When LOFTy Ideals Meet Clinical Trial Reality

Despite occurring in approximately one third of patients with heart failure with reduced ejection fraction (HFrEF) and being recognized as a harbinger of poor outcomes, including increased mortality, the optimal management of central sleep apnea (CSA) in HFrEF remains elusive (1–3). Although optimization of cardiac function through guideline-directed medical therapy can improve CSA and Cheyne-Stokes respiration (CSR), it may be difficult to achieve in practice, and CSA frequently persists (4, 5). Patients with HF and CSA have worse outcomes than those with HF alone. The worse outcomes are thought to result from repetitive arousal from sleep, sympathetic nervous system activation, and recurrent hypoxemia (6).

Several CSA-directed treatments have been proposed to improve cardiovascular outcomes in this population. Continuous positive airway pressure (CPAP) did not clearly improve transplant-free survival in the CAN-PAP (Canadian Continuous Positive Airway Pressure for Patients with Central Sleep Apnea and Heart Failure) trial, but in a secondary analysis appeared to provide benefit when the apnea–hypopnea index (AHI) was suppressed by CPAP (7). This finding led to the development and study of adaptive servo ventilation (ASV), designed specifically to regulate CSA-CSR. Surprisingly, compared with medical management, ASV was shown to be harmful in the SERVE-HF (Treatment of Sleep-Disordered Breathing with Predominant Central Sleep Apnea by Adaptive Servo Ventilation in Patients with Heart Failure) trial in patients with reduced ejection fraction and CSA (8). The ADVENT HF (Effect of Adaptive Servo-Ventilation on Survival and Cardiovascular Hospital Admissions in Patients with Heart Failure and Sleep Apnea) trial, enrolling patients with obstructive sleep apnea (OSA) or CSA, evaluated a different ASV algorithm that did not appear to be harmful; however, the number of participants with CSA was relatively small (9). Supplemental oxygen therapy has emerged as a potential alternative supported by biological rationale and preliminary evidence demonstrating efficacy in eliminating CSA in a few small studies, but with wide variability in treatment effect (10). For all of the reasons above, a large trial evaluating the efficacy and safety of oxygen therapy for CSA-CSR in heart failure with reduced ejection fraction was needed.

In this issue of AnnalsATS, Redline and colleagues (pp. 1951–1960 ) report the result of LOFT-HF (Impact of Low Flow Nocturnal Oxygen Therapy on Hospital Admissions and Mortality in Patients with Heart Failure and Central Sleep Apnea), a multicenter randomized trial that evaluated supplemental oxygen versus a sham control device in patients with HFrEF and CSA (11). The key outcomes were mortality and time to first cardiovascular event: priority outcomes that have been lacking in trials of other CSA therapies such as phrenic nerve pacing (12). Adherence to the sham device that provided air flow without supplemental oxygen was also carefully monitored. After years of planning, the trial suffered the operational realities of the coronavirus disease (COVID-19) pandemic. Although the authors pivoted to streamlined procedures (e.g., simplified oxygen titration procedures and remote research visits), the trial was terminated after randomizing just 98 participants (48 to supplemental oxygen therapy) of the 858 planned.

Given the small numbers enrolled, what can we conclude about the use of oxygen for CSA in HF? First, there does not seem to be a compelling reason to prescribe oxygen to improve cardiovascular or patient-reported outcomes in those with CSA-CSR based on what is now the largest study of oxygen therapy in HF to date. More broadly, we still lack data that CSA-CSR is a modifiable risk factor of cardiovascular outcomes (13). Second, oxygen therapy showed a trend toward harm. Although not statistically significant, this finding warrants caution and may indicate mechanisms by which hyperoxia may increase oxidative stress and adverse outcomes. Third, oxygen therapy—or indeed, any therapy for sleep disordered breathing—may be associated with low adherence, limiting its real-world effectiveness. CPAP is criticized for its low adherence, yet multiple other therapies also show less than ideal adherence or use, including the use of supplemental oxygen in the LOFT-HF trial, in which it was used for only 3.5 hours per night. Overall, in a patient with CSA-CSR and HF, the preferred approach might be to first optimize fully cardiac function. Then, in the absence of clear data to improve cardiovascular outcomes, patients and providers can discuss therapies such as PAP, oxygen, acetazolamide, or phrenic nerve pacing, ideally within the context of a randomized controlled trial. Although current evidence is insufficient to guide decision-making regarding cardiovascular outcomes, treatment may be justified for symptomatic benefit in the absence of demonstrable harm. Clearly, more work is needed in this area.

The landscape of sleep disordered breathing management is changing. Diagnostic modalities to identify sleep disordered breathing are increasing (i.e., wearables, nearables, and smartphone-based platforms), and additional devices and drugs to treat OSA and CSA are under evaluation. Thus, it is of great value that the authors share their lessons learned about how to conduct such clinical trials, with challenges including difficulty in obtaining data and interacting with participants remotely, slow site activation, capitated budgets, and the stigma associated with oxygen therapy that may have contributed to poor adherence. LOFT-HF also draws attention to the critical need for pilot and feasibility data to inform estimates of enrolment, adherence, outcome ascertainment, and attrition so that trials can be successfully completed and inform practice. For example, the actual event rate was approximately 22%, whereas a control event rate of 40% was used in power calculations. The trial would have been underpowered to assess the primary outcome even if the target sample size had been achieved unless the true effect size was considerably larger than expected. Overestimation of sample size at the planning stage is well documented in the trial literature, with a recent systematic review of major cardiovascular trials finding that >80% of trials overestimated the effect size of the tested intervention (14).

In other fields, different approaches are being used. One such approach that is gaining popularity because of its flexibility and potential for impactful results is to use a Bayesian statistical framework rather than the classical frequentist framework. A frequentist framework assumes fixed treatment effects, and inference depends on the probability of observing a treatment difference as large as, or larger than, the one seen if the null hypothesis were true (i.e., the P value as a long-run probability across repeated randomizations). By contrast, a Bayesian framework treats treatment effects as random variables with probability distributions that represent uncertainty. As data accumulate, prior beliefs are updated to yield a posterior distribution that allows direct probability statements about the likelihood of benefit. Bayesian trials are also naturally sequential and easily allow for flexible adaptations such as sample size reestimation or adaptive allocation if an intervention is shown to be beneficial or poorly tolerated as trial data accumulate (15). Embedding implementation science principles at the outset in these future trials is especially important with complex interventions to identify barriers and facilitators early on. Other strategies to improve adherence such as real-time adherence monitoring coupled with personalized, motivation-type feedback may improve treatment uptake, but should themselves be scalable across different sites and diverse patient populations. Future trials can also pilot these strategies before using them in definitive studies. Finally, networks to facilitate research in OSA would be helpful. Some of the largest trials in the field have recruited from cardiology (SAVE) or weight management (SURMOUNT OSA) clinics that do not clearly reflect those patients who present to sleep medicine clinics (16, 17). For more rare disorders like CSA, single centers are unlikely to be able to conduct impactful trials. With the use of master protocols and platform trials, perpetual ecosystems of respiratory research can be created where clinical trials are more closely embedded into clinical care.

The technology available for the management of sleep disordered breathing is evolving at a rapid pace. Methods to evaluate these new therapies will need to keep up.