Poor NOAC Adherence in AF Wreaks Havoc on Efficacy vs Warfarin

May 17, 2018

BOSTON — No one expected a new oral anticoagulant (NOAC) comparison with warfarin featuring "real-world" data from patients with atrial fibrillation (AF) to turn clinical trial results on their head.

But that's what happened, though inconclusively, in an analysis of administrative data from a large but highly selected cohort of patients prescribed a NOAC or warfarin for nonvalvular AF in the United States from 2015 to 2016.

The analysis, with too many limitations to disturb guidelines but likely providing useful lessons for clinical practice, saw a slightly but significantly lower rate of thromboembolic events in patients prescribed warfarin than in those prescribed a NOAC.

That was surprising and contrary to NOAC efficacy outcomes that were at least similar to those in patients receiving warfarin in randomized trials comparing the two treatments in nonvalvular AF.

Consistent with those studies, however, the key safety endpoints of hemorrhagic stroke and other bleeding complications were overall less frequent in patients prescribed the NOACs.

Table 1. Hospitalizations or Emergency Department Visits, Rate Per 100 Years, Warfarin vs NOAC Prescriptions

Event Warfarin (n = 42,639) NOAC (n = 40,529) P Value
Thromboembolic events 3.16 3.37 .011
Hemorrhagic stroke 0.88 0.54 <.001
Any stroke 3.73 3.73 .96
Bleeding 5.92 4.90 <.001


Of special note in the analysis of more than 80,000 patients, about one fourth of those prescribed NOACs and more than one third of those prescribed warfarin were found to be poorly adherent to their oral anticoagulation (OAC) regimen.

That means they were receiving their appropriate dosage only 40% to 80% of days on the prescription; they were highly adherent if they were covered for more than 80% of days (patients with less than 40% adherence had been excluded from the analysis).

The NOACs included dabigatran (Pradaxa, Boehringer Ingelheim), rivaroxaban (Xarelto, Bayer/Janssen Pharmaceuticals), apixaban (Eliquis, Bristol-Myers Squibb/Pfizer), and edoxaban (Savaysa/Lixiana, Daiichi Sankyo). The analysis compared them collectively to warfarin.

For patients prescribed one of those agents or warfarin, whether they were highly or poorly adherent to their appropriate regimen made a big impact on how they fared, and not always in expected ways.

Poorly adherent patients prescribed NOACs or warfarin had more thromboembolic events and strokes than highly adherent patients. But poorly adherent patients prescribed NOACs had higher rates of both compared to poorly adherent patients prescribed warfarin.

Even among highly adherent patients, warfarin had an edge over the NOACs for the endpoints of freedom from thromboembolic events (P < .001) and any stroke (P = .032), observed Dhanunjaya R. Lakkireddy, MD, from the University of Kansas, Kansas City.

"The Highest Stroke Rate of Any Group"

Among low-adherence patients, the greater benefit for warfarin- vs NOAC-prescribed patients was "pretty dramatic," with a wider gap starting at about 6 months compared to highly adherent patients, for freedom from thromboembolic events (P < .001) and all strokes (P < .001).

Strikingly, patients with low adherence to NOACs "had the highest stroke rate of any group," Lakkireddy said when presenting the findings here at the Heart Rhythm Society (HRS) 2018 Scientific Sessions.

And a bit counterintuitively, patients poorly adherent to either form of OAC showed a jump in both hemorrhagic stroke and bleeding complications compared to highly adherent patients. However, Lakkireddy said, NOACs remained slightly more protective than warfarin against those events.

So the analysis as a whole suggests that when adherence to prescribed therapy is lower in practice than in the clinical trials — essentially always the case — the treatment's benefits can't be expected always to carry over to practice, he observed.

Table 2. Hospitalizations or Emergency Department Visits by Adherence, Rate per 100 years, Warfarin vs NOAC Prescriptions

Event Warfarin NOAC
High Adherence (n = 27,309) Low Adherence (n = 15,330) High Adherence (n = 29,933) Low Adherence (n = 10,596)
Thromboembolic events 3.05 3.37 3.13 4.05
Hemorrhagic stroke 0.69 1.21 0.43 0.67
Any stroke 3.46 4.20 3.41 4.63
Bleeding 5.28 7.02 4.24 6.77


In the current analysis, whenever adherence was low, those prescribed warfarin "actually did much better" than those prescribed a direct oral anticoagulant, Lakkireddy told theheart.org | Medscape Cardiology. The observed differences were small in absolute terms, but across clinical practice any such differences would likely be clinically important, he said.

"It really shook my fund of knowledge, what I've known about these drugs, and really opens up the debate as to what we need to really learn more about these medications, rather than simply assume."

Clinical Trials vs Clinical Practice

Patients who enroll in trials are typically better educated and more motivated than those broadly across clinical practice, Lakkireddy pointed out, "and that means their compliance is going to be higher." And it's well appreciated that in trials, patients have more frequent contact with providers, who likely remind them more often to take their meds.

And that, he said, "falsely elevates the overall adherence of the drug with a patient. That's the reason why, all of a sudden when you roll the drug out into real-life practice, the compliance creeps down, and that's what leads to the inefficacies of something that has performed so well in the trials that doesn't do as well in real life."

The invited discussant for Lakkireddy's formal presentation of the analysis commented on some of its quirky findings. "We have seen that this real-world data set showed a lower overall efficacy for NOACs compared to vitamin K antagonists in the overall population. This is in contrast to almost all other real-world data sets," said Hein Heidbuchel, MD, PhD, from Antwerp University, Belgium.

It was seen in both highly and lowly adherent patients. "This is really a new finding, which contradicts prior findings, and which requires further analysis to explain."

However, the bleeding risk was lower with NOACs than warfarin, "especially in the high-adherence patients, and that's in line with the randomized controlled trials and real-world data sets we've seen."

But why would the bleeding risk go up with poor adherence to either therapy, when less anticoagulant presumably would mean less bleeding risk? The question highlights why it's important to consider the profile of poorly adherent patients, Heidbuchel said.

The reason is probably that patients who are poorly adherent to either warfarin or a NOAC are already intrinsically at increased risk for bleeding and other events. "Bias by indication may be the explanation for these findings, and not necessarily a cause and effect relationship," he said.

"Those patients at highest risk for stroke were somehow implicitly put on the drug believed to have a lower risk for stroke," and then they didn't take it as prescribed, he proposed. "So in the end you can have a paradoxical finding."

Low-Adherence Impact on OAC Choice

"It's rather ironic that we would postulate the notion that a nonadherent patient should go on warfarin. It's sort of backwards. But the data sort of suggests that maybe they're better off, if they're that kind of person," Andrew D. Krahn, MD, Sauder Family and Heart & Stroke Foundation of BC, Vancouver, British Columbia, Canada, said at a media briefing on the analysis.

With warfarin, he said, time in the therapeutic international normalized ratio range (TTR) is a proven predictor of clinical outcomes. So if a patient's TTR is varying widely at repeated testing, "saying warfarin's better for you doesn't make any sense, it's not biologically plausible. Whereas, if they're nonadherent, and you know that, but most of the time their blood tests are okay, there's some reason to believe they may have a net benefit," he said, referring to the current findings.

"And this is why I think there's limitations to these administrative data sets, that we don't try to make a lot of therapeutic recommendations from them."

Also at the media briefing, Andrea M. Russo, MD, agreed that "it's hard to make firm conclusions from something like this study." But in the present NOAC age, selection bias may be enhancing apparent outcomes among patients getting warfarin compared to the newer drugs, said Russo, from Cooper University Hospital, Camden, New Jersey.

"I don't think its completely out of the question that we've already preselected some patients who are more adherent because they were doing great all along, and nobody really said they had to switch over."

If patients say they're doing well on warfarin and don't want to change their regimen, "I don't force them to switch. So I've already preselected someone who might be doing well. I don't know if that's the case, but that's a possibility."

The analysis looked at ad­ministrative claims for patients in the IBM Watson Health Market Scan databases who were written a prescription of warfarin or a NOAC for nonvalvular AF over 18 months starting in 2015. It included only adults with a CHA2DS2-VASc score of 2 or higher and excluded those with only transient AF or with only short-term OAC use in the previous year.

Role of NOAC Dosing Schedules

Heidbuchel pointed out some limitations to the analysis, including a lack of propensity matching to attenuate confounding, exclusion of transient AF even if the patients otherwise had an elevated stroke risk, no formal adjudication of events, no distinguishing among the different NOACs, and exclusion of patients with prior short-term OAC therapy — who could have stopped taking the drugs because they experienced one of the evaluated endpoints.

Knowing the prevalence of each NOAC in the high- and low-adherence groups would show how many patients were contending with once-daily or twice-daily dosing schedules, Heidbuchel observed.

The recommended, evidence-based schedules in nonvalvular AF are twice daily for dabigatran and apixaban, and once daily for rivaroxaban and edoxaban.

"While adherence may be a little bit higher for the once-daily NOACs, there might also be more of an impact of underprotection when they forget to take the dose." And that, he said, could have affected outcomes in the current analysis.

Lakkireddy discloses receiving honoraria or research grants from or serving as a consultant for AtriCure, Boston Scientific, Biotronik, St. Jude Medical/Abbott, Janssen, Bristol-Meyers Squibb, Pfizer, Estech, SentreHeart, Boehringer Ingelheim, and Biosense Webster. Heidbuchel discloses being on the speakers' bureau for or receiving research grants from Boehringer Ingelheim, Bayer HealthCare, Daiichi Sankyo, Bristol-Meyers Squibb, Merck, Pfizer, Medtronic, St. Jude Medical/Abbott, and Boston Scientific. Russo discloses compensation for services from Boston Scientific, Biotronik, St. Jude Medical, Medtronic, and Zoll Medical; receiving research grants from Boston Scientific, Boehringer Ingelheim, and Medtronic; and receiving fellowship support from Medtronic. Krahn discloses receiving compensation for services for Medtronic and receiving research grants from Boston Scientific and Medtronic.

Heart Rhythm Society (HRS) 2018 Scientific Sessions. Abstract B-LBCT02-03. Presented May 1, 2018. 

Follow Steve Stiles on Twitter: @SteveStiles2. For more from theheart.org | Medscape Cardiology, follow us on Twitter and Facebook.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.