Journal of Applied Physiology

Effect of different levels of hyperoxia on breathing in healthy subjects

Heinrich F. Becker, Olli Polo, Stephen G. McNamara, Michael Berthon-Jones, Colin E. Sullivan


Becker, Heinrich F., Olli Polo, Stephen G. McNamara, Michael Berthon-Jones, and Colin E. Sullivan. Effect of different levels of hyperoxia on breathing in healthy subjects. J. Appl. Physiol. 81(4): 1683–1690, 1996.—We have recently shown that breathing 50% O2 markedly stimulates ventilation in healthy subjects if end-tidal PCO2 ( PETCO2 ) is maintained. The aim of this study was to investigate a possible dose-dependent stimulation of ventilation by O2 and to examine possible mechanisms of hyperoxic hyperventilation. In eight normal subjects ventilation was measured while they were breathing 30 and 75% O2 for 30 min, with PETCO2 being held constant. Acute hypercapnic ventilatory responses were also tested in these subjects. The 75% O2 experiment was repeated without controlling PETCO2 in 14 subjects, and in 6 subjects arterial blood gases were taken at baseline and at the end of the hyperoxia period. Minute ventilation (V˙i) increased by 21 and 115% with 30 and 75% isocapnic hyperoxia, respectively. The 75% O2 without any control on PETCO2 led to a 16% increase inV˙i, but PETCO2 decreased by 3.6 Torr (9%). There was a linear correlation (r = 0.83) between the hypercapnic and the hyperoxic ventilatory response. In conclusion, isocapnic hyperoxia stimulates ventilation in a dose-dependent way, withV˙i more than doubling after 30 min of 75% O2. If isocapnia is not maintained, hyperventilation is attenuated by a decrease in arterial PCO2 . There is a correlation between hyperoxic and hypercapnic ventilatory responses. On the basis of data from the literature, we concluded that the Haldane effect seems to be the major cause of hyperventilation during both isocapnic and poikilocapnic hyperoxia.

  • oxygen
  • ventilatory response
  • respiration
  • Haldane effect

in healthy subjects,O2 has been extensively used and its effects studied in exercise physiology, space and aviation medicine, as well as diving medicine. In patients, O2 has been widely used both in the setting of acute hypoxemia and for long-term therapy in chronically hypoxic patients. Breathing O2-enriched air transiently decreases minute ventilation (V˙i) by ∼10–20% (9, 22). This effect, which has been attributed to a decrease in carotid body output, lasts for ∼2–3 min (22). Thereafter, V˙i returns to the baseline value, although carotid body activity remains suppressed. It has long been assumed that apart from this transient initial decrease in ventilation, breathing O2-enriched air or pure O2 for up to 1 h in healthy subjects has no effect on breathing (19) or leads to a mild increase in V˙i by up to 15% (21, 28,30). We have reported a marked increase inV˙i in healthy subjects who were breathing O2-enriched air (50% O2) while end-tidal PCO2 ( PETCO2 ) was maintained and despite a constant arterial PCO2 (PaCO2 ) (3). The aim of this study was to investigate the effects of different levels of O2 on breathing and to evaluate some possible mechanisms of hyperoxia-induced stimulation of breathing.



Eight healthy men (mean age 32.8 ± 2 yr, range 23–40 yr) volunteered for the study. All subjects were familiar with the theoretical and practical aspects of ventilatory response tests. They were aware that experiments would involve exposure to either room air, O2, or hypercapnia but were not aware of the particular challenge being undertaken during testing. The protocol was approved by the Ethics Committee of the University of Sydney, and all subjects gave informed consent.

Breathing circuit.

To control inspiratory O2 and PETCO2 and to measure ventilation, the tests were done by using a closed biased-flow circuit, which has been described in detail previously (4). In brief, the system was composed of a fixed-speed blower (50 l/min), a bypassable soda-lime absorber, and a 6-liter flow-through bag that was encased in a Perspex box. A Fleisch no. 3 pneumotachograph, which was coupled to a differential pressure transducer (model DP-45, Validyne, Northridge, CA), was connected to the box. The PETCO2 was controlled by a computer-regulated system that adjusted the amount of expired air passing through the soda-lime absorber according to the PETCO2 of the preceding breath.


Subjects were seated in a comfortable chair throughout the test period and distracted with classical music. They breathed through a mouthpiece on the closed biased-flow circuit. The nose was occluded with a noseclip. The experiments were performed in random order and on different days.

The hypercapnic ventilatory response was tested by using the Read rebreathing method (26). The hyperoxic experiments consisted of three phases: a 5-min baseline period with the subjects breathing room air, a 30-min hyperoxic period (30 or 75% O2), and a 15-min recovery period with the subjects breathing room air (21 ± 2% O2) again. After the baseline period, O2 was rapidly added so that the fractional concentration of O2 in the inspired air ( FIO2 ) target level of 0.3 or 0.75 ± 0.02 was reached within 90 s. FIO2 was then maintained for 30 min by adding O2 at a low flow rate that compensated for the subject’s O2 consumption. By flushing the breathing circuit with room air, FIO2 was then lowered to 0.21 ± 0.02 within 90 s and maintained at that level for the 15-min recovery period. During both the hyperoxia and the recovery periods, PETCO2 was maintained at the level observed during the last 2 min of the baseline period.

An additional experiment using 75% O2, following the above-mentioned protocol, was performed in 14 subjects (the 8 subjects mentioned above and an additional 4 male and 2 female healthy subjects with a mean age of 30.6 yr). This experiment differed in that PETCO2 was not controlled. The recovery period was shortened to 10 min for subject comfort. In eight of these subjects, mean inspiratory PCO2 ( PICO2 ) was calculated from the raw data signal of flow and PCO2 for four to six breaths in the 2nd and 5th min of the control period and after 21 and 28 min of the hyperoxia. In another 6 of these 14 subjects, arterial blood gases were sampled at the end of the control period and at the end of the hyperoxia period. Blood was sampled from the radial artery, and blood-gas analyses were performed immediately. Local anesthesia had been administered at the site of the puncture before the beginning of the test.


Tidal volume, inspiratory and expiratory time, respiratory rate, and breath-to-breath V˙i were calculated from the flow signal produced by the pneumotachograph. FIO2 and PETCO2 were measured at the mouthpiece by a fast-response paramagnetic O2 analyzer (Datex Multicap CNO-103, Datex, Helsinki, Finland) and an infrared CO2 analyzer (model 47210A, Hewlett-Packard, Waltham, MA). Arterial O2 saturation ( SaO2 ) and heart rate were measured by using a pulse oximeter via an ear probe (model 3700e, Ohmeda). Arterial blood pressure was monitored noninvasively (Finapress 2300, Ohmeda, Boulder, CO). Blood-gas analyses were performed by using a calibrated Corning blood- gas analyzer (model 178, Ciba Corning, Medfield, MA).

Mean values for SaO2 , heart rate, and blood pressure were calculated, and PETCO2 was measured for each breath. All data were acquired with a 12-bit analog-to-digital converter sampling at 125 Hz.

Data processing.

In the hypercapnic test the slope of the ventilatory response (ΔV˙i/ ΔPCO2 ) was calculated by using linear regression. In the hyperoxic experiment, mean values for all measured variables were calculated for the last 5 min of the baseline period, the hyperoxic period, and the recovery period. For the 75% hyperoxia experiment where PETCO2 was not maintained, mean V˙i and mean PETCO2 were calculated for each minute of the test.

Statistical analyses.

The experimental design of the 30 and 75% isocapnic hyperoxia tests is a fully crossed repeated-measures factorial, with factor one being O2 concentration used during the period of hyperoxia (2 levels: 30 and 75% O2) and factor two being testing period (3 levels: control, hyperoxia, and recovery). Because treatment effects were expected to be proportional rather than additive, a two-factor repeated-measures analysis of variance (ANOVA) was performed on the log-transformed data. The following planned linear comparisons were made for both the 30 and 75% hyperoxia experiments: hyperoxia vs. control, hyperoxia vs. recovery, and control vs. recovery (14). A modified Bonferroni’s correction was made (14). To obtain more than two data points on a dose-response curve ofV˙i during hyperoxia, we have included the previously reported data on V˙iduring an identical isocapnic hyperoxia experiment using 50% O2 that had been done in the same subjects (3) .

For the poikilocapnic experiment, a pairedt-test was used to analyze the differences in arterial blood gases, PETCO2 , andV˙i between control and hyperoxia. The correlation coefficient between the hyperoxic and hypercapnic ventilatory response was calculated. One-factor ANOVA for repeated measures was used to analyze the changes in PICO2 . All values are expressed as means ± SE. Statistical significance was assumed atP values < 0.05.


Mean V˙i increased after 30 min of isocapnic hyperoxia in a dose-dependent manner (Fig.1). With 30% hyperoxiaV˙i increased on average by 21.4 ± 4.7% (P < 0.05), with 50% O2 by 61.3 ± 7.4% (P < 0.01), and with 75% hyperoxia by 114.7 ± 9.6% (P< 0.001) [the previously published results obtained during 50% isocapnic hyperoxia in the same subjects (3) have been included for convenience of comparision]. The difference in responses between the three levels of hyperoxia was significant (P < 0.001). The individual changes inV˙i during 30 and 75% isocapnic hyperoxia compared with the respective control values are shown in Fig.2. With 75% isocapnic hyperoxia, the increase in V˙i ranged from 43 to 287%, this being due to an increase in both tidal volume from 0.67 ± 0.06 to 1.08 ± 0.05 liters (P < 0.001) and breathing rate from 13.2 ± 0.8 to 17.0 ± 1.5 breaths/min (P < 0.01) (Table1). In contrast, only tidal volume increased significantly, from 0.63 ± 0.05 to 0.71 ± 0.04 liters (P < 0.01), during 30% isocapnic hyperoxia. After 15 min of recovery from 75% isocapnic hyperoxia, V˙i had decreased compared with hyperoxia (P < 0.001) but was still increased compared with the control period (P < 0.01).

Fig. 1.

Minute ventilation (V˙i) during 30, 50, and 75% isocapnic hyperoxia. Values are means ± SE calculated for last 5 min of each period. Previously published results obtained during 50% isocapnic hyperoxia (3) have been included for convenience of comparison.

Fig. 2.

Individual increase in V˙i during last 5 min of 30% (open bars) and 75% (filled bars) isocapnic hyperoxia compared with respective V˙i during control period.

View this table:
Table 1.

Effect of 30 and 75% isocapnic hyperoxia on respiratory and cardiovascular variables

PETCO2 decreased by 0.8 ± 0.2 and 0.85 ± 0.1 Torr during 30 and 75% isocapnic hyperoxia, respectively (bothP < 0.001; Table 1). SaO2 increased during both levels of hyperoxia (both P < 0.001) as expected, but there was no difference in the noninvasively measured SaO2 between the 30 and 75% isocapnic hyperoxia (Table 1).

Heart rate did not change during the experiment. In both isocapnic hyperoxia tests, systolic blood pressure increased during the hyperoxic and the recovery period compared with baseline (bothP < 0.01; Table 1).

The average ventilatory response to hypercapnia was 3.0 ± 0.4 l ⋅ min−1 ⋅ Torr−1(range 1–4.3 l ⋅ min−1 ⋅ Torr−1). There was a significant linear correlation (P < 0.01) between the ventilatory response to hypercapnia and 75% isocapnic hyperoxia (r = 0.83; Fig.3).

Fig. 3.

Correlation between ventilatory response to hypercapnia and increase inV˙i during 75% isocapnic hyperoxia. Δ, change.

In the experiment using 75% O2where PETCO2 was not controlled, V˙i also increased in all subjects, the average increase being 15.7 ± 3.4% (P < 0.001). In this test, mean PETCO2 decreased from 41.3 ± 0.7 Torr during the control period to 37.7 ± 0.7 Torr during the last 5 min of hyperoxia, a decrease of 3.6 ± 0.3 Torr (P < 0.001). The results of the blood-gas analyses for the six subjects are shown in Table2. PETCO2 corresponded well with PaCO2 , the former being 42.2 ± 0.6 Torr during the control period and 38.8 ± 0.5 Torr during the hyperoxia period for the six subjects in whom blood-gas analyses were done. pH increased during hyperoxia, but the change did not reach statistical significance (P = 0.11).

View this table:
Table 2.

Effect of nonisocapnic hyperoxia on arterial Pco 2 and pH

The time course of the ventilatory response and the PETCO2 throughout this experiment for all subjects is shown in Fig.4. Dead space of the breathing circuit was 56 ml and was unchanged in all tests. Because of the dead space of the circuit, mean PICO2 was 3.7 ± 0.2 and 3.3 ± 0.3 Torr during the 2nd and 5th min of the control period and 4.1 ± 0.4 and 4.0 ± 0.5 Torr during the 21st and 28th min of hyperoxia. Statistically, this small increase was not significant.

Fig. 4.

Time course of V˙i and end-tidal Pco 2( Embedded Image ) during 75% nonisocapnic hyperoxia. Values are means ± SE calculated for each minute of the experiment. Open circles, 21% O2; closed circles, hyperoxia.


Breathing O2-enriched air while isocapnia was maintained led to hyperventilation in healthy subjects. This increase in V˙i was dose dependent. The stimulation of breathing was mild while subjects were breathing 30% O2, butV˙i more than doubled with 75% O2. When PETCO2 was not controlled during the experiment using 75% O2,V˙i increased by 16%, but both PETCO2 and PaCO2 decreased significantly. The hyperoxia-induced increase in V˙i was much less pronounced when PETCO2 was not maintained, presumably because the resulting decrease in PaCO2 counterbalances the O2-induced stimulation of breathing. However, this result clearly indicates that hyperoxia has a respiratory stimulatory effect independent of any potential effects of PETCO2 control. There was a significant correlation between the ventilatory response to hyperoxia and hypercapnia.

Technical issues.

In previously published work (3), we have shown that ventilation does not significantly change during 45 min of normoxic ( FIO2 0.21 ± 0.02) isocapnic sham experimental recording. This is evidence against the possibility that either discomfort or minor fluctuations in PICO2 are responsible for the hyperventilation seen during hyperoxia. We have previously shown by blood-gas sampling that the rise in ventilation during 50% isocapnic hyperoxia was not due to a change in PaCO2 (3).

All subjects were blinded as to the test protocol that was performed on any given day so that an anticipation of the reaction by the subjects was excluded. The control of PETCO2 was crucial in our experiments. The system reacts to changes in PETCO2 and does not anticipate them. PETCO2 was actually slightly lower during hyperoxia compared with baseline (due to finite gain of the servo-controller), a fact that would have biased our results against an increase inV˙i.

Another technical factor might have led to the unexpected increase inV˙i seen during hyperoxia. PETCO2 measurement by using infrared devices is influenced by the PO2 such that these analyzers will underestimate PETCO2 at increasing O2 levels. Maintenance of the measured value of PETCO2 in this condition would erroneously increase the inspired PCO2 . The capnograph used for our experiments (model 47210A, Hewlett-Packard) was equipped to function at different levels of hyperoxia so that during the hyperoxic tests the subject’s PETCO2 was not underestimated. Furthermore, the decrease in both PETCO2 and PaCO2 during the nonisocapnic hyperoxia experiment clearly demonstrates that controlling PETCO2 does not induce the ventilatory or blood gas changes but only unmasks the effect of hyperoxia. The blood-gas analyses demonstrate that PETCO2 reflects PaCO2 under the conditions of our experiments.

The transition to hyperoxia was done by emptying the circuit’s reservoir bag and refilling it rapidly with O2. This period, which lasted for up to 90 s, was excluded from the analysis because the values were artificial. Therefore, we were not able to study the previously reported transient depression of ventilation caused by acute O2 administration.

Comparison with previous work.

Numerous experiments have been performed that studied the effect of O2 breathing on PaCO2 or PETCO2 . It has long been known (8, 10, 19) that breathing O2 leads to a decrease in alveolar PCO2 in healthy subjects. The same effect has been demonstrated in animals (5, 12).

Although this decrease in PETCO2 is generally agreed on, there is some controversy concerning PaCO2 during O2 breathing. Lambertsen et al. (18) reported a decrease in PaCO2 of 2 Torr after 1 h of breathing 100% O2 at normobaric conditions. Our study demonstrated a fall in both PaCO2 and PETCO2 after 30 min of hyperoxia without PETCO2 being controlled, in concordance with this previous report. Experiments in animals have also demonstrated a decrease in PaCO2 during O2 breathing (5). In disagreement with these data, Larson and Severinghaus (19) found no change in PaCO2 after 10 min of O2 breathing despite a decrease in PETCO2 . However, the exposure to hyperoxia was shorter than in the above-mentioned studies. Lenfant (20) reported an increase in PaCO2 during O2 breathing. This discrepant result is likely to be due to the fact that the subjects studied included both patients with chronic obstructive pulmonary disease (COPD) and healthy subjects. In patients with COPD, PaCO2 increases during O2 breathing because of increasing ventilation-perfusion mismatch and decreased ventilation (1). These mechanisms are negligible in normal subjects (9, 29). The published data thus strongly suggest that PETCO2 and PaCO2 decrease during sustained poikilocapnic O2breathing in normal subjects. However, if PETCO2 is maintained, PaCO2 does not change during hyperoxia in healthy subjects (3).

i has been shown to either not change (19) or slightly increase by up to 15% during poikilocapnic O2 breathing in normal individuals (21, 27). We demonstrated a 16% increase inV˙i after 30 min of 75% O2 breathing without PETCO2 being maintained, in concordance with previous results (21, 27).

Holtby et al. (13) studied the effect of O2 on the ventilatory response to repetitive hypoxia using isocapnic hyperoxia ( FIO2 ∼1.0) for 5–10 min. They reported a 25% increase inV˙i after 5 min of O2 breathing and interpreted this as being in concordance with previous studies. In our experiment using 75% isocapnic hyperoxia, the increase inV˙i by 18.7% after 5 min was similar to that reported after 5 min of hyperoxia by Holtby et al. However, 30 min of hyperoxia led to a more than fourfold increase inV˙i compared with 5 min. Thus the exposure time to O2 also seems crucial for its effect on ventilation.

The changes in heart rate were not significant. The small increase in systolic blood pressure during both the hyperoxia and the recovery period seems to be related to the duration of the test, because it occurred during 30% and 75% isocapnic hyperoxia and in the previously reported sham experiment (3).


The mechanisms causing the mild increase inV˙i seen in previous nonisocapnic hyperoxia experiments until this point remain controversial. Our experiments have confirmed this small increase in V˙i during nonisocapnic hyperoxia but have demonstrated a marked dose-dependent hyperventilation caused by isocapnic hyperoxia. Several mechanisms that might explain the stimulation of breathing during nonisocapnic hyperoxia, such as the Haldane effect, a change in cerebral blood flow, ventilation-perfusion mismatching, direct activation of respiratory neurons, stimulation of irritant receptors in the lung, a change in metabolic rate, or other factors, have been suggested. Taking the well-documented effects of O2 on CO2 transport, ventilation-perfusion mismatching and cerebral blood flow in normal subjects into account, we have calculated the effect of these changes on PCO2 and the resulting ventilatory effect.

Under physiological conditions, 0.6 mmol/l (13.3 ml/l) of the 1.8 mmol/l (40 ml/l) of CO2 eliminated in the lung comes from carbamino carriage in the hemoglobin of venous blood (25). Nonoxygenated hemoglobin has a higher transport capacity for CO2 than does oxygenated hemoglobin because of the reduced carbamino carriage and the decreased buffering capacity of oxygenatedblood [Christiansen-Douglas-Haldane effect, or in short the Haldane effect (6)]. During the breathing of room air, mixed venous O2 saturation ( Sv¯ O2 ) is ∼70% and SaO2 95% (v and a, respectively, in Fig. 5). During the breathing of 100% O2, Sv¯ O2 increases by ∼10% (25). On the basis of the blood-gas values obtained during 75% nonisocapnic hyperoxia, we calculated an increase in Sv¯ O2 by 8% in our experiment. Because of the Haldane effect the amount of CO2 in carbamino carriage will decrease by approximately one-third, that is, 0.2 mmol/l (4.4 ml/l). The higher acidity of the oxygenated hemoglobin leads to a reduction in buffering capacity of blood as shown by the reduced slope of the 78% O2 saturation ( SO2 ) line compared with the 70% SO2 line [Fig.5; based on in vivo human data from Lambertsen et al. (16, 17)]. If V˙i remained unchanged, brain venous blood PCO2 would move to point h in Fig. 5, thus increasing by 2.8 Torr. In healthy subjects, however, the increase in brain tissue PCO2 due to the Haldane effect will only be transient, because the higher PCO2 causes an increase inV˙i. Assuming, for simplicity, perfect regulation of brain tissue PCO2 by the central chemoreceptors, the actual brain venous PCO2 will be at point v′ and the arterial point at a′ due to the increase in V˙i caused by the Haldane effect. Although there is some discussion concerning the relative contribution of carbamino carriage and change in buffering capacity in the Haldane effect (see Ref. 15 for references), this does not influence the above-stated calculations.

Fig. 5.

CO2 dissociation curve and carbamino carriage for physiological Embedded Image range. Carbamino carriage and buffering capacity of blood decrease with increasing O2 saturation, as demonstrated for values of 70, 78, and 95% (according to data from Refs. 16, 17, and25). Reduced buffering capacity of more oxygenated blood is expressed by reduced slope of curve, and reduction in carbamino carriage as downward shift of curve. a and v, arterial and mixed venous values during room air breathing, respectively; h, change in venous Embedded Image caused by increase in mixed venous O2 saturation of 8%; X, difference in arteriovenous CO2content. Ensuing increase in V˙i returns venous Embedded Image to baseline (v′) and shifts arterial points (a′) downward, thus maintaining the same arteriovenous CO2 content difference (X′). During poikilocapnic hyperoxic experiment there is a reduction in cerbral blood flow, increasing arteriovenous CO2 content difference (Y). This results in a further reduction in arterial Embedded Image (a”).

During the 75% nonisocapnic hyperoxia experiment, PETCO2 decreased by 3.6 Torr. This will reduce cerebral blood flow by ∼10% (11). From the Fick equation it can be calculated that this will lead to an increased difference in arteriovenous CO2 content, thus to a decrease in PaCO2 by 1.5 Torr (a” in Fig. 5).

Thus, during 75% nonisocapnic hyperoxia, the known effects of O2 on CO2 transport and cerebral blood flow will lead to an increase in both the difference between arteriovenous CO2 content and PCO2 . According to our calculations for humans, PETCO2 should decrease by 4.3 Torr from its baseline value. The measured decrease in PETCO2 in the nonisocapnic hyperoxia experiment was 3.6 Torr. There was, however, a small increase in PICO2 by 0.5 Torr from the control to the hyperoxic phase, which may account for the difference in the measured and calculated values.

As has been shown in the literature, the difference between venous Pco 2 and PaCO2 increases during O2 breathing (17). In the poikilocapnic experiment, the hyperventilation resulting from the increase in central chemoreceptor PCO2 decreases PaCO2 , thus returning the former toward its original value. In the isocapnic experiments, however, PETCO2 and thus PaCO2 are kept constant. Therefore, this experiment will lead to an increase in tissue PCO2 at the central chemoreceptor by 2.8 Torr (as PaCO2 remains constant, cerebral blood flow will not decrease). The mean ventilatory response to hypercapnia in our subjects was 3 l ⋅ min−1 ⋅ Torr−1. The predicted increase in V˙i due to the calculated 2.8-Torr increase in tissue PCO2 during 75% isocapnic hyperoxia would therefore have been 8.4 l/min. The measured increase inV˙i in the experiment was 9.7 l/min.

Because we have not actually measured cerebral venous blood gases, we had to base our calculations on data from the literature (16, 17) for normal subjects, which might, however, not be true in our subjects. Further direct measurements of the central venous blood gases during hyperoxia are certainly needed to confirm the data obtained by Lambertsen and co-workers (16, 17), which form the basis of our calculations. If the available data are correct, then our calculations suggest to us that the Haldane effect seems to be the most important mechanism in causing hyperventilation during both isocapnic and poikilocapnic hyperoxia. The clearly dose-dependent hyperventilation can also be explained by this mechanism, because SV¯O2 will be higher with increasing FIO2 levels. The high correlation between the hypercapnic and the hyperoxic ventilatory response further supports the suggested mechanism, i.e., in both tests hypercapnia is the relevant stimulus. Our results and hypothesis emphasize the need to measure central venous, in addition to arterial, blood gases, because PaCO2 may not reflect changes in venous PCO2 and thus changes at the central chemoreceptor.

It is interesting to note that in intact unanesthetized cats, hyperoxia does not alter ventilation (12, 23) but leads to a decrease in PETCO2 by ∼2 Torr (12). After chemodenervation, hyperoxia caused an increase inV˙i by 16 and ∼30%, respectively (12,23). These results suggest that the hyperoxic hyperventilation is caused centrally, a finding that might be in concordance with the Haldane effect. Miller and Tenney (23) hypothesize that in the absence of the carotid bodies O2“raises the excitatory state of the central chemoreceptor.” This hypothesis could be reconciled with our suggestion that the Haldane effect plays an important role in hyperoxic hyperventilation. Acidosis, which is present in the chemodenervated animals, will shift the CO2 binding curve to the right and lower its slope (6). The carbamino transport of CO2 is substantially reduced with acidosis (2). The additional lowering of the CO2 transport capacity during hyperoxia would then have to be compensated by a bigger increase inV˙i compared with the intact animal. Again, experiments are necessary to verify this hypothesis.

Several other mechanisms have been suggested that might explain the mild increase in V˙i seen in previous nonisocapnic hyperoxia experiments. However, when the present data are considered, most of the suggested mechanisms seem very unlikely to contribute significantly to hyperoxic hyperventilation. Metabolic rate, which seems to be increased during hyperoxia in newborns (24), does not change in adults (31). In normal subjects breathing pure O2, a shunt of ∼1% of cardiac output will develop (29). However, this has a negligibly small effect on PaCO2 .

Hyperoxia might lead to a stimulation of irritant lung receptors because of the toxic effect of O2. Activation of these receptors, however, characteristically leads to rapid shallow breathing and coughing (7). The subjects in our study did not report any such symptoms, and tidal volume increased during hyperoxia. These findings suggest that stimulation of irritant receptors does not play an important role in our experimental results.

Importance for research and patient management.

O2 is widely used both as a therapeutic agent and in research experiments in cardiovascular and respiratory medicine. Whereas the implication for the treatment of patients remains to be studied, our data have implications for all experiments using hyperoxia. The results suggest that the stimulation of breathing during hyperoxia and the increase in tissue PCO2 have to be considered. Breathing high O2 concentrations, even for a short period, will decrease PETCO2 and PaCO2 , if PETCO2 is not maintained. As a result of arterial hypocapnia, cerebral perfusion and vascular tone will be altered. V˙i will be increased by high O2 levels, which may therefore influence results of ventilatory response tests. If PETCO2 is maintained, PCO2 at the central chemoreceptor will be increased and this will lead to marked hyperventilation.


In conclusion, isocapnic hyperoxia leads to hyperventilation. This effect increases with increasing FIO2 , is substantial with high levels of O2 and is revealed when PaCO2 is held constant during hyperoxia. However, even when PETCO2 is not maintained and is allowed to fall during O2breathing, there is still clear evidence that hyperoxia stimulates ventilation. There is a close correlation between the hyperoxic and hypercapnic ventilatory response, suggesting that the hyperventilation caused by hyperoxia is caused by hypercapnia in the central chemoreceptor. Although further measurements of central venous blood gases during hyperoxia are needed, the available data suggest that the Haldane effect seems to be the most important mechanism leading to the increase inV˙i during both isocapnic and poikilocapnic hyperoxia. In the poikilocapnic experiment, reduced cerebral blood flow due to a decrease in PaCO2 is an additional factor leading to a small increase in V˙i, thus decreasing PETCO2 . In experiments using O2 breathing, these effects have to be taken into account.


We are grateful to Prof. Iven Young for valuable comments and to Monique Aarts and Natalie Edwards for technical assistance. We thank the subjects for their cooperation.


  • Address for reprint requests: H. Becker, Schlafmedizinisches Labor, Dept. of Medicine, University of Marburg, 35033 Marburg, Germany.

  • This work was supported by Deutsche Forschungsgemeinschaft Grant Be 1608/1-1.


View Abstract