INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

SLEEP APNEA IS TWO TO THREE TIMES more common in men than in women (5, 27, 41). This may be due to gender differences in pharyngeal structure and/or function or ventilatory control stability.

The female pharynx, despite some evidence of greater upper airway muscle activity (25), is anatomically smaller (6, 19, 28) or of similar size (7, 16) to the male airway during both wakefulness and sleep. Because a small airway generally predisposes one to sleep apnea, the smaller female airway is unlikely to be protective. Furthermore, pharyngeal collapsibility, as well as the sleep-related increase in upper airway resistance, are not systematically different between the sexes, as conflicting results have been reported (26, 28-31). Thus, it is difficult to conclude that structural and/or functional differences alone are responsible for the male predominance in obstructive sleep apnea (OSA).

On the other hand, differences in respiratory control system stability may be a contributing factor. First, sleep apnea in women, unlike in men, is almost exclusively related to obesity (5), suggesting that factors such as chemical control instability (CCI) may play a greater role in male sleep apnea. Second, some evidence suggests that males have greater awake hypoxic and hypercapnic ventilatory responses (HVR and HCVR, respectively) (36), although a subsequent study failed to reproduce this (1). Third, men are more susceptible to hypocapnic ventilatory inhibition during sleep (higher apnea threshold) (42). However, interpretation of these traditional control of breathing tests is limited because they do not assess the dynamic aspects of chemical control, nor do they assess the plant or feedback gains, which are major components of stability (Fig. 1). Consequently, it is still unclear whether gender differences in CCI exist.

View larger version (12K): [in this window] [in a new window]   Fig. 1.   Simplified chemical feedback loop. A ventilatory disturbance [E(D)] produces a change in alveolar PCO2 (PACO2) according to the gain of the plant (the lungs and body tissues). The PACO2 signal is delayed (circulatory delay) and attenuated by mixture with arterial blood that fills the space between the alveoli and the chemoreceptors (feedback gain). The change in PCO2 at the chemoreceptors (PcrCO2 then causes a change in ventilation [E(R)], which opposes the original ventilatory disturbance. Proportional-assist ventilation (PAV) increases controller gain. Loop gain is the ratio of the ventilatory response to the disturbance, which is in turn the product of the controller, plant, and feedback gains [LG = E(R)/E(D) = controller gain × plant gain × feedback gain].

Recently, a technique for measuring CCI during sleep has been developed that encompasses all components of the chemical feedback loop, including the dynamic response characteristics (40). This technique measures loop gain (LG) at the frequency of periodic breathing and takes advantage of the ability of proportional-assist ventilation (PAV) to enhance respiratory controller gain (Fig. 1). LG is a term well known to control engineers and is a measurement of the propensity for a system governed by feedback loops to develop unstable behavior (for a review of LG, see Ref. 13). It has both a magnitude and a dynamic component. The magnitude of LG is determined by the ratio of the corrective response to a disturbance (which is also the product of the controller, plant, and feedback gains). The dynamic component consists of the circulatory delay and the time constants of the three loop components. Instability occurs when the corrective response, 180° out of phase with the disturbance, is equal to or greater than the disturbance. Consequently, by incorporating the gain of all loop components as well as the phase angle, LG is a more complete metric of stability than traditional steady-state (or progressive) gain tests (e.g., HVR, HCVR) and apneic threshold measurements.

We hypothesized that men have intrinsically higher LG than women and that this may partially explain the gender discrepancy in sleep apnea. We tested this hypothesis by incrementally increasing LG magnitude with PAV in both genders until periodic breathing or awakening occurred. This ability to manipulate LG tells us how close a subject is to spontaneous oscillation; individuals requiring less ventilatory augmentation with PAV to induce periodic breathing have greater CCI and vice versa.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects. Normal, healthy volunteers were recruited from the community for this study. Women were required to have normal menses and not be pregnant or taking oral contraceptive pills. Subjects were excluded if they had a chronic medical condition (except for treated hypertension) or complained of snoring or daytime sleepiness. Twenty-six individuals gave informed written consent to participate in the study. However, 7 developed claustrophobia or were unable to sleep on PAV, leaving 19 participants (10 men, 9 women) who completed the study. Women were studied in both the follicular (days 5-11, with day 1 being the final day of menses) and luteal (days 17-22) phases of the menstrual cycle. The study was approved by the human subjects committee at Brigham and Women's Hospital.

Equipment. Electroencephalography (EEG; C₃/A₂, O₂/A₁), electrooculography, and submental electromyography were assessed for sleep staging. Oxygen saturation was monitored by using a pulse oximeter attached to the index finger (BCI Capnograph Series 9000, Waukesha, WI), and PCO₂ was sampled at the mask by using a calibrated infrared CO₂ analyzer (BCI). Airflow was measured using a pneumotachometer (Hans-Rudolph, Kansas City, MO) placed between the mask and the intentional leak port, and the airflow signal was integrated for tidal volume (VT). Airway pressure was measured at the mask using a pressure transducer (Validyne, Northridge, CA). All data were displayed on a Grass model 78E Polygraph (West Warwick, RI) as well as a computer (Spike 2, Cambridge Electronic Design, Cambridge, UK). LG measurements were made by using a BiPAP Vision (Respironics, Murrysville, PA) mechanical ventilator capable of delivering CPAP alone or in varying combinations with PAV.

Protocol. Subjects reported to the sleep laboratory at about 9:30 PM. Sleep staging equipment was attached to each subject, and PAV was connected by using a nasal mask held in place with straps. Before application of the mask, each nostril was decongested with one application of oxymetazoline HCl. Once all of the equipment was attached and the mouth taped shut, subjects were allowed to fall asleep in either the supine or lateral position. Manipulation of the ventilator was begun once several minutes of sustained non-rapid eye movement (NREM) sleep had occurred.

1

Data analysis. Sleep staging and arousals were scored using standard criteria (26). Arousals were defined as an abrupt increase in EEG frequency to 8 Hz or greater for at least 3 s. Data were analyzed from NREM sleep only, with the majority coming from stage 3/4 sleep because subjects tended to arouse easily at high levels of assist in stage 2 sleep. Because VT tended to vary, particularly at high levels of assist, the three VT values before the unassisted breath were averaged and used for the numerator of VTAF. The VTAFs for each assist level were then averaged. The average VTAF at the level immediately preceding the development of periodic breathing or runaway was used for LG calculation.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Ten men (age 24-48 yr) and nine women (age 22-36 yr) completed the study. The average body mass index was 23.2 for the men and 22.6 for the women. With the exception of one woman, data were collected in both the follicular and luteal phases of all women.

A representative example of periodic breathing in female subject 7 is shown in Fig. 2. When the assist was increased to 75%, sustained periodic breathing at a frequency of 0.02 Hz appeared. Cyclic breathing usually began within 30-60 s of the PAV increase. VT (0.60 liter) and PCO₂ (38 Torr), just before cycling, were not substantially different from baseline even at 70% assist, signifying both a compensatory reduction in respiratory muscle pressure and an intact chemical control system.

View larger version (36K): [in this window] [in a new window]   Fig. 2.   Representative tracing of a subject who developed periodic breathing on PAV. Approximately 50 s after the percent assist was increased from 70 to 75% (arrow), the subject developed sustained periodic breathing at a frequency of 0.02 Hz. PETCO2, end-tidal PCO2; VT, tidal volume.

Figure 3 shows data from female subject 6, who was resistant to periodic breathing. An increase in the percent-assist level led to a failure to transition into expiration at end-neural inspiration (i.e., runaway breathing). In this instance, runaway occurred before LG could be brought to 1 with PAV. VTAF at the preceding percent-assist level was 1.82, meaning the subject's underlying LG had to be <0.55.

View larger version (13K): [in this window] [in a new window]   Fig. 3.   An example of runaway breathing on PAV. Runaway is identified by failure of inspiratory flow to terminate (solid arrow) at the usual end-inspiratory time established by previous breaths. This occurs when the pressure given by the ventilator is greater than the resistive and elastic recoil pressures of the respiratory system. Note that the percent assist was decreased to 0 for 1 breath (single-breath reloading test) (open arrow) to measure the VT amplification factor. Paw, airway pressure.

Men. Of the 10 men, 1 developed periodic breathing at a VTAF of 1.50 (LG = 0.67) (Table 1). All others experienced runaway breathing that precluded further increases in controller gain and consequently LG quantification. We were able to demonstrate, however, that LG was at most <0.46 in 7 of 10 men, 4 of whom had LGs of <0.37. The average of the highest VTAF achieved in each man was 2.49, which was not significantly different from the follicular-phase female value of 2.06 (P = 0.22). LG between men and follicular-phase women was not statistically different either (P = 0.38).

Women. For women in the follicular phase, two of nine experienced periodic breathing at a VTAF of 1.28 (LG = 0.78) and 2.26 (LG = 0.44) (Table 1). LG was <0.52 in five follicular-phase women, three of whom were <0.45. When studied in the luteal phase, the same two women who developed periodic breathing in the follicular phase did so again in the luteal phase at VTAF levels of 1.79 (LG = 0.56) and 2.06 (LG = 0.48). Five of eight luteal-phase women had LGs of <0.55, two of whom were <0.44. The average of the highest VTAF achieved in each follicular-phase women was 2.06, which was not significantly different compared with the luteal-phase value of 1.98 (P = 0.68).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The purpose of this study was to determine whether gender differences in CCI exist. We utilized a recently described technique for quantifying the LG of the respiratory control system with PAV. Our data indicate that LG is low in normal men and women, because most subjects from both gender groups were resistant to periodic breathing despite levels of PAV high enough to induce runaway. These results, however, do not definitively answer the question of whether gender differences in LG exist in these lower ranges. To do so would require higher volume amplification factors than we were able to achieve in this study. Potential reasons for this are discussed below. Nonetheless, previous data, in which PAV induced periodic breathing in only 5 of 24 normal people (17, 18), as well as our results suggest that LG is low in the majority of normal healthy individuals, regardless of gender. Our data also suggest that LG is not considerably increased by circulating progesterone, because LG during the luteal phase was similar to or less than LG during the follicular phase (in the 2 women in whom LG could be quantified). Although our results do not exclude LG as a component contributing to the male predominance in OSA, they suggest that other factors, like pharyngeal structure and/or tissue characteristics or state-dependent muscle function, may be more important.

There are two possible explanations for why we were unable to find a gender difference in LG. First, the combination of physiological variables known to affect LG magnitude (15, 37, 40) and phase angle (e.g., circulation delay) must yield a sufficiently stable respiratory control system in both sexes such that "large" amplifications of controller gain (to the point that runaway breathing occurs) do not produce instability. Thus there may be no actual difference in CCI between the genders. On the other hand, we may have failed to demonstrate a difference because we could not measure LG in the lower ranges. When LG was less than ~0.50, it was difficult to quantify chemical control stability with PAV. Therefore, differences in the lower ranges could still exist.

Role of LG in OSA. There appears to be an anatomic continuum for the risk of CCI-induced upper airway obstruction. It is known that individuals with moderately compromised upper airway anatomy and/or physiology (e.g., snorers) develop repetitive pharyngeal obstruction to a greater degree than nonsnorers at the nadir of hypoxia-induced periodic breathing (10, 20, 32). Indeed, Cheyne-Stokes respiration is often associated with upper airway occlusion (2). This occurs because the pharyngeal dilator muscles, like the diaphragm, receive input from medullary respiratory neurons (21, 32). During periods of increased respiratory drive, when output from these neurons is high, pharyngeal dilator muscles are activated (4, 22) and help maintain upper airway patency. During low drive, however, these muscles relax (3), making the pharynx more collapsible and prone to occlusion, particularly in the setting of deficient upper airway anatomy (20). Thus the oscillating respiratory drive during periodic breathing provides an ideal milieu for recurrent collapse of the upper airway. Consequently, depending on the anatomic substrate, a LG in the moderate range (e.g., 0.4-0.5) may produce enough oscillation in respiratory drive to "tip" a snorer over into the OSA category or worsen existing OSA (40).

Gender differences in chemoresponsiveness. Components of the ventilatory control system (HVR and HCVR) have been compared between men and women both awake and asleep. During wakefulness, White et al. (34, 35) found HVR to be greater in men, although a subsequent study showed the opposite (awake HVR greater in women) (1). Nevertheless, HVR appears to be similar in men and women during sleep (34). When HCVR is measured during wakefulness, men appear to have greater chemoresponsiveness than women (1, 9, 34). During sleep, however, HCVR is either similar (34) or substantially greater in men (42), depending on the study.

Gender differences in apnea threshold. There is recent evidence of a gender difference in apnea threshold, with men being more susceptible to hypocapnic-induced ventilatory inhibition than women (42). This could have important effects on LG for the following reason. A high apnea threshold (small difference between eupneic and apneic PCO₂) would tend to amplify the ventilatory undershoot that occurs in response to a preceding hyperpnea, which would increase LG (by increasing the response-to-disturbance ratio). Importantly, the PAV technique incorporates this apnea threshold into its LG measurement. Thus, if men have a higher apnea threshold than women (and all other factors that influence LG are equal), one would predict that cycling would be easier to induce (and thus LG would be higher) in men. Because the men in our study were as resistant to periodic breathing as the women, we would conclude that gender differences in apnea threshold are not substantial enough to translate into gender differences in LG, at least within the range that we could measure it.

LG in normal subjects. Our findings agree with previous PAV studies in normal people. In the first study to introduce PAV as a tool for measuring ventilatory stability, Meza and Younes (18) attempted to induce periodic breathing in 12 normal subjects (7 nonsnorers, 5 snorers). None of the 12 developed sustained periodic breathing on PAV either spontaneously or after a ventilatory perturbation (percent assist was turned to 0 for 3-4 breaths). Baseline LG in these subjects was reported to be <0.3. According to Meza et. al (17), however, this study used an inaccurate amplification factor to calculate LG. In a follow-up study that used a different amplification factor (but not VTAF), 5 of 12 normal sleep-deprived subjects (2 of whom received 0.5 mg lorazepam) developed periodic breathing on PAV, and LG was reported to be on average <0.28 (17). Because VTAF was not used to calculate LG in either of these two studies, and because some subjects were sleep deprived or taking benzodiazepines, direct comparisons with our data are difficult.

Technical considerations. There are two potential reasons why we did not achieve as high a VTAF as desired in each subject. First, we did not set pressure or VT limits on the ventilator, because we reasoned that this may have blunted ventilatory responsiveness. However, this made it difficult to distinguish true runaway (runaway occurs with almost every breath) from spontaneously large VT values, which occur singly. Because it was necessary to turn down the assist after a large breath (as a result of subject awakening), we could not confirm that true runaway had occurred. However, we were careful not to call a large VT breath runaway unless it had a typical "saddle-shaped" flow pattern (Fig. 3) (39). Furthermore, we achieved runaway on several occasions before terminating the study. Nonetheless, it is likely that we could have maintained subjects on a higher level of assist for a longer period of time, and possibly have achieved higher volume amplification, if pressure and volume limits were set at 25 cmH₂O and 1.5 liters, respectively (M. Younes, personal communication).

Limitations. First, we measured LG in a group of relatively young men and women without comorbidities, a population that has a low incidence of OSA.

Conclusions. We found that, insofar as LG was low in the majority of our subjects, CCI was not substantially different between the genders. Therefore, CCI may not play as large a role in male OSA as we originally hypothesized. Higher VTAFs than were achieved in this study, however, would be necessary to determine whether gender differences exist in this lower range.