INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Obstructive sleep apnea (OSA) is characterized by repeated episodes of upper-airway occlusion during sleep. Recurrent cessation of airflow is associated with hypoxia, hypercapnia, and increasing inspiratory effort against the obstructed airway, all of which ultimately lead to a brief arousal, restoration of airway patency, and recovery of blood gases (1, 2). It would be anticipated that the nightly recurrence of such events could with time lead both to changes in ventilatory control and changes in the arousal responses to chemoreceptor stimuli (1, 3). Altered respiratory control, as evidenced by daytime hypercapnia and hypoxemia and diminished ventilatory responses to hypoxia and hypercapnia, has been reported in a proportion of OSA patients (1, 3). However, it is clear that even some patients with severe OSA show little evidence of such abnormalities (1, 5). The factors that predispose to altered respiratory control in OSA, and the mechanisms by which these changes are produced remain unclear. The investigation of these mechanisms has been difficult, partly because of the inherent variability of measures of ventilatory control and the practical difficulties in performing longitudinal studies with human subjects.

We have recently described an animal model of OSA (7) in which dogs are subjected to repeated episodes of upper-airway occlusion during sleep on a nightly basis over periods of months. The model is designed specifically for study of the consequences of OSA, and thus affords a unique method of evaluating the effects of repeated airway occlusion during sleep on ventilatory control in previously normal animals that serve as their own controls. The aim of the studies reported here was to assess the effects of application of the OSA model on resting ventilation, and on the ventilatory and arousal responses to hypoxia and hypercapnia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animals

Five adult dogs of mixed short-haired breeds (weight: 20 to 26 kg) were prepared with a permanent tracheostomy. In a separate operation, a two-channel amplifier-telemetry unit (Model TL10M2-D70-EE or TLM11M3D70-CCP; Data Sciences, Akron, OH) was implanted subcutaneously for electroencephalographic (EEG) and nuchal electromyographic (EMG) recording in a manner slightly modified from that previously reported (7). The ends of the EEG and ground electrodes were threaded through perforated stainless steel washers, which were then screwed onto the skull with stainless steel screws. The EEG electrodes were implanted in the temporoparietal area and the ground electrode was attached to the posterior aspect of the sinus ridge. The upper surface of the washer and the end of the EEG lead at the attachment point to the washer were then covered with dental acrylic. A single loop, approximately 1 cm in diameter, was created on the end of the EMG wires and was sewn down onto a dorsal deep strap muscle of the neck on either side of the midline at the midneck level. The rest of the model equipment was unchanged from that described previously (7).

Ventilatory Studies

During the studies, the dogs were intubated through the tracheostomy with a cuffed endotracheal tube, and airflow was directed through a pneumotachograph (Fleisch No. 2; Roxon, Toronto, Canada) connected to a pressure transducer (MP-45 or P300D; Validyne, Inc., Northridge, CA). End-tidal PCO₂ (PET_CO2) was measured with a calibrated infrared analyzer (LB-2; Beckman Instruments, Mountain View, CA). End-tidal PO₂ (PET_O2) was measured with a calibrated O₂ analyzer (Ametek S-3/A-1; Thermox, Pittsburgh, PA, or Beckman OM-11), and Sa_O2 was then derived from the PET_O2 values, using the equation for the canine hemoglobin-O₂ dissociation curve of Rossing and Cain (10, 11). The flow signal was processed by computer to provide breath-by-breath measurements of tidal volume (VT) (by integration of the flow signal) and respiratory timing variables. Tracheal pressure was measured at the proximal end of the endotracheal tube with a pressure transducer (Statham PM5; Oxnard, CA, or Validyne P300D) to ensure that flow resistance did not increase at low-rebreathing-bag volumes. The EEG and nuchal EMG signals from the implanted telemetry unit were continuously monitored by visual inspection during the ventilatory studies, and sleep state and arousal were identified according to standard EEG and behavioral criteria (11, 12). The EEG signals were displayed on paper (Beckman R711 or Gould TA240; Cleveland, OH), and all other signals were displayed either on paper or on a computer screen (CODAS; Dataq Instruments, Akron, OH).

Progressive Hypercapnia and Hypoxia

Hyperoxic progressive hypercapnia and isocapnic progressive hypoxia were induced by rebreathing according to standard techniques (11). A minimum of three trials were performed for each sleep state in each experimental condition, with four or more trials obtained for wakefulness (W) and slow-wave sleep (SWS) and five or more trials obtained for rapid-eye-movement sleep (REM). For hypoxic trials, careful attention was paid to the CO₂ level, and trials were rejected in which PET_CO2 deviated from the control value by more than 3 mm Hg for two or more breaths during the initial part of the run, or by more than 1 mm Hg after the mixed venous plateau for PET_CO2 had been reached. Rebreathing was continued until the point of movement due to discomfort (W) or arousal from sleep. Trials in which a change in sleep state occurred during rebreathing prior to arousal were discarded.

Protocol

Two of the dogs were studied at McGill University and the other three were studied at the University of Toronto. A virtually identical protocol was followed for application of the OSA model, and all ventilatory measurements were made in a standardized manner. After the animals had completely recovered (> 10 d) from implantation of the telemetry unit, measurements of ventilatory and arousal responses were obtained during daytime sleep studies over a period of approximately 2 wk prior to application of the OSA model (baseline measurements). Once these measurements were completed, the animals were connected to the airway occlusion apparatus, which was applied 7 d per wk for up to 17 wk. The dogs were extubated daily and removed from the apparatus by midmorning, and were taken from the housing facility to the laboratory, where they were kept under constant observation and were prevented from sleeping other than during physiologic studies. During these daytime periods the animals underwent several physiologic measurements in the laboratory during both wakefulness and sleep, including the studies of ventilatory control. Ventilatory control studies were performed within the same 2- to 3-hr period in the middle of the day, and no other studies were performed on the days on which the ventilatory measurements were made. Hypoxic and hypercapnic trials were conducted in random order, and measurements for a given intervention period were made on at least two and on as many as six different days. At the end of the OSA intervention period, the animals were removed from the model system and allowed to resume undisturbed nocturnal sleep. Ventilatory control studies were then repeated 4 wk after withdrawal from the model system (recovery measurements). A consistent daily schedule was maintained throughout the pre-OSA baseline, OSA period, and post-OSA recovery period.

Data Analysis

For resting ventilation, a minimum of 30 breaths from at least three different measurement sessions were averaged for each animal under each experimental condition. Variability of resting ventilation was compared through statistical comparison of the coefficients of variation (CVs) for each variable across the three experimental periods. For rebreathing studies, instantaneous minute ventilation (I) was calculated for all breaths, beginning with the mixed venous plateau, and was displayed graphically against the Sa_O2 or PET_CO2 level. The inflection point of the curve was identified visually by one investigator, and all breaths from the linear portion of the curve, up to the point of movement or arousal, were included for analysis. Sighs and the two breaths following sighs were excluded from the analysis. For each rebreathing trial, the slope and intercept of the curve were determined by simple linear regression, and were recorded along with the inflection and arousal CO₂ or Sa_O2 values. The mean values for I and tidal volume (VT) for the last five breaths prior to arousal were also recorded for each trial as the "arousal I" and "arousal VT."

For comparisons of ventilatory pattern during rebreathing, all breaths for an individual dog for trials under a given experimental condition were grouped together and sorted according to CO₂ or Sa_O2 level. Mean values for each level of chemical stimulus were then calculated for I, VT, breathing frequency, inspiratory time (TI), expiratory time (TE), total respiratory cycle time (Ttot), inspiratory duty cycle (TI/Ttot), and mean inspiratory flow rate (VT/TI). The mean values for individual dogs were then averaged to obtain the group mean. Owing to differences between animals in inflection and arousal points, mean values were not available for all dogs at the same levels of chemical stimulus. Group mean values for ventilatory variables under a given condition (e.g., SWS hypercapnia) were therefore calculated for levels of chemical stimulus (e.g., PET_CO2 = 56 mm Hg) for which there were data points for at least four of the five animals in all three experimental periods (i.e., baseline, OSA and recovery).

Statistical comparisons for most variables were performed using two-way analysis of variance (ANOVA), with the two factors of dog and experimental period. Post-ANOVA comparisons were made through the StudentNewman-Keuls method. For the analysis of ventilatory pattern during hypercapnic and hypoxic stimulation, a three-way ANOVA was first performed to detect a significant effect of CO₂ or Sa_O2 level, respectively. If the effect of CO₂ or Sa_O2 level was significant, post hoc two-way ANOVA was performed for individual levels of CO₂ or Sa_O2 (cf. Figures 2 and 4). A two-value of p < 0.05 was used for statistical significance.

View larger version (34K): [in this window] [in a new window]   Figure 2.   Group mean values for breathing frequency (F, breaths/min), tidal volume (VT, L) and minute ventilation ( I, L/min) at successive levels of hypoxia during rebreathing for W (upper panels), SWS (middle panels) and REM (lower panels). The symbols are the same as in Figure 1. The single asterisk below a group of points indicates that all points are p < 0.05 OSA versus baseline and recovery. When more than one asterisk appears on a graph, the symbols relate only to the adjacent data point.

View larger version (32K): [in this window] [in a new window]   Figure 4.   Similar plot to that in Figure 2 for ventilatory pattern during hypercapnic rebreathing. Meaning of single versus multiple asterisks is the same. *p < 0.05 OSA versus baseline and recovery; #p < 0.05 for recovery versus baseline and OSA.

View larger version (11K): [in this window] [in a new window]   Figure 1.   Group mean data (± SE) for hypoxic ventilatory responses at baseline (open circles), the end of the OSA period (closed circles), and during recovery (triangles) for W, SWS, and REM. The data are plotted from the inflection point of the ventilatory response to the point of movement (W) or arousal (SWS, REM). *p < 0.05 OSA versus baseline and recovery for rightward shift of the curve; **p < 0.05 OSA versus baseline and recovery for both decrease in slope and rightward shift of the curve.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Function of OSA Model

The OSA model functioned well throughout the experimental period and was well tolerated in all animals. Apart from occasional missed nights due to minor technical problems, the dogs underwent nightly airway occlusions during sleep for the entire intervention period, which ranged from 12 to 17 wk (mean ± SD: 15.5 ± 1.7 wk). During the last week of the OSA period, which was typical of the latter segment of the intervention period, the mean apnea index was 57.5 ± 4.5 occlusions/h of sleep. The accuracy of the computerized algorithm for detection of sleep-wake state and arousal was similar to that previously reported (7), and remained at a comparable level throughout the intervention period, with only occasional minor adjustment of the sleep-detection program required every several weeks to maintain accuracy.

Resting Ventilation

Each dog underwent an evaluation of ventilation and ventilatory pattern prior to application of the OSA model, and none demonstrated evidence of sleep-disordered breathing or any other respiratory disturbance. The data for resting ventilation during wakefulness for each of the three experimental periods are shown in Table 1. The findings were qualitatively and quantitatively similar during SWS and REM. There were only minor changes in breathing at rest during the OSA phase. During both wakefulness and sleep there was a shift to a slightly slower, deeper respiratory pattern (i.e., a decrease in respiratory rate due to prolongation of both TI and TE, and a small increase in VT). During W, there was a small but significant decrease in I, whereas during SWS and REM the decreased respiratory rate was offset by the increased VT, such that I did not change significantly. During the recovery period there was an increase in respiratory rate from OSA toward baseline values, and a partial return of VT to baseline, although the net result was that ventilation was increased above both OSA and baseline values.

There were no significant changes over the three experimental periods in Sa_O2 during either W or sleep. There was a statistically significant but small decrease in end-tidal PCO₂ during the OSA phase, which appears to have been due to the change in the pattern of ventilation, as I decreased slightly (Table 1). Application of the OSA model was therefore not associated with a progressive deterioration in blood gases during W. There was no evidence of periodic breathing during W or sleep when the animals were not attached to the airway occlusion apparatus during the OSA or recovery phases. Furthermore, there were no consistent differences in the coefficients of variation for ventilatory measures during quiet breathing across the three experimental periods for W, SWS, or REM.

Ventilatory Responses to Hypoxia

The group mean values for the ventilatory responses to hypoxia are shown in Figures 1 and 2. The data shown in Figure 1 are the group mean values for ventilation calculated from regression equations, using the measured mean inflection and arousal point for each condition. The most striking finding was a decline during OSA in the ventilatory response to hypoxia during W. This change was characterized both by a significant decrease in the slope and a shift to the right of the ventilatory response curve. The hypoxic responses during sleep were also decreased, although less prominently so. There was a significant shift to the right of these curves at lower levels of ventilation in both SWS and REM. During recovery, hypoxic responsiveness during W returned to baseline, and hypoxic responses during sleep increased as compared with both their baseline and OSA values, owing to a left-shift of the curve.

The pattern of breathing during hypoxia for the three experimental periods is depicted in Figure 2. It can be seen that the reduced ventilatory response to hypoxia was due exclusively to a reduction in the breathing frequency response to hypoxia. Further analysis demonstrated that this was due to significant increases in TE. These graphs further demonstrate that although the changes in hypoxic ventilation overall were less prominent during sleep than during W, there is a clear effect of the OSA intervention on breathing frequency during hypoxia for SWS and REM, as well as for W.

Ventilatory Response to CO₂

The ventilatory responses to CO₂ over the three experimental periods are shown in Figures 3 and 4. In contrast to the hypoxic findings, the hypercapnic response during OSA was significantly increased during W and SWS, due to a leftward shift of the curve without a change in slope. During recovery, these responses partially reverted to baseline during W and completely reverted for SWS. No significant changes in hypercapnic responses were observed for REM across the three experimental periods. The pattern of breathing for hypercapnic responses is shown in Figure 4. It can be seen that the changes were less consistent overall than those for hypoxia, with the exception of SWS, during which there was a small but consistent increase in ventilation, which was largely due to an augmented frequency response to hypercapnia. There were small and statistically nonsignificant changes in both TI and TE during the OSA period.

View larger version (11K): [in this window] [in a new window]   Figure 3.   Similar plot to that in Figure 1 for hypercapnic responses during the three experimental periods. *p < 0.05 OSA versus baseline for leftward shift of the curve.

Arousal Responses to Hypercapnia and Hypoxia

The values for mean Sa_O2 and PCO₂ at arousal from hypoxic and hypercapnic rebreathing, respectively, during SWS and REM for the three experimental periods are shown in Table 2. There was a small but significant decrease in Sa_O2 at arousal for both SWS and REM during OSA, and this change was reversed during recovery. There were no significant changes in the CO₂ arousal threshold over the three experimental periods.

The values for I and VT at arousal during rebreathing studies for the three study periods are shown in Table 3. For both hypoxic and hypercapnic rebreathing in SWS and REM, there was a significant increase in either arousal I or VT or both for OSA as compared with baseline and recovery. A striking observation was that despite the very discrepant I values at arousal for hypoxia versus hypercapnia, VT values at arousal from SWS were identical for the two stimuli in each of the experimental periods.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the present study, we demonstrated the feasibility of long-term application of our model of OSA in five dogs over a 12- to 17-wk period. All components of the model system functioned well throughout the intervention periods.

We found that during the phase of OSA, there were minor changes in resting ventilation during both W and sleep (Table 1). These consisted of a slowing of respiratory rate, secondary to prolongation of both TI and TE, and a small increase in VT. This resulted in a small decrease in I during W, but no significant change in ventilation during sleep. Following 4 wk of recovery from OSA, there was an increase in respiratory rate to baseline, but VT tended to remain slightly elevated, such that I was slightly higher than at baseline or during OSA. (In two animals studied at 8 wk of recovery, these changes reverted to baseline). There was no significant decrease in Sa_O2 with OSA, and there was a small decrease in PET_CO2. Furthermore, when the animals were permitted to sleep undisturbed by airway occlusion during the daytime sleep studies, there was no evidence of periodic or unstable breathing as a result of the OSA intervention.

The mechanism and significance of the changes in resting ventilation during W and sleep are unclear. The decrease in ventilation during W could conceivably represent a more "sleeplike" pattern of respiration, due perhaps to the effect of OSA-associated sleep fragmentation. However, similar changes were observed during both SWS and REM. The findings therefore suggest that OSA resulted in a specific change in the control of respiratory-pattern generation at rest. The change in breathing pattern may account for the seemingly paradoxical finding of decreases in both I and PET_CO2 during the OSA period (0.6 to 1 mm Hg decrease in group mean values for OSA versus baseline in W and sleep). The increased VT may have led to an increase in alveolar ventilation due to a decrease in dead space-to-VT ratio. The small reduction in PET_CO2 raises the possibility that central chemosensitivity to CO₂ was increased, which then led to a change in respiratory pattern, thus lowering PET_CO2 (13). This would be consistent with the increase in the ventilatory sensitivity to CO₂ that we observed during rebreathing trials in the OSA phase (discussed subsequently).

We did not observe significant hypoxia or hypercapnia during W or undisturbed sleep in any of the animals during the OSA period. This may not be surprising, in that studies with human OSA patients indicate a strong association of hypoventilation during W with underlying abnormalities of pulmonary function (3). The blood-gas abnormalities in such cases are often out of proportion to the degree of lung dysfunction (1, 4, 5), and improve following treatment of OSA. These observations suggest that hypoventilation during wakefulness in OSA results from an interaction between abnormal respiratory mechanics and repeated airway occlusion during sleep. Studies involving the addition of a chronic mechanical load to the airway occlusion valve (14) would permit investigation of this interaction.

The most striking finding in the present study was the selective decline in the ventilatory response to progressive hypoxia during the OSA intervention period. Impaired hypoxic ventilatory responses have been described in OSA patients, although affected patients typically also demonstrate blood-gas abnormalities during W and an impaired ventilatory response to hypercapnia (4). In our animals, blood gases remained normal, and the hypercapnic ventilatory response was either unchanged or increased during the OSA intervention. Thus, there may be different mechanisms involved in the impairment of hypoxic and hypercapnic ventilatory responses in OSA.

The present study does not establish the mechanism of the altered hypoxic responsiveness with the OSA model. One contributing factor may have been the sleep fragmentation resulting from repeated end-apneic arousal (6). Sleep deprivation has been reported to impair hypoxic ventilatory responses in humans (6, 15). However, the OSA model is associated with sleep fragmentation rather than sleep deprivation, and previous work has shown that short-term acoustically-induced sleep fragmentation does not alter the ventilatory responses to hypoxia or hypercapnia (16). Furthermore, a specific depressive effect of sleep fragmentation on hypoxic responsiveness would have to be postulated in view of the findings during hypercapnia. Nonetheless, the effects of long-term sleep fragmentation on hypoxic ventilatory responsiveness have not been studied specifically. Future investigations, using a modification of our OSA model in which isolated sleep fragmentation is induced without airway occlusion (9), should help resolve these issues.

The hypoxic response is affected by the PCO₂ level at which the response is determined (17, 18). Since there were no significant differences in mean PET_CO2 in the three experimental periods during hypoxic trials, changes in PCO₂ do not account for the change in hypoxic response. Metabolic factors could also have contributed, in that a reduction in metabolic rate with OSA could have reduced the rate of induction of hypoxia during the rebreathing trial (which for a constant rebreathing-bag volume and initial FI_O2 is determined by metabolic rate). This in turn could have reduced the ventilatory response (18). Although we did not specifically measure minute O₂ consumption or CO₂ production in this group, a change in metabolic rate would have been expected to similarly delay the hypoxic and hypercapnic responses, which was not the case. Furthermore, calculation of the rate of change in Sa_O2 and CO₂ during rebreathing in two of the animals showed no differences between the three experimental periods, further suggesting that no major change in metabolic rate occurred during the OSA period.

It seems more probable that the diminished hypoxic response represents a specific adaptation to the repeated hypoxia induced during apneas (17, 19). An extensive literature on ventilation during sustained hypoxia (17, 20) indicates that a reduction in ventilatory response occurs during both short- and very long-term hypoxic exposure. This is believed to represent an "adaptive" response to the hypoxic environment (17, 20). Unfortunately, only very limited data exist on ventilatory changes during repeated hypoxia (19). However, the decline in hypoxic ventilatory response during the OSA period could conceivably also represent an adaptive response similar to that observed during more sustained hypoxia.

If this is so, the mechanisms involved may be similar. Previous work indicates that adaptation to sustained hypoxia may result from changes in either carotid chemoreceptor or central hypoxic sensitivity. The latter may be due either to altered central processing of afferent carotid-body stimuli or to a change in direct central nervous system (CNS) sensitivity to hypoxia (17, 20, 21). In that the acute hypoxic ventilatory response is mediated by the peripheral chemoreceptor, an effect on carotid-body function or central processing of carotid-body stimuli would seem more likely in our study. The decreased hypoxic response was due exclusively to a decline in breathing frequency during hypoxia. Changes in carotid-body discharge typically affect both respiratory timing and VT; an isolated change in respiratory timing may therefore be more likely to be due to a brainstem controller effect (17, 21). Further investigation will be required to determine the mechanisms by which these changes are produced.

The increased hypercapnic ventilatory response observed during W and SWS during OSA was somewhat surprising in that hypercapnic responses in human OSA have previously been reported to be either normal or depressed. We are not aware of any reported measurements of hypercapnic responses before and after treatment for OSA patients who are initially normocapnic with a normal pretreatment response. Our findings raise the possibility that the hypercapnic response may be relatively increased in such individuals, which may not previously have been appreciated because of the wide "normal range" for human hypercapnic responses (22).

The augmented hypercapnic response during the OSA phase in our study was characterized by a leftward shift of the ventilation-CO₂ relationship without a change in its slope. The increased ventilation was due to changes in respiratory timing and a tendency toward an increase in VT (Figure 4). The lack of a demonstrable change in hypercapnic responses during REM may have been due to the greater variability of ventilatory pattern during that state (22). The acute hyperoxic hypercapnic ventilatory response is believed to be largely mediated by the central chemoreceptor. The augmented hypercapnic ventilatory response is therefore likely to have resulted from either increased central chemoreceptor sensitivity to CO₂ or increased sensitivity of pontomedullary respiratory control centers to input from central chemoreceptors (13).

A number of factors may have contributed to these alterations. However, we proposed earlier that the decreased hypoxic ventilatory responsiveness during the OSA phase may represent an adaptation to the repeated hypoxemic episodes, and that the mechanisms involved could be similar to those involved in the ventilatory adaptation to sustained hypoxia. In this regard, it is interesting that an increase in the ventilatory response to hypercapnia (at comparable levels of hypoxia) has been reported in both humans and animals during acclimatization to altitude (i.e., adaptation to sustained hypoxia) (23, 24). Furthermore, the increase in hypercapnic responsiveness has most often been reported to be due to a leftward shift of the ventilation-CO₂ relationship without a change in its slope (23, 24), as we observed in our animals during the OSA phase. The changes in hypercapnic response might therefore represent one aspect of an integrated "adaptive" response to the recurrent apnea-associated hypoxemia during application of the OSA model.

Assessment of the arousal thresholds to hypoxia and CO₂ during rebreathing for the three experimental periods showed no significant change in the arousal threshold to CO₂, and a small but significant decrease in Sa_O2 at arousal for both SWS and REM during the OSA period. An increase in the arousal threshold to respiratory stimuli during application of the model could result from altered arousability caused by sleep fragmentation (6, 9, 16, 25), or may represent specific habituation to the stimuli generated repeatedly during apneas (26).

The changes in hypoxic arousal threshold could be linked to the changes in ventilatory response. We have previously provided evidence that mechanical stimuli generated during ventilatory efforts play an important role in mediating arousal from sleep in response to chemoreceptor stimulation (11, 12, 26). A reduced ventilatory response could therefore lead to an increase in the time to attain a level of ventilatory effort sufficient to provoke arousal.

In support of the effort-mediated arousal hypothesis (11, 12, 26), it is interesting to note that VT (an indirect reflection of ventilatory effort) at the point of arousal was virtually identical for hypoxia and hypercapnia during SWS in each of the three experimental periods (Table 3). The observation that VT and/or I at arousal for both hypoxic and hypercapnic rebreathing were significantly increased during the OSA period suggests that application of the OSA model led to an impairment in the arousal response to effort-related stimuli (26). This is also supported by the dramatically increased time to arousal and level of inspiratory effort at arousal in response to airway occlusion that we observed during the OSA phase (9).

In summary, our model of OSA has successfully been applied over a period of 3 to 4 mo in five animals. Application of the model was associated with only minor changes in resting ventilation and with no significant change in resting blood gases. There was a selective decrease in the ventilatory response to progressive hypoxia and a small increase in the ventilatory response to hypercapnia. These changes may represent an integrated adaptive response to the recurrent hypoxemia induced during the OSA period. The arousal threshold to hypoxia but not hypercapnia increased during OSA, as did the I and VT at arousal for both hypoxia and hypercapnia. Many of the changes produced during application of the OSA model appear closely analagous to abnormalities reported in human OSA. This suggests that further investigation of the phenomena produced during application of our model may provide important insights into the pathophysiology of human OSA.