Control of ventilation is dependent on both metabolic (automatic, i.e., chemoreceptors, stretch receptors) and behavioral (wake-dependent) control systems (7). During NREM sleep, ventilation is critically dependent on the metabolic control system. Under metabolic control, ventilation can be stimulated through medullary brain stem centers by CO₂ (likely via H⁺ concentration) and the carotid body by PO₂ and PCO₂. During wakefulness, an additional nonspecific “wakefulness drive” maintains ventilation even if the PCO₂ is reduced below the eucapnic level. With the onset of sleep, the wakefulness drive is lost and other factors, such as increased upper airway resistance, contribute to a normal rise in PCO₂ of 4 to 6 mm Hg above the wakefulness eucapnic level (8,9). In addition, if the PCO₂ falls below a certain level (the apneic threshold) during NREM sleep, ventilatory effort ceases (10). The apneic threshold is typically 2 to 6 mm Hg below the sleeping eucapnic PaCO₂ level and usually corresponds roughly to the awake eucapnic PaCO₂ level or slightly lower (11). This apnea threshold may, in part, explain the dysrhythmic breathing frequently seen at sleep onset (see below, sleep transition central apneas), even in normal individuals. Once a stable sleep stage is reached, ventilation should become regular under stable metabolic control.

Since the CSA syndromes are heterogeneous, the pathophysiology is also quite varied (Table 23-2). In some patients with hypercapnic CSA, ventilatory control is abnormal. These individuals have low or absent hypoxic/hypercapnic responsiveness. Respiration during wakefulness is maintained by behavioral/wakeful stimuli plus automatic mechanisms. However, during sleep, when these wakefulness-dependent mechanisms are no longer operative, there is little residual drive to ventilate because metabolic drive is attenuated. As a result, hypoventilation and central apneas frequently ensue (4).

Based on the comments above, the slope of the ventilatory response to hypercapnia (frequently measured
during wakefulness) may be an important variable in the development of central apneas during sleep (12). In patients with markedly diminished or absent chemosensitivity, some form of sleep-disordered breathing (SDB), frequently central apnea, would be expected. If carbon dioxide sensitivity is very low or absent, there would be little ventilation stimulus during sleep and central apneas with sustained hypoventilation would be predictable. Patients with disorders such as central alveolar hypoventilation (“Ondine’s curse”) and the obesity-hypoventilation (Pickwickian) syndrome would fall into this category (13,14 and 15). Similarly, recent reports have suggested a high prevalence of CSA among patients chronically taking narcotic medications (which can inhibit chemosensitivity) (16).

On the other hand, increased hypercapnic sensitivity can also lead to ventilatory instability during sleep. These “high-drive” CSA patients often have low arterial PCO₂
levels during wakefulness and a high hypercapnic ventilatory response (17). The unusually steep hypercapnic responsiveness can lead to respiratory instability with a fluctuating pattern of ventilation in the wake—sleep transition. This instability is believed to be secondary to intermittent overventilation (yielding hypocapnia) alternating with underventilation (yielding hypercapnia). The cyclic nature of ventilation probably results from the marked fluctuations in ventilation that occur with modest changes in PaCO₂. Thus, elevated chemosensitivity contributes to cessations in airflow by producing ventilatory overshoots during which PaCO₂ falls below the apnea threshold.

The term “loop gain” is used to describe the net stability of the ventilatory control system (18,19,20,21,22,23,24 and 25). A high loop gain refers to one that is intrinsically unstable, whereas a low loop gain implies intrinsic stability. Loop gain is defined as the ratio of the response to the stimulus (ventilatory response/ventilatory disturbance) in a feedback control system (i.e., a large response to a minor perturbation is destabilizing). Thus, an intrinsically stable system would experience a minor response to a perturbation and continue relatively stably thereafter. On the other hand, an intrinsically unstable system may have a large response to a minor perturbation, yielding subsequent instability. In the case of breathing, a robust ventilatory response to a trivial increase in PaCO₂ would be destabilizing to the control system. This loop gain concept is useful in determining the various factors that can contribute to the development of CSA.

A unifying phenomenon in all forms of CSA is a complete loss of ventilatory drive. This cessation can be recognized by an absence of electromyographic activity in the respiratory muscles or an absence of pleural pressure swings (estimated using esophageal pressure). In nonhypercapnic CSA, the basic mechanism underlying central apnea is that the PaCO₂ falls below the apneic threshold (26). Following an arousal, which typically increases ventilation, apnea can occur if sleep resumes before the PCO₂ rises above the apneic threshold (i.e., sleep transition apneas) or if there is a ventilatory overshoot due to control instability (i.e., high loop gain). The ventilatory response to arousal may indeed be a mechanism driving unstable ventilation in those patients with an unstable state (i.e., wake to sleep to wake transitions) (27). As an individual changes from wakefulness to stage 1 or 2 sleep, the PCO₂ level that was adequate to stimulate ventilation during wakefulness may be inadequate to do so during sleep, and an apnea may occur. This apnea may arouse the individual, and the process repeats itself over and over. Thus, there is ventilatory instability with oscillating levels of ventilatory drive. The causes of this instability include high controller gain (hypercapnic sensitivity and responsiveness), high plant gain (efficient CO₂ excretion), delay in information reaching the controllers (long circulation time, i.e., mixing gain), and a PaCO₂ level close to the apnea threshold (11,28,29). The overall ventilatory loop gain can therefore be characterized by the product of the individual gains (i.e., controller, plant, and mixing) (30).

The most common type of nonhypercapnic CSA is called Cheyne—Stokes respiration (CSR) (Fig. 23-2) (31). This disorder is characterized by a waxing and waning pattern of breathing most commonly seen in individuals with congestive heart failure (CHF) and left ventricular systolic dysfunction (32). This type of breathing is quite distinct, with a crescendo—decrescendo ventilatory pattern with a central apnea or hypopnea at the nadir. Thus, it is somewhat different from the aforementioned central apnea, which has a more abrupt onset and offset. Typically, the ventilatory phase between central apneas is longer in CSR than in other types of repetitive central apneas because of a prolonged circulation time (33). Another difference is that arousal tends to occur at the peak of the ventilatory effort in CSR rather than at event termination (34). The etiology of this breathing pattern in CSR has never been fully understood but is probably a product of respiratory control system instability (high loop gain with out-of-phase signal at the chemoreceptor) resulting from a prolonged circulation time and increased ventilatory responsiveness to rising PCO₂. In an anesthetized animal, lengthening normal circulation time can induce Cheyne—Stokes breathing (CSR) (35). Such an increased circulation time may produce unstable ventilation due to the delay that occurs between mechanical changes in respiration (hyperpnea or hypopnea) and receptor stimulation resulting from changes in arterial blood gases. If a patient hypoventilates or has an apnea, a substantial period of time will pass before increasing PCO₂ or decreasing PO₂ is detected at the appropriate receptor. As a result, the apnea or hypopnea is prolonged. When the deoxygenated, hypercapnic blood does reach the receptor, ventilation is stimulated for a disproportionately long period of time because the corrected blood gases are again not presented to the receptor for an extended period of time. Thus, ventilation can wax and wane, with actual apneas occurring at the nadir of this cycle. Whether this is the mechanism of Cheyne—Stokes ventilation (CSR) seen with CHF is speculative, because a severalfold increase in circulation time is necessary to produce such breathing dysrhythmias in animals (36). This is probably a greater change than commonly occurs with heart failure. When carefully matched groups of CHF patients with and without CSR have been compared, the groups do not systematically differ in circulation time. On the other hand, patients with CSR in CHF do have high ventilatory drive based on increased chemosensitivity and stimulation of ventilation through pulmonary stretch receptors (37). Thus, the group with CSR does generally have a lower awake arterial PCO₂ than CHF patients without CSR. One cause of the higher drive appears to be a higher wedge pressure (pulmonary arterial occlusion pressure) (38). Presumably, pulmonary congestion results in
the stimulation of J receptors in the lung, which also increases ventilatory drive. Thus, a prolonged circulation time probably must be combined with an increased ventilatory drive to yield (CSR) (39). Patients with CHF and CSR tend to maintain their waking eupneic PCO₂ levels during sleep, thus keeping the PaCO₂ close to the apnea threshold. Therefore, rather minor overshoots in ventilation could dramatically reduce ventilatory drive (40,41). These mechanisms (circulation time, CO₂ responsiveness, CO₂ set point) explain much of the variability in CSR occurrence in heart failure patients.

CSR has also been reported in patients with neurologic disease, primarily cerebrovascular disorders. However, the actual ventilatory pattern in these patients has been less well characterized than in patients with CHF, and the mechanisms remain poorly understood. In many cases, a cardioembolic etiology for the cerebrovascular event has not been carefully excluded, leading many to suspect underlying cardiac dysfunction in such patients. In addition, fluctuations in the level of consciousness in stroke patients have been associated with central apneas (state instability), similar in mechanism to the sleep-transition apneas described earlier. Other mechanisms leading to increased ventilatory drive have also been proposed to explain CSR in stroke patients. A reduction in or loss of tonic inhibition of ventilation could lead to this instability but has not been fully characterized. Thus, the major mechanisms of CSRs in cerebrovascular disease remain to be elucidated (4).