Background: Biological rhythm disturbance is involved in the course of bipolar disorder (BD) and a biomarker of biological rhythm instability would be clinically useful. Actigraphy has received insufficient because it is the complex endpoint of circadian, sleep and social factors. We postulated that, as long as this inferential complexity is recognized, actigraph data has potential as a biomarker of BD.
Methods: In Study 1, traitlike affective instability was measured on the General Behaviour Inventory, and two groups of high (N = 35) and low (N = 35) vulnerability healthy young adults were compared. In Study 2, a clinical sample (N = 15) and matched healthy controls (N = 15) were compared. In studies 1 and 2, wavelet analysis was used to generate a multi-scale activity parameter for each person. In Study 3, daily actigraph parameters were used to predict within-subject fluctuations of mood state in a clinical sample (N = 11) monitored longitudinally for an average of 133.7 days.
Results: High and low vulnerability subjects were significantly differentiated by the shape parameter of a multi-scale function F(1,68) = 12.15, p <.01 (Study 1). Study 2 data is currently being analysed. In Study 3, long-term actigraphy was found to be well tolerated, and ΔActivity significantly predicted ΔMood in hierarchical linear modelling analyses, t(755) = 3.84, p < .001.
Conclusions: Non-linear signals in actigraph data may act as nonintrusive trait and state biomarkers of BD. Although actigraph data cannot be parsed into constituent drivers, it contains signals that are clinically important.