Cortisol dynamics in depression: Application of a continuous-time process model

Background: The temporal dynamics of cortisol may be altered in depression. Optimally studying these dynamics in daily life requires specific analytical methods. We used a continuous-time multilevel process model to study set point (rhythm-corrected mean), variability around this set point, and regulation strength (speed with which cortisol levels regulate back to the set point after any perturbation). We examined the generalizability of the parameters across two data sets with different sampling and assay methods, and the hypothesis that regulation strength, but not set point or variability thereof, would be altered in depressed, compared to non-depressed individuals. Methods: The first data set is a general population sample of female twins (n = 523), of which 21 were depressed, with saliva samples collected 10 times a day for 5 days. The second data set consists of pair-matched clinically depressed and non-depressed individuals (n = 30), who collected saliva samples 3 times a day for 30 days. Set point, regulation strength and variability were examined using a Bayesian multilevel Ornstein-Uhlenbeck (OU) process model. They were first compared between samples, and thereafter assessed within samples in relation to depression. Results: Set point and variability of salivary cortisol were twice as high in the female twin sample, compared to the pair-matched sample. The ratio between set point and variability, as well as regulation strength, which are relative measures and therefore less affected by the specific assay method, were similar across samples. The average regulation strength was high; after an increase in cortisol, cortisol levels would decrease by 63 % after 10 min, and by 95 % after 30 min, but depressed individuals of the pair-matched sample displayed an even faster regulation strength. Conclusions: The OU process model recovered similar cortisol dynamics for the relative parameters of the two data sets. The results suggest that regulation strength is increased in depressed individuals. We recommend the presented methodology for future studies and call for replications with more diverse depressed populations.