The Professor of Medicine at Duke University School of Medicine discusses prescribing DOACs and his opinion on a recent meta-analysis in the Annals of Internal Medicine.
Manesh Patel, MD
In any condition with multiple approved therapies from a single class, there will always be debate and a desire to compare the therapies to figure which is most appropriate for patients.
Treating nonvalvular atrial fibrillation is no different.
While multiple studies have come out in the past seeking to compare the multitude of direct oral anticoagulants (DOACs) approved by the US Food and Drug Administration (FDA) for treatment, no randomized trials have ever been performed to compare the various members of the class against one another.
A more recent example of this was a retrospective analysis using claims data published in the Annals of Internal Medicine, which concluded apixaban showed a more favorable safety profile in terms of bleeding than rivaroxaban. While the study has been widely shared and received praise, it has also received its fair share of criticisms for design and what some saw as flawed methodology.
To discuss the apparent flaws and learn more about the challenge physicians face when attempting to prescribe the most appropriate DOAC in a clinical setting, HCPLive® conducted an interview on the study and prescribing practices in a real-world setting with Manesh R. Patel, MD, of the Duke University School of Medicine and ROCKET-AF steering committee member.
HCPLive: What are some of the issues you see with the retrospective comparison of DOACs published in the Annals of Internal Medicine?
Patel: I guess I was surprised that the Annals (of Internal Medicine) had published it. I would say just because I think the decision to prescribe a DOAC is very patient-specific and there are measured and unmeasured confounders. So, aside from just saying, "It's retrospective, and it's observational,"—I think it's also important to recognize that the rates of stroke and systemic embolism were identical in the two arms.
It's only after they started to do statistical analysis to say, “We think one population was higher risk than the other because we're going to adjust for this and propensity-match for that," that they come to the conclusion the rates were lower with apixaban, which may be true and maybe a true finding, but it's hard to know when the observed rates were the same and unmeasured confounders, like renal function and left ventricular function, are not captured in the database.
So from my perspective, some things that I clinically would see all the time that I think about our patients—what's their ventricular function, how big is their atrium, what is a renal creatinine, and some other things that might drive how I think about it.
Another flaw in the study I think was the prevalence of prior warfarin. So what they did is that they said, “Well, warfarin was more often frequently previously used in the rivaroxaban group". So they propensity-match for people who had the same rate of previously being on warfarin. If you fail warfarin and you go to something else—that's a very different patient than a new start patient who gets one of these therapies because we know that once you start one of these therapies, if you haven't been on them before your rates may be different than if you failed warfarin before.
So, I guess my way of thinking about that paper are: yes, we have to make clinical decisions and, of course, I have a bias. I've worked on ROCKET-AF, I have obviously been thinking about these drugs a lot, but I would have had a problem whether they said rivaroxaban or apixaban was better when the observed rates were very similar and the patient population demographics didn't look that different.
Another big problem I have for the Annals paper is they only chose the patients that got the 5-mg dose of apixaban or the 20-mg dose of rivaroxaban when we know there's a lot of dose adjustment going on in practice—with a lot of patients getting a lower dose just because people are worried about bleeding. So for me, the higher comorbid patient or the patient with renal insufficiency, which is more represented in rivaroxaban patient with a higher rate of heart stroke, systemic embolism, diabetes, those are the patients that would think of a head of apixaban where the CHA₂DS₂-VASc Score was 2 and they were a lower risk in the trial.
HCPLive: What about the less commonly prescribed DOACs, such as dabigatran and edoxaban? Where do they fit alongside apixaban and rivaroxaban in a clinical setting?
Patel: So with dabigatran, the big pull is that it is the one agent that is probably better than the others for reduced ischemic stroke, at least in the trial. The biggest issue for me has been that dabigatran in the United States is 150 mg and 75 mg. It's not in the 110 mg dose that was studied in practice. So, that's been a limitation and it's 80% renally cleared.
Edoxaban is also fairly effective in that intermediate-range group and I think is effective, but its got another issue, which is as your renal function gets really well, the lower dose did not get approved. So, I think there's the dose, at least in edoxaban that is effective, but it's probably not used as much in our practices, because (apixaban and rivaroxaban) are getting used so much.
HCPLive: Is it possible to get what you consider a reliable comparison without putting the DOACs in a head-to-head randomized trial?
Patel: That's the problem, right? I think people want to want to know and I know clinicians say, "I want to make sure my patients get the best drug." I understand why, but I guess it's just hard to imagine how with just observational registered data or meta-analyses or any of those things, that we haven't already applied our preconceived notions or clinical practice to those places.
So, whether it says any one of these drugs is better than the other—it's really hard to believe that. To be frank, when you look at the rough meta-analysis, they're all really similar. They have a few different characteristics, right? But what I often tell people is those characteristics that we saw in the trials even with that driven by the drug or driven by the patients that were studied.
The fact that the bleeding rate is much higher in one of the trials, which is the higher comorbid patient probably driving what we see and competing risk or older patients die from cancer and other things. So, I do think it's important to recognize that I think the drugs are generally very well-tolerated and they're all generally pretty similar. When that's the case, how do you use them? That's why I said I would use them as they were studied.