Article Text


Digitalis: a dangerous drug in atrial fibrillation? An analysis of the SPORTIF III and V data
  1. K Gjesdal1,
  2. J Feyzi2,
  3. S B Olsson3
  1. 1
    Ullevål University Hospital, University of Oslo, Norway
  2. 2
    Statistical Data Analysis Center, Department of Biostatistics and Medical Informatics, University of Wisconsin, Madison, WI, USA
  3. 3
    Department of Cardiology, University of Lund, Sweden
  1. Professor K Gjesdal, Ullevål University Hospital, 0407 Oslo, Norway; knut.gjesdal{at}


Objective: In heart failure, digitalis increases exercise capacity and reduces morbidity, but has no effect on survival. This raises the suspicion that the inotropic benefits of digitalis may be counteracted by serious adverse effects. Patients with atrial fibrillation (AF) were studied to clarify this.

Design: In the Stroke Prevention using an ORal Thrombin Inhibitor in atrial Fibrillation (SPORTIF) III and V studies, 7329 patients with AF at moderate-to-high risk were randomised to preventive treatment of thromboembolism, either with warfarin or the oral direct thrombin inhibitor ximelagatran. The survival of users and non-users of digitalis was investigated.

Results: At baseline, 53.4% of the study population used digitalis, and these patients had a higher mortality than non-users (255/3911 (6.5%) vs 141/3418 (4.1%), p<0.001; hazard ratio (HR) = 1.58 (95% CI 1.29 to 1.94)). Digitalis users also had more baseline risk factors. After multivariate risk factor adjustment, the increased mortality persisted (p<0.001; HR = 1.53 (95% CI 1.22 to 1.92 vs 1.23 to 1.92)).

Conclusions: The results suggest that digitalis, like other inotropic drugs, may increase mortality. This may be concealed in heart failure, but be revealed in patients with AF, who need the rate-reducing effect of digitalis, but do not benefit much from an increased inotropy. Cautious interpretation of the data is mandatory since the patients were not randomised with respect to digitalis use.

View Full Text

Statistics from

The American College of Cardiology/American Heart Association/European Society of Cardiology practice guidelines advise the use of digitalis for heart rate control in persistent or permanent atrial fibrillation (AF) as a class I indication.1 New inotropic drugs are desired since the present ones relieve symptoms in heart failure, but often at the cost of increased mortality.2 Fortunately, digitalis has been extensively studied in heart failure, and the largest study, the Digitalis Investigation Group (DIG) study, concludes that there is no evidence of excess mortality caused by digitalis in patients with heart failure.3 In AF, however, safety data are sparse. The ventricular rate is decreased by digitalis, and this may increase exercise capacity and quality of life. These patients often have no heart failure, and may not benefit from the positive inotropic effects of digitalis. If digitalis should have serious adverse effects, the AF patient group could fare worse when given this treatment than patients with heart failure in whom clinical improvement with the drug could counterbalance lethal adverse effects, if such were present.

In the Stroke Prevention using an ORal Thrombin Inhibitor in atrial Fibrillation (SPORTIF) III4 and V5 studies, patients with moderate-to-high risk AF were treated either with dose-adjusted warfarin or the thrombin inhibitor ximelagatran. Before study closures, a substudy was planned to look at the potential effect of concomitant digitalis use upon mortality.


Design of the SPORTIF III and V studies

The multinational randomised SPORTIF III and V studies were designed to study the efficacy of ximelagatran compared with warfarin in the prevention of thromboembolic complications in patients with moderate-to-high risk AF. The SPORTIF III study4 was open labelled and SPORTIF V5 was double blind. A total of 7329 patients participated.

Statistical methods

Analyses of all-cause mortality and the primary outcome (strokes and systemic embolic events) were performed according to the patients’ baseline use or non-use of digitalis. Time-to-first event was analysed using Cox proportional hazards modelling, both with baseline digitalis use as the only covariate and then adjusting the effect for factors thought to have an impact on the outcome. The factors used for adjustment were all measured at baseline: study treatment (ximelagatran or warfarin), trial, age, race, gender, weight, body mass index, systolic blood pressure, diastolic blood pressure, prior stroke or transient ischaemic attack, prior systemic embolic event, diabetes mellitus, history of hypertension, history of coronary heart disease, years since first AF diagnosis, smoking status, baseline aspirin use, baseline β-blocker use and left ventricle dysfunction. The proportionality assumption was checked for all covariates by testing the interaction term of each covariate with time. The adjusted results for digitalis, presented below, are from the model including all covariates above together with significant covariate/time interaction terms.

Adjustment for multiple testing was not done because of the exploratory nature of the analysis—digitalis use would need to be subject to a randomised comparison for results to be valid. To compare baseline characteristics for digitalis users with those of non-users, the Kruskal–Wallis test was used for continuous variables and the Pearson χ2 test for categorical variables.


Baseline and mortality data were obtained for all patients. Tests for interactions were performed using the interaction terms of the covariate with both study (SPORTIF III and V) and treatment group (ximelagatran/warfarin) included in the Cox proportional hazards model. Neither baseline use of digitalis, nor the effect of digitalis on outcomes differed significantly between the studies, or between the treatment groups. Hence the data were pooled over both trials and study treatments.

Baseline data

Table 1 compares digitalis users (53.4% of the patients) and non-users at baseline.

Table 1 Baseline characteristics by digitalis use at baseline*

Clinical outcomes

Thromboembolic events

Strokes and systemic embolic event (intention to treat) were the primary events of the SPORTIF III and V studies. The rate of primary end points (all strokes and systemic embolic event) among digitalis users was 1.76%/patient-year (105 patients with events, follow-up of 5982 patient-years) vs 1.50%/patient-year in non-users (79 events, follow-up of 5251 patient-years) (p = 0.282; hazard ratio (HR) = 1.17 (95% CI 0.88 to 1.57)). After adjusting for all covariates above, baseline digitalis use was still not significant for the primary end point (p = 0.411; HR = 1.14 (95% CI 0.83 to1.57 vs 0.84 to 1.57).

All-cause mortality

Two hundred and fifty-five deaths occurred over 6047 patient-years (4.22%/patient-year) for digitalis users vs 141 deaths over 5300 patient-years (2.66%/patient-year) for non-users (p<0.001; HR = 1.58 (95% CI 1.29 to 1.94)) (fig 1). After adjusting for all covariates above, baseline digitalis use was still significant for the outcome of all-cause mortality (intention to treat) (p<0.001; HR = 1.53, 95% CI 1.22 to 1.92 vs 1.23 to 1.92). Figure 2 shows that digitalis use was associated with a worse outcome in almost all subgroups. This was even the case among those who had baseline left ventricular dysfunction (NS, data not shown).

Figure 1 Kaplan–Meier plot of time to event (all-cause mortality) in 7329 patients with moderate-to-high risk atrial fibrillation. Digitalis users had a significantly higher event rate than non-users (p<0.0001; hazard ratio, 1.58 (95% CI 1.29 to 1.94)).
Figure 2 Hazard ratio with 95% confidence intervals for all-cause mortality in baseline subgroups among 7329 patients with moderate-to-high risk atrial fibrillation. TIA, transient ischaemic attack.

Table 2 shows that total mortality was highest among digitalis users who were not receiving a β-blocker and lower among those receiving both drugs. The interaction between digitalis and β-blocker usage was not significant (p = 0.62 for the interaction effect), which indicates that β-blocker usage does not have an impact on the effect of digitalis on all-cause mortality.

Table 2 Effects of β-blocker use on primary end points (stroke and systemic embolic events (SEE)) and all-cause mortality

Digitalis users were more likely to have a vascular-related death than non-users (p = 0.004). The incidence of strokes, systemic embolic episodes and major non-central nervous system bleeds did not differ significantly, but myocardial infarcts occurred more often in digitalis users (8.6% vs 5.7%, p = 0.026), as did “other vascular deaths” that included failure and sudden death (21.6% vs 20.6%, p = 0.027).


The main and disturbing finding of this study was that the death rate among patients with AF who were digitalis users was significantly higher than that among non-users (6.5% vs 4.1%, p<0.001; HR = 1.58 (95% CI 1.29 to 1.94)). A number of baseline risk factors were, however, more prevalent in digitalis users, but baseline digitalis use was still significant for the outcome of all-cause mortality after adjustment (p<0.001; HR = 1.53 (95% CI 1.22 to 1.92)).

The present study has several strong features. It is large, including 7329 patients of whom 53% were digitalis users. The patients were at moderate-to-high risk for thromboembolic events, usually with high age and/or manifest cardiovascular disease in addition to their AF. The number of deaths was accordingly high (n = 396), justifying subgroup analysis. This analysis was planned before study closure and hence was not a post hoc analysis. The major weakness is that the study was not randomised for the use of digitalis. At baseline, digitalis users had significantly more risk factors than non-users. Even if we adjusted the results for the influence of a series of risk factors, those who were prescribed digitalis may have additional confounding risk factors not accounted for in our analysis. Another important limitation is that we lack data on symptoms and mechanisms of heart failure and on concomitant drug use, as well as data on the indication for prescription of digitalis. Socioeconomic status and the type of care provider should also have been included in the analysis, had they been available. This type of analysis can, however, only demonstrate associations, and cause and effect can only be demonstrated with randomised clinical trials.

The use of β-blocker treatment might be a possible confounder. Since both digitalis and β-blockers are used for rate control, those who were well controlled with digitalis would not receive the β-blocker with its cardioprotective effects, and those receiving β-blocker treatment would less often need digitalis in addition. However, our analysis argues against such an interaction.

The present results fit remarkably well with the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) study of rhythm versus rate control in AF.6 There, 2796 patients were analysed, also with respect to drug treatment. The design was a post hoc analysis of the non-randomised use of digitalis in AF. Despite multivariate risk adjustments, digoxin users had a relative risk of death of 1.42 (99% CI 1.09 to 1.86). The authors, however, did not focus on this in their discussion. Patients with heart failure fare better. A neutral effect upon survival in heart failure was found in the largest study, the randomised placebo-controlled DIG trial.3 In the non-randomised post-infarct β-Blockade Heart Attack Trial (BHAT), digoxin users had a higher mortality than non-users; however, multiple risk factor adjustment removed the difference.7

If digitalis does have prothrombotic effects, patients with AF may be particularly vulnerable since thromboembolism is their major threat, affecting up to 10%/year.1 This might be easily overlooked in a population prone to thromboembolism and atherosclerosis. Unfortunately, the classification system of deaths in the SPORTIF studies was rather crude. However, the increase in deaths was confined to vascular causes, in particular to myocardial infarctions and “other vascular disease”, including heart failure and sudden death.

Proarrhythmic events due to digitalis?

In the SPORTIF studies there was no recording of possible digitalis overdose symptoms and no measurement of drug concentrations in blood. Although antiarrhythmic drugs can induce deadly proarrhythmias, such effects have not been reported for digitalis at normal serum levels.8 However, even though the drug levels for 4–10% of digoxin users are reported to be too high, proarrhythmia is suspected in only 0.25%.8 9 The incidence of digoxin-induced arrhythmia was reported to be dose related: 10% at a level of 1.7 ng/ml and 50% at 2.5 ng/ml.10 In the DIG study, death rates were higher in women11 and, further, the higher the dose of digoxin given, the higher was the mortality in heart failure.12 In a recent analysis of the Studies of Left Ventricular Dysfunction (SOLVD) data, however, no excess mortality was found in female digitalis users.13 Similarly, in the present study we did not find an increased risk in female compared with male digitalis users. Adverse effects of digitalis include atrioventricular conduction disturbances, and accelerated atrial activity and malignant ventricular arrhythmias also occur.14 There is, however, little reason to think this is a serious problem with digitalis use since, if so, dizzy spells should often precede syncope and cardiac arrest. This is not the clinical picture.

Platelet activation by digitalis?

Digitalis inhibits the membrane-bound Na+/K+-ATPases, thereby finally increasing intracellular Ca2+.15 16 Also, blood platelets are influenced: ouabain causes an increased aggregation after agonist stimulation in aggregometer studies.17 Whereas digitalis can activate platelets in vitro, that was not the case in our in vivo study in healthy young volunteers.18 In AF, however, platelets are activated.19 20 In patients with AF who are not anticoagulated, the platelet contribution to thromboembolism may have clinical importance, as shown by the (moderate) protection provided by aspirin,21 and further supported by studies demonstrating increased platelet activation in patients with AF who have experienced embolic events.22 23 Digitalis might influence preactivated platelets in AF even if normal platelets are not affected; this has recently been demonstrated.24 Digitalis did not increase the number of thromboembolic events in the present study. This was, however, as expected since all our patients received anticoagulation treatment, either with warfarin or ximelagatran.

What are the clinical implications?

The prevalence of AF is high, especially among the elderly.25 In our study, more than half the patients used digitalis. If digitalis does have serious adverse effects, the number of patients affected might be large. The question of the potentially serious adverse effect of digitalis in AF should be approached by a prospective randomised study. For ethical reasons, a randomised placebo-controlled withdrawal study may be preferable.


This large, prospective, observational analysis showed that when treated with digitalis, patients with moderate-to-high risk AF had a hazard ratio of 1.58 for mortality, and this excess mortality persisted after multifactorial risk adjustment. The study suffers, however, from two major limitations: it is not randomised for the use of digitalis and the data provide no insight into the pathophysiological mechanisms involved.


View Abstract


  • Competing interests: None declared

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles