The effect of digoxin on mortality and morbidity in patients with heart failure

Summary

The efficacy of digoxin, particularly its effect on mortality, in patients with congestive heart failure in normal sinus rhythm has been debated for years. This large, randomized, placebo-controlled study was designed to look at the effect of digoxin on mortality and on hospitalization.

Methods

The study was divided into a main trial (6800 patients with CHF and an ejection fraction of 0.45 or less) and an ancillary trial (988 patients with CHF but with an EF greater than 0.45). The results presented here concern the main trial.

• Subjects: Patients with congestive heart failure (by symptoms, signs or chest x-ray), in normal sinus rhythm, with an EF of 0.45 or less (except for the ancillary trial, as noted above). Treatment with digoxin was not a reason for exclusion. Patients were enrolled from 302 centers in the United States and Canada.

• Intervention: Between February, 1991 and August, 1993 patients were randomized to either digoxin (dose determined by a multi-factorial algorithm) or placebo. Patients were followed up every four months. The use of ACE inhibitors was encouraged. If CHF could not be controlled by adjusting other therapy, the study medication could be discontinued and digoxin given. Follow-up ended on December 31, 1995.

• Endpoints: The primary endpoint was mortality. Secondary outcomes included cardiovascular mortality, death from worsening CHF, hospitalization for CHF and hospitalization for other causes.

  Pre-defined subgroup analyses included: EF (<0.25 and >0.25), heart size by CXR, ischemic vs. non-ischemic CHF, NYHA functional class and digoxin use prior to randomization.

• Subjects: In the main trial, 3397 patients were assigned to digoxin, 3403 to placebo. Mean duration of follow-up was 37 months. Baseline characteristics were well matched between the two groups. The mean EF was 0.28, median duration of CHF was 16.5 months. 78% were men, 27% were over the age of 70. NYHA class I -- 13%; class II -- 54%; class III -- 31%; class IV -- 2%. Ischemic cardiomyopathy was the cause of CHF in 70%. ACE inhibitors were prescribed in 94% of patients, diuretics in 82%. The daily digoxin dose was 0.25 mg in 70%.

• Mortality: There was no significant difference in the number of deaths between the two groups -- 1181 in the digoxin group (34.8%), vs. 1194 in the placebo group (35.1%). Most of these deaths were cardiovascular (1016 and 1004, respectively). There was a strong trend towards lower mortality from worsening CHF in the digoxin group, however (394 vs. 449, p=0.06). Although not specified in advance, the number of deaths from "other cardiac causes" (arrhythmia, cad, cardiac surgery) was higher in the digoxin group (508 vs. 444, P=0.04).

• Hospitalization: In the digoxin group, there were fewer hospitalizations for worsening heart failure (1927 vs. 2553) as well as fewer patients hospitalized for worsening heart failure (910 vs. 1180; RR 0.72).

  There was no significant difference in the number of patients hospitalized for cardiac arrest or ventricular arrhythmias (142 vs. 145).

  There were fewer all-cause hospitalizations in the digoxin group -- 64.3% of digoxin treated patients were hospitalized for any reason, vs. 67.1% in the placebo group.

• Subgroup analyses: The effect of digoxin in the prespecified subgroups is reported in terms of the combined outcome of death from CHF or hospitalization for CHF. In the overall study population, the RR for this outcome was 0.75 (95% CI 0.69 to 0.82), digoxin vs. placebo. Subgroup analysis revealed that the beneficial effect of digoxin was more pronounced in the "sicker" patients and in those with nonischemic CHF:
  • EF 0.25-0.45: RR 0.80
    EF <0.25: RR 0.68

  • NYHA class I or II: RR 0.78
    NYHA class III or IV: RR 0.70

  • Ischemic etiology: RR 0.79
    Nonischemic etiology: RR 0.69

• Digoxin: Mean serum digoxin levels at one month were between 0.8 and 1.0 for all digoxin doses (between 0.125 and 0.5 mg daily). At the final visit, 71% of patients randomized to digoxin were taking the study drug and an additional 10% were taking open-label digoxin; 68% of those randomized to placebo were taking placebo, and 16% were taking open-label digoxin.

• Ancillary trial: In the ancillary trial (patients with EF > 0.45), the results were similar to those of the main trial described above.

Comment

A note on combining endpoints

One objection I have to the way the results are presented here is the use of the combined endpoints "death or hospitalization for heart failure" and "death from heart failure or hospitalization for heart failure".

Combining endpoints always entails a loss of information. Nevertheless, it sometimes makes good sense to combine specific endpoints. When two or more events are of similar clinical importance and often occur as a trade-off or cannot be reliably distinguished, combining them makes perfect sense. Thus, when looking at various causes of death, the combined endpoint of "all-cause mortality" makes good sense, since it is not always possible to accurately determine the actual cause of death, and the clinical impact of death from one cause is very similar to that of death from another cause. What counts is "dead or alive"; whether dead from cardiovascular disease or another cause matters little to the patient. Another example of a useful combining of endpoints is "death or severe deficit at 6 months" when looking at thrombolytic therapy for stroke. Thrombolytic therapy for stroke might convert some patients who would have had a severe deficit to either a minor deficit or to death from intracerebral bleeding. Since a severe deficit after a stroke is an outcome that is at least in the same order of magnitude of severity as death, it makes sense to consider "death or severe disability" as a combined endpoint.

On the other hand, death is a substantially more serious outcome than non-fatal myocardial infarction, for example. Many studies of lipid-lowering therapy look at this combined endpoint (death or non-fatal MI), which makes little sense, in my opinion. Ten deaths plus one MI is not comparable to ten MI's plus one death. I suspect that the reason this combined endpoint is used is because many trials have shown a significant reduction in MI without being able to show a reduction in mortality. By using the combined endpoint, "death or MI", the beneficial effect on MI can be "extended" to the word death, making the results seem more impressive. But this is spin-doctoring. We are better served with the data on death and MI presented individually, not aggregated.

Similarly, in the study being reviewed here, the combined outcome of "death or hospitalization for heart failure" makes little sense. Ten deaths plus one hospitalization is not comparable to ten hospitalizations plus one death. The authors note that, in patients with more severe congestive heart failure, the effect of digoxin on the combined endpoint of "death from CHF or hospitalization for CHF" was stronger than in the overall study population. But this does not tell us whether there was a significant effect on mortality in this subgroup, or whether the stronger effect was due to a greater reduction in hospitalizations only. Perhaps the authors will publish another paper entitled "the effect of digoxin on mortality in patients with severe congestive heart failure", and we will then have the answer to this question.

February 27, 1997

Reader comments

Regarding combined endpoints for heart failure, I agree with you. However, in the example that you mentioned, death and non-fatal MI, we can also face this association in a different way: if death due to coronary disease and non-fatal MI share the same pathophysiological mechanism, then it would make sense to have them combined.

It is true that death and non-fatal MI often share the same pathogenetic mechanism, and from this standpoint it makes sense to combine them as an endpoint. My point is that, from the standpoint of clinical importance, they are so different that the combination can be misleading.

Example: treatment A leads to 10 deaths and 2 nonfatal MI's; treatment B leads to 2 deaths and 10 nonfatal MI's (out of 100 patients, say). Even if all of the deaths are due to MI's, would we say that the treatments are equivalent? Yes, based on the combined outcome and only looking at the incidence of MI (12 in both groups); no, based on considering the endpoints separately (10 deaths vs 2). -- mwj

Submit a comment about this article

References

Return to Journal Club home page