The Mediating Role of Insulin Resistance in Obesity and The Alpha-1 Acid Glycoprotein Association in Women: An NHANES Study with Machine Learning

Background Nearly 20% of the global population is affected by obesity, which is associated with numerous health complications, including chronic inflammation and insulin resistance. This study utilized surrogate indices of insulin resistance to analyze whether obesity mediates the development of chronic inflammation induced by insulin resistance. Methods This cross-sectional study integrated data from the NHANES between 2015 and 2023. We employed data analysis and machine learning techniques to examine the correlations between the degree of obesity, surrogate indices of insulin resistance, and the inflammatory marker alpha-1 acid glycoprotein (AGP). The data were analyzed via generalized additive models (GAMs). For variables exhibiting nonlinear correlations, threshold and saturation effect analyses were further conducted. Receiver operating characteristic (ROC) curves were employed to evaluate these associations. Causal mediation analysis was applied to assess the mediating relationships among the three factors. Results The study included 828 participants. After adjusting for confounders and applying sampling weights, multivariate logistic regression revealed significant positive associations of BMI (OR: 1.28, 95% CI: 0.83 ~ 1.73), WHtR (OR: 103.53, 95% CI: 72.47 ~ 134.59), WHR (OR: 113.08, 95% CI: 57.87 ~ 168.28), waist circumference (OR: 0.60, 95% CI: 0.41 ~ 0.79), hip circumference (OR: 0.66, 95% CI: 0.43 ~ 0.89), and insulin (OR: 0.10, 95% CI: 0.01 ~ 0.19) with AGP levels. Conversely, QUICKI (OR: -55.90, 95% CI: -92.60~-19.21) showed a significant inverse association. HOMA-IR and HOMA-β became nonsignificant after adjustment. Generalized additive models revealed significant nonlinear positive associations of waist circumference, WHtR, BMI, weight, hip circumference, WHR, insulin, and HOMA-IR with AGP, with the WHtR demonstrating the strongest predictive ability (Adj. R²: 0.4431). The QUICKI showed a significant nonlinear negative association. Conclusion Obesity itself, or potentially through mediating insulin resistance, leads to elevated levels of AGP, highlighting its potential utility as a biomarker for obesity-related chronic inflammation.