How to disentangle psychobiological stress reactivity and recovery: A comparison of model-based and non-compartmental analyses of cortisol concentrations

TL;DR

This study compares model-based and non-compartmental methods for analyzing cortisol stress responses, developing a physiologically plausible model to better distinguish reactivity and recovery components, and evaluates their effectiveness and statistical power.

Contribution

It introduces a comprehensive pharmacokinetic model for cortisol response analysis and validates non-compartmental parameters against this model, highlighting their limitations and proposing improved measures.

Findings

Model explains 99% of cortisol data variance.

Minimum-maximum cortisol difference effectively proxies reactivity.

Minimum cortisol level is a good indicator of recovery.

Abstract

This article seeks to address the prevailing issue of how to measure specific process components of psychobiological stress responses. Particularly the change of cortisol secretion due to stress exposure has been discussed as an endophenotype of many psychosomatic health outcomes. To assess its process components, a large variety of non-compartmental parameters (i.e., composite measures of substance concentrations at different points in time) like the area under the concentration-time curve (AUC) are utilized. However, a systematic evaluation and validation of these parameters based on a physiologically plausible model of cortisol secretion has not been performed so far. Thus, a population pharmacokinetic (mixed-effects SDE) model was developed and fitted to densely sampled salivary cortisol data of 10 males from Montreal, Canada, and sparsely sampled data of 200 mixed-sex participants from Dresden, Germany, who completed the Trier Social Stress Test (TSST). Besides the two major process components representing (1) stress-related cortisol secretion (reactivity) and (2) cortisol elimination (recovery), the model incorporates two additional, often disregarded components: (3) the secretory delay after stress onset, and (4) deviations from the projected steady-state concentration. The fitted model (R2 = 99%) was thereafter used to investigate the correlation structure of the four individually varying, and readily interpretable model parameters and eleven popular non-compartmental parameters. Based on these analyses, we recommend to use the minimum-maximum cortisol difference and the minimum concentration as proxy measures of reactivity and recovery, respectively. Finally, statistical power analyses of the reactivity-related sex effect illustrate the consequences of using impure non-compartmental measures of the different process components that underlie the cortisol stress response.