This exploratory analysis identified three novel insomnia subtypes based on PSG-derived sleep macroarchitecture and EEG spectral power variables. These subtypes have distinct objective sleep duration, coupled with differences in delta power during NREM and REM sleep and importantly differences in subjective sleep. Only the SSDD subtype showed aberrant sleep macroarchitecture along with reductions in EEG-delta power during NREM and REM sleep. The NSDD subtype also showed this reduced EEG-delta power, but this did not result in aberrant sleep macroarchitecture. Furthermore, using acute sleep restriction to reduce sleep opportunity, both SSDD and NSDD subtypes showed improvements in SE, SOL, WASO, delta power and subjective sleep ratings. Finally, the NNS subtype seemed to have “normal” sleep with no apparent differences in sleep macrostructure or EEG-power as compared to controls and showed no changes in delta power with sleep restriction. Given the absence of abnormal neurophysiological findings in NNS, we speculate that this subtype will only show neurophysiological abnormalities using higher granularity EEG measures, such as high-density EEG (HdEEG) studies to explain their insomnia symptomology (Lecci et al., 2020). Collectively, our exploratory findings are the first to show insomnia subtypes that have distinctive patterns in sleep duration, coupled with differences in delta power and subjective reports as summarised in supplementary Tables 3 and 6.
Insomnia subtypes
Our data-driven approach identified at least one insomnia subtype with short (< 6 hr, SSDD) and two with normal sleep duration (> 6hr, NSDD and NNS). The SSDD in our study aligns to some extent with the proposed short-sleep insomnia phenotype based on objective sleep duration (Vgontzas et al., 2013). This phenotype was originally proposed to explain a biologically severe phenotype (< 6 hr TST) with significant hyperarousal and increased morbidity risk. Subsequently, there have been a number of studies that have shown increased cardiometabolic risk (Bathgate et al., 2016; Johnson et al., 2021), attenuated cognitive behavioural therapy response (Troxel et al., 2013), and brain metabolism differences (Miller et al., 2016) between the short-, and normal-sleep duration insomnia.
Both insomnia SSDD and NSDD subtypes significantly underestimated sleep duration compared with controls during a habitual sleep night. They also showed greater sleep misperception coupled with reduced delta power in NREM sleep compared to NNS and controls. Some have proposed that sleep-state misperception may represent one insomnia subtype (Manconi et al., 2010), but our observations suggest that both SSDD and NSDD report misperception. However, recent evidence shows that individuals with insomnia who underestimate sleep duration may actually be accurately reporting sleep disturbance (Lecci et al., 2020). This study used HdEEG in 10 participants and found that sleep was associated with a shift from low to high frequency spectral power in central and posterior brain regions indicative of wake-like activity. Our findings support these results and suggest that sleep misperception, based on sleep duration, may need to be reconsidered as sleep mismeasurement, especially in relation to subtype differences in insomnia. Although NNS reported insomnia (Insomnia Severity Index (ISI) and Pittsburgh Sleep Quality Index (PSQI)), there was no clear evidence of SSM and similar PSG and EEG spectral power findings as controls. It is possible that this subtype has a lesser degree of abnormality on HdEEG testing but this is entirely speculative and will require larger studies of insomnia subtypes.
Acute sleep restriction
With the acute sleep restriction, we found insomnia SSDD and NSDD not only increased delta power but also showed improvements in several clinical outcomes, such as objective sleep efficiency, subjective sleep quality and sleep misperception. These findings align with research showing that sleep restriction therapy improves both objective and subjective sleep in insomnia (Fernando III et al., 2013; Galbiati et al., 2019; Miller et al., 2018). Our results support the hypothesis that the impaired homeostatic sleep function in insomnia may be corrected using behavioural treatment, such as sleep restriction, to improve sleep quantity and quality (Cervena et al., 2004).
Interhemispheric asymmetry
Interhemispheric asymmetry did not contribute meaningfully to identifying insomnia subtypes. Previous work (Tamaki et al. 2016) found reduced SWA in the left hemisphere default mode network (DMN) compared to the right hemisphere in healthy young participants experiencing their first night in a sleep laboratory (first night effect). This suggests that interhemispheric asymmetry maybe more related to immediate disturbed sleep in young healthy controls rather than to chronic sleep disturbances. It is possible we did not observe interhemispheric asymmetry as we used the whole hemisphere which may not adequately identify regional asymmetry between subtypes. Whilst we did not find associations between interhemispheric asymmetry and subjective sleep quality, further research is needed to clarify the role of interhemispheric asymmetry in insomnia and by exploring regional EEG brain activity using HdEEG (Colombo et al. 2016).
There are several limitations in this study. This was an exploratory study using a clustering approach to differentiate insomnia subtypes was based on a single night PSG. Spectral EEG activity may have been affected by first night effect or influenced by reverse first-night effect in insomnia. The reliability of neurophysiological subtypes in insomnia would need to be explored in future studies by using multiple sleep nights. Our data are from cross-sectional analyses and the stability of the subtypes needs to be assessed by using longitudinal designs. We observed three subtype clusters from 99 insomnia patients, however, a larger dataset of participants and controls will be required to confirm the subtypes and clinical significance with response to various types of therapies and may show differences in subjective sleep quality and sleep misperception. There may have been an order effect based on participants completing a habitual sleep night followed by a consecutive sleep restricted night. Although others have examined sleep restriction and found similar results (Cervena et al., 2004). The controls were younger than the insomnia participants and this may reflect the differences in SWA. This needs to be tested in larger samples.
Implications for treatment and conclusion
It remains to be seen whether this or other ID subtyping can assist in targeting specific insomnia therapies to patients. For example, SSDD may require a shorter period of hypnotic medication to improve the impaired homeostatic sleep drive, and slow wave enhancement therapy for both SSDD and NSDD based on the spectral power deficits. Clearly clinical trials are needed to test these hypotheses.
In conclusion, our data-driven classifications of objective PSG sleep duration and EEG spectral power revealed three neurophysiological insomnia subtypes, highlighting potential neural mechanisms underlying sleep misperception in insomnia disorder. The results suggest that SWA may be deficient in some subtypes and can be related to the subjective complaints reported in insomnia. In addition to the two subtypes with deficits in delta power during NREM sleep, we observed a third insomnia subtype with normal PSG and EEG spectral power patterns. This will provide insights into the biological mechanisms underpinning insomnia presentations, especially related to misperception, a central tenet of many insomnia patients.