Article Text
Statistics from Altmetric.com
Dual antiplatelet therapy (DAPT) with aspirin and a P2Y12 antagonist improves cardiovascular outcomes in patients with acute coronary syndrome but at a cost of an increased risk of bleeding complications.1 There is marked variability in the thrombotic and bleeding risk between individuals with multiple patient and procedural factors identified but not yet fully understood. Generic recommendations for DAPT duration inevitably expose some patients to an excessive duration of treatment and bleeding risk while simultaneously disadvantaging other patients by withdrawing therapy that could protect them from atherothrombotic events. This is increasingly recognised as a major clinical problem, and current European and North American Acute Coronary Syndromes (ACS) guidelines now acknowledge that shorter durations of DAPT may be considered in patients at high risk of bleeding while extended treatment (>12 months) is an option in selected patients. To implement these guideline recommendations, robust prediction tools are required to facilitate the accurate identification of individuals who are more or less likely to benefit from shorter or longer durations of treatment.
The association between bleeding and mortality has been a consistent feature of acute coronary syndrome trials, irrespective of the intervention being assessed. The present paper by Alfredsson and colleagues2 fills an important gap since it provides the first longitudinal long-term bleeding risk score (0–30 months) for medically managed ACS patients treated with DAPT. Previous post-ACS bleeding risk tools have informed on short-term bleeding risk in patients predominantly managed with percutaneous coronary intervention3–5 or at a single time point (2 years) in a population with a majority of stable coronary artery disease (~70%).6 From 12 preselected candidate variables and using the Targeted Platelet Inhibition to Clarify the Optimal Strategy to Medically Manage (TRILOGY) ACS study database, the authors identified 10 baseline predictors of Global Use of Strategies to Open Occluded Arteries (GUSTO) moderate/severe/life-threatening and five baseline predictors of thrombolysis in myocardial infarction (TIMI) major or minor bleeding (table 1). While many of the predictors for bleeding were common to those previously reported, other variables such as a history of peptic ulcer disease and angiography before randomisation (unique to the TRILOGY ACS database) were also identified. In sharp contrast to results from short-term bleeding risk tools and studies of long-term bleeding in mixed populations,6 7 female gender predicted a reduced long-term bleeding risk. These and other differences highlight the potentially time-dependent and population-specific contribution of risk factors for bleeding, emphasising the possible hazard of applying prediction tools to patient groups distinct from the reference cohort.
The bleeding risk prediction models derived by Alfredsson and colleagues showed moderate to good predictive accuracy with C-indices of 0.78 (95% CI 0.75 to 0.80) for the GUSTO model and 0.67 (95% CI 0.65 to 0.69) for the TIMI model. Internal validation with bootstrapping gave bias-adjusted C-index values of 0.77 and 0.65 for the GUSTO and TIMI models, respectively, comparing favourably to existing prediction tools. However, if risk scores are to be of clinical utility they must also be easy to estimate, generalisable to populations treated in clinical practice and perhaps most importantly, actionable.
All the current risk models for bleeding are based on metrics measured at a set time point, typically within the index presentation. Measures of admission blood pressure or heart rate and cardiovascular or periprocedural medication can be difficult and time consuming to retrieve, particularly as patients are often transferred to and from a tertiary centre for angiography. Furthermore, given the high turnover of most cardiology departments, formally estimating an individual’s relative risk and benefit of continuing DAPT will often default to the outpatient clinic. Thus, even at the expense of a degree of model discrimination, prediction tools with a modest number of easily identifiable risk factors (eg, TIMI minor or major bleeding model in the current paper—five variables; Patterns of Non-Adherence to Anti-Platelet Regimens in Stented Patients (PARIS) registry risk score—six variables) may be of greater clinical utility, as they are more likely to be adopted in the ‘real world’.
Selected populations included in randomised controlled trials have lower rates of bleeding and non-cardiovascular death than the general population, since patients with any history of bleeding or major comorbidity are often specifically excluded from such trials. Applying risk tools derived from clinical trial data to broader populations may therefore underestimate true bleeding hazards. A relative strength of the prediction models derived by Alfredsson and colleagues is the high number of female patients (~40%), patients aged over 75 years (>22%) and important comorbidities in the study population, comparable to those found in large observational ACS registries.1 External validation in a predominantly invasively managed ACS trial population (Thrombin-Receptor Antagonist Vorapaxar in Acute Coronary Syndromes trial) demonstrated good generalisability of both risk tools to other acute coronary syndrome (trial) populations with a C-index that was essentially unchanged for the TIMI minor or major bleeding model (0.68) and modestly attenuated for the GUSTO moderate/severe/life-threatening bleeding model (0.69).
Patients with cerebrovascular disease are particularly sensitive to the bleeding risks of DAPT and a past history of transient ischaemic attack (TIA) or stroke is both a commonly encountered comorbidity and important risk factor for intracranial haemorrhage (ICH) in ACS populations. Unfortunately, many clinical trials of DAPT, including TRILOGY ACS, exclude such patients from the study population. Management of ACS patients with an indication for anticoagulation remains unclear with considerable heterogeneity in clinical practice reflecting the limited available data. This is even more problematic given the recent widespread uptake of contemporary direct oral anticoagulants. In the PARIS registry risk score,6 triple therapy (DAPT + oral anticoagulant) at discharge was associated with a near twofold increased risk of major bleeding within 2 years; although given the recruitment window was between 2009 and 2010, it is unlikely that individuals were receiving an oral anticoagulant other than a vitamin K antagonist. Prediction tools that inform on the short-, medium- and long-term risks and benefits of triple therapy are to be strongly welcomed as well as ongoing trials to assess the long-term discontinuation or resumption of antithrombotic therapies in patients with a history of ICH [ISRCTN71907627].
The CHA2DS2-VASc score has a pooled C-statistic of approximately 0.68 for predicting the risk of ischaemic stroke in non-anticoagulated patients, lower than most bleeding prediction scores. Nevertheless, this clinical tool is widely used because not only is it is easy to calculate, the results are easy to translate into patient management. Presently, this is a shortcoming of all bleeding risk scores. Optimal DAPT duration depends on the balance between preventing future atherothrombotic events and the increased risk of bleeding from continued treatment. Without knowing the consequences of discontinuing therapy, how then can a 2-year risk of major bleeding, for example of 5%, inform treatment? This is even more challenging if we consider that many of the risk factors for bleeding (eg, age, gender, type of presentation, renal dysfunction) predict higher atherothrombotic risk and vice-versa. On the basis of a high bleeding risk alone, prescribing shorter durations of DAPT will inevitably disadvantage some patients who stand to gain the most anti-ischaemic benefit. Clinical tools that can predict the trade-off between ischaemic event reduction and occurrence of major bleeding events from continuing DAPT may be of greater practical use. Using data from the DAPT trial, the DAPT score was developed to help determine the net clinical benefit of extending dual antiplatelet therapy from 12 to 30 months in both ACS and non-ACS patients who underwent PCI.7 However this and other similar risk/benefit tools remain to be validated in broader populations and are perhaps too immature at present for mainstream use.
Currently, bleeding and thrombotic risk prediction models for DAPT in post-ACS patients are based on variables collected at a single time point and therefore provide a static determination of the risk of future events. In reality though the bleeding risk for an individual is dynamic, modified by the accumulation of positive or negative predictors of risk and the relative contributions of these factors over time. The consequences of an event can also be time dependent, and it is important to keep in context the merits to the individual of increasing the risk of one type of event to prevent another. For example, the risk of death associated with early stent thrombosis is markedly higher than the risk of death associated with very late stent thrombosis.8 In contrast, there is much less temporal variability in the mortality from bleeding.9
Ideally, risk prediction tools are developed that allow modifiable longitudinal risk estimates to be generated at any given point, taking into account known dynamic variables and the time dependent severity of an event. Whether it is feasible to assimilate all these factors into one unifying and implementable model remains to be determined. Until then, the TRILOGY ACS bleeding risk models presented by Alfredsson and colleagues represent a significant step forward in predicting long-term bleeding risks with DAPT after ACS to facilitate clinical decision making on an individual level.
Footnotes
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.