American Journal of Respiratory and Critical Care Medicine

We studied interrelationships between exercise endurance, ventilatory demand, operational lung volumes, and dyspnea during acute hyperoxia in ventilatory-limited patients with advanced chronic obstructive pulmonary disease (COPD). Eleven patients with COPD (FEV1.0 = 31 ± 3% predicted, mean ± SEM) and chronic respiratory failure (PaO2 52 ± 2 mm Hg, PaCO2 48 ± 2 mm Hg) breathed room air (RA) or 60% O2 during two cycle exercise tests at 50% of their maximal exercise capacity, in randomized order. Endurance time (Tlim), dyspnea intensity (Borg Scale), ventilation (V˙ e), breathing pattern, dynamic inspiratory capacity (ICdyn), and gas exchange were compared. PaO2 at end-exercise was 46 ± 3 and 245 ± 10 mm Hg during RA and O2, respectively. During O2, Tlim increased 4.7 ± 1.4 min (p < 0.001); slopes of Borg, V˙ e, V˙ co 2, and lactate over time fell (p < 0.05); slopes of Borg–V˙ e, V˙ e–V˙ co 2, V˙ e–lactate were unchanged. At a standardized time near end-exercise, O2 reduced dyspnea 2.0 ± 0.5 Borg units, V˙ co 2 0.06 ± 0.03 L/min, V˙ e 2.8 ± 1.0 L/min, and breathing frequency 4.4 ± 1.1 breaths/min (p < 0.05 each). ICdyn and inspiratory reserve volume (IRV) increased throughout exercise with O2 (p < 0.05). Increased ICdyn was explained by the combination of increased resting IRV and decreased exercise breathing frequency (r2 = 0.83, p < 0.0005). In conclusion, improved exercise endurance during hyperoxia was explained, in part, by a combination of reduced ventilatory demand, improved operational lung volumes, and dyspnea alleviation.

Ambulatory oxygen therapy has been shown in several controlled studies to improve exercise performance and to relieve exertional dyspnea in patients with chronic obstructive pulmonary disease (COPD) (1-5). However, responses to this intervention are highly variable and are unpredictable in any given individual (6-9). The mechanisms of improvement when breathing oxygen are complex and poorly understood. Ultimately, the success of ambulatory oxygen therapy in COPD likely depends on its net effect on integrated cardiopulmonary function and symptom generation. Previous studies have identified several potential contributing factors that include (1) altered central perception of dyspnea, independent of the drop in ventilation; (2) reduced ventilatory demand; (3) improved respiratory and peripheral muscle function; and (4) possible cardiovascular effects (1-9).

It is a common clinical observation that some patients with COPD and unequivocal ventilatory limitation to exercise show marked improvements in exercise performance with ambulatory oxygen. To gain new insights into the mechanisms of this improvement, we examined the effects of supplemental oxygen in a group of patients with COPD whose exercise was limited primarily by ventilatory insufficiency, that is, patients with severe lung hyperinflation and a limited ability to increase respired volume or flow with exercise. Our hypothesis was that oxygen therapy would reduce ventilatory demand, reduce the rate of dynamic hyperinflation and, therefore, reduce the stress on the ventilatory system during exercise, thus improving exercise endurance. On the basis of previous work (10), we further postulated that relatively small changes in operational lung volumes, as a result of reduced ventilation and altered breathing pattern, would convey important clinical benefit in this group who breathe at lung volumes close to their TLC (11).

Using a randomized, double-blind, cross-over design, we compared the acute effects of room air and 60% oxygen on ventilation, operational lung volumes, breathing pattern, dyspnea intensity, and metabolic parameters in hypoxemic patients with stable, advanced COPD during constant-load exercise. We explored potential mechanisms of improvement in exercise endurance by studying interrelationships between the above-listed dynamic physiological and psychological variables.


We studied 11 clinically stable patients with advanced COPD (FEV1 < 50% predicted) who met medical criteria for ambulatory O2 in Ontario (Ministry of Health's Home Oxygen Program): (1) PaO2 ⩽ 55 mm Hg or oxygen saturation ⩽ 88% at rest or (2) PaO2 between 56 and 60 mm Hg at rest with desaturation to ⩽ 88% for ⩾ 2 min during exercise. Patients also had severe activity-related dyspnea with a score of ⩽ 6 on the modified Baseline Dyspnea Index (12). Patients with other significant disorders that could contribute to dyspnea or exercise limitation were excluded.

Study Design

This study was a randomized, double-blind, placebo-controlled, cross-over trial with local university/hospital research ethics approval. After giving written informed consent, patients were familiarized with all testing procedures and completed a symptom-limited incremental exercise test. In a subsequent visit, subjects performed two constant-load exercise tests at approximately 50% of their previously determined maximal work rate while breathing either 60% O2 or room air (RA, 21% O2), in randomized order, with a 60- to 90-min washout or recovery period between tests. Subjects were blinded to the oxygen concentration being breathed, as was the investigator evaluating subjective responses and performing data analysis.


Subjects performed pulmonary function and cycle exercise tests as previously described (7). In addition, subjects described their breathing discomfort at the end of exercise by selecting descriptor phrases from a questionnaire modified from that of Simon and coworkers (13).

Operational lung volumes. Assuming that TLC did not change during exercise (14), measurements of dynamic inspiratory capacity (ICdyn) were used to derive end-expiratory lung volume (EELVdyn = TLC − ICdyn) and inspiratory reserve volume (IRV = ICdyn − tidal volume [Vt]). Tidal flow–volume loops were also placed relative to each subject's TLC, using concurrent IC measurements. IC maneuvers were carried out at the end of each 10-min resting baseline until three reproducible efforts were achieved (within 5%), every 2–3 min during exercise, and at peak exercise. This has been found to be a reliable and responsive method of tracking acute changes in lung volume (10, 15, 16).

Statistical Analysis

Results are presented as means ± SEM. A statistical significance of 0.05 was used for all analyses, with appropriate Bonferroni corrections for multiple comparisons. Before treatment comparisons were made, the possibility of sequence effects was evaluated (17). Treatment comparisons were made using paired t tests. Exercise endurance was evaluated as total cumulative work performed [Σ work rate [% predicted maximum] × minutes); exercise slopes were expressed as means of individual regression lines; isotime exercise was defined as the highest equivalent minute of exercise completed during both tests.

Pearson correlations were used to establish associations between change in endurance (ΔWork) with hyperoxia and concurrent changes in relevant independent variables (i.e., dyspnea, leg discomfort, minute ventilation [V˙e], respiratory frequency [fr], Vt, inspiratory capacity [IC], IRV, oxygen consumption [V˙o 2], carbon dioxide production [V˙co 2], lactate, PaO2 , PaCO2 ). Stepwise multiple regression analysis was carried out with these variables and possible covariates (baseline lung function and gas exchange) to establish the best equation for improvement in exercise endurance. Similar analyses were carried out to examine interrelationships between changes in dyspnea intensity, ventilation, operational lung volumes, breathing pattern, and other relevant cardioventilatory parameters.

Subject characteristics are summarized in Table 1 (18-22). No significant sequence effects were found in this study with respect to exercise responses (i.e., measurements of exercise endurance, symptom intensity, ventilation and metabolic responses), thereby allowing a valid analysis of treatment effects in response to supplemental O2.


Age, yr 68 ± 2
Height, cm163 ± 2
Weight, kg 68 ± 6
Body mass index, kg/m2 25.5 ± 1.9
Modified baseline dyspnea index4.5 ± 0.3 (severe)
Peak V˙ o 2, L/min (% predictededicted maximum)0.47 ± 0.09 (38)
Peak V˙ o 2, ml/kg/min7.2 ± 1.1
Pulmonary function and gas exchange (% of predicted normal)
 FEV1, L0.65 ± 0.06 (31)
 FVC, L1.59 ± 0.11 (53)
 FEV1/FVC, %41 ± 3 (59)
 TLC, L6.86 ± 0.51 (127)
 RV, L5.07 ± 0.50 (237)
 FRC, L5.64 ± 0.50 (190)
 IC, L1.23 ± 0.12 (50)
 Pi max, cm H2O42 ± 4 (59)
 SRaw, cm H2O · s26.7 ± 2.5 (666)
 Dl CO, ml/min/mm Hg6.9 ± 0.8 (36)
 PaO2 (room air), mm Hg52.4 ± 2.2
 PaCO2 (room air), mm Hg48.5 ± 2.1
 pH (room air)7.41 ± 0.02
 HCO3 (room air), mM28.1 ± 1.7

* n = 11. Values represent means ± SEM. Pulmonary function variables in parentheses represent the percentage of predicted normal values. Predicted normal values for spirometry, lung volumes, Dl CO, and Pi max were those of Morris and associates (18), Goldman and Becklake (19), Gaensler and Wright (20), and Hamilton and associates (21), respectively. Peak V˙ o 2 was compared with predicted normal values of Jones (22).

Symptom-limited Exercise Endurance

Symptom-limited endurance exercise at 23 ± 3 W (26 ± 5% of the predicted maximum work rate) on RA was severely curtailed at 38 ± 7% predicted maximum V˙o 2 after 4.1 ± 0.9 min (Table 2). Although endurance time increased significantly by 4.7 ± 1.4 min (p < 0.001) during 60% O2, peak values for V˙o 2 (n = 7, where technically satisfactory) and V˙co 2 did not change significantly with added O2 (Table 2). Breathing 60% O2 during exercise did not result in any significant change in peak Borg ratings (23) of dyspnea or leg discomfort; however, slopes of Borg ratings of both dyspnea and leg discomfort over time fell significantly during hyperoxia (p < 0.01) (Figure 1).


Room Air60% O2
Endurance time, min4.1 ± 0.98.8 ± 1.3*
Dyspnea intensity, Borg rating5.2 ± 0.75.0 ± 0.8
Leg discomfort, Borg rating4.1 ± 0.84.5 ± 0.8
Reason for stopping exercise, number of subjects:
 Leg discomfort03
HR, beats/min112 ± 5110 ± 4
SaO2 , % 82 ± 2 99 ± 0.1*
PaO2 , mm Hg45.9 ± 2.8244.7 ± 10.4*
PaCO2 , mm Hg53.3 ± 3.058.0 ± 5.0*
Lactate, mM2.9 ± 0.42.8 ± 0.4
co 2, L/min0.54 ± 0.090.53 ± 0.09
e, L/min23.0 ± 3.522.4 ± 2.9
e/V˙ co 2 45.0 ± 2.245.4 ± 4.2
fr, breaths/min30.0 ± 2.028.6 ± 2.2
Vt, L0.77 ± 0.110.80 ± 0.10
ICdyn, L1.07 ± 0.131.25 ± 0.16
IRV, L0.30 ± 0.040.45 ± 0.08
Dynamic lung hyperinflation (DH), L0.26 ± 0.070.21 ± 0.08
EILV, %TLC 95 ± 193 ± 1

*p < 0.01, significant difference between room air and 60% O2.

Other reasons for stopping exercise included general “tiredness” (n = 1) and discomfort on the bicycle seat (n = 1).

p < 0.05, significant difference between room air and 60% O2.

All patients reported breathing discomfort as a primary reason for stopping exercise while breathing RA: eight subjects stopped exercise because of breathlessness alone, whereas the remaining three stopped because of a combination of both breathing and leg discomfort. At the end of RA exercise, the main descriptors used to describe dyspnea were “I feel a need for more air” (82% of subjects), “I cannot get enough air in” (64%), and “breathing in requires effort” (64%). Breathing discomfort was still a primary reason for stopping exercise during O2 in six patients (five because of breathlessness alone, and one because of the combination of breathing and leg discomfort); however, three subjects then stopped because of leg discomfort and two for other reasons (one because of general “tiredness,” one because of discomfort from sitting on the bicycle seat).

Of the patients who stopped exercise because of dyspnea during both RA and O2 tests (n = 6), there were no changes in peak measurements of ventilation, operational lung volumes, or breathing pattern. Interestingly, in the subgroup (n = 5) that continued exercise and stopped because of a new reason during O2 (i.e., leg discomfort, other), there was a significant improvement in IRV (p < 0.05) and IC (p = 0.05) throughout exercise and at peak exercise.

Ventilatory Responses to Exercise

While breathing room air, mean peak V˙e was low at only 23.0 ± 3.5 L/min; the small increase in V˙e during exercise was accomplished almost exclusively through an increase in fr, because Vt was severely constrained from above (minimal IRV) and below (increased EELVdyn) (Figure 2). In all subjects at rest and throughout exercise, tidal expiratory flows met or exceeded isovolume maximal flows throughout Vt (Figure 3A).

Gas exchange and various other parameters measured at rest on RA and 60% O2 are presented in Table 3. Gas exchange responses to 60% O2 during exercise are shown in Figure 4 and are reported in Tables 2 and 4. Exercise response slopes that fell significantly when expressed over time included (Figures 2 and 3) V˙e (p = 0.015), lactate (p = 0.14), V˙co 2 (p = 0.021), breathing frequency (p = 0.002), IRV/predicted TLC (p = 0.011), and IC% predicted (p = 0.031) (Table 5). To highlight these reductions that became more apparent in the latter part of exercise with O2, variables were compared at a time near exercise cessation on RA with those at isotime during exercise on O2 (Table 4). V˙o 2 responses to hyperoxia at isotime were variable, and thus, we could not evaluate the possibility of a shift in substrate utilization during exercise (n = 7): V˙o 2 was similar or tended to go down in five of seven subjects, but increased in the other two subjects, with no significant change on average. At isotime, the fall in V˙e correlated strongly with the fall in V˙co 2 (r2 = 0.86, p < 0.0005). The fact that V˙e/V˙co 2 and V˙e/lactate slopes were not altered with hyperoxia also supports the notion that these variables changed in proportion to each other (Figure 5).


Room Air60% O2
Dyspnea intensity, Borg rating 0.8 ± 0.2 0.9 ± 0.3
HR, beats/min 94 ± 3 87 ± 2*
BP systolic, mm Hg123 ± 6125 ± 5
BP diastolic, mm Hg 74 ± 3 73 ± 4
SaO2 , %88.5 ± 1.998.8 ± 0.1
PaO2 , mm Hg 56 ± 3244 ± 9
PaCO2 , mm Hg 44 ± 3 51 ± 4*
pH7.41 ± 0.017.39 ± 0.01
co 2, L/min0.21 ± 0.030.16 ± 0.02
e, L/min11.0 ± 1.110.0 ± 1.1
fr, breaths/min19.3 ± 1.518.3 ± 2.3
Te, s2.28 ± 0.262.42 ± 0.28
Vt, L0.60 ± 0.070.56 ± 0.06
ICdyn, L1.33 ± 0.131.46 ± 0.18
IRV, L0.73 ± 0.130.91 ± 0.18
EILV, %TLC88 ± 286 ± 3

*p < 0.05, significant difference between room air and 60% O2.

p < 0.01, significant difference between room air and 60% O2.


Room Air60% O2 Mean Difference
Dyspnea intensity, Borg rating 4.9 ± 0.52.9 ± 0.6−2.0
Leg discomfort, Borg rating3.9 ± 0.72.7 ± 0.8−1.2
HR, beats/min114 ± 4103 ± 4−11
co 2, L/min0.50 ± 0.080.45 ± 0.06−0.06
e, L/min21.5 ± 2.818.7 ± 1.9−2.8
e/V˙ co 2 45.3 ± 2.245.3 ± 3.10.04
fr, breaths/min28.4 ± 1.623.9 ± 1.8−4.4
Ti, s0.72 ± 0.050.85 ± 0.070.13
Te, s1.46 ± 0.091.81 ± 0.170.35
Ti/Ttot0.33 ± 0.010.33 ± 0.020.00
Midtidal expiratory flow, L/s0.72 ± 0.130.54 ± 0.10−0.18
Midtidal inspiratory flow, L/s1.51 ± 0.171.42 ± 0.140.09
Vt, L0.76 ± 0.100.80 ± 0.090.04
Vt/IC, %73 ± 363 ± 4−10
ICdyn, L1.05 ± 0131.34 ± 0.170.29
DH from rest, L0.28 ± 0.060.13 ± 0.04−0.16§
IRV, L0.29 ± 0.050.53 ± 0.110.25
EILV/TLC, %95 ± 192 ± 2−3
Vd/Vt, %37 ± 441 ± 44
Lactate, mM2.7 ± 0.42.3 ± 0.3−0.5
PaO2 , mm Hg45.3 ± 2.6245.9 ± 9.3200.6
PaCO2 , mm Hg53.3 ± 3.057.6 ± 3.44.4
Pet CO2 43.1 ± 2.047.2 ± 1.94.1

*3.6 ± 0.9 min.

p < 0.1, significant difference between room air and 60% O2. 

p < 0.05, significant difference between room air and 60% O2.

§p < 0.065, significant difference between room air and 60% O2.


Room Air* 60% O2 * p Value
Symptom intensity
 Dyspnea–time, Borg/min1.47 ± 0.290.53 ± 0.130.004
 Dyspnea–V˙ e, Borg/L/min0.79 ± 0.340.32 ± 0.08NS
 Leg discomfort–time, Borg/min1.20 ± 0.250.70 ± 0.150.006
 Leg discomfort–V˙ co 2 10.2 ± 7.010.5 ± 3.5NS
Slopes over time
 V˙ e–time, L/min/min3.13 ± 0.661.54 ± 0.310.015
 V˙ co 2–time0.09 ± 0.020.04 ± 0.010.021
 Lactate–time 0.5 ± 0.1 0.3 ± 0.10.014
 fr–time 3.2 ± 0.7 1.4 ± 0.40.002
 Vt–time0.03 ± 0.030.02 ± 0.01NS
 IC–time−0.10 ± 0.04−0.02 ± 0.010.055
 IC% predicted–time−3.9 ± 1.4−0.0 ± 0.00.031
 IRV–time−0.14 ± 0.04−0.03 ± 0.020.013
 IRV% predicted TLC–time−2.5 ± 0.7−0.7 ± 0.30.011
 HR% predicted max–time 3.3 ± 0.7 1.9 ± 0.50.002
Slopes expressed against ventilation
 F–V˙ e 1.41 ± 0.570.63 ± 0.19NS
 Vt–V˙ e 0.00 ± 0020.03 ± 0.01NS
 IC–V˙ e −0.08 ± 0.05−0.00 ± 0.01NS
 IRV–V˙ e −0.08 ± 0.04−0.03 ± 0.01NS
 V˙ e–lactate 9.9 ± 3.311.3 ± 4.2NS
 V˙ e–V˙ co 2 33.0 ± 2.933.3 ± 6.9NS

*Values represent means ± SEM.

NS = no significant difference between room air and 60% O2.

Operational Lung Volumes

Along with the decrease in ventilation during O2, there was a corresponding decrease in midtidal expiratory flow rates that allowed tidal flow–volume loops to shift rightward within the maximal loop (Figure 3); that is, there was an increase in IC and IRV at rest (see above) and throughout exercise. At isotime during exercise with O2, there was also a significant reduction in the extent of dynamic hyperinflation (−ΔICdyn): ICdyn was decreased by 0.28 ± 0.06 and 0.13 ± 0.04 L from rest at isotime exercise on RA and O2, respectively (p < 0.05). IC (and IRV) were also increased for any given ventilation during exercise with O2 due, in part, to alterations in breathing pattern. By stepwise multiple regression analysis, the fall in IC at isotime was best predicted by the combination of an increase in resting IRV and a decrease in isotime breathing frequency (r2 = 0.83, p < 0.0005). Of note, the increase in IRV at rest occurred, in part, as a result of associated small reductions in resting V˙co 2 (ΔIRV versus ΔV˙co 2, r = −0.81, p = 0.003) and, in turn, V˙e (ΔIRV versus Δ V˙e, r = −0.75, p = 0.008).

Mechanisms of Improved Exercise Endurance and Relief of Breathlessness

Improvements in exercise endurance (ΔWork) correlated best with changes in the slopes of V˙e/time (r = −0.626, p = 0.039) and lactate/time (r = −0.625, p = 0.040), and with baseline resting maximal midexpiratory flow (MMEF)% predicted (r = 0.656, p = 0.028) and resting IC% predicted (r = 0.603, p = 0.049); that is, the subjects who improved exercise endurance the most were those with the least baseline constraints on tidal volume or flow expansion, and with greater reductions in exercise ventilation or lactate. By stepwise multiple regression analysis, the combination of the baseline MMEF% predicted and Δlactate/time accounted for 71% of the variance in ΔWork (r = 0.841, p < 0.0005). In a subgroup of five subjects who improved exercise endurance by less than 2 min with O2, there was no significant change in operational lung volumes at rest (ΔIRV = 0.00 ± 0.07 L, ΔIC = 0.02 ± 0.04 L) or at isotime during exercise on O2 (ΔIRV = 0.08 ± 0.11 L, ΔIC = 0.19 ± 0.11 L). This was in contrast to the remaining six subjects who had significantly larger increases in endurance with O2 (by 7.4 ± 4.7 min), as well as significant improvements in operational lung volumes at rest (ΔIRV = 0.32 ± 0.17 L, ΔIC = 0.23 ± 0.11 L; p < 0.05 each) and at isotime during exercise (ΔIRV = 0.39 ± 0.16 L, ΔIC = 0.37 ± 0.15 L; p < 0.05 each).

Dyspnea–ventilation relationships fit a single-exponential model that did not change significantly with the addition of supplemental O2 (Figure 6): equations for this model were Borg = 0.16e0.16( e) (r2 = 0.98) and 0.24e0.13(V˙ e) (r2 = 0.99) on RA and O2, respectively. Therefore, Borg ratings of dyspnea intensity and ventilation were reduced proportionally within this relationship during hyperoxia.

The slopes of dyspnea over exercise time fell inversely to the slopes on RA; that is, the greater the intensity of exertional dyspnea on RA, the greater the reduction with O2 (r = −0.895, p < 0.0005). Improvement in respiratory sensation did not correlate with baseline pulmonary function, gas exchange, PaO2 , PaCO2 , or ventilatory mechanics. Although perceived leg discomfort during exercise fell significantly during hyperoxia, it was not the primary symptom limiting exercise on RA. Thus, reductions in leg discomfort were not as important with respect to improvements in exercise endurance.

The novel aspects of this study are as follows. First, hyperoxia had profound effects on dyspnea and exercise endurance in this group of patients with chronic ventilatory insufficiency. Second, improved exercise performance was primarily related to reduced ventilatory demand, which, in turn, led to improved operational lung volumes and a delay in the attainment of limiting ventilatory constraints on exercise and the onset of intolerable dyspnea. Third, modest changes in submaximal ventilation and dynamic ventilatory mechanics resulted in relatively large improvements in symptom intensity and exercise capacity. Fourth, reduced submaximal ventilation was likely linked, in part, to altered metabolic requirements under hyperoxic conditions. Finally, hyperoxia resulted in additional, and potentially important, effects on cardiovascular function and perceived leg discomfort.

Effects on Ventilatory Responses

Ventilatory limitation and the attendant severe breathing discomfort were the primary factors curtailing exercise in these severely hyperinflated patients with chronic respiratory failure. Perusal of the tidal and maximal flow–volume loops on room air confirmed their limited ability to increase ventilation further when challenged with the increasing metabolic demands of exercise (Figure 3). Significant ventilatory constraints were evident at peak exercise on RA: end-inspiratory lung volume (EILV)/TLC was 95%, and dynamic IRV was only 0.3 L on average.

Oxygen therapy resulted in modest, but consistent, reductions in submaximal ventilation, by an average of 11% at isotime, with a significant time delay in reaching the peak ventilation attained under RA conditions. Reduced submaximal ventilation was achieved primarily by a reduction in breathing frequency (by 16% at isotime), because Vt in these severely hyperinflated patients was relatively fixed.

The mechanism by which hyperoxia results in a reduction in submaximal ventilation throughout exercise is debated. Many previous studies have suggested that the primary mechanism is direct reduction of peripheral chemoreceptor activation with consequent reduced central medullary motor drive (i.e., loss of the hypoxic stimulus to breathe) (1-4). V˙e/V˙co 2 slopes were identical on RA and oxygen; therefore, the reduction of V˙e at isotime correlated strongly with simultaneous V˙co 2 reduction (Figure 5). Thus, reduced V˙co 2 at rest and during low levels of exercise could have been reflective of reduced ventilatory drive and the attendant reduced work of breathing, which is high in such patients. Reduction in V˙co 2, particularly toward the end of exercise, may additionally reflect reduced acid buffering effects because of reduced metabolic acidemia (7). With hyperoxia, oxygen delivery to the active peripheral muscles is increased for a given blood flow with less reliance on anaerobic glycolysis. In our subjects, hyperoxia delayed metabolic acidemia, of which blood lactate is a marker. Lactate levels at the breakpoint of exercise during RA and oxygen were identical; therefore, oxygen delayed further accumulation of lactate despite the large increases in cumulative work performed. However, the relative contribution of reduced hypoxic drive and altered metabolic loading to reduced submaximal V˙e during hyperoxia could not be determined in this study, but both mechanisms are likely to be instrumental.

Effects on Operational Lung Volumes

In conjunction with reduced ventilation–time slopes, oxygen also resulted in a delay in dynamic hyperinflation during exercise: at isotime, the change in EELVdyn from rest was 0.13 L, on average, which was half the magnitude of the change in EELVdyn on room air at this point of comparison. Thus, while the changes in EELVdyn at peak exercise on RA and oxygen were similar, the time to reach this level of hyperinflation was greatly increased (i.e., by greater than 2-fold) when breathing oxygen. The reduced dynamic lung hyperinflation (DH) was associated with less restrictive mechanical constraints: IRV was significantly increased at isotime, thus delaying the onset of ventilatory limitation. There was a spectrum of responses to hyperoxia among study subjects; those individuals who had only minimal changes in endurance time were the ones whose measured ICdyn did not change on oxygen compared with RA. Patients with the greatest response to oxygen were those with the largest reserves for tidal volume and flow generation at baseline, and who had the greatest reductions in V˙e–time and lactate–time slopes during oxygen.

The extent of DH in these patients with advanced disease is less than that reported previously in patients with less severe COPD, who had average increases of 0.3 to 0.6 L, but with considerable variation in the range (10, 14). The relatively reduced rate of DH in our patients likely reflects the marked resting hyperinflation. The extent of DH in COPD depends on the resting level of hyperinflation, the extent of expiratory flow limitation (EFL), the ventilation during exercise, and the breathing pattern for a given ventilation (24, 25). It follows that possible mechanisms of delay in dynamic hyperinflation during oxygen are (1) reduced ventilation for a given level of EFL; (2) altered breathing pattern (i.e., increased Te) at a given ventilation with unchanged EFL; (3) increased maximal and tidal volume-matched expiratory flows with enhanced lung emptying (i.e., bronchodilator effect); or (4) any combination of the above.

Comparison of tidal flow–volume loops at isotime during exercise showed that midtidal expiratory flow rates were diminished on oxygen compared with RA, suggesting a persistence of expiratory flow limitation during oxygen, but at lower operational lung volumes, as dictated by the reduced ventilatory demands. Previous studies have suggested a direct, but mild, bronchodilator effect of hyperoxia in COPD (11). However, the lack of a significant increase in tidal expiratory flow rates during hyperoxia makes this mechanism less likely. Thus, there was no evidence to suggest that improved dynamic airway function during exercise was responsible for the reduced operational lung volumes during hyperoxia.

It is interesting to note that ICdyn was also reduced at a given ventilation on RA and oxygen: in the absence of a bronchodilator effect, this could be explained by alterations in respiratory timing (i.e., prolonged Te) for a given ventilation during oxygen. Using multiple regression analysis, the increased ICdyn at isotime was explained primarily by reductions in breathing freqency and improved resting inspiratory reserve volume (r2 = 0.83, p < 0.0005).

Effects of Hyperoxia on Exertional Symptoms

Patients with the greatest intensity of exertional dyspnea (i.e., Borg–time slopes) on RA were those who had the greatest alleviation of dyspnea during hyperoxia. Moreover, the responses to 60% oxygen in our current study patients were much more dramatic than those achieved previously, when 60% oxygen was delivered to a group of patients with less severe COPD (FEV1.0 = 37% predicted) but with only mild exercise hypoxemia (7). Indeed, it is remarkable that relatively modest reductions in submaximal V˙e (by approximately 3 L/ min) and operational lung volumes (by approximately 0.3 L) have such clinically important effects on symptom intensity and exercise performance. Perusal of the Borg–V˙e slopes in the previously studied group of patients with less severe disease (7) and our present study subjects helped to explain the differences in the magnitude of response to 60% oxygen (Figure 6). Borg–V˙e slopes become steeper as baseline mechanical and gas exchange abnormalities increase. It follows that small reductions in submaximal ventilation, on the order of 3 L/min, result in relatively greater reductions in dyspnea in patients with more severe mechanical impairment and hypoxemia.

It is interesting to note that in the setting of less restrictive ventilatory mechanics at peak exercise on oxygen, dyspnea was displaced by leg discomfort (or some other complaint) as the primary exercise–limiting symptom in several patients. In a subgroup of five patients whose more prolonged exercise was limited by a new symptom on hyperoxia, there were significant improvements in IRV and IC throughout exercise and at peak exercise. Reductions in operational lung volumes in this severely mechanically compromised group translate into reduced elastic and threshold loading of functionally weakened inspiratory muscles, reduced muscular effort of breathing, and enhanced neuromechanical coupling of the respiratory system (10). Collectively, these mechanical changes would be expected to reduce perceived respiratory discomfort at any given work rate.

The relative importance of reduced chemoreceptor activation in contributing to dyspnea relief in this study was impossible to quantify. The extent to which hypoxia directly contributed to unpleasant respiratory sensations during exercise, independent of ventilatory muscle activity, is debated (5-8). While there is strong evidence that hypercapnia can cause unpleasant sensations of “air hunger” independent of muscle activation in healthy subjects paralyzed by neuromuscular blockade, and in ventilated quadriplegics with high cervical spine transection, direct effects of hypoxia on respiratory sensation appear to be inconsistent (26, 27). It is noteworthy that while breathing 60% oxygen, PaCO2 rose at peak exercise by an average of 7 mm Hg above an already increased resting value, reflecting a combination of respiratory depression, increased ventilation–perfusion mismatching (note increased physiological deadspace in the setting of a preserved Vt), and the Haldane effect (28). However, this was not associated with an increase in perceived dyspnea intensity at a similar ventilation, perhaps reflecting effective compensatory buffering in these patients with chronic respiratory failure. Moreover, “air hunger” was not selected as a representative qualitative descriptor of dyspnea by any of the study subjects, either under RA or hyperoxic conditions despite acute on chronic hypercapnia during exercise.

It must be emphasized that the effects of hyperoxia are multifactorial and involve many integrated mechanisms. In addition to reducing the stress on the ventilatory system, hyperoxia may (1) improve oxygen delivery to the peripheral muscles and possibly the ventilatory muscles, and thus delay fatigue; (2) improve cardiovascular function; (3) improve central nervous system function; (4) modify afferent inputs from peripheral chemoreceptors; and (5) alter the perception of symptom intensity. During 60% oxygen in this study, exercise tachycardia was consistently reduced despite patients achieving higher levels of cumulative work. In addition, 60% oxygen significantly reduced perceived leg discomfort throughout exercise. However, the relative contributions of these various physiological effects of hyperoxia to improved exercise endurance were impossible to quantify using this study design.

In summary, hyperoxia resulted in relatively large improvements in exercise endurance in patients who were severely disabled by advanced COPD. Improvement was multifactorial and ultimately reflected the integrated effects of hyperoxia on ventilatory drive, the metabolic load, and improved dynamic ventilatory mechanics, which together resulted in a delay in the attainment of critical ventilatory constraints and the attendant intolerable respiratory discomfort. The important clinical implication of our study is that in advanced COPD, modest reductions in ventilation and in operational lung volumes translate into clinically important dyspnea alleviation and enhanced exercise capabilities.

The authors acknowledge Dr. Emma Hollingworth for valuable assistance in patient recruitment and testing in this study.

Supported by the Ontario Thoracic Society. Denis O'Donnell holds a career scientist award from the Ontario Ministry of Health.

1. Stein DA, Bradley BL, Miller WMechanisms of oxygen effects on exercise in patients with chronic obstructive pulmonary disease. Chest811982610
2. Bradley BL, Garner AE, Billiu D, Mestas JM, Forma JOxygen-assisted exercise in chronic obstructive lung disease: the effect on exercise capacity and arterial blood gas tensions. Am Rev Respir Dis1181978239243
3. Bye PTP, Esau SA, Levy RO, Shiner RJ, Macklem PT, Martin JG, Pardy RLVentilatory muscle function during exercise in air and oxygen in patients with chronic air-flow limitation. Am Rev Respir Dis1321985236240
4. Dean NC, Brown JK, Himelman RB, Doherty JJ, Gold WM, Stulbarg MSOxygen may improve dyspnea and endurance in patients with chronic obstructive pulmonary disease and only mild hypoxemia. Am Rev Respir Dis1481992941945
5. Swinburn CR, Wakefield JM, Jones PWRelationship between ventilation and breathlessness during exercise in chronic obstructive airway disease is not altered by prevention of hypoxemia. Clin Sci671984146149
6. Lane R, Crockcroft A, Adams L, Guz AArterial oxygen saturation and breathlessness in patients with chronic obstructive airways disease. Clin Sci761987693698
7. O'Donnell DE, Bain DJ, Webb KAFactors contributing to relief of exertional breathlessness during hyperoxia in chronic airflow limitation. Am J Respir Crit Care Med1551997530535
8. Chronos N, Adams L, Guz AEffect of hyperoxia and hypoxia on exercise-induced breathlessness in normal subjects. Clin Sci741988531537
9. Woodcock AA, Gross ER, Geddes DMOxygen relieves breathlessness in “pink puffers.” Lanceti1981907909
10. O'Donnell DE, Webb KAExertional breathlessness in patients with chronic airflow limitation: the role of lung hyperinflation. Am Rev Respir Dis148199313511357
11. Libby DM, Briscoe WA, King TKCRelief of hypoxia-related bronchoconstriction by breathing 30% oxygen. Am Rev Respir Dis1231981171175
12. Stoller JK, Ferranti R, Feinstein ARFurther specification and evaluation of a new clinical index for dyspnea. Am Rev Respir Dis135198611291134
13. Simon PM, Schwartzstein RM, Weiss JW, Fencl V, Teghtsoonian M, Weinberger SEDistinguishable types of dyspnea in patients with shortness of breath. Am Rev Respir Dis142199010091014
14. Stubbing DG, Pengelly LD, Morse JLC, Jones NLPulmonary mechanics during exercise in subjects with chronic airflow limitation. J Appl Physiol491980511515
15. O'Donnell DE, Lam M, Webb KAMeasurement of symptoms, lung hyperinflation and endurance during exercise in chronic obstructive pulmonary disease. Am J Respir Crit Care Med158199815571565
16. Yan S, Kaminski D, Sliwinski PReliability of inspiratory capacity for estimating end-expiratory lung volume changes during exercise in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med15619975559
17. Grizzle JEThe two-period change-over design and its use in clinical trials. Biometrics211965467480
18. Morris JF, Koski A, Johnson LCSpirometric standards for healthy non-smoking adults. Am Rev Respir Dis10319715767
19. Goldman HI, Becklake MRRespiratory function tests: normal values at median altitudes and the prediction of normal results. Am Rev Tuberculosis791959457467
20. Gaensler EA, Wright GWEvaluation of respiratory impairment. Arch Environ Health121966146189
21. Hamilton AL, Killian KJ, Summers E, Jones NLMuscle strength, symptom intensity, and exercise capacity in patients with cardiorespiratory disorders. Am J Respir Crit Care Med152199520212031
22. Jones NJ. Clinical exercise testing, 3rd ed. Philadelphia: W.B. Saunders; 1988.
23. Borg GAVPsychophysical basis of perceived exertion. Med Sci Sports Exerc141982377381
24. Dodd DSBrancatisano T, Engel LA. Chest wall mechanics during exercise in patients with severe chronic airflow obstruction. Am Rev Respir Dis12919843338
25. Pride NB, Macklem PT. Lung mechanics in disease. In: AP Fishman, ed. Handbook of physiology, vol. III, sect. 3, part 2: The respiratory system. Bethesda, MD: American Physiological Society; 1986. p. 659–692.
26. Banzett RB, Lansing RW, Reid MS, Adams L, Brown R“Air hunger” arising from increased PCO2 in mechanically ventilated quadriplegics. Respir Physiol7619895367
27. Banzett RB, Lansing RW, Brown R“Air hunger” from increased PCO2 persists after complete neuromuscular block in humans. Respir Physiol811990117
28. Robinson TD, Freiberg DB, Regnis JA, Young IHThe role of hypoventilation and ventilation–perfusion redistribution in oxygen-induced hypercapnia during acute exacerbations of chronic obstructive pulmonary disease. Am J Respir Crit Care Med161200015241529
Correspondence and requests for reprints should be addressed to Denis O'Donnell, M.D., Richardson House, 102 Stuart Street, c/o Kingston General Hospital, Kingston, ON, K7L 2V7 Canada. E-mail:

Presented, in part, at the ALA/ATS International Conference, Toronto, May 5–10, 2000.


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record