Abstract Introduction Women with OSA are frequently underdiagnosed, in part due to sex-specific symptom patterns and milder disease presentation. To address this gap, we developed a female-specific OSA screening questionnaire and conducted a pilot study to understand which characteristics predict OSA risk in women. Methods Participants undergoing diagnostic testing for suspicion of OSA at clinics in Australia and the US were asked to complete the female-specific screening questionnaire prior to their sleep study. Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied to identify the most informative predictors of OSA. LASSO was performed separately for data using AHI and ODI as outcomes. Variables selected through LASSO were then entered into two multivariable logistic regression models to assess the likelihood of a patient having a risk of mild OSA (AHI ≥ 5, ODI ≥ 5). Odds ratios (OR 95% confidence interval) are reported. Results Responses from 437 (39% female; mean age: 48 ± 14 years; age range: 18–86 years) participants have been analyzed. Male gender (vs. female) and higher BMI classification (vs. normal weight BMI) were significant predictors of OSA in both models. Compared to pre-menopausal women, natural post-menopausal (51.02 3.15, 826.42, p = 0.006) women had higher odds of mild OSA risk in the AHI model, but using ODI as a metric revealed that both post- (15.60 1.43, 170.23, p = 0.02) and peri-menopausal women (67.26, 2.98, 1517.21, p = 0.008) had higher odds of developing mild OSA compared to pre-menopausal women. When using ODI thresholds as a metric, higher BMI classifications (vs. normal weight BMI) were significant predictors for mild OSA risk. Conclusion Our preliminary findings highlight the importance of menopausal stage and BMI to predict OSA risk. The results underscore the importance of sex-specific symptom patterns in OSA assessment and the promise of ODI as a sensitive and complementary metric to AHI in HST studies. Additional data from non-clinical populations are required to improve dataset balance and enhance model generalizability in the development of a female OSA screener. Support (if any) Funded by Resmed
Wimms et al. (Fri,) studied this question.