He Ming-Yu, Du Xin-Jie, Liu Yi-Ming
Ultrasonography Department, Longyan First Hospital, Longyan, 364300, Fujian, China.
Thyroid and Breast Surgery Department, Longyan First Hospital, Longyan, 364300, Fujian, China.
BMC Gastroenterol. 2025 Jan 17;25(1):20. doi: 10.1186/s12876-025-03612-9.
Metabolic dysfunction-associated fatty liver disease (MAFLD) is increasingly prevalent, and systemic inflammation markers may play a role in its pathogenesis. This study aimed to investigate the relationship between neutrophil-albumin ratio (NAR) and MAFLD.
This population-based study was performed using data from NHANES 2017-2018 and included 4526 individuals with a median age of 44 years old, and the males account for 46.13% (n = 2088). Ultrasound-defined MAFLD was diagnosed using a controlled attenuation parameter (CAP) threshold of ≥ 285 dB/m. Differences in baseline characteristics between patients with CAP ≥ 285 dB/m and < 285 dB/m were analyzed. A generalized additive model (GAM) and restricted cubic splines (RCS) were applied to explore the nonlinear relationship between NAR and CAP, followed by generalized linear models (GLMs). Threshold effect analysis was performed to identify the inflection point in the nonlinear relationship. CAP-related variables were ranked using XG Boost and random forest algorithms, and predictive models were developed and evaluated.
The study population included 1,503 patients with CAP ≥ 285 dB/m. NAR was significantly elevated in subjects with CAP ≥ 285 dB/m (P < 0.001), and nonlinear relationships between NAR and CAP were observed. NAR was positively associated with CAP in three GLMs, and this relationship remained after adjusting for confounding factors or dividing NAR into tertiles. Additionally, when NAR < 1.436, a one-unit rise in NAR was linked to a 3.304-fold increase in the risk of NAFLD (OR = 3.304, 95% CI: 2.649-4.122). The NAR-based random forest model showed the best predictive performance with AUC values of 0.978 (training) and 0.813 (validation).
NAR is positively associated with CAP, and the NAR-based random forest model is optimal for predicting MAFLD risk, highlighting the importance of NAR in predicting MAFLD.
代谢功能障碍相关脂肪性肝病（MAFLD）日益普遍，全身炎症标志物可能在其发病机制中起作用。本研究旨在探讨中性粒细胞与白蛋白比值（NAR）与MAFLD之间的关系。
本基于人群的研究使用了2017 - 2018年美国国家健康与营养检查调查（NHANES）的数据，纳入了4526名个体，中位年龄为44岁，男性占46.13%（n = 2088）。使用≥285 dB/m的受控衰减参数（CAP）阈值通过超声诊断MAFLD。分析了CAP≥285 dB/m和<285 dB/m的患者基线特征的差异。应用广义相加模型（GAM）和受限立方样条（RCS）来探索NAR与CAP之间的非线性关系，随后进行广义线性模型（GLM）分析。进行阈值效应分析以确定非线性关系中的拐点。使用XG Boost和随机森林算法对与CAP相关的变量进行排序，并建立和评估预测模型。
研究人群包括1503名CAP≥285 dB/m的患者。CAP≥285 dB/m的受试者中NAR显著升高（P < 0.001），并观察到NAR与CAP之间的非线性关系。在三个GLM中，NAR与CAP呈正相关，在调整混杂因素或将NAR分为三分位数后，这种关系仍然存在。此外，当NAR < 1.436时，NAR每升高一个单位，非酒精性脂肪性肝病（NAFLD）风险增加3.304倍（OR = 3.304，95% CI：2.649 - 4.122）。基于NAR的随机森林模型显示出最佳预测性能，训练集AUC值为0.978，验证集AUC值为0.813。
NAR与CAP呈正相关，基于NAR的随机森林模型在预测MAFLD风险方面最优，突出了NAR在预测MAFLD中的重要性。
