Related Experiment Video
Updated: Jun 9, 2026
08:38
Extraction and Analysis of Cortisol from Human and Monkey Hair
Published on: January 24, 2014
29.0K
Use of indirect methods and machine learning algorithms for the estimation of reference intervals, taking cortisol
Fatma Demet Arslan1, Georg Hoffmann2
1Department of Medical Biochemistry, Faculty of Medicine, Bakırçay University, Izmir, Türkiye.
Clinical Chemistry and Laboratory Medicine
|October 21, 2025
Summary
This study determined reliable total cortisol (TC) reference intervals (RIs) in adults, finding age and collection time significantly impact results. Machine learning validated these RIs, crucial for accurate clinical interpretation.
Area of Science:
- Endocrinology
- Clinical Chemistry
- Biostatistics
Background:
- Accurate reference intervals (RIs) are essential for interpreting hormone levels.
- Total cortisol (TC) levels exhibit diurnal variation and can be influenced by age.
- Existing RIs may not fully account for these physiological factors.
Purpose of the Study:
- To establish reliable RIs for total cortisol (TC) in adults.
- To investigate the influence of age and blood collection time on TC levels.
- To evaluate the utility of indirect methods and machine learning for RI determination.
Main Methods:
- Serum TC was measured using the Roche Elecsys Cortisol II kit.
- Indirect methods (refineR, reflimR) and machine learning (mclust, rpart) were employed.
- Estimated RIs were compared against the manufacturer's established interval.
Main Results:
- Machine learning methods provided RIs (e.g., 55.8-187 μg/L with mclust) that better reflected TC variability.
- Clustering revealed TC levels are highest in the early morning (08:00-08:45) and in younger adults (18-35 years).
- Age stratification (≤35 and >35 years) and time stratification (three periods) improved RI accuracy.
Conclusions:
- Age and blood collection time are critical factors for accurate TC interpretation.
- Indirect methods and machine learning effectively verify hormone RIs, especially for those with known heterogeneity.
- This approach enhances the reliability of diagnostic testing for conditions related to cortisol levels.
