cardiovascular field is the medical discipline publishing the largest number of randomized clinical trials (RCTs) paving the way from experience-based to evidence-based clinical decision-making. RCTs represent the best approach to learn about the efficacy and safety of a particular therapeutic strategy. Indeed, in the era of ‘evidence-based medicine’, RCTs have been enthroned at the highest level of the hierarchy of what has been scientifically proven. Clinical practice guidelines advocate the use of this approach to support their recommendations. However, RCTs include only relatively selected patients, raising concerns on the generalizability of the results to routine clinical practice.1 Moreover, even for therapies with well-established clinical efficacy, the absolute number of patients that actually obtain clinical benefit is relatively small, and some patients may suffer from serious treatment-related adverse effects.1 , 2 Moreover, the clinical characteristics of cardiovascular patients are very heterogeneous, yet they play a key role in treatment effects. Therefore, the main challenge for clinicians remains to apply wisely the scientific evidence stemming from large-scale RCTs (namely the average group-level effect) to individual patients.1 , 2 In this regard, the use of prognostic variables able to identify likely responders is gaining widespread acceptance to move away from the ‘one-size-fits-all’ concept. A classical approach to address this conundrum has been the use of subgroup analyses from RCTs. However, subgroup analyses, even when the relevant variables are pre-specified or stratified and formal tests for interaction are performed, have major limitations, including the inherent multiplicity problem. Moreover, single variables displayed on classical forest plot subgroup analyses are not independent of each other.1 , 2 Accordingly, subgroup analyses should be considered just as hypothesis generating until confirmed by dedicated studies. Likewise, multiple risk scores have been devised to enhance prediction of benefits and harms using combinations of well-selected baseline characteristics.1 , 2 However, this methodology only provides relative treatment effects, thus representing a humble approach to ‘precision medicine’.1 , 2 Perception that clinical experience is sufficient and that the use of risk models leads to unacceptable omission of potentially beneficial therapy remains a barrier preventing their widespread clinical implementation. Ideally, however, patient-tailored treatment decisions, leading to selective treatment of patients for whom the greatest benefit and the least harm can be expected, should be pursued.1 , 2 Currently, multivariable prediction models may be derived from landmark RCTs by using multiple baseline clinical characteristics to provide patient-level estimates of the absolute risk reduction in cardiovascular events under the assigned therapy.1 The individual number needed to treat (1/absolute risk reduction) provides additional practical insights. Altogether these strategies advance knowledge and represent important new steps forward to personalized medicine.1 , 2 The optimal duration of DAPT after coronary stenting remains a controversial topic.3–9 This is particularly relevant considering the safer profile of new-generation drug-eluting stents. Multiple RCTs and several subsequent meta-analyses have recently focused on this issue, yet the optimal therapeutic strategy remains unsettled.4 , 7 , 8 Overall, evidence from these trials demonstrated that short (3–6 months), as compared with the standard 12-month, DAPT duration is associated with similar rates of ischaemic events but reduced risk of major bleeding.3–9 Currently 12-month DAPT is recommended on clinical grounds for all patients with acute coronary syndrome, thus the short strategy mainly remains an option for stable patients and those with a very high bleeding risk.6 On the other hand, an ‘extended’ (>12 months) duration of DAPT therapy (18–48 months) reduces the risk of ischaemic events [myocardial infarction (MI), stent thrombosis, and combined outcome measures] but increases the odds of major bleeding.4 , 7 , 8 The prolonged use of DAPT was particularly supported by the results of the DAPT trial, the largest study specifically addressing this question.4 In this RCT, however, the extended strategy was associated with higher non-cardiovascular mortality.4 A number of risk scores have been advocated to predict the risk of coronary thrombosis and major bleeding in patients undergoing coronary stenting.10–12 A DAPT trial score >2 suggest a clear benefit of DAPT extension on ischaemic events with a negligible risk of bleeding.10 Of note, both clinical presentation as MI and previous MI contribute equally to this score.10 Bleeding risk scores are also very useful to inform clinical decisions regarding DAPT duration.12 Patients with a previous MI are at increased risk for recurrent cardiovascular events. Therefore, major research efforts have focused on the critical assessment of therapeutic modalities that may curtail this risk. The potential benefit of extending DAPT beyond 1 year after MI has been recently suggested. In a post-hoc analysis of the CHARISMA RCT focused in patients with previous MI, the primary endpoint was significantly reduced by long-term DAPT (clopidogrel and asprin) compared with the control group.13 Likewise, in a subgroup analysis of the DAPT RCT, the risk–benefit ratio in favour of extended DAPT duration was particularly favourable in the subset of patients with a previous MI.14 Finally, the PEGASUS-TIMI 54 RCT allocated stable patients with previous MI (median time 1.7 years after MI) to either long-term DAPT (aspirin and ticagrelor) or aspirin alone.15 At 3-year follow-up, the primary clinical endpoint was significantly reduced in the DAPT arm although the rate of major bleeding was increased.15 Interestingly, in this trial, patients with renal failure showed an increased risk, and particularly benefited from extended DAPT.16 A systematic review focused in patients treated with drug-eluting stents suggested that the risk–benefit ratio of prolonged DAPT was better in patients with than in those without MI.3 Finally, a recent meta-analysis including >33 000 patients suggested that in those with a preious MI, extended DAPT significantly reduced ischaemic events and also cardiovascular death at the price of increasing major haemorrhages.17 Considering the bleeding risk associated with prolonged DAPT, the extended regimen should probably be reserved for post-MI patients with a high ischaemic risk and low bleeding risk who had a good tolerance for this combined therapy during the first year.4 , 7 , 8 Patients’ characteristics (i.e. age, weight, diabetic status, and specific genotypes) drastically influence the pharmacokinetics, pharmacodynamics, and, eventually, the clinical effects of antiplatelet therapy.3–9 Expectations that platelet function testing might be used to guide this therapy have not yet been fulfilled.18 In addition, in real-world populations, both ischaemic and bleeding risk are much higher than those observed in the highly selected cohorts of patients included in most RCTs.3–9 Registry data suggest that a large number of patients actually receive DAPT beyond 1 year with significant variability among countries.19 However, the ‘risk–treatment paradox’ suggests that in clinical practice patients with the highest risk (therefore with the greatest potential benefit) tend to be treated less frequently than patients at lower risk. Recent clinical practice guidelines suggest that prolonged DAPT ‘may be considered’ (recommendation IIb, level of evidence A) after careful individualized assessment of ischaemia and bleeding risks.9 Accordingly, balancing ischaemic and bleeding risks on an individual basis remains essential to address this major therapeutic dilemma properly. Could personalized medicine help to disentangle this ‘Gordian knot’? In this issue of the journal, Pasea et al.20 sought to assess the value of population-based electronic health records (EHRs) in patients evaluated 1 year after acute MI to develop (12 694 patients from 159 general practices) and validate (5613 patients from 61 general practices) prognostic models for (i) cardiovascular events (cardiovascular death, MI, and stroke); and (ii) bleeding events. Prognostic models for cardiovascular events (c-index 0.75) and bleeding events (c-index 0.72) showed good discrimination and were well calibrated. Overall, the 3-year risk of cardiovascular events was 16.5% and that of major bleeding 1.7%. Relative risks for efficacy and safety for ticagrelor 60 mg vs. placebo from the PEGASUS-TIMI 54 trial were applied to the validation cohort to estimate events prevented and events caused per 10 000 treated patients per year. Importantly, predicted net benefits were also calculated under multiple risk–benefit clinical scenarios. These investigators estimated that for every 10 000 patients treated per year, 249 cardiovascular events and 134 major bleeding events could have been prevented in the highest risk patients. Overall, there was a clear net clinical benefit of prolonged DAPT (in 63–99% patients) largely depending on how benefits and harms were weighted. Personalized estimated ischaemic and bleeding risks at 5 years with and without extended dual antiplatelet therapy (DAPT). Top: 50-year-old man without coronary risk factors and no significant co-morbidities. Bottom: 80-year-old female with several co-morbidities including coronary risk factors, atrial fibrillation, anaemia, and previous heart failure and renal failure. This figure was generated using the web-based tool developed by the authors (Pasea et al.) to provide a personalized risk prediction20 (http://www.caliberresearch.org/prolonged_dapt_benefits_harms_risks). CI, confidence interval; CV, cardiovascular; MI, myocardial infarction. Some important issues, however, deserve further consideration. First, CALIBER is a well-established research platform providing EHR linkages between primary care data, hospital data, disease registry data, and cause-specific mortality in England20 (www.ucl.ac.uk/health-informatics/caliber). Reassuringly, several previous studies have demonstrated the validity of this population-based tool to estimate cardiovascular endpoints.20 From a methodological perspective, the models discriminated well among risk groups and were highly calibrated. In addition, net clinical benefit was estimated under various assumptions (namely with equal or different weighting of efficacy and safety events). Secondly, size represents a major strength of the use of administrative databases for clinical research purposes. Alternatively, the inherent caveats of these databases remain accuracy, underascertainment, and the reliance on International Classification of Diseases (ICD) codes that may be non-specific or imprecise regarding some clinical diagnoses (including the diagnosis of MI). In the present study,20 the authors nicely acknowledge that these population-based EHR data lack potentially relevant anatomical- and procedural-related information. Actually, very limited procedural data were analysed. Only the performance of an intervention was captured (yes/no), but no additional information (type, length, diameter and number of stents, and target vessel, left main involvement, number of diseased vessels, completeness of revascularization, or left ventricular ejection fraction) was available. Although the investigators highlighted the limited predictive value of many of the absent angiographic data in previous studies, this issue remains controversial. Indeed, classical clinical studies indicate a clear prognostic role for relevant anatomical and procedural-related covariates. Thirdly, this study confirms that patients’ characteristics with potential prognostic value may change from the index MI to 1 year post-MI. Importantly, the differences in clinical characteristics over time influenced the number of patients that were considered eligible for prolonged DAPT. Eventually, 20 and 18 prognostic factors were identified as valuable to be included in the efficacy and safety models, respectively.20 Fourthly, most prognostic factors had a concordant effect on efficacy and safety outcome measures. However, identification of biomarkers able to separate ischaemic cardiovascular events from bleeding risk would be of major value to refine further the clinical decision-making process. Likewise, the relative value of the different new P2Y12 antagonists in extended tailored DAPT regimens requires additional investigation. Finally, personalized estimates of risks of cardiovascular events and bleeding among several risk subgroups were provided. Even when avoiding the risk of bleeding was considered as more important (using different assumptions) than preventing cardiovascular events, a net clinical benefit of prolonged DAPT was observed for most patients.20 This provides additional evidence—from a real-world patient population—on the benefit of the extended therapeutic strategy. In addition, as the rate of cardiovascular events observed in this unselected large patient cohort was higher than that seen in the PEGASUS-TIMI 54 placebo group, the authors dare to suggest that the potential benefit of prolonged DAPT might be even greater.20 Extension of DAPT duration beyond 1 year entails a major trade-off. The use of a patient-targeted approach for prolonging DAPT beyond 1 year after MI is generating major expectations. Further studies are warranted to confirm the benefits of personalized medicine to optimize DAPT duration and to improve clinical outcomes in patients with a previous MI. Conflict of interest: none declared.