Treating Central Sleep Apnoea in Heart Failure: Is Pull Better Than Push?

This article refers to ‘Phrenic nerve stimulation to treat patients with central sleep apnoea and heart failure’ by M.R. Costanzo et al., published in this issue on pages xxx. Heart failure (HF) is frequently accompanied by sleep-disordered breathing. As HF becomes more severe, the prevalence of Cheyne–Stokes respiration with central sleep apnoea (CSR-CSA) (characterised by an oscillatory pattern of ventilation with hyperventilation-induced hypocapnia) increases markedly.1 Such patients are at increased risk of both hospitalisation due to HF decompensation, and mortality.1, 2 The other main type of sleep apnoea is obstructive sleep apnoea, which is also frequent in patients with HF, particularly if they are overweight, diabetic, have a thick neck or retro- or pro-gnathism: this form of apnoea is due to collapse of the upper airway. Positive pressure mask therapy to apply pressure support to the airway and to smooth out breathing patterns has been studied extensively in a wide range of patients with sleep-disordered breathing, including patients with HF and predominantly CSR-CSA. Two major randomised trials of positive airway pressure (PAP) have been reported: the Canadian Continuous Positive Airway Pressure for Patients with Central Sleep Apnea and Heart Failure (CANPAP) in 258 patients with HF with reduced ejection fraction (HFrEF, ejection fraction < 40%) followed up for a mean of 2 years,3 and the Treatment of Sleep-Disordered Breathing with Predominant Central Sleep Apnea by Adaptive Servo Ventilation in Patients with Heart Failure (SERVE-HF) with the use of adaptive servo-ventilation in 1325 patients with HFrEF (ejection fraction ≤ 45%) followed for an average of 31 months4 (Table 1). CANPAP3, 4 Recruited Dec 1998- May 2004 CPAP vs. control 18–79 years HF with NYHA II–IV LVEF <40% Optimal medical therapy for ≥1 month AHI ≥15 with >50% central events AHI: –19* O2 sats: Mean + 1.2%* Lowest +4%* No changes in: - Sleep time - Distribution of sleep stages - Arousal index Total of 55 deaths: NS difference between the two groups. No difference in hospitalisations O2 sats Mean 93 ± 3% Lowest 82 ± 7% SERVE-HF4, 7 Recruited Feb 2008- May 2013 ASV vs. control ≥22 years HF with NYHA III or IV, or II with hospitalisation within 24 months LVEF ≤45% Optimal medical therapy for ≥1 month AHI ≥15 with >50% central + central AI ≥10 events/h AHI: –25 * CAHI: –22 * ODI3: –22.9* O2 sats: Mean + 1%* Lowest +6%* t < 90%: –32.7 min* (all at 24 months, but results very similar from 3–48 months) No difference in QoL between groups (MLHFQ improved by 4 points by 3 months in both groups, then returned to no change from baseline gradually up to 48 months) ESS improved by 1 point from a baseline of 7 (P = 0.08 compared to control) Total of 425 deaths: All-cause mortality increased by 28% [6–55%] P = 0.01 CV mortality by 34% [9–65%] P = 0.006 No difference in hospitalisations 48% O2 sats Mean 93 ± 2% Lowest 80% ± 7 t < 90% remedē® Pivotal Trial11, 12 Recruited April 2013- May 2015 Transvenous unilateral phrenic stimulation vs. control ≥18 years HF in opinion of investigator NYHA I–III Optimal medical therapy for ≥1 month AHI ≥20 with ≥50% of apnoeas central OAI ≤20% of AHI 96 with HF (total n = 151) AHI: –25.0 (18.1)** CAI: –22.8 (17.8)** ODI4: –22.7 (17.8)** t < 90%: –6.2% (11.9) P = 0.002 % sleep in REM: +2.4% (7.9) P = 0.02 Arousal index: –15.2 (18.5)** ESS –3.7** (from baseline of 9) Marked or moderate improvement in PGA: 55% (40–68)** Non-randomised post-hoc: ESS –3.1 (4.7) at 12 months* Marked or moderate improvement in PGA 55% MLHFQ –6.8 (20) or –4 [–18 to +8] P = 0.005 Non-randomised, SR only: LVEF +4% [–1 to +8] P = 0.004 or mean 3.3 (7.6) Total of 3 deaths in 151 patients in randomised phase of study. 6-month rate of HF hospitalisation was 4.7% in treated patients and 17% in control patients (P = 0.065) 72% 15.7 ± 17 (% of sleep time) CANPAP reported improvement at 3 months in respiratory metrics [the apnoea–hypopnoea index (AHI) dropped by 21 points in the active treatment group compared with only by 2 in the control group (P < 0.001), and mean nocturnal oxygen saturation increased by 1.6% compared with 0.4% (P < 0.001)] but this improvement in surrogate outcomes was not reflected in any change in the primary endpoint of transplant-free survival. Post-hoc analysis suggested that if the CSR-CSA could be controlled effectively (AHI < 15) there was an improvement in the primary endpoint (P = 0.043).5 SERVE-HF took a form of ventilatory support that more effectively ameliorates CSR-CSA, adaptive servo-ventilation,1, 2 and included many more patients and followed them up for longer than CANPAP. Once again, there was improvement in respiratory metrics, with a major reduction in AHI [31.2 dropping to 6.7 throughout the trial (P < 0.001), and the time with an oxygen saturation < 90% (t < 90%) dropping from 51 min to around 19 min (P < 0.001), but an unexpected and highly significant increase in cardiovascular mortality was reported (34% increase, 95% confidence interval 9–65%, P = 0.006)]. The increased mortality was driven not by an increase in the risk of HF hospitalisation but by an increase in cardiovascular death without a preceding hospitalisation — presumably related to sudden death.6 Data from a sub-study looking at ventricular reverse remodelling and change in biomarkers reported no difference between those randomised to adaptive servo-ventilation and control.7 The mechanism of harm from PAP in patients with central sleep apnoea associated with HFrEF remains unexplained. A harmful effect of PAP on cardiac haemodynamics has been proposed, particularly in those with a very low ejection fraction4 — but no change in left or right ventricular dimensions or function was reported out to 12 months in the SERVE-HF substudy.7 It has also been proposed that CSA may be, at least partially, protective for arrhythmia — with amelioration of CSR-CSA leading to an increased risk of sudden death.8 A further randomised trial (ADVENT-HF), recruiting both patients with predominantly obstructive sleep apnoea and a minority with predominantly CSR-CSA, has been enrolling for some years and may shed further light on the issue, albeit in a much smaller number of patients with CSR-CSA than SERVE-HF.9 Into this scene comes a very different method of treating CSA: an implantable transvenous unilateral phrenic nerve stimulator. Unlike PAP, intermittent stimulation of the diaphragm has been shown elsewhere to generate physiological levels of negative (not positive) intrathoracic pressure, simulating the normal act of breathing that‘sucks’ air into the chest.10 The remedē® system has been studied in a randomised trial in 151 patients from 31 hospital-based centres in Germany, Poland and the USA, with only 96 (64%) giving a history of HF.11 In this issue of the Journal, Costanzo and colleagues12 report a post-hoc exploratory analysis of these 96 patients. It is important to realise that most of the comparisons reported are non-randomised, are in a small number of patients, and lack a control group. But, nonetheless, the data are intriguing (Table 1). The remedē® algorithm is fine-tuned to the patient to minimise any sensation of stimulation, and to prevent any cross-talk with other implantable technology such as a defibrillator. The technology ameliorates CSA: in the randomised comparison of all patients out to 6 months, the central apnoea index drops from a median of 30.1/h to 1.4/h compared with no change in the control group with the implant switched off (P < 0.0001). However, hypopnoeas remain (AHI only dropped from 48.9 to 21.3/h) indicating a significant degree of CSA persisted. Similarly, the oxygen desaturation index (number of events per hour when oxygen desaturation was 4% or more from the overnight baseline) halved from 41 to 19.5 compared with little change in the control group (P < 0.0001). Results from the post-hoc non-randomised data out to 12 months shows a very similar picture in the patients with HF, with t < 90% improved from 9.8% to 4.4% of sleep time, and arousal index dropping from 43 times per hour to 24 per hour.11 Most of the patients studied with the remedē® system to date have little in the way of obstructive apnoea, which at least in theory could be worsened by applying negative intrathoracic pressure to a collapsible upper airway. So — if we accept that the therapy appears to control central apnoeic (if not hypopnoeic) episodes in people with HF — are there suggestions that there might be benefits beyond respiratory or sleep metrics? While daytime somnolence (measured by the Epworth Sleepiness Scale) fell to a greater extent in the treatment group, the baseline level was within the normal range, as is typical in CSA, so the clinical importance of this finding is arguable. It is always difficult to assess robustly health-related quality of life, particularly in a non-randomised comparison, but both the pivotal trial (randomised out to 6 months)11 and the post-hoc non-randomised analyses from the trial population (out to 12 months) suggest a potential beneficial impact of the order of 7 points on average on the Minnesota Living with HF questionnaire — where 5 points is usually taken as the minimally important difference. How does this compare with mask therapy? In SERVE-HF, both the active and the control group showed an improvement of around 3 points on this questionnaire out to 12 months, but the difference was no longer significant by 24 months, presumably as any placebo effect or the effect of being in a trial wore off. The effect on New York Heart Association (NYHA) class appeared to last longer, but was the same in both randomised groups. Evidence of reverse remodelling would also be promising. In the non-randomised exploratory data analysis (and only in those who were in sinus rhythm and completed follow-up to 12 months — a total of only 50 patients), there was some evidence that ejection fraction had modestly improved over this time (average increase of 4%) — although one cannot account for any changes in drug prescription or compliance during this period. The number of patients with HF in the remedē® trial (n = 96) and the number of hospitalisations or deaths (only 3 deaths in total in the 6 months of the randomised comparison, compared with 425 in SERVE-HF) is too small to draw any conclusions on harder outcomes. Compared to patients randomised in SERVE-HF, the patients in the remedē® study would appear to have milder HF and presumably a lower risk of death: they were younger, had a higher ejection fraction, and substantially less severe symptoms (19% with NYHA class I symptoms — not enrolled in SERVE-HF — and only 39% had NYHA class III compared to 69% in SERVE-HF). A higher proportion had an implantable cardioverter defibrillator (55% compared with 48% in SERVE-HF). Any assessment of the impact on HF hospitalisations or mortality must await a larger and longer-term trial. Unlike positive pressure mask therapy (where compliance can be an issue) there will also be the inevitable procedure-related serious adverse events — these happened in 8% of patients in the pivotal trial, including two patients requiring explantation of the system and two patients with lead displacement.11 Despite Food and Drug Administration clearance for the treatment of moderate-to-severe CSR-CSA in adults,13 it is unlikely that HF guideline writers will consider the evidence base strong enough to make a recommendation for the treatment of CSR-CSA with an implantable phrenic nerve stimulator as yet. Central sleep apnoea remains a risk marker in HF: evidence that any specific intervention targeted at treating the centrally-triggered respiratory disturbance improves the outcome for patients remains circumstantial. We should continue to treat HF according to best evidence-based guidelines. Further adequately powered randomised trials with hard endpoints are necessary to determine if treating CSA, either with positive pressure mask therapy or with an implantable unilateral phrenic nerve stimulator, improves the outcome of our patients. Some improvement in respiratory, sleep and health-related quality of life metrics is a positive signal — but we have been here before with other therapies in HF, and patients and the physician community require firmer evidence before intervention can be justified. Conflict of interest: none declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.