Central sleep apnea pathophysiology

Patients with central sleep apnea (CSA) have a disorder of the mechanisms that control breathing. As opposed to OSA, CSA is characterised by repetitive cessation of ventilation during sleep resulting from lack of ventilatory drive to breathe. Normally, ventilation is tightly regulated to ensure levels of arterial oxygen (PaO2) and carbon dioxide (PaCO2) are maintained within narrow ranges. This is achieved by feedback loops that involve peripheral and central chemoreceptors, intrapulmonary vagal receptors, the respiratory control centres in the brainstem and the respiratory muscles (figure 16).

Normal breathing regulation

Figure 16. Normal breathing regulation (Cherniack et al., 2005).

During wakefulness, behavioural control (signals from cortical areas of the brain influencing respiration like talking or swallowing), involving non-chemical stimuli and awake stimulation to ensure normal arterial blood gases. However, during sleep this is lost and chemical control (particularly PaCO2) becomes the key mechanism in regulating ventilation. CSA is most often seen at the onset of sleep or during non–rapid eye movement (NREM) sleep, when behavioural influence is least (Baillieul et al., 2019).

CSA can be primary (idiopathic CSA) or secondary. Secondary CSA can arise because of Cheyne-Stokes breathing in patients with congestive heart failure, certain medical conditions such as stroke, a drug or substance, or high-altitude periodic breathing (AASM, 2014).

Polysomonographic patterns of central sleep apnea

Analysis of the central breathing pattern can help in identifying the underlying pathophysiology and aetiology (figure 17).

Flow traces in two types of CSA

Figure 17. Flow traces in two types of CSA (Baillieul et al., 2019).
(a) Cheyne-Stokes breathing (CSB)* is the stereotypical breathing pattern of CSA in congestive heart failure. CSB is characterised by repetitive, cyclic, waxing and waning changes in tidal volume interspersed by central apnea with a prolonged cycle time (from 45–75 seconds). (b) Central apnea with a short cycle time (20–30 seconds) can be quite regular (idiopathic CSA) or ataxic and irregular (following opioids). *Historical term: recent guidance is to define as periodic breathing and add the underlying disease. For example, ‘periodic breathing in heart failure’.

Specific subtypes of CSA can also be distinguished based on waking levels of CO2 (Eckert et al., 2007). In eucapnic/hypocapnic CSA, the underlying mechanism is instability in ventilatory control (Hernandez & Patil, 2016). This manifests typically as low PaCO2 and chronic hyperventilation and high respiratory drive. The breathing cycles show distinct alternation between central apneas or hypopneas (waning) and hyperpneas (waxing) (Baillieul et al., 2019).

Hypercapnic CSA and hypoventilation can arise from:

  • functional or anatomical lesions at the level of the respiratory centres reducing ventilatory drive (hypoventilation syndromes and drug-induced hypoventilation)
  • an inability to translate the central drive into adequate ventilation (neuromuscular disorders)

In CSA, there are two main determinants of ventilatory control instability, high loop gain and a narrow CO2 reserve (figure 18) (Baillieul et al., 2019).

The role of high loop gain in the pathogenesis of CSA

Figure 18. The role of high loop gain in the pathogenesis of CSA (Baillieul et al., 2019).
(a) CO2 levels in a patient with ventilatory instability and a narrow CO2 reserve; (b) Polysomnographic trace of CSA.

High loop gain is defined as a disproportional response to a given stimulus which promotes instability. When there is a disturbance, such as hyperventilation, PaCO2 will drop (i) and once the blood gas change is detected, a central apnea event will be initiated to counter the disturbance (ii). This will result in an increase in PaCO2 levels. With high loop gain, each response to a disturbance is greater than the initial disturbance resulting in a corrective hyperpnea (iii) which will send PaCO2 levels towards or below the apnea threshold (Baillieul et al., 2019).