The ventilatory responsiveness to CO2 below eupnoea as a determinant of ventilatory stability in sleep


  • Jerome A. Dempsey,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • Curtis A. Smith,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • Tadeuez Przybylowski,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • Bruno Chenuel,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • Ailiang Xie,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • Hideaki Nakayama,

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author
  • James B. Skatrud

    1. The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA
    Search for more papers by this author

Corresponding author J. A. Dempsey: The John Rankin Laboratory of Pulmonary Medicine, Department of Population Health Sciences, University of Wisconsin-Madison, Madison, WI, 53726-2368, USA. Email:


Sleep unmasks a highly sensitive hypocapnia-induced apnoeic threshold, whereby apnoea is initiated by small transient reductions in arterial CO2 pressure (PaCO2) below eupnoea and respiratory rhythm is not restored until PaCO2 has risen significantly above eupnoeic levels. We propose that the ‘CO2 reserve’ (i.e. the difference in PaCO2 between eupnoea and the apnoeic threshold (AT)), when combined with ‘plant gain’ (or the ventilatory increase required for a given reduction in PaCO2) and ‘controller gain’ (ventilatory responsiveness to CO2 above eupnoea) are the key determinants of breathing instability in sleep. The CO2 reserve varies inversely with both plant gain and the slope of the ventilatory response to reduced CO2 below eupnoea; it is highly labile in non-random eye movement (NREM) sleep. With many types of increases or decreases in background ventilatory drive and PaCO2, the slope of the ventilatory response to reduced PaCO2 below eupnoea remains unchanged from control. Thus, the CO2 reserve varies inversely with plant gain, i.e. it is widened with hyperventilation and narrowed with hypoventilation, regardless of the stimulus and whether it acts primarily at the peripheral or central chemoreceptors. However, there are notable exceptions, such as hypoxia, heart failure, or increased pulmonary vascular pressures, which all increase the slope of the CO2 response below eupnoea and narrow the CO2 reserve despite an accompanying hyperventilation and reduced plant gain. Finally, we review growing evidence that chemoreceptor-induced instability in respiratory motor output during sleep contributes significantly to the major clinical problem of cyclical obstructive sleep apnoea.

Sleep effects on ventilatory control stability

Sleep presents major challenges to the stability of ventilatory control and to the homeostatic regulation of O2 and CO2 transport. All healthy humans hypoventilate to a significant degree at sleep onset and throughout the night (Pmath formula, 2–8 mmHg > awake). Furthermore, resistance to flow through the upper airway is increased variably but significantly in a very large percentage of healthy subjects, and cyclical overshoots and undershoots of ventilation occur in a significant segment of the population who do not experience ventilatory instability in wakefulness.

Several factors contribute to these effects of sleep. First, the loss of supra-pontine neural input to the medullary respiratory pattern generator (‘wakefulness drive’) results in reduced rhythmic and tonic activation of both phrenic and hypoglossal motor neurones (Orem et al. 1985). (The magnitude of the respiratory motor output attributable to ‘wakefulness’ drives has been included in theoretical models of respiratory control (Longobardo et al. 2002). However, this wakefulness input has no fixed magnitude. Rather, its most important characteristic is its marked variability, as opposed to the highly reproducible responses obtained in NREM sleep, in response to such perturbations as changes in airway resistance or hypocapnia (Skatrud & Dempsey, 1983; Wilson et al. 1984).) Second, sleep state is not constant, which greatly enhances the opportunity for changing respiratory drives and airway resistance. Third, the supine position and the associated reduction in functional residual capacity occurring in sleep will reduce caudal traction on and therefore narrow the upper airway, and will also cause greater arterial O2 desaturation and CO2 retention for any given apnoeic length.

There are multiple, complex causes of sleep-induced periodic breathing. The propensity for unstable, periodic breathing in a given subject or condition has been assessed using the concept of ‘loop gain’ (GL), the combination of three types of gain, namely controller (Gc), plant (Gp), and mixing (Gm) gains (Khoo, 2000; Younes et al. 2001). The higher the GL, the greater the probability of periodic breathing. This quantitative approach to predicting ventilatory instability and periodicity has provided a valuable tool for understanding the complex, multifaceted determinants of periodic breathing. However it is a concept that was developed to characterize the dynamic behaviour of linear systems (Khoo, 2000). Thus the usefulness of the GL index as a predictor of instability only applies over the range of linear changes in the ventilatory stimulus–response slope and does not consider the mechanisms and/or conditions which predispose to apnoea. (Loop gain (GL) is determined by: (a) plant gain (Δ pulmonary capillary inline image) as determined by the position of inline imageversusPmath formula on the iso-metabolic line (see Fig. 2 below), and also by inline image distribution and the prevailing functional residual capacity (FRC) (b) controller gain inline image as determined by chemosensitivity (above eupnoea), lung and airway mechanics and respiratory muscle power output; and (c) mixing gain (Δ chemoreceptor Pmath formula pulmonary capillary Pmath formula), as determined by circulatory delay time, thoracic blood volume and brain extracellular fluid volume (Younes et al. 2001).) Accordingly, when Pmath formula falls below eupnoea and apnoea ensues, controller gain, and therefore loop gain would actually be reduced (i.e. the CO2 ventilatory response slope is zero), thereby predicting a stable breathing pattern even in the face of apnoea.

Figure 2.

The effects of changing background ventilatory drive on the gain of the ventilatory responsiveness to CO2 below eupnoea, on ‘plant gain’ and on the CO2 reserve ( ΔPmath formula eupnoea – apnoea) in sleeping dogs and humans
Data are plotted on separate isometabolic lines for dogs (CO2 flow inline image and for humans inline image. The diagonal dashed or continuous lines join eupnoeic and apnoeic points and their slopes indicate the gain below eupnoea of the ventilatory response to hypocapnia in each condition. The height of the vertical bar above the isometabolic line indicates the increase in inline image required to reduce the Pmath formula to the apnoeic threshold (i.e. the inverse of plant gain). The CO2 reserve is the difference in Pmath formula between eupnoea and the apnoeic threshold. Note in the sleeping dog: (a) that the slope of the CO2 response below eupnoea remained unchanged during hyper- or hypoventilation; (b) the required increase in inline image to reach apnoea under background conditions of hyperventilation via metabolic acidosis inline image is about twice that required in control and almost five times that required with the background condition of hypoventilation via metabolic alkalosis (0.3 l min−1); (c) the greater protection against apnoea during these backgrounds of hyper- versus hypoventilation are due solely to the change in Gp. Similarly, almitrine infusion (see bottom panel) caused hyperventilation, with no significant influence on the slope of the ventilatory response to CO2 below eupnoea. However, in hypoxia (bottom panel) note the increased slope (versus control) of inline image relationship between eupnoea and the apnoeic threshold in dogs and especially in humans and in chronic heart failure (CHF) patients with Cheyne-Stokes periodic respiration (CSA). Thus in these latter two examples the CO2 reserve is narrowed significantly despite the background hyperventilation and reduced plant gain. (Data compiled from Xie et al. 2001, 2002; Nakayama et al. 2003; and from author's laboratory.)

We believe a neglected but important element to ‘controller’ gain is the CO2 responsiveness below eupnoea, i.e. over the range between eupnoea and apnoea. Certainly, understanding how apnoeas occur is an important part of understanding periodic breathing, for several reasons: (a) apnoeas are an integral ingredient of many types of periodic breathing; (b) apnoeas permit the accumulation of large amounts of chemoreceptor sensory stimuli which in turn will promote arousal and cause subsequent ventilatory overshoots; (c) the marked reductions in respiratory motor output which initiate the apnoea may also trigger airway muscle hypotonicity, and cause airway narrowing and even closure (see examples below under ‘Ventilatory control instability and obstructive sleep apnoea’). In this brief review we focus on the causes, lability, and consequences of the gain of the CO2 response below eupnoea and the associated change in the CO2 reserve between eupnoea and the apnoeic threshold.

Our view of the multiple factors affecting transient ventilatory overshoots (hyperventilation) and undershoots (hypoventilation) in NREM sleep are shown in Fig. 1 and are based on the observation that significant transient overshoots in ventilation and blood pressure precede and terminate most apnoeic and/or hypopnoeic events. An increased drive to breathe, depending on its magnitude and duration, also causes a persistent, centrally mediated excitatory ‘memory’ effect or short-term potentiation (STP) which would normally prevent a sudden ventilatory undershoot and maintain ventilatory stability (Eldridge & Millhorn, 1986). There are several factors which oppose this stabilizing effect of STP during this post-hyperpnoeic period (see Fig. 1). However, as outlined below, during NREM sleep it is the transient hypocapnia initiated during the ventilatory overshoot which is the dominant inhibitory influence required to cause the apnoea.

Figure 1.

Schematic representation of the factors influencing sleep-induced breathing instabilities
A, Cheyne-Stokes respiration in heart failure; B, periodic breathing clusters in hypoxia. C and D isolated ventilatory overshoots followed by hypopnea (C) or apnoea (D). The ventilatory overshoot is determined by: (a) increased magnitude of chemoreceptor stimuli, which in turn are dependent upon the duration of any preceding apnoea or hypopnoea, the functional residual capacity and the metabolic rate; (b) the gain of the chemoreceptor response, which in part depends upon state of consciousness; and (c) the translation of respiratory drive to ventilation which is dependent upon the prevailing upper airway resistance. Following the ventilatory overshoot a continued excitatory short-term potentiation (STP) of respiratory motor output favours stabilization of breathing pattern. The effectiveness of STP as a stabilizer will depend upon the magnitude and duration of the preceding stimulus. Inhibitory effects opposing STP (and causing apnoea) include: lung stretch at high VT which will have a carry-over inhibitory effect, baroreceptor stimulation coincident with the rising systemic pressure during the overshoot phase which also contributes to a reduced respiratory motor output to both the pump and upper airway musculature, and transient hypocapnia. ‘Inertia’ refers to the continued short-term inhibition of inspiratory drive, prolongation of apnoea and delayed resumption of rhythmic respiration in the face of a Pmath formula which increases to above eupnoeic levels.

The apnoeic thresholds

The ‘unmasking’ of a highly sensitive apnoeic threshold in NREM sleep, some 2–5 mmHg Pmath formula below (sleeping) eupnoeic Pmath formula and within 1–2 mmHg of the normal waking eupnoeic Pmath formula has been documented in healthy sleeping subjects in several ways. First, positive-pressure mechanical ventilation through a nasal mask was used to progressively reduce Pmath formula below eupnoea, which caused progressive, linear reductions in the amplitude of diaphragm EMG until apnoea occurred (Skatrud & Dempsey, 1983). Mechanical ventilation also elicits a non-chemical, neuro-mechanical inhibitory effect on the amplitude of respiratory motor output (Henke et al. 1988), but this can be minimized by using a subject-triggered pressure support form of mechanical ventilation and a background of continuous positive airway pressure (CPAP) to minimize upper airway resistance (Meza et al. 1998). Second, mechanical ventilation was used to control airway and chest wall mechanics in normal subjects during NREM sleep and wakefulness, to demonstrate a sleep-induced increase of 2–4 mmHg in the set-point for Pmath formula which was independent of any sleep-induced changes in airway resistance (Simon et al. 1993). Third, periodic breathing caused by chronic heart failure, hypoxia, or idiopathic central apnoea was alleviated by small increments in Pmath formula (via increased inspired CO2 fraction (Fmath formula); Berssenbrugge et al. 1983; Xie et al. 1997). Finally, Semple et al. (Semple et al. 1999) recently used mechanical ventilation to control tidal volume at waking, eupnoeic levels in order to prevent the hypoventilation and increase in Pmath formula normally experienced upon transition from awake to light sleep. At sleep onset, apnoeas of up to 15 s ensued. These observations demonstrate the marked CO2 dependence of ventilatory control at sleep onset and the critical importance of the normal sleep-induced hypoventilation as a deterrent to apnoea and instability.

Apnoea really has two Pmath formula‘thresholds’, namely the apnoeic initiation threshold (as outlined above) and the higher Pmath formula at which apnoea is terminated and respiratory rhythm re-initiated. In real life when apnoeas are initiated by transient reductions in alveolar Pmath formula (Pmath formula), some of the delay in resumption of breathing rhythm is likely to be due to a persistent alkalinity in the environment of the brain chemoreceptors at a time when arterial Pmath formula (Pmath formula) is restored to eupnoeic levels. However, even when apnoeas are caused experimentally by purely mechanical feedback at normocapnic Pmath formula (using controlled mechanical ventilation or upper airway negative pressure) the Pmath formula will rise 3–5 mmHg or so above spontaneous eupnoea before breathing is re-initiated (Leevers et al. 1994; Satoh et al. 2001). What causes this persistent, prolongation of apnoea to occur in the face of rising chemoreceptor stimuli, i.e. a so-called ‘inertial’ or short-term inhibition effect, following the cessation of rhythmic respiratory motor output? Expiratory muscle EMG consistently shows tonic activity throughout the apnoea (Satoh et al. 2001), suggesting that inspiratory neuronal activity is actively and reciprocally inhibited (Sears et al. 1982). Recordings from medullary respiratory neurones during apnoeas may (Remmers et al. 1986) or may not (Orem & Vidruk, 1998) also show tonic activity of expiratory or post-inspiratory neurones and may even show persistent rhythmic activity of some bulbar inspiratory neurones (Batsel, 1967). So, apnoeas are clearly not quiet, passive events but are alive with a variety of tonic and even rhythmic neural activities. Together, these two Pmath formula thresholds in sleep mean that even small transient reductions in Pmath formula will initiate apnoea and that prolongation of apnoea will exacerbate hypercapnia and hypoxaemia leading to transient arousals and ventilatory overshoots upon apnoea termination.

Lability of the apnoeic threshold and the CO2 reserve

The apnoeic threshold and the CO2 reserve are not fixed quantities. They will change inversely with both the sensitivity of the ventilatory response to CO2 below eupnoea and the plant gain. Thus, increasing the drive to breathe and lowering the prevailing steady-state Pmath formula does not mean that one is at greater risk of inducing apnoea and/or ventilatory instability. Similarly, simply increasing the background Pmath formula via sustained hypoventilation does not protect against crossing the apnoeic threshold. In fact quite the opposite occurs. For example, unstable, periodic ventilatory patterns during sleep in short-term hypoxia or chronic heart failure (CHF) are stabilized when further hyperventilation and hypocapnia are induced by increasing background drive by pharmacological means (Sutton et al. 1979; Javaheri et al. 1996) or by sojourning for longer periods at high altitude (Berssenbrugge et al. 1984).

Influence of background drive

There appear to be two reasons for the increased protection against apnoea and instability afforded by increased ventilatory drive, as illustrated in Fig. 2. First, hyper- or hypoventilation means relocation along the isometabolic inline image hyperbola. Accordingly, with increased background drive and therefore high inline image and low Pmath formula, plant gain is reduced, i.e. the increase in inline image required for a given reduction in Pmath formula is increased. Second, the slope of the decrease in inline image with a reduction in end-tidal CO2 pressure (Pmath formula) below eupnoea showed no significant change from control in the background of hyper- or hypoventilation attending these metabolic acid–base derangements. Thus, the CO2 reserve (i.e. ΔPmath formula required between eupnoea and the apnoeic threshold) changed inversely with the change in plant gain, as ΔPmath formula averaged 1.8 times higher in metabolic acidosis (6.7 ± 0.8 mmHg) than in metabolic alkalosis 3.7 ± 1.0 mmHg) (Nakayama et al. 2002).

These changes in plant gain accompanied by no change in the ventilatory sensitivity to reductions in CO2 below eupnoea markedly enhance the likelihood that the apnoeic threshold is crossed (with further transient ventilatory perturbations) in metabolic alkalosis and greatly reduces this likelihood in metabolic acidosis. For example, as shown by contrasting Fig. 3A and B, the enhanced susceptibility to apnoea and to periodic breathing during the hypoventilation of metabolic alkalosis was manifested: (a) consistently in response to small transient reductions in Pmath formula accompanying the transient increases in tidal volume (VT) at even the lowest levels of pressure support ventilation (PSV); and (b) occasionally during spontaneous eupnoea (i.e. with no PSV applied) – periodic breathing occurs over several minutes sometimes triggered by a single augmented breath but other times with no apparent triggering other than the background hypoventilation.

Figure 3.

Apneic threshold lability
A, polygraph record in a tracheostomized dog during NREM sleep of a pressure support trial (PSV) during 5 h of metabolic acidosis achieved via oral acetazolamide administration (arterial pH = 7.34, [HCO3] 16 mequiv l−1, Pmath formula 30 mmHg). The protocol used to determine the apnoeic threshold required several trials of incremental increases in PSV. Note that with metabolic acidosis a tracheal pressure (PM) of over 20 cmH2O, an increase in VT of over 2× baseline eupnoea and a reduction in Pmath formula of 7 mmHg was required to cause apnoea and subsequent periodic breathing. B, polygraph record of a pressure support trial in NREM sleep during 1 h of metabolic alkalosis via i.v. NaHCO3 (arterial pH = 7.51 [HCO3]= 35 mequiv l−1, Pmath formula= 44 mmHg). Note that a tracheal pressure of 9 cmH2O induced via PSV was required to cause apnoea and subsequent periodic breathing. Contrast the CO2 reserve of 3 mmHg (eupnoea – apnoeic threshold ΔPmath formula) in this condition of background hypoventilation with that of 7 mmHg in the condition of background hyperventilation via metabolic acidosis (Fig. 3A). C, polygraph record of a pressure support trial in NREM sleep after ∼1 h of hypoxia (Pmath formula= 47 mmHg, Pmath formula= 31 mmHg). Despite a similar level of background hyperventilation as with metabolic acidosis (see Fig. 3A), in hypoxia the CO2 reserve (eupnoeic Pmath formula– apnoeic threshold Pmath formula) was reduced by one-half, because hypoxia increased the slope of the CO2 response below eupnoea. (Also see Fig. 2, bottom panel.) Note, in the 3 PSV trials shown in Fig. 3A–C, that diaphragm EMG amplitude is reduced on the first breath of PSV, an effect which we also found when Pmath formula was held normocapnic throughout PSV trials. These non-chemoreceptor, neuro-mechanical inhibitory effects of PSV averaged reductions of 5–20% in EMGdi and 30% prolongations of TE (Nakayama et al. 2002).

Influence of the type of background drive

In hypoxia, the accompanying hyperventilation and reduced plant gain – by itself – protects against apnoea and instability, just as occurred with increased non-hypoxic stimuli to breathe (e.g. pharmacological carotid chemoreceptor stimulation and metabolic acidosis – see Figs 2 and 3A). However, in hypoxia the stabilizing factor of reduced plant gain is counteracted by an increase in the slope of the ventilatory response to reduced Pmath formula below eupnoea, thereby greatly reducing the magnitude of the CO2 reserve between eupnoea and apnoea (see Fig. 2; bottom panel and Fig. 3C) (Xie et al. 2001; Nakayama et al. 2002). The effect of even relatively mild hypoxia (Pmath formula 54–64 mmHg) was especially evident in the sleeping human in whom the average CO2 reserve between eupnoea and apnoea was reduced by two-thirds from 3.4 to 1.1 mmHg ΔPmath formula. In the presence of hypoxia, the ventilatory response to CO2above eupnoea is also increased in slope (Cunningham et al. 1986) and this would cause greater ventilatory overshoots in response to transient chemical stimuli, thereby further enhancing opportunities to reach the apnoeic threshold.

Many congestive heart failure (CHF) patients experience highly unstable and periodic ‘Cheyne-Stokes’ respiration in sleep and this often coincides with chronic hyperventilation and hypocapnia. As explained above, by itself, this hyperventilation in CHF will decrease plant gain and stabilize breathing pattern. However, several influences apparently override this protection. Many investigators have reported that severe CHF is accompanied by increased peripheral chemo-responsiveness (Sun et al. 1999) and prolonged circulation times, both of which would contribute to ventilatory overshoot and periodicity. In addition the CO2 reserve below eupnoea is markedly reduced to less than one-half normal in these patients, for two reasons (Xie et al. 2002). First, as is the case with hypoxia, the ventilatory response slope to CO2 below eupnoea is significantly increased in CHF patients (see Fig. 2). Second, these patients experience little or no significant hypoventilation and therefore no increase in Pmath formula in the transition from awake to NREM sleep (also see below).

Influence of REM sleep

In (phasic) REM sleep the CO2 reserve is widened and much more variable than in NREM sleep (Xi et al. 1993). Accordingly, central apnoeas and periodic breathing following a ventilatory overshoot are rare in REM sleep, just as are augmented ventilatory responses to chemoreceptor stimuli (Phillipson & Bowes, 1986). Thus, while tachypnea and a variable VT are common in REM sleep, apnoea and periodic breathing occur very rarely, even in CHF or hypoxia or central sleep apnoea syndrome (Berssenbrugge et al. 1983; Hanly et al. 1989). In REM sleep, ponto-medullary inspiratory neurones are markedly excited (Orem, 1994) and apparently (like behavioural inputs in wakefulness) cause dissociation of the respiratory pattern generator from the usual excitatory or inhibitory inputs from chemo- and mechano-receptors (Phillipson & Bowes, 1986; Smith et al. 1997).

Causes of changes in ventilatory responsiveness below eupnoea and reductions in the CO2 reserve

Why do conditions of sustained hyperventilation elicited by hypoxia and CHF cause an increased slope of the ventilatory response to CO2 below eupnoea and a reduced CO2 reserve in sleep? A popular concept suggests that a dominance of peripheral over central chemoreceptor contributions to the total ventilatory drive creates instability (Cherniack & Longobordo, 1994; Khoo, 2000). This opinion is based on evidence of ventilatory instability in conditions such as hypoxia, medullary chemoreceptor ablation and carotid chemoreceptor sensitization (via dopamine receptor blockade). However, this hypothesis was not supported by more direct tests in the sleeping animal. First, as shown in Fig. 2, non-hypoxic pharmacological stimulation of carotid chemoreceptors and ventilation using almitrine (which, like hypoxia increases carotid body and ventilatory sensitivity to CO2; Nishino & Lahiri, 1981), did not mimic the effects of hypoxia by increasing the slope of the ventilatory response below eupnoea and narrowing the CO2 reserve (see Fig. 2). Rather, with almitrine-induced hyperventilation, the slope of the CO2 response below eupnoea remained unchanged from control and the CO2 reserve was widened in proportion to the decrease in plant gain – just as occurred with (normoxic) metabolic acidosis.

The same rationale led us to try the reverse experiment. We recently used low-dose dopamine infusion to cause hypoventilation (i.e. increased Gp) and to reduce the CO2 responsiveness of the carotid chemoreceptors (Lahiri et al. 1980, 1989; Nishino & Lahiri, 1981; Sabol & Ward, 1987). Like metabolic alkalosis (see above) the slope of the CO2 response below eupnoea remained unchanged from control (see Fig. 2); thus the CO2 reserve was narrowed in proportion to the increase in plant gain. During dopamine infusion, brief apnoeas and periodic breathing occurred occasionally during spontaneous breathing in sleep and consistently with small transient perturbations in VT and Pmath formula provided by low levels of pressure support ventilation. This instability in ventilatory control during dopamine infusion contrasted with the marked resistance to apnoea and periodic breathing found during PSV-induced transient hypocapnia following carotid body denervation (CBX) (Nakayama et al. 2003). Perhaps this difference between dopamine and CBX reflects the importance to hypocapnic-induced apnoea and periodicity of at least some level of carotid body CO2 responsiveness – even if markedly subnormal – in the face of a high plant gain and reduced CO2 reserve.

So, contrary to predictions based on the peripheral–central chemoreceptor imbalance concept (Khoo, 2000), non-hypoxic stimulation of the carotid chemoreceptors combined with hypocapnic inhibition of the central chemoreceptors, widened the CO2 reserve and enhanced ventilatory stability. Furthermore, non-hypoxic inhibition of the peripheral chemoreceptors combined with hypercapnoeic stimulation of the central chemoreceptors narrowed the CO2 reserve and made the control system more susceptible to apnoea and periodicity.

Through what other mechanisms might hypoxia increase the slope of the ventilatory response to reduced Pmath formula below eupnoea and move the eupnoeic Pmath formula so close to the apnoeic threshold? First, perhaps a unique sensory output from the carotid chemoreceptors is elicited by hypoxia which is quite different from pharmacological stimuli. Certainly different stimuli such as hypoxia, Pmath formula and pH have been shown to activate quite different types of receptors within the carotid chemoreceptor (Summers et al. 2002).

A second potential mediator of these effects might be found in the central inhibitory effects of hypoxia on ventilatory control, as has been well documented in anaesthetized animals (Neubauer & Bisgard, 1994), but at present the findings on this question in unanaesthetized preparations are contradictory. On the one hand, the well-known cerebral vasodilatory effects of hypoxia would be expected to inhibit breathing by reducing medullary Pmath formula for any given arterial Pmath formula. For example, 30–40% swings in middle cerebral artery blood flow were recently found to accompany periodic breathing in NREM sleep (mostly Stage II) in hypoxic humans (see Fig. 4). The observed increases in cerebral blood flow correspond to large reductions of 6–8 mmHg in the Pmath formula difference between arterial and jugular venous blood (Chapman et al. 1979). In awake humans we reduced the slope of the cerebral blood flow response to hypercapnia by about 65% via indomethicin administration, and this caused an average 50% increase in the steady-state ventilatory response to inhaled CO2 (Xie et al. 2003). This study and other earlier studies using fixed reductions in cerebral blood flow in awake goats (Chapman et al. 1979) show that changes in cerebral blood flow clearly influence the chemical control of breathing, presumably acting via changes in brain chemoreceptor Pmath formula. On the other hand, in sleeping unanaesthetized dogs in which the isolated and extra-corporally perfused carotid chemoreceptors were maintained normocapnic and normoxic (Curran et al. 2000), cerebral hypoxia per se was found to stimulate – not depress – eupnoeic ventilation. A similar intact, isolated carotid chemoreceptor preparation used in the awake goat showed that short-term potentiation of ventilatory output immediately following carotid chemoreceptor stimulation was not affected by CNS hypoxia (Engwall et al. 1994). This fundamental question concerning the influence of CNS hypoxia and cerebral blood flow on ventilatory control needs to be further addresed in physiological preparations – both awake and asleep.

Figure 4.

The effects of one night of hypoxic exposure ( F math formula= 0.11 ) on periodic breathing and middle cerebral artery (MCA) blood flow during NREM sleep in the healthy human
Shown are signal averaged data over 150 periodic breathing cycles in a single healthy subject. These data are representative of those found in six additional subjects. Each periodic cycle averaged 22 s in duration and consisted of three hyperpnoeic tidal breaths followed by an 8–12 s apnoea. Note the average 30% increase in MCA blood velocity (CBF) as determined by Doppler ultra-sound measurements and the increase in mean arterial pressure (MAP) which began at the termination of the apnoea and peaked during the ventilatory overshoot phase. Immediately following the ventilatory overshoot, CBF fell 10% below baseline control (data from authors' laboratory). A transfer function analysis of the gain, phase and coherence between MCA flow velocity and MAP changes showed that changes in cerebral blood flow throughout the cycle always led those in MAP. Based on this correlative analysis and previous ganglionic blockade studies of breath-hold apnoea in awake subjects (Przybylowski et al. 2003), we attribute the increases in CBF to the transient hypoxaemia caused by the apnoeas and the subsequent decreases in CBF below baseline to the transient hypocapnia caused by the ventilatory overshoot. These marked increases in CBF may exacerbate brain extra-cellular fluid (ECF) hypocapnia, thereby reducing the CO2 reserve and precipitating apnoea and periodic breathing. SaO2, arterial O2 saturation.

Why a reduced CO2 reserve in CHF?

The reduced CO2 reserve observed with CHF patients cannot be attributed to arterial hypoxaemia which usually is absent in CHF. Correlative evidence suggests that elevated pulmonary capillary wedge pressures may stimulate a vagally mediated hyperventilation in these patients (Solin et al. 1999). Whether this specific stimulus would, like hypoxia, sensitize the slope of the CO2 response below eupnoea and reduce the CO2 reserve remains unknown.

Another promising explanation for the reduced CO2 reserve in CHF might be found in a blunted cerebrovascular endothelial response to CO2. The effects of CO2 on cerebral blood flow provide an important counter-regulatory mechanism which serves to minimize changes in brain [H+], thereby stabilizing the breathing pattern in the face of perturbations in Pmath formula. For example, hypocapnia normally causes marked cerebral vasoconstriction and reduced cerebral blood flow which attenuates the fall in brain Pmath formula relative to that in arterial blood. Accordingly, ventilatory inhibition in response to reduced Pmath formula will be lessened because of the attenuated decrease in central chemoreceptor [H+]. Patients with CHF (Georgiadis et al. 2000) and an animal model of CHF (Caparas et al. 2000) show reduced cerebral blood flow responses to CO2. More specifically, our recent data showed that CHF patients with periodic breathing in sleep demonstrate a significantly attenuated reduction in cerebral blood flow response to hypocapnia relative to CHF patients without periodicity (Xie et al. 2003).

Again, an important potential role for cerebral blood flow in the regulation of breathing and breathing stability is implicated. Its true contributions will depend upon the relative importance of central chemoreceptor Pmath formula in causing apnoea. In this regard, the elegantly documented increase in carotid chemoreceptor responsiveness in an animal model of CHF (Sun et al. 1999) points to sensitization of peripheral chemoreceptor inputs as a potentially important determinant of a sensitized apnoeic threshold and reduced CO2 reserve.

To summarize the findings of changing susceptibilities to apnoea and periodic breathing, we believe there is no easily determined, single index such as loop gain, at least as presently conceived, which predicts ventilatory instability in sleep. Rather, at a minimum it is important to consider plant gain together with several elements of controller gain in response to ΔPmath formula both above eupnoea (i.e. the gain of the response to increased ventilatory stimuli – both chemical and state-related) and below eupnoea. The change in the spontaneous, eupnoeic Pmath formula in the transition from wakefulness to sleep may also have an important bearing on the magnitude of the CO2 reserve and ventilatory stability during sleep. All of these variables are significant determinants of the propensity for both ventilatory overshoot and undershoot.

Ventilatory control instability and obstructive sleep apnoea

How might the occurrence of instability in central respiratory motor output during sleep relate to the major clinical problem of repetitive airway obstruction and high resistance hypopnoeas in the ‘obstructive sleep apnoea syndrome’ (OSA)? Conventional opinion would cite the highly significant risk factors of obesity and the volume of upper airway soft tissue structures and cranio-facial abnormalities in OSA (Dempsey et al. 2002; Schwab, 2003) to support the idea that the inherent passive collapsibility of upper airway is the dominant determinant of OSA and that sleep promotes the obstruction by removing the abnormally augmented activation of upper airway dilator muscles present during wakefulness in OSA patients (White, 2002). However, there is accumulating evidence to demonstrate that neuro-mechanical control of the stability and coordination of motor output to both the respiratory pump muscles and the resistance muscles of the upper airway during sleep is also likely to be a significant player in many types of repetitive sleep apnoeas. For example, many OSA patients display an enhanced controller (and therefore loop) gain during sleep, which would predispose to periodic breathing (Younes et al. 2001). When these types of patients were tracheostomized in order to bypass the upper airway, marked periodic breathing during sleep was revealed (Onal & Lopata, 1982). Furthermore, the severe OSA patient's ability to effectively compensate (via effective neural control of the respiratory muscles of both the upper airway and the pump, without EEG arousal) for airway narrowing and increased mechanical load was shown by correlational analysis to be a more important determinant of the degree of cycling behaviour of airway patency and ventilation than was the inherent passive collapsibility of the airway (Younes, 2003).

An implication of these clinical data is that an inherently collapsible airway may allow for significant airway narrowing and even obstruction in sleep, but any subsequent repetitive cycling behaviour in airway patency and ventilation is critically dependent upon neuro-chemical control mechanisms. What is not known from these reports in long-standing OSA patients is the extent to which their predisposition towards instability in neural control is inborn and/or is exacerbated by chronic intermittent hypoxia – a well-known cause of increased carotid chemoreceptor gain and CNS neuro-transmitter turnover (Ling et al. 2001; Peng et al. 2001).

In addition to these clinical examples outlined above, there are many experimentally documented links between respiratory neural control and airway patency. First, unstable oscillations in respiratory motor output produced in sleep during the transition from normoxia to hypoxia (and from steady-state hypoxia to normoxia) will cause airway obstruction which occurs at the nadir of the cycling respiratory motor output (Warner et al. 1987). Likewise, in OSA patients the nadir of upper airway muscle EMG activity coincides with the onset of airway obstruction (Suratt et al. 1985). We emphasize that an oscillating respiratory motor output precipitated by transient hypoxia caused airway obstruction during periods of low respiratory motor output only in those subjects who had upper airways that were highly compliant and already narrowed prior to the induction of an oscillating respiratory drive (e.g. a snorer with an airway resistance (Rua) of 20–50 cmH2O l−1 s−1 during sleep). However, even in these subjects, if the hypoxia was continued for several minutes leading to full-blown, cluster-type periodic breathing, airway obstruction did not persist. On the contrary, upon re-initiation of breathing rhythm at the termination of each apnoea, inspiratory motor output and flow rate were very high and Rua was reduced to near waking levels (Warner et al. 1987). Apparently the steep crescendo in chemoreceptor (plus arousal)-driven inspiratory motor output upon apnoea cessation must include marked and near-simultaneous activation of both pharyngeal dilator and inspiratory pump muscles.

Second, bronchoscopy performed during induced or naturally occurring central apnoeas, revealed that upper airway obstruction occurs early (< 10 s) during the central apnoeic period and even in the absence of inspiratory efforts (Badr et al. 1995). Apparently, all so-called ‘central’ apnoeas have an ‘obstructive’ airway component, but this is usually not detectable in the absence of inspiratory effort.

Third, obstructive apnoeas will precipitate central apnoeas and oscillatory ventilation. For example, with airway obstruction, accumulation of chemoreceptor stimuli will cause ventilatory overshoots and hypocapnia, followed by central apnoea or hypopnoea (Iber et al. 1986). Transient arousals at apnoea termination are likely to enhance the magnitude of the overshoot and subsequent cycling. In addition, the mechanical events occurring in a narrowed or obstructed airway, such as negative pressure, pharyngeal tissue deformation and/or palatal tissue vibration at high frequency (simulating snoring) are excitatory to pharyngeal muscle activity but inhibitory to respiratory pump muscles (Eastwood et al. 1999). These dual reflex responses to airway pressures serve to preserve or restore airway patency (Brouillette & Thach, 1979) but also act to reduce respiratory motor output to the pump muscles and will even cause central apnoea if the airway pressure changes occur during the expiratory phase (Harms et al. 1996).

These many links between the stability of respiratory motor output and airway resistance have implications for both the diagnosis and treatment of OSA. For example, routine non-invasive polysomnography does not include independent, sensitive, measures of respiratory motor output or effort; nor does it provide accurate assessment of airway calibre during an apnoea in the absence of inspiratory effort (Kryger, 1994; Morrell et al. 1995). Accordingly, reductions in central respiratory drive preceding an obstruction or the prevalence of truly ‘mixed’ (central plus obstructive) apnoeas are likely to be underestimated. Future treatment of OSA may need to focus on both optimizing neural drive to the upper airway muscles and on preventing subsequent apnoeas and cycling by stabilizing respiratory motor output (Strohl, 2003).



Original research from the author's laboratory reported in this review was supported by NHLBI, the American Heart Association, and the Veterans Administration Merit Review. We are indebted to Kathy Henderson and Ben Dempsey for technical assistance.