To understand the pathogenesis of central sleep apnea (CSA) in patients with congestive heart failure (CHF), we measured the end-tidal carbon dioxide pressure (PET(CO2)) during spontaneous breathing, the apnea-hypopnea threshold for CO2, and then calculated the difference between these two measurements in 19 stable patients with CHF with (12 patients) or without (7 patients) CSA during non-rapid eye movement sleep. Pressure support ventilation was used to reduce the PET(CO2) and thereby determine the thresholds. In patients with CSA, 1.5-3% CO2 was supplied temporarily to stabilize breathing before determining the thresholds. Unlike patients without CSA whose eupneic PET(CO2) increased during sleep (37.7 +/- 1.4 mm Hg versus 40.2 +/- 1.5 mm Hg, p < 0.01), patients with CSA showed no rise in PET(CO2) from wakefulness to sleep (37.5 +/- 0.9 mm Hg versus 38.2 +/- 1.0 mm Hg, p = 0.2). Patients with CHF and CSA had their eupneic PET(CO2) closer to the threshold PET(CO2) than patients without CSA (DeltaPET(CO2) [eupneic PET(CO2) - threshold PET(CO2)] was 2.8 +/- 0.3 mm Hg versus 5.1 +/- 0.7 mm Hg for apnea, p < 0.01; 1.7 +/- 0.7 versus 4.1 +/- 0.5 mm Hg for hypopnea, p < 0.05). In summary, patients with CHF and CSA neither increase their eupneic PET(CO2) during sleep nor proportionally decrease their apnea-hypopnea threshold. The resultant narrowed DeltaPET(CO2) predisposes the patient to the development of apnea and subsequent breathing instability.