Repeated hypoxic exposures change respiratory chemoreflex control in humans


Corresponding author J. Duffin: Department of Physiology, University of Toronto, Toronto, Ontario, Canada M5S 1A8., Email:


  • 1A group of seven volunteers (5 male, 2 female) were exposed to 20 min isocapnic (eucapnic) hypoxia once daily for 14 consecutive days. Their chemoreflexes were measured before and after each exposure. The same volunteers repeated the exposures with air substituted for the hypoxic gas mixture in a pseudorandom crossover design.
  • 2On day 1 an initial ventilatory response to hypoxia and subsequent decline was discernible in two volunteers, but the mean response for all volunteers at this stage was not significant. However, the response gradually increased, and by day 14 was discernible in six volunteers making the mean response for all volunteers significant. No change was observed over the 14 days of air exposure.
  • 3Only the chemoreflex threshold measured in iso-oxic (hypoxic) modified rebreathing tests changed significantly, and only for the series of exposures to hypoxia.
  • 4Over 14 days, the mean ±s.e.m. threshold for all volunteers fell proportionately, from 42 ± 1.1 mmHg on day 1 to 39 ± 1.0 mmHg on day 14. By contrast, the mean ±s.e.m. threshold, for all volunteers and all days, rose from 40 ± 0.4 mmHg before to 42 ± 0.5 mmHg after the hypoxic exposures.
  • 5We conclude that the enhancement of the initial ventilatory response to hypoxia induced by repeated hypoxic exposure is produced by a decrease in chemoreflex threshold. However, the decline in the ventilatory response during a single exposure is produced by an increase in the chemoreflex threshold. Since threshold changes were only found for hypoxic (iso-oxic) modified rebreathing tests, we conclude that only the peripheral chemoreflex changed.

Exposing adult humans to hypoxia produces changes in the chemoreflex control of breathing; two examples of this process are the decline of the initial ventilatory response in the 20 min following exposure, even if hypocapnia is prevented (e.g. Easton et al. 1986), and the enhancement of the initial ventilatory response by repeated exposures (e.g. Serebrovskaya et al. 1999). The mechanisms mediating these changes in respiratory control are unknown but we hypothesise that changes in the peripheral chemoreflex are responsible.

Although a number of hypotheses have been proposed to account for the decline in the initial ventilatory response to hypoxia (for a review see Robbins, 1995), no consensus has been reached. One proposal, that a reduction in central chemoreceptor stimulation due to a withdrawal of central chemoreflex stimulation as central carbon dioxide partial pressure declines with the hypoxia-induced increase in cerebral blood flow, is not supported by measurements of cerebral blood flow (Poulin & Robbins, 1998). Another proposal is that hypoxia produces a rise in the central chemoreflex threshold due to central acidosis (Sato et al. 1992). Some investigations implicate a decline in peripheral chemoreflex activity (Robbins, 1995; Garcia et al. 2000a). Also considered is a peripheral chemoreceptor-mediated depression of central respiratory drive (Dahan et al. 1994; Kimura et al. 1998). The hypotheses involving the peripheral chemoreflex receive support from the observation that subjects with bilaterally resected carotid bodies lack a response to hypoxia (Honda, 1992; Kimura et al. 1998).

Bascom et al. (1990) measured the ventilatory responses to brief pulses of additional hypoxia and hypercapnia during 23 min of sustained isocapnic hypoxia. Since both declined, they reasoned that the carotid bodies were involved, because a depression of the peripheral chemoreflex should affect both responses. By contrast, Georgopoulos et al. (1990) found that the ventilatory response to 5 min normoxic carbon dioxide was unchanged after 25 min of normocapnic hypoxia. Berkenbosch et al. (1992) concurred; they also measured the normoxic ventilatory response to carbon dioxide before and after 25 min of isocapnic hypoxia, and found that while both the peripheral and central chemoreflex sensitivities to carbon dioxide were somewhat larger after hypoxia than before, they were not significantly so. These investigators suggested that a rise in the peripheral chemoreflex threshold for carbon dioxide might account for the decline in the initial ventilatory response to hypoxia. We tested this hypothesis by measuring this threshold directly, before and after exposure to 20 min isocapnic (eucapnic) hypoxia.

Hypotheses to account for the augmentation of the initial ventilatory response to hypoxia by repeated exposures are scarce, possibly because studies of the effects of repeated exposures to hypoxia on chemoreflex control in humans are few. Repeated exposures to hypoxia can take the form of either several brief episodes within a short period (e.g. 4 bouts of 5 min separated by 10 min) or daily hypoxic exposures for several days (e.g. 20 min day−1 for 14 days).

For the first type, recent work in goats, dogs and cats shows that several brief bouts of hypoxia within 1 h induces both a progressive augmentation in ventilatory responses to hypoxia and a long-term respiratory chemoreflex facilitation; by contrast, these changes were not found in humans (for a review see Powell et al. 1998). Nevertheless, increased ventilatory responses to hypoxia have been reported in obstructive sleep apnoea patients, who are subject to many short hypoxic episodes while sleeping (Narkiewicz et al. 1999).

For the second type, involving daily exposures to hypoxia, several studies in humans have demonstrated that the initial ventilatory response to hypoxia is enhanced (Levine et al. 1992; Savourey et al. 1996; Katayama et al. 1998; Serebrovskaya et al. 1999; Garcia et al. 2000b). Most of these studies involved subjects spending time in hypobaric chambers at simulated altitudes: 30 min daily for 6 days at 4 500 m (Katayama et al. 1998) or 8 h daily for 5 days, day 1 at 4500 m, day 5 at 8500 m (Savourey et al. 1996), or 45 min daily for 5 days week−1 for 5 weeks (Levine et al. 1992), with the latter complicated by exercise during the exposures. However, two studies involved hypoxic exposures at normal barometric pressures, that of Serebrovskaya et al. (1999) in subjects exposed to three 5 min bouts of progressive hypoxia daily for 14 consecutive days, and that of Garcia et al. (2000a) with 2 h daily exposures to 13 % inspired oxygen for 12 days. Such findings also implicate changes in the peripheral chemoreflex. We tested this hypothesis by measuring the chemoreflex characteristics in humans for a period of 14 days of daily 20 min isocapnic (eucapnic) exposures to hypoxia.



This study was approved by the human ethics committee of the University of Toronto and conforms to the Declaration of Helsinki. A group of seven healthy volunteers (5 male, 2 female) gave their informed consent after the experimental protocols and possible risks had been explained. The volunteers were non-smokers taking no medication, and with no history of cardiorespiratory diseases. Their mean ±s.d. age, height and body mass were 23.7 ± 3.1 years, 174.2 ± 4.7 cm and 65.0 ± 3.7 kg, respectively.


Each volunteer was familiarised with all procedures by performing trial runs on each set of apparatus during an introductory session. Approximately 1 week later, they began 14 consecutive days of exposures to air or hypoxia (assigned pseudorandomly). Volunteers were seated during the ≈1 h testing sessions which began with a 5 min rest, followed by the first modified rebreathing test. After a 10 min rest, volunteers were exposed to hypoxia or air for 20 min, and immediately afterwards undertook the second modified rebreathing test. The same testing schedule was repeated after a minimum of 1 month with the alternate exposure to hypoxia or air in a pseudorandom crossover design. During the quiet periods of these experiments volunteers usually spent their time reading, sitting quietly or listening to the radio.

Modified rebreathing

The modification of Read's original rebreathing test (Read, 1967), first described by Casey et al. (1987) and Duffin & McAvoy (1988), has recently been updated (Mohan & Duffin, 1997), as well as the interpretation of its results (Duffin et al. 2000). In its hyperoxic form the modified rebreathing test measures central chemoreflex sensitivity equivalent to those measured using Read's technique and the end-tidal forcing technique (Mohan et al. 1999).

Two modifications are made to Read's rebreathing technique. First, rebreathing is preceded by 5 min of hyperventilation to lower stores of carbon dioxide, thereby allowing the chemoreflex threshold to be discerned as carbon dioxide rises through it. This modification also permits measurement of the subthreshold ventilation independent of the chemoreflexes, ‘basal’ ventilation estimating the ‘wakefulness drive’ (Fink, 1961).

In the second modification, iso-oxia is maintained during rebreathing by supplying oxygen to the rebreathing bag, so that both hyperoxic (150 mmHg) and hypoxic (50 mmHg) ventilatory responses to carbon dioxide can be examined. Comparison of hypoxic and hyperoxic measurements determine if changes occurred in peripheral or central chemoreflexes or both, making assumptions discussed in Duffin et al. (2000). The daily pairs of modified rebreathing tests before and after the exposures were either both hypoxic (50 mmHg) or both hyperoxic (150 mmHg); the choice was alternated between days so that over the 14 day period each volunteer undertook seven pairs of each type.

Each rebreathing test consisted of 5 min of hyperventilation, during which volunteers were coached to maintain an end-tidal partial pressure of carbon dioxide between 19 and 25 mmHg. The rebreathing bag initially contained oxygen at the desired iso-oxic partial pressure, and carbon dioxide at a partial pressure of about 42 mmHg. Rebreathing began at the end of expiration and was followed by three deep breaths, producing rapid equilibration of carbon dioxide partial pressures in the bag, lungs and arterial blood to that of mixed venous blood. The equilibration was verified by observation of a plateau in end-tidal carbon dioxide partial pressure, and was a prerequisite for continuing the test. Rebreathing was terminated either when ventilation exceeded 100 l min−1 or the end-tidal partial pressure of carbon dioxide exceeded 60 mmHg, although some volunteers, females and small males, had tests terminated at lower levels of ventilation because of discomfort.

During rebreathing, the volunteers wore nose-clips and breathed through a mouthpiece connected to one side of a wide-bore Y-valve (Collins P-319, 80 ml dead space) that permitted switching from room air to the rebreathing bag. The 5 l rebreathing bag was enclosed in a rigid container with a 50 mm diameter tube connected to a dry rolling-seal spirometer (Spiroflow model 130, Morgan) to monitor breath-by-breath changes in ventilation. A sample flow of 90 ml min−1 was drawn from the mouthpiece side of the Y-valve (gas sample tube UD 5037, Brüel and Kjær) for end-tidal monitoring (anaesthetic gas monitor type 1304, Brüel and Kjær,) of carbon dioxide partial pressure (resolution of 1 mmHg) and oxygen partial pressure (resolution of 4 mmHg), and returned to the bag during rebreathing. Iso-oxia was maintained by a flow of oxygen to the rebreathing bag side of the Y-valve under computer control.

Hypoxic or air exposures

During the 20 min exposure to hypoxia or air, isocapnia at eucapnic partial pressures was maintained using a method described by Sommer et al. (1998). Volunteers wore a facemask (8930 series, Hans Rudolph Inc.) attached to a non-rebreathing valve (8932 T-shape 2WNRBV, Hans Rudolph Inc.). Expired volume was measured from the outlet of the non-rebreathing valve with a dry gas meter (Parkinson-Cowan), and respired gases were sampled at the mouthpiece as for the rebreathing test.

At first, the inspired gas was supplied from a 3 l reservoir bag filled continuously from a compressed gas cylinder containing room air; with the flow manually set to match each volunteer's average resting ventilation during the first 5 min. This average flow was supplied from a compressed gas cylinder containing room air for the air exposures, but containing 10 % oxygen for the hypoxic exposures. If inspired flow exceeded that supplied, the reservoir bag emptied and a demand valve (Mares, Beta, modified to reduce its opening pressure) supplied flow from another compressed gas cylinder containing 6 % carbon dioxide (approximately mixed venous carbon dioxide partial pressure) and 10 % oxygen. This arrangement maintained isocapnia at eucapnic partial pressures despite increases in ventilation produced by hypoxia.

Data acquisition

A 16-bit analog to digital converter (National Instruments, AT-MIO-16XE-50) digitised the analog signals for on-line computer analysis using specially written software (LabVIEW, National Instruments). The software calculated tidal volumes, inspiratory and expiratory times, ventilation, and end-tidal partial pressures of carbon dioxide and oxygen on a breath-by-breath basis. The volume changes and partial pressures of oxygen and carbon dioxide were also written on a chart recorder (Lineacorder mark VII WR 3101, Graphtec) to monitor the initial rebreathing equilibration. This measurement system was calibrated before each experimental session using gases of known concentrations (room air and a primary standard mixture of 5 % carbon dioxide, 10 % oxygen), and a calibrated 1 l volume syringe (model 5540, Hans Rudolph Inc.).

Data analysis

The data accumulated during the hypoxic and air exposures were analysed using a specially designed spreadsheet (Microsoft Excel). For display purposes, the breath-by-breath ventilation and end-tidal partial pressures of carbon dioxide and oxygen for each exposure were first averaged over 60 s to allow mean responses for all volunteers to be calculated for any day of exposure. For statistical comparisons, these 60 s averages were then combined to calculate means for all volunteers over 5 min epochs corresponding to the 5 min prior to hypoxic or air exposure (T1) and the first (T2) and last (T3) 5 min of hypoxic or air exposure for any day. A repeated measures analysis of variance with the significance level set at 5 % was used to detect significant differences, and individual parameters were contrasted using Student-Newman-Keuls comparisons (SigmaStat 2.0, SPSS).

The data accumulated from the rebreathing tests were analysed using a specially designed spreadsheet (Microsoft Excel). First, breaths from the initial three breath equilibration, as well as sighs, swallows and breaths incorrectly detected by the software were excluded from further analysis. Next, the breath-by-breath end-tidal partial pressures of carbon dioxide were plotted against time and fitted with a least-squares regression line. The equation for this line provided a predicted value of end-tidal partial pressure of carbon dioxide vs. time, thereby minimising interbreath variability due to measurement. Subsequently, tidal volume (ml body temperature and pressure, saturated, BTPS), respiratory rate (breaths min−1) and ventilation (l min−1 BTPS) were plotted against the predicted end-tidal partial pressure of carbon dioxide (mmHg). Figure 1B and C shows examples.

Figure 1.

Comparison of respiratory responses

A, mean (±s.e.m.) end-tidal partial pressures of oxygen and carbon dioxide and ventilatory responses to isocapnic (eucapnic) hypoxia are greater on day 14 (filled symbols) than on day 1 (open symbols). Variables were binned (1 min) and averaged for all volunteers. B, the ventilatory response to hypoxic iso-oxic rebreathing of a representative volunteer before (open symbols) and after (filled symbols) a single 20 min exposure to isocapnic (eucapnic) hypoxia on day 8, showing a post-hypoxic increase in threshold. C, the ventilatory response to iso-oxic (hypoxic) rebreathing of a representative volunteer before a single 20 min exposure to isocapnic (eucapnic) hypoxia on day 1 (open symbols) and day 14 (filled symbols), showing a decrease in chemoreflex threshold (E,T1) with daily hypoxic exposures.

Each of these plots was fitted with a model made up of the sum of two or three segments separated by one or two breakpoints, respectively (Duffin et al. 2000). Model fitting was based on minimising the sum of least squares for non-linear regressions using commercial software (SigmaPlot 5.0, SPSS). The first segment was an exponential decline to a final value, the latter taken as a measure of the basal ventilation, basal tidal volume and basal breathing frequency (E,B, VT,B and fB characteristics, respectively). The exponential decline was chosen to fit any waning of short-term potentiation of ventilation produced by hyperventilation that might have occurred; it was observed in only 17 of 384 (4.4 %) of tests. In those tests without such a trend the decay constant of the exponential decline reverted to a value less than 1, so that the basal values were equivalent to the mean values below the first breakpoint.

The second and possible third segments were straight lines from the first to the second breakpoint, and above the second breakpoint, respectively. The first breakpoint was taken as a measure of the chemoreflex threshold for the ventilation, tidal volume and breathing frequency responses to carbon dioxide (E,T1, VT,T1 and fT1 characteristics, respectively). The second breakpoint was taken as the point at which the ventilatory response pattern changed, in terms of tidal volume and frequency, but was observed in so few tests that it is not reported here. The slopes of the first straight-line portions were taken as the chemoreflex sensitivity for the ventilation, tidal volume, and breathing frequency responses (E,S1, VT,S1 and fS1 characteristics, respectively).

The three chemoreflex characteristics, measured 7 times over the course of 14 days, were first classified into groups, according to: type of exposure series (air or hypoxia), iso-oxia of rebreathing test (hyperoxic or hypoxic) and time of measurement (before or after exposure). To detect significant changes, hyperoxic and hypoxic rebreathing test results were compared separately using two-way repeated measures analyses of variance on 2 × 7 matrices, with the significance level set at 5 % (SigmaStat 2.0). To test for changes over the 14 days the two factors were air vs. hypoxic exposure and the seven factors were days. To test for changes due to a single exposure the two factors were before vs. after and the seven factors were days. When significance was observed, individual parameters were contrasted using Student-Newman-Keuls comparisons, and trends were specified using regression (SigmaStat 2.0).

In addition to these comparisons we also examined the variability of the parameters from all the air exposure sessions by calculating their coefficient of variation over the 14 day period, and then their correlation using the Pearson product moment correlation coefficient (SigmaStat 2.0).


Response to daily exposures

All volunteers completed the entire protocol, and there were no significant differences in eucapnic end-tidal carbon dioxide partial pressures measured over 14 days of hypoxic or air exposures. While the initial ventilatory response to hypoxia and subsequent decline was only discernible in two volunteers on day 1, a statistical examination of the ventilation epochs for all days showed that it gradually increased over the 14 days (r = 0.630, P = 0.0157), becoming discernible in six volunteers by day 14, so that by the end of 14 days a clear acute hypoxic ventilatory response and subsequent decline were discernible in the mean response for all volunteers. Figure 1a compares the mean ventilatory response of the seven volunteers to 20 min isocapnic (eucapnic) hypoxia on day 1 with that on day 14.

A comparison of the 5 min epochs of ventilation (Table 1) on days 1 and 14 showed that the increase in response was significant (P = 0.042). While there were no differences on day 1, differences became significant by day 7 and onward, and on the last exposure (day 14), the initial ventilation increase was significantly larger, 14.8 ± 0.5 l min−1, than either before hypoxia, 11.0 ± 0.7 l min−1, or after 15 min hypoxia, 10.4 ± 0.5 l min−1. By contrast, ventilation in the air exposure experiments remained unchanged throughout the 14 day period.

Changes in chemoreflex characteristics over 14 days

No changes in any chemoreflex characteristics were found during the series of exposures to air. By contrast, during the series of exposures to hypoxia the chemoreflex threshold (E,T1), measured with hypoxic (iso-oxic) rebreathing tests before exposure, progressively decreased over the 14 days. This decrease is illustrated in Fig. 1C showing hypoxic iso-oxic rebreathing tests from a representative volunteer performed before the exposure to hypoxia. The threshold of the ventilatory response to carbon dioxide observed on day 14 is lower than that on day 2. A gradual decrease in the threshold over the 14 days is apparent in the mean thresholds for all volunteers shown in Fig. 2 and was significant (P = 0.029), resulting in a significant negative correlation (r = -0.925, P = 0.003).

Figure 2.

Chemoreflex threshold changes

The variation of mean ±s.e.m. chemoreflex threshold (E,T1), determined from the ventilatory responses to hypoxic (iso-oxic) rebreathing tests for all volunteers, with day of exposure, before (□) and after (□) daily 20 min exposures to isocapnic (eucapnic) hypoxia. A strong negative correlation was found between threshold and days of exposure before the exposures but not after. The post-hypoxia thresholds were significantly higher than the pre-hypoxia thresholds from day 8 onwards.

Mean ±s.e.m. E,T1 for all volunteers decreased from 42 ± 1.1 mmHg on day 1 to 39 ± 1.0 mmHg on day 14. By comparison, E,T1 determined from the corresponding hypoxic (iso-oxic) rebreathing tests for the room air exposure protocol) showed no correlation with days (r = 0.045, P = 0.930); mean ±s.e.m. E,T1 was 40 ± 1.3 and 40 ± 1.1 mmHg on the first and last day, respectively. Neither did E,T1 values determined from hypoxic (iso-oxic) rebreathing tests done after the hypoxic exposures (r = -0.316, P = 0.489); mean ±s.e.m. E,T1 was 41 ± 1.5 and 39 ± 1.5 mmHg on the first and last day, respectively.

Changes in chemoreflex characteristics with a single exposure

No significant changes in any chemoreflex characteristic were found for single exposures to air (Table 2). By contrast, single exposures to hypoxia raised the chemoreflex threshold (E,T1) measured with hypoxic (iso-oxic) rebreathing tests significantly (P = 0.003). This increase is illustrated in Fig. 1B showing hypoxic iso-oxic rebreathing tests from a representative volunteer performed before and after a single hypoxic exposure on day 8. The threshold of the ventilatory response to carbon dioxide observed after the exposure is higher than that before. The mean ±s.e.m. E,T1 for all volunteers and all days rose from 40 ± 0.4 mmHg before exposure to 20 min hypoxia to 42 ± 0.5 mmHg after (Table 3).

In addition, VT,T1 and fT1, measured with hypoxic rebreathing tests, also rose significantly (P = 0.002 and P = 0.039, respectively). Mean ±s.e.m. VT,T1 for all volunteers and all days rose from 39 ± 0.4 mmHg before exposure to 20 min hypoxia to 41 ± 0.5 mmHg after (Table 3). Similarly, mean ±s.e.m. fT1 for all volunteers and all days rose from 42 ± 0.6 to 43 ± 0.5 mmHg (Table 3). By contrast, no such differences were observed for E,T1, VT,T1 or fT1 before and after exposure to air (Table 2).

Correlations between chemoreflex characteristics

Because no differences were detected in the chemoreflex characteristics measured during the air exposures, we combined the before and after measurements and tested them for their variability with the days of the experiment. They were classified according to the iso-oxia of the rebreathing tests as hypoxic or hyperoxic. The coefficients of variation over the 14 days for E,B (0.833 and 0.866) and E,S1 (0.599 and 0.859), for hypoxic and hyperoxic tests, respectively, were similar, but the coefficient of variation for E,T1 was much less (0.076) for both hypoxic and hyperoxic tests.

To test for correlations among these chemoreflex characteristics we averaged them over the 14 days for each volunteer, again classifying them according to the iso-oxia of the rebreathing tests with which they were measured, and tested them using Pearson's product moment correlation. Neither E,T1 and E,B or E,T1 and E,S1 were significantly correlated. However, E,S1 was positively correlated with E,B (r = 0.861, P = 0.013 and r = 0.922, P = 0.003) for hypoxic and hyperoxic tests, respectively.



The main findings of this study are threefold. (1) Fourteen days of daily 20 min exposures to isocapnic (eucapnic) hypoxia enhanced the initial ventilatory response to hypoxia because of a decrease in the chemoreflex threshold. (2) During single 20 min exposures to isocapnic (eucapnic) hypoxia the ventilatory response slowly declined because of a rise in the chemoreflex threshold. (3) Significant changes in chemoreflex threshold were measured only with hypoxic (iso-oxic) rebreathing tests. These findings are now discussed in detail.

Methodological considerations

Isocapnic breathing

The breathing circuit (Sommer et al. 1998) used to maintain isocapnia during the exposures was the same whether the exposure was to air or hypoxia, and end-tidal carbon dioxide partial pressure was maintained to within ≈1 mmHg of resting partial pressures throughout the exposures; thus, although the isocapnia was different for each volunteer it did not differ between exposures. Therefore, changes in isocapnia either during a single exposure or over the 14 days would not account for our findings.

We chose to maintain isocapnia at each volunteer's eupnoeic partial pressure so as to avoid the complication of hypercapnia during the exposures. In addition, eucapnic isocapnia provided a comfortable exposure period for the volunteers. However, as a result, the ventilatory response to hypoxia was less than that observed by other investigators where end-tidal partial pressures of carbon dioxide are often maintained 2-3 mmHg above resting (e.g. Bascom et al. 1990; Berkenbosch et al. 1992; Tansley et al. 1998). Nonetheless, the repeated hypoxic exposures were still effective; demonstrating that hypoxia alone exerted an effect on the chemoreflexes, so that by the 14th day the initial ventilatory response to hypoxia and its subsequent decline were clearly discernible.


The use of this modified rebreathing method to measure chemoreflex characteristics has been discussed extensively in previous reports (Mohan & Duffin, 1997; Duffin et al. 2000). With respect to its use here, the rebreathing tests did subject the volunteers to regular, although brief, exposures to hypocapnia (during the hyperventilation), as well as hypercapnia and alternately hypoxia or hyperoxia (during rebreathing). However, since only the series of exposures to hypoxia produced changes in chemoreflex characteristics, we are confident that hypoxia and not the modified rebreathing method used to measure the chemoreflex characteristics instigated the changes.

The modified rebreathing technique provided a direct measurement of the chemoreflex threshold to carbon dioxide by least-squares regression. By contrast, other methods extrapolate a fitted line backwards to the zero ventilation point on the carbon dioxide partial pressure axis and this intercept is often termed the apnoeic threshold (Berkenbosch et al. 1992; Dwinell et al. 1997). If it were a true estimate of the apnoeic threshold, then E,T1, as measured in the present investigation, would depend on E,B, but we found no such correlation. Indeed, increases in basal ventilation with moderate exercise do not change E,T1 (Duffin & McAvoy, 1988). A further drawback to the use of extrapolation to determine the chemoreflex threshold is the inevitable correlation that will exist between it and the sensitivity to carbon dioxide (E,S1). Volunteers with a low sensitivity to carbon dioxide will have much lower apnoeic thresholds than those with high sensitivity to carbon dioxide. We found no such correlation between E,T1 and E,S1. We therefore suggest that an observable threshold that is not correlated with other parameters provides a better estimate of the chemoreflex threshold.

Timing of tests.

Two timing considerations are relevant to our findings. First, we performed the daily experiments on each volunteer at approximately the same time of day; circadian effects on respiratory control (Stephenson et al. 2000) would therefore not account for our findings. Second, the first rebreathing test was performed a maximum of 20 min before the isocapnic (eucapnic) exposure to either hypoxia or room air, and the second test immediately after the exposure protocol and completed within 15 min.

Although no appreciable recovery of the acute response to hypoxia occurs during 7 min of breathing room air, recovery is complete after approximately 1 h (Easton et al. 1988). If the hyperventilation of the modified rebreathing test accelerated this recovery, then the changes in chemoreflex threshold produced by exposure to hypoxia may have been attenuated and therefore be greater than those we observed.

Changes in chemoreflex characteristics

The repeated daily exposures to hypoxia produced a marked enhancement of the initial ventilatory response to hypoxia by the end of the 14 days, a finding similar to that of others using a variety of repeated exposure protocols (Levine et al. 1992; Savourey et al. 1996; Katayama et al. 1998; Serebrovskaya et al. 1999; Garcia et al. 2000b). We also observed that during each 20 min exposure to hypoxia the ventilatory response declined slowly, again a finding similar to that of others (e.g. Easton et al. 1986; Bascom et al. 1990; Berkenbosch et al. 1992; Sato et al. 1992; Garcia et al. 2000a).

In contrast to many previous studies, we also measured the chemoreflex characteristics before and after each exposure, throughout the 14 days. The chemoreflex threshold, measured before each daily exposure, became progressively lower as the 14 days passed. However, the threshold measured after each exposure did not change; it was consistently higher than the chemoreflex threshold measured before each exposure, and the difference between them increased as the 14 days passed. No other chemoreflex characteristic changed significantly, including the basal ventilation, which estimates non-chemoreflex drives to ventilation. From the latter observation we concluded that repeated episodes of hypoxia did not facilitate any other drive to breathe, in agreement with McEvoy et al. (1996).

Our observations corroborate the contention of Dahan et al. (1994) that stimulation of the peripheral chemoreflexes is necessary to provoke a decline in the ventilatory response to hypoxia. In addition, we observed that the decline was proportional to the initial response, as have others (e.g. Khamnei & Robbins, 1990; Kimura et al. 1998). Considering these findings, we suggest that the decline may be viewed as a waning or accommodation of the ventilatory response to hypoxia rather than a hypoxia-induced inhibition of ventilation. The lack of any ventilatory response to hypoxia in volunteers with bilaterally resected carotid bodies (Honda, 1992; Kimura et al. 1998) supports this view.

We further suggest that the changes we measured in the chemoreflex threshold account for both the decline in the ventilatory response to hypoxia observed during single exposures to hypoxia, and the enhancement of the initial ventilatory response to hypoxia following repeated hypoxic exposures. As the chemoreflex threshold falls progressively with the repeated hypoxic exposures, the same isocapnic hypoxia produces an increased ventilatory response, but as the hypoxia continues during a single hypoxic exposure the chemoreflex threshold rises and the ventilatory response therefore declines. The latter contention receives support from previous investigations. Bascom et al. (1990) and later Berkenbosch et al. (1992) and Sato et al. (1992) suggested that the sensitivity of the ventilatory response to carbon dioxide is unaltered after hypoxic exposure, and posited that the threshold was raised, although these studies were unable to observe directly the threshold due to limitations in the testing procedures used.

Changes in chemoreflex threshold may result from changes in either the central or the peripheral chemoreflex pathways. To assist in distinguishing between the two we used both hyperoxic and hypoxic (iso-oxic) rebreathing tests, with the peripheral chemoreflex response enhanced in the latter. Any chemoreflex parameter changes determined from hypoxic rebreathing tests, which are not also observed for parameters determined from hyperoxic tests, may be attributed solely to altered peripheral chemoreflexes. This situation was observed and so we concluded that only the peripheral chemoreflex had changed.

Since only the threshold changed and not the sensitivity, ventilation drive for a given iso-oxic hypercapnia was changed, but not the change in ventilation for a given change in iso-oxic hypercapnia. We therefore suggest that the alterations in the peripheral chemoreflex were mediated either through a change in the general level of activity of the peripheral chemoreceptors or in the effectiveness of the transmission of their afferent signals. Garcia et al. (2000b) too have recently suggested that hypoxic ventilatory decline in humans involves a decrease in hyperoxic ventilatory drive without a change in hypoxic sensitivity. For example, we speculate that a depletion of stored neurotransmitter could explain the decrease in hypoxic responsiveness after a 20 min exposure to hypoxia. However, by repeating the hypoxic stimulus daily, the production of neurotransmitter could be up-regulated; on subsequent exposures to hypoxia more stored neurotransmitter would be available for release and an enhanced acute ventilatory response to hypoxia would result. Nevertheless, despite the enhanced stores, neurotransmitter depletion again produces a decline in the response.


We observed that both single and repeated hypoxic exposures produced changes in the peripheral chemoreflex threshold. This study therefore implicates hypoxia as a stress to the peripheral chemoreflex capable of inducing alterations in its responsiveness. We suggest that any environmental condition or respiratory disorder that involves bouts of hypoxia may provoke such changes, and predict that they would be observed in conditions such as sleep apnoea.


We acknowledge the expert advice of the Respiratory Research Group at the University of Toronto. This research was supported by a grant from the Ontario Thoracic Society.