We compared the prognostic power of end-tidal CO 2 pressure (PETCO 2 ) during exercise, an index of arterial CO 2 pressure, with those of established respiratory gas indexes during exercise testing in patients with left ventricular dysfunction. Seventy-eight consecutive patients with a left ventricular ejection fraction (LVEF) B40% were enrolled in the study. All the patients performed a symptom-limited incremental exercise test with respiratory gas measurements. PETCO 2 at peak exercise, peak O 2 uptake ( _ VO 2 ), the ratio of the increase in ventilation to the increase in CO 2 output ( _ VE/ _ VCO 2 slope), and the ratio of the increase in _ VO 2 to the increase in work rate (D _ VO 2 / DWR) were measured. PETCO 2 at peak exercise was significantly correlated with peak _ VO 2 , _ VE/ _ VCO 2 slope and D _ VO 2 /DWR. During a prospective follow-up period of 992 ± 570 days, 14 cardiac deaths occurred. As compared to survivors, non-survivors had a significantly lower LVEF, lower PETCO 2 at peak exercise, lower peak _ VO 2 , lower D _ VO 2 /DWR and a higher _ VE/ _ VCO 2 slope. Among these indexes, only PETCO 2 at peak exercise was found to be an independent predictor for cardiac death. PETCO 2 at peak exercise is useful in predicting poor prognosis in patients with left ventricular systolic dysfunction.