Association of Serum Digoxin Concentration and Outcomes in Patients With Heart Failure FREE

The Digitalis Investigation Group (DIG) trial, a randomized, double-blinded, placebo-controlled study evaluating the efficacy of digoxin therapy for patients with heart failure, found that digoxin therapy had no effect on mortality but modestly reduced hospitalizations due to worsening heart failure.¹ Patients were randomly assigned to digoxin with the objective of achieving a therapeutic serum digoxin concentration (SDC) in the range of 0.5 to 2.0 ng/mL.^1- 2 However, this range was defined to avoid concentrations associated with digoxin toxicity and was not specified as a result of evaluations of the relative efficacy of different SDCs.³

Since the publication of the DIG trial, concerns have been raised regarding the relative efficacy of SDCs greater than 1.0 ng/mL. Higher SDCs in this range have been associated with improved left ventricular function^4- 6 but have not shown a beneficial effect on neurohormonal function, hemodynamic profile, or exercise tolerance.^4- 5,7 Consequently, current clinical practice guidelines of the American College of Cardiology and the American Heart Association state "large doses of digoxin may not be more effective than small doses in the treatment of heart failure,"⁸ and some textbooks suggest that clinicians pursue target SDCs lower than the upper therapeutic range accepted in the DIG trial.^9- 10

Published evaluations of the efficacy of various SDCs have relied on intermediate end points, such as neurohormonal function or risk of worsening heart failure,^4- 5,7 rather than risks of hospitalization or mortality.^4- 5,7 Prior studies have not accounted for clinical characteristics, such as age or renal function, that may confound the relationship between SDC and patient outcomes.¹¹ The 4 published studies of SDC and the efficacy of digoxin therapy have been based on fewer than 210 total patients, drawn from single centers, and evaluated outcomes after digoxin withdrawal, thereby limiting their generalizability to other patient populations and their power to detect differences in efficacy across a range of SDCs.^4- 7 Accordingly, we undertook a post hoc evaluation of the DIG trial to evaluate the association of SDC, risk of mortality, and risk of hospitalization among patients with heart failure and left ventricular dysfunction.

We obtained a public-use copy of the database of the DIG trial by submitting a written request to the National Heart, Lung, and Blood Institute. None of the authors were members of the DIG Investigators, and thus none were involved in the conduct or initial analysis of the trial.

The design and results of the DIG trial have been previously reported.^1- 2 The study was conducted from August 1991 to December 1995. Patients with stable, clinically confirmed heart failure, a left ventricular ejection fraction (LVEF) of 45% or less, and were in sinus rhythm were enrolled between August 1991 and February 1993. Patients were randomly assigned to digoxin or placebo based on a published algorithm that accounted for age, sex, weight, and renal function.¹² Modifications in initial dosing were permitted to allow for prior digoxin dosages and concomitant pharmacotherapy.

Because we have previously identified a sex and digoxin therapy interaction in the DIG trial¹³ and only a small number of women had SDCs measured, we restricted our main analysis to men enrolled in the main DIG trial (N = 5281) (Figure 1). Patients were divided into those randomly assigned to digoxin (n = 2642) and those randomly assigned to placebo (n = 2639). Serum digoxin concentrations were measured in a random group of men in the digoxin arm 1 month after randomization (n = 1409). We selected patients randomly assigned to digoxin who had valid SDCs assessed at 1 month after randomization with blood samples drawn at least 6 hours after their previous digoxin dose (n = 1171). Among the 2639 patients randomly assigned to placebo, we excluded 28 subjects who died in the month following randomization to replicate the same selection process for patients randomly assigned to digoxin. Thus, the study sample consisted of 3782 patients; 2611 randomly assigned to placebo and 1171 randomly assigned to digoxin.

Patients randomly assigned to digoxin were further divided into 3 groups based on their SDC measured 1 month after randomization using published cut points of clinical interest: 0.5 to 0.8 ng/mL, 0.9 to 1.1 ng/mL, and 1.2 ng/mL and higher.^3,7 No patient in the DIG trial database had a reported SDC that was less than 0.5 ng/mL (Table 1).

The primary end point of the DIG trial was all-cause mortality within 37 months (range, 24-48 months) of randomization.¹ We also examined 3 prespecified secondary outcomes in the DIG trial²—death due to cardiovascular causes, death due to worsening heart failure, and hospitalization for worsening heart failure—as well as all-cause hospitalization and hospitalization for suspected digoxin toxicity. Causes of death and hospitalization were recorded by local investigators who were blinded to patients' treatment assignments.

Baseline characteristics were compared between patients randomly assigned to placebo and the 3 SDC groups using χ² and Wilcoxon rank sum tests. Kaplan-Meier curves were plotted for all-cause mortality for patients in the placebo and the 3 SDC groups. χ² Analyses were conducted to compare rates of mortality and hospitalization between patients randomly assigned to placebo and the 3 SDC groups and to test for trends in outcomes across the SDC groups.

A multivariable Cox proportional hazards analysis was conducted to determine the independent association of SDC and all-cause mortality. Patient characteristics incorporated in our multivariable analysis included age, race, body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters), LVEF, New York Heart Association class, cardiothoracic ratio, number of heart failure signs and symptoms, systolic blood pressure, heart rate, estimated glomerular filtration rate, duration of heart failure, primary cause of heart failure, history of myocardial infarction, angina, diabetes, hypertension, prior use of digoxin, and use of potassium-sparing diuretics, all other diuretics, angiotensin-converting enzyme inhibitors, nitrates, hydralazine, and other vasodilators.¹⁴ Multivariable analyses were repeated for the end points of death due to cardiovascular causes, death due to worsening heart failure, all-cause hospitalization, and hospitalization due to worsening heart failure.

Analyses were repeated modeling SDCs as a continuous variable with risk-adjusted all-cause mortality rates estimated using fractional polynomial modeling.¹⁵ Secondary analyses also were conducted by excluding patients with an SDC greater than 2.0 ng/mL, which was outside of the DIG trial's prespecified therapeutic range. Although the sample size did not confer adequate power, we also evaluated the association of SDC and all-cause mortality among the 330 women who had valid SDCs evaluated at 1 month and compared their outcomes with the 756 women in the DIG trial randomly assigned to placebo.

The study was approved by the Yale University School of Medicine Human Investigation Committee. All analyses were conducted using SAS 8.02 statistical software (SAS Institute Inc, Cary, NC).

Of the 1171 men with SDCs assessed at 1 month following randomization, 572 (49%) patients had an SDC of 0.5 to 0.8 ng/mL, 322 (27%) patients had an SDC of 0.9 to 1.1 ng/mL, and 277 (24%) patients had an SDC of 1.2 ng/mL or greater. Patients with higher SDCs were older, had greater numbers of heart failure signs or symptoms, and had lower median BMIs and estimated glomerular filtration rates. A greater proportion of patients with higher SDCs were in New York Heart Association class III/IV and had used non–potassium-sparing diuretics and nitrates. There were no clinically meaningful differences observed in LVEF, primary cause of heart failure, prior use of digoxin, medical history, or digoxin dosages of 0.375 mg/d or higher among patients with different SDCs (Table 1).

There was no difference in all-cause mortality among patients randomly assigned to placebo and patients randomly assigned to digoxin who had SDCs assessed (36.2% placebo vs 36.6% digoxin, P = .80). Kaplan-Meier survival curves, however, differed based on patients' SDCs (P<.001) (Figure 2). Higher SDCs were associated with higher crude all-cause mortality (P = .006 for trend) and cardiovascular mortality (P<.001 for trend), but not mortality due to worsening heart failure (P = .68 for trend) (Table 2).

The effectiveness of digoxin therapy compared with placebo varied based on patients' SDCs. Patients with SDCs of 0.5 to 0.8 ng/mL had a 6.3% (95% confidence interval [CI], 2.1%-10.5%) lower rate of all-cause mortality, a 3.7% (95% CI, 0.4%-7.7%) lower rate of cardiovascular mortality, and a 4.7% (95% CI, 2.1%-7.3%) lower rate of mortality due to worsening heart failure compared with patients randomly assigned to placebo. Patients with SDCs of 0.9 to 1.1 ng/mL had similar rates of all-cause mortality, cardiovascular mortality, and mortality due to worsening heart failure as patients randomly assigned to placebo. In contrast, patients with SDCs of 1.2 ng/mL and higher had an 11.8% (95% CI, 5.7%-18.0%) higher all-cause mortality rate and an 11.5% (95% CI, 5.4%-17.5%) higher cardiovascular mortality rate than patients randomly assigned to placebo, but no increased risk of mortality due to worsening heart failure (1.9%, 95% CI -2.6% to 6.3%) (Table 2).

Patients with an SDC of 0.5 to 0.8 ng/mL remained at lower risk of all-cause mortality (hazard ratio [HR], 0.80; 95% CI, 0.68-0.94) and mortality due to worsening heart failure (HR, 0.66; 95% CI, 0.49-0.89) and showed a trend toward a lower risk of cardiovascular mortality (HR, 0.86; 95% CI, 0.72-1.02) after multivariable adjustment. (Table 3) In contrast, patients with an SDC of 0.9 to 1.1 ng/mL had mortality risks comparable with those of patients randomly assigned to placebo. Multivariable adjustment attenuated the increased mortality risk among patients with an SDC of 1.2 ng/mL and higher, but a trend toward increased risk of all-cause and cardiovascular mortality persisted.

The association of SDC and all-cause mortality was similar when plotting SDC as a continuous variable (Figure 3) and when excluding patients with an SDC greater than 2.0 ng/mL (results not shown).

Higher SDCs were associated with higher crude rates of all-cause hospitalization, hospitalization for worsening heart failure, and hospitalization for suspected digoxin toxicity (Table 2). Patients with an SDC of 0.5 to 0.8 ng/mL had a 5.9% (95% CI, 1.5%-10.2%) lower crude rate of all-cause hospitalization than patients randomly assigned to placebo. This relationship primarily reflected a 14.0% (95% CI, 10.2%-17.8%) lower absolute rate of hospitalization for worsening heart failure. Patients with an SDC of 0.9 to 1.1 ng/mL and 1.2 ng/mL and higher had all-cause hospitalization rates comparable with patients randomly assigned to placebo. Hospitalization for suspected digoxin toxicity was higher among patients with an SDC of 1.2 ng/mL and higher (2.4%; 95% CI, 0.3%-4.5%) compared with patients randomly assigned to placebo. Rates of hospitalization for suspected digoxin toxicity were comparable with placebo in the other SDC groups (Table 2). Findings were similar after multivariable adjustment, although the higher risk of hospitalization due to worsening heart failure was reduced among patients with SDCs of 0.9 to 1.1 ng/mL and 1.2 ng/mL and higher (Table 3).

Of the 330 women who had SDCs measured at 1 month following randomization, 131 (39.6%) had an SDC of 0.5 to 0.8 ng/mL, 96 (29.1%) had an SDC of 0.9 to 1.1 ng/mL, and 103 (31.1%) had an SDC of 1.2 ng/mL and higher. Women also exhibited a pattern of increased all-cause mortality rates at higher SDCs (0.5-0.8 ng/mL, 25.2%; 0.9-1.1 ng/mL, 33.3%; ≥1.2 ng/mL, 41.8%; P = .009 for trend). Compared with the 756 women randomly assigned to placebo, higher SDCs continued to be associated with all-cause mortality after multivariable adjustment (0.5-0.8 ng/mL: HR, 0.89; 95% CI, 0.61-1.29; 0.9-1.1 ng/mL: HR, 1.18; 95% CI, 0.81-1.72; and ≥1.2 ng/mL: HR, 1.26; 95% CI, 0.90-1.76), although the findings were not statistically significant.

Our results indicate that the effectiveness of digoxin in the DIG trial varied according to patients' SDCs, such that higher SDCs were associated with higher rates of mortality and hospitalization. Furthermore, our analyses suggest that previously accepted SDCs of 1.2 ng/mL and higher^3,16- 17 may be harmful and that currently recommended therapeutic SDCs of approximately 1.0 ng/mL^9,18 may not provide any clinical benefit compared with placebo. Our findings suggest that an SDC of 0.5 to 0.8 ng/mL likely constitutes the optimal therapeutic range for digoxin therapy among men with stable heart failure and left ventricular dysfunction.

Questions concerning the relative efficacy of digoxin therapy as a function of SDC were raised when the DIG trial was first published.¹⁹ The discordance between our findings and previous evaluations of SDCs and patient outcomes^4- 7 may be attributable to several factors. Previous attempts to examine the efficacy of SDCs and patient outcomes focused on intermediate outcomes rather than major clinical end points, and thus only examined SDCs in terms of relative efficacy (ie, changes in left ventricular contractility), not possible harm (ie, mortality).^4- 7 Furthermore, these studies were likely unable to detect variations in outcomes based on SDCs because they included small numbers of patients observed during brief follow-up periods.^4- 7 Prior studies also did not account for the sex and digoxin therapy interaction we identified in a prior evaluation of the DIG trial,¹³ indicating earlier findings concerning SDCs may represent an artificial combination of sex-specific variations in digoxin's effect.

The association of SDC and patient outcomes may reflect the dissociated neurohormonal and inotropic effects of digoxin observed at different concentrations.^11,20- 22 Neurohormonal modulation is believed to contribute to digoxin's symptomatic benefits among patients with stable heart failure in sinus rhythm.^5,20,22- 24 These benefits are achieved at an SDC of 0.7 to 0.8 ng/mL and do not appear to increase at higher SDCs.³ In contrast, digoxin-associated harms are believed to reflect inotropic-associated increases in myocardial oxygen consumption and arrhythmogenesis at higher serum concentrations.^5,25- 26 Thus, our findings are consistent with the hypothesis that digoxin provides a neurohormonal benefit with minimal inotropic-associated risk at lower SDCs.^{3,20- 21,25- 26} As the SDC increases, the inotropic action of digoxin becomes stronger and begins to offset the therapeutic benefits provided by neurohormonal modulation and may account for the increased risk of mortality at concentrations of 1.2 ng/mL and higher observed in our study and previous evaluations.^27- 28 Thus, an SDC of 0.5 to 0.8 ng/mL may represent the most clinically efficacious balance of digoxin's competing neurohormonal and inotropic effects.

Our findings challenge the belief that digoxin is not associated with mortality among patients with heart failure in sinus rhythm and suggest that the neutral association of digoxin and all-cause mortality reported in the main DIG trial publication likely obscured an association between SDC and mortality. Digoxin-associated mortality risks have previously been shown to be dose-dependent in patients recovering from acute myocardial infarction²⁹ and are similar to serum concentration–dependent mortality risks reported for other positive inotropic agents.³⁰ Digoxin, however, was associated with mortality reduction at an SDC of 0.5 to 0.8 ng/mL, suggesting that digoxin may offer a survival advantage in a narrow therapeutic window. Because we cannot definitively exclude the possibility of residual confounding, our findings should be considered hypothesis generating. A digoxin-associated mortality reduction may be plausible if the neurohormonal actions of digoxin observed at low serum concentrations were sufficient to reduce the rate of heart failure progression by attenuating neuroendocrine abnormalities associated with increased mortality.^22,31 Trials of other inotropic agents also have shown trends toward a reduction in mortality at low doses, but increased mortality at higher doses.^32- 33 Clearly, additional research is needed to confirm these findings and clarify the mechanisms underlying the digoxin-associated mortality effects we observed.

The association of SDC and mortality in women merits particular attention. We previously identified a sex and digoxin therapy interaction such that women randomly assigned to digoxin were at greater risk of all-cause mortality than women randomly assigned to placebo.¹³ Although the 131 women with an SDC of 0.5 to 0.8 ng/mL had a mortality risk point estimate below that of women randomly assigned to placebo, our analysis was insufficiently powered to determine whether digoxin use was associated with benefit, harm, or had a neutral effect for women in this or any SDC range. These data are consistent with our prior findings of a digoxin-associated harm in women¹³ in that they suggest there is no serum concentration range in which digoxin may provide a benefit to women nor a range in which a digoxin-associated increased mortality risk in women can be definitively excluded.

Our study has certain issues to consider. First, because our study is a post hoc analysis of nonrandomized treatment subgroups in the DIG trial, we cannot exclude the influence of unmeasured factors. However, our analysis accounted for BMI and estimated glomerular filtration rate, 2 of the primary determinants of digoxin absorption and excretion. We used SDCs measured at 1 month after randomization, thus excluding a small group of patients who died in the month following randomization. However, 1 month provided a reasonable time frame in which to ensure serum levels provided an accurate measure of patients' steady state SDCs. Furthermore, SDC levels were generally stable during the course of the DIG trial.¹ Serum digoxin concentrations may not accurately estimate digoxin tissue concentrations given variations in serum level measurement.³⁴ However, SDCs drawn at least 6 hours after the most recent dose in a cohort with a fixed digoxin dose for 30 days are unlikely to reflect inaccurate serum concentration measurements.

Data concerning the use of β-blockers was not collected as part of the DIG trial, making it unclear whether the overall association of SDC and patient outcomes would be altered by the routine use of β-blockers. A randomized controlled trial that incorporated contemporary management strategies would be necessary to definitively establish the current incremental value of digoxin, and of different digoxin concentrations, in patients with left ventricular systolic dysfunction. In the absence of such a trial, these data provide the best available evidence concerning the relative value of digoxin therapy in the heart failure armamentarium.

Our study is the first, to our knowledge, to examine the association of different SDCs and major clinical end points. Although based on a post hoc analysis of the DIG trial, our results suggest that an SDC of 0.5 to 0.8 ng/mL was associated with lower rates of mortality and hospitalization. Given that no study has demonstrated any substantive clinical benefit for higher SDCs, prudent practice would support an SDC of 0.5 to 0.8 ng/mL as a revised therapeutic range. The feasibility of achieving SDCs within this range in daily clinical practice is unclear. Only a randomized controlled trial can confirm this recommendation; however, we believe our data provide sufficient grounds for consideration of lower target SDCs for men with stable heart failure and left ventricular dysfunction.