Nonlinear association of body roundness index with female infertility and the mediating effect of NHHR: A cross-sectional study

The Body Roundness Index (BRI) is an innovative anthropometric measure reflecting body shape and obesity, increasingly recognized as a potential predictor of infertility, warranting further exploration. This research employed a cross-sectional approach using National Health and Nutrition Examination Survey data collected between 2013 and 2018. A weighted multivariate regression model was applied to examine the connection between BRI and infertility risk. To analyze nonlinear trends, weighted restricted cubic splines were implemented, while generalized additive models helped detect threshold effects and inflection points. Subgroup and interaction analyses, adjusted by weights, evaluated specific subgroup roles. Mediation analysis investigated how non-HDL to HDL cholesterol ratio (NHHR) (non-HDL/HDL cholesterol ratio) acts as a mediator. The dataset was split into 70% training and 30% testing subsets. Characteristic variables were identified using LASSO regression, logistic regression, and the BORUTA algorithm. Prediction models were validated in both training and testing datasets to assess their predictive performance. The study analyzed data from 2576 participants. Infertility risk rose by 8% per unit increase in BRI after adjusting for covariates (odds ratio [OR] = 1.08; 95% confidence interval [CI]: 1.02-1.16; P = .018). Categorizing BRI into tertiles showed higher infertility risks for the second (OR = 1.71; 95% CI: 1.03-2.83; P = .038) and third tertiles (OR = 2.46; 95% CI: 1.37-4.43; P = .005) compared to the first tertile. Weighted restricted cubic spline analysis indicated a nonlinear link between BRI and infertility, while threshold effect analysis revealing a significant correlation for BRI < 6.4664 (P < .001). Weighted subgroup analysis showed significant interactions between BMI, age, and BRI. Mediation analysis demonstrated that NHHR mediates the association between BRI and infertility. Prediction models incorporating selected characteristic variables showed good clinical utility in both the training and testing datasets. BRI positively correlates with infertility risk, mediated by NHHR. Machine learning models suggest that incorporating BRI with other predictors can improve the management of high-risk infertility populations.