Author + information
- †Division of Cardiology, Lahey Hospital & Medical Center, Burlington, Massachusetts
- ‡Harvard Clinical Research Institute, Boston, Massachusetts
- ↵∗Reprint requests and correspondence:
Dr. Matthew R. Reynolds, Lahey Hospital & Medical Center, 41 Mall Road, Burlington, Massachusetts 01805.
Despite—or perhaps because of—a very long history of clinical use, high-quality evidence regarding the long-term clinical outcomes with digoxin and related cardiac glycosides is limited. It took until the late 1990s to complete a large randomized trial that clarified digoxin’s role in the treatment of heart failure (HF) (decreased hospitalization, no effect on mortality) (1). With respect to atrial fibrillation (AF), digoxin’s other main therapeutic target, no such randomized outcomes trial has ever been completed. As a result, the best available evidence on clinical outcomes with digoxin in AF patients comes from post-hoc analyses of clinical trials designed to answer other questions, or from observational studies.
Patients treated with digoxin tend to be older and sicker than those not treated with digoxin (2). In several studies, digoxin treatment in AF patients has been associated with increased crude mortality rates, raising the question: Is it the patients, or the drug? The key to answering this question lies in fully accounting for the differences between the patients who do and do not receive the drug. There are different ways to adjust for such differences in observational research. However, even when similar analysis methods are used, data elements can vary depending on data sources. Even “exposure” to a drug can be defined in more than 1 way. It should not come as a surprise then, that previous studies of this topic have produced conflicting results. Two studies from Sweden have reached differing conclusions regarding a possible link between digoxin and mortality in AF patients after adjustment for confounding factors, one reporting an association (2), the other finding none (3). Likewise, 2 simultaneously published post-hoc analyses of the AFFIRM (Atrial Fibrillation Follow-Up Investigation of Rhythm Management) trial, using different analytic methods, reported opposite conclusions as to whether or not digoxin increases mortality (4,5).
Into this fray enters a new analysis of outcomes in AF patients treated with digoxin, based on a large national sample of patients treated in the Veteran’s Administration (VA) health system (6). In a program called TREAT-AF (The Retrospective Evaluation and Assessment of Therapies in AF), Turakhia et al. (6), in this issue of the Journal, using novel linkages between clinical, pharmacy, claims, and mortality databases, have assembled a cohort of more than 120,000 veterans with incident AF between 2004 and 2008. This unique database was used to explore the potential association between digoxin use and mortality over a roughly 3-year period.
A remarkable 23% of the study cohort was prescribed digoxin within 90 days of the first AF encounter, with <25% of them carrying a diagnosis of HF. The investigators found the expected crude relationship between digoxin use and mortality with an unadjusted hazard ratio (HR) of ∼1.37. The strength of this association was attenuated in multivariable Cox models both in the full study population (HR: 1.26) and in a propensity-matched subset that included 93% of the digoxin-treated patients (HR: 1.21). The investigators additionally provided a sensitivity analysis in which the prevalence of a potential confounder that would obviate their results is plotted against the necessary strength of association between that confounder and mortality.
This new, well-conducted study has several notable strengths. The sample size is at least 20 times larger than any prior study on the topic, the sample is drawn from many clinical centers around the United States, and the time frame of observation is more contemporary than some previous studies (e.g., the AFFIRM trial). The source data may therefore be more representative of current U.S. practice than that of previous analyses. State-of-the-art methods for observational research were used, within the constraints of the available data. The sensitivity analysis suggests that the main study finding may be difficult to ascribe to a single unmeasured confounder.
A few important limitations of the study also require discussion. Studies that are based on VA populations underrepresent women, an issue that matters in this case as a post-hoc analysis of the DIG (Digitalis Investigation Group) trial showed an important interaction between sex and outcomes in patients with HF (7). However, in that study, the risk was higher in women than in men. The investigators did not have information on digoxin doses or serum digoxin levels, but were able to calculate estimated glomerular filtration rate and did not find any difference across glomerular filtration rate strata.
More importantly, findings from observational studies always suffer from the reality that treatment decisions in medicine are nonrandom and frequently made on the basis of factors that cannot be measured, and therefore cannot be used in multivariable adjustment models or propensity score calculations. Findings from observational studies must therefore always be interpreted with caution. For example, a recent Swedish observational study reported that patients with AF and HF who were treated with dronedarone had lower mortality than AF patients with HF who did not receive dronedarone (8). Given previous clinical trial evidence (9), it would be erroneous to conclude that dronedarone reduces mortality in HF patients—rather, it must be the case that, in accordance with recommendations, dronedarone is selectively used in lower-risk HF patients.
In the case of the study from Turakhia et al. (6), the investigators did not have access to data on left ventricular ejection fraction or New York Heart Association functional class, and therefore, they could not adjust for HF severity. Given its accepted role in the treatment of HF, it seems likely that digoxin is selectively used in higher-risk patients, and that these and possibly other unmeasured factors could mediate the reported relationship between digoxin and mortality, such that the “true” hazard ratio is likely to be less than the reported ∼1.2. How close to unity is the “true” hazard ratio? The only way to know for sure is to test the hypothesis in a randomized trial.
What should clinicians do with this (and other recent) data regarding possible risks with digoxin in AF patients? It has been realized for some time that digoxin is no longer the first choice for rate control in AF patients, because other drugs are both safer and more effective. It is now widely recognized that digoxin exerts its rate-controlling effects via enhancement of vagal tone, a mechanism that is easily overcome by exercise and other high catecholamine states (10). It has also long been known that digoxin has a narrow therapeutic window, and that adverse effects, including mortality, are related to higher serum levels (11). Safety continues to be an issue: more patients are hospitalized for adverse effects from digoxin than for any cardiovascular medicine other than antiplatelet and anticoagulant drugs (12). If digoxin is to be used, it is clear that dosing should be conservative, particularly in the elderly.
The main implication of the Turakhia et al. (6) paper and related data is that digoxin should be used selectively and with care in AF patients. This view is reflected in the recently updated AF treatment guidelines, where beta-blockers and non-dihydropyridine calcium channel blockers were given a Class I recommendation for rate control, and digoxin received no specific recommendation at all (13).
Should the use of digoxin for rate control in AF be abandoned altogether? Such a recommendation cannot be made on the basis of this kind of observational data. Going forward, the role of digoxin in AF treatment may continue to diminish. For now, there are still clinical circumstances (HF, difficult rate control, low blood pressure) where this old herbal remedy remains useful.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Reynolds has reported that he has no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- Friberg L.,
- Hammar N.,
- Rosenqvist M.
- Gheorghiade M.,
- Fonarow G.C.,
- van Veldhuisen D.J.,
- et al.
- Whitbeck M.G.,
- Charnigo R.J.,
- Khairy P.,
- et al.
- Turakhia M.P.,
- Santangeli P.,
- Winkelmayer W.C.,
- et al.
- Friberg L.
- Farshi R.,
- Kistner D.,
- Sarma J.S.,
- et al.
- January C.T.,
- Wann L.S.,
- Alpert J.S.,
- et al.