Ten Thoughts on the 2019 AF Treatment Guidelines

John Mandrola, MD


February 08, 2019

The American College of Cardiology and American Heart Association have published a focused update of their 2014 atrial fibrillation (AF) treatment guidelines.[1] First some general comments and then some words on the specifics of this document.

1. The Expanding Role of Guidelines

Clear language early on draws your attention: "Guidelines are intended to define practices meeting the needs of patients in most, but not all, circumstances, and should not replace clinical judgment."

As proliferating guidelines crowd out judgment, this common-sense statement could not be more important. It seems self-evident that guidelines should guide, not decree, but, sadly, and to the detriment of patients, the clinician who deviates from guideline-directed therapies incurs risk. The nod to clinical judgment, therefore, is necessary but increasingly irrelevant. More on this later.

2. Can Guidelines Be Too Simple?

Efforts to make the document "user friendly" worry me. The authors laud the fact that their recommendations come in neatly color-coded tables. Yes, this style "optimally facilitates dissemination of information to clinicians at the point of care." But, dangerously, it reduces emphasis on critical appraisal of the evidence.

The simple format and neat categories of recommendations (class I, II, III) belie the complexity and uncertainty surrounding the care of patients with AF. I often let it slip to inquisitive patients that many treatment decisions in AF have two answers: the "guideline-directed" answer and the truthful one. The truth of course is that we often don't really know.

3. Aspirin Gone With Barely a Whisper

Aspirin use in low-risk patients is out. For patients with AF and a CHA2DS2VASc score of 1 in men and 2 in women, the authors write that "prescribing an oral anticoagulant to reduce thromboembolic stroke risk may be considered."

This includes no mention that omitting aspirin from this sentence is different from the 2014 guideline. I agree with their call but highlight the lack of clarifying language because aspirin use is ensconced in people's minds. A clear statement not to use aspirin would have helped clinicians.

4. Valvular Heart Disease Clarified

When the US Food and Drug Administration (FDA) approved the direct-acting oral anticoagulants (DOACs) or non–vitamin K oral anticoagulants (NOACs), they specified use only in patients with nonvalvular AF. They meant that the new drugs shouldn't be used in patients with moderate to severe rheumatic mitral stenosis or mechanical heart valves because these patients were excluded from the clinical trials.

The nuance of language plus the epidemic of overdiagnosis created a problem. Echocardiography reports almost always note some degree of valve thickening or regurgitation. This raised the puzzle of how one defines "nonvalvular" valve disease.

The authors did a great job explaining why the only valvular exclusion for DOACs is rheumatic mitral stenosis and mechanical heart valves. The short version: In the DOAC trials, approximately 20% of patients had various valvular defects, including bioprosthetic replacements, and multiple post hoc studies[2] and meta-analyses[3] upheld the efficacy and safety of DOACS vs warfarin in these patients.

5. Left Atrial Appendage Closure

After the FDA and the Centers for Medicare & Medicaid Services (CMS) approved percutaneous left atrial appendage closure with the Watchman device (Boston Scientific), the guideline writers had to weigh in. Most notable to me was their emphasis on differences in wording from the FDA and CMS. The FDA restricts the device to patients who are deemed suitable for long-term warfarin but had appropriate rationale for seeking a nonpharmacologic alternative. CMS says the device is an option for patients who are suitable for short-term warfarin but deemed unable to take long-term oral anticoagulation. You can read this again but it gets no clearer.

The authors settled on the lowest strength of recommendation (class IIb) while using the text to highlight the fact that patients unable to take anticoagulation were excluded from the Watchman trials. Translation: It's approved, but we have no idea if it works for these patients. My negative views on percutaneous left atrial appendage (LAA) closure have not changed. In fact, recent data on device-related thrombosis,[4,5,6] which are not mentioned in the document, reinforce my concern that this procedure will become one of cardiology's biggest mistakes.

Surgical appendage closure at the time of heart surgery also gets a class IIb recommendation. The text in the table next to this recommendation is interesting: "Surgical occlusion of the LAA may be considered"— not for reduction of stroke or death, but"as a component of an overall team approach to the management of AF." The reason for this hedge is that surgical LAA occlusion is backed only by observational studies, which are fatally flawed by selection bias. The good news is that the Canadian-led randomized controlled trial (RCT) called LAAOS-III will soon shed light on this add-on surgical procedure. 

6. AF Ablation in Heart Failure

Full disclosure: I do AF ablation. Until 2018, critics of AF ablation rightly pointed to the lack of hard outcomes data for this procedure. Then Nassir Marrouche from the University of Utah and colleagues published the CASTLE-AF trial—an RCT comparing AF ablation to medical therapy in patients with symptomatic AF and systolic heart failure.[7] After more than 3 years of follow-up, AF ablation resulted in a 12% absolute risk reduction of both death and heart failure admissions. This finding was not an outlier—the AATAC trial of AF ablation vs amiodarone in similar patients showed a trend toward lower death rate with ablation.[8]

The guideline writers were not convinced. They gave AF ablation for this indication the weakest recommendation possible, class IIb, citing the trials' small sample sizes and highly selected patient cohorts. They called for more studies.

CASTLE-AF and AATAC have limitations—all RCTs do—but to give ablation in these select patients the same class of recommendation as apixaban in patients with end-stage renal disease (ESRD) or LAA closure is an injustice and casts doubt on the validity of the strength of recommendation system. As for the call for more studies, how do you convince patients with heart failure and AF that is refractory to medical therapy to be randomly assigned in a trial that includes taking more of an ineffective medication?

7. DOACs Preferred Over Warfarin

The least controversial part of this guideline is the preference for DOACs over warfarin. Multiple large RCTs and meta-analyses support this recommendation. I was a slow adopter of DOACs and admit that the absolute differences in efficacy and safety of DOACs vs warfarin are small, but after years of using the new drugs, I have come to appreciate their convenience.

My preference for DOACs is not based on evidence alone. I consider DOACs far more compassionate than warfarin. I ascribe to the teachings of Mayo Clinic endocrinologist Victor Montori, MD, who urges clinicians to be mindful of how much our treatments increase the work of being a patient.[9] Time spent being a patient, Montori argues, is less time to enjoy life. Few drugs are more burdensome than warfarin.

8. Female Sex Alone Is Not a Risk Factor

The writers clarify that female sex should not influence the decision to recommend oral anticoagulation unless the patient has other risk factors in the CHA2DS2VASc scoring system. This recommendation stems from recent studies showing that in the absence of other risk factors, women have a similarly low risk for stroke as men. They cite an analysis using three nationwide registries that found the excess risk associated with female sex was especially evident among patients with two or more non–sex-related risk factors.[10]

9. Weight Loss Gets Noticed

The work of the Adelaide group led by Professor Prash Sanders has changed electrophysiology. An RCT showed that weight loss in patients with AF who are overweight or obese is effective,[11] a finding bolstered by multiple observational studies.[12,13,14] Risk factor modification reduces AF episodes, improves AF ablation success, and reduces the progression of AF to more advanced forms.[15]

But two bigger stories from this work have been (1) the gains in basic knowledge about the role of electrical and structural atrial disease in promoting AF[16] and (2) showing patients and doctors the striking improvements that risk factor management can deliver in lipids, blood pressure, and glycemic control—with fewer meds! The latter point bears on public health policy.

10. Apixaban Gets a Nod in Patients With ESRD

Treating patients with AF and ESRD is tough because these patients have a higher baseline risk for stroke and a higher risk for bleeding with oral anticoagulation.  RCTs do not enroll patients with ESRD, and a recent large meta-analysis of observational studies concluded that warfarin therapy was not associated with a higher risk for mortality and stroke/thromboembolism compared with no warfarin use but significantly increased the risk for major bleeding.[17]

The FDA recently approved apixaban use in patients with ESRD on dialysis. Guideline authors cite an observational study of Medicare dialysis patients with AF who started warfarin or apixaban (n = 25,000). This matched cohort analysis suggested that a standard dose of apixaban (5 mg twice daily) may be associated with a lower risk for major bleeding than warfarin and may lower thromboembolic and mortality risk.[18]

The guidelines give warfarin or apixaban use in patients with ESRD a class IIb recommendation. That is problematic because without an RCT, we have no idea whether anticoagulation provides a net benefit or, if it does, which agent is best.


Overall, the writing in this focused update is clear and some specific recommendations are good.

That said, the misstep on AF ablation in heart failure exposes the central flaw of a system of simple color-coded tables with classes of recommendations. If you give a treatment that has multiple RCTs showing benefit the same class of recommendation as treatments with no RCT evidence (surgical LAA closure, apixaban in ESRD), then you might as well cut to the truth and go to two categories: may use and don't use.

The authors would counter this point by saying the level-of-evidence labeling further delineates their recommendations. While this may be true in the academy, in the real world, it is ignored. I have multiple colleagues who dismiss the fact that a recommendation is only a class IIb and may have no RCT. They are busy and care only that a treatment is in the guidelines.

Making decisions simple when they are not should also be reconsidered. Stop and think about our allegiance to the CHA2DS2VASc score, which is simple and easy of use, but at its core distills a decidedly continuous risk for a future event down to an integer. That number then determines right care from wrong care, good docs from bad docs.

Enter Quinn and colleagues with a systematic review of more than 30 studies of AF that shows massive variations in baseline stroke risk of untreated patients with the same CHA2DS2VASc score.[19] Translation: We have no idea of the risk for stroke in untreated patients. The guideline authors quietly cite Quinn et al but keep the CHA2DS2VASc score in the tables of recommendations without a hint of uncertainty.

I have been part of writing committees for guidelines. Getting the wording exactly right in the text boxes is hard. People want to be fair, and the authors surely know that guidelines have gained enormous power—they are used in quality measures, payers' decisions, and malpractice cases.

Reading and commenting on this version of AF treatment guideline gave me two ideas. One is to add a category, say U, for unknown benefit/harm. If we are truthful, many treatments for AF fall into this category. I realize this doesn't help the quality people or payers, but overconfidence at the bedside seems worse to me than saying we don't know.

The other thought I could not push out of my head was for the authors to skip nearly all of the simplified color-coded boxes of recommendations and just give an evidence review. Let clinicians read the evidence and come to their own conclusions.

While this sounds scary, the danger of codifying every aspect of medicine might be to atrophy the all-important skill of critical appraisal. Young children are increasingly unable to sign their name because they have not been taught cursive. We wouldn't want a generation of clinicians who are unable to appraise evidence on their own.



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.