Patients with heart failure with reduced ejection fraction and predominant central sleep apnea pose treatment challenges. Jeffrey Voigt, a systematic review and meta-analysis of randomized controlled trials (RCTs) were undertaken. Electronic searches of digital repositories, journals, specialty society and manufacturer websites, manual searches of reference sections of RCTs, and published clinical guidelines were performed. Studies were graded for bias. Meta-analytic random-effects models were used. Outcomes of interest included: sleep, cardiovascular, mortality, and quality of life (QoL). Grading of recommendations assessment, development, and evaluation was performed. Nineteen randomized studies were identified that met the inclusion criteria of apnea hypopnea index (AHI) ≥10, predominant central sleep apnea (CSA), and heart failure with reduced ejection fraction (HFrEF) ≤50%. Most trials examined adaptive servo ventilation (ASV) (8 studies) and continuous positive airway pressure (CPAP) (9 studies). Bias existed in that: 15 of 19 (79%) of the trials lacked blinding, 10 of 19 were manufacturer funded, and with attrition in 8 of 19 studies. In meta-analysis, ASV performed better than control on sleep but not on QoL or cardiovascular outcomes, including mortality. CPAP demonstrated positive short-term outcomes on sleep, cardiovascular, and QoL (3 months). Longer-term cardiovascular and mortality data did not show benefit. Drug therapies demonstrated a positive clinical effect short term on sleep outcomes only. Transvenous phrenic nerve stimulation (TPNS) demonstrated positive treatment on sleep and QoL at 6 months. Evidence suggests improvement in cardiovascular outcomes with TPNS. In conclusion, ASV and CPAP therapies improve sleep, but long-term QoL or cardiovascular benefit was lacking. TPNS exhibited positive outcomes on sleep and QoL at 6 months with positive trends in CV outcomes.
Periods of apnea (absence of inspiratory airflow) and hypopnea (a decrease in airflow leading to a drop in arterial oxygen saturation and/or arousal) are common in patients with heart failure with reduced ejection fraction (HFrEF). Obstructive sleep apnea, the more common form of sleep apnea in the general population, is characterized by persistent respiratory efforts against an occluded airway. In contrast, central sleep apnea (CSA) occurs most often in cardiovascular patients and is characterized by a complete or partial reduction in the neural outflow from the brainstem to the respiratory muscles during sleep.1 This reduction leads to a characteristically long cycle of cessation/reduced respiratory effort (upwards of 40 seconds or more) followed by periods of rapid, deep respiratory effort. Both obstructive sleep apnea and CSA may coexist in the same patient. A strong association exists between HFrEF and CSA, with greater than 50% of CSA patients having a concurrent diagnosis of HFrEF.1 Treatment options for CSA include therapies such as adaptive servo ventilation (ASV) and continuous positive airway pressure (CPAP) supplemental nocturnal oxygen, medications, and transvenous phrenic nerve stimulation (TPNS). Randomized clinical trials (RCTs) evaluating oxygen and medications have been limited by their small sample sizes, thus limiting strong, evidence-based clinical recommendations. To date, systematic reviews have focused on investigating treatments for patients with CSA7 or with HFrEF for specific individual treatments and individual outcomes, but not for the overall treatment effect of patients with CSA plus HFrEF. The goal of this manuscript is to provide a systematic review and meta-analysis of available RCTs examining the totality of therapies for CSA in the HFrEF population.
Methods
The current review followed the preferred reporting and Cochrane methodology in assessing and reporting on systematic reviews and meta-analyses. Additionally, the current analysis utilized the grading of recommendations, assessment, development, and evaluation methodology in assessing the certainty of the evidence.
RCTs evaluating one therapy versus other therapies or placebo for the treatment of CSA were identified. Studies or analyses included in this systematic review and meta analysis required the following criteria. Untreated control groups are identified as guideline-directed medical therapy.
Outcomes of primary interest include general sleep and respiratory parameters such as oxygen saturation, apnea hypopnea index (AHI), central apnea-hypopnea index or central apnea index per the American Academy of Sleep Medicine; daytime sleepiness such as the Epworth Sleepiness Scale; cardiac events; and quality of life (as measured by a validated instrument for either general health or disease/condition-specific). Other endpoints evaluated specifically related to heart failure evaluation and management include quality of life instruments specific to heart failure (e.g., Minnesota Living with Heart Failure Questionnaire; New York Heart Association functional classification; Chronic Heart Failure Questionnaire) and left ventricular ejection fraction. Other secondary outcomes of interest included hospital admissions for a worsening heart failure (planned or unplanned), death from any cause, cardiovascular death, compliance with therapy (if a positive airway pressure therapy), and life-saving cardiovascular intervention. Lastly, healthcare costs and cost-effectiveness were explored as outcomes but none were identified.
The data was extracted by JV and reviewed independently by RG. Data included in the analysis items for systematic reviews and meta-analysis guidelines were extracted via a data collection form. In the case of ambiguous or missing data, an attempt was made to contact the corresponding author(s) of the publication. Full-length papers were used for data extraction where available. Abstracts were used only if a full-length paper was not available. Unpublished results were also used if available and authors were willing to share them.
Assessment of the risk of bias was performed using the risks of bias tables and summary generated by Cochrane methodology.
Data synthesis and statistical meta-analysis (when possible) were carried out using Cochrane Review Manager
software (v 5.1). Random effects models were used mainly based on the assumption that the studies were performed by different people in different locations at different times. Continuous data were evaluated using the mean difference. Continuous data were evaluated using the mean difference with 95% confidence intervals. Discrete data were analyzed using the Mantel-Haenszel random or fixed effects methods. The I2 statistic was calculated when there were at least 3 studies used in the meta-analysis to estimate the heterogeneity of the study findings. Sensitivity analysis was employed by removing one of the studies and observing its effect on the result. Forest plots were generated when at least 2 studies were identified of the same therapy evaluating the same outcome. If only one study was identified for a specific therapy, the statistical data from the study were presented descriptively. Conversion from one statistic to another (e.g., median and interquartile range to mean and standard deviation) used statistical methods as outlined in the literature. The grading of recommendations assessment, development, and evaluation was also evaluated using the GRADEpro guideline development tool.
Results
The 2,377 hits using the above search terms and evaluating the sources listed above resulted in 1,009 records after duplicates were removed. Of the 1,009 records remaining, 505 abstracts were screened with 183 full-text articles obtained for review. The 322 records not addressed (1,009 less 505 less 183) did not relate to either the CSA or heart failure disease state or were not RCTs. Ultimately 164 of the 183 articles were excluded leaving 19 for qualitative review and synthesis (i.e., risk of bias assessment) and 12 for meta-analytic purposes. Details of each of the 19 included studies can be found in methods, study design, participants, treatment(s), and outcomes
Of the 19 articles included, the breakout by therapy comparison in each of the RCTs was as follows: Nineteen trials met inclusion criteria with the majority being single-center trials. Single-center trials took place in Germany(1), Japan(2), United States(3), Canada(5), France (1), and New Zealand(1). Multicenter trials took place in Canada (1), United Kingdom (1), and France (1). Additionally, there were 3 multinational trials. In total, there were 13 single-center trials and 6 multicenter trials.
The trials broken out by the type of therapy evaluated or compared with were as follows:
In 79% (15 of 19) of the studies, there was a high risk of bias noted in those participants and/or clinicians monitoring the patients during the trial who were not blinded to the treatment.
 were undertaken.){kind=link}