Modeling Psychosocial Stress-Cortisol Interactions in Menstrual Cycle Dynamics through Machine Learning: A Systems Engineering Approach

Technical Summary: Modeling Psychosocial Stress-Cortisol Interactions in Menstrual Cycle Dynamics

Problem Statement
Menstrual health is governed by a complex interplay of endocrine regulation, psychological state, and environmental factors, primarily mediated by the hypothalamic-pituitary-ovary (HPO) axis. While extensive evidence links psychosocial stress and elevated cortisol levels to menstrual dysfunction (such as irregularity, altered flow, and dysmenorrhea), significant inter-individual variability remains unexplained. Current literature often treats stress and cortisol as universal drivers of pathology, overlooking constitutional moderators that determine individual stress responsivity and reproductive outcomes. This study addresses the gap in understanding how "temperament" (Mizaj)—a concept from Unani medicine representing a stable biopsychological state combining metabolic rate and emotional responsiveness—moderates the translation of psychosocial stress into endocrine and menstrual phenotypes.

Methodology
The study employed a cross-sectional observational design with a cohort of 119 women aged 19–28 years. Participants were stratified into two dominant temperament groups based on a standardized Unani medicine questionnaire: Safrawi (hot-dry, n=59) and Balghami (cold-moist, n=60).

• Data Collection: Researchers gathered data on psychosocial stress (using the ADSSBSPSA scale), serum cortisol levels (measured via chemiluminescent immunoassay in the follicular phase, 08:00–10:00 h), and menstrual parameters (duration, flow intensity, blood loss, pain severity, and regularity). Physiological metrics including BMI, pulse, and blood pressure were also recorded.
• Analytical Framework: A dual-analytic approach was utilized:
  1. Classical Statistics: Independent-samples t-tests and Mann-Whitney U tests were used to compare group differences. Pearson correlation coefficients assessed linear relationships between continuous variables. Binary logistic regression was applied to model the probability of cycle regularity.
  2. Machine Learning (ML): Logistic Regression and Random Forest (RF) classifiers/regressors were trained on an 80:20 train-test split to predict menstrual outcomes (regularity, pain intensity, flow heaviness, and blood loss). Model performance was evaluated using accuracy, precision, recall, F1-score, ROC-AUC, and Precision-Recall (PR) curves. Feature importance analysis was conducted to identify dominant predictors.

Key Results

• Temperament and Menstrual Dynamics: Statistical analysis revealed significant differences between temperament groups. Safrawi women exhibited shorter menstrual durations (mean ~5 days), heavier flow, and higher blood loss compared to Balghami women (mean ~6 days, lighter flow). Logistic regression identified temperament as the primary predictor of cycle regularity (B = +2.89, p < 0.001), whereas psychosocial stress scores did not significantly predict regularity.
• Stress and Pain: In contrast to regularity, menstrual pain intensity showed a strong positive correlation with psychosocial stress (r = 0.64, p < 0.001) across both temperament groups. This suggests stress acts as a universal aggravator of dysmenorrhea regardless of constitutional type.
• Machine Learning Predictions:
  • Cycle Regularity: Both Logistic Regression and Random Forest models achieved 78% accuracy. Feature importance analysis indicated that serum cortisol and BMI were the strongest predictors of regularity, surpassing perceived stress and temperament in multivariate non-linear contexts.
  • Pain Intensity: The Random Forest classifier for pain (low vs. high) achieved 81% accuracy with an ROC-AUC of 0.87. Serum cortisol and stress scores were the most significant predictors.
  • Flow and Blood Loss: ML models identified serum cortisol and BMI as the top predictors for flow heaviness and blood loss. While the models performed well in detecting extreme phenotypes (e.g., heavy blood loss), they faced challenges with class imbalance in intermediate categories.

Key Contributions

1. Operationalization of Mizaj: The study successfully translates the traditional Unani concept of temperament into a statistically tractable construct, demonstrating its role as a moderating variable in reproductive physiology.
2. Triadic Framework: It establishes a triadic model where psychosocial stress is the perturbing force, cortisol is the biochemical mediator, and temperament is the regulatory matrix determining individual vulnerability or resilience.
3. Hybrid Analytics: By integrating classical inferential statistics with machine learning, the study reveals non-linear interactions and predictive hierarchies that classical methods alone might miss, specifically highlighting the dominance of physiological markers (cortisol, BMI) over subjective stress in predicting cycle regularity.

Significance and Claims
The paper posits that menstrual variability is not merely a consistent pathological reaction to stress but an individualized physiological expression of the interaction between stress and constitution. The findings support a shift toward precision medicine in reproductive health, where biological individuality (temperament) conditions reproductive outcomes.

The authors claim that this approach provides a scalable framework for risk stratification, allowing for the identification of individuals susceptible to stress-induced menstrual disruptions through the combination of constitutional profiling and psychophysiological biomarkers. While the study highlights the potential for "temperament-informed computational modelling" in integrative reproductive health, it modestly notes that further research with longitudinal tracking and broader physiological signals is required to enhance prediction across the full spectrum of menstrual phenotypes. The work aims to move beyond reductionist, single-factor approaches to a holistic, systems-engineering perspective on menstrual health.