Pathogenesis of Central and Complex Sleep Apnoea

Abstract

Central sleep apnoea (CSA)—the temporary absence or diminution of ventilator effort during sleep—is seen in a variety of forms including periodic breathing in infancy and healthy adults at altitude and Cheyne-Stokes respiration in heart failure. In most circumstances, the cyclic absence of effort is paradoxically a consequence of hypersensitive ventilatory chemoreflex responses to oppose changes in airflow, i.e. elevated loop gain, leading to overshoot/undershoot ventilatory oscillations. Considerable evidence illustrates overlap between CSA and obstructive sleep apnoea (OSA), including elevated loop gain in patients with OSA and the presence of pharyngeal narrowing during central apnoeas. Indeed, treatment of OSA, whether via CPAP, tracheostomy, or oral appliances, can reveal CSA, an occurrence referred to as complex sleep apnoea. Factors influencing loop gain include increased chemosensitivity (increased controller gain), reduced damping of blood gas levels (increased plant gain) and increased lung to chemoreceptor circulatory delay. Sleep-wake transitions and pharyngeal dilator muscle responses effectively raise the controller gain and therefore also contribute to total loop gain and overall instability. In some circumstances, for example apnoea of infancy and central congenital hypoventilation syndrome, central apnoeas are the consequence of ventilatory depression and defective ventilatory responses, i.e. low loop gain. The efficacy of available treatments for CSA can be explained in terms of their effects on loop gain, e.g. CPAP improves lung volume (plant gain), stimulants reduce the alveolar-inspired PCO₂ difference, supplemental oxygen lowers chemosensitivity. Understanding the magnitude of loop gain and the mechanisms contributing to instability may facilitate personalised interventions for CSA.

1. Introduction and Definitions

Central sleep apnoea (CSA) is characterized by the absence of airflow accompanying the cessation of ventilatory effort during sleep. In most forms, CSA is cyclic in nature manifesting as phases of hyper-ventilation alternating with apnoea: CSA can be classified into cyclic/periodic forms characterized by an oscillatory nature versus more sustained or irregular forms. CSA in periodic forms is seen commonly in preterm and term infants in the first weeks of life¹, in adults sojourning to high altitude², and in about a third of patients with heart failure³. CSA also occurs in ~5% of patients with obstructive sleep apnoea when pharyngeal patency is restored with intervention, a phenomenon termed complex sleep apnoea^4–7. Periodic CSA is also seen in the form of idiopathic or primary CSA⁸. CSA is a common side effect of opioids⁹ and can be either periodic or “ataxic” in nature. Finally, we note that CSA can also occur in the form of isolated or prolonged, non-periodic central apnoeas, such as those seen with apnea of infancy/prematurity¹⁰, congenital central hypoventilation syndrome¹¹ and respiratory muscle weakness¹².

CSA is of clinical concern as it causes arterial oxygen desaturation, hypercapnia, post-apnoeic arousals from sleep, surges in ventilatory drive and negative intrathoracic pressure, sensation of dyspnoea, swings in arterial blood pressure and sympathetic excitation^13–15. In patients with heart failure, CSA can promote cardiac arrhythmia, reduced cardiac function, and is strongly associated with mortality^{16, 17}. In this review, we summarize the definitions of CSA, the mechanisms contributing to this affliction, and how it is transformed into stable breathing with treatment.

Criteria used to diagnose CSA vary somewhat depending on the patient population, the suspected aetiology and whether central hypopnoeas are scored. In adults, CSA is often defined as the presence of at least 5 central apnoeas per hour. In patients with heart failure, CSA is typically diagnosed as at least 15 events per hour with at least 50% of these being central events, but central hypopneas are included. Central hypopnoeas are generally defined as a 30–90% reduction in airflow due to a reduction in ventilatory effort; yet since effort is not directly measured (i.e. via oesophageal pressure/diaphragm EMG), non-invasive signals are used to infer the absence of pharyngeal obstruction. Signals indicative of pharyngeal obstruction include the flattening or scooping of the inspiratory flow shape, thoracoabdominal paradox (typically inward motion of the ribcage in concert with outward motion of the abdomen indicative of raised respiratory system resistance), or the presence of snoring indicating a flow-limited upper airway. It should be noted however that American Academy of Sleep Medicine scoring rules state that distinction between central and obstructive hypopnoeas is not required and can be challenging, and thus central events may be underreported¹⁸.

The duration of respiratory events is also employed in the diagnosis of CSA. In adults, as with obstructive events, apnoeas/hypopnoeas need to be at least 10 s in duration (~2–3 breaths). In preterm infants short apnoeas can yield severe desaturation or bradycardia (up to 50% reduction in saturation in 6 s), so the definition is broadened for neonates (≥20 s, or less if accompanying desaturation or bradycardia occurs), but hypopnoeas are typically ignored.

2. General Background

Introduction to ventilatory control

The primary features of the ventilatory control feedback loop that determine ventilatory effort are described as follows: Increased PCO₂ and reduced PO₂ are sensed at the carotid bodies located at the carotid bifurcation, making up the peripheral chemoreceptors. These chemoreceptors are well perfused and positioned to detect fast changes in PCO₂/PO₂ levels and are generally thought to dominate the response to transient changes in these variables. Increased PCO₂ (in the form of H+) is also sensed at the medulla and pons, particularly at the retrotrapezoid nucleus in the ventrolateral medulla, making up the central chemoreceptors. The central chemoreceptors also typically dictate the baseline level of ventilatory effort. Both sets of inputs are integrated and act on the respiratory pattern generator to determine the strength and frequency of the efferent neural signals to the inspiratory muscles, namely the diaphragm and external intercostals. If the respiratory mechanics are normal, these efferent signals generate a level of inspiratory muscle pressure that yields a tidal volume excursion in direct proportion.

Traditionally, CSA has been considered a simple failure of this apparatus, described broadly as the controller, to generate ventilatory effort during sleep, akin to a severe yet temporary respiratory depression. Indeed, during normal sleep, ventilatory drive is reduced and reflex ventilatory responses to changes in PCO₂ and PO₂ are diminished^{19, 20} leading to the view that CSA is an extension of this diminution in ventilatory drive. Yet, as we discuss below, CSA in most cases is paradoxically the consequence of hypersensitivity of this chemoreceptor system.

Introduction to loop gain

To understand the negative feedback control system, we also consider the effect that ventilation has on PCO₂ in the lungs and in the pulmonary venous blood leaving the lungs (arterial PCO₂). An increase in arterial PCO₂ will act on chemoreceptors to cause a rise in ventilation that will subsequently lead to a corrective reduction in arterial PCO₂ i.e. as indicated by the metabolic hyperbola (the increase in ventilation is roughly proportional to the percent rise in PCO₂). Normally, an equilibrium is achieved whereby ventilation and PCO₂ levels are relatively steady. Yet on the time-scale of CSA, a fluctuation in ventilation such as a temporary hyperpnoea accompanying arousal can wash CO₂ out of the lungs, leading to a temporary fall in arterial PCO₂. After a circulation time, the hypocapnic arterial blood reaches the chemoreceptors, yielding a temporary reduction in ventilatory drive. But because of time delay between this disturbance and its effect on the control system, the ventilatory drive response will typically yield a ventilatory undershoot. This reflex undershoot will, in turn, raise alveolar/arterial PCO₂ to elicit a delayed reflex ventilatory overshoot and so on.

The loop gain of this system, which describes the ratio of this ventilatory response (e.g. undershoot) to a prior disturbance (e.g. overshoot), ultimately determines whether the oscillation will grow into periodic central apnoeas (loop gain >1) or damp out (loop gain <1)^{21, 22}.

It is also apparent that the ventilatory response to a disturbance has two distinct components (see Figure 1). Consider that the temporary rise in ventilation (5 L/min) washed CO₂ out of the lungs such PCO₂ falls by 5 mmHg (plant gain of 1 mmHg/L.min). After a lung to chemoreceptor circulatory delay, this reduction in PCO₂ elicits a temporary 6 L/min reduction in ventilation (controller gain is 1.2 L/min/mmHg), such that the undershoot is larger than the initial disturbance (loop gain = 1.2) and periodic CSA will occur. CSA could be avoided if the CO₂ damping was improved (lowered plant gain via increased lung volume) or if the chemoreflexes were less sensitive. Reducing circulatory delay also lowers loop gain.

Figure 1.

Computer simulation of CSA in heart failure illustrating the impact of loop gain >1. Loop gain was set to 1.2 at the time denoted by the asterisk. As CSA builds-up, each undershoot in ventilation is ~1.2-times larger than the prior ventilatory overshoot (see text for details). The example also illustrates the spectrum of central events, from mild hypopnoeas to more severe apnoeas (left to right).

Since controller gain describes the change in ventilation due to changes in PCO₂ (or PO₂), the controller gain can be modified by sleep state transitions. During the CSA cycle, as ventilatory drive rises, there is often an accompanying arousal that provides an additional increase in ventilatory drive^{20, 23–26} that in turn further increases the ventilatory overshoot. The effective gain relevant for the pathogenesis of CSA now becomes the chemoreflex response plus the arousal response per change in PCO₂ throughout the cycle. Evidence that this effect plays a role includes: (1) CSA occurs more commonly at sleep onset or in light sleep (stage 1 non-REM) compared with during wake or deeper non-REM sleep^{27, 28}, and (2) sedatives can improve CSA in some patients²⁹. The increase in ventilation during arousal might relate to intrinsically greater ventilatory drive (for the same PCO₂) observed during the awake state when compared to sleep, and possibly a reflex arousal ventilation, akin to a startle response²⁶.

In principle, upper-airway effects may also promote CSA^30–32. For example, changes in pharyngeal patency that occur in parallel with PCO₂ will raise controller gain. In this case the overall controller gain (chemoresponsiveness) is equal to the intrinsic gain (chemosensitivity) multiplied by the effectiveness of the upper airway. In this context, some authors make a distinction between chemosensitivity (which reflects the ventilatory drive response to PCO₂) and chemoresponsiveness (which reflects the change in actual ventilation in response to a PCO₂ stimulus). The reason that the upper airway is considered a component of controller gain is that controller gain is essentially synonymous with chemoresponsiveness. This concept has two implications: An airway that tends to collapse as drive is reduced (i.e. via loss of muscle tone) will tend to yield a greater undershoot, thereby increasing the effective loop gain. Likewise, the same individual will exhibit a greater increase in ventilation as ventilatory drive is increased and muscle tone is re-established.

3. Pathogenesis in Patient Populations

Cheyne-Stokes respiration in congestive heart failure

Cheyne-Stokes respiration is perhaps the most widely recognised form of CSA, occurring in a substantial proportion of patients with heart failure (see example trace taken from a recent study³³ in Figure 2). A reduced cardiac output and resultant increase in the circulatory delay between the lungs and chemoreceptors is believed to play an important role in the pathogenesis of CSA. Indeed, patients with a reduced cardiac output, worsened systolic function, atrial fibrillation, and prolonged lung to chemoreceptor delays are more likely to exhibit CSA^34–37. Furthermore, heart failure therapies such as cardiac resynchronization and afterload reduction improve ventilatory stability^{38, 39}. While many patients with CSA have prolonged circulatory delays, the presence of prolonged delay alone does not appear sufficient to generate CSA, highlighting the importance of increased chemosensitivity^40–43.

Figure 2.

Illustrative example trace of CSA during non-rapid eye movement (NREM) sleep in a male patient with heart failure and atrial fibrillation. Note the resolution of CSA with the transition to REM.

The specific causes of increased chemosensitivity in CSA are unclear, and may differ between individuals^{40, 43}. Elevated pulmonary capillary pressures are associated with presence of CSA and its severity, while diuresis improves CSA within individual patients^{34, 39, 44–47}. Overnight shifts in fluid from the legs may provide another source of pulmonary congestion, with ventilatory instability more likely with increasing volume of mobilized fluid⁴⁸. Left atrial distension may also drive increased chemosensitivity and CSA irrespective of pulmonary vascular congestion⁴⁹. Notably, a few studies have called into question the role of pulmonary congestion in development of CSA^{50, 51}. Recent evidence from animal models suggests abnormalities at the level of the carotid body may play an important role, leading to both enhanced chemosensitivity and sympathetic hypertonia, which might propagate CSA and worsen heart failure^{52, 53}. Identifying the precise sources of enhanced chemosensitivity will likely provide for new therapeutic targets for CSA. For example, pharmacological reversal of the signalling mechanisms causing carotid chemoreflex hyperactivity (e.g. purinergic⁵⁴) might help suppress CSA.

Idiopathic central sleep apnoea

The presence of CSA in patients without any identifiable cardiac or neurological cause is termed idiopathic CSA. The cycling period in idiopathic CSA is ~30–40 seconds and appears to be driven largely by elevated chemosensitivity to PCO₂^{40, 55}. Arousals typically occur at the peak of hyperventilation and likely contribute to ventilatory overshoot, enhancing chemoresponsiveness⁵⁶. Circulatory delay is by definition normal in these patients and therefore unlikely to contribute to CSA.

Periodic breathing at altitude

At high altitude, low total barometric pressure with a relatively stable fraction of oxygen results in a decreased PO₂, leading to CSA of a periodic nature^{57, 58}. Although there is variation in the altitude at which CSA will develop, CSA occurs in virtually all lowlanders at arrival to altitude⁵⁹. Hypoxia promotes instability via hypoxic augmentation of the chemoreflex response to CO₂ and via an increase in hypoxic chemoresponsiveness while on a steeper portion of the hypoxic ventilatory response curve⁶⁰. In contrast to sojourners, highlanders are less susceptible to CSA, suggesting that genetic or adaptive factors likely play an important role in these responses⁵⁷. Interestingly, the hypoxic chemosensitivity increases over days-to-weeks after arrival at altitude, facilitating an increase in ventilation and improvement in PO₂, but further increasing loop gain^{59, 61, 62} (i.e. differences in instability are typically attributable to differences in the hypoxic ventilatory response rather than those in the magnitude of arterial hypoxemia). This instability appears to persist in lowlanders living at altitude beyond 1 year⁶³. Adaptive increases in the hypoxic ventilatory response with acclimatization appear to improve symptoms of acute mountain sickness (via raising PO₂) but come at the cost of exacerbating CSA⁶⁴. CSA is also linked with hypoxemia and pulmonary hypertension accompanying chronic mountain sickness, likely acting via hypoxemic effects on chemosensitivity in such cases⁶⁵.

Other factors beyond chemosensitivity may play some role in CSA at altitude. Decreases in plant gain due to hyperventilation with resulting hypocapnia, and increases in cardiac output with short circulatory delays would act as compensatory mechanisms to stabilize breathing^{62, 66} and thus individuals with less strong compensatory mechanisms may have more severe CSA. Subclinical pulmonary oedema appears to occur relatively frequently in sojourners and might lower lung volumes which would exacerbate CSA⁶⁷. Recent research has suggested that cerebral blood flow reactivity may be important in ventilatory instability at altitude via regulation (damping) of cerebral PCO₂ levels ⁵⁹.

Periodic breathing in newborn infants

Periodic breathing is almost ubiquitous in term infants and those born prematurely in the first weeks of life and its high prevalence has led to the assumption that it is non-pathological⁶⁸. However, in some cases, periodic breathing in preterm infants can lead to profound oxygen desaturation¹⁴ that may have serious consequences. Treatment of periodic breathing in such cases is warranted in light of the associations between reduced oxygen levels and mortality in neonatal intensive care⁶⁹. Periodic breathing is rare in the first days after birth but becomes progressively more prevalent over the next 2–4 weeks before a steady decline over the first year^{1, 70}. The increased CSA prevalence likely results from the raised hypoxic chemosensitivity that accompanies chemoreceptor “resetting” in the days after birth⁷¹, followed later by a reduced chemosensitivity with development. CSA is also thought to be due partly to hypoxemia^{72, 73} consequent to ventilation-perfusion heterogeneity in the developing lungs. Lower lung volumes (relative to body weight/metabolic rate) especially in preterm infants are also expected to play a role in some infants⁷⁴.

Opioid-induced central sleep apnoea

Use of opioids has become a major public health issue that has garnered considerable media attention. Studies suggest that roughly a third of chronic opioid administered patients have some form of central sleep apnoea⁷⁵. This breathing pattern has several important characteristics: First, opioids are sometimes associated with bradypnoea i.e. very low respiratory rates and attendant hypoventilation, hypercapnia and hypoxemia⁷⁶. Second, breathing is often erratic in nature, often described as ‘ataxic’⁹, attributable to effects at the central respiratory pattern generator⁷⁷. Third, severe CSA in opioid users often exhibits a periodic pattern remarkably similar to CSA at altitude, with a cycle period similar to idiopathic CSA (~30–40 s)⁷⁶ suggesting that elevated loop gain is responsible. Detailed mechanistic studies in chronic opioid patients are relatively sparse but possible causes of elevated loop gain include: 1) an elevated alveolar PCO₂ which would be expected to reduce CO₂ damping (elevated plant gain), 2) severe hypoventilation and concomitant hypoxemia⁷⁶ that will presumably raise hypoxic chemosensitivity, and 3) a doubling of the slope of the hypoxic ventilatory response independent of the prevailing hypoxemia⁷⁸. These factors likely combine to yield an elevated loop gain and promote CSA. An increased loop gain with opioids appears paradoxical given that ventilatory drive is typically reduced, highlighting the important distinction between baseline ventilatory drive and the responsiveness to changes in drive.

Treatment of opioid induced CSA is challenging. CPAP may improve sleep apnoea in some patients, but often fails to improve it in others^{76, 79}. Opioid effects on CSA are thought to be dose dependent such that breathing pattern may actually normalize with reduced doses⁸⁰. Adaptive servo-controlled ventilation has been used effectively in small studies, but the use of this therapy clinically in this context remains to be defined ^{76, 79}. There is a mechanistic basis for use of ventilatory stimulants (acetazolamide) or oxygen but the efficacy of such therapies is unproven. Of particular interest is a case study illustrating that acetazolamide improved opioid-induced CSA in a patient on CPAP therapy, but oxygen was ineffective⁸¹.

Overlap between and obstructive and central sleep apnoea

Obstructive sleep apnoea is a very common condition affecting roughly 10% of the US population. The details of obstructive sleep apnoea are covered elsewhere in this Review Series, but we review important concepts to give a more complete view of central apnoea pathogenesis. Obstructive sleep apnoea is known to be due to multiple underlying mechanisms: While some patients have primarily an anatomical problem, others have issues with control of upper airway dilator muscles while still others have unstable ventilatory control (high loop gain)^{82, 83}. Some patients have multiple mechanisms underlying apnoea. In theory, treatment directed at the underlying mechanism is likely to yield improvement in apnoea using a personalized approach.

Among obstructive sleep apnoea patients with high loop gain, the question arises as to why they develop obstructive sleep apnoea rather than CSA. In reality, many patients have features of both obstructive sleep apnoea and CSA or can change features during the course of an overnight recording, emphasizing that the distinction between these two conditions can be challenging. A number of lines of evidence suggest considerable overlap between obstructive sleep apnoea and CSA:

• Patients with more severe obstructive sleep apnoea have been shown to have higher loop gain than milder obstructive sleep apnoea or controls using multiple different measurement techniques^{82, 83}. Presumably the fluctuations in output from the central pattern generator lead to upper airway collapse when output to the upper airway dilator muscles is at its nadir in those who are anatomically predisposed.

• Agents which lower loop gain such as oxygen and acetazolamide improve obstructive sleep apnoea in some individuals, particularly those with high loop gain^84–86.

• Tracheostomy in patients with obstructive sleep apnoea can transform obstructive sleep apnoea into CSA, i.e. complex sleep apnoea (see below).

• In patients with CSA, the forced oscillatory technique and direct visualization of the airway have shown evidence of upper airway narrowing/closure during central apneas and hypopneas^{31, 32}.

• Some patients have mixed apnoeas with features of both obstructive sleep apnoea and CSA e.g. patients can have minimal respiratory effort during a portion of a respiratory event but evidence of obstructive physiology during the same respiratory event. Thus, some patients are difficult to classify as strictly obstructive sleep apnoea or CSA.

• REM sleep is a period of blunted chemosensitivity that often is associated with improvements in CSA including Cheyne Stokes breathing or periodic breathing at high altitude. Similarly, some obstructive sleep apnoea patients have worse breathing disturbance in NREM, presumably driven by a higher loop gain in this sleep stage.

Although obstructive sleep apnoea and CSA clearly have some similar features, this overlap is particularly evident in people with congestive heart failure. In some patients, investigators have observed an overnight conversion of obstructive sleep apnoea to CSA in conjunction with a reduction in PCO₂ and an increase in circulatory delay suggesting an overnight deterioration of cardiac function and rise in chemosensitivity⁸⁷. CSA can also convert to obstructive sleep apnoea with improvement in cardiac function and circulatory delay over time⁸⁸, and can also revert to obstructive sleep apnoea with cardiac transplant⁸⁹. CSA and obstructive sleep apnoea can both be improved with heart failure treatment in the form of cardiac resynchronization therapy⁹⁰. Given the evidence of considerable overlap, the use of strict cut-offs based on percentage of central events (e.g. >50%) to define CSA may be inappropriate. We therefore favour the use of the more general term sleep apnea to encompass both central and obstructive sleep apnea manifestations of disordered ventilatory control.

Treatment emergent central sleep apnoea: “Complex sleep apnoea”

The overlap between CSA and obstructive sleep apnoea is particularly relevant for patients with obstructive sleep apnoea who exhibit CSA when the upper airway is made patent with therapies. This phenomenon has been labelled complex sleep apnoea. The conversion from obstructive sleep apnoea to CSA was reported in classic studies in the context of tracheostomy. Likewise, during CPAP titrations, removal of upper airway obstruction results in CSA in some patients. This form of CSA often resolves over time with ongoing CPAP treatment, although in individuals with higher loop gain, complex sleep apnoea may persist. Oral appliance therapy can also yield CSA in patients treated for obstructive sleep apnoea.^4–7

Although unproven, a likely mechanism of complex sleep apnoea involves the relief of inspiratory flow limitation. In some patients with pharyngeal compromise, increasing ventilatory drive with increasing CO₂ may not yield increased airflow due to the prevailing upper airway mechanics (i.e. ineffective muscle responses). That is, the presence of flow limitation can markedly reduce the controller gain (i.e. no rise in airflow for increasing effort). When inspiratory flow limitation is then relieved with treatment, a high chemosensitivity can be unmasked to yield CSA. In this context, application of CPAP would increase chemoresponsiveness without changing chemosensitivity per se.

Depressed ventilatory drive and chemoresponsiveness as a mechanism of CSA

In contrast to the elevated loop gain mechanism of periodic breathing, prolonged apnoeas consequent to reduced ventilatory drive and depressed chemoresponsiveness (i.e. extremely low loop gain) occur in some cases, highlighting the importance of having an intact chemoreflex control system. Infants with prolonged apnoeas (apnoea of prematurity, apnoea of infancy, and apparent life threatening events)—as opposed to periodic breathing—have been found to have a reduced chemosensitivity and a depressed ventilatory drive¹⁰ consistent with a reduced or less robust ventilatory drive response to apnoea/hypoventilation. The neonatal tendency to respond to hypoxia with further ventilatory depression (i.e. negative chemoresponsiveness) may serve to further reinforce an event once initiated. Patients with central congenital hypoventilation syndrome exhibit profound hypoventilation and hypoxemia during sleep consequent to reduced ventilatory drive and virtually non-existent sleep-related ventilatory responses to CO₂ and hypoxia¹¹. In patients with neuromuscular weakness, apnoeas may be seen particularly during REM sleep due to a combination of low chemosensitivity and severe muscle weakness during REM atonia¹². While the absence of effort during these events classifies them as central, some have advocated for the terminology “pseudo-central” or “diaphragmatic” to emphasize the primary role of muscle weakness.

4. Personalising Treatments

The loop gain concept integrates several different components into a combined parameter which is helpful in determining overall stability and the likelihood of CSA. Several individual components (chemosensitivity, plant gain, and circulatory delay) interact in a multiplicative manner to yield the overall loop gain. Thus, in many cases, the improvement of any given component, strictly speaking, does not require knowledge of the particular mechanism of CSA. For example, instability caused by elevated circulatory delay could be resolved with CPAP/or lateral positioning to improve lung volume (plant gain).

However, we emphasise that identification of the underlying mechanism may help to guide therapy for a number of reasons. The isolation of an underlying abnormality is critical since such abnormalities may well be the most amenable to improvement with therapy. For instance, a normal circulatory delay may be difficult to improve, but a markedly elevated chemosensitivity may well respond to appropriate interventions (oxygen/pharmacological agents). In addition, the potential for toxicity of an intervention might be heightened if a normal value is being manipulated. For example, a patient with a normal chemosensitivity (but high plant gain and increased delays) may be more likely to exhibit hypoventilation in certain circumstances (e.g. REM) if efforts to suppress chemosensitivity are successful.

The overall loop gain can also be considered the sum of the separate feedback loops for PO₂ and PCO₂²¹. Thus, identifying whether increased chemosensitivity is driven by hypoxic versus hypercapnic hyperreflexia may have important implications for therapy. For example, CSA driven by hypoxic feedback (e.g. infants/altitude) is expected to be readily resolved with supplemental oxygen administration. However, supplemental oxygen is effective in some heart failure patients with CSA but not in others⁹¹, consistent with findings that some patients with heart failure have increased responses to PO₂ whereas others have increased responses to PCO₂⁴³. Thus, we view a mechanistic understanding of control of breathing critical for meaningful progress towards individualized therapies to occur.

Finally, the magnitude of loop gain is also important, regardless of the particular factor destabilizing breathing. For example, it is harder to lower loop gain to below 1 and resolve CSA in a patient with a loop gain of 1.9 at baseline than if it is 1.1. Recognising the magnitude of instability may inform which treatments have the scope to resolve CSA^{22, 33}. Clinicians could combine interventions if an individual intervention does not have sufficient potential, e.g. CPAP plus acetazolamide, or bed elevation plus oxygen. Further investigation along these lines is needed.

5. Physiological Mechanisms of Treatments

The following interventions are considered in terms of their mechanistic effects on ventilatory control:

• Continuous positive airway pressure (CPAP) undoubtedly increases lung volume and consequently improves CO₂ damping (reducing plant gain)²². There is little direct evidence that CPAP improves circulatory delay, but cardiac function can be improved and a preferential benefit may occur in those with increased filling pressures in whom there is a mechanistic basis for improved cardiac output^20,134.

• Supplemental oxygen has a profound impact on CSA in infants⁹² and at altitude, and improves CSA in some patients with heart failure⁹¹. Increased arterial PO₂ is known to lower carotid-body chemosensitivity⁹³. Other beneficial effects on stability are unlikely, as supplemental oxygen is expected to increase plant gain (for feedback control of PO₂) and circulatory delay.

• Respiratory stimulants (e.g. inhaled carbon dioxide, rebreathing, acetazolamide, theophylline) act to increase CO₂ damping (reduce “plant gain”) by making alveolar PCO₂ less susceptible to changes due to fluctuations in ventilation³³. This phenomenon is encapsulated by a reduction in the difference between alveolar and inspired PCO₂. For patients with CSA due to ventilatory depression rather than high loop gain, respiratory stimulants may act to prevent apnoea by restoring baseline ventilatory drive.

• Sleeping position can have as profound an impact on CSA as CPAP. Sleeping lateral or with bed elevation can improve CSA and is likely to act in part by raising lung volume^{27, 94}. Improvements in upper airway collapsibility may also contribute in some individuals.

• Bi-level positive airway pressure with a backup rate and phrenic nerve stimulation seek to provide an additional non-chemical source of actual ventilation independent of a subject’s own ventilatory drive. The expected effect on loop gain is complex: Loop gain will be reduced by increasing ventilation and lowering the alveolar-inspired PCO₂ gradient. A baseline source of ventilation will minimise the possible amplitude of hypopnoea for any reduction in ventilatory drive, thereby effectively lowering controller gain^{95, 96}. For patients with CSA due to ventilatory depression, these interventions act to prevent apnoea by providing an additional source of ventilatory effort.

• Dynamic interventions, including adaptive servo-controlled ventilation⁹⁷ and dynamic inspired CO₂ delivery⁹⁸, seek to clamp ventilation or PCO₂ levels to resolve CSA. If these interventions were completely effective, loop gain would be lowered to zero.

In treating CSA that is secondary to heart failure or opioids, we note that the primary focus must be on resolving the underlying pathophysiology causing CSA.

6. Conclusions

Central sleep apnoea, defined as the temporary absence of ventilatory effort during sleep, is seen in a variety of forms across the life span. Paradoxically, in most cases, the reduction in ventilatory effort is a consequence of hypersensitive ventilatory effort responses to changes in PCO₂/PO₂, i.e. elevated loop gain (overshoot/undershoot). In other patients, central apnoeas are the consequence of depressed/absent ventilatory effort responses, i.e. extremely low loop gain. Treatments for CSA can be explained in terms of effects on loop gain, e.g. CPAP improves lung volume (plant gain), stimulants reduce the alveolar-inspired CO₂ difference, supplemental oxygen lowers chemosensitivity. A greater understanding of the pathophysiology in subgroups of patients may provide insight into which interventions will have the greatest beneficial impact.

Biography

Dr. Jeremy Orr is an Assistant Clinical Professor at the University of California San Diego. His research investigates control of breathing mechanisms in sleep disordered breathing with a particular focus on patients with pulmonary hypertension, chronic lung disease, and neuromuscular disease. Professor Atul Malhotra is the Chief of the Division of Pulmonary and Critical Care Medicine, University of California San Diego. His research interests include the pathophysiology and cardiovascular/metabolic consequences of sleep apnoea and mechanical ventilation in acute respiratory distress syndrome. Dr. Scott Sands is Instructor in Medicine at the Brigham and Women’s Hospital and Harvard Medical School. His research investigates sleep apnea pathophysiology with a focus on the development of methods for personalised treatment.