INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Breath components display considerable breath-to-breath variability in healthy human subjects (1). Mathematical tools, such as time series analysis, fast Fourier transformation (FFT), and the multiregressive model developed by Modarrasezadeh and colleagues (3), provide powerful methods for investigating the nature of this variability. With these tools, total variational activity of breathing can be partitioned into correlated, oscillatory, and random fractions. Moreover, these tools can detect ventilatory oscillations not obvious to the naked eye. The nonrandom (i.e., correlated and oscillatory) and random fractions of the total variability appear to carry different physiological implications. In healthy volunteers, we have shown that stimulation of the central chemoreceptors, with hyperoxic hypercapnia, and the mechanoreceptors, with elastic and resistive loading, cause specific and unique changes in different fractions of variational activity of breathing (5).

During non-rapid eye movement (NREM) sleep, Berssenbrugge and coworkers (8) demonstrated that hypoxia induced periodic breathing and also increased the gross variability of breath components, quantified in terms of their coefficients of variation. The hypoxia and sleep were accompanied by hypocapnia, and the alterations in breath variability did not occur when hypocapnia was prevented (8). Even when combined with hypocapnia, hypoxia produced periodic breathing in only one of the subjects during wakefulness. Moreover, Berssenbrugge and coworkers (8) did not study the effect of hypoxia during isocapnia while the subjects were awake. That ventilatory oscillations occur when hypoxia is combined with either hypocapnia or sleep is consistent with the current understanding of the part played by chemical feedback in the respiratory control system. Hypoxia stimulates the fast-acting peripheral chemoreceptors, whereas, when combined with hypocapnia, it decreases the contribution of the slowly responding central chemoreceptors (9, 10). Potentiation of the peripheral chemoreceptor contribution and a reduction in that of the central chemo-receptors will result in faster dynamicsparticularly during sleep when PCO₂ is just above the apneic threshold (11). The faster dynamics, in turn, predispose to ventilatory oscillations (12); it is not known, however, whether oscillations occur in the absence of hypocapnia and sleep.

To investigate the effect of isocapnic hypoxia on the variational activity of breathing, we recorded ventilation in awake, healthy volunteers before and during a steady-state decrease in oxygen saturation while carbon dioxide tension (PCO₂) was kept constant. The subjects inhaled a hypoxic mixture through an open-ended plastic hood and ventilation was measured nonobtrusively with an inductive plethysmograph to avoid the confounding effects resulting from instrumentation attached to the face (13).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Fourteen, nonsmoking volunteers (10 men and 4 women) with a mean age of 34 (range, 22-45) yr participated in this study. All had normal pulmonary function. Informed consent was obtained from all subjects, and the study was approved by the Human Studies Subcommittee of Edward Hines Jr., Veterans Affairs Hospital.

Respiratory Inductive Plethysmograph

Ventilation was measured nonobtrusively with a respiratory inductive plethysmograph (Non-Invasive Monitoring Systems, Miami Beach, FL). The inductive plethysmograph was calibrated against spirometry by the least-squares method (1, 13). Validation was checked during tidal breathing in the horizontal and semirecumbent positions. Inductive plethysmographic data were considered acceptable if they displayed a  10% difference from spirometry at both the beginning and end of the study. Validation of the inductive plethysmograph against spirometry in the horizontal and semirecumbent positions revealed mean arithmetic differences of 4.0 ± 3.2% (SD) and 3.1 ± 2.8%, respectively, before the experiments and 6.1 ± 2.6 and 6.3 ± 3.9%, respectively, after the experiments. The signals from the inductive ple-thysmograph were recorded on a microprocessor system (Non-Invasive Monitoring Systems, Miami Beach, FL), which sampled the data at 20 Hz. On a breath-by-breath basis, the microprocessor continuously calculated the values of minute ventilation (I), tidal volume (VT), inspiratory time (TI), and expiratory time (TE). The data were processed by taking the duration of the respiratory cycle (Ttot) as the unit of time.

Oxygen Saturation and End-tidal CO₂ Measurements

Oxygen saturation (Sa_O2) was measured with a pulse oximeter (Criticare 3740, Boulder, CO) and end-tidal PCO₂ (PET_CO2) was measured with an infrared analyzer (Ohmeda 4700, Louisville, CO) using cannulae inserted in both nostrils. Continuous breath-by-breath recordings of Sa_O2 and PET_CO2 were stored on computer disk.

Inspired Gas Administration

To avoid the significant changes in breathing pattern induced by instrumentation attached to the face (13), an open-ended hood was employed for the delivery of gases. The hood consisted of transparent material that was not in direct contact with subject's head and did not produce uncomfortable feelings such as claustrophobia. The hood was loosely positioned over the subject's head and it was ventilated with the desired gas composition. Cylinders of compressed nitrogen and CO₂ were connected to a flowmeter (Matheson Co., Montgomeryville, PA), and the outlet of the latter was connected to the hood. To eliminate rebreathing and entrainment of room air, the bias flow rate from the flowmeter exceeded 1.5 L/s.

Protocol

The study was conducted while the subject lay semirecumbent on a bed in a quiet room. Following successful validation of the inductive plethysmograph, the plastic hood was positioned over the subject's head while he or she rested quietly for 15 min before the initiation of data recording. Subjects were instructed to remain motionless, keeping their arms by the side, and to remain awake for the duration of the study while watching a documentary recording on a video tape player. Subjects first breathed room air through the hood for 60 min. Then, nitrogen was delivered into the hood and the flow was adjusted to achieve a steady-state Sa_O2 of 80%. In reality, Sa_O2 decreased from 97.8 ± 1.0 to 79.1 ± 1.0 (SD)%; this level of Sa_O2 was maintained for at least 60 min. During the introduction of hypoxia, supplemental CO₂ was added to the hood to achieve a PET_CO2 equivalent to that during air breathing: the values for PET_CO2 during air and isocapnic hypoxia were 34.4 ± 3.3 and 34.1 ± 3.0 mm Hg, respectively. Breathing pattern, Sa_O2 and PET_CO2 were continuously recorded on a computer and stored on disk for later analysis (Figure 1).

View larger version (30K): [in this window] [in a new window]   Figure 1.   Time-series plot of continuous measurements of oxygen saturation (SO2), and breath-by-breath measurements of end-tidal carbon dioxide tension (PETCO2) and minute ventilation ( I) in a subject breathing air and an isocapnic hypoxic mixture. The displayed values of PETCO2 are the measurements at the end of the plateau of each breath, and values of SO2 represent a running 6-s average. The arrows indicate the onset and termination of the hypoxic challenge. The segments of data selected for analysis were from 17 to 76 min during air and from 95 to 158 min during isocapnic hypoxia. Hypoxia produced a decrease in SO2 from 98 to 80% and a variable increase in I.

Data Analysis

Paper tracings and computerized calculations of breath components during air and isocapnic hypoxia were inspected, and artifacts (due to cough, swallowing, movement, etc.) identified and deleted; this constituted < 1% of the data. For each experimental condition, 700 consecutive breaths in which the values of I, VT, TI, and TE did not display large deviations from the mean on visual inspection were chosen for analysis. Because of the well-described biphasic response to sustained hypoxia (i.e., roll-off phenomenon), data collected during the first 20- 25 min of isocapnic hypoxic exposure were excluded in the assessment of the effect of hypoxia on variational activity of breathing; instead, analysis was confined to the subsequent 700 breaths. This approach was undertaken to ensure that breath components displayed a steady statea necessary condition for time-series analysis.

Breath components in a breath series display breath-to-breath variability, which may consist of correlated, oscillatory, and random components. Autocorrelation and spectral analysis enable the determination of the relative magnitudes of these components. Mean values of breath components as well as the hypoxic ventilatory response were also calculated.

Mean changes of breath components. The mean value of the 700 breaths was calculated for each breath component during air and isocapnic hypoxia.

Gross variability. The standard deviation (SD), viz. the square root of the variance, for each breath component was calculated in each subject during air and isocapnic hypoxia as a measure of gross breath-to-breath variability. The coefficient of variation (CV, that is SD divided by the mean) of each breath component was taken as a measure of the relative variability. To ensure that the SDs and CVs approximated a Gaussian distribution, they were logarithmically transformed. Then, the log-transformed data for each breath component during air were compared with those during isocapnic hypoxia using paired t tests.

Autocorrelation analysis. Autocorrelation analysis was employed to determine what fraction of variational activity is correlated on a breath-to-breath basis as previously described (5). By its ability to extract correlated activity from data obscured by random noise, autocorrelation analysis can determine if there is a strong relationship between one breath and another at some interval (or lag) away. It also determines the relative strength of "short-term memory" for each breath component (14). "Short-term memory" refers to the number of consecutive lags, starting at a lag of one breath, that displays autocorrelation coefficients that are statistically different from zero at p < 0.01 levels (2). Although memory is the term used to describe this statistical relationship, it does not necessarily signify that the consecutive serial autocorrelation coefficients have a neural origin (2, 15).

Spectral analysis. Power spectral analysis can also be used to quantitate breath-to-breath variability in a breath component. The power spectrum expresses the variance of a signal as a function of frequency. The presence of a significant peak (5, 7) in the spectrum indicates that some of the variance in the data is due to a periodic oscillation with a period equal to the inverse of the frequency of the peak. The area inscribed by the peak (amount of power) reflects the degree of variability in the signal resulting from fluctuations at that frequency.

Although spectral analysis and autocorrelation analysis are mathematically related, periodic oscillations as detected by spectral analysis can represent physiological mechanisms other than autoregressive behavior (3); in particular, spectral analysis can reveal low-frequency (slow) oscillations of breath components that might be missed by autocorrelation analysis; it has also been shown (3) that both types of analysis should be done in order to correctly partition the total variability of breathing without corrupting the autocorrelation coefficients.

Fractionation of variational activity of breathing. The variational activity of breathing was partitioned into autoregressive, periodic, and random fractions employing the approach of Modarreszadeh and coworkers (3). This model enables the quantification of each fraction and its contribution to the entire variational activity of breathing. Total variational activity is modeled as a compound consisting of both random and nonrandom fractions. The correlated and oscillatory fractions are quantified using autocorrelation and spectral analysis, and the random (white-noise) fraction is derived as the remainder of the total variance (3, 5, 7). Fractionation of variational activity allows the influence of isocapnic hypoxia on the composition of the total variational activity to be investigated.

Hypoxic ventilatory response. The mean values of the last 10 min of air breathing and the first 5 min and last 5 min of isocapnic hypoxia were calculated for each breath component. A crude measure of the ventilatory response to hypoxia was obtained by calculating the difference in the mean value of I between the last 10 min of air breathing and the first 5 min of hypoxia.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Mean Change in Breathing Component

For the 700 breaths of air breathing, the mean value VT was 0.52 ± 0.16 (SD) L, and it increased to 0.62 ± 0.31 L in the 700 breaths during steady-state isocapnic hypoxia (p < 0.05); the respective values for I were 7.81 ± 2.14 and 8.06 ± 2.58 L/min (p = 0.48); the respective values for frequency were 15.8 ± 1.8 and 14.6 ± 2.6 breaths/min (p = 0.08); the respective values for TI were 1.6 ± 0.2 and 1.8 ± 0.6 s (p = 0.11); and the respective values for TE were 2.4 ± 0.3 and 3.1 ± 1.1 s (p < 0.01).

Gross Variability of Breath Components

Frequency histograms of I, VT, TI, and TE during air and isocapnic hypoxia in a subject are shown in Figure 2. Standard deviations of the breath components in each subject for the 700 breaths during air breathing and the 700 breaths during isocapnic hypoxia are shown in Figure 3. The standard deviation of I for the group increased from 1.78 ± 0.69 L/min during air breathing to 3.22 ± 2.15 L/min during isocapnic hyp-oxia (p < 0.002); the respective values for VT were 0.169 ± 0.090 and 0.309 ± 0.218 L (p < 0.0044); the respective values for TI were 0.37 ± 0.19 and 0.96 ± 1.23 s (p < 0.003); and the respective values for TE were 0.65 ± 0.33 and 1.59 ± 1.36 s (p < 0.0002).

View larger version (22K): [in this window] [in a new window]   Figure 2.   Frequency histograms of minute ventilation ( I), tidal volume (VT), inspiratory time (TI), and expiratory time (TE) during air and hypoxia in a subject. Hypoxia broadened all of the histograms. The standard deviation for I increased from 3.3 L/min during air breathing to 5.4 L/min during hypoxia; respective values for VT were 0.384 and 0.497 L, for TI 0.50 and 1.28 s, and for TE 0.64 and 1.69 s.

View larger version (30K): [in this window] [in a new window]   Figure 3.   Standard deviations (SDs) of minute ventilation ( I), tidal volume (VT), inspiratory time (TI), and expiratory time (TE) during air and isocapnic hypoxia in 14 subjects (mean of 700 breaths for each component in each subject). Hypoxia produced an increase in SDs of I (p < 0.002), VT (p < 0.004), TI (p < 0.003), and TE (p < 0.0002). Circles and bars represent group mean ± SD.

The mean CVs for I for the group increased from 0.23 ± 0.07 in the 700 breaths during air breathing to 0.39 ± 0.21 in the 700 breaths during steady-state isocapnic hypoxia (p < 0.0005); the respective values for VT were 0.32 ± 0.12 and 0.48 ± 0.24 (p < 0.0045); the respective values for TI were 0.23 ± 0.09 and 0.44 ± 0.36 (p < 0.0006); and the respective values for TE were 0.27 ± 0.10 and 0.45 ± 0.25 (p < 0.0001).

Autocorrelation Analysis

Autocorrelograms for TE during air and isocapnic hypoxia in a representative subject are shown in Figure 4. Isocapnic hypoxia decreased the autocorrelation coefficient at a lag of one breath and the number of breath lags with significant (p < 0.01) serial correlations. The autocorrelation coefficients at a lag of one breath for each breath component during air and isocapnic hypoxia in the 14 subjects are shown in Figure 5. The autocorrelation coefficient for I tended to decrease from 0.284 ± 0.167 during air breathing to 0.150 ± 0.279 during isocapnic hyp-oxia (p = 0.07); the respective values for VT were 0.260 ± 0.166 and 0.194 ± 0.193 (p = 0.19); the respective values for TI were 0.192 ± 0.137 and 0.139 ± 0.174 (p = 0.37); and the respective values for TE were 0.307 ± 0.158 and 0.102 ± 0.259 (p < 0.008).

View larger version (17K): [in this window] [in a new window]   Figure 4.   Autocorrelograms of expiratory time (TE) in a subject during air and isocapnic hypoxia. The autocorrelation coefficient at a lag of one breath decreased from 0.18 during air breathing to 0.01 during isocapnic hypoxia. Two breath lags exhibited significant serial correlations (p < 0.01) during air breathing, and these disappeared during isocapnic hypoxia. On the autocorrelograms, points lying outside the inner pair of isopleths are statistically different than zero, at p < 0.05, and outside the outer pair of isopleths, at p < 0.01.

View larger version (17K): [in this window] [in a new window]   Figure 5.   Autocorrelation coefficients at a lag of one breath for minute ventilation ( I), tidal volume (VT), inspiratory time (TI), and expiratory time (TE) during air and isocapnic hypoxia in 14 subjects. The autocorrelation coefficient for TE decreased (p < 0.008) during hypoxia, a tendency toward the same was noted for I (p = 0.07), while those for VT and TI did not change. Circles and bars represent group mean ± SD.

The number of consecutive breath lags displaying significant autocorrelation coefficients for I were 5.7 ± 11.3 during air breathing and 3.6 ± 6.2 during isocapnic hypoxia (p = 0.58); the respective values for VT were 4.7 ± 10.4 and 3.1 ± 4.5 (p = 0.62); the respective values for TI were 2.8 ± 3.5 and 2.5 ± 4.2 (p = 0.86); and the respective values for TE were 4.4 ± 4.5 and 2.1 ± 3.2 (p < 0.03).

Spectral Analysis

Power spectra of I during air and isocapnic hypoxia in a representative subject are shown in Figure 6. Isocapnic hypoxia caused a shift in the oscillations for I to higher frequencies. The centroid frequency, i.e., the mathematically weighted median frequency of the entire spectrum (0.0-0.5 cycle/breath), for I increased from 0.14 ± 0.06 cycle/breath during air breathing to 0.19 ± 0.11 cycle/breath during isocapnic hypoxia (p < 0.05); the respective values for VT were 0.15 ± 0.06 and 0.18 + 0.08 cycle/breath (p = 0.15); the respective values for TI were 0.17 ± 0.05 and 0.18 ± 0.08 cycle/breath (p = 0.84); and the respective values for TE were 0.13 ± 0.05 and 0.19 ± 0.09 cycle/breath (p < 0.05).

View larger version (20K): [in this window] [in a new window]   Figure 6.   Power spectra of minute ventilation during air (dashed tracing) and isocapnic hypoxia (solid tracing) in a subject. A significant peak was observed at 0.03 cycle/breath during air breathing, whereas one occurred at 0.26 cycle/breath during isocapnic hypoxia. The centroid frequency I for increased from 0.18 cycle/breath during air breathing to 0.22 cycle/breath during isocapnic hypoxia.

The number of subjects displaying at least one significant peak in either the low-frequency (0.0-0.2 cycle/breath) or high-frequency (0.2-0.5 cycle/breath) band of the spectra for VT increased from four during air breathing to nine subjects during isocapnic hypoxia (p < 0.05); the respective values for I were eight and nine subjects (p = 0.70); the respective values for TI were five and five subjects (p = 0.99); and the respective values for TE were 7 and 10 subjects (p = 0.25). The frequency and power of the significant oscillations for each breath component in the combined low- and high-frequency bands are listed in Table 1. If a subject had more than one significant oscillation, the oscillation with the highest power was selected for statistical analysis. Isocapnic hypoxia caused a significant shift in the oscillations for I (p = 0.03) to higher frequencies, while it had no effect on those of VT, TI, and TE. Isocapnic hypoxia increased the power of the oscillations for I (p < 0.02) and tended to increase those for TI (p = 0.08) and TE (p = 0.08), while it had no effect on the power of the oscillations for VT.

Fractionation of Variational Activity of Breathing

The fractions of variational activity of breathing secondary to oscillatory behavior, correlated behavior, and random [w(n)] behavior for each breath component during air and isocapnic hypoxia are listed in Table 2. During air breathing, random [w(n)] behavior was the predominant fraction of the variational activity of each breath component (p < 0.01), and the correlated fraction exceeded the oscillatory fraction for each breath component (p < 0.01 in each instance). During isocapnic hypoxia, random [w(n)] behavior was also the predominant fraction of variational activity in each breath component (p < 0.01), and again the correlated fraction was greater than the oscillatory fraction for each breath component (p < 0.01 in each instance).

Isocapnic hypoxia increased the total variational activity of I, VT, and TE (p < 0.02 in all instances) through increases in both their absolute random fractions (p < 0.001 for I and TE; p < 0.05 for VT) and their absolute oscillatory fractions (p < 0.03 for I and TE; p < 0.01 for VT); the increase in total variational activity of TI (p < 0.002) was achieved solely through an increase in its random fraction (p < 0.002). The absolute correlated behavior of each breath components was not altered by isocapnic hypoxia (Figure 7).

View larger version (46K): [in this window] [in a new window]   Figure 7.   Partitioning of total variance of minute ventilation ( I), tidal volume (VT), expiratory time ( TE), and inspiratory time ( TI) into fractions of oscillatory (white columns), correlated (light gray columns), and random (dark gray columns) during air and isocapnic hypoxia in 14 subjects. Isocapnic hypoxia increased the random fractions for I, VT, TE, and TI (all p < 0.05) and oscillatory fractions for I, VT, and TE (all p < 0.03).

Hypoxic Ventilatory Response

I increased from 7.3 ± 2.3 L/min during the last 10 min of air breathing to 12.3 ± 5.4 L/min during the first 5 min of hypoxia (p < 0.002), and then decreased to 8.6 ± 3.0 L/min (p < 0.01) during the last 5 minthe latter I was higher than that during air breathing (p < 0.008). The decline in I between the first 5 min and the last 5 min of hypoxia was strongly correlated with a crude measure of ventilatory sensitivity to hypoxia, viz. the increase in I between the last 10 min of air breathing and the first 5 min of hypoxia (r = 0.95, p < 0.001). VT increased from 0.495 ± 0.193 L during the last 10 min of air breathing to 0.848 ± 0.421 L during the first 5 min of hypoxia (p < 0.01), and then decreased to 0.607 L during the last 5 min of the sustained hypoxia (p < 0.02); during the last 5 min of hypoxia, VT was higher than that during air breathing (p < 0.01). Frequency was 15.7 ± 2.4 breaths/min during the last 10 min of air breathing, 15.4 ± 3.1 breaths/min during the first 5 min of hypoxia, and 15.3 ± 2.3 breaths/min during the last 5 min of the sustained hypoxia (p = 0.89).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The gross variability of each breath component was increased by isocapnic hypoxia. Isocapnic hypoxia decreased the autocorrelation coefficient at a lag of one breath for TE, it tended to have the same effect on I, and it decreased the number of consecutive breath lags having significant autocorrelation coefficients for TE. The number of subjects displaying oscillations in VT increased with isocapnic hypoxia, and the cycle time of oscillations in I decreased by about ~ 4-fold. Partitioning of the total variational activity of breathing revealed that alterations resulted from changes in both the random fractions and oscillatory fractions. This is the first report of the effect of isocapnic hypoxia on random and nonrandom fractions of variational activity of breathing in awake healthy subjects.

Critique of Methods

An important requirement in this study was obtaining accurate measurements of breathing pattern over a prolonged period of time. By using inductive plethysmography, it is possible to measure ventilation nonobtrusively and avoid the spurious alterations resulting with the use of a mouthpiece (13). The calibration of the inductive plethysmograph was verified in two postures at both the beginning and end of the study; the values obtained always displayed a less than 10% difference from simultaneous spirometry. To distract the subjects from focusing on the experiment, they watched a video. To avoid dramatic changes in video content throughout the experiment, a recording of a documentary program was selected.

Subjects were exposed to air and hypoxia in a sequential, rather than a randomized, manner. This approach was deliberately chosen because of the well-known residual effects of hypoxia on ventilation. Exposing the subjects to hypoxia before the air breathing recordings might have confounded the latter measurements. Passage of time over the course of an experiment might influence breath-by-breath variability. In a previous study in healthy subjects, however, we found that the autocorrelation coefficient at a lag of one breath for VT, TI, and TE were similar for the first 350 breaths and last 350 breaths of air breathing (2).

Effect of Hypoxia on Mean Change and Gross Variability of Breathing

Isocapnic hypoxia produced an increase in the mean value of VT, but, because of an accompanying decrease in frequency, I did not change. The absence of change in I during sustained hypoxia may seem surprising, but relates to the fact that the initial increase in I during the first ~ 20-25 min of exposure to isocapnic hypoxia was excluded; data analysis was confined to the subsequent 700 breaths, thus containing the segment that includes the reduction in I observed with sustained hypoxia. It is well recognized that I displays a biphasic response to sustained hypoxia: an increase within the first 3-5 min of exposure followed by a decrease to a new steady-state level that usually slightly exceeds the normoxic baselinethe so-called "roll off" phenomenon (16, 17). We likewise observed a biphasic response to sustained hypoxia; the value of I at the point of the secondary decline was slightly, but significantly, greater than the value during room air breathing, as has been reported by other investigators (16). Like us, previous investigators have noted that the biphasic response in I during hypoxia is mediated by VT, with frequency remaining unaltered (16, 17).

Isocapnic hypoxia increased the gross variability, quantitated in terms of either CVs or standard deviations, of all breath components. The effect of sustained hypoxia on the gross variability of ventilation has been studied by few investigators. Berssenbrugge and coworkers (8) employed a hypobaric chamber to induce hypocapnic hypoxia (Sa_O2 = 78%) in healthy volunteers, and found that the mean intrasubject CVs of VT, TI, and TE increased by 7, 5, and 13%, respectively, over normoxic values. The increases in CVs with hypoxia in the subjects of Berssenbrugge and coworkers (8) are less than in our subjects, who exhibited increases of 15, 22, and 19% in the CVs of VT, TI, and TE, respectively, over normoxic values. The difference in the two sets of data may have occurred because the subjects of Berssenbrugge and coworkers (8) also developed hypocapnia (PCO₂ decreased by 7 mg Hg during hypoxia)which may have blunted the increase in the CVs. In addition, the longer duration of hypoxic exposure in the subjects of Berssenbrugge and coworkers (8), ~ 9 h versus 1-1.5 h in our subjects, may have contributed to the lesser changes in variability.

Effect of Hypoxia on Variational Activity of Breathing

An early method for studying the variability of breathing was the calculation of "run lengths"; a value of 1.50 breaths per run is consistent with a random series, whereas a value above 1.5 indicates a nonrandom series (18). In a study of three subjects, Bolton and Marsh (19) reported that hypocapnic hypoxia increased the run length in two of their subjects, indicating that the breath series is not entirely random and contains a nonrandom component; of interest, these two subjects also displayed periodic breathing. When the subjects were studied under conditions of hypoxia combined with eucapnia or hypercapnia, the run length did not appear to change from that observed with air breathing. The small number of subjects and wide scatter in the data preclude any rigorous statistical analysis; nevertheless, these data suggest that the overall variability of breathing may not be altered by hypoxia unless it is combined with hypocapnia.

Unlike earlier studies in which hypoxic-induced hypocapnia was not prevented, we studied awake subjects while maintaining isocapnic conditions. We found increases in the total variational activity of each breath component, which were due to increases in random behavior in all components and to increases in oscillatory behavior of all components with the exception of TI. The alterations in random behavior may represent behaviorally mediated adaptations of the respiratory controller to hypoxia. Studies in animals have demonstrated that brief hypoxic stimulation has a direct action on higher brain centers independent of its effect on chemoreceptors (20, 21). The modulating influence of the higher brain centers on the respiratory controller, as reflected by an increase in random behavior, may be a useful strategy for minimizing discomfort associated with hypoxia (22).

Employing the same mathematical technique (3), we examined the effects of elastic (6) and resistive (7) loading on the variational activity of breathing in healthy subjects. Elastic loading decreased the total variational activity of VT and TE due to changes in their random fractions. Resistive loading increased the total variational activity of TI due to alterations in its random fraction. The differing responses to mechanical loading suggest that alterations in variational activity of breathing have different physiological implications. In particular, unstructured random variability may be a measure of behavioral (cortical) influences on the respiratory controller, whereas the structured correlated fraction may represent automatic (subcortical) influences on the controller.

The strength of the relationship of I and TE between immediately adjoining breaths was decreased by isocapnic hypoxia; the number of breath lags displaying significant autocorrelation coefficients for TE also decreased. The decreased dependence of breath components on the characteristics of the preceding cycle may be related to ablation of afterdischarge with sustained hypoxia. In healthy awake subjects, Georgopoulos and coworkers (23) found that the increase in I during brief exposure (35-51 s) to hypoxia took several breaths before returning to baseline I, suggesting that hypoxia may cause activation of the afterdischarge mechanism. When the exposure to hypoxia was extended to 25 min, however, termination of the hypoxic challenge was accompanied by an immediate decrease in I to values below those during baseline room air breathing (23); the latter observation suggests that sustained hypoxia inactivates the afterdischarge mechanism. The mechanism whereby sustained hypoxia might abolish afterdischarge is not clear. Possibly, it relates to the accumulation of adenosine (24), a depressant neuromodulator, or an increase in blood flow that, in turn, decreases medullary extracellular PCO₂ and thereby depresses the heightened respiratory drive anticipated with hypoxic stimulation of the peripheral chemoreceptors (25).

Another factor contributing to the decrease in strength of the relationship of I and TE between neighboring breaths during isocapnic hypoxia may be activation of the peripheral chemoreceptors. Peripheral chemoreceptors respond rapidly (26), and can adjust deviated values of I from one breath to the next toward the mean level; such a process is likely to cause either a decrease or no change in correlated behavior. We previously demonstrated that hyperoxic hypercapnia increased the autocorrelation coefficient at a lag of one breath for I in healthy subjects (5). We attributed the increase in correlated behavior to activation of the central chemoreceptors with CO₂ and blunting of the peripheral chemoreceptors with hyperoxia. Removal of the rapid and fine-adjustment actions of the peri-pheral chemoreceptors with hyperoxia caused dominance of the slowly responding central chemoreceptors, with the result that deviations in I returned more slowly to the mean level (5).

Isocapnic hypoxia increased the number of subjects displaying significant oscillations in VT and the oscillatory fraction of variational activity in I, VT, and TE; the cycle time of oscillations in I was also decreased. To our knowledge, this is the first description of the development of hidden ventilatory oscillationsoscillations not obvious to the naked eyein awake human subjects during isocapnic hypoxia. Several investigators have demonstrated that hypoxia induces periodic breathing, although usually when combined with hypocapnia and often in the presence of sleep. In studies at actual or simulated high altitude, oscillations have been observed in I, VT, and Ttot (27, 28); in these studies, however, some subjects fell asleep and no attempt was made to prevent hypocapnia. The fact that hypoxia combined with either hypocapnia or sleep induces periodic breathing is consistent with the influence of the dynamic properties of the chemical feedback control system (12). Hypocapnia and sleep can increase loop gain, and combined with hypoxia, they serve as considerable catalysts for the development of ventilatory oscillations.

What are the potential mechanisms for the development of periodic breathing with isocapnic hypoxia? Employing a mathematical model, Khoo and colleagues (12) demonstrated that ventilatory oscillations result when the loop gain of the respiratory control system is increased secondary to delays in the circulation time between the lung and the chemoreceptors. Such a mechanism is unlikely in our subjects since hypoxia increases cardiac output and shortens circulation time. A decrease in functional residual capacity can lead to ventilatory oscillations, but this possibility is also unlikely because end- expiratory lung volume increases with hypoxia (29). Khoo and colleagues (12) emphasize the importance of activation of the peripheral chemoreceptors in causing oscillations during hypoxia. Hypoxia, however, not only affects the peripheral chemoreceptors but also acts on the central chemoreceptors. Specifically, sustained isocapnic hypoxia causes ventilatory depression, which may abolish afterdischarge (16, 23); this possibility is suggested by the decrease in the correlated behavior for I and TE in our subjects (Figure 5). If, as Younes (9) has suggested, afterdischarge plays an important role in the prevention of oscillations, the corollary may hold that the removal of afterdischarge during sustained hypoxia contributes for the development of oscillations. As such, central mechanisms may play a role in addition to peripheral chemoreceptors in the genesis of hypoxia-associated oscillations.

Additional support for the likely involvement of central mechanisms in the development of oscillations in VT is provided by the study of Brown and coworkers (30). In awake ponies with intact and denervated carotid bodies, periodic oscillations in I, VT, TI, and TE were noted during normoxia. Oscillations in I and VT were increased in intact ponies during hypocapnic hypoxia, whereas the response was variable in the denervated ponies. The occurrence of periodic breathing in the denervated ponies indicates that peripheral chemoreceptors are not essential for development of periodic breathing; instead, the investigators speculated that periodic breathing originates in the central pattern generators under modulation by several neurotransmitters.

Isocapnic hypoxia caused a 4-fold increase in the frequency of significant oscillations in I (Table 1). In other words, the number of breaths per cycle decreased from 20 during air breathing to 5 during isocapnic hypoxia; this can also be viewed as a decrease in cycle time from ~ 70 to 18 s (assuming an average breath duration of 3.5 s [1]). This shortening of the cycle time of the oscillations in I during isocapnic hypoxia is compatible with the faster response of the peripheral chemoreceptors versus the more slowly responding central chemoreceptors (26). Of interest, the reported response time of the peripheral chemoreceptors (~ 17.5 s) is remarkably close to the cycle length of the high-frequency oscillations during isocapnic hypoxia in our subjectssuggesting that activation of the peripheral chemoreceptors plays an important role in their development (31). In contrast, we previously showed that hyperoxic hypercapnia produced low-frequency oscillations in I that had a cycle time (80 s) consistent with central chemoreceptor activation (5).

In summary, isocapnic hypoxia increased the total variational activity of breath component, mediated by increases in their random and oscillatory behavior. Isocapnic hypoxia increased the number of subjects displaying significant oscillations in VT and decreased the cycle time of oscillations in I. In conclusion, hypoxia induces ventilatory oscillations even in the absence of hypocapnia and sleep, suggesting that neural responses may have a more important role in the genesis of hypoxic-induced oscillations than previously reported.