To determine if end-tidal carbon dioxide tension (PETCO2) is a clinically reliable indicator of arterial carbon dioxide tension (PaCO2) under conditions of heterogeneous tidal volumes and ventilation-perfusion inequality, we examined the expiratory gases of 25 postcardiotomy patients being weaned from ventilator support with intermittent mandatory ventilation. Using a computerized system that automatically sampled airway flow, pressure, and expired carbon dioxide tension, we were able to distinguish spontaneous ventilatory efforts from mechanical ventilatory efforts. The PETCO2 values varied widely from breath to breath, and the arterial to end-tidal carbon dioxide tension gradient was appreciably altered during the course of several hours. About two-thirds of the time, the PETCO2 of spontaneous breaths was greater than that of ventilator breaths during the same 70-second sample period. The most accurate indicator of PaCO2 was the maximal PETCO2 value in each sample period, the correlation coefficient being 0.768 (P less than 0.001) and the arterial to end-tidal gradient being 4.24 +/- 4.42 mm Hg (P less than 0.01 compared with all other measures). When all values from an 8-minute period were averaged, stability was significantly improved without sacrificing accuracy. We conclude that monitoring the maximal PETCO2, independent of breathing pattern, provides a clinically useful indicator of PaCO2 in postcardiotomy patients receiving intermittent mandatory ventilation.