BACKGROUND Auto-CPAP machines used in the treatment of obstructive sleep apnoea (OSA) are designed to vary the treatment pressure automatically in order always to apply the actually needed pressure. Consequently they should be able to achieve at least identical therapeutic effects as conventional constant pressure CPAP with a lower mean treatment pressure. The present study was designed to evaluate the therapeutic efficacy and the treatment pressure of an auto-CPAP machine (REM + auto, SEFAM) in comparison with a conventional CPAP device. METHODS Following CPAP titration, 16 patients with OSA were allocated to receive conventional CPAP and auto-CPAP treatment under polysomnographic control in a randomised order. After each treatment the patients were asked to assess the therapy using a questionnaire; a vigilance test was also carried out and subjective daytime sleepiness was evaluated using the Epworth Sleepiness Scale (ESS). RESULTS The mean (SD) apnoea/hypopnoea index (AHI) during auto-CPAP treatment was comparable with that during conventional CPAP treatment (4.2 (5.1) versus 3.6 (4.0)). Neither an analysis of sleep architecture nor the arousal index (7.4 (4.1) versus 7.0 (4.3)) revealed any significant differences. Daytime sleepiness measured with the ESS was also comparable (5.3 (3.4) versus 6.5 (4.2)). The vigilance test showed normal values after both treatments in all patients with no significant differences. The mean pressure during auto-CPAP treatment (8.1 (2.9) mbar), however, was significantly higher than that employed in conventional CPAP treatment (7.6 (2.7) mbar; mean difference 0.5 mbar; 95% CI 0.1 to 0.9 mbar; p < 0.05). CONCLUSIONS Auto-CPAP was equally as effective as conventional CPAP with respect to therapeutic efficacy. The aim of reducing the treatment pressure with auto-CPAP, however, was not achieved.