The accuracy of transcutaneous CO2 monitoring (PtcCO2) was studied in 22 subjects suspected of having sleep-related breathing disorders, by comparison with arterial CO2 measurements (PaCO2). At rest 40 simultaneous sets of PaCO2 and PtcCO2 were obtained. The mean PaCO2 (+/-SD) was 5.3 +/- 0.9 kPa and PtcCO2 was 5.7 +/- 1.0 kPa (r = 0.79). The ventilatory response to CO2 was evaluated by a CO2 rebreathing method, and simultaneous measurements of PaCO2, PtcCO2 and end-tidal PCO2 (PETCO2) were made every min. Both PaCO2 and PETCO2 increased more during the first min of CO2 rebreathing than PtcCO2 (p less than 0.001). Between 1 to 5 min after the start of rebreathing there were no significant differences between the three methods. During sleep there was an increase in PtcCO2 (by 0.1-0.3 kPa) with each apneic event, the magnitude of the increase depending on the length and distribution of these events. With repeated long apneas there was a cumulative increase in PtcCO2, especially during REM sleep. Continuous PtcCO2 monitoring proved useful in monitoring and diagnosing sleep-related breathing disorders.