Abstract
Background
In patients with heart failure (HF), anxiety or insomnia is prevalent and associated with poor clinical outcomes. Benzodiazepines (BZDs) are one of the most commonly prescribed medications for anxiety or insomnia in Taiwan. Evidence regarding the effects of BZDs on patients with heart failure and reduced ejection fraction (HFrEF) is inconclusive.
Objectives
To evaluate whether BZDs can mitigate the adverse effects of anxiety or insomnia on the prognosis of patients with HFrEF.
Methods
Patients with HFrEF were identified from the Chang Gung Research Database between January 1, 2007 and December 31, 2018. Those who received BZD prescriptions were defined as the BZD group; patients in the BZD group were then paired with those who had never been prescribed BZDs after matching for age, sex, and baseline left ventricular ejection fraction, defined as the no-BZD group. Propensity score matching was used to balance baseline characteristics. Cox proportional hazards model and the Fine-Gray subdistribution hazard model were used to examine the association between BZD prescription and the risks of adverse cardiovascular outcomes.
Results
After propensity score matching, there were 1,941 patients in both BZD and no-BZD groups. The composite of cardiovascular (CV) death or HF hospitalization (HFH) occurred in 64.4% and 54.4% of the patients in the BZD and no-BZD groups, respectively [hazard ratio (HR): 1.44; 95% confidence interval (CI): 1.32-1.56], which was mainly driven by HFH (HR: 1.52; 95% CI: 1.39-1.67).
Conclusions
In the patients with HFrEF, those who received BZD were at a higher overall risk of CV death and HFH.
Keywords: Anxiety, Benzodiazepines, Heart failure, Insomnia
INTRODUCTION
Anxiety or sleep disorders are more prevalent in patients with heart failure (HF) than in the general population.1 Studies have demonstrated that 13% of HF patients fulfill the diagnostic criteria for anxiety disorder, with 30% experiencing clinical anxiety2 and over half reporting insomnia symptoms.3,4 Anxiety has been reported to have adverse effects on various HF-related outcomes, including rehospitalization, quality of life, and functional status.5,6 Furthermore, a prospective observational study reported that insomnia was an independent predictor of cardiac death in HF patients.7
Benzodiazepines (BZDs) are among the most commonly prescribed classes of drugs for managing anxiety and sleep disorders in Taiwan.8 Several studies have reported the benefits of BZDs for relieving symptoms and improving general well-being in patients with cardiovascular (CV) diseases, including angina, hypertension, and silent myocardial ischemia.9 In addition, a cohort study observed an independent association between BZDs and a decreased risk of cardiac mortality and HF hospitalization (HFH) in patients with recent myocardial infarction (MI).10 However, evidence regarding the effects of BZDs on the prognosis of patients with HF and anxiety or insomnia is limited and inconclusive.11,12 Notably, patients enrolled in such studies have been characteristically heterogeneous: 1) both outpatients with chronic stable HF and hospitalized patients with acute decompensated HF have been included; 2) HF patients with preserved or reduced ejection fraction have all been enrolled. Thus, the present study sought to investigate whether BZD use can mitigate the negative effects of anxiety or insomnia in patients with HF and reduced ejection fraction (HFrEF).
METHODS
Data source
This retrospective cohort study used data from the Chang Gung Research Database (CGRD), which contains de-identified information from the original electronic medical records at Chang Gung Memorial Hospital (CGMH). The CGMH system, the largest health care provider in Taiwan, encompasses four tertiary medical centers and three major teaching hospitals (with 10,000 hospital beds in total), to which approximately 280,000 patients are admitted annually. In 2015, approximately 500,000 emergency department visits and 8,600,000 outpatient department visits to CGMH were recorded, accounting for approximately 10% of Taiwan’s medical services that year.13 The CGRD contains comprehensive patient data, including information on diagnoses, orders, imaging, laboratory examinations, medications, and procedures. An encryption procedure is consistently applied to ensure that the personal information of all patients remains anonymous. Details on the CGRD have been described elsewhere.14,15 The study protocol was approved, and the requirement for informed consent was waived by the Institutional Review Board of Chang Gung Medical Foundation (approval number: 202002062B0).
Study design
Figure 1 shows the flowchart of the patient selection process. Included patients were adults with HFrEF — defined as the combination of having a principal diagnosis of HF and a left ventricular ejection fraction (LVEF) of less than 40% by echocardiography — between January 1, 2007 and December 31, 2018. Patients with HFrEF who had been prescribed with BZDs at least twice within 3 consecutive months were identified as the BZD group. The cohort entry date of the BZD group was defined as the date when BZDs were prescribed for the first time in the database. Patients in the BZD group were then paired with one or more patients (who had never been prescribed BZDs) according to age, sex, and baseline LVEF, defined as the no-BZD group. After this process, there were 7,511 and 7,317 patients in the BZD and no-BZD groups, respectively. The cohort entry date for patients in the no-BZD group was defined as the same date as their paired counterparts in the BZD group to avoid immortal time bias.
Figure 1.
Flowchart of the patient selection process. BZD, benzodiazepine; HF, heart failure; HFrEF, heart failure and reduced ejection fraction; LVEF, left ventricular ejection fraction; MI, myocardial infarction; PSM, propensity score matching.
Patients who had immunological diseases, active cancer, or recent MI (defined as MI within the previous 3 months before cohort entry date), as well as those who had received cardiac surgery within the previous 3 months, been hospitalized for HF within the previous 6 months, or had a history of suicidal ideation or attempt, were excluded. Patients who were lost to follow-up within 90 days after the cohort entry date or had fewer than two BZD prescriptions within 3 consecutive months were also excluded.
Study outcomes and follow-up
The primary outcome was the composite of CV death and HFH during the follow-up period. Secondary outcomes comprised all-cause mortality, CV death, HFH, nonfatal stroke, nonfatal MI, and the composite CV outcome (i.e., CV death, HFH, nonfatal stroke, and nonfatal MI). The outcome of CV death met the criteria of Standardized Definitions for CV and Stroke Endpoint Events in Clinical Trials developed by the Standardized Data Collection for Cardiovascular Trials Initiative and the US Food and Drug Administration in 2017.16 HFH was defined as having a principal discharge diagnosis of HF and having received treatment with diuretics, nitrates, or inotropic agents during hospitalization. A principal discharge diagnosis using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes before 2016 [and International Classification of Diseases, 10th Revision (ICD-10-CM) codes after 2016] was used to defined nonfatal stroke and nonfatal MI. The diagnostic codes of HFH, stroke, and AMI using claims data in Taiwan have been validated in previous studies.17 The follow-up period was defined as the period from the index date until the first occurrence of any study outcome, death, or the most recent visit date recorded in the CGRD (outpatient, emergency department, or inpatient), whichever occurred first.
Covariates
The covariates were age, sex, physical characteristics (height, body weight, body mass index, systolic and diastolic blood pressure, and heart rate), LVEF at baseline, comorbidities (hypertension, diabetes, dyslipidemia, coronary arterial disease, peripheral arterial disease, chronic obstructive pulmonary disease, renal function status, epilepsy/seizure, and atrial fibrillation), history of events (permanent pacemaker implantation, MI, and stroke), coexisting mental disorders or related symptoms (major depression, schizophrenia, insomnia, cognitive impairment, bipolar disorder, substance abuse disorder, and anxiety disorder), 12 laboratory indicators (Table 1), use of 16 types of CV disease medications (Table 1), and the use of antidepressant or other psychotropic medications [including selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), zolpidem or other anti-psychotics]. Most diagnostic codes of the diseases mentioned above using claims data in Taiwan have been validated in previous studies.17
Table 1. Baseline characteristics of HFrEF patients according to the use of BZD.
| Variable | Valid number | Before imputation and PSM | After imputation and PSM | ||||
| BZD (n = 3,197) | No-BZD (n = 4,397) | STD | BZD (n = 1,941) | No-BZD (n = 1,941) | STD | ||
| Age, year | 7,594 | 68.6 ± 13.6 | 66.7 ± 14.5 | 0.13 | 68.4 ± 13.6 | 68.6 ± 14.1 | -0.01 |
| Male | 7,594 | 2,003 (62.7) | 3,152 (71.7) | -0.19 | 1,278 (65.8) | 1,290 (66.5) | -0.01 |
| BZD type | NA | NA | |||||
| Lorazepam | 7,594 | 932 (29.1) | 0 (0.0) | 546 (28.1) | 0 (0.0) | ||
| Alprazolam | 7,594 | 850 (26.6) | 0 (0.0) | 554 (28.5) | 0 (0.0) | ||
| Fludiazepam | 7,594 | 656 (20.5) | 0 (0.0) | 430 (22.2) | 0 (0.0) | ||
| Clonazepam | 7,594 | 642 (20.1) | 0 (0.0) | 338 (17.4) | 0 (0.0) | ||
| Diazepam | 7,594 | 117 (3.7) | 0 (0.0) | 73 (3.8) | 0 (0.0) | ||
| Physical examination | |||||||
| Height, cm | 4,468 | 159.9 ± 11.3 | 160.7 ± 11.5 | -0.08 | 160.3 ± 9.7 | 160.3 ± 8.6 | 0.01 |
| Body weight, kg | 5,806 | 63.0 ± 14.2 | 65.1 ± 15.9 | -0.14 | 64.0 ± 13.2 | 63.9 ± 13.4 | 0.01 |
| Body mass index, kg/m2 | 4,464 | 24.5 ± 4.7 | 24.8 ± 5.1 | -0.07 | 24.7 ± 4.2 | 24.7 ± 4.2 | 0.01 |
| Systolic blood pressure, mmHg | 7,172 | 132.0 ± 25.0 | 131.0 ± 23.0 | 0.04 | 131.0 ± 24.2 | 130.9 ± 22.8 | 0.01 |
| Diastolic blood pressure, mmHg | 7,172 | 74.0 ± 15.0 | 74.0 ± 15.0 | < 0.01 | 73.6 ± 14.3 | 73.4 ± 14.6 | 0.01 |
| Heart rate, beat/min | 7,047 | 82.1 ± 17.8 | 79.9 ± 16.8 | 0.13 | 81.0 ± 17.0 | 81.0 ± 16.5 | < 0.01 |
| LVEF, % | 7,594 | 32.1 ± 6.7 | 32.1 ± 6.5 | -0.01 | 32.1 ± 6.5 | 32.0 ± 6.6 | 0.03 |
| Number of heart failure hospitalizations in the previous 3 years | 7,594 | 0.8 ± 1.3 | 0.8 ± 1.2 | < 0.01 | 0.8 ± 1.3 | 0.8 ± 1.3 | -0.05 |
| History of event | |||||||
| Permanent pacemaker implantation | 7,594 | 215 (6.7) | 263 (6.0) | 0.03 | 131 (6.8) | 127 (6.5) | 0.01 |
| Myocardial infarction | 7,594 | 714 (22.3) | 1,080 (24.6) | -0.05 | 459 (23.7) | 467 (24.1) | -0.01 |
| Stroke | 7,594 | 702 (22.0) | 658 (15.0) | 0.18 | 379 (19.5) | 387 (19.9) | -0.01 |
| Comorbidity | |||||||
| Hypertension | 7,594 | 2,300 (71.9) | 2,976 (67.7) | 0.09 | 1,353 (69.7) | 1,363 (70.2) | -0.01 |
| Diabetes | 7,594 | 1,408 (44.0) | 1,868 (42.5) | 0.03 | 856 (44.1) | 846 (43.6) | 0.01 |
| Dyslipidemia | 7,594 | 1,462 (45.7) | 2,161 (49.2) | -0.07 | 933 (48.1) | 925 (47.7) | 0.01 |
| Coronary arterial disease | 7,594 | 1,755 (54.9) | 2,454 (55.8) | -0.02 | 1,092 (56.3) | 1,075 (55.4) | 0.02 |
| Peripheral arterial disease | 7,594 | 354 (11.1) | 427 (9.7) | 0.04 | 205 (10.6) | 206 (10.6) | < 0.01 |
| Chronic obstructive pulmonary disease | 7,594 | 644 (20.1) | 679 (15.4) | 0.12 | 358 (18.4) | 357 (18.4) | < 0.01 |
| Renal function status | |||||||
| Non-CKD or CKD stage 1-2 | 7,594 | 1,464 (45.8) | 2,632 (59.9) | -0.28 | 1,004 (51.7) | 988 (50.9) | 0.02 |
| CKD stage 3a | 7,594 | 518 (16.2) | 632 (14.4) | 0.05 | 302 (15.6) | 290 (14.9) | 0.02 |
| CKD stage 3b | 7,594 | 356 (11.1) | 474 (10.8) | 0.01 | 220 (11.3) | 238 (12.3) | -0.03 |
| CKD stage 4 | 7,594 | 261 (8.2) | 319 (7.3) | 0.03 | 170 (8.8) | 169 (8.7) | < 0.01 |
| CKD stage 5 or dialysis | 7,594 | 598 (18.7) | 340 (7.7) | 0.33 | 245 (12.6) | 256 (13.2) | -0.02 |
| Epilepsy/seizure | 7,594 | 95 (3.0) | 78 (1.8) | 0.08 | 49 (2.5) | 44 (2.3) | 0.01 |
| Atrial fibrillation | 7,594 | 732 (22.9) | 1,138 (25.9) | -0.07 | 474 (24.4) | 502 (25.9) | -0.03 |
| Previous treatment of heart failure | |||||||
| Cardiac resynchronization therapy | 7,594 | 26 (0.81) | 63 (1.4) | -0.06 | 19 (0.98) | 25 (1.3) | -0.03 |
| Implantable cardioverter defibrillator | 7,594 | 59 (1.8) | 61 (1.4) | 0.04 | 42 (2.2) | 31 (1.6) | 0.04 |
| Co-existing mental disorder | |||||||
| Major depression | 7,594 | 265 (8.3) | 58 (1.3) | 0.33 | 65 (3.4) | 53 (2.7) | 0.04 |
| Schizophrenia | 7,594 | 28 (0.88) | 9 (0.20) | 0.09 | 6 (0.3) | 7 (0.4) | -0.01 |
| Insomnia | 7,594 | 525 (16.4) | 141 (3.2) | 0.46 | 143 (7.4) | 138 (7.1) | 0.01 |
| Cognitive impairment | 7,594 | 294 (9.2) | 236 (5.4) | 0.15 | 151 (7.8) | 139 (7.2) | 0.02 |
| Bipolar disorder | 7,594 | 110 (3.4) | 17 (0.39) | 0.22 | 15 (0.8) | 17 (0.9) | -0.01 |
| Substance abuse disorder | 7,594 | 93 (2.9) | 81 (1.8) | 0.07 | 43 (2.2) | 49 (2.5) | -0.02 |
| Anxiety disorder | 7,594 | 622 (19.5) | 173 (3.9) | 0.5 | 165 (8.5) | 162 (8.4) | 0.01 |
| Laboratory data | |||||||
| Blood urea nitrogen, mg/dL | 4,926 | 35.5 ± 24.7 | 30.8 ± 21.2 | 0.21 | 32.4 ± 20.7 | 33.0 ± 19.9 | -0.03 |
| Sodium, mg/dL | 5,139 | 138.0 ± 4.4 | 138.8 ± 4.2 | -0.18 | 138.4 ± 3.9 | 138.4 ± 3.6 | < 0.01 |
| Potassium, mg/dL | 5,888 | 4.2 ± 0.7 | 4.2 ± 0.6 | -0.11 | 4.2 ± 0.6 | 4.2 ± 0.5 | 0.02 |
| B-type natriuretic peptide, pg/mL | 2,399 | 868 [327, 1940] | 673 [250, 1700] | 0.11 | 1069 [597, 1565] | 1073 [615, 1594] | -0.01 |
| Hemoglobin, g/dL | 5,902 | 12.0 ± 2.4 | 12.4 ± 2.4 | -0.19 | 12.2 ± 2.2 | 12.2 ± 2.1 | < 0.01 |
| Hematocrit, % | 5,883 | 36.1 ± 6.6 | 37.5 ± 6.7 | -0.2 | 36.9 ± 6.2 | 36.9 ± 6.0 | < 0.01 |
| HbA1c, % | 4,913 | 6.9 ± 1.6 | 7.0 ± 1.7 | -0.05 | 6.9 ± 1.4 | 6.9 ± 1.4 | < 0.01 |
| Low-density lipoprotein cholesterol, mg/dL | 5,828 | 98.9 ± 42.8 | 99.4 ± 45.2 | -0.01 | 98.5 ± 37.2 | 99.2 ± 39.5 | -0.02 |
| High-density lipoprotein cholesterol, mg/dL | 5,681 | 42.2 ± 13.5 | 42.6 ± 12.7 | -0.02 | 42.3 ± 12.1 | 42.0 ± 11.2 | 0.03 |
| Total cholesterol, mg/dL | 6,056 | 165.0 ± 42.0 | 165.0 ± 42.0 | < 0.01 | 164.7 ± 38.6 | 164.5 ± 37.1 | < 0.01 |
| Triglyceride, mg/dL | 5,988 | 132.0 ± 84.0 | 135.0 ± 87.0 | -0.04 | 130.5 ± 76.9 | 131.2 ± 74.2 | -0.01 |
| Uric acid, mg/dL | 5,218 | 7.2 ± 2.2 | 7.2 ± 2.2 | < 0.01 | 7.2 ± 1.9 | 7.2 ± 1.9 | -0.02 |
| Medication | |||||||
| ACEi | 7,594 | 938 (29.3) | 1,385 (31.5) | -0.05 | 580 (29.9) | 596 (30.7) | -0.02 |
| ARBs | 7,594 | 1,930 (60.4) | 2,689 (61.2) | -0.02 | 1,202 (61.9) | 1,171 (60.3) | 0.03 |
| ARNI | 7,594 | 27 (0.84) | 65 (1.5) | -0.06 | 23 (1.18) | 17 (0.9) | 0.03 |
| Beta-blocker | 7,594 | 1,930 (60.4) | 3,166 (72.0) | -0.25 | 1,279 (65.9) | 1,252 (64.5) | 0.03 |
| MRA | 7,594 | 721 (22.6) | 1,269 (28.9) | -0.14 | 480 (24.7) | 464 (23.9) | 0.02 |
| Ivabradine | 7,594 | 52 (1.6) | 89 (2.0) | -0.03 | 35 (1.8) | 37 (1.9) | -0.01 |
| Loop diuretics | 7,594 | 1,911 (59.8) | 2,705 (61.5) | -0.04 | 1,198 (61.7) | 1,215 (62.6) | -0.02 |
| Thiazide | 7,594 | 193 (6.0) | 191 (4.3) | 0.08 | 96 (5.0) | 97 (5.0) | < 0.01 |
| Statin | 7,594 | 1,350 (42.2) | 2,020 (45.9) | -0.07 | 867 (44.7) | 850 (43.8) | 0.02 |
| Digoxin | 7,594 | 633 (19.8) | 1,024 (23.3) | -0.08 | 406 (20.9) | 433 (22.3) | -0.03 |
| Amiodarone | 7,594 | 448 (14.0) | 592 (13.5) | 0.02 | 292 (15.0) | 306 (15.8) | -0.02 |
| OHA | 7,594 | 1,139 (35.6) | 1,505 (34.2) | 0.03 | 692 (35.7) | 674 (34.7) | 0.02 |
| Insulin | 7,594 | 757 (23.7) | 809 (18.4) | 0.13 | 430 (22.2) | 436 (22.5) | -0.01 |
| Aspirin | 7,594 | 1,792 (56.1) | 2,560 (58.2) | -0.04 | 1,120 (57.7) | 1,100 (56.7) | 0.02 |
| Clopidogrel | 7,594 | 987 (30.9) | 1,210 (27.5) | 0.07 | 586 (30.2) | 581 (29.9) | 0.01 |
| Ticagrelor | 7,594 | 84 (2.6) | 113 (2.6) | < 0.01 | 60 (3.1) | 54 (2.8) | 0.02 |
| Antidepressant or other psychotropic medications | |||||||
| SSRI | 7,594 | 108 (3.4) | 36 (0.82) | 0.18 | 36 (1.9) | 28 (1.44) | 0.03 |
| SNRI | 7,594 | 64 (2.0) | 10 (0.23) | 0.17 | 12 (0.6) | 10 (0.52) | 0.01 |
| Zolpidem | 7,594 | 508 (15.9) | 167 (3.8) | 0.41 | 122 (6.3) | 140 (7.2) | -0.04 |
| Others (Mirtazapine, Quetiapine) | 7,594 | 404 (12.6) | 230 (5.2) | 0.26 | 172 (8.9) | 154 (7.9) | 0.03 |
| Follow up years | 7,594 | 3.5 ± 2.8 | 2.8 ± 2.4 | 0.26 | 3.0 ± 2.5 | 3.1 ± 2.7 | -0.06 |
ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; ARNI, angiotensin receptor-neprilysin inhibitor; BZD, benzodiazepine; HFrEF, heart failure with reduced LVEF; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonists; OHA, oral hypoglycemic agent; PSM, propensity score matching; SNRI, serotonin-norepinephrine reuptake inhibitors; SSRI, selective serotonin reuptake inhibitors; STD, standardized difference.
Conditions were considered comorbidities or coexisting mental disorders if at least two outpatient diagnoses or one inpatient diagnosis had been made before the cohort entry date. Data from physical examinations and laboratory tests were extracted from records made 3 months before the cohort entry date. The medication information was extracted from the first 90-day follow-up.
Statistical analysis
After performing frequency matching by age, sex, and baseline LVEF, we constructed a propensity score-matched cohort using propensity scores derived from multivariable logistic regression. The propensity scores were calculated using the values of all covariates except for the follow-up duration. The matching was performed using a greedy nearest-neighbor algorithm with a caliper of 0.2 times the standard deviation of the logit of the propensity score, with random matching order and without replacement. Matching quality was assessed using the absolute value of the standardized difference between the groups after propensity score matching, where a value of less than 0.1 was considered a negligible difference. Because of the substantial number of missing observations in laboratory data, we imputed missing values using single expectation-maximization before creating the propensity score-matched cohort.
Regarding the time-to-event outcomes of all-cause mortality and CV death, a Cox proportional hazards model was used to compare the risk between groups. The incidence rates of the time-to-event outcome of HFH between the groups were compared using the Fine-Gray subdistribution hazard model, which considers all-cause mortality a competing risk. The group assignment (BZD vs. no-BZD) constituted the only explanatory variable in these survival analyses. The assumption of proportional hazard was tested using the correlation between the Schoenfeld residuals and survival time. In addition, post hoc subgroup analyses of primary outcomes were conducted to assess whether the effect of BZD use on the outcomes was consistent across different levels of subgroup variables. The selected subgroup variables were age (dichotomized by 65 years), sex, body mass index (dichotomized by 27 kg/m2), LVEF (dichotomized by 30%), any coexisting mental disorders or related symptoms, and use of antidepressants or sedative medications.
We finally compared the risk of the primary composite outcome according to the BZD prescription pattern by applying a Cox model to the cohort on which expectation maximization imputation was performed before propensity score matching. The medications were classified for dosage and timing: daytime BZDs, nighttime BZDs (i.e., BZDs taken before bed), and no-BZD medications. The following formula was used for the equivalent dosage determination of each BZD (with diazepam, 5 mg used as the reference value): lorazepam 1 mg = alprazolam 0.5 mg = fludiazepam 0.5 mg = clonazepam 0.25 mg.18,19 BZD users were divided into four groups according to the daily total equivalent dosage of diazepam, with each group containing a similar number of BZD users (> 70 mg vs. 43-70 mg vs. 24-42 mg vs. < 24 mg vs. no-BZD). The Fine and Gray subdistribution hazard model was adjusted for all baseline characteristics included in Table 1, except for the type of BZD. Statistical significance was defined as a two-sided p value of < 0.05. All analyses were performed using SAS software, version 9.4 of the SAS System for Unix (SAS Institute, Cary, NC, USA). Direct-adjusted survival was derived from the multivariable Cox model using the SAS macro "ADJSURV%."20
RESULTS
Baseline characteristics
A total of 7,594 HFrEF patients were included (3,197 in the BZD group and 4,397 in the no-BZD group). Table 1 presents their clinical characteristics. The most commonly prescribed BZDs were lorazepam (29.1%), followed by alprazolam (26.6%), fludiazepam (20.5%), clonazepam (20.1%), and diazepam (3.7%).
Several between-group differences in baseline characteristics were observed. Notably, compared with the no-BZD users, the BZD users were more likely to have a diagnosis of major depressive disorder (8.3% vs. 1.3%), insomnia (16.4% vs. 3.2%), bipolar disorder (3.4% vs. 0.39%), or anxiety disorder (19.5% vs. 3.9%). They were also more likely to be prescribed antidepressants or anxiolytics, and more likely to be female, older, have end-stage renal disease (and undergoing dialysis for the condition), and have a history of stroke. After expectation maximization imputation and propensity score matching, each group comprised 1,941 patients, and no substantial between-group differences in any parameters were noted.
Outcomes
BZD use in the patients with HFrEF was associated with an increased risk of the primary outcome of the composite of CV death and HFH [64.4% vs. 54.4%; hazard ratio (HR): 1.44, 95% confidence interval (CI): 1.32-1.56; Table 2 and Figure 2A]. Regarding the secondary outcomes, the BZD group had higher risks of HFH [57.9% vs. 45.8%; subdistribution HR (SHR): 1.52; 95% CI: 1.39-1.67; Table 2 and Figure 2B], nonfatal stroke (9.7% vs. 6.8%; SHR: 1.26; 95% CI: 1.001-1.57), nonfatal MI (9.9% vs. 6.2%; SHR: 1.34; 95% CI: 1.07-1.68), and the composite CV outcome (66.0% vs. 56.0%; HR: 1.45; 95% CI: 1.33-1.57). In addition, the correlation coefficients between the Schoenfeld residuals and survival time for each outcome were low (ranging from -0.12 to 0.12), indicating that the assumption of proportional hazard was not violated (data not shown).
Table 2. Outcomes of HFrEF patients according to the use of BZD in the propensity score matched cohort.
| Outcome | BZD (n = 1,941) | No-BZD (n = 1,941) | HR or SHR of BZD (95% CI) | p value |
| Primary outcome: composite of cardiovascular death and HFH | 1,250 (64.4) | 1,056 (54.4) | 1.44 (1.32-1.56) | < 0.001 |
| Secondary outcome | ||||
| All-cause death | 441 (22.7) | 402 (20.7) | 1.14 (0.995-1.30) | 0.059 |
| Cardiovascular death | 397 (20.6) | 368 (19.0) | 1.12 (0.97-1.29) | 0.118 |
| HFH | 1,123 (57.9) | 889 (45.8) | 1.52 (1.39-1.67) | < 0.001 |
| Non-fatal stroke | 188 (9.7) | 132 (6.8) | 1.26 (1.001-1.57) | 0.049 |
| Non-fatal myocardial infarction | 192 (9.9) | 120 (6.2) | 1.34 (1.07-1.68) | 0.011 |
| Composite cardiovascular outcome* | 1,281 (66.0) | 1,087 (56.0) | 1.45 (1.33-1.57) | < 0.001 |
BZD, benzodiazepine; CI, confidence interval; HFH, heart failure hospitalization; HFrEF, heart failure with reduced LVEF; HR, hazard ration; SHR, subdistribution hazard ration.
* Anyone of cardiovascular death, heart failure hospitalization, non-fatal stroke and non-fatal myocardial infarction.
Figure 2.
Plot of the cumulative event rate of the composite of cardiovascular death and HFH (A) and cumulative incidence function of HFH (B) of patients with or without BZD use in the propensity score-matched cohort. The plot was truncated at the sixth year because the number at risk was under 10% after that point. BZD, benzodiazepine; HFH, heart failure hospitalization.
Subgroup analysis
Figure 3 presents the results of the subgroup analysis of the primary composite outcome using the selected baseline characteristics. No interactions were detected regarding sex (p = 0.165), body mass index (p = 0.400), any coexisting mental disorder or related symptom (p for interaction = 0.339), or the concomitant use of antidepressants/anxiolytics (p = 0.284). However, the association between BZDs and adverse clinical outcomes was more significant in the patients aged under 65 years (p = 0.001) and those with an LVEF of < 30% (p = 0.005).
Figure 3.
Subgroup analysis of the composite of CV death and HFH by selected baseline characteristics in the propensity score-matched cohort. BZD, benzodiazepine; CI, confidence interval; CV, cardiovascular; HR, hazard ratio; LVEF, left ventricular ejection fraction.
BZD prescription pattern
As shown in Figure 4A, compared with not taking BZDs, both the use of daytime and nighttime BZDs was associated with a significantly higher risk (adjusted one minus survival) of the composite outcome of CV death and HFH [adjusted HR (aHR): 1.49; 95% CI: 1.37-1.62; aHR: 1.42; 95% CI: 1.32-1.53, respectively]. However, the risk of the use of BZDs with different dosage patterns (daytime and nighttime) was comparable (aHR: 0.95; 95% CI: 0.84-1.04). Patients with different daily doses in the BZD group had a comparable event rate, which was significantly higher than in the no-BZD group (p for trend < 0.001; Figure 4B).
Figure 4.
The adjusted one minus occurrence of the composite outcome of CV death and HFH, stratified by BZD timing (A) and dosage (B) in the cohort before matching. BZD, benzodiazepine; CV, cardiovascular; HFH, heart failure hospitalization.
DISCUSSION
Data examining the impact of pharmaceutical therapy for HFrEF patients with anxiety or insomnia are limited. The main findings of the current study are as follows. First, BZD use in patients with HFrEF and anxiety or insomnia was associated with an increased risk of the composite of CV death and HFH, which was mainly driven by HFH. Second, the composite outcome occurred more frequently in patients aged under 65 or those whose LVEF was below 30%. Finally, the daily dose and timing of BZD use did not affect the occurrence of the composite outcome.
Only two studies have investigated the prognostic effect of BZD on patients with HF, both of which showed inconclusive results. The prospective observational cohort study by Diez-Quevedo et al.11 reported an association between BZD prescription and lower mortality risk in 2,139 patients with HF, and the retrospective observational cohort study by Sato et al.12 found that compared with zolpidem, zopiclone, and/or eszopiclone, BZD use was associated with a higher risk of rehospitalization for HF after propensity score matching. Both studies included patients with normal or decreased LVEF. Heart failure with preserved ejection fraction is considered to be a clinically distinct disease from HFrEF, with different demographics, outcomes, and evidence-based therapies. Thus, only patients with HFrEF (confirmed by baseline echocardiography) were included in the present study, with most of them receiving guideline-directed medical therapies. Furthermore, the baseline characteristics and comorbidities of the study population were comparable to those of patients with HFrEF enrolled in previous studies, including participants recruited for randomized trials and individuals whose data were extracted from registries.
HF patients with insomnia, depression, or anxiety have been linked with an adverse prognosis in previous studies;2,21 however, the underlying mechanisms of how the diseases interact and the role of psychotropic agents in these situations remain unclear. In this study, the patients receiving BZD were associated with a higher overall risk of CV death or HFH. There are several possible explanations for this finding. First, recent studies have shown that BZDs may not have a clear benefit in the long-term treatment of anxiety22 or when used concomitantly with antidepressants for depression.23 Therefore, the symptoms of these mental disorders may not be alleviated under BZD use and may thus exert persistent adverse effects on patients with HF. Second, BZDs are misused in a high proportion of patients24 and they have numerous adverse effects, including an increased risk of falling25 and BZD tolerance. Long-term BZD use may lead to BZD tolerance,26 which may cause increased neurohormonal activation12 and potentially exacerbate HF. The association of BZD with poor clinical outcomes was more significant in the patients with an LVEF < 30% and age under 65 years in our study. Insomnia experienced by patients with HF may be caused by symptoms of HF,27 such as orthopnea or paroxysmal nocturnal dyspnea, however BZDs do not play a role in these conditions. This may explain the stronger association between BZD use and poor outcomes in the subgroup of patients with an LVEF of < 30%. As shown in Figure 2, the absolute number of events was much higher in the patients aged over 65 years in both the BZD and no-BZD groups. This can be attributed to the associations of older age with more comorbidities and greater physical and cognitive impairment.28 The association of BZD with poor outcomes was nonsignificant in the patients aged over 65 years because older age is assumed to be a much stronger predictor of negative outcomes. In Figure 4A and 4B, there were no differences in outcomes regarding different BZD dosage patterns or daily total BZD dosage. This may be because insomnia in HFrEF patients may be caused by a deterioration in HF condition. Therefore, prescribing BZDs may be an indicator of worsening HF, and not related to dosage patterns or daily total BZD dosage.
This study has some major limitations. First, this is a retrospective study, and although we made great efforts to match the covariates and exclude conditions that may have interfered with the results, insomnia in patients with HFrEF is sometimes caused by worsening HF, therefore causing selection bias and confounding by indication. Second, the patients with HFrEF in this study were most likely to receive all prescriptions, including BZDs, from the outpatient cardiology clinic. Thus, the prevalence of anxiety and insomnia in the BZD group may have been underestimated using the diagnostic coding in the claims data. Third, this retrospective study only used data from the CGRD, and we did not have the outcome data of patients who were followed up at other hospitals. Due to this limitation, we may have underestimated the outcomes in the study. Fourth, the CGRD does not contain information on some patient information, such as New York Heart Association functional classification, physical activity levels, anxiety scale score, the system through which depression was scored, and scores on the insomnia severity index.
CONCLUSIONS
In the patients with HFrEF in this study, those who received BZD were at a higher overall risk of CV death and HFH. When prescribing BZD medications, physicians need to be aware that insomnia in patients with HFrEF may be related to worsening HF.
Acknowledgments
The authors are grateful for the statistical assistance and support from the Maintenance Project of the Center for Big Data Analytics and Statistics (Grant CGRPVVM 0011) at Chang Gung Memorial Hospital, regarding help with the study design and monitoring, data analysis, and interpretation.
DECLARATION OF CONFLICT OF INTEREST
All the authors declare no conflict of interest
AUTHOR CONTRIBUTIONS
Chi Chuang and Fu-Chih Hsiao contributed equally to the study manuscript. FC Hsiao and PH Chu conceptualized and designed the study. CP Lin collected and analyzed the data. YC Tung and CT Wu assisted with the statistical analysis and interpretation of the data.
All authors have read and approved the final manuscript.
FUNDING
This study was supported by Grants CORPG3H0311 and CORPG5H0051 from Chang Gung Memorial Hospital, Linkou, Taiwan.
REFERENCES
- 1.Easton K, Coventry P, Lovell K, et al. Prevalence and measurement of anxiety in samples of patients with heart failure: meta-analysis. J Cardiovasc Nurs. 2016;31:367–379. doi: 10.1097/JCN.0000000000000265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Celano CM, Villegas AC, Albanese AM, et al. Depression and anxiety in heart failure: a review. Harv Rev Psychiatry. 2018;26:175–184. doi: 10.1097/HRP.0000000000000162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Redeker NS, Jeon S, Muench U, et al. Insomnia symptoms and daytime function in stable heart failure. Sleep. 2010;33:1210–1216. doi: 10.1093/sleep/33.9.1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Javaheri S, Redline S. Insomnia and risk of cardiovascular disease. Chest. 2017;152:435–444. doi: 10.1016/j.chest.2017.01.026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Volz A, Schmid JP, Zwahlen M, et al. Predictors of readmission and health related quality of life in patients with chronic heart failure: a comparison of different psychosocial aspects. J Behav Med. 2011;34:13–22. doi: 10.1007/s10865-010-9282-8. [DOI] [PubMed] [Google Scholar]
- 6.Chen CY, Wang CL. Psychiatric comorbidity and psychosocial factors matter in heart failure. Acta Cardiol Sin. 2016;32:62–64. doi: 10.6515/ACS20150723A. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kanno Y, Yoshihisa A, Watanabe S, et al. Prognostic significance of insomnia in heart failure. Circ J. 2016;80:1571–1577. doi: 10.1253/circj.CJ-16-0205. [DOI] [PubMed] [Google Scholar]
- 8.Huang WF, Lai IC. Patterns of sleep-related medications prescribed to elderly outpatients with insomnia in Taiwan. Drugs Aging. 2005;22:957–965. doi: 10.2165/00002512-200522110-00005. [DOI] [PubMed] [Google Scholar]
- 9.Balon R, Rafanelli C, Sonino N. Benzodiazepines: a valuable tool in the management of cardiovascular conditions. Psychother Psychosom. 2018;87:327–330. doi: 10.1159/000493015. [DOI] [PubMed] [Google Scholar]
- 10.Wu CK, Huang YT, Lee JK, et al. Anti-anxiety drugs use and cardiovascular outcomes in patients with myocardial infarction: a national wide assessment. Atherosclerosis. 2014;235:496–502. doi: 10.1016/j.atherosclerosis.2014.05.918. [DOI] [PubMed] [Google Scholar]
- 11.Diez-Quevedo C, Lupon J, de Antonio M, et al. Benzodiazepine use and long-term mortality in real-life chronic heart failure outpatients: a cohort analysis. Psychother Psychosom. 2018;87:372–374. doi: 10.1159/000491879. [DOI] [PubMed] [Google Scholar]
- 12.Sato Y, Yoshihisa A, Hotsuki Y, et al. Associations of benzodiazepine with adverse prognosis in heart failure patients with insomnia. J Am Heart Assoc. 2020;9:e013982. doi: 10.1161/JAHA.119.013982. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Chan YH, Chuang C, Chan CC, et al. Glycemic status and risks of thromboembolism and major bleeding in patients with atrial fibrillation. Cardiovasc Diabetol. 2020;19:30. doi: 10.1186/s12933-020-01005-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tsai MS, Lin MH, Lee CP, et al. Chang Gung Research Database: a multi-institutional database consisting of original medical records. Biomed J. 2017;40:263–269. doi: 10.1016/j.bj.2017.08.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Shao SC, Chan YY, Kao Yang YH, et al. The Chang Gung Research Database - a multi-institutional electronic medical records database for real-world epidemiological studies in Taiwan. Pharmacoepidemiol Drug Saf. 2019;28:593–600. doi: 10.1002/pds.4713. [DOI] [PubMed] [Google Scholar]
- 16.Hicks KA, Mahaffey KW, Mehran R, et al. 2017 cardiovascular and stroke endpoint definitions for clinical trials. Circulation. 2018;137:961–972. doi: 10.1161/CIRCULATIONAHA.117.033502. [DOI] [PubMed] [Google Scholar]
- 17.Cheng CL, Chien HC, Lee CH, et al. Validity of in-hospital mortality data among patients with acute myocardial infarction or stroke in National Health Insurance Research Database in Taiwan. Int J Cardiol. 2015;201:96–101. doi: 10.1016/j.ijcard.2015.07.075. [DOI] [PubMed] [Google Scholar]
- 18.Inada T, Inagaki A. Psychotropic dose equivalence in Japan. Psychiatry Clin Neurosci. 2015;69:440–447. doi: 10.1111/pcn.12275. [DOI] [PubMed] [Google Scholar]
- 19.Tyrer PJ, Silk KR. Cambridge Textbook of Effective Treatments in Psychiatry. Cambridge University Press; Jan 24, 2008. p. 920 pgs (pg 301). [Google Scholar]
- 20.Zhang X, Loberiza FR, Klein JP, et al. A SAS macro for estimation of direct adjusted survival curves based on a stratified Cox regression model. Comput Methods Programs Biomed. 2007;88:95–101. doi: 10.1016/j.cmpb.2007.07.010. [DOI] [PubMed] [Google Scholar]
- 21.Larsson SC, Markus HS. Genetic liability to insomnia and cardiovascular disease risk. Circulation. 2019;140:796–798. doi: 10.1161/CIRCULATIONAHA.119.041830. [DOI] [PubMed] [Google Scholar]
- 22.Rickels K, Moeller HJ. Benzodiazepines in anxiety disorders: reassessment of usefulness and safety. World J Biol Psychiatry. 2019;20:514–518. doi: 10.1080/15622975.2018.1500031. [DOI] [PubMed] [Google Scholar]
- 23.Bushnell GA, Stürmer T, Gaynes BN, et al. Simultaneous antidepressant and benzodiazepine new use and subsequent long-term benzodiazepine use in adults with depression, United States, 2001-2014. JAMA Psychiatry. 2017;74:747–755. doi: 10.1001/jamapsychiatry.2017.1273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Janhsen K, Roser P, Hoffmann K. The problems of long-term treatment with benzodiazepines and related substances. Dtsch Arztebl Int. 2015;112:1–7. doi: 10.3238/arztebl.2015.0001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hung CY, Wu TJ, Wang KY, et al. Falls and atrial fibrillation in elderly patients. Acta Cardiol Sin. 2013;29:436–443. [PMC free article] [PubMed] [Google Scholar]
- 26.Buscemi N, Vandermeer B, Friesen C, et al. The efficacy and safety of drug treatments for chronic insomnia in adults: a meta-analysis of RCTs. J Gen Intern Med. 2007;22:1335–1350. doi: 10.1007/s11606-007-0251-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gharzeddine R, McCarthy MM, Yu G, et al. Insomnia and insomnia symptoms in persons with heart failure: an integrative review. J Cardiovasc Nurs. 2020 doi: 10.1097/JCN.0000000000000719. [DOI] [PubMed] [Google Scholar]
- 28.Warraich HJ, Kitzman DW, Whellan DJ, et al. Physical function, frailty, cognition, depression, and quality of life in hospitalized adults ≥ 60 years with acute decompensated heart failure with preserved versus reduced ejection fraction. Circ Heart Fail. 2018;11:e005254. doi: 10.1161/CIRCHEARTFAILURE.118.005254. [DOI] [PMC free article] [PubMed] [Google Scholar]