The purpose of this study was (1) to examine the factors that account for the variability in continuous positive airway pressure (CPAP) levels required to abolish obstructive sleep apnea (OSA) in patients with this disorder, and (2) to examine the feasibility of predicting the lowest effective pressure (CPAPmin) from simple anthropometric and polysomnographic variables easily available in all patients considered for home CPAP therapy. To accomplish these tasks we studied a group of 208 patients with OSA all of whom were treated with nasal CPAP at home. We first analyzed a model set of 38 patients all of whom had at least two polysomnographic studies (the diagnostic one and a subsequent one to determine CPAPmin for home use), anthropometric measurements (including body mass index, neck circumference, and waist circumference), pulmonary function measurements (lung volumes, airways resistance, and flow-volume curves), pharyngeal and glottic cross-sectional areas at functional residual capacity and residual volume, and nasal airflow resistances. We compared patients requiring CPAP > 10 cm H2O with those who required CPAP < 5 cm H2O. The high CPAP group was characterized by a greater degree of obesity, more severe sleep apnea, and more collapsible pharynx. Multiple linear regression analysis using principal components and Mallows C(P) statistics revealed that the optimal set of predictors for CPAPmin consisted of only three variables: apnea/hypopnea index, body mass index, and neck circumference. This model accounted for 76% of the variability in CPAP.(ABSTRACT TRUNCATED AT 250 WORDS)