A forced desynchrony correction: Accounting for longitudinal confounders
Files
(Published version)
Date
2011
Authors
Matthews, R.W.
Ferguson, S.A.
Zhou, X.
Roach, G.
Editors
Kennedy, G.
Sargent, C.
Sargent, C.
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Conference paper
Citation
Little clock big clock: molecular to physiological clocks, 2011 / Kennedy, G., Sargent, C. (ed./s), pp.43-47
Statement of Responsibility
Conference Name
Australasian Chronobiology Society, 8th Annual scientific meeting (17 Sep 2011 : Melbourne, Australia)
Abstract
Aim: The forced desynchrony (FD) protocol is used widely in animal and human studies to investigate endogenous circadian and homeostatic processes. With different circadian phases systematically sampled on different experimental days, any longitudinal effect across the experiment can confound the findings. This study aims to assess a method of accounting for longitudinal effects within three FD data sets. It is hypothesised that the endogenous circadian rhythm of core body temperature will have a higher correlation to the circadian performance rhythm after the longitudinal effect is removed.
Methods: Two dummy circadian rhythms (sine waves) were plotted and cross-correlated. A linear effect was added to one rhythm and their correlation was reassessed. The previous steps were repeated in reverse order on three FD data sets that contained longitudinal effects. Firstly, the cross-correlations between core body temperature and performance were assessed. A regression of day on performance revealed the linear longitudinal effect for each data set. This was then removed by subtracting the relative effect associated with each day. The cross-correlations between core body temperature and performance were then reassessed
Results: The cross-correlations between the dummy rhythms were higher before the longitudinal effect was added (from r =.75 to r =1). Similarly, the cross-correlations between core body temperature and performance improved after the longitudinal effect was removed from each day (from r =.74, .86 and .74 to r =.96, .93 and .92 respectively).
Discussion: After the longitudinal effect was removed, the mean circadian rhythm of performance shared almost all of its variance with the average core body temperature – a measure of the endogenous circadian rhythm. This correction could explain the skewed performance rhythms depicted in many FD papers, as well as the discrepancy between the timing of performance nadirs reported in laboratory studies and in real world observations.