Chapter 6 184 participate in the objective sleep measures or less likely to participate in the follow-up measurements altogether. Secondly, we opted not to conduct full-head sleep polysomnography (PSG) measurements, which means we could not study REM sleep according to the “golden standard” measure, which includes, amongst other additional channels, an eye movement electrode, and we did not have the raw EEG data to measure sleep EEG microarchitecture, such as micro-arousals, sleep spindles and eye movement. Furthermore, for EEG processing we relied on the automated sleep staging of the EEG device, which has an embedded deep neural network-based sleep stager. In a validation study, a (nonembedded) version of the sleep stager was compared to manual scoring of the EEG signal in (Bresch et al., 2018). This resulted in a Kappa score of 0.73, which is close to the human inter-rater agreement of about 0.75 (DankerHopfe et al., 2009). The devices used in this study included EOG electrodes, but these were only used for the manual scoring and not used by the automatic sleep stager, allowing the neural network in the stager to be trained to detect REM sleep from single channel EEG input only. When compared to full PSG manual scoring, the automatic stager showed even better agreement (κ = 0.76). As shown in (Bresch et al., 2018), the sensitivity for REM sleep was 0.81, meaning the automated stager performs adequately in assessment of REM sleep. The current device did not equip EOG electrodes to assess REM sleep, relying on the automatic sleep stager to classify REM sleep. The performance of the automatic sleep stager in the device was compared to manual scoring of the devices’ raw EEG data in (Garcia-Molina et al., 2019). This study showed kappa scores of 0.65 to 0.67, compared to an inter-rater agreement of 0.69, indicating that the device’s automated stager performed equally to manual scoring. Our current methodology also did not equip oximetry measurements, meaning that possible changes in sleep apnea would not have been detected by the sleep EEG device, and we did not use any sleep apnea screening questionnaires. There have been cases of transgender men developing sleep apnea after starting testosterone(Earl & Brown, 2019; Robertson et al., 2019), and previous work has linked higher testosterone levels to increased risk of sleep apnea and reduced slow wave amplitude (Morselli et al., 2018). It is possible that the sleep architecture of the TM group after three months of GAHT was affected by sleep apnea, but we could not account for this possibility in the current study. For further study of sex hormones on sleep,
RkJQdWJsaXNoZXIy MTk4NDMw