Assessing the association between multiple indicators of inflammation and sleep disorders in young and middle-aged women: insights from traditional and machine learning approaches.

Abstract

Background: Interactions between inflammation and sleep disorders are increasingly recognized; however, limited research comprehensively evaluates the association between multiple inflammatory indicators and sleep disorders.

Methods: This cross-sectional study utilized data from the National Health and Nutrition Examination Survey (NHANES, 2015-2020) involving 2,342 participants. Machine learning algorithms were employed to identify inflammatory indicators with potential predictive value for sleep disorders, followed by Shapley value analysis to quantify their contributions. Weighted logistic regression and restricted cubic spline models were applied to examine associations between key inflammatory markers and sleep disorders. Mediation analysis was conducted to assess the role of depression in these relationships. Receiver operating characteristic (ROC) curves were generated to compare the predictive performance of individual inflammatory markers. Sensitivity analyses using E-values were performed to evaluate the robustness of findings against unmeasured confounding.

Results: Alpha-1-acid glycoprotein, C-reactive protein, and Naples Prognosis Score all showed significant positive correlations with sleep disorders. Among these, AGP and CRP contributed most significantly to the model (Shap value≈0.23). Furthermore, mediation analysis indicated that depression mediated 15.1% of the total effect of AGP on sleep disorders.

Conclusions: The study confirms a significant positive association between serum AGP levels and sleep disorders. Among the inflammatory markers evaluated, AGP exhibited the strongest correlation, underscoring its potential clinical relevance in the pathophysiology of sleep disturbances.