Anti-thrombotic treatment in patients with atrial fibrillation (AF) undergoing percutaneous coronary intervention (PCI) harbours the risk of inducing relevant bleeding complications when an oral triple therapy is prescribed to prevent embolic stroke and coronary stent thrombosis at the same time [1]. Oral anticoagulation is highly efficient in reducing stroke risk in AF [2, 3], but inefficiently prevents stent thrombosis following PCI [4, 5]. On the contrary, dual anti-platelet therapy (DAPT) is highly efficient in preventing stent thrombosis following PCI [4, 5], but inefficiently reduces stroke risk in AF [6]. The dilemma is that for a long time, guidelines recommended triple anti-thrombotic treatment as the default strategy following PCI in AF to reduce the ischemic and embolic risks [7‐9], but that strategy bears an enormous risk of relevant bleeding complications [10]. Over the last years, several studies have shown that more intense anti-thrombotic treatment in patients with acute coronary syndromes (ACS) does not automatically lead to fewer ischemic events [11‐14] and less intense anti-thrombotic regimens were specifically tested after PCI in AF patients [15‐18]. As most of the studies are not comparable between each other, we want to highlight the individual specificities of the trials and to compare similar outcome data where available.

Anti-platelet therapy is a corner stone in acute and chronic treatment of ischaemic cardiovascular diseases comprising stable coronary artery disease (CAD) as well as ACS. DAPT is required for a defined period of time following stent implantation during elective (usually 6 months) as well as urgent (usually 12 months) PCI to prevent stent thrombosis and subsequent myocardial infarction [19]. Generally, clopidogrel is the most commonly used P2Y₁₂ inhibitor prescribed together with acetylsalicylic acid (ASA) for DAPT following elective PCI. The European Society of Cardiology (ESC) published a “Focused Update DAPT 2017”, which summarises the current standard in individualised anti-platelet therapy [19]. Current ESC guidelines (2018 myocardial revascularisation [8], 2015 NSTE-ACS [20], 2017 STEMI [21], 2017 DAPT [19]) demand an increasingly differentiated anti-thrombotic treatment for ACS patients to maximise treatment efficacy and to reduce bleeding risk. Nevertheless, triple therapy is still the default option following PCI in patients on oral anticoagulation (OAC), at least in European recommendations [8].

When bleeding is a serious concern, shortening DAPT duration even to only 1 month while using modern drug-eluting stents, did not increase the rate of stent thrombosis or myocardial infarction in the ZEUS trial compared to short DAPT duration with the use of bare-metal stents [22]. Therefore, operators are encouraged to use drug-eluting stents also in patients with uncertain ability for longer DAPT treatment. Current guidelines recommend to shorten DAPT duration, if clinical co-variables indicate a potential bleeding risk, e.g. if the PRECISE-DAPT score is ≥ 25 [8, 19].

In the early times of PCI, several trials documented that OAC using vitamin K antagonists (VKA) even with target INRs above 2.0–3.0 were not as efficient as DAPT consisting of ASA and ticlopidine in preventing stent thrombosis but caused much higher rates of severe bleeding [4, 5]. Therefore, when a patient has an indication for OAC, OAC alone might not be sufficient to prevent stent thrombosis after PCI. However, it is often forgotten that those older trials did not use VKA alone, but combined VKA with ASA indicating that ASA might not have such a great effect on the prevention of stent thrombosis as it was used in both groups, those with and those without stent thrombosis occurring [4, 5].

Ischemic or embolic strokes are the major disabling complications in patients with atrial fibrillation (AF). OAC can prevent most of these events and is, therefore, widely used for stroke prevention in AF (SPAF) [23]. Excluding patients with very low stroke risks, the superiority of OAC compared to non-treatment is overwhelming, and based on current ESC “2016 Guidelines for the management of atrial fibrillation”, should be used in SPAF [9]. Furthermore, the European Heart Rhythm Association (EHRA) published their “2018 EHRA Practical Guide on the use of non-vitamin K antagonist oral anticoagulants in patients with atrial fibrillation”, which elucidates the advantage of non-vitamin K oral anticoagulants (NOACs) compared to vitamin K antagonists (VKA) [24]. Previously, a potential benefit of certain NOACs compared to VKA has been described in a meta-analysis including differing indications for anticoagulation [25]. NOACs are safe and effective during cardioversion [26‐28], AF ablation [29‐32], following transcatheter aortic valve implantation [33] and even when left atrial appendage thrombi have been formed on OAC their resolution appears to be not different on NOACs and VKA [28, 34, 35]. Upon availability of NOACs in Germany, AF patients were more frequently anticoagulated and stroke rate decreased without an increase in bleeding complications [36] and may exert differential effects on cardiac function than VKA [37].

ASA had previously been considered as a potential alternative to OAC in elderly patients due to a presumed lower bleeding risk than VKA. In the AVERROES trial randomising 5,599 patients to apixaban vs ASA, apixaban had a similar bleeding risk compared to ASA, but reduced stroke and systemic embolism (SSE) by 55% [38]. More specifically, in elderly patients apixaban had a similar bleeding risk and reduced embolic risk compared to ASA dramatically (e.g. 86% risk reduction for patients ≥ 85 years of age) [39]. Consequently, current guidelines clearly state that single anti-platelet treatment is not recommended for SPAF [9]. DAPT is not an option to replace OAC for SPAF in general due to its limited efficacy resulting in a 44% higher ischemic event rate than VKA in the ACTIVE W trial randomising 6706 patients either to DAPT combining acetylsalicylic acid (ASA) plus clopidogrel or VKA [6]. DAPT with more potent P2Y₁₂ inhibitors has not been evaluated for SPAF.

When a patient with AF requires PCI with consecutive DAPT, combining ASA + OAC + clopidogrel can increase the risk of bleeding by 3–4-fold compared to OAC alone [10, 40]. Thus, there had been a quest for alternative anti-thrombotic combination therapies with less bleeding risk. First, gastrointestinal bleeding is reduced by proton-pump inhibitors whenever a combination of anti-platelet and OAC medications is prescribed [41, 42]. Second, a combination of OAC with either prasugrel or ticagrelor is not recommended due to an increased bleeding risk [19, 43, 44]. Third, choosing a NOAC instead of VKA might reduce the bleeding risk as all four NOAC strategies tested in and approved for SPAF use an anti-thrombotic intensity that is comparable to sub-therapeutic dosing of low-molecular weight heparins or VKA. Fourth, dropping one out of three anti-thrombotic drugs early after PCI might further reduce bleeding risk, if it is safe regarding prevention of stent thrombosis. As trials mentioned in the upcoming sections have different bleeding definitions, the different definitions are summarized in Table 1.

There have been several trials comparing additive NOACs in addition to DAPT in patients with PCI for ACS, and the results are diverse. In RE-DEEM, adding dabigatran to DAPT was associated with an increased ischaemic event rate despite more intense anti-thrombotic therapy on lower doses and more ISTH-major bleedings on the usual dabigatran doses [14]. The mechanism of increased rates of myocardial infarctions on dabigatran as described in the RE-LY trial for SPAF [45] or in meta-analyses including other direct thrombin inhibitors [46] is still under debate, however, functional experiments suggest that there is a distinct mechanism in direct thrombin inhibitors which contributes to increased platelet activation and adhesion [47]. In ATLAS-ACS 2 TIMI 51 adding rivaroxaban 2 × 2.5 mg/day or 2 × 5 mg/day to DAPT, ischaemic events were lower on additive rivaroxaban, but non-CABG-associated TIMI-major bleedings were increased to a similar extent as ischaemic events were lowered. It is important to note that neither of the two bi-daily dosing regimens had ever been proven effective for SPAF [48]. The APPRAISE-2 trial adding full-dose apixaban to DAPT was terminated prematurely because of an increase in major bleeding events with apixaban in the absence of a counterbalancing reduction in recurrent ischemic events [12]. In summary, there is no evidence of clinical net-benefit by adding OAC with NOACs on top of DAPT in ACS patients. In most studies, there is even an increase in harm measured by major bleeding complications.

Five studies have investigated strategies of early discontinuation or even only peri-procedural application of ASA in patients receiving a P2Y₁₂ inhibitor with OAC: WOEST [49] dropped ASA on a VKA background, whereas PIONEER AF [16], RE-DUAL PCI [15], AUGUSTUS [17], and ENTRUST-AF PCI [18] evaluated dual strategies on the different NOAC backgrounds (Fig. 1, Table 2) [50, 51]. However, all studies were not sufficiently powered to demonstrate efficacy regarding stroke prevention. Using a network meta-analysis of these trials, there was no indicator of increased ischaemic risk when using a NOAC in combination with only a P2Y₁₂ inhibitor. In particular, there was no increase in stroke risk when using a NOAC-based dual anti-thrombotic regimen compared to a VKA-based triple regimen [18, 52]. The specific trials are discussed in more detail:

In elderly patients, NOACs may be related to less side effects due to medication interaction than VKA. For most of the above mentioned trials, the primary bleeding endpoint has been reported for elderly patients, but for major bleedings or ischemic events there is yet insufficient data reported [58]. In general, clinically significant bleedings in elderly patients were lower on NOAC/dual compared to VKA/triple strategies [15‐18, 49]. In summary, it seems reasonable for elderly patients with AF, who require PCI, to administer ASA peri-procedurally followed by OAC plus clopidogrel. OAC can principally mean NOAC, but limitations apply to the currently published studies with dabigatran (trend towards more ischemic events in lower dose, only lower dose for elderly patients) and rivaroxaban (excluding patients with higher stroke risk, dose reduction without proven clinical efficacy) [59]. Data for apixaban do suggest that full-dose NOAC instead of VKA is safe and efficient in elderly patients. Current ESC recommendations emphasize that if NOACs are used they should be given without empiric dose reduction [19]. The advantage of all combinations of NOAC and clopidogrel are significantly lower bleeding rates compared to conventional triple therapy using VKA [59]. Another important contribution to prophylaxis of bleeding especially for elderly patients is to avoid permanent combinations of ASA or clopidogrel with (N)OAC for CAD and AF following the usual post PCI duration of DAPT or additive P2Y₁₂ inhibition [9, 19].

Due to the rather low number of participants in most of the trials, the primary endpoint was relatively soft considering the composite of major bleeding and clinically-relevant non-major bleedings, which could potentially cause some bias when trying to balance risk and benefit from the primary endpoints. In contrast to previous analyses [60] we focused on primarily reported data on TIMI-major bleedings (Fig. 2). Similarly, efficacy was not even evaluated as a primary endpoint, but in most cases only considered hypothesis generating. Even then, the composite ischemic endpoint was not always restricted to major adverse cardiovascular events (MACE: CV death, non-fatal myocardial infarction and non-fatal stroke), but often included hospitalisations and/or revascularisation. Therefore, we collected data on major bleedings according to TIMI classification (Table 1) with a separate note on intracranial bleedings as well as on MACE (Table 3) from the four trials. The reassuring information on aggregate was that NOAC- compared to VKA-based strategies first reduced the risk for TIMI-major bleeding by 39%, second reduced intracranial bleedings by 66%, and third did not increase MACE rate (Fig. 3). To balance the absolute risk over all trials, we calculated the incremental risk and the incremental benefit over all trials for less intense (NOAC-based dual) compared to more intense (VKA-based triple) anti-thrombotic treatments [61]. While NOAC-based dual strategies were associated overall with an absolute 0.4% incremental risk for MACE, they were also associated with an absolute 1.1% benefit for TIMI-major bleedings compared to VKA-based triple strategies. Therefore, there was an absolute 0.7% risk–benefit [62] in favour of NOAC-based dual strategies based on the higher impact of preventing TIMI-major bleedings compared to MACE, which occurred more often in the population, but did not increase as much during less intense anti-thrombotic treatment. When looking into the randomisation strata of the individual trials, there appears to be a modest risk benefit in most of the analysed strategies (Table 3).

As the primary data from AUGUSTUS compared VKA to NOAC and independently ASA to placebo, we provide the same calculation for the VKA-triple compared to the apixaban-dual group from data published in a meta-analysis by the AUGUSTUS group [52]. Again, the NOAC-dual arm inherits some risk for more MACE, however, reduction of TIMI-major bleedings was even stronger (Table 3). Recently, a subanalysis from AUGUSTUS indicated that the major prevention of MACE by triple therapy occurred during the first 30 days, whereas triple anti-thrombotic treatment beyond 30 days was related to more significant bleeding events without reducing MACE [56].

The very early discontinuation of ASA raised the question about an inherent risk of stent thrombosis. PIONEER AF was the only of the four trials that did not have a proper definition and adjudication of stent thrombosis events. Given that limitations, no significant differences were reported between the rivaroxaban dual and the VKA-triple group [0.8% vs 0.7%, HR 1.20 (0.32–4.45)] [16]. RE DUAL-PCI reported definite stent thrombosis to be increased by 86% on the dual dabigatran 110 mg regime [1.5% vs. 0.8%, HR 1.86 (0.79–4.40)] and to be similar on the dual dabigatran 150 mg regime [0.9% vs. 0.9%, HR 0.99 (0.35–2.81)] compared to VKA-triple therapy [15]. AUGUSTUS reported definite or probable stent thrombosis with no relevant difference between apixaban and VKA [1.3%/year vs. 1.6%/year, HR 0.77 (0.38–1.56)] and a marked difference between ASA and placebo [1.0%/year vs. 1.9%/year, HR 0.52 (0.25–1.08)] [17]. Regarding stent thrombosis in AUGUSTUS, at no time had the NOAC been inferior to VKA regarding the occurrence of stent thrombosis and most stent thromboses occurred in the placebo (compared to ASA) group and did so within the first 30 days [63]. ENTRUST AF-PCI reported definite stent thrombosis with some difference between edoxaban dual and VKA-triple regimens [1.16%/year vs. 0.88%/year, HR 1.32 (0.46–3.79)] [18]. Given the low event rates resulting in all trials being underpowered to detect a significant difference, several meta-analyses have been performed [18, 52, 60]. However, the major problem is the reporting of different definitions of stent thrombosis. Some studies do not report probable events, the ones reported from PIONEER AF-PCI as “any” are most likely “definite” ones. If mixing all strategies the question remains whether the lower dose of dabigatran in RE-DUAL PCI should be included in such comparisons. The differing numbers and definitions of stent thrombosis are provided in Table 4. Given the limitations in different definitions of stent thrombosis and mixing those for meta-analyses, there is signal about a hazard ratio of 1.5, which in some analyses is just significant, in others just misses significance. Nevertheless, these aggregate results would suggest that immediate dual anti-thrombotic therapy with ASA only given at the time of procedure might indeed pose some risk regarding stent thrombosis.

Still the picture remains obscure, as the original AUGUSTUS publication reports two more events than the subsequent publication on stent thrombosis, one each for apixaban and VKA, but both in the placebo group [17, 63]. If one would compare only the full-dose NOAC-based dual to VKA-based triple anti-thrombotic therapies from AUGUSTUS and ENTRUST AF-PCI for definite or probable stent thrombosis, the resulting relative risk ratio is 1.32 (0.69–2.51). When additively including the reported definite stent thrombosis from RE-DUAL PCI in the high-dose group, which is recommended to be used by guidelines now [8], the resulting relative risk ratio for stent thrombosis on full-dose NOAC-based dual compared to VKA-based triple anti-thrombotic therapies is 1.22 (0.70–2.11). The overall rate of stent thrombosis in all trials is reassuringly low. The number needed to treat to prevent one stent thrombosis by adding ASA in AUGUSTUS was 250. The number needed to harm to induce one major bleeding by adding ASA in AUGUSTUS was 55 using the ISTH definition and 91 using the TIMI definition [17]. Therefore, even eventually significant increases in relative risk for stent thrombosis by NOAC-based dual compared to VKA-based triple anti-thrombotic therapies might not translate into clinical net harm as more major bleedings are prevented at the same time by dual anti-thrombotic regimens. This thought is supported by the clinical net-benefit analyses.

In summary, the accumulated data suggest that ASA beyond 30 days is not required in patients on OAC plus a P2Y₁₂ inhibitor following PCI. For the first month, however, severe bleeding and ischemic events have to be weighed against each other and the treatment regimen has to be adopted on an individual basis including patient-specific ischemic and bleeding risk.

Current guidelines state that “the use of ticagrelor or prasugrel is not recommended as part of triple anti-thrombotic therapy with aspirin and an OAC” [8]. However, there is no clear recommendation regarding those P2Y₁₂ inhibitors, which are more potent than clopidogrel, in the context of dual anti-thrombotic therapy, e.g. when ASA is dropped very early. Their use in the contemporary OAC PCI trials was low (Fig. 1) and most probably the patients treated with them might have had different baseline characteristics (e.g. younger, lower bleeding risk) compared to patients being treated with clopidogrel. Therefore, one has to be very cautious when interpreting reported bleeding rates, which were sometimes higher, sometimes lower on more potent P2Y₁₂ inhibitors added to OAC (Suppl. Table 1). One major consistent finding through all four trials, however, was that the bleeding on the more potent P2Y₁₂ blockers was always lower in the NOAC/dual compared to the VKA/triple strategy,

In general the available data from all studies suggest that bleeding can be lowered significantly by reducing the intensity of anti-thrombotic treatment. In PCI patients with AF as an indication for OAC, NOACs in the same dose as approved for stroke prevention have shown the most reliable protection regarding stroke prevention and coronary ischaemia. NOAC-based strategies inherit a lower risk for major bleedings than VKA-based regimes and should, therefore, be preferred. Beyond 30 days, there are no conclusive data supporting triple therapy with additive ASA. For the first month, there are concerns regarding potential stent thrombosis, even so the rates appear to be low. Dedicated analyses from AUGUSTUS suggest that individual balancing of ischaemic vs bleeding risk is required to tailor the optimal anti-thrombotic regimen for AF-PCI patients requiring OAC for stroke prevention. Regarding elderly patients with higher stroke risk, strategies with deliberate dose reductions should be avoided as they pose a higher risk for ischemic/ embolic events.