August 11th, 2014

What’s the Next Phase for Digoxin in Patients with Atrial Fibrillation?

CardioExchange’s Harlan M. Krumholz interviews Mintu P. Turakhia, lead author of an observational study of digoxin use and mortality risk in patients with atrial fibrillation, using data from the TREAT-AF study. The article is published in JACC.


Researchers identified 122,465 outpatients in the VA health system who were seen for newly diagnosed, nonvalvular AF during VA fiscal years 2004 through 2008 (23.4% received digoxin). Cumulative mortality rates were significantly higher for digoxin-treated patients than for untreated patients (95 vs. 67 per 1000 person-years). Digoxin use was independently associated with mortality after multivariate adjustment and propensity matching, even after adjustment for drug adherence.

Click here for CardioExchange’s news coverage of the study and a central illustration from the article.


Krumholz: You conducted an observational study within a randomized clinical trial. How confident can we be that digoxin causes an increased risk for mortality in patients with atrial fibrillation (AF)?

Turakhia: TREAT-AF is retrospective cohort study (not a trial) of VA patients with newly diagnosed AF from a combination of linked data sources: VA EMR, VA claims, linked Medicare claims, and vital status. As in any study with nonrandomized treatment assignments, the data are vulnerable to confounding and bias, which we tried to minimize using several techniques:

  1. We used a new disease (AF) cohort to minimize survival bias, which can occur in studies of prevalent disease.
  2. We looked at digoxin prescribed within the first 90 days of AF diagnosis (i.e., initial or early AF therapy) to minimize confounding by indication.
  3. We propensity-matched patients in a way that allowed almost all the digoxin-treated patients (93%) to be matched while minimizing baseline differences in the matched cohort. And we further adjusted for drug adherence. In all of our models, digoxin use was associated with greater mortality risk.

Nevertheless, we cannot exclude the possibility that an unidentified confounder was the causal factor. In measuring that uncertainty, we found that that such a confounder is very unlikely to explain the causal association.

Krumholz: What do we know about the digoxin blood levels in this study?

Turakhia: Prior studies have indeed shown an association between serum digoxin levels and mortality (Harlan, you did much of this great work.) We did not look at digoxin levels because they are not routinely measured in practice, and sicker patients would be more likely to have their digoxin levels measured. However, we did evaluate kidney function, and my hypothesis was that we would see a mortality signal only in patients with kidney disease. In contrast, we found a consistent association with mortality risk across the spectrum of kidney function.

Krumholz: Should we put a moratorium on digoxin use in these patients?

Turakhia: Absolutely not. I don’t want readers to conclude that all patients should stop the drug and that all clinicians should cease using it. There may be perfectly valid reasons to use digoxin, and some studies over the years have shown benefit in heart failure, at least for the outcome of hospitalization. However, in light of the many safer drugs that can be used for rate control, clinicians need to ask whether digoxin should be the treatment of choice. We examined digoxin as early or initial therapy for new AF — not treatment failures or other nuanced scenarios — and we still found a mortality signal.

Krumholz: Do you use digoxin in your practice?

Turakhia: Well, I stop digoxin more than I start it, but that is not just because of a perception of safety. People come to me in an AF clinic generally because they are experiencing treatment failure with whatever they are taking. In selected patients, I do continue digoxin under careful observation.

Krumholz:  Do we need another study of this medication?

Turakhia: Future work that clarifies cause-specific mortality and potential mechanisms of death would provide greater context for our observations. However, I don’t think that calling for a randomized trial is necessary every time a well-conducted observational study shows a signal. We are entering an era when large-scale observational studies will outnumber mega-trials. We must consider the context and the uncertainty of the result, not just the result itself.

Prior to our study, I believe that, with respect to digoxin in AF patients, the needle had already moved from “is probably good” to “might not be good.” Perhaps our study nudges the needle a bit more. The take-home message, in my view, is to carefully reevaluate the role of digoxin in AF management. In the U.S., digoxin is used in 20% to 40% of AF patients, many of whom might be better served by other drugs or treatment strategies.


How do the findings from Dr. Turakhia’s study affect your perception of digoxin use in AF patients?

5 Responses to “What’s the Next Phase for Digoxin in Patients with Atrial Fibrillation?”

  1. Leon Hyman, Ms M.D. says:

    We have come a long way since Dr. Withering published his seminal report in 1785 on the use of foxglove in patients with edema.

  2. Brett Forge, MB BS FRACP says:

    I suspect that most doctors only use digitalis when the patients don’t tolerate a beta blocker or verapamil. The most common reason is hypotension. So that in my practice pts on digitalis are more likely to be hypotensive and to have poor LV fn and cannot get adequate control off af on the beta blocker they tolerate.
    I would therefore expect a higher mortality in these pts, but it is due to their underlying condition rather than the effect of the drug.

  3. Jose Gros-Aymerich, MD says:

    Hi!: when dealing with any information regarding the clinical use of Digitalis preparations, an important question is how often Blood Digitalis levels were assessed, and, as the: ‘Window effect’ for Digitalis level testing may not be well described, what are the chances that in any sampling time, part of the treated patients may be in toxic levels, too high levels of Digoxin being much more noxious than sub-therapeutic dosing.

    Just from the variable hydratation condition coming from weather changes, noticeable changes in Digitalis levels could be expected, from concomitant foods or drugs too, and also time of dosing matters, some propose a night Digitalis dosing to have standardized levels taken the next morning. Digitalis was the Drug where the concept that ‘Generic’ drugs many times don’t match the original product bioavailability was shown, and also that bioavailability greatly differs from one Digitalis brand to another, and thus, the subject of Digitalis levels is a must in any information about a Clinical Trial or an Observational Study involving this line of drugs.

    Are there data about how often patients on Digoxin and related products have their blood Digitalis levels measured in everyday life, and how many are found being out of therapeutic range?
    Thanks. Salut †

  4. Mintu Turakhia, MD MAS says:

    All of you raise excellent points. Dr. Forge, our study looked at digoxin used as early or initial therapy, that is, within 90 days of new AF. There are many clinicians worldwide (not just in the U.S.) who do consider and prescribe digoxin as early therapy. In the situation where other drugs are not suitable (hypotension, poor LV function), then I agree that digoxin is reasonable, but only as a last resort, since it has never been studied in such patients.

    Dr. Gros-Aymerich, using digoxin levels is probably a good idea when prescribing digoxin, but I believe that the arrhythmic risk is probabilistic, not deterministic. Therefore, you will still have some probability of having arrhythmias, even with lower levels of digoxin, when you would have been less likely to had you not been taking it. So it’s unclear if levels solves the problem entirely and the practice of level-based tailored digoxin therapy has also not been systematically evaluated on a large scale.

  5. Jean-Pierre Usdin, MD says:

    Thank you for asking the members to share with you their opinion.

    This retrospective study was done before the worldwide acceptation of CHADS² score (if I remember well it was roughly in 2010).
    I Wonder if we will have the same bad results with Digoxin as a first line drug in AF nowadays. With patients under safer conditions with “general” use of anticoagulant?