American Journal of Respiratory and Critical Care Medicine

In severe chronic obstructive pulmonary disease (COPD), carbon dioxide retention during exercise is highly variable and is poorly predicted by resting pulmonary function and arterial blood gases or by tests of ventilatory control. We reasoned that in patients with compromised gas exchange capabilities, exercise hypercapnia could be explained, in part, by the restrictive consequences of dynamic lung hyperinflation. We studied 20 stable patients with COPD (FEV1 = 34 ± 3 percent predicted; mean ± SEM) with varying gas exchange abnormalities (PaO2 range, 35 to 84 mm Hg; PaCO2 range, 31 to 64 mm Hg). During symptom-limited maximum cycle exercise breathing room air, PaCO2 increased 7 ± 1 mm Hg (p < 0.05) from rest to peak exercise (range, −6 to 25 mm Hg). We measured the change in PaCO2 after hyperoxic breathing at rest as an indirect test of ventilation–perfusion abnormalities. The change in PaCO2 from rest to peak exercise correlated best with the acute change in PaCO2 during hyperoxia at rest (r2 = 0.62, p < 0.0005) and with resting arterial oxygen saturation (r2 = 0.30, p = 0.011). During exercise, the strongest correlates of serial changes in PaCO2 from rest included concurrent changes in end-expiratory lung volume expressed as a percentage of total lung capacity (partial correlation coefficient [r] = 0.562, p < 0.0005) and oxygen saturation (partial r = 0.816, p < 0.0005). In severe COPD, the propensity to develop carbon dioxide retention during exercise reflects marked ventilatory constraints as a result of lung hyperinflation as well as reduced gas exchange capabilities.

Previous studies have shown that in patients with chronic obstructive pulmonary disease (COPD), the development of hypercapnia during exercise cannot reliably be predicted from resting pulmonary function tests or arterial blood gas measurements (14). Ultimately, exercise hypercapnia must reflect integrated alterations in minute ventilation (V·e), physiologic dead space (Vd/Vt), and carbon dioxide output (V·co2), as dictated by the mass balance equation for alveolar ventilation (V·a). The relative importance of these key factors likely varies among individuals. Reduced V·e may reflect reduced central respiratory drive or, alternatively, mechanical limitation and/or inspiratory muscle dysfunction. Recent studies have provided evidence that respiratory drive is preserved or actually amplified in hypercapnic COPD patients (57), and there is little convincing evidence to suggest that acute inspiratory muscle weakness or fatigue is an important contributor to exercise hypercapnia (1, 8). No study to date has examined the effects of mechanical factors such as lung hyperinflation on CO2 retention during exercise. Resting and dynamic hyperinflation influence breathing patterns, inspiratory muscle function, and neuroregulatory control during exercise and, therefore, are relevant to the understanding of CO2 retention. We postulated that CO2 retention during exercise occurs in patients with COPD who have higher ventilatory demands as a result of greater ventilation–perfusion (·V/Q·) abnormalities, in the setting of earlier mechanical limitation of ventilation due to lung hyperinflation. The demonstration of a close correlation between increases in PaCO2 and end-expiratory lung volume (EELV) during exercise would lend support to this hypothesis.

A normal resting PaCO2 can obscure extensive ·V/Q· inhomogeneity in patients with COPD. For example, in some hypoxemic patients, resting PaCO2 (on room air) may become “normalized” by compensatory strategies such as regional hypoxic vasoconstriction and/or increased peripheral chemosensitivity (i.e., increased hypoxic drive). Together, these adaptations serve to minimize ·V/Q· inequalities and to preserve V·a and PaCO2 in the normal range. Based on a previous study by Sassoon and colleagues (9), using hyperoxia in clinically stable patients with hypoxia, intrapulmonary vasoregulatory adjustments are likely to be important in maintaining resting eucapnia (9). We therefore used the hyperoxia test to evaluate indirectly the extent of ·V/Q· inequalities present at rest in our patients. We reasoned that although compensatory mechanisms (e.g., hypoxic vasoconstriction) can ensure satisfactory CO2 elimination at rest, these are quickly overwhelmed during exercise in the setting of a rapid mechanical restriction of ventilation and increased CO2 production. Therefore, we wished to determine whether those who develop CO2 retention at rest during hyperoxia (caused by release of hypoxic vasoconstriction) were more likely to develop hypercapnia during exercise.

We studied 20 patients with clinically stable, severe COPD who exhibited a range of gas exchange abnormalities at rest and during exercise. To determine whether mechanical abnormalities were associated with CO2 retention, we studied the physiologic correlates of CO2 retention during both rest and exercise. Finally, we compared ventilatory responses to exercise in subgroups of CO2 retainers and nonretainers.


We studied 20 clinically stable patients with advanced COPD (FEV1 < 60% predicted) and varying degrees of hypoxemia. All patients were ventilatory limited during exercise and experienced severe activity-related breathlessness with a modified Baseline Dyspnea Index focal score (10) of six or less. Exclusions included clinical evidence of significant cardiovascular disease, other pulmonary disease, including cor pulmonale, and other disorders that could contribute to dyspnea or exercise limitation.

Study Design

The study was a randomized, blinded, placebo-controlled, crossover trial with local university/hospital research ethics approval. After giving written informed consent, patients were familiarized with testing procedures and completed a symptom-limited incremental exercise test. In a subsequent visit, subjects performed two constant-load cycle exercise tests. The order of tests was randomized so that subjects breathed either 60% O2 or room air (21% O2) for at least 10 minutes at rest before exercising at approximately 50% of their previously determined maximal work rate while breathing the same oxygen mixture. There was a 60- to 90-minute washout period after testing on one oxygen mixture and then a crossover to the alternate mixture for identical testing. Subjects were blinded to the oxygen concentration being breathed.

In this article, we only consider the resting data during room air and O2 and the exercise tests on room air. Changes in exercise responses with hyperoxia will be discussed in a subsequent article.


Subjects performed pulmonary function tests as previously described (2). Measurements during rest and symptom-limited cycle exercise tests were performed as previously described (2), with patients breathing either room air or 60% O2 on demand from a 200-L Douglas bag reservoir. In addition to the usual measurements collected during cardiopulmonary exercise testing, subjects rated their intensity of dyspnea (defined as breathing discomfort) and perceived leg discomfort using the modified Borg Scale (11); operating lung volumes were derived from dynamic inspiratory capacity (IC) measurements (12), and arterial blood gas samples were obtained from the radial artery (2).

Statistical Analysis

Results are reported as means ± SEM. A statistical significance of 0.05 was used for all analyses. Subgroup comparisons were made using unpaired t tests. Comparisons between measurements on room air and 60% O2 were made using paired t tests. For subgroup comparisons and regression analyses, pulmonary function measurements were standardized as a percentage of predicted normal values (1317). Pearson correlations were used to establish associations between the change (Δ) in PaCO2 from rest to peak exercise and concurrent changes in relevant independent variables, which included V·e, V·e/V·co2, V·a, V·co2, breathing frequency, inspiratory time, expiratory time, tidal volume (Vt), IC, inspiratory reserve volume, EELV, end-inspiratory lung volume (EILV), oxygen saturation, PaO2, and Vd/Vt. Forward stepwise multiple regression analysis was performed with these variables and relevant covariates (baseline lung function, gas exchange measurements, and the ΔPaCO2 response to the hyperoxic test at rest) to establish the best predictive equation for ΔPaCO2 from rest to peak exercise. Similar analyses were performed to examine inter-relationships between (1) hyperoxia-induced ΔPaCO2 at rest and concurrent changes in a similar pool of candidate variables and (2) resting PaCO2, baseline lung function, and gas exchange measurements.

Finally, to determine factors associated with progressive hypercapnia during exercise, we used a multivariable linear model with the change (Δ) in PaCO2 from rest as the dependent variable and concurrent changes in relevant independent variables (similar to those described previously here). Because of dependency between serial measurements within subjects, “subjects” (categorical variable) were treated as random effects in the regression model.


Subjects with COPD included 10 males and 10 females with severe expiratory airflow obstruction, lung hyperinflation, functional inspiratory muscle weakness (decreased inspiratory muscle strength) and reduction of diffusing capacity of the lung for carbon monoxide (DlCO) (Table 1)

TABLE 1. Subject characteristics (n = 20)



Age, yr 67 ± 2
Height, cm 165 ± 3
Weight, kg67.8 ± 6.6
Body mass index, kg/m224.8 ± 2.1
Modified Baseline Dyspnea Index4.9 ± 0.4Severe
Pulmonary function (percentage
     predicted normal)
FEV1, L0.78 ± 0.08(34)
FVC, L1.88 ± 0.15(58)
FEV1/FVC, % 41 ± 2(59)
FEV50%, L/s0.35 ± 0.04(9)
TLC, L6.84 ± 0.32(122)
RV, L4.74 ± 0.32(219)
FRC, L5.47 ± 0.31(179)
IC, L1.37 ± 0.10(53)
PImax, cm H2O 43 ± 3(58)
DlCO, ml/min/mm Hg7.7 ± 0.8(41)
DlCO/Va, min/mm Hg2.53 ± 0.28(68)
Symptom-limited peak exercise
Dyspnea, Borg Scale5.0 ± 0.4Severe
Leg discomfort, Borg Scale4.0 ± 0.5Somewhat severe
SaO2, % 83 ± 3
Heart rate, beats/min 114 ± 4
V·o2, ml/kg/min7.8 ± 0.9
V·entilation, L/min25.8 ± 3.5
0.34 ± 0.03

Definition of abbreviations: DlCO = diffusing capacity of the lung for carbon monoxide; FRC = plethysmographic function residual capacity; IC = inspiratory capacity; IRV = inspiratory reserve volume; PImax = inspiratory muscle strength; RV = residual volume; SaO2 = oxygen saturation; TLC = total lung capacity; Va = alveolar volume; V·o2 = oxygen consumption.

. There were varying levels of gas exchange abnormalities at rest on room air; that is, PaO2 ranged from 35 to 84 mm Hg, and PaCO2 ranged from 31 to 64 mm Hg. All subjects had “severe” exertional dyspnea (peak Borg Scale ratings = 5.0 ± 0.4), with 14 out of 20 patients stopping exercise because of breathing discomfort, three because of a combination of breathing and leg discomfort, two because of leg discomfort, and one because of discomfort with the mouthpiece and nose clips. Exercise capacity was severely reduced (symptom-limited peak oxygen consumption [V·o2] = 24 ± 5% predicted maximum) because of ventilatory limitation. All patients showed significant expiratory flow limitation with a limited ability to increase ventilation at the end of exercise; that is, tidal expiratory flows overlapped the maximal expiratory flow-volume envelope, and EILV closely approached the TLC envelope.

Exercise-induced Hypercapnia

The magnitude of change (Δ) in PaCO2 from rest to the end of symptom-limited exercise while breathing room air was variable (range, −6 to 25 mm Hg). One subject showed a fall in PaCO2 (−6 mm Hg). One subject had no change in PaCO2 (1 mm Hg). Nine subjects experienced small increases in PaCO2 of between 3 and 5 mm Hg, and the remaining nine subjects increased PaCO2 by greater than 5 mm Hg (i.e., by a mean of 11.6 ± 2.2 mm Hg). The change in PaCO2 from rest to peak exercise did not correlate with the resting PaCO2 (r = −0.03, p = 0.91). Likewise, the magnitude of change in PaCO2 from rest to peak exercise was similar in a subgroup of patients with resting hypercapnia (PaCO2 ⩾ 45 mm Hg, n = 7) as in the patients with resting eucapnia (PaCO2 < 45 mm Hg, n = 13). PaCO2 increased by 7 ± 3 and 6 ± 2 mm Hg, respectively (p = 0.77). However, compared with patients with resting eucapnia, the subgroup of individuals with resting hypercapnia had greater resting hyperinflation (FRC was 206 ± 15% versus 165 ± 7% predicted, p = 0.01), a limited ability to increase ventilation during exercise (peak V·e was 19 ± 2 versus 29 ± 5 L/minute), greater Vd/Vt at rest and throughout exercise (peak Vd/Vt was 47 ± 2 versus 40 ± 2%, p < 0.05), and greater PaCO2 values throughout exercise (peak PaCO2 was 58 versus 44 mm Hg, p = 0.001). Likewise, resting PaCO2 correlated best with the baseline residual volume (RV) percentage predicted (r = 0.725, p = 0.001) and FRC expressed as either percent predicted (r = 0.666, p = 0.001) or as percentage TLC (r = 0.659, p = 0.001), but did not correlate with baseline Vd/Vt, V·e, V·e/V·co2, or functional inspiratory muscle strength.

The best correlate of the magnitude of change (Δ) in PaCO2 from rest to peak exercise was the hyperoxia-induced change in resting PaCO2 (r = 0.79, p < 0.0005). Those with the greatest hyperoxia-induced hypercapnia had the greatest exercise-induced hypercapnia. The ΔPaCO2 from rest to peak exercise did not correlate with any resting measurements of pulmonary function or gas exchange other than FEF25–75% (expressed percentage predicted; r = −0.58, p = 0.007) and oxygen saturation on room air (r = −0.55, p = 0.011). These relationships were maintained even after removing the one outlier from the analysis; this patient had developed significantly greater hypercapnia during both exercise (25 mm Hg) and hyperoxic breathing (27 mm Hg) than the rest of the sample.

After accounting for serial measurements within each subject, the strongest correlates of serial ΔPaCO2 measurements during exercise included concurrent measurements of change in oxygen saturation (partial correlation coefficient [R] = 0.816, p < 0.0005) and ΔEELV expressed as a percentage of TLC (partial r = 0.562, p < 0.0005) or as a percentage of predicted TLC (partial r = 0.562, p < 0.0005) (Figure 1)


Hyperoxia-induced Hypercapnia

After breathing the 60% oxygen mixture for 10 minutes at rest, there were two distinct PaCO2 response patterns (Figure 2)

: (1) “Retainers” (n = 9) significantly increased resting PaCO2 by at least 3.6 mm Hg (i.e, outside the 95th confidence interval of the PaCO2 measured on room air) during hyperoxia, and (2) “nonretainers” (n = 11) had no significant change, or a fall, in PaCO2 during hyperoxia (i.e., ⩽ 3 mm Hg change). These subgroups were well matched with respect to sex, body mass index, baseline levels of airflow obstruction, resting ventilatory mechanics, and resting PaCO2 on room air (Table 2)

TABLE 2. Chronic obstructive pulmonary disease subgroups: hyperoxia-induced carbon dioxide retainers versus nonretainers

CO2 Retainers
 (n = 9)

 (n = 11)
FEV1, % predicted34 ± 535 ± 4
FVC, % predicted58 ± 756 ± 5
FEV1/FVC, % predicted56 ± 462 ± 3
TLC, % predicted 126 ± 7 118 ± 4
FRC, % predicted188 ± 16 172 ± 8
IC, % predicted52 ± 554 ± 3
PImax, % predicted55 ± 360 ± 7
DlCO, % predicted46 ± 437 ± 4
Steady-state resting measurements (room air)
PaO2, mm Hg 57.8 ± 4.7*69.9 ± 3.2
PaCO2, mm Hg43.2 ± 3.541.6 ± 1.4
Vd/Vt, %44 ± 441 ± 3
V·a/V·co2, %34 ± 233 ± 2
V·e, L/min10.7 ± 1.011.9 ± 1.6
Vt, % predicted VC20 ± 317 ± 2
F, breaths/min18.4 ± 1.620.8 ± 1.2
Symptom-limited peak exercise (room air)
Dyspnea intensity, Borg 5.2 ± 0.8 4.7 ± 0.4
Heart rate, beats/min 113 ± 6 115 ± 4
V·o2, ml/kg/min 7.7 ± 1.2 7.8 ± 1.3
V·co2, ml/kg/min 8.0 ± 1.3 8.8 ± 1.1
Hydrogen ion, mEq/L45.3 ± 1.342.0 ± 2.0
PaCO2, mm Hg53.7 ± 3.9 (p = 0.06)45.4 ± 2.0
  ΔPaCO2, from rest, mm Hg 10.5 ± 2.4* 3.8 ± 1.2
SaO2, % 77 ± 5*89 ± 2
  ΔSaO2 from rest, % −10 ± 2* −5 ± 1
PaO2, mm Hg 46.8 ± 3.5*60.8 ± 3.9
Vd/Vt, %44 ± 341 ± 2
V·e, L/min22.0 ± 3.628.9 ± 3.9
F, breaths/min28.6 ± 2.230.4 ± 2.7
Ti/Ttot 0.33 ± 0.02 0.37 ± 0.02
Vt, % predicted VC25 ± 427 ± 3
IC, % predicted46 ± 648 ± 5
  ΔIC rest to peak, L−0.30 ± 0.11−0.22 ± 0.07
95 ± 1
95 ± 1

*p < 0.05 significant difference between subgroups.

Definition of abbreviations: CO2 = carbon dioxide; DlCO = diffusing capacity of the lung for carbon monoxide; EILV = end-inspiratory lung volume; IC = inspiratory capacity; PImax = functional inspiratory muscle strength; SaO2 = oxygen saturation;Ti/Ttot = inspiratory duty cycle; TLC = total lung capacity; V·a = alveolar ventilation; Vd/Vt = physiologic dead space; V·e = minute ventilation; V·co2 = carbon dioxide output; V·o2 = oxygen consumption.

; however, the mean PaO2 measured at rest on room air was significantly less in the retainers group (58 ± 5 mm Hg) than the nonretainer group (70 ± 3 mm Hg, p < 0.05). In the CO2 retaining subgroup, seven out of the nine subjects were on long-term home oxygen. Four of the 11 patients who did not retain CO2 were on home oxygen.

Mean resting changes in gas exchange, breathing pattern, and various other cardioventilatory parameters in response to hyperoxia are reported in Table 3

TABLE 3. Steady-state resting breathing: changes in response to hyperoxia

CO2 Retainers

PaCO2, mm Hg 10 ± 2* 0 ± 1
PaO2, mm Hg174 ± 19* 175 ± 9*
SaO2, %10 ± 2* 7 ± 1*
HCO3, mM 4.2 ± 2.4–1.1 ± 0.5
H+, mM 2.4 ± 0.9 2.1 ± 0.3*
pH −0.03 ± 0.01* −0.02 ± 0.03*
Vd/Vt, % 10 ± 3* 5 ± 4
V·co2, ml/kg/min −1.1 ± 0.6 −0.5 ± 0.4
V·a, L/min −2.1 ± 0.9 −0.7 ± 0.6
V·e, L/min −1.9 ± 1.5 −0.7 ± 0.8
F, breaths/min −0.6 ± 0.8 −0.5 ± 0.9
Ti, S 0.00 ± 0.13 0.09 ± 0.07
Te, S 0.11 ± 0.26 0.07 ± 0.16
Vt, L −0.09 ± 0.09 −0.03 ± 0.04
EELV, % TLC0 ± 1−1 ± 1
IC, L0.03 ± 0.09 0.10 ± 0.08
IRV, L0.12 ± 0.12 0.13 ± 0.09
Heart rate, beats/min
 −6 ± 3
 −3 ± 2

*p < 0.01.

p < 0.05, significant change from room air.

p = 0.00, significant difference betwen subgroup responses.

Definition of abbreviations: CO2 = carbon dioxide; EELV = end-expiratory lung volume; HCO3 = bicarbonate; IC = inspiratory capacity; IRV = inspiratory reserve volume; SaO2 = oxygen daturation; Ti = inspiratory time; TLC = total lung capacity; V·a = alveolar ventilation; V·co2 = carbon dioxide output; Vd/Vt = physiologic dead space; V·e = minute ventilation; Vt = tidal volume.

. Aside from the differences in PaCO2 responses to hyperoxia, Vd/Vt increased significantly (p < 0.01), and V·a decreased significantly (p < 0.05) in the retainer subgroup but did not change in the nonretainer subgroup. The decrease in V·a was directly proportional to the decrease in V·co2 (r = 0.89, p < 0.0005), as was the change in V·e (r = 0.82, p < 0.0005). In addition, the arterial-end tidal difference in Pco2 increased in response to O2 in retainers but did not change in nonretainers (p = 0.058).

The hyperoxia-induced change in resting PaCO2 correlated significantly with resting PaO2 on room air (r = −0.50, p = 0.04), baseline FEF75% (r = −0.48, p= 0.33), and the concurrent changes in V·a (r = −0.45, p = 0.05) and Vd/Vt (r = 0.44, p = 0.05). After accounting for the baseline PaO2, the combination of FEF75% and the change in Vd/Vt best predicted the oxygen-induced change in PaCO2 (r2 = 0.64, p = 0.001).

Exercise Responses in CO2 Retainers Versus Nonretainers

To illustrate differences in exercise response patterns between CO2 retainers and nonretainers, subgroups were divided on the basis of the PaCO2 response to hyperoxia (discussed previously here) (18). Aside from gas exchange parameters, variables measured at rest and at peak exercise were similar for both subgroups while breathing room air (Table 2). Gas exchange and ventilatory responses are shown relative to the metabolic load (V·co2) in Figure 3

. PaCO2 responses during exercise were clearly different within subgroups. The arterial-end tidal difference in Pco2 response to exercise was also different between subgroups, widening by 7 ± 4 mm Hg (p = 0.09) from rest to peak exercise in CO2 retainers, but not changing (−1 ± 2 mm Hg) in nonretainers. Nevertheless, slopes of exertional dyspnea intensity over V·co2 were not significantly different between subgroups (Figure 4) .

In both subgroups, exercise was ventilatory limited due to mechanical constraints on tidal volume expansion, that is, Vt approached the dynamic IC as EILV encroached on the TLC envelope. Peak Vt correlated very strongly with the concurrent IC (r = 0.97, p < 0.0005). However, CO2 retainers reached this mechanical limitation earlier and tended to have a lower peak V·e and V·co2 than CO2 nonretainers (Figure 5)

. Peak symptom-limited V·co2 (ml/kg/min) correlated best with parameters reflecting dynamic lung hyperinflation and the mechanical constraints on Vt expansion at the end of exercise (EELV/TLC, r = −0.77, p < 0.0005; IC percentage predicted, r = 0.71, p < 0.0005; EILV/TLC, r = −0.64, p = 0.002; inspiratory reserve volume/predicted TLC, r = 0.59, p = 0.006) and with the peak Vt attained (r = 0.68, p = 0.001).

The novel findings of this study are as follows: First, patients who developed hypercapnia during hyperoxic breathing at rest were more likely to develop significant CO2 retention with exercise. Second, PaCO2 at rest correlated well with measures of resting lung hyperinflation, whereas changes in PaCO2 during exercise correlated best with the level of arterial oxygen desaturation and with concomitant increases in dynamic EELV/TLC. The subgroup of patients who retained CO2 in response to hyperoxic breathing at rest demonstrated greater ·V/Q· inequalities and greater mechanical constraints on ventilation for a given V·co2 during exercise than the subgroup of patients who did not retain CO2.

Correlation between Hyperoxic and Exercise Hypercapnia

This is the first study to show a strong correlation between the hypercapnic response to hyperoxia and the change in PaCO2 from rest to peak exercise. In chronically hypoxemic patients with COPD, resting eucapnia can be maintained by the development of adaptive mechanisms, which include (1) regional intrapulmonary vasoregulatory adjustments to minimize ·V/Q· inequalities (i.e., hypoxic vasoconstriction) and (2) increased respiratory drive as the result of increased peripheral chemosensitivity (i.e., hypoxic drive). That these adaptations were present in many of our patients is evidenced by their response to hyperoxic breathing. In agreement with the study of Sassoon and colleagues (9), which was conducted in a similar group of clinically stable patients, hyperoxia caused release of hypoxic vasoconstriction with worsening of ·V/Q· mismatching, expansion of dead space ventilation, and reduction of CO2 elimination. Our study similarly showed that the change in PaCO2 while breathing 60% oxygen was mainly explained by the induced increase in Vd/Vt, as V·e and V·co2 (other determinants of PaCO2) fell in proportion and by a small amount. Both studies suggest that CO2 retention during hyperoxia primarily reflects worsening ·V/Q· relationships rather than suppression of hypoxic drive per se. The hyperoxic test is, therefore, an indirect measure of the extent of prevailing ·V/Q· inequalities at rest. Previously, studies using multiple inert gas elimination techniques have demonstrated that when hyperoxia caused release of vasoconstriction in previously underperfused and poorly ventilated alveolar units, blood was diverted from alveolar units with relatively well preserved ·V/Q· ratios (18). These latter units became converted into units with high ·V/Q· ratios, which contributed to increased Vd/Vt. In addition, it has been proposed that increased oxygenation of low ·V/Q· units will increase PaCO2 via the Haldane effect (19). This effect was not measured in this study but is likely to be small (19, 20).

In hypoxic healthy individuals living at high altitude, acute hyperoxia may result in a small decrease in V·e (21). During hyperoxia in our group of COPD patients, ventilation fell by a mean of 3% (and by 8% in the subgroup retaining CO2), likely reflecting suppression of hypoxic drive. The inability to augment further ventilation could not be explained by mechanical constraints, as our CO2 retainers were capable of increasing ventilation, at least to an average peak of 22 L/minute during exercise. It is possible that reduced sensitivity of the central respiratory controller may partly explain CO2 retention during both hyperoxia and exercise.

Seven of the nine patients in the CO2 subgroup were receiving long-term continuous home oxygen. Fleetham and colleagues clearly demonstrated a measurable decline in the ventilatory response to added CO2 in such individuals after 12 months of receiving home oxygen therapy (22). However, no definitive conclusions could be made about ventilatory control status in our patients as we do not know to what extent suppression of hyperoxic drive during hyperoxia counterbalanced the stimulatory effects of the increased CO2. Any assessment of ventilatory control during exercise is extremely difficult in the setting of coexistent severe mechanical constraints in COPD patients. Therefore, the relative contribution of reduced central respiratory drive to exercise CO2 retention could not be ascertained in this study.

Clearly, intrapulmonary vasoregulation was an important mechanism responsible for preventing resting hypercapnia on room air in these hypoxic COPD patients. We postulate that this compensatory mechanism was quickly overwhelmed during exercise in the setting of increased V·co2, worsening ·V/Q· relations (i.e., lack of the usual improvement in ·V/Q· relations), and critical dynamic mechanical constraints on ventilation.

Correlates of PaCO2 during Exercise

All of our patients had advanced COPD with clinical and physiological features of emphysema. They had severe resting hyperinflation with further dynamic hyperinflation during exercise. At a peak symptom-limited V·o2 of only 24% predicted, IC decreased by 17% of its already reduced resting value. As in a previous study (12), symptom-limited peak V·co2 correlated well with peak Vt, which in turn correlated strongly with both the resting and peak exercise IC (expressed as the percentage predicted). This indicates that in the group as a whole, mechanical factors formed the predominant limitation to exercise.

There was a good correlation between resting PaCO2 and measures of resting hyperinflation, suggesting that mechanical factors play a role in CO2 retention in patients with the worst ·V/Q· inequalities. In addition, this is the first study to show a strong statistical correlation between exercise-induced increases in PaCO2 and simultaneous increases in dynamic EELV/TLC in patients with advanced COPD (Figure 1). In fact, as much as 56% of the variance in exercise ΔPaCO2 was explained by this mechanical factor alone. The strong association between exercise arterial oxygen desaturation and exercise-induced changes in PaCO2 was not surprising, as the progressive arterial oxygen desaturation primarily reflects the effects of worsening alveolar hypoventilation as a result of dynamic mechanical restriction. It is noteworthy that other important exercise variables such as Vd/Vt, breathing pattern, and V·co2 did not additionally contribute to the variance in exercise-induced hypercapnia in the group as a whole.

Exercise Responses in CO2 Retainers and Nonretainers

In accordance with a previous study by Robinson and colleagues (18), subgroups were arbitrarily divided on the basis of their PaCO2 response to hyperoxic breathing at rest. This approach permitted a clear demarcation of the subgroups (Figures 2 and 3). Hyperoxic CO2 retainers had greater mechanical constraints on ventilation than nonretainers and reached their maximal EILV of 95% of the TLC at a peak V·e of only 22 L/minute. This peak V·e was substantially less (by an average of 7 L/minute) than that of nonretainers at a similar peak EILV. The lower resting PaO2, greater hyperoxic hypercapnia, and more pronounced arterial oxygen desaturation with a widened arterial-end tidal difference in Pco2 difference during exercise also indicates greater ·V/Q· inequalities in those who developed exercise hypercapnia. In keeping with previous studies (1), breathing pattern responses to exercise were almost identical in these subgroups. There appeared to be a greater disparity between ventilatory capacity and ventilatory demand in the CO2 retainer subgroup at any given V·co2. Clearly, this subgroup of patients was incapable of increasing ventilation further despite marked ventilatory stimulation as a result of the combination of profound hypercapnia, hypoxia, and acute metabolic acidosis, as well as stimulatory inputs from peripheral ergoreceptors and the cortical “locomotor” center.


In COPD patients with chronic hypoxia, a near-normal resting PaCO2 on room air does not preclude the existence of extensive ·V/Q· abnormalities. Such abnormalities can be uncovered by a hyperoxic breathing test, which acutely reverses the adaptive vasoregulatory adjustments that normally help to maintain resting eucapnia. These patients who have more compromised gas exchange capabilities are more likely to develop hypercapnia when challenged with increasing V·co2 and progressive mechanical restriction during exercise. In advanced COPD, the propensity to develop CO2 retention during exercise in patients with marked ·V/Q· inequalities primarily reflects the severe mechanical constraints on ventilation as a consequence of lung hyperinflation.


The authors would like to thank Dr. Miu Lam (Community Health and Epidemiology, Queen's University) for his assistance with the statistical analysis.

Supported by the Ontario Thoracic Society. A career scientist award is held from the Ontario Ministry of Health (D. O.).

1. Montes de Oca M, Celli BR. Respiratory muscle recruitment and exercise performance in eucapnic and hypercapnic severe chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;161:880–885.
2. O'Donnell DE, Bain DJ, Webb KA. Factors contributing to relief of exertional breathlessness during hyperoxia in chronic airflow limitation. Am J Respir Crit Care Med 1997;155:530–535.
3. Gimenez M, Servera E, Candina R, Mohan Kumar T, Bonnassis JB. Hypercapnia during maximal exercise in patients with chronic airflow obstruction. Bull Eur Physiopathol Respir 1984;20:113–119.
4. O'Donnell DE, D'Arsigny C, Webb KA. Effects of hyperoxia on ventilatory limitation during exercise in advanced COPD. Am J Respir Crit Care Med 2001;163:892–898.
5. Montes de Oca M, Celli BR. Mouth occlusion pressure, CO2 response and hypercapnia in severe obstructive pulmonary disease. Eur Respir J 1998;12:666–671.
6. Scano G, Spinelli A, Duranti R, Gorini M, Gigliotti F, Goti P, Milic-Emili J. Carbon dioxide responsiveness in COPD patients with and without hypercapnia. Eur Respir J 1995;8:78–85.
7. DeTroyer A, Leeper JB, McKenzie D, Gandevia S. Neural drive to the diaphragm in patients with severe COPD. Am J Respir Crit Care Med 1997;155:1335–1340.
8. Mador MJ, Kufel TJ, Pineda LA, Sharma GK. Diaphragmatic fatigue and high intensity exercise in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;161:118–123.
9. Sassoon CSH, Hassell KT, Mahutte CK. Hyperoxic-induced hypercapnia in stable chronic obstructive pulmonary disease. Am Rev Respir Dis 1987;135:907–911.
10. Stoller JK, Ferranti R, Feinstein AR. Further specification and evaluation of a new clinical index for dyspnea. Am Rev Respir Dis 1986;135: 1129–1134.
11. Borg GAV. Psychophysical basis of perceived exertion. Med Sci Sports Exerc 1982;14:377–381.
12. O'Donnell DE, Revill SM, Webb KA. Dynamic hyperinflation and exercise intolerance in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2001;164:770–777.
13. Morris JF, Koski A, Temple WP, Claremont A, Thomas DR. Fifteen-year interval spirometric evaluation of the Oregon predictive equations. Chest 1988;92:123–127.
14. Knudson RJ, Lebowitz MD, Holberg CJ, Burrows B. Changes in the normal maximal expiratory flow-volume curve with growth and aging. Am Rev Respir Dis 1983;127:725–734.
15. Crapo RO, Morris AH, Clayton PD, Nixon CR. Lung volumes in healthy nonsmoking adults. Bull Eur Physiopathol Respir Dis 1982;18:419–425.
16. Burrows B, Kasik JE, Niden AH, Barclay WR. Clinical usefulness of the single-breath pulmonary diffusing capacity test. Am Rev Respir Dis 1961;84:789–806.
17. Hamilton AL, Killian KJ, Summers E, Jones NL. Muscle strength, symptom intensity, and exercise capacity in patients with cardiorespiratory disorders. Am J Respir Crit Care Med 1995;152:2021–2031.
18. Robinson TD, Freiberg DB, Regnis JA, Young IH. The roles of hypoventilation and ventilation-perfusion redistribution in oxygen-induced hypercapnia during acute exacerbations of chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;161:1524–1529.
19. Lenfant C. Arterial-alveolar difference in PCO2 during air and oxygen breathing. J Appl Physiol 1966;21:1356–1362.
20. Dick CR, Sassoon CSH, Berry RB, Mahutte CK. O2 induced change in ventilatory drive in COPD. Am J Respir Crit Care Med 1997;155:609–614.
21. Lefrancois R, Gautier H, Pasquis S. Ventilatory oxygen drive in acute and chronic hypoxia. Respir Physiol 1968;4:217–228.
22. Fleetham JA, Bradley CA, Kryger MH, Anthonisen NR. The effect of low flow oxygen therapy on chemical control of ventilation in patients with hypoxemic COPD. Am Rev Respir Dis 1980;122:833–840.
Correspondence and requests for reprints should be addressed to Dr. Denis O'Donnell, Richardson House, 102 Stuart Street, Kingston, ON, K7L 2V6 Canada. E-mail:


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