Deng Chenyuan, Jiang Yu, Lin Yuechun, Liang Hengrui, Wang Wei, He Jianxing, Huang Ying
Department of Thoracic Surgery and Oncology, State Key Laboratory of Respiratory Disease, National Clinical Research Center for Respiratory Disease, Guangzhou Institute of Respiratory Health, the First Affiliated Hospital of Guangzhou Medical University, Guangzhou, 510120, China.
Lipids Health Dis. 2025 Aug 1;24(1):258. doi: 10.1186/s12944-025-02674-8.
Preserved ratio impaired spirometry (PRISm) has been identified as a potential precursor to chronic obstructive pulmonary disease (COPD) and demonstrates a significant correlation with unfavorable clinical outcomes. Modification of PRISm-related risk factors is a higher priority in public health than treating PRISm itself. Dietary fatty acids (FAs) affect human health through a variety of physiological pathways. However, no prior research has investigated the associations of FAs and their subclasses with PRISm, particularly the combined effects of different types of FAs.
Data analysis was conducted on 8,836 individuals drawn from the NHANES dataset spanning the years 2007 to 2012. Logistic regression and smooth curve fitting were first used to assess relationships of individual FA intake with PRISm. Multiple comparisons were adjusted using the Benjamini-Hochberg (BH) correction. Threshold effect analysis was conducted to explore potential nonlinear associations. Subsequently, innovative implementation of the principal component analysis (PCA), Weighted Quantile Sum (WQS) regression, and Bayesian Kernel Machine Regression (BKMR) approaches were employed to assess the joint impact of the various intake of FAs, as well as total saturated, monounsaturated, and polyunsaturated FAs on PRISm. To facilitate the prediction of PRISm, six distinct machine learning algorithms were constructed, followed by the application of SHAP analysis to elucidate the contribution of individual predictors. For improved clinical utility, the most effective model was further implemented as an online tool.
The weighted prevalence of PRISm observed in this study was 8.81%. The results from the single-exposure models demonstrated that most FAs were negatively associated with PRISm, and these associations remained significant after BH correction. In all three models, saturated FAs revealed impressive protective associations with PRISm. LightGBM was identified as the most effective machine learning model. Among all variables, race was the most influential factor and butyric acid (SFA 4:0) was identified as the most critical FA subclass.
Adequate dietary intake of FAs may reduce the prevalence of PRISm. Furthermore, an interactive Web-based application enables healthcare professionals to estimate individuals' odds of having PRISm and to design personalized dietary interventions based on their specific needs.
肺功能测定比值降低（PRISm）已被确定为慢性阻塞性肺疾病（COPD）的潜在先兆，并且与不良临床结局显著相关。在公共卫生领域，改变与PRISm相关的风险因素比治疗PRISm本身具有更高的优先级。膳食脂肪酸（FAs）通过多种生理途径影响人类健康。然而，此前尚无研究调查FAs及其亚类与PRISm的关联，尤其是不同类型FAs的联合作用。
对2007年至2012年美国国家健康与营养检查调查（NHANES）数据集中的8836名个体进行数据分析。首先使用逻辑回归和平滑曲线拟合来评估个体FA摄入量与PRISm的关系。使用Benjamini-Hochberg（BH）校正对多重比较进行调整。进行阈值效应分析以探索潜在的非线性关联。随后，创新性地运用主成分分析（PCA）、加权分位数和（WQS）回归以及贝叶斯核机器回归（BKMR）方法，评估各种FA摄入量以及总饱和、单不饱和和多不饱和FAs对PRISm的联合影响。为便于预测PRISm，构建了六种不同的机器学习算法，随后应用SHAP分析来阐明各个预测因子的贡献。为提高临床实用性，将最有效的模型进一步开发为在线工具。
本研究中观察到的PRISm加权患病率为8.81%。单暴露模型的结果表明，大多数FAs与PRISm呈负相关，并且在BH校正后这些关联仍然显著。在所有三个模型中，饱和FAs显示出与PRISm令人印象深刻的保护关联。LightGBM被确定为最有效的机器学习模型。在所有变量中，种族是最有影响力的因素，丁酸（饱和脂肪酸4:0）被确定为最关键的FA亚类。
充足的膳食FA摄入量可能会降低PRISm的患病率。此外，一个基于网络的交互式应用程序使医疗保健专业人员能够估计个体患PRISm的几率，并根据他们的特定需求设计个性化的饮食干预措施。
