We undertook the present study with the following objectives: (1) to compare the difference between the end-tidal and the arterial carbondioxide concentration (P[ETa] CO2) gradients at rest and during exercise in normal subjects and patients with COPD; and (2) to analyze the factors contributing to this gradient. We studied seven normal subjects and seven patients with COPD using a symptom-limited exercise test on a cycle ergometer. Our results show that the P(ET-a)CO2 increased progressively as the individuals went from rest to higher workloads in both the normal group and in the COPD group. The P(ET-a)CO2 in patients with COPD both at rest (-3.24 +/- 2.78 mm Hg) and during exercise (1.03 +/- 2.23 mm Hg) is significantly lower than that in normal individuals at rest (1.84 +/- 3.68 mm Hg) and during exercise (10.3 +/- 6.5 mm Hg) (p < 0.01). However, the slope for the relationship between the P(ET-a)CO2 and the workload is actually significantly steeper in the patients with COPD. Although the P(ET-a)CO2 correlated significantly with the workload in both normal subjects (r = 0.63, p < 0.001) and patients (r = 0.55, p < 0.005), the P(ET-a)CO2 was much more closely correlated with the ratio of dead space to tidal volume (VD/VT) (r values of -0.86 and -0.77, respectively). Moreover, when multiple regression analysis was performed, addition of any other physiologic measure (eg, oxygen consumption [VO2], carbon dioxide production [VCO2], minute ventilation [VE], or workload) as a second independent variable after the VD/VT did not improve the correlation. This indicates that the correlation between the P(ET-a)CO2 and the workload is probably related to the dependence of the VD/VT on the workload. The PaCO2 in normal subjects and in the COPD group correlated significantly with the partial pressure of end-tidal carbon dioxide (PETCO2). Using multiple regression analysis, with the PaCO2 as the dependent variable and the PETCO2 (along with other physiologic measures) as the independent variables, we found that the standard error of the estimate was still above 2.1 mm Hg in normal subjects and in patients with COPD. We conclude that (1) during exercise, the P(ET-a)CO2 in normal subjects and in patients with COPD increases significantly, (2) the P(ET-a)CO2 gradient is more closely correlated with the VD/VT than any other physiologic variable, and (3) changes in the PETCO2 during exercise are not correlated closely with changes in the PaCO2.