Values of end-tidal CO2 (PETCO2) approximate PaCO2 in spontaneous breathing normal subjects and in stable patients receiving mechanical ventilatory support (MVS). Because marked inequality of ventilation/perfusion ratios in critically ill patients might affect this correlation, we assessed changes of PETCO2 in predicting changes in PaCO2 (delta PaCO2) and changes in minute ventilation (delta Ve) in this patient population. Twenty consecutive intubated patients 38 to 89 yr of age (mean, 70 yr) with respiratory failure while receiving MVS with indwelling arterial lines were studied. Settings on the mechanical ventilator were varied for frequency and tidal volume, and after a minimum of 5 to 10 min equilibration, PaCO2 and PETCO2 were measured. Vt and Ve were recorded from the digital indicator of the pneumotachygraph within the mechanical ventilator and corrected for compression volume in the respirator circuit. A total of 116 simultaneous measurements were performed. PETCO2 correlated well with PaCO2 (r = 0.78, p less than 0.001). The 95% confidence interval for the mean difference in PaCO2-PETCO2 was 4.0 +/- 0.97 mm Hg. However, delta PETCO2 (measured from baseline) did not correlate as well with delta PaCO2 (r = 0.58, p = less than 0.001). In four patients, the trend in their PETCO2 during changes in mechanical ventilation were in the opposite direction from the trend in their PaCO2. Thus, many critically ill patients, who cannto be preidentified, have an inconstant PaCO2-PETCO2 gradient with changes of ventilation. Utilization of PETCO2 as a noninvasive monitoring substitute for trends in PaCO2 in critically ill patients may be misleading despite establishing an initial PaCO2-PETCO2 relationship.