Pablo Lamelas –
AUGUSTUS was the last (until mid 2019) large randomized controlled trial evaluating antithrombotic therapy in patients with atrial fibrillation with indication of oral alticoagulation and concomitant indication of dual antipletelet therapy. Here we present methodological comments, then how it fits in context with two key questions: 1) Should Apixaban 5mg/BID be the NOAC used with triple therapy? 2) Time to move to dual (NOAC + Clopidogrel) therapy?
This trial, of people needing dual anti-platelet therapy (not necessarily a PCI) and having concomitant atrial fibrillation requiring anticoagulation, evaluated two different questions with a factorial design: 1) if apixaban was non-inferior (or superior) than anti-Vit-K agents and 2) Aspirin vs placebo (triple vs dual regimen) for the main outcome of bleeding. Apixaban resulted superior in reducing bleeding compared to anti-VitK, while aspirin (triple therapy) was also associated with increased bleeding compared to placebo (dual therapy).
Time from index event to randomization
The protocol stated that randomization should happen within 14 days from presentation, and this ended being 6 days median. This is a key aspect of the study, since the majority of patients, even those allocated to dual therapy, received aspirin (therefore, likely triple therapy) periprocedurally.
Why is important? For external validity purposes, is unknown if dropping ASA the day of the procedure would result in higher stent-related thrombotic events, since is well known that the highest risk of stent thrombosis is within 24 hours of the PCI (specially those presenting as acute coronary syndromes) and then decays markedly as time goes on.
Bleeding combined outcome
The primary outcome was 6-month major bleeding (very important bleedings) or clinically relevant non-major bleeding (hospitalization, medical or surgical intervention, unscheduled clinic visit, change in antithrombotic therapy) as per ISTH.
This combined outcome may be a good estimate of the burden of bleeding. However, this combined estimate may result troublesome when putting it into the balance against thrombotic events. For example, from patients perspective, a relevant but non-major bleeding may be wighted as lesser importance compared to coronary thrombotic complications, specially stent thrombosis. Dr Devereaux showed that patients are much more likely to weight more ischemic events than bleeding events when compared to doctors.
Why this is important? Think we all agree that a major bleeding can be consider as or even more important than many coronary thrombotic complications, but the majority of relevant non-major bleedings likely less important to patients vs coronary thrombotic events.
In AUGUSTUS authors conclude that NNT to avoid bleeding superior to possible NNH dropping ASA supporting dual regimen, but if we take away non-major bleedings and compare mayor bleeding-only vs MI (more comparable events) major bleed 4.7 vs 2.9%, MI 2.9 vs 3.6%, both non-significant.
Same non-statistical trend observed for stent thrombosis. I know that this could be chance, but study underpowered for this patient-important outcome, and go in the pre-specified direction and is about the safety of a dual regimen, raising concern rather than reassurance in my opinion. Especially because 1 in 4 patients had no stents placed in index event, and that may be an important diluting factor, and these patients (those who did not have a fresh stent placed) were not excluded from the denominator in such analysis (or even sensitivity analysis).
On the other hand, Apixaban did not show this trend vs Anti-VitK for CV death, MI, stent thrombosis or urgent revascularization, with reasonable confidence intervals in some of the individual outcomes.
Factorial designs are great to test more than one research question in the same study, two in one in this case (triple vs double therapy, and NOAC vs Anti-vitK). To accomplish this you need to assume no interaction between study arms. In other words, that one study question (or intervention) does not potentiate or reduce the efficacy of the other study question (or intervention). If there is no interaction, the study can be interpreted simply like two separate studies. But, if there is interaction, then this study will be underpowered (if doing the two-in-one approach). However, the best way to test an interaction is doing a factorial design, but with double sample size!
If this is confusing think about this hypothetical example: if you need 4.000 to adequately test dual vs triple therapy, and you need 4,600 people to test NOAC Vs anti-vitK, assuming no interaction, then a factorial study with 4,600 would be ok. However, if there is interaction (one of the interventions affects the efficacy of the other), doing a 4,600 will be underpowered: the best way to demonstrate and quantify the interaction effects is doing a large factorial design that will need close to 8,000 people.
Why this is important in this case? because this study was done assuming no interaction and when they tested them they couldn’t find evidence of it, so analysis can be interpreted as two studies in one. If interaction was there, then this study will be underpowered for the outcomes powered for. One may argue that there may still be an interaction that was not detected because is inherently underpowered for that given the sample size, and thats a valid point. Indeed, the WOEST trial did suggest some degree of interaction (adding ASA was not just additive, but exponential in bleeding rates), that was not clear in the rest of the NOAC trials (RE-DUAL and PIONEER-AF). But, on the other hand, is reasonable to assume no interaction since data did not suggest that.
Wording: Safety Vs Efficacy
Wording is important to understand and then transmit the results of clinical research. Should we change the wording from constantly calling efficacy = ischemic and safety = bleeding, and adapt to context? If intervention is to drop ASA to reduce bleeding, then efficacy is reducing bleeding, and safety not paying the price of higher ischemic events.
Bleeding presented as 100patient/years. The issue I find interpreting is that bleeding hazard not constant over time (highest at the beginning). Look bleeding survival figures, around 3/4 of the bleeding events happen within 90 days (1/2 trial), so not very constant and translatable to years.
You will see that in tables the 100 patient/year estimate is roughly double the absolute 6-month raw percent. That is ok if hazard is constant, but here is not the case and after 6 months likely to reach the plateau similar to 90 to 180 days, so 100 patient/years estimates likely exaggerated.
Research in context
Based on AUGUSTUS on top of prior literature:
Question 1: Is apixaban 5mg/BID the standard when using triple therapy and NOAC preferred?
AUGUSTUS was the only randomized trial that used triple therapy with NOAC with a dose approved for stroke. So, are the results of triple therapy valid for the other NOACs? this is unknown, but seems reasonable to assume that the effect seen in bleeding would parallel what apixaban showed here, with some degree of indirectness.
Question 2: Is time to move to a dual regimen after PCI and AF?
In this trial, even the dual therapy arm used aspirin up to a median of 6 days before allocation. Therefore, recommending dropping ASA just after the PCI is not fully supported by this analysis. Not stating if it is unsafe or not, just unkown. However, after this early period seems that in cases without a high risk of stent thrombosis (ie, Left-main PCI, two-stent bicurcation PCI, PCI with suboptimal result, etc) early drop of aspirin seems the way to go. If high risk of stent thrombosis is suspected, then we will need more data, hopefully individual patient data meta-analysis, to better understand this issue.