INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

IT IS NOW WELL ESTABLISHED that ventilatory instability during sleep onset can be driven purely by state instability (6, 7, 23, 31). There are also data suggesting that the magnitude of state-related ventilatory fluctuations during sleep onset is related to the number of hypopneas and apneas experienced during established sleep (30). In addition, two recent experiments have shown that the magnitude of state-related ventilatory fluctuations is amplified over the course of the sleep-onset period (10, 30). Dunai et al. (10) also demonstrated that this amplification effect is due, at least in part, to peripheral chemoreceptor activity, because hyperoxic peripheral chemoreceptor inhibition reduced it.

A feature of both of these experiments (10, 30) was the presence of considerable intersubject variability in both the magnitude of amplification and the peripheral chemoreceptor contribution to the amplification. Thus, even though all subjects experienced amplification of state effects on ventilation under normoxic conditions, in some subjects the effect was small, whereas in others it resulted in very large state-related fluctuations in ventilation by the attainment of stable stage 2 sleep. Similarly, the effect of hyperoxia on the amplification ranged from a very small reduction to almost complete elimination. Furthermore, subjects who experienced the greatest amplification under normoxic conditions were also those who experienced the greatest reduction in amplification during hyperoxia. These data suggest that individuals with a large peripheral chemoreceptor contribution to the amplification are likely to experience greater state-related ventilatory instability and may, therefore, be predisposed to develop sleep-disordered breathing later in established sleep.

One factor that could lead to differences in the peripheral chemoreceptor contribution to the amplification is differences in peripheral chemosensitivity, which is known to vary considerably within the normal population (2, 9, 11, 29, 33). This hypothesis is consistent with the explanation for the amplification effect outlined by Dunai et al. (10), which predicts that individuals with high peripheral chemosensitivity will have a larger peripheral chemoreceptor contribution to the amplification and will, therefore, experience greater amplification of state-related ventilatory instability. This is because feedback delays to peripheral chemoreceptors result in delayed ventilatory responses to state-induced perturbations, leading to progressive increases in the amplitude of fluctuations in peripheral chemoreceptor drive (PCD), and therefore, to progressive amplification of state-related fluctuations in ventilation. A corollary of this is that people with stronger PCD will experience greater amplification than people with low PCD. This is because a larger chemical component is added to state effects at each state transition, producing greater hyperventilation for a given level of chemical drive, greater hypoventilation with the return of sleep, and even greater hyperventilation at a subsequent arousal.

However, although differences in peripheral chemosensitivity are theoretically capable of explaining intersubject variability in the amplification of state-related ventilatory fluctuations, there is little supporting empirical evidence. Despite consistent evidence of a relationship between peripheral chemosensitivity and respiratory instability during sleep at high altitude (22, 34), and the results of modeling studies identifying peripheral chemoreceptors as the most likely source of respiratory instability under normoxic conditions (20), data obtained in sea-level experiments have been contradictory. Chapman et al. (3) were unable to demonstrate a relationship between hypoxic sensitivity and the degree of respiratory instability induced in healthy subjects by artificially increasing the peripheral chemoreceptor response to hypoxia at sea level, although they did observe a relationship with combined hypoxic hypercapnic and hypercapnic sensitivity. Garay et al. (12) were unable to demonstrate a relationship between hypoxic or hypercapnic ventilatory responses and severity of obstructive apnea in obstructive sleep apneic patients. However, the failure to find a relationship between hypoxic sensitivity and respiratory instability in both of these experiments may have been due to a preponderance of subjects with low to moderate hypoxic sensitivity in the samples. Mean hypoxic ventilatory responses (HVRs) in stable and unstable subjects in the study by Chapman et al. (3) were 0.61 ± 0.24 and 0.5 ± 0.34 l · min¹ · % arterial O₂ saturation (Sa_O2)¹, respectively, compared with group means ranging from 0.56 to 1.07 l · min¹ · % Sa_O2¹ reported in previous studies (e.g., 2, 9, 11, 25, 29, 33). Similarly, HVRs obtained in Garay et al. (12) ranged from 0.1 to 0.9 l · min¹ · % Sa_O2¹ compared with the authors' reported laboratory range of 0.2 to 2.8 l · min¹ · % Sa_O2¹ and HVRs ranging from 0.03 to 1.67 l · min¹ · % Sa_O2¹ reported in previous experiments (11, 25, 29).

Experiments comparing chemosensitivity in sleep apneic patients and healthy control subjects have also produced highly inconsistent results. Compared with those in normal control subjects, HVRs have been reported to be lower or higher in central apneic patients, lower or the same in obstructive apneic patients, higher in mixed apneic patients, and the same in unclassified sleep apneic patients (17, 28). Results that were similarly inconsistent were obtained for central hypercapnic sensitivity, which was elevated in central apneic, the same in obstructive apneic, possibly higher in mixed apneic, and lower in unclassified apneic patients (17, 28, 35).

It is difficult to come to any firm conclusion concerning the relationship between peripheral chemosensitivity and respiratory instability from these results. Although theoretical considerations and modeling and high-altitude studies suggest a relationship, the results of sea-level investigations have been inconsistent, although, as indicated above, this may be due to the restricted range of peripheral chemosensitivities in the samples used in these experiments. For this reason, the purpose of the present experiment was to investigate the relationship between peripheral chemosensitivity and the amplification of state effects on ventilation by comparing normoxic and hyperoxic amplification as a function of PCD in subjects with high and low peripheral chemosensitivity. It was predicted that subjects with high peripheral chemosensitivity would experience greater amplification of state effects on ventilation under normoxic conditions and greater reduction of the amplification under hyperoxic conditions, than would subjects with low peripheral chemosensitivity.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The experiment consisted of two parts. In part 1, waking HVRs were measured in a large sample of subjects so that subjects with low and high PCD could be identified. This method of subject selection was adopted because of a lack of sufficient normative HVR data using the presently most widely accepted method of HVR assessment. In part 2 of the experiment, ventilation was measured throughout sleep onset under normoxic and hyperoxic conditions in both high- and low-PCD subjects. The experiment was approved by the University of Melbourne Human Ethics Committee, and all subjects gave written informed consent.

Part 1: Measurement of HVRs

Subjects. Forty male psychology undergraduate students, aged between 18 and 29 yr (mean 21.05 yr), participated in part 1 of the experiment as part of course requirements. All subjects were healthy nonsmokers with no history of respiratory or sleep-related disorders.

Equipment. HVRs were measured by using the rebreathing method developed by Rebuck and Campbell (24). The test apparatus consisted of a 10-liter Douglas bag inside a box. A Vacumed adult mouthpiece was connected to the bag via two Hans Rudolph two-way nonrebreathing valves (model 2600) and a port through the box. The two-way nonrebreathing valves were used to isolate expiratory gas so that end-tidal PCO₂ (PET_CO2) could be measured. A port in the section of tubing containing expiratory gas was connected to an Ametek CO₂ analyzer (model CD-3A). A Dynavac variable-flow vacuum pump (model G12/09) drew bag gas through a soda lime CO₂ absorber and returned it through a port in the end of the bag farthest from the mouthpiece. Pump flow rate was regulated by a controller connected to the CO₂ analyzer. Pump activity was decreased when PET_CO2 fell below a preset value and increased when PET_CO2 rose above it. Airflow was measured by using a Morgan pneumotachograph attached to a port in the opposite end of the box to the mouthpiece and a Validyne differential pressure transducer (model DP-45). Sa_O2 was measured by using an Ohmeda Biox 3700 pulse oximeter and ear probe.

Procedure. Subjects were tested over a 2-h period during the daytime and were requested to refrain from drinking coffee or alcohol before the test session. Waking HVRs were used as the index of PCD because, according to the model described at the beginning of this study, the amplification effect is driven by ventilatory responses to Sa_O2 during arousals from sleep (i.e., during wakefulness). Subjects were not permitted to eat or drink (apart from water) during the 2-h test period. Testing consisted of one baseline measurement period and four HVR tests, each lasting 5-8 min and separated by a 20-min rest period. Baseline measurements consisted of 5 min of normal quiet breathing, during which resting minute ventilation (E) and PET_CO2 were continuously measured. Average PET_CO2 for this 5-min period was used as a subject's baseline level during HVR tests. For the HVR tests, the CO₂ controller was set to an individual subject's baseline PET_CO2, and the bag was filled with a gas mixture consisting of 21% O₂-remainder N₂. Subjects were asked to begin with three large breaths to facilitate mixing contents between bag and lungs and then to breathe as naturally as possible until requested to stop. The test was terminated when Sa_O2 had decreased to 75%. The bag was then emptied, flushed with test gas, and refilled before the commencement of the next HVR test.

Data analysis. Data from the first HVR test were discarded, and the slope of the flow-Sa_O2 response line was determined for each of the three remaining rebreathing tests by using regression analysis. Individual subjects' HVRs were defined as the average slope from these three rebreathing tests. Analysis of CO₂ data confirmed that PET_CO2 was maintained within 1.5 Torr of baseline levels in all subjects.

Part 2: Sleep Studies

Selection of subjects. The purpose of part 2 was to select two groups of subjects, a low-PCD group, consisting of the six subjects with the lowest HVRs, and a high-PCD group, consisting of the six subjects with the highest HVRs. However, not all of the subjects with the most extreme HVRs agreed to participate in the experiment. Thus the low-PCD group consisted of the subjects with the second, third, fourth, fifth, eighth, and eleventh lowest HVRs, and the high-PCD group consisted of the subjects with the first, third, fourth, fifth, sixth, and seventh highest HVRs. Nonetheless, the two groups had very different hypoxic sensitivities. The mean HVR for the low-PCD group was 0.36 l · min¹ · % Sa_O2¹ (range 0.14 to 0.57 l · min¹ · % Sa_O2¹), whereas the mean HVR for the high-PCD group was 2.14 l · min¹ · % Sa_O2¹ (range 1.58 to 3.34 l · min¹ · % Sa_O2¹). Results of a group-by-HVR ANOVA confirmed that these group differences were highly significant [F(1,10) = 45.69, P < 0.001]. Mean ages for low- and high-PCD subjects were 19.67 ± 1.51 and 21.83 ± 2.32 yr, respectively. Mean weights for low- and high-PCD subjects were 68.83 ± 6.74 and 72.33 ± 10.63 kg, respectively.

Design. Ventilation was measured throughout sleep onset under normoxic and hyperoxic conditions in low- and high-PCD subjects. Ten of the subjects were studied for two nonconsecutive nights per condition. The remaining two subjects were studied for one night per condition. The order of normoxic and hyperoxic conditions was counterbalanced so that one-half of the subjects experienced the normoxic condition first, and the other one-half experienced the hyperoxic condition first.

Procedure. On experimental days, subjects were required to abstain from coffee and alcohol from 1200 onward. They arrived at the sleep laboratory at ~2200 so that electrodes could be attached and were generally in bed by ~2300. Before the commencement of data collection, a mask (CIG rubber anesthetic mask) was checked for leaks by instructing subjects to exhale strongly against an obstructed expiratory line while checking for air escaping around the edge of the mask. Detected leaks were corrected by tightening and/or repositioning the mask. Data collection did not commence until all existing leaks had been eliminated. Subjects were asked to notify the experimenter if they become aware of any subsequent leakage. Because subjects were awakened frequently as part of experimental procedures, the experimenter was able to check the mask regularly. As a consequence, unidentified mask leakage was unlikely.

Respiratory measurements. Subjects wore the CIG rubber anesthetic mask (size 5 or 6), which covered the face and nose and was held on by a headstrap. A Hans Rudolph two-way nonrebreathing valve (model 2600) was attached to the mask. Mask plus valve dead space was ~127 ml depending on the individual subject's facial configuration. Expiratory airflow was measured by using a heated Morgan pneumotachograph connected to a Validyne differential pressure transducer (model DP-45). The output of the pressure transducer was converted to a voltage signal by a Validyne carrier demodulator (model CD15). Subjects' expired air was vented out of the bedroom via 152 cm of 35-mm-ID tubing. Airflow was calibrated by using a Sho-rate flowmeter (model 1355). CO₂ and O₂ percentages were measured by using an Ametek CO₂ analyzer (model CD-3A) and Ametek O₂ analyzer (model S-3A/I), respectively, connected to the breathing valve by 1-mm-ID tubing. Gas analyzers were calibrated by using ambient fresh air and gases with known CO₂ and O₂ concentrations. Sa_O2 was measured by using an Ohmeda Biox 3700 pulse oximeter and ear probe. Temperature was maintained at between 20 and 24°C.

Control of inspired O₂ levels. On hyperoxic nights, subjects inspired through 86 cm of 35-mm-ID tubing that connected the inspiratory side of the breathing valve to a 15.14-liter high-density polyethylene barrel. The tubing extended ~1.5 cm into the barrel through a tightly sealed hole in the lid. A second 240-cm length of 35-mm-ID tubing was inserted ~23 cm into the barrel through an adjacent hole in the lid. The other end of this piece of tubing was left open to room air. A port 152 cm along this piece of tubing was connected by 4 m of 5-mm-ID tubing to a tank containing 100% O₂ situated outside the bedroom. O₂ entered the tubing under pressure from the tank. By using this setup, normal inspiration drew both room air and 100% O₂ from the second piece of tubing into the barrel, where it was mixed by turbulent flow and then inhaled via the tubing connecting the barrel to the breathing valve. The amount of O₂ bled into the inspiratory tubing was adjusted, via a flow rate meter attached to the O₂ tank, until the subject's inspired O₂ levels were raised to 40%. The level of inspired O₂ fraction indicated by the O₂ analyzer was monitored constantly throughout data collection to ensure that it remained between 40 and 44%. On normoxic nights, subjects inspired room air through the same inspiratory tubing but with the oxygen tank turned off.

Electrophysiological recordings. To permit the identification of state, electroencephalographic (EEG) activity was recorded from scalp electrodes attached at central (C₃/A₂) and occipital (O₁/A₂) sites, electrooculographic activity from two electrodes attached above and below the outer canthi of the eyes, and electromyographic activity from two electrodes over the submental muscles. The occipital recording was used for automated discrimination between alpha and theta EEG activity.

Preliminary data analysis. RESPIRATORY VARIABLES. Breath-by-breath values of tidal volume, estimated expiratory E, total duration (TT), PET_CO2, and average Sa_O2 were calculated as described previously (10).

Analyses of ventilatory variables. TRANSITION ANALYSES. Data from sleep onsets were analyzed via software that used sleep/wake state and breath number to identify sequences of consecutive breaths associated with transitions between specified sleep/wake states. This permitted calculation of the magnitude of the change in ventilation associated with each state transition (_diff) by using the following procedure. First, average pretransition E (E_pre) and posttransition E (E_post) were calculated for each transition. E_pre was defined as the mean of the second and third last breaths preceding a state transition, and E_post was defined as the mean of the second and third breaths after a state transition. _diff was then calculated by substrating E_post from E_pre. _diff values were calculated for transitions between alpha and theta activity in phases 2 and 3. The second and third last breaths before a transition and the second and third after a transition were used to define the amplitude of the state effect as previous work has indicated that these points best define the maximum amplitude (31). Note that _diff values were calculated for both possible transition types, i.e., transitions from alpha to theta and from theta to alpha activity.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Amplification During Normoxia as a Function of Peripheral Chemosensitivity

View larger version (18K): [in this window] [in a new window]   Fig. 1.   Magnitude of state-related fluctuations in ventilation (diff) during phases 2 and 3 during normoxia (A) and hyperoxia (B) in subjects with low (left) and high (right) peripheral chemoreceptor drive (PCD).  to , Alpha to theta;  to , theta to alpha. Bars, SE.

View this table: [in this window] [in a new window]   Table 1.   Average normoxic and hyperoxic alpha-to-theta and theta-to-alpha diff values in individual subjects

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Normoxic breath-by-breath changes in expiratory minute ventilation over alpha-to-theta and theta-to-alpha state transitions in low (left)- and high-PCD subjects (right) during phases 2 (A) and 3 (B). Bars, SE.

Effect of Hyperoxia on Amplification as a Function of Peripheral Chemosensitivity

View larger version (27K): [in this window] [in a new window]   Fig. 3.   Average breath-by-breath values of minute ventilation over phase 2 and phase 3 state transitions during normoxia and hyperoxia in low (left)- and high-PCD subjects (right). A: phase 2 alpha-to-theta transitions. B: phase 2 theta-to-alpha transitions. C: phase 3 alpha-to-theta transitions. D: phase 3 theta-to-alpha transitions. Bars, SE.

Two analyses were conducted to determine whether the additional editing procedure applied to hyperoxic data had biased results in favor of the experimental hypothesis (for example, by leading to the exclusion of large state-related ventilatory fluctuations from high-PCD subjects' data). First, analyses of the effects of group, condition, and phase by using unedited hyperoxic data (i.e., including apneas and hypopneas) were conducted. These analyses revealed the same pattern of significant results (although smaller in magnitude), suggesting that the effect of excluding transitions with very low theta ventilation simply strengthened the existing pattern of results. There was a significant main effect of group [F(1,10) = 7.01, P = 0.024], and significant group-by-phase and group-by-condition interactions [F(1,10) = 8.53, P = 0.015] and [F(1,10) = 6.54, P = 0.029], respectively. Second, the numbers of transitions deleted from low- and high-PCD subjects were compared. The mean numbers of transitions deleted from phase 2 data were 3.17 ± 3.07 in low-PCD subjects and 4.08 ± 3.00 in high-PCD subjects. The mean numbers of transitions deleted from phase 3 data were 10.08 ± 14.59 in low-PCD subjects and 13.25 ± 14.35 in high-PCD subjects. Results of a group-by-phase-by-transition-type analysis of variance, with the number of deleted transitions as the dependent variable, confirmed the absence of significant effects of group [F(1,10) = 0.23, P = 0.642], group-by-phase [F(1,10) = 0.09, P = 0.775], group-by-transition type [F(1,10) = 2.69, P = 0.132], or group-by-phase-by-transition type [F(1,10) = 1.25, P = 0.289], indicating that the results of the experiment were not biased by preferential exclusion of large _diff values from phase 3 data in high-PCD subjects only. The results of these analyses demonstrate that the exclusion of transitions with very-low-theta ventilation did not bias results in favor of the experimental hypothesis and was in fact essential for accurate assessment of the effects of hyperoxia: clearly, hyperoxia can have no effect on ventilation if it is not inhaled.

Hyperoxic Hyperventilation

Comparison of the tonic effects of hyperoxia (i.e., averaged over all breaths within states and phases) also suggested the occurrence of hyperventilation and revealed marked group differences in the pattern of effects as a function of condition, state, and phase (see Figs. 4 and 5). In low-PCD subjects, hyperoxia increased E and decreased PET_CO2 in all states and phases (Figs. 4A and 5A). In contrast, the tonic effect of hyperoxia in high-PCD subjects was similar to the effect observed by Dunai et al. (10). Hyperventilation was evident during theta activity in phases 2 and 3 but not during alpha activity. During theta activity, average E was increased and average PET_CO2 decreased in both phases 2 and 3 (2 and 3 in Figs. 4B and 5B). However, during alpha activity, E and PET_CO2 were unchanged in phase 2, and E was decreased and PET_CO2 was unchanged in phase 3 (2 and 3, Figs. 4B and 5B). Although group differences in the tonic effects of hyperoxia on E were not statistically significant [F(1,10) = 0.6, P = 0.455], group differences in the effect of hyperoxia on PET_CO2 were statistically significant, as indicated by a group-by-condition-by-state interaction [F(1,10) = 5.6, P = 0.039]. Thus examination of E and PET_CO2 data averaged both across and within low- and high-PCD subjects suggests the occurrence of a mild degree of hyperoxic hyperventilation consistent in magnitude with the numerous studies that have investigated hyperoxic hyperventilation during wakefulness (e.g., Refs. 1, 15). Although the degree of hyperventilation was small, it has been reported in some detail, as its occurrence has important implications for the interpretation of results.

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Average normoxic and hyperoxic minute ventilation as a function of phase and state. A: low-PCD subjects. B: high-PCD subjects. 2, Phase 2 alpha electroencephalographic (EEG) activity; 3, phase 3 alpha or rearousal EEG activity; 2, phase 2 theta EEG activity; 3, phase 3 theta or stage 2 EEG activity. Bars, SE.

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Average normoxic and hyperoxic end-tidal PCO2 (PETCO2) as a function of phase and state. A: low-PCD subjects. B: high-PCD subjects. Bars, SE. Hyperoxic PETCO2 values were significantly lower than those for normoxic PETCO2 (P = 0.02).

Sa_O2

View larger version (28K): [in this window] [in a new window]   Fig. 6.   Average normoxic and hyperoxic breath-by-breath values of arterial O2 saturation (SaO2) over phase 2 and phase 3 state transitions in low (left)- and high-PCD subjects (right). A-D are as described in Fig. 3. Bars, SE.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of the present experiment clearly demonstrate greater amplification of state-related ventilatory fluctuations during sleep onset in individuals with high peripheral chemosensitivity. This is evident from both normoxic data, which indicated far greater amplification in high-PCD compared with low-PCD subjects, and hyperoxic data, which showed that hyperoxia considerably reduced the amplification in high-PCD but not low-PCD subjects. In addition, the larger state-related ventilatory fluctuations experienced by high-PCD subjects during normoxia were associated with large state-related fluctuations in Sa_O2, which were greatly reduced during hyperoxia, whereas the much smaller state-related ventilatory fluctuations experienced by low-PCD subjects during normoxia were not associated with state-related fluctuations in Sa_O2, and the only effect of hyperoxia was a tonic increase in both ventilation and Sa_O2. Considered together, these findings indicate that the greater degree of amplification experienced by high-PCD subjects under normoxic conditions was largely the result of peripheral chemoreceptor activity. In contrast, the much smaller degree of amplification experienced by low-PCD subjects was clearly not related to the level of peripheral chemoreceptor activation. These results demonstrate that peripheral chemoreceptor activity can contribute to respiratory instability in individuals with high peripheral chemosensitivity, who may therefore be predisposed to experience considerable state-related respiratory instability over the course of the sleep-onset period.

Although the above results indicate that peripheral chemoreceptor activity can contribute to normoxic respiratory instability in certain individuals, they conflict with earlier experiments reporting either the absence of a relationship between peripheral chemosensitivity and respiratory instability or the failure of hyperoxia to improve normoxic respiratory instability. However, it is possible to explain this discrepancy as follows. First, examination of available HVR data from experiments investigating the relationship between peripheral chemosensitivity and respiratory instability indicates low to moderate hypoxic sensitivity in the samples used (3, 12). Thus the absence of a relationship between respiratory instability and peripheral chemosensitivity may have occurred because only subjects unlikely to experience peripheral-chemoreceptor-induced respiratory instability were studied. In support of this interpretation, there is considerable evidence suggesting that chemically induced respiratory instability is likely when peripheral chemoreceptor activity is high, as in subjects with high PCD, or after artificial elevation of peripheral chemosensitivity, under both hypoxic and normoxic conditions (22). Thus Chapman et al. (3) were able to induce respiratory instability in human subjects with low to moderate PCD by artificially augmenting the ventilatory response to hypoxia. Similarly, Lahiri et al. (21) and Cherniack et al. (4) were able to induce respiratory instability in animals by artificially increasing peripheral chemoreceptor gain and to abolish it by administering hyperoxic gas. These findings suggest that high peripheral chemoreceptor activity can be the mechanism underlying both normoxic and hypoxic respiratory instability.

Second, in experiments reporting the failure of hyperoxia to reduce respiratory instability, differences in peripheral chemosensitivity were not controlled for. Thus the samples used may have included subjects with a range of peripheral chemosensitivities. As a result, any positive effects of hyperoxia in high-PCD subjects may have been masked by the absence of positive effects in low-PCD subjects. However, although the failure to control for individual differences in peripheral chemosensitivity can account for the absence of a beneficial effect of hyperoxia in some experiments, it cannot explain consistently observed differences in the effects of hyperoxia on central and obstructive apneas. Hyperoxia typically lengthens the duration of obstructive apneas but decreases both the duration and number of central apneas (16, 18). A possible explanation for this anomaly is that, in addition to the positive effect of potentially reducing the amplitude of fluctuations in chemical drive, hyperoxia may also have the negative effect of exacerbating airway obstruction. In support of this contention, there is evidence that hyperoxic peripheral chemoreceptor inhibition increases upper airway resistance (UAR) because it produces a greater reduction in upper airway muscle activity relative to diaphragm activity (13, 32). Hyperoxia would also tend to lengthen obstructive apneas because the higher prevailing Sa_O2 would increase the time between apnea onset and a fall in Sa_O2 sufficient to stimulate ventilation strongly enough to reinstitute breathing against an obstructed airway. For these reasons, the elimination of PCD during hyperoxia is likely to increase both the incidence and duration of obstructive apnea in individuals susceptible to airway collapse, exacerbating rather than alleviating respiratory instability. This interpretation does not imply that peripheral chemosensitivity is irrelevant for the development of obstructive respiratory instability. Rather, it is possible that the exaggerated pattern of alternating over- and underbreathing induced by hyperoxia in patients with obstructive apnea will be greater in individuals with high PCD. In support of this, Sforza et al. (27) found a positive relationship between peripheral (but not central) chemosensitivity and respiratory effort during obstructive apnea. Thus it is possible that the failure of hyperoxia to stabilize obstructive apnea occurs not because PCD is unrelated to obstructive apnea, but rather because it is the result of the deleterious effect of hyperoxic peripheral chemoreceptor inhibition on UAR in individuals predisposed to upper airway obstruction.

Some aspects of the results of this experiment require further explanation. An important consideration is the manner in which hyperoxia appeared to reduce the amplification in high-PCD subjects. As in Dunai et al. (10), reduced amplification of state-related fluctuations in ventilation appeared to result mainly from increased phase 3 theta ventilation with little change in the level of alpha ventilation (see Fig. 3). However, it might have been anticipated that the effect of hyperoxia would have been to reduce state-related fluctuations in ventilation by increasing ventilation during theta activity and decreasing it during alpha activity. That is, hyperoxia would be expected to alter both the peak (alpha ventilation) and the trough (theta ventilation) of ventilatory fluctuations. That this did not occur may be explained by noting that hyperoxia had two separate effects. One, which was the phenomenon under investigation, was hyperoxic peripheral chemoreceptor inhibition. This, as predicted, reduced the amplitude of ventilatory fluctuations associated with transitions between alpha and theta activity in high-PCD subjects. The other effect, hyperoxic hyperventilation, increased ventilation during both states. The combined effect of these two components was to reduce the amplitude of state-related ventilatory fluctuations but move them to a higher overall level of ventilation. Thus ventilation increased during phase 3 theta activity because the trough of ventilatory fluctuations (theta ventilation) was higher (hyperoxic peripheral chemoreceptor inhibition) and because the fluctuations occurred at a higher level of ventilation (hyperoxic hyperventilation). In contrast, ventilation was unchanged during phase 3 alpha activity because, although the peak of ventilatory fluctuations (alpha ventilation) was reduced (hyperoxic peripheral chemoreceptor inhibition), hyperoxic hyperventilation shifted the fluctuations to a higher level of ventilation. Thus during alpha activity the two effects of hyperoxia canceled each other out, whereas during theta activity they were additive. This hypothesis is supported by two aspects of the present results. First, average E and PET_CO2 data indicate that a mild degree of tonic hyperoxic hyperventilation was evident in both high- and low-PCD subjects. Second, group differences in the effects of hyperoxia are also consistent with this hypothesis. In low-PCD subjects the predominant effect of hyperoxia was mild tonic hyperventilation, with no effect on the magnitude of amplification, whereas in high-PCD subjects the predominant effect of hyperoxia was a reduction in amplification, which appeared to result mainly from increased theta ventilation with little or no change in alpha ventilation.

Another aspect of the results that requires clarification is hyperoxic reduction of the amplification during phase 2 as well as phase 3 in high-PCD subjects (see Fig. 1). This was an unexpected finding, as the results of Dunai et al. (10) indicated no effect of hyperoxia on phase 2 _diff values despite a substantial phase 3 effect. However, although unexpected, these results are not inconsistent with peripheral chemoreceptor involvement in the amplification of state effects on ventilation. This is because a corollary of the model of peripheral chemoreceptor amplification described at the beginning of this study is that amplification occurs gradually throughout the entire sleep-onset period, as a result of slightly larger fluctuations in PCD at successive state transitions. Hyperoxic reduction in the magnitude of phase 2 _diff values is entirely consistent with this scenario. The absence of a phase 2 effect in Dunai et al. and the strong phase 2 effect in high-PCD subjects in the present experiment can be accounted for by the different levels of PCD in the two groups of subjects. Subjects with high PCD are likely to experience more rapid amplification of state effects on ventilation because of their stronger peripheral chemoreceptor responses to a given level of arterial PO₂. This means that amplification of state effects will occur more rapidly, and therefore become evident earlier, in sleep onset in high-PCD compared with low-PCD subjects. For this reason, it is not surprising that subjects with high peripheral chemosensitivity begin to show signs of significant amplification before the attainment of phase 3 sleep. As Dunai et al. did not assess peripheral chemosensitivity, it is likely that subjects with a range of hypoxic sensitivities were studied. As a result, average data may have masked hyperoxic reduction of phase 2 _diff values in subjects with higher PCD.

The failure to completely eliminate amplification of state effects on ventilation is also of interest. Although hyperoxia had a marked effect on the magnitude of phase 3 _diff values in high-PCD subjects, the amplification effect was not completely abolished in any subject. This pattern was also observed by Dunai et al. (10), who suggested three possible explanations: failure of 40% O₂ to completely inhibit peripheral chemoreceptors; greater increases in UAR during phase 3; and the operation of a phase 3 neural arousal mechanism. Evidence from numerous earlier experiments suggests that the first explanation is unlikely (e.g., Ref. 14). The occurrence of greater increases in airway resistance during phase 3 is consistent with data showing larger increases in airway resistance during theta activity in phase 3 compared with phase 2 (19). Thus it is possible that amplification of state effects on UAR partially offsets hyperoxic reduction of peripheral-chemoreceptor-induced respiratory instability. This is particularly likely given evidence that hyperoxic peripheral chemoreceptor inhibition increases UAR (13, 32). The operation of an additional arousal-related mechanism during phase 3 was suggested by the results of Dunai et al. (10), who reported less hyperoxic reduction of the amplification at arousals (theta-to-alpha transitions) than at wakefulness-to-sleep (alpha-to-theta) transitions during phase 3. However, the data from the present experiment did not support this interpretation.

The absence of data suggestive of an arousal-related mechanism, and the magnitude of the reduction in amplification during hyperoxia in high-PCD subjects, suggests that much of the increase in ventilation at transient arousals can be explained by the interaction between state and chemical effects, without invoking the existence of an arousal complex (8). However, this contention does not rule out the possibility that greater amplification in high-PCD subjects is the result of a larger number of hypoxia-induced arousals due to lower hypoxic arousal thresholds, rather than high peripheral chemosensitivity per se. There is evidence indicating that hypoxic arousal responses are the result of peripheral chemoreceptor activity, which suggests that subjects with stronger HVRs may also have stronger arousal responses to hypoxic stimuli (5). This could occur either as a consequence of greater hypoxic sensitivity generally or because of a lower hypoxic arousal threshold (i.e., arousal at a higher Sa_O2). If this were the case, the stabilizing effect of hyperoxia in high-PCD subjects could be the result of a reduction of the number of hypoxia-induced arousals rather than a decrease in the magnitude of state-related fluctuations in Sa_O2 (and therefore in ventilation). However, the results of the present experiment do not support this hypothesis. Although hyperoxia reduced the number of arousals in both high- and low-PCD subjects, it did so to the same extent in both groups. Furthermore, the number of arousals experienced by low- and high-PCD subjects was not significantly different during normoxia [t(10) = 0.03, P = 0.98] or hyperoxia [t(10) = 0.05, P = 0.96]. This indicates that hyperoxic reduction of the amplification in high-PCD subjects was not the result of a reduction in the number of hypoxia-induced arousals. Similarly, it is unlikely that group differences in hypoxic arousal thresholds were responsible for either the greater degree of state-related respiratory instability during normoxia or the greater reduction of the amplification during hyperoxia in high-PCD subjects. This is because the level of Sa_O2 at arousals did not differ as a function of peripheral chemosensitivity. This was the case during normoxia [t(10) = 1.17, P = 0.13] and hyperoxia [t(10) = 1.49, P = 0.08]. Thus lower hypoxic arousal thresholds in high-PCD subjects cannot explain either the greater degree of state-related respiratory instability under normoxic conditions or the stabilizing effect of hyperoxia in these subjects. The data from this experiment are therefore consistent with the hypothesis that an interaction between state and chemical effects, which is independent of the effects of arousal per se, can account for much of the increase in ventilation at arousals. This does not rule out the possibility of an additional contribution from arousal mechanisms, which may be responsible for the residual amplification observed during hyperoxia in high-PCD subjects.

In summary, the results of the present experiment demonstrate greater amplification of state-related ventilatory fluctuations during normoxia in high-PCD subjects compared with low-PCD subjects. The results also support the hypothesis that this is related to peripheral chemoreceptor activity because hyperoxia decreased the amplification in high-PCD subjects but had no effect in low-PCD subjects. These findings indicate that peripheral chemoreceptor activity can contribute to respiratory instability under normoxic conditions and suggest that individuals with high peripheral chemosensitivity may be predisposed to develop respiratory instability during sleep.