We hypothesized that reductions in arterial PCO2 (PaCO2) below the apnea threshold play a key role in the pathogenesis of idiopathic central sleep apnea syndrome (ICSAS). If so, we reasoned that raising PaCO2 would abolish apneas in these patients. Accordingly, patients with ICSAS were studied overnight on four occasions during which the fraction of end-tidal CO2 and transcutaneous PCO2 were measured: during room air breathing (N1), alternating room air and CO2 breathing (N2), CO2 breathing all night (N3), and addition of dead space via a face mask all night (N4). Central apneas were invariably preceded by reductions in fraction of end-tidal CO2. Both administration of a CO2-enriched gas mixture and addition of dead space induced 1- to 3-Torr increases in transcutaneous PCO2, which virtually eliminated apneas and hypopneas; they decreased from 43.7 +/- 7.3 apneas and hypopneas/h on N1 to 5.8 +/- 0.9 apneas and hypopneas/h during N3 (P < 0.005), from 43.8 +/- 6.9 apneas and hypopneas/h during room air breathing to 5.9 +/- 2.5 apneas and hypopneas/h of sleep during CO2 inhalation during N2 (P < 0.01), and to 11.6% of the room air level while the patients were breathing through added dead space during N4 (P < 0.005). Because raising PaCO2 through two different means virtually eliminated central sleep apneas, we conclude that central apneas during sleep in ICSA are due to reductions in PaCO2 below the apnea threshold.