Construction of a predictive model for cerebral small vessel disease MRI burden based on β2-microglobulin and lipoprotein(a)
10.12025/j.issn.1008-6358.2025.20250085
- VernacularTitle:基于β2-微球蛋白和脂蛋白a的脑小血管病MRI总负荷预测模型构建
- Author:
Xiaoyan LI
1
;
Hongge JI
2
;
Tingting WANG
2
;
Yingying LI
2
;
Xijing ZHA
1
;
Bin LI
1
;
Dan JIANG
1
Author Information
1. Department of Neurology, Anhui No.2 Provincial People’s Hospital, Hefei 230041, Anhui, China.
2. Graduate School, Bengbu Medical University, Bengbu 233033, Anhui, China.
- Publication Type:Shortarticle
- Keywords:
cerebral small vessel disease;
total MRI burden;
β2-microglobulin;
lipoprotein(a);
predictive model
- From:
Chinese Journal of Clinical Medicine
2025;32(4):634-641
- CountryChina
- Language:Chinese
-
Abstract:
Objective To construct a predictive model for cerebral small vessel disease (CSVD) MRI burden based on β2-microglobulin (β2-MG) and lipoprotein(a) [Lp(a)], analyze its predictive value, and validate the model. Methods A total of 138 CSVD patients admitted to Anhui No.2 Provincial People’s Hospital from February 2023 to August 2024 were enrolled. Patients were divided into a low-burden group (n=63) and a moderate/severe burden group (n=75) according to the CSVD MRI burden scoring criteria. The related clinical data were compared between the two groups. Binary logistic regression analysis was used to identify independent factors for CSVD moderate/severe MRI burden. A nomogram predictive model was constructed based on these factors and its performance was evaluated. Results The proportions of male patients, as well as those with a history of diabetes or hypertension, were significantly higher in the moderate/severe burden group than those in the low burden group. Additionally, the age of patients in the moderate/severe burden group was significantly older, and the levels of β2-MG, Lp(a), and homocysteine (Hcy) were higher than those in the low burden group (P＜0.01). Binary logistic regression analysis revealed that hypertension, diabetes, β2-MG, and Lp(a) were independent factors for CSVD moderate/severe MRI burden (P＜0.05). The nomogram predictive model based on these four factors had a cut-off value of 0.467 0, with an area under curve (AUC) of 0.838 7 (95%CI 0.760 8-0.916 6) in the training set (n＝97) and 0.854 1 (95%CI 0.742 1-0.966 1) in the internal validation set (n＝41) . The calibration curve demonstrated good agreement between predicted and observed values. Decision curve analysis (DCA) indicated that the nomogram model had good clinical utility. Conclusions The nomogram model based on β2-MG and Lp(a) has high predictive performance in assessing the risk of CSVD moderate/severe MRI burden, with good discrimination and calibration.
