Abstract
Methods for predicting circadian phase have been developed for healthy individuals. It is unknown whether these methods generalize to clinical populations, such as delayed sleep–wake phase disorder (DSWPD), where circadian timing is associated with functional outcomes. This study evaluated two methods for predicting dim light melatonin onset (DLMO) in 154 DSWPD patients using ~ 7 days of sleep–wake and light data: a dynamic model and a statistical model. The dynamic model has been validated in healthy individuals under both laboratory and field conditions. The statistical model was developed for this dataset and used a multiple linear regression of light exposure during phase delay/advance portions of the phase response curve, as well as sleep timing and demographic variables. Both models performed comparably well in predicting DLMO. The dynamic model predicted DLMO with root mean square error of 68 min, with predictions accurate to within ± 1 h in 58% of participants and ± 2 h in 95%. The statistical model predicted DLMO with root mean square error of 57 min, with predictions accurate to within ± 1 h in 75% of participants and ± 2 h in 96%. We conclude that circadian phase prediction from light data is a viable technique for improving screening, diagnosis, and treatment of DSWPD.
Highlights
Methods for predicting circadian phase have been developed for healthy individuals
There were no significant differences between the test and training datasets for sex, age, dim light melatonin onset (DLMO) time, desired bedtime-DLMO phase angle, bed and wake times, composite morningness–eveningness questionnaire, or clinical global impression (CGI) scale
State-of-the-art methods for predicting circadian timing have yet to be tested in clinical populations
Summary
Methods for predicting circadian phase have been developed for healthy individuals. It is unknown whether these methods generalize to clinical populations, such as delayed sleep–wake phase disorder (DSWPD), where circadian timing is associated with functional outcomes. These models have been developed and tested in healthy individuals under laboratory and field conditions to predict circadian phase markers such as core body temperature minimum (CBTmin)[3,10] and dim light melatonin onset (DLMO)[4] It is currently unknown, whether these models can accurately predict circadian phase in clinical populations, such as those with circadian rhythm sleep disorders. We have shown previously that 43% of DSWPD patients, using current diagnostic criteria, do not have a circadian phase delay relative to the desired sleep–wake schedule[14], with others reporting similar findings[15] This apparent discrepancy can occur because the relationship between the onset of the evening rise in melatonin (a gold-standard circadian phase marker16) and sleep is highly variable, with an inter-individual range of up to 5 h between melatonin onset and sleep onset in healthy p opulations[17,18,19]. While simpler biochemical measures are being developed[22,23,24], this clinical gap could be addressed by less invasive techniques that attempt to predict circadian timing using ambulatory monitoring of activity and light, given that light is the primary synchronizing agent for the circadian c lock[25]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
