基于IGF-1及影像评分对伴认知障碍 脑小血管病的预测模型的构建

Development of a Predictive Model for Cerebral Small Vessel Disease with Cognitive Impairment Based on IGF-1 and Imaging Scores

摘要: 目的：探讨胰岛素样生长因子-1 (IGF-1)、脑小血管病(CSVD)负荷及内侧颞叶萎缩(MTA)等多模态指标对CSVD伴轻度认知障碍/痴呆的预测价值。方法：纳入认知正常(NC) 50例、轻度认知障碍(MCI) 44例及痴呆(DG) 44例。比较三组人口学特征、血管危险因素、血清IGF-1水平及CSVD影像学指标(Fazekas评分、腔隙灶、脑微出血、基底节PVS及CSVD总负荷评分)。采用单因素及多因素Logistic回归分析认知受损的独立相关因素，并构建多模态预测模型，采用5折交叉验证评估AUC、灵敏度及特异度。结果：三组年龄、教育年限与性别构成差异无统计学意义(均P > 0.05)，与NC组相比，MCI组与DG组糖尿病、高血压、高脂血症及冠心病发生率显著升高(均P < 0.05)。随认知损害加重，血清IGF-1水平逐渐降低(P = 0.003)，MMSE及MoCA评分显著下降(均P < 0.001)，HAMD评分升高(P = 0.002)；CSVD总负荷评分及MTA评分呈进行性升高(均P < 0.001)。多因素分析显示，教育年限(OR = 0.88, 95% CI: 0.80~0.97)、IGF-1 (每增加10 ng/ml，OR = 0.86，95% CI：0.76~0.96)为保护因素；糖尿病(OR = 3.85, 95% CI: 1.42~8.76)、高血压(OR = 2.94, 95% CI: 1.18~6.91)、CSVD总负荷评分(OR = 1.98, 95% CI: 1.24~3.12)及MTA评分(OR = 2.36, 95% CI: 1.34~4.08)为独立危险因素(均P < 0.05)。多模态模型区分认知受损(MCI + DG)与NC的5折交叉验证AUC为0.661 ± 0.060 (OOF AUC = 0.654)，判别能力中等；量表模型(MMSE + MoCA) AUC为0.922 ± 0.054；联合模型(多模态 + 量表)AUC为0.918 ± 0.039。结论：CSVD相关认知障碍与血管代谢危险因素负担、IGF-1水平下降及结构性脑损伤程度密切相关。以IGF-1、CSVD总负荷及MTA为核心的多模态模型具有中等判别效能，结合认知量表可进一步提升早期识别与风险分层能力。

Abstract: Objective: To investigate the predictive value of multimodal indicators, including serum insulin-like growth factor-1 (IGF-1), cerebral small vessel disease (CSVD) burden, and medial temporal lobe atrophy (MTA), for mild cognitive impairment (MCI) and dementia associated with CSVD. Methods: A total of 138 participants were enrolled, including 50 cognitively normal controls (NC), 44 patients with MCI, and 44 patients with dementia (DG). Demographic characteristics, vascular risk factors, serum IGF-1 levels, and neuroimaging markers of CSVD (Fazekas score, lacunes, cerebral microbleeds, basal ganglia perivascular spaces [PVS], and total CSVD burden score) were compared among the three groups. Univariate and multivariate logistic regression analyses were performed to identify independent factors associated with cognitive impairment. A multimodal predictive model was subsequently constructed, and its performance was evaluated using five-fold cross-validation to assess the area under the curve (AUC), sensitivity, and specificity. Results: There were no statistically significant differences among the three groups in age, years of education, or sex distribution (all P > 0.05). Compared with the NC group, the incidences of diabetes mellitus, hypertension, hyperlipidemia, and coronary heart disease were significantly higher in the MCI and DG groups (all P < 0.05). With worsening cognitive impairment, serum IGF-1 levels gradually decreased (P = 0.003), MMSE and MoCA scores declined significantly (both P < 0.001), whereas HAMD scores increased (P = 0.002). Meanwhile, both the total CSVD burden score and MTA score showed a progressive increase (both P < 0.001). Multivariate analysis showed that years of education (OR = 0.88, 95% CI: 0.80~0.97) and IGF-1 (per 10 ng/ml increase, OR = 0.86, 95% CI: 0.76~0.96) were protective factors, whereas diabetes mellitus (OR = 3.85, 95% CI: 1.42~8.76), hypertension (OR = 2.94, 95% CI: 1.18~6.91), total CSVD burden score (OR = 1.98, 95% CI: 1.24~3.12), and MTA score (OR = 2.36, 95% CI: 1.34~4.08) were independent risk factors (all P < 0.05). The multimodal model yielded a 5-fold cross-validated AUC of 0.661 ± 0.060 (OOF AUC = 0.654) for distinguishing cognitively impaired subjects (MCI + DG) from NC, indicating moderate discriminatory performance. The scale-based model (MMSE + MoCA) achieved an AUC of 0.922 ± 0.054, while the combined model (multimodal + scales) achieved an AUC of 0.918 ± 0.039. Conclusion: CSVD-related cognitive impairment is closely associated with vascular–metabolic risk burden, decreased serum IGF-1 levels, and the extent of structural brain damage. A multimodal model incorporating IGF-1, total CSVD burden score, and MTA demonstrates moderate discriminative performance, and its integration with cognitive assessment scales may further enhance early identification and risk stratification.