We hypothesized that sleep-disordered breathing (SDB), defined by the apnea index (AI), the apnea + hypopnea index (A + H/I), or the desaturation event frequency (number of desaturations greater than 5%/h slept) (DEF), could be diagnosed after less than full-night polysomnography (PSG). Forty-eight consecutive full-night PSG sessions were evaluated by separately analyzing the first half (PSG-1/2) and the total (PSG-T) sleep time: 134.42 +/- 35.7 and 277.15 +/- 56.5 min (mean +/- SD), respectively. PSG-1/2 and PSG-T were not different with respect to AI. The DEF was statistically but not clinically higher during PSG-1/2 (41.72 +/- 41 versus 37.95 +/- 35.8, p = 0.04). Sensitivities, specificities, and predictive values of each PSG-1/2 parameter were determined by comparing the values with those measured during PSG-T, using cutoff frequencies of both 5 and 10 events/h slept to define SDB. At a cutoff frequency of 10, sensitivities and positive predictive values were high for all PSG-1/2 parameters (range, 94.6 to 96.9%). The specificities of the DEF and AI during PSG-1/2 were also high (93.75 and 95%, respectively), but the negative predictive value of the DEF was substantially better (93.75 versus 83.3%). With SDB defined by 5 events/h slept, there also were high sensitivities (87.9 to 93%), positive predictive values (93.6 to 100%), and specificities (86.7 to 100%) for all parameters. Negative predictive values, however, were substantially reduced (62.5 to 76.5%). We conclude that PSG during 2 h of sleep is an appropriate method for evaluating SDB.(ABSTRACT TRUNCATED AT 250 WORDS)