In order to determine which patients with congestive heart failure (CHF) develop Cheyne-Stokes respiration (CSR) during sleep, we compared the cardiorespiratory profiles of CHF patients with CSR to those of CHF patients without CSR. Overnight polysomnography and continuous transcutaneous PCO2 (tc PCO2) monitoring, estimation of left ventricular ejection fraction (LVEF), pulmonary function tests, and chest radiograph were performed on 16 consecutive patients with chronic, stable CHF. The tc PCO2 monitor (Kontron 7640) was calibrated so that measurements reflected arterial PCO2 values. A mean value was calculated for wakefulness (W) and total sleep time (TST). Circulation time (CT) from the lung to the carotid body was estimated from the end of an apnea or voluntary breath-hold to the nadir of oxygen desaturation recorded on an ear oximeter. The duration of CSR was expressed as a percent of TST. Nine patients developed CSR during sleep (52.5 +/- 31.6 percent TST) (group 1) and 7 did not (group 2). All patients were male and both groups were a similar age (64 +/- 8 vs 63 +/- 4 years) and weight (body mass index, 28.1 +/- 3.5 vs 25.4 +/- 3.4 kg/m2). There were no significant intergroup differences between LVEF (22 +/- 5.2 vs 24.1 +/- 5.2 percent), CT (19.1 +/- 3.6 vs 15.9 +/- 6.7 s), SaO2 (W) (94 +/- 1.2 vs 92.4 +/- 2.1 percent), and SaO2 (TST) (90.8 +/- 2.7 vs 92.4 +/- 2.1 percent). The tc PCO2 (W) was lower in group 1 (34.4 +/- 3.5 vs 38.1 +/- 1.9 mm Hg), increased during sleep by a similar amount in both groups (1.6 +/- 1.5 vs 2.1 +/- 2.2 mm Hg), and was significantly lower during sleep in group 1 (36.1 +/- 3.4 vs 40.2 +/- 2.2 mm Hg). We conclude that CHF patients with CSR hyperventilate during sleep and wakefulness and that CHF patients with awake hypocapnia are more likely to develop CSR during sleep. These findings indicate that arterial PCO2 is important in determining which CHF patients develop CSR.