Influence of sex hormone use on sleep architecture in a transgender cohort 173 manual scoring: κ=0.65-0.67 on the same single channel from the device (close to the inter scorer agreement, κ=0.69; Garcia-Molina et al., 2018), and even better agreement for the automatic stager compared to manual scoring from EEG and EOG signals, and from full PSG (κ=0.73 and κ=0.76, respectively; Bresch et al., 2018). Participants were asked to record their sleep at home. After completing a week of measurements, the device was brought or sent back to the hospital and the data on the device was uploaded to the study database by the local research team. Measurements were cleaned to exclude incomplete nights or nights with poor measurement quality: a measurement night was considered incomplete if the participant reported that they took off the headband during the night in the sleep diary, if the measured TST, WASO and SOL together were shorter than 240 minutes or if the contact impedance was higher than the devices’ internal threshold limit threshold for good quality recordings. For the analyses, the first night of each sleep architecture measurement week was not used for analysis, since the first night was considered a habituation night. The SE was calculated by dividing the TST by the total duration of the sleep episode (e.g. the sum of the TST, SOL and WASO) and multiplying this ratio by 100 to get a percentage. In 26 of the 121 measurements, the device was inadvertently set in active mode (as described in Garcia-Molina et al., 2018). The active mode is designed to increase slow wave amplitude in sleep restricted populations, but was shown to not affect sleep architecture (e.g. total sleep duration, WASO, SWS; Garcia-Molina et al., 2018). Sensitivity analyses incorporating device setting as covariate show that the active mode did not significantly affect measurements, as shown in the Supplementary Materials 6.2. Hence, upon consideration, we deemed this not to be a significant factor for our research questions. Statistical analyses Continuous outcomes were summarized with means and standard deviations (SDs) if they were normally distributed and medians, 25th percentile and 75th percentile if they were not normally distributed. Categorical outcomes were summarized based on counts and percentages. Outcome variables with a non-normal distribution in residuals were log-
RkJQdWJsaXNoZXIy MTk4NDMw