We investigated whether cephalometric measurements, nocturnal indices of negative intrathoracic pressure, or the frequency of sleep-related breathing disorders were related to the level of effective continuous positive airway pressure (CPAP) in patients with obstructive sleep apnea (OSA). We examined 22 OSA patients who underwent two consecutive polysomnographic recordings, the first for diagnosis and the second for CPAP titration. Cephalometric measurements, spirometric data, and blood-gas analysis results were available for all subjects. In the diagnostic polysomnography, at least 30 apneas were analyzed during non-rapid-eye-movement (NREM) sleep and 10 apneas during rapid eye movement (REM) sleep for each patient. Swings in esophageal pressure (Pes) during the preapneic period and during the beginning and the end of obstructive apneas were calculated as the average of three consecutive breaths (or ineffective efforts). The difference in Pes from the minimal initial to the maximal final apneic respiratory effort (DPes) and the rate of increase in Pes (RPes = DPes/apnea duration) during apnea were computed. Within an apnea, the lowest Pes always occurred during the first three occluded breaths and the highest during the last three, with a more marked difference in NREM sleep. The level of effective CPAP was correlated with the length of the soft palate (r = 0.69, p = 0.000), RPes (r = 0.55, p = 0.008), and DPes (r = 0.49, p = 0.02). The correlations of effective CPAP level with body mass index and apnea + hypopnea index were not significant. A model including length of the uvula, DPes, and RPes accounted for 56 to 59% of the variability in effective CPAP.(ABSTRACT TRUNCATED AT 250 WORDS)