To evaluate the accuracy of noninvasive estimates of VD/VT in clinical exercise testing, we compared measurements of standard VD/VT with estimates based either on end-tidal CO2 (VD/VTET) or a published estimate of arterial PCO2 (VD/VTest) at peak exercise in 68 patients. Using regression analysis, we identified highly significant differences (p < 0.001) between each method and VD/VTstand across a broad range of observed VD/VT. Assuming a normal exercise VD/VT < or = 0.30, estimate methods were specific but were insensitive (50 percent for VD/VTET and 57 percent for VD/VTest) for identifying patients with abnormal gas exchange during exercise. Separate analysis of subgroups based on resting pulmonary function did not identify any group for which either method was acceptable. Our analysis showed that errors in estimating PaCO2, which are amplified by the Bohr equation when calculating VD/VT, are responsible for the inaccuracies of each noninvasive method. We conclude that noninvasive estimates of PaCO2 cannot replace measured arterial PCO2 for calculation of VD/VT during exercise.