Background Chronic obstructive pulmonary disease (COPD) and non-alcoholic fatty liver disease share common metabolic and inflammatory pathways. This study aims to elucidate the dual association of the fatty liver index (FLI) with COPD. Methods A dual-cohort study was conducted using data from the National Health and Nutrition Examination Survey (NHANES) 2017–2020 ( n = 15,560) and a clinical cohort from the Wenzhou Third Affiliated Hospital, China ( n = 296). Feature selection was performed using three machine learning algorithms. Multivariate logistic regression, restricted cubic splines were used to examine associations. The receiver operating characteristic curves, confusion matrix evaluated model accuracy. A nomogram was developed to visualize the contribution of key predictive factors for COPD. Results After adjustment for machine learning-selected covariates, FLI was positively associated with COPD prevalence in the NHANES cohort (OR = 1.013, 95% CI: 1.005–1.021, p = 0.002) but inversely associated with COPD severity in the Wenzhou cohort (OR = 0.567, 95% CI: 0.336–0.955, p = 0.033). A non-linear threshold relationship was identified in the Wenzhou cohort ( p -non-linear = 0.002), with an inflection point at FLI = 1.045. Machine learning-optimized models demonstrated good predictive accuracy, with the NHANES model achieving an AUC of 0.819. Conclusion FLI exhibits a dual association with COPD, serving as a risk factor for its prevalence in the general population yet being inversely associated with disease severity among diagnosed patients. These findings support FLI as a clinically relevant, non-invasive biomarker for stratifying COPD risk and assessing severity.
Huang et al. (Tue,) studied this question.