Original Article
Digoxin use and risk of mortality in hypertensive patients with atrial fibrillation Peter M. Okin a, Darcy A. Hille b, Kristian Wachtell c, Sverre E. Kjeldsen d, Kurt Boman e, Bjo¨rn Dahlo¨f f, and Richard B. Devereux a See editorial comment on page 1371
Background: Digoxin is widely used for rate control of atrial fibrillation. However, recent studies have reported conflicting results on the association of digoxin with mortality when used in patients with atrial fibrillation. Moreover, the relationship of digoxin use to mortality in hypertensive patients with atrial fibrillation has not been examined. Methods and results: All-cause mortality was examined in relation to in-treatment digoxin use in 937 hypertensive patients with ECG left ventricular hypertrophy in atrial fibrillation at baseline (n ¼ 134) or who developed atrial fibrillation during follow-up (n ¼ 803), randomly assigned to losartan or atenolol-based treatment, in post-hoc analysis of a substudy of the Losartan Intervention For Endpoint Reduction in hypertension (LIFE) trial. During 4.7 1.1 years of mean follow-up, 167 patients died (17.8%) and 372 (39.7%) were treated with digoxin. In univariate Cox analyses, in-treatment digoxin use, entered as a time-varying covariate, was associated with a 61% higher risk of dying (hazard ratio 1.61, 95% confidence interval 1.18–2.19, P ¼ 0.003). After adjusting for other univariate predictors of death in this population, including age, diabetes, history of ischemic heart disease, stroke, or heart failure, baseline Cornell product, QRS duration, heart rate, serum glucose, creatinine and high-density lipoprotein cholesterol, and a propensity score for digoxin use entered as standard covariates, and for in-treatment heart rate, pulse pressure, and Sokolow–Lyon voltage treated as timevarying covariates, digoxin use was no longer a significant predictor of mortality (hazard ratio 1.04, 95% confidence interval 0.73–1.48, P ¼ 0.839). Conclusion: In hypertensive patients with ECG left ventricular hypertrophy with existing or new atrial fibrillation, digoxin use is not associated with a significantly increased risk of all-cause mortality after adjusting for other independent predictors of death and for the factors associated with the propensity to use digoxin in this population. These findings suggest that factors other than digoxin use may account for the increased mortality found with digoxin use in some studies. Clinical Trials Registration: http://clinicaltrials.gov/ct/ show/NCT00338260?order=1 Keywords: electrocardiography, fibrillation, hypertension, hypertrophy Abbreviation: LVH, left ventricular hypertrophy
INTRODUCTION
A
trial fibrillation is a common arrhythmia [1,2] that is increasing in prevalence [2]. The incidence of atrial fibrillation increases with age [1] and is increased in patients with heart failure, left ventricular hypertrophy (LVH), and coronary heart disease, and is particularly strongly related to hypertension [3–12]. The increased risk of death [3–5], sudden cardiac death [6], heart failure [5], and stroke [3,7,8] in patients with atrial fibrillation and the substantial risks associated with antithrombotic therapies aimed at decreasing the risk of embolic sequelae [13] long made maintenance of sinus rhythm a goal of therapy for patients with atrial fibrillation. However, both the Atrial Fibrillation Follow-Up Investigation of Rhythm Management (AFFIRM) study [14] and the Rate Control versus Electrical Cardioversion for Persistent Atrial Fibrillation study [15] demonstrated no significant survival benefits of a rhythm-control as compared with a rate-control strategy in the management of patients with atrial fibrillation. Indeed, current guidelines for the management of atrial fibrillation recommend the use of digoxin for rate control in patients with atrial fibrillation [16]. Although digoxin has been widely used in the treatment of patients with chronic heart failure due to left ventricular systolic dysfunction [17–22] and for rate control of atrial fibrillation [23–26], the effect of digoxin therapy on mortality remains controversial. Among patients with heart failure, although the randomized, placebo-controlled Digitalis Investigation Group (DIG) study [17] found no effect of digoxin on survival in 6800 patients, subsequent nonrandomized studies found increased all-cause mortality with digoxin use in patients with heart failure [18–20].
Journal of Hypertension 2015, 33:1480–1486 a Greenberg Division of Cardiology, Weill Cornell Medical College, New York, bMerck Research Laboratories, West Point, Pennsylvania, USA, cDepartment of Medicine, Glostrup University Hospital, Glostrup, Denmark, dDepartment of Cardiology, University of Oslo, Ulleva˚l Hospital, Oslo, Norway, eResearch Unit, Department of Medicine, Skelleftea˚ Hospital Institution of Public Health and Clinical Medicine, Umea˚ University, Umea˚, Sweden and fDepartment of Medicine, Sahlgrenska University Hospital/O¨stra, Go¨teborg, Sweden
Correspondence to Peter M. Okin, MD, Weill Medical College of Cornell University, 525 East 68th Street, New York, NY 10065, USA. Tel: +1 212 746 4688; fax: +1 212 746 8473; e-mail: [email protected] Received 9 October 2014 Revised 2 February 2015 Accepted 3 February 2015 J Hypertens 33:1480–1486 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000559
1480
www.jhypertension.com
Volume 33 Number 7 July 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Digoxin and mortality in atrial fibrillation
Conversely, post-hoc analyses of patients from the DIG study demonstrated reduced mortality among patients with serum digoxin levels in the 0.5–0.8 or 0.9 ng/ml range [21,22]. In the treatment of patients with atrial fibrillation, digoxin was associated with increased mortality in an early post-hoc analysis of AFFIRM data [23], in a subset of a prospective cohort study with atrial fibrillation and no heart failure [24], and in a large, retrospective analysis of patients with newly diagnosed atrial fibrillation in the Veterans Affairs Healthcare System [25]. However, two recent post-hoc analyses from AFFIRM reported conflicting results on the association of digoxin with mortality depending on how digoxin use was modeled and how the likelihood of being treated with digoxin was adjusted for in their analyses [26,27]. As a consequence, whether rate control with digoxin in atrial fibrillation has a significant mortality risk remains unclear. Moreover, the relationship of digoxin use to mortality in high-risk hypertensive patients with atrial fibrillation has not been examined. Therefore, the purpose of this study was to examine whether digoxin therapy is associated with an increased risk of all-cause mortality in hypertensive patients with ECG LVH who were in atrial fibrillation at study baseline or developed atrial fibrillation during study follow-up, adjusting for univariate predictors of death in this population and for the propensity to be on digoxin therapy in this post-hoc analysis of a substudy of the Losartan Intervention For Endpoint Reduction in hypertension (LIFE) trial.
METHODS Patients The LIFE study [3,6,28–31] enrolled 9193 hypertensive patients with ECG LVH by Cornell voltage-duration product [32] and/or Sokolow–Lyon voltage criteria [6] on a screening ECG in a prospective, double-blind randomized study that compared cardiovascular morbidity and mortality with losartan as opposed to atenolol-based treatment [30], as previously described [3,6,28–31]. There were 134 patients who were in atrial fibrillation on their LIFE baseline ECG and 803 patients who developed atrial fibrillation during follow-up and had baseline data on digoxin use, resulting in 937 patients with existing or new atrial fibrillation in the present post-hoc, retrospective analysis (445 women and 497 men, mean age 70 6 years).
Treatment regimens Blinded treatment was begun with losartan 50 mg or atenolol 50 mg daily and matching placebo of the other agent, with a target pressure of 140/90 mmHg or lower. During clinic visits at frequent intervals for the first 6 months and at 6-month intervals thereafter, study therapy could be uptitrated by addition of 12.5 mg hydrochlorothiazide, followed by increase in blinded losartan or atenolol to 100 mg daily. In patients whose blood pressure was still not controlled, additional open-label upward titration of hydrochlorothiazide and, if necessary, institution of therapy with a calcium channel blocker or additional other medications (excluding AT1 or b-blockers or angiotensin-converting enzyme inhibitors) was added to the double-blind treatment regimen [30]. Journal of Hypertension
Electrocardiography Study ECGs were obtained at baseline, at 6 months, and at yearly follow-up intervals until study termination or patient death and were interpreted as previously reported in detail [6,28,29]. Cornell product higher than 2440 mmms or Sokolow–Lyon voltage above 38 mm was used to identify LVH [6,28–32].
Endpoint determination All-cause mortality, and cardiovascular and sudden cardiac death were ascertained and each case was reviewed and verified by the Endpoint Committee that was blinded to other clinical information when classifying deaths [30,31]. Sudden cardiac death was a prespecified secondary endpoint in LIFE and was defined as death that was sudden and unexpected, including observed arrhythmic deaths and those not attributable to myocardial infarction, intractable heart failure, or other identifiable cause, that occurred within 24 h of symptom onset, or when the patient was last seen alive if sudden cardiac death was unwitnessed [6].
Statistical analyses Data management and analyses were performed by the investigators using SPSS version 22.0 (IBM, Inc., Armonk, New York, USA). Data are presented as mean standard deviation for continuous variables and proportions for categorical variables. Differences in mean values between patients grouped according to ever versus never digoxin use were compared using unpaired t test; comparison of proportions between groups was performed using x2 tests. The risk of all-cause mortality, and cardiovascular and sudden cardiac death to digoxin use was assessed using Cox proportional hazards models in which digoxin use was entered as a time-varying covariate to take into account changing digoxin use over time. Because patients who took digoxin at some point during the study differed significantly from patients who were never on digoxin with respect to demographic and clinical variables that are potential predictors of mortality (Tables 1 and 2), a propensity score for digoxin use at any time during the study was calculated using the logistic regression analysis in which digoxin use was the dependent variable and variables that were significantly different between groups in Tables 1 and 2 were the independent variables. To test the independence of digoxin use as a predictor of mortality, digoxin use was entered as a time-varying covariate into a multivariable Cox model that also included other statistically significant univariate predictors of all-cause mortality, including age, diabetes, history of heart failure, ischemic heart disease, or stroke, baseline serum creatinine and glucose, baseline heart rate, QRS duration and Cornell product, and the propensity score for digoxin use entered as standard covariates, and in-treatment heart rate, pulse pressure, and Sokolow–Lyon voltage LVH treated as time-varying covariates by choosing the last measurement before death or final follow-up. To illustrate the results of time-varying analyses, all-cause mortality rate was plotted as a function of in-treatment digoxin use using a univariate modified Kaplan–Meier method [33], implemented in SAS Release 8.2 (SAS Institute, Inc., Cary, North Carolina, USA) on the WIN_PRO platform. www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
1481
Okin et al. TABLE 1. Baseline demographic and clinical characteristics in relation to in-treatment digoxin use Variables Age (years) Sex (% female) Race (% black) Randomized to losartan (%) History of ischemic heart disease (%) History of myocardial infarction (%) History of heart failure (%) History of stroke (%) Diabetes (%) Current smokers (%) Prior antihypertensive treatment (%) Incident myocardial infarction (%) Incident heart failure (%) BMI (kg/m2) HDL cholesterol (mmol/l) Total cholesterol (mmol/l) Serum glucose (mmol/l) Serum creatinine (mmol/l) Urine albumin/creatinine ratio (mg/mmol)
No digoxin (n ¼ 565)
Any digoxin (n ¼ 372)
P Value
69.40 6.7 44.1 3.9 45.0 20.9 6.9 1.2 5.0 14.9 15.0 77.2 11.0 9.6 27.9 4.5 1.49 0.43 5.94 1.12 6.04 202 89.1 22.7 14.6 57.0
71.1 6.0 52.7 1.9 47.6 29.6 12.4 10.2 9.7 21.2 14.8 79.6 9.4 21.5 28.0 5.4 1.37 0.41 5.80 1.16 6.47 2.72 89.2 20.8 18.6 83.2
Digoxin use and risk of mortality in hypertensive patients with atrial fibrillation Peter M. Okin a, Darcy A. Hille b, Kristian Wachtell c, Sverre E. Kjeldsen d, Kurt Boman e, Bjo¨rn Dahlo¨f f, and Richard B. Devereux a See editorial comment on page 1371
Background: Digoxin is widely used for rate control of atrial fibrillation. However, recent studies have reported conflicting results on the association of digoxin with mortality when used in patients with atrial fibrillation. Moreover, the relationship of digoxin use to mortality in hypertensive patients with atrial fibrillation has not been examined. Methods and results: All-cause mortality was examined in relation to in-treatment digoxin use in 937 hypertensive patients with ECG left ventricular hypertrophy in atrial fibrillation at baseline (n ¼ 134) or who developed atrial fibrillation during follow-up (n ¼ 803), randomly assigned to losartan or atenolol-based treatment, in post-hoc analysis of a substudy of the Losartan Intervention For Endpoint Reduction in hypertension (LIFE) trial. During 4.7 1.1 years of mean follow-up, 167 patients died (17.8%) and 372 (39.7%) were treated with digoxin. In univariate Cox analyses, in-treatment digoxin use, entered as a time-varying covariate, was associated with a 61% higher risk of dying (hazard ratio 1.61, 95% confidence interval 1.18–2.19, P ¼ 0.003). After adjusting for other univariate predictors of death in this population, including age, diabetes, history of ischemic heart disease, stroke, or heart failure, baseline Cornell product, QRS duration, heart rate, serum glucose, creatinine and high-density lipoprotein cholesterol, and a propensity score for digoxin use entered as standard covariates, and for in-treatment heart rate, pulse pressure, and Sokolow–Lyon voltage treated as timevarying covariates, digoxin use was no longer a significant predictor of mortality (hazard ratio 1.04, 95% confidence interval 0.73–1.48, P ¼ 0.839). Conclusion: In hypertensive patients with ECG left ventricular hypertrophy with existing or new atrial fibrillation, digoxin use is not associated with a significantly increased risk of all-cause mortality after adjusting for other independent predictors of death and for the factors associated with the propensity to use digoxin in this population. These findings suggest that factors other than digoxin use may account for the increased mortality found with digoxin use in some studies. Clinical Trials Registration: http://clinicaltrials.gov/ct/ show/NCT00338260?order=1 Keywords: electrocardiography, fibrillation, hypertension, hypertrophy Abbreviation: LVH, left ventricular hypertrophy
INTRODUCTION
A
trial fibrillation is a common arrhythmia [1,2] that is increasing in prevalence [2]. The incidence of atrial fibrillation increases with age [1] and is increased in patients with heart failure, left ventricular hypertrophy (LVH), and coronary heart disease, and is particularly strongly related to hypertension [3–12]. The increased risk of death [3–5], sudden cardiac death [6], heart failure [5], and stroke [3,7,8] in patients with atrial fibrillation and the substantial risks associated with antithrombotic therapies aimed at decreasing the risk of embolic sequelae [13] long made maintenance of sinus rhythm a goal of therapy for patients with atrial fibrillation. However, both the Atrial Fibrillation Follow-Up Investigation of Rhythm Management (AFFIRM) study [14] and the Rate Control versus Electrical Cardioversion for Persistent Atrial Fibrillation study [15] demonstrated no significant survival benefits of a rhythm-control as compared with a rate-control strategy in the management of patients with atrial fibrillation. Indeed, current guidelines for the management of atrial fibrillation recommend the use of digoxin for rate control in patients with atrial fibrillation [16]. Although digoxin has been widely used in the treatment of patients with chronic heart failure due to left ventricular systolic dysfunction [17–22] and for rate control of atrial fibrillation [23–26], the effect of digoxin therapy on mortality remains controversial. Among patients with heart failure, although the randomized, placebo-controlled Digitalis Investigation Group (DIG) study [17] found no effect of digoxin on survival in 6800 patients, subsequent nonrandomized studies found increased all-cause mortality with digoxin use in patients with heart failure [18–20].
Journal of Hypertension 2015, 33:1480–1486 a Greenberg Division of Cardiology, Weill Cornell Medical College, New York, bMerck Research Laboratories, West Point, Pennsylvania, USA, cDepartment of Medicine, Glostrup University Hospital, Glostrup, Denmark, dDepartment of Cardiology, University of Oslo, Ulleva˚l Hospital, Oslo, Norway, eResearch Unit, Department of Medicine, Skelleftea˚ Hospital Institution of Public Health and Clinical Medicine, Umea˚ University, Umea˚, Sweden and fDepartment of Medicine, Sahlgrenska University Hospital/O¨stra, Go¨teborg, Sweden
Correspondence to Peter M. Okin, MD, Weill Medical College of Cornell University, 525 East 68th Street, New York, NY 10065, USA. Tel: +1 212 746 4688; fax: +1 212 746 8473; e-mail: [email protected] Received 9 October 2014 Revised 2 February 2015 Accepted 3 February 2015 J Hypertens 33:1480–1486 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI:10.1097/HJH.0000000000000559
1480
www.jhypertension.com
Volume 33 Number 7 July 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Digoxin and mortality in atrial fibrillation
Conversely, post-hoc analyses of patients from the DIG study demonstrated reduced mortality among patients with serum digoxin levels in the 0.5–0.8 or 0.9 ng/ml range [21,22]. In the treatment of patients with atrial fibrillation, digoxin was associated with increased mortality in an early post-hoc analysis of AFFIRM data [23], in a subset of a prospective cohort study with atrial fibrillation and no heart failure [24], and in a large, retrospective analysis of patients with newly diagnosed atrial fibrillation in the Veterans Affairs Healthcare System [25]. However, two recent post-hoc analyses from AFFIRM reported conflicting results on the association of digoxin with mortality depending on how digoxin use was modeled and how the likelihood of being treated with digoxin was adjusted for in their analyses [26,27]. As a consequence, whether rate control with digoxin in atrial fibrillation has a significant mortality risk remains unclear. Moreover, the relationship of digoxin use to mortality in high-risk hypertensive patients with atrial fibrillation has not been examined. Therefore, the purpose of this study was to examine whether digoxin therapy is associated with an increased risk of all-cause mortality in hypertensive patients with ECG LVH who were in atrial fibrillation at study baseline or developed atrial fibrillation during study follow-up, adjusting for univariate predictors of death in this population and for the propensity to be on digoxin therapy in this post-hoc analysis of a substudy of the Losartan Intervention For Endpoint Reduction in hypertension (LIFE) trial.
METHODS Patients The LIFE study [3,6,28–31] enrolled 9193 hypertensive patients with ECG LVH by Cornell voltage-duration product [32] and/or Sokolow–Lyon voltage criteria [6] on a screening ECG in a prospective, double-blind randomized study that compared cardiovascular morbidity and mortality with losartan as opposed to atenolol-based treatment [30], as previously described [3,6,28–31]. There were 134 patients who were in atrial fibrillation on their LIFE baseline ECG and 803 patients who developed atrial fibrillation during follow-up and had baseline data on digoxin use, resulting in 937 patients with existing or new atrial fibrillation in the present post-hoc, retrospective analysis (445 women and 497 men, mean age 70 6 years).
Treatment regimens Blinded treatment was begun with losartan 50 mg or atenolol 50 mg daily and matching placebo of the other agent, with a target pressure of 140/90 mmHg or lower. During clinic visits at frequent intervals for the first 6 months and at 6-month intervals thereafter, study therapy could be uptitrated by addition of 12.5 mg hydrochlorothiazide, followed by increase in blinded losartan or atenolol to 100 mg daily. In patients whose blood pressure was still not controlled, additional open-label upward titration of hydrochlorothiazide and, if necessary, institution of therapy with a calcium channel blocker or additional other medications (excluding AT1 or b-blockers or angiotensin-converting enzyme inhibitors) was added to the double-blind treatment regimen [30]. Journal of Hypertension
Electrocardiography Study ECGs were obtained at baseline, at 6 months, and at yearly follow-up intervals until study termination or patient death and were interpreted as previously reported in detail [6,28,29]. Cornell product higher than 2440 mmms or Sokolow–Lyon voltage above 38 mm was used to identify LVH [6,28–32].
Endpoint determination All-cause mortality, and cardiovascular and sudden cardiac death were ascertained and each case was reviewed and verified by the Endpoint Committee that was blinded to other clinical information when classifying deaths [30,31]. Sudden cardiac death was a prespecified secondary endpoint in LIFE and was defined as death that was sudden and unexpected, including observed arrhythmic deaths and those not attributable to myocardial infarction, intractable heart failure, or other identifiable cause, that occurred within 24 h of symptom onset, or when the patient was last seen alive if sudden cardiac death was unwitnessed [6].
Statistical analyses Data management and analyses were performed by the investigators using SPSS version 22.0 (IBM, Inc., Armonk, New York, USA). Data are presented as mean standard deviation for continuous variables and proportions for categorical variables. Differences in mean values between patients grouped according to ever versus never digoxin use were compared using unpaired t test; comparison of proportions between groups was performed using x2 tests. The risk of all-cause mortality, and cardiovascular and sudden cardiac death to digoxin use was assessed using Cox proportional hazards models in which digoxin use was entered as a time-varying covariate to take into account changing digoxin use over time. Because patients who took digoxin at some point during the study differed significantly from patients who were never on digoxin with respect to demographic and clinical variables that are potential predictors of mortality (Tables 1 and 2), a propensity score for digoxin use at any time during the study was calculated using the logistic regression analysis in which digoxin use was the dependent variable and variables that were significantly different between groups in Tables 1 and 2 were the independent variables. To test the independence of digoxin use as a predictor of mortality, digoxin use was entered as a time-varying covariate into a multivariable Cox model that also included other statistically significant univariate predictors of all-cause mortality, including age, diabetes, history of heart failure, ischemic heart disease, or stroke, baseline serum creatinine and glucose, baseline heart rate, QRS duration and Cornell product, and the propensity score for digoxin use entered as standard covariates, and in-treatment heart rate, pulse pressure, and Sokolow–Lyon voltage LVH treated as time-varying covariates by choosing the last measurement before death or final follow-up. To illustrate the results of time-varying analyses, all-cause mortality rate was plotted as a function of in-treatment digoxin use using a univariate modified Kaplan–Meier method [33], implemented in SAS Release 8.2 (SAS Institute, Inc., Cary, North Carolina, USA) on the WIN_PRO platform. www.jhypertension.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
1481
Okin et al. TABLE 1. Baseline demographic and clinical characteristics in relation to in-treatment digoxin use Variables Age (years) Sex (% female) Race (% black) Randomized to losartan (%) History of ischemic heart disease (%) History of myocardial infarction (%) History of heart failure (%) History of stroke (%) Diabetes (%) Current smokers (%) Prior antihypertensive treatment (%) Incident myocardial infarction (%) Incident heart failure (%) BMI (kg/m2) HDL cholesterol (mmol/l) Total cholesterol (mmol/l) Serum glucose (mmol/l) Serum creatinine (mmol/l) Urine albumin/creatinine ratio (mg/mmol)
No digoxin (n ¼ 565)
Any digoxin (n ¼ 372)
P Value
69.40 6.7 44.1 3.9 45.0 20.9 6.9 1.2 5.0 14.9 15.0 77.2 11.0 9.6 27.9 4.5 1.49 0.43 5.94 1.12 6.04 202 89.1 22.7 14.6 57.0
71.1 6.0 52.7 1.9 47.6 29.6 12.4 10.2 9.7 21.2 14.8 79.6 9.4 21.5 28.0 5.4 1.37 0.41 5.80 1.16 6.47 2.72 89.2 20.8 18.6 83.2
