Abstract
Background:
Digoxin has been shown to reduce heart failure hospitalizations with a neutral effect on mortality. It is unknown whether there is heterogeneity of treatment effect for digitalis therapy according to predicted risk of heart failure hospitalization.
Methods and Results:
We conducted a post hoc analysis of the Digitalis Investigator Group (DIG) studies, randomized controlled trials of digoxin versus placebo in participants with heart failure and left ventricular ejection fraction less than or equal to 45% (main DIG study, n=6800) or >45% (ancillary DIG study, n=988). Using a previously derived multistate model to risk stratify DIG study participants, we determined the differential treatment effect on hospitalization and mortality outcomes. There was a 13% absolute reduction in the risk of any heart failure hospitalizations (39% versus 52%; odds ratio [OR] = 0.58 (0.47 −0.71) in the digoxin versus placebo arms in the highest risk quartile compared to a 3% absolute risk reduction for any heart failure hospitalization (17% versus 20%; odds ratio [OR] = 0.84 (0.66– 1.08) in the lowest risk quartile. There were 12 fewer total all-cause hospitalizations per 100 person-years in the highest risk quartile compared with an increase of 8 hospitalizations per 100 person-years in the lowest risk quartile. There was neutral effect of digoxin on mortality in all risk quartiles and no interaction between baseline risk and the effect of digoxin on mortality (p=0.94).
Conclusions:
Participants in the DIG study at higher risk of hospitalization as identified by a multistate model were considerably more likely to benefit from digoxin therapy to reduce heart failure hospitalization.
Subject Codes: Heart Failure, Mortality/Survival, Cardiomyopathy, Quality and Outcomes
Introduction:
Heart failure is the leading cause of hospitalization in the United States among those 65 and older, leading to substantial morbidity and an estimated 20.9 billion dollars in direct costs each year1. In addition to the morbidity and costs associated with hospitalization, patients often experience a myriad of adverse events in the immediate post-discharge period and all-cause readmission rates after heart failure hospitalization approach 50% at 6 months2. Thus, interventions that safely decrease hospitalizations for heart failure have the potential to improve quality of life for patients and their caregivers as well as reduce overall health care costs.
The cardiac glycoside, digoxin, has been used in clinical cardiology since 1785. In patients with heart failure in sinus rhythm, the Digitalis Investigator Group (DIG) study showed a reduction in hospitalizations for worsening heart failure and all-cause hospitalizations with a neutral effect on overall mortality3. A meta-analysis of the 7 randomized controlled studies of digoxin for the treatment of heart failure showed a reduction in heart failure hospitalizations and no change in mortality4 and smaller studies have shown that digoxin also improved exercise tolerance, health related quality of life and symptoms, although no trials have been done with digoxin in the setting of current medical and device therapy5–7. Digoxin toxicity can lead to excess morbidity, mortality and hospitalizations, however this risk can be reduced by ensuring patients have achieved digoxin levels less than 1ng/ml8,9.
In situations where the risk:benefit ratio is not clearly weighted towards the treatment of all patients, a risk prediction model can be used to explore which patients may derive the greatest benefit from the therapy. We recently reported the derivation and external validation of the Tufts MC HF Risk Model10, a multistate model to predict heart failure hospitalization and death that accounts for the semi-competing risk of the two outcomes and allows for prediction of each outcome individually as well as the composite. We applied the Tufts MC HF Risk Model, to the DIG study population with the hypothesis that the patients at highest risk for heart failure hospitalization would derive the greatest benefit from digoxin to reduce hospitalization without increased mortality.
Methods:
Study population:
The Digitalis Investigator Group (DIG) main study enrolled 6800 patients with heart failure and a left ventricular ejection fraction less than or equal to 45%3 and an ancillary study enrolled 988 patients with heart failure and left ventricular ejection fraction >45% from 1991 to 199311. Full inclusion and exclusion criteria have previously been described.12 Briefly, in both studies participants had to be in normal sinus rhythm without significant renal insufficiency (creatinine > 3.0mg/dl), severe liver disease, hypokalemia (potassium <3.2mmol/l), hyperkalemia (potassium >5.5 mmol/l) or clinically relevant bradyarrhythmias without a pacemaker. Most participants (95%) were receiving angiotensin-converting-enzyme inhibitors but patients were not on routine beta-blocker or aldosterone antagonist therapy. In the main trial, digoxin was associated with reduction in heart failure hospitalizations (HR 0.72, 95% CI 0.66–0.79) with no effect on mortality (HR 0.99, 95% CI 0.91–1.07)3. In the ancillary trial, there was a non-significant trend towards reduced heart failure hospitalizations, similar in magnitude to that seen in the main study (HR 0.79, 95% CI 0.59–1.04), and no effect on mortality (HR 0.99, 95% CI 0.76–1.28)3,11. All patients gave informed consent for the DIG studies and the Tufts Medical Center Institutional Review Board approved this post-hoc analysis.
Prediction Model:
Tufts MC HF Risk Model derivation and external validation has previously been reported10. The Tufts MC HF Risk Model is an illness-death multistate model that uses routinely collected clinical and laboratory variables to predict the semi-competing risks of heart failure hospitalization and death in outpatients with heart failure.. In this model, all participants are in the initial state of prevalent heart failure and are at risk of a heart failure hospitalization (transition 1) or death without a preceding heart failure hospitalization (transition 2). In addition, those who were hospitalized for heart failure are also at risk for death after a heart failure hospitalization (transition 3). The model was developed from the Heart Failure Endpoint evaluation of Angiotensin II Antagonist Losartan (HEAAL) study cohort10,13, an international, multicenter, randomized trial of low dose (50mg) versus high dose (150mg) losartan in stable outpatients with heart failure with reduced ejection fraction. Four of the Tufts MC HF Risk Model predictors - serum sodium, weight, history of stroke and history of atrial fibrillation - were not available in the DIG cohort. Thus a “DIG-compatible” model with the remaining eight predictors - age, gender, New York Heart Association class, left ventricular ejection fraction, serum creatinine (mg/dl), systolic blood pressure (mmHg), history of diabetes and ischemic heart disease - was derived in the original HEAAL derivation cohort. Model performance of the DIG-compatible Tufts MC HF Risk Model with 8 variables was compared with the full Tufts MC HF Risk Model in the derivation cohort and then applied to the DIG cohort to calculate predicted probabilities.
Predictor and Outcome Ascertainment:
The 8 predictors in the DIG-compatible Tufts MC HF Risk Model were assessed in the DIG study cohort at the time of enrollment. LVEF was measured by radionuclide left ventricular angiography, echocardiogram or left ventricular contrast angiography. New York Heart Association class, history of diabetes and ischemic heart disease were all reported by the site investigator based on medical records and current symptoms. The DIG study captured all hospitalizations and deaths throughout the duration of follow up with reasons for hospitalization or death determined by the site investigator based upon medical record review and discussions with patient or family members. This analysis was restricted to the patient’s first heart failure hospitalization after enrollment. There was little missing data in the DIG cohort (<1%), thus a complete case analysis was used.
Digoxin treatment effect stratified by baseline risk:
The estimated treatment effect of digoxin on the risk of any heart failure hospitalization, total number of all-cause hospitalizations, death and suspected digoxin toxicity were analyzed by quartile of predicted risk of heart failure hospitalization (Transition 1 of the multi-state model). Risk was analyzed on the relative scale (odds ratios for heart failure hospitalization, death and suspected digoxin toxicity and risk ratios for total number of all-cause hospitalizations) and on the absolute scale (absolute risk difference). A formal test of interaction between the digoxin treatment effect over predicted outcome risk was performed to test whether risk (expressed as the linear predictor) influenced the odds ratio of the treatment effect. The pre-specified primary analysis was the effect of digoxin by quartile of predicted risk in the main trial (left ventricular ejection fraction ≤ 45%). A sensitivity analysis explored the effect of digoxin on the same outcomes in the pooled DIG cohort.
Statistical analysis:
Cox proportional hazards regression was used to model the effect of covariates on the cause-specific hazards of the 3-state transitions with separate (stratified) nonparametric baseline hazards for transitions into the hospitalization state and into the death state14. Discrimination was assessed by the c-index for time-to-event data with censoring15. For both the HEAAL derivation and DIG cohorts, the c-index was calculated for the overall model as well as for hospitalization and death separately. Calibration was assessed in the overall DIG cohort by dividing the cohort into deciles of predicted risk and comparing model predicted outcomes to observed outcomes at 4 years of follow-up. The extreme quartile risk ratio was calculated as the ratio of the observed outcome rate in the quartile with the highest predicted risk divided by the observed outcome rate in the quartile with the lowest predicted risk16,17. All analyses were performed using R version 3.1.2 and the mstate and survival packages for multistate modeling.
Results:
Study Population:
Baseline characteristics of the HEAAL model development and DIG main study and ancillary study cohorts are shown in Table 1. In general, baseline characteristics of the two cohorts were similar (Table 1). Consistent with years of enrollment, baseline medical and device therapy were different in the two studies with the HEAAL study having a high use of beta-blockers (72%), moderately high use of aldosterone antagonists (38%) and fairly low use of implantable cardioverter-defibrillators (5%) at the time of study enrollment and with none of these therapies used at the time of DIG study enrollment. All patients in HEAAL were on an angiotensin II receptor blocker and 94% of patients in the DIG main study were on angiotensin- converting enzyme inhibitor therapy.
Table One:
Baseline characteristics of the HEAAL derivation cohort, DIG Main study and DIG Ancillary Study cohorts
| Derivation Cohort (HEAAL) n=3786 | HEAAL missing data: no. (%) | DIG Main Study Cohort n=6800 | DIG main study missing data: no. (%) | DIG ancillary study Cohort n=988 | DIG ancillary study missing data: no. (%) | |
|---|---|---|---|---|---|---|
| Age (yrs): mean ±SD | 64 ± 12 | 0 (0) | 64 ± 11 | 0 (0) | 67 ± 10 | 0 (0) |
| Female: n(%) | 1143 (30) | 0 (0) | 1519 (22) | 0 (0) | 407 (41) | 0 (0) |
| Race/Ethnicity: | ||||||
| White: n(%) | 2321 (61) | 0 (0) | 5809 (85) | 0(0) | 851 (86) | 0 (0) |
| Asian: n(%) | 856 (22) | |||||
| Multi: n(%) | 354 (9) | |||||
| Hispanic: n(%) | 212 (5) | |||||
| Other: n(%) | 103 (3) | |||||
| Left ventricular ejection fraction (%): median (IQR) | 33 (27–37) | 1(0.02) | 29 (22–35) | 0(0) | 53 (49–60) | 0(0) |
| New York Heart Association Class: | ||||||
| I: n(%) | 2 (<0.01) | 1 (0.02) | 907 (13) | 6(0.1) | 196 (20) | 1(0.1) |
| II: n(%) | 2657 (69) | 3664 (54) | 573 (58) | |||
| III: n(%) | 1152 (30) | 2081 (31) | 206 (21) | |||
| IV: n(%) | 22 (0.5) | 142 (2) | 12 (1) | |||
| Ischemic Heart Disease: n(%) | 2456 (64) | 0 (0) | 4803 (71) | 18 (0.2) | 557 (57) | 5 (0.5) |
| Diabetes: n(%) | 1199 (31) | 0 (0) | 1933 (28) | 0 (0) | 285 (29) | 1 (0.1) |
| Atrial Fibrillation: n(%) | 1070 (28) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| Stroke: n(%) | 307 (8) | 0 (0) | N/A | N/A | N/A | N/A |
| Systolic Blood Pressure (mmHg): mean ±SD | 126 ± 18 | 1 (0.02) | 126 ± 20 | 3 (0.04) | 138 ± 21 | 0 (0) |
| Heart Rate (bpm): mean ±SD | 73± 12 | 5 (0.1) | 79± 13 | 8 (0.1) | 76± 12 | 0 (0) |
| Weight (kg): mean ±SD | 76 ± 17 | 5 (0.1) | N/A | N/A | N/A | N/A |
| Serum Sodium (mEq/L): mean ±SD | 140 ± 4 | 16 (0.4) | N/A | N/A | N/A | N/A |
| Creatinine (mg/dl): mean ±SD | 1.2 ± 0.3 | 16 (0.4) | 1.3 ± 0.4 | 0 (0) | 1.3 ± 0.4 | 0 (0) |
| ACEi/ARB: n (%) | 3786 (100) | 0 (0) | 6422 (94) | 0 (0) | 852 (86) | 0 (0) |
| Beta blocker: n (%) | 2758 (72) | 0 (0) | N/A | N/A | N/A | N/A |
| Aldosterone antagonist: n (%) | 1436 (38) | 0 (0) | N/A | N/A | N/A | N/A |
| Diuretics: n(%) | 3071 (80) | 0 (0) | 5325 (78) | 3 (0.04) | 751 (76) | 0 (0) |
| Aspirin: n (%) | 2138 (56) | 0 (0) | N/A | N/A | N/A | N/A |
| Statin: n (%) | 1511 (39) | 0 (0) | N/A | N/A | N/A | N/A |
| Calcium channel Blocker: n (%) | 450 (12) | 0 (0) | N/A | N/A |
N/A | N/A |
| ICD at time of enrollment n(%) | 172 (4.5) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| Study enrollment period | 2001–2005 | 1991–1993 | ||||
| Median follow-up | 4.7 years | 3.2 years | 3.2 years | |||
| HF hospitalization n(%) | 944 (25) | 2090 (31) | 197 (20) | |||
| Death n(%) | 1285 (34) | 2375 (35) | 231(23) |
HEAAL indicates Heart Failure Endpoint evaluation of Angiotensin II Antagonist Losartan trial; DIG, Digitalis Investigator Group trial; SD, standard deviation; IQR, interquartile range; ACEi, angiotensin-converting inhibitor; ARB, angiotensin II receptor blocker; ICD, implantable cardioverter-defibrillator; HF, heart failure
“DIG-compatible” Tufts MC HF Model Specification and Performance:
The multistate model transitions and states are shown in Supplemental Figure 1. The multivariate effects of the 8 predictors of the DIG-compatible model in the HEAAL derivation cohort are shown in Table 2 and are similar to the full model10. In the derivation cohort, model discrimination was similar with the DIG-compatible model as compared to the full model; discrimination of the DIG-compatible model was relatively well maintained on external validation in the DIG cohorts (Table 3). In all the cohorts, discrimination was better for the transition states into death than for heart failure hospitalization. Calibration plots by deciles of predicted risk show slight underestimation of the risk of both heart failure hospitalization and death in the older DIG cohort (supplemental Figure 2).
Table Two:
Transitional hazard ratios for Simplified Tufts MC HF Risk Model:
| Predictor | T1: Prevalent HF to HF hospitalization | T2: Prevalent HF to Death | T3: HF hospitalization to Death | |||
|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | |
| Age per 10 year increase | 1.13 | 1.07–1.21 | 1.34 | 1.24–1.44 | 1.12 | 1.03–1.22 |
| Female gender | 0.88 | 0.75–1.02 | 0.76 | 0.64–0.91 | 1.12 | 0.91–1.39 |
| NYHA III vs II | 1.64 | 1.43–1.88 | 1.45 | 1.25–1.69 | 1.28 | 1.07–1.52 |
| LVEF per 10% increase | 0.72 | 0.66–0.79 | 0.74 | 0.66–0.82 | 0.86 | 0.75–0.99 |
| Creatinine per 1mg/dl increase | 2.22 | 1.82–2.65 | 1.58 | 1.26–1.97 | 1.66 | 1.29–2.13 |
| Systolic Blood Pressure per 10mmHg increase | 0.88 | 0.85–0.92 | 0.95 | 0.91–0.99 | 0.91 | 0.86–0.95 |
| Diabetes mellitus | 1.45 | 1.27–1.66 | 1.39 | 1.20–1.62 | 1.06 | 0.89–1.27 |
| Ischemic etiology | 0.87 | 0.76–1.00 | 1.14 | 0.97–1.33 | 1.25 | 1.03–1.51 |
HR indicates hazard ratio, CI confidence interval, HF heart failure, NYHA New York Heart Association, LVEF left ventricular ejection fraction
Table Three:
Model discrimination in HEAAL derivation cohort and on external validation in the DIG cohorts
| Full model (12 predictors) in HEAAL derivation cohort | DIG compatible model (8 predictors) in HEAAL derivation cohort | DIG compatible model in main DIG trial (LVEF ≤ 45%) | DIG compatible model in pooled DIG trials (all LVEF) | |||||
|---|---|---|---|---|---|---|---|---|
| c-statistic | 95% CI | c-statistic | 95% CI | c-statistic | 95% CI | c-statistic | 95% CI | |
| Overall model | 0.72 | 0.66–0.70 | 0.71 | 0.69–0.73 | 0.69 | 0.67–0.70 | 0.69 | 0.68–0.70 |
| Death | 0.75 | 0.73–0.76 | 0.73 | 0.72–0.75 | 0.72 | 0.71–0.73 | 0.72 | 0.71–0.73 |
| HF hospitalization | 0.68 | 0.66–0.70 | 0.67 | 0.65–0.69 | 0.65 | 0.63–0.66 | 0.65 | 0.64–0.67 |
HEAAL indicates Heart Failure Endpoint evaluation of Angiotensin II Antagonist Losartan trial; DIG, Digitalis Investigator Group trial; CI, confidence interval; HF, heart failure
Outcome Risk Heterogeneity:
In the DIG main study cohort the overall heart failure hospitalization risk in the placebo arm was 35% over a median follow up of 3.1 years with a substantial difference in risk between the lowest (20%) and highest risk quartiles (52%) when the cohort was divided into quartiles of predicted risk of heart failure hospitalization (Table 4). The extreme quartile risk ratio of heart failure hospitalization was 2.6 for the placebo arm and 2.3 for the digoxin arm. Similar risk heterogeneity was seen for death with the lowest quartile having a risk of 17% and the highest quartile a risk of 56% and an extreme quartile risk ratio of 3.3. In a sensitivity analysis, similar results were seen in the pooled DIG cohort (Supplemental Table 1).
Table Four –
Effect of digoxin on HF hospitalization, Death, Total hospitalizations and suspected digoxin toxicity by quartile of predicted risk of HF hospitalization in the DIG Main Study
| Placebo Risk | DigoxinRisk | RR** | 95% CI | OR* | 95% CI | ARD | 95% CI | ||
|---|---|---|---|---|---|---|---|---|---|
| HF Hosp | Overall | 0.35 | 0.27 | 0.69 | 0.62–0.76 | 0.08 | 0.06–0.10 | ||
| Q1 | 0.20 | 0.17 | 0.84 | 0.66–1.08 | 0.03 | −0.01–0.06 | |||
| Q2 | 0.28 | 0.22 | 0.73 | 0.59–0.91 | 0.06 | 0.02–0.10 | |||
| Q3 | 0.39 | 0.29 | 0.63 | 0.52–0.78 | 0.10 | 0.06–0.15 | |||
| Q4 | 0.52 | 0.39 | 0.58 | 0.48–0.71 | 0.13 | 0.09–0.18 | |||
| *p-value interaction = 0.11 | |||||||||
| Death | Overall | 0.35 | 0.35 | 0.99 | 0.89–1.09 | 0.00 | −0.02 −0.03 | ||
| Q1 | 0.17 | 0.18 | 1.05 | 0.82–1.35 | −0.01 | −0.04−(0.03) | |||
| Q2 | 0.28 | 0.27 | 0.95 | 0.77–1.18 | 0.01 | −0.03–(0.05) | |||
| Q3 | 0.39 | 0.39 | 0.98 | 0.81–1.19 | 0.00 | −0.04–(0.05) | |||
| Q4 | 0.56 | 0.55 | 0.97 | 0.80–1.17 | 0.01 | −0.04–(0.06) | |||
| *p-value interaction = 0.94 | |||||||||
| Total Hosp/person-year | Overall | 0.68 | 0.64 | 0.94 | 0.91–0.96 | 0.04 | 0.02–0.07 | ||
| Q1 | 0.42 | 0.50 | 1.18 | 1.09–1.27 | −0.08 | −0.11–(−0.04) | |||
| Q2 | 0.59 | 0.54 | 0.92 | 0.85–0.98 | 0.05 | 0.01–0.09 | |||
| Q3 | 0.77 | 0.65 | 0.85 | 0.80–0.94 | 0.11 | 0.07–0.16 | |||
| Q4 | 1.05 | 0.93 | 0.89 | 0.83–0.94 | 0.12 | 0.06–0.18 | |||
| ** p-value interaction <0.001 | |||||||||
| Suspected Dig Toxicity | Overall | 0.01 | 0.02 | 2.19 | 1.43–3.36 | −0.01 | −0.02–0.00 | ||
| Q1 | 0.00 | 0.01 | 5.91 | 1.32–26.51 | −0.01 | −0.02–0.00 | |||
| Q2 | 0.01 | 0.01 | 1.91 | 0.64–5.72 | −0.01 | −0.01–0.00 | |||
| Q3 | 0.01 | 0.02 | 1.64 | 0.77–3.50 | −0.01 | −0.02–0.00 | |||
| Q4 | 0.02 | 0.03 | 2.16 | 1.11–4.20 | −0.02 | −0.03–0.00 | |||
| *p-value interaction = 0.43 | |||||||||
HF heart failure; RR Risk ratio; OR odds ratio; ARD absolute risk difference; Q1 quartile 1; Q2 quartile 2; Q3 quartile 3; Q4 quartile 4. Hosp hospitalization; Dig digoxin. ARD is calculated as placebo-digoxin study group for all analyses
Heterogeneity of Digoxin Treatment Effect:
Heart Failure Hospitalizations:
Patients at higher risk of hospitalization were considerably more likely to benefit from digoxin therapy for the outcome of heart failure hospitalization. In the DIG main study cohort (left ventricular ejection fraction ≤ 45%), the overall absolute risk reduction with digoxin for any heart failure hospitalization was 8%; however, there was a 13% absolute reduction in the highest risk quartile compared to just a 3% absolute risk reduction in the lowest risk quartile (Table 4 and Figure 1). While the odds ratio for heart failure hospitalization reduction favored digoxin in all quartiles, there was a non-significant trend towards greater benefit at higher baseline risk (p=0.11 for interaction). When patients from the ancillary study were included, results were similar and the interaction became statistically significant (p < 0.01), indicating greater proportional effect in high-risk patients (Supplemental Table 1).
Figure 1:
Odds ratios (OR) and absolute risk reduction (ARR) of heart failure hospitalization across quartile of predicted risk of heart failure hospitalization in the DIG main study (left ventricular ejection fraction ≤ 45%). Interaction p value =0.11
Total all-cause hospitalizations:
Overall in the non-risk stratified analysis of the DIG main study, there were 4 fewer total hospitalizations per 100 person years in the digoxin arm compared with the placebo arm. However, significant heterogeneity of treatment effect was seen with 12 fewer hospitalizations per 100 person years in the digoxin arm compared with the placebo arm in the high risk group but an increase of 8 hospitalizations per 100 person years with digoxin in the low risk quartile (Table 4 and Figure 2). There was a greater proportional reduction in all cause hospitalizations at higher baseline predicted risk (interaction p<0.001).
Figure 2:
Risk ratios (RR) and absolute risk reduction (ARR) of total all-cause hospitalizations across quartile of predicted risk of heart failure hospitalization in the DIG main study (left ventricular ejection fraction ≤ 45%). Interaction p value <0.001
Mortality and toxicity:
Digoxin had no effect on mortality overall or in any of the quartiles. While patients at higher risk quartiles in both digoxin and placebo arms had numerically more hospitalizations for suspected digoxin toxicity, there was no interaction with baseline risk (p=0.28). Similar findings were seen when the data from the DIG ancillary study was also included in the analysis.
Discussion:
Using a multi-state prediction model to risk stratify patients, we found clinically significant differences in heart failure hospitalization and death according to quartile of predicted risk of heart failure hospitalization (heterogeneity in baseline risk). In addition, the effect of digoxin on both the risk of any heart failure hospitalizations and the total number of all-cause hospitalizations was greater on both the relative and absolute scales in patients at higher predicted risk of heart failure hospitalization (heterogeneity of treatment effect). In the highest risk quartiles, digoxin was associated with a greater relative and absolute reduction in heart failure hospitalizations and all-cause hospitalizations compared to the effect of digoxin in the lowest risk quartiles. The magnitude of the benefit in those at high risk was substantially higher than that seen in the overall trial results.
Almost 6 million adults in the United States have a diagnosis of heart failure and this number is projected to increase to more than 8 million by 203018. Safely preventing unnecessary heart failure hospitalizations should help patients by reducing the stress, morbidity and direct costs to patients associated with hospitalizations. In addition, there is a significant societal cost, as 80% of the direct costs associated with the care of patients with heart failure is due to hospitalization1. Our analysis suggests that digoxin could be used in patients at high risk for heart failure hospitalization without an increased risk of death or digoxin toxicity. We did not have access to serum digoxin levels for this analysis; however, previous analyses suggest that digoxin levels in the range of 0.5–0.8 or 0.5–0.9ng/ml are associated with reduced mortality compared with higher levels8,9. We thus propose that a strategy of risk-stratified selection of patients most likely to benefit from digoxin therapy be combined with close monitoring of serum levels.
A previous analysis of the DIG study reported the effect of digoxin stratified by subsets of patients with higher New York Heart Association Class, lower left ventricular ejection fraction and greater cardiothoracic ratio19. Our analysis adds to this prior work by using a prediction model, which allows for improved discrimination of risk by pooling multiple risk factors, avoids the pitfalls of single variable subgroup analyses20 and includes a test of interaction between treatment effect and baseline risk. As seen in other studies17, this analysis highlights the clinically meaningful heterogeneity in risk and effects typically seen across a clinical trial population. The difference in the absolute risk reduction of heart failure hospitalization and the total number of all-cause hospitalizations with digoxin, as compared with placebo, was large and clinically meaningful in the highest risk groups, but negligible in those at low risk (and there was a trend for higher total hospitalizations with digoxin in this low risk group). Our findings suggest a value to more routinely reporting such analyses stratified by baseline risk21–24.
Finally, we externally validated a simplified version of the Tufts MC HF Risk Model to risk stratify the patients. The Tufts MC HF Risk Model generates predictions of both heart failure hospitalization and death in the same model. Consistent with our prior description of the model10, we found better discrimination for the mortality outcome than for heart failure hospitalization, though discrimination across all transitions remained relatively stable in the DIG trial. It’s worth noting that despite only fair discrimination for heart failure hospitalization, we found clinically meaningful heterogeneity of risk in our analysis (even with the reduced “DIG-compatible” model used in this study), suggesting the model can be useful in identifying patients with heart failure at high risk for heart failure hospitalization.
Study Limitations:
The DIG study was conducted prior to trials showing that routine beta-blocker, aldosterone antagonist, implantable cardioverter-defibrillator, cardiac resynchromization and angiotensin II receptor-neprilysin inhibitor therapies were proven to reduce heart failure hospitalizations and mortality in selected patients with heart failure and reduced ejection fraction, and thus the validity of extrapolating our findings to a contemporary population is unknown. In addition, this represents a post-hoc analysis of the DIG trial cohort.
Conclusions:
In conclusion, digoxin reduced heart failure hospitalization rates in all risk groups but with a greater relative and absolute risk reduction in patients at higher predicted risk of heart failure hospitalization. While digoxin use has declined substantially since the DIG trial, the potential benefits of this therapy in those at high risk for hospitalization should be considered.
Supplementary Material
Acknowledgements:
We would like to acknowledge all of the participants and investigators who participated in HEAAL and DIG studies as well as the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repository Information Coordinating Center for their assistance in providing the de-identified DIG database. This article does not necessarily reflect the opinions or views of the DIG study investigators or the National Heart, Lung, and Blood Institute.
Funding Sources: This work was supported by the Patient-Centered Outcomes Research Institute (PCORI) [grant number: 1IP2PI000722] and the National Institutes of Health (NIH) [grant number: U01 NS086294].
Footnotes
Conflict of Interest: The authors have no relevant conflicts of interest to disclose
References:
- 1.Heidenreich PA, Albert NM, Allen LA, et al. Forecasting the impact of heart failure in the United States: a policy statement from the American Heart Association. Circulation. Heart failure May 2013;6(3):606–619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. The New England journal of medicine. April 2 2009;360(14):1418–1428. [DOI] [PubMed] [Google Scholar]
- 3.The Effect of Digoxin on Mortality and Morbidity in Patients with Heart Failure. New England Journal of Medicine. 1997;336(8):525–533. [DOI] [PubMed] [Google Scholar]
- 4.Ziff OJ, Lane DA, Samra M, et al. Safety and efficacy of digoxin: systematic review and meta-analysis of observational and controlled trial data. Bmj. 2015;351:h4451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Packer M, Gheorghiade M, Young JB, et al. Withdrawal of digoxin from patients with chronic heart failure treated with angiotensin-converting-enzyme inhibitors. RADIANCE Study. The New England journal of medicine. July 1 1993;329(1):1–7. [DOI] [PubMed] [Google Scholar]
- 6.Lee DC, Johnson RA, Bingham JB, et al. Heart failure in outpatients: a randomized trial of digoxin versus placebo. The New England journal of medicine. March 25 1982;306(12):699–705. [DOI] [PubMed] [Google Scholar]
- 7.Hood WB Jr., Dans AL, Guyatt GH, Jaeschke R, McMurray JJ. Digitalis for treatment of congestive heart failure in patients in sinus rhythm: a systematic review and meta-analysis. J Card Fail. April 2004;10(2):155–164. [DOI] [PubMed] [Google Scholar]
- 8.Rathore SS, Curtis JP, Wang Y, Bristow MR, Krumholz HM. Association of serum digoxin concentration and outcomes in patients with heart failure. JAMA : the journal of the American Medical Association. February 19 2003;289(7):871–878. [DOI] [PubMed] [Google Scholar]
- 9.Adams KF Jr., Gheorghiade M, Uretsky BF, Patterson JH, Schwartz TA, Young JB. Clinical benefits of low serum digoxin concentrations in heart failure. Journal of the American College of Cardiology. March 20 2002;39(6):946–953. [DOI] [PubMed] [Google Scholar]
- 10.Upshaw JN, Konstam MA, Klaveren D, Noubary F, Huggins GS, Kent DM. Multistate Model to Predict Heart Failure Hospitalizations and All-Cause Mortality in Outpatients With Heart Failure With Reduced Ejection Fraction: Model Derivation and External Validation. Circulation. Heart failure August 2016;9(8). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ahmed A, Rich MW, Fleg JL, et al. Effects of digoxin on morbidity and mortality in diastolic heart failure: the ancillary digitalis investigation group trial. Circulation. August 1 2006;114(5):397–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Rationale, design, implementation, and baseline characteristics of patients in the DIG trial: a large, simple, long-term trial to evaluate the effect of digitalis on mortality in heart failure. Controlled clinical trials. February 1996;17(1):77–97. [DOI] [PubMed] [Google Scholar]
- 13.Konstam MA, Neaton JD, Dickstein K, et al. Effects of high-dose versus low-dose losartan on clinical outcomes in patients with heart failure (HEAAL study): a randomised, double-blind trial. Lancet. November 28 2009;374(9704):1840–1848. [DOI] [PubMed] [Google Scholar]
- 14.de Wreede LC, Fiocco M, Putter H. The mstate package for estimation and prediction in non- and semi-parametric multi-state and competing risks models. Computer methods and programs in biomedicine. September 2010;99(3):261–274. [DOI] [PubMed] [Google Scholar]
- 15.Harrell FE Jr., Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluating the yield of medical tests. JAMA : the journal of the American Medical Association. May 14 1982;247(18):2543–2546. [PubMed] [Google Scholar]
- 16.Ioannidis JP, Lau J. Heterogeneity of the baseline risk within patient populations of clinical trials: a proposed evaluation algorithm. American journal of epidemiology. December 1 1998;148(11):1117–1126. [DOI] [PubMed] [Google Scholar]
- 17.Kent DM, Nelson J, Dahabreh IJ, Rothwell PM, Altman DG, Hayward RA. Risk and treatment effect heterogeneity: re-analysis of individual participant data from 32 large clinical trials. International journal of epidemiology. July 3 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mozaffarian D, Benjamin EJ, Go AS, et al. Heart disease and stroke statistics--2015 update: a report from the American Heart Association. Circulation. January 27 2015;131(4):e29–322. [DOI] [PubMed] [Google Scholar]
- 19.Gheorghiade M, Patel K, Filippatos G, et al. Effect of oral digoxin in high-risk heart failure patients: a pre-specified subgroup analysis of the DIG trial. European journal of heart failure. May 2013;15(5):551–559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kent DM, Lindenauer PK. Aggregating and disaggregating patients in clinical trials and their subgroup analyses. Annals of internal medicine. July 06 2010;153(1):51–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kent DM, Rothwell PM, Ioannidis JP, Altman DG, Hayward RA. Assessing and reporting heterogeneity in treatment effects in clinical trials: a proposal. Trials. 2010;11:85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Dorresteijn JA, Visseren FL, Ridker PM, et al. Estimating treatment effects for individual patients based on the results of randomised clinical trials. Bmj. 2011;343:d5888. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Sussman JB, Kent DM, Nelson JP, Hayward RA. Improving diabetes prevention with benefit based tailored treatment: risk based reanalysis of Diabetes Prevention Program. Bmj. 2015;350:h454. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kent DM, Hayward RA, Griffith JL, et al. An independently derived and validated predictive model for selecting patients with myocardial infarction who are likely to benefit from tissue plasminogen activator compared with streptokinase. The American journal of medicine. August 01 2002;113(2):104–111. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.