# Mathematical modeling of the cortisol stress response to develop indicators that are applicable across studies

## TL;DR

This paper introduces a new mathematical method to analyze cortisol stress responses that works well even when data collection times vary.

## Contribution

The novel contribution is a model-based approach for cortisol indicators that accommodates variable sampling schedules in stress studies.

## Key findings

- Model-based indicators showed higher accuracy than conventional AUC measures in simulations with variable sampling.
- The method enables harmonization of cortisol data across studies for IPD meta-analyses.
- Validated with both simulated and independent datasets, showing robustness.

## Abstract

Growing interest in interindividual differences in stress neuroendocrinology has created a need to combine data from multiple laboratory acute stress induction studies to allow large-scale individual participant data (IPD) meta-analyses. However, established cortisol stress response indicators such as area under the curve (AUC) are inherently affected by sampling timing and duration, although to what extent remains unknown. Here, we leveraged a large, combined dataset (STRESS-EU; n = 1295) to develop novel model-based indicators that can accommodate variability in sampling schedules. These were based on modeled individual response curves that achieve full data inter- and extrapolation. We validated this method with simulated and independent data. Crucially, combined data simulations particularly showed higher accuracy for model-based versus conventional ‘observation-based’ AUC indicators with variability in sampling duration. In conclusion, our novel method harmonizes cortisol response indicator estimates for combined data, yielding opportunities for IPD meta-analyses of acute stress test studies that could greatly advance the field.