LAUSR

Sleep fragmentation is typical in OSA, which is commonly co-morbid with insomnia and depression. A complex interaction between these conditions may be also gender-dependent. Moreover, self-report measures of sleep quality and insomnia, such as PSQI and ISI, may relate to depression symptoms more than polysomnographic sleep disturbance. The present aim is to ascertain relative contributions of polysomnographic variables and depression symptoms to PSQI and ISI in a large sample of OSA patients. The interaction between depressive symptomatology and gender in their relationships with subjective sleep is also analyzed. A total of 1,166 patients (923 women, 1136 minorities, 18-97 y.o., age M=53.1±15.2, BMI M=34.4±8.7) undergoing an overnight PSG filled out the Center for Epidemiologic Studies Depression Scale-Revised (CESDR), ISI and PSQI. ISI and PSQI were separately regressed onto age, sex and BMI, followed by PSG variables meeting p<0.1 criterion when tested individually, followed by CESDR and CESDR-by-sex interaction. Mean AHI=29.6±34.7, range 0-167/hr, 72.3% of patients had AHI≥5. The PSQI final model included total sleep time (TST), sleep efficiency (SEF), WASO, PLM index, CESDR and CESDR-by-sex. Only CESDR and CESDR-by-sex were significant (p<0.001, p=0.023, respectively). Higher CESDR predicted higher PSQI in both sexes (both p<0.001), accounting for a greater portion of PSQI variance in men (R2=39%) than in women (R2=29%). The ISI final model included TST, N3%, REM%, SEF, WASO, total arousal index, AHI, PLM index, CESDR and CESDR-by-sex. Higher ISI related to lower TST (p=0.042, R2<1%), higher REM% (p=0.016, R2<1%), and higher CESDR (p<0.001, R2=42%). CESDR-by-sex was not significant. In this large sample, after controlling for demographic variables, PSG parameters had only minimal relationship with self-report insomnia and sleep quality measures. Higher depressive symptomatology was associated with higher subjective sleep disturbance on PSQI and worse insomnia symptoms on ISI in both sexes, accounting for 29-42% of the variance. none