预测 MacroB12：基于数据驱动的预 PEG B12 效用评估及临床注意事项
摘要
MacroB12 干扰在维生素 B12 状态诊断中 presents a significant challenge, potentially masking true deficiency. 为建立稳健的预测模型并量化 pre-聚合物酚 (PEG) B12 水平的效用, 我们对 875 名血清巴雷维尼 (greater than 1000 pg/mL) 的个体进行了回顾性分析, 采用多重插补处理缺失数据. 多变量回归模型显示, MacroB12 阳性 (小于 30% PEG 恢复) 与提示自身免疫性特征的 profiles 相关, 包括 older age (调整 odds 比 [aOR] 1.03 每年, P<0.001) 和 Rheumatologic/Autoimmune 诊断 (aOR 2.96, P=0.01). 我们还识别出一种 counter-intuitive 的血小板特征, 高血红蛋白 (aOR 1.14, P=0.02) 和平均球细胞容积 (MCV) (aOR 1.04, P=0.01) 预测 MacroB12. 普鲁士曲线 (ROC) 分析 pre-PEG B12 浓度仅显示 moderate 的判别能力 (Area Under the Curve [AUC] = 0.744, 95% CI 0.707-0.782). 最优阈值为 1584.0 pg/mL (灵敏度 71.3%, 特异度 69.7%) 用于分层临床怀疑, 但其性能 confirms it cannot replace definitive confirmatory testing.我们的 findings 定义了 suspected MacroB12 的临床和实验室表型, 但强调 confirmatory PEG 沉淀仍是 accurate 诊断 essential, 尤其是在 older 患者和具有自身免疫病史者中.
引用
@article{arxiv.2510.00178,
title = {Predicting MacroB12: Data-Driven Assessment of Pre-PEG B12 Utility and Clinical Caveats},
author = {Carmen Frías-Ruiz and José María Gálvez-Navas and Isabel García-Calcerrada and Silvia Heras Flórez and Rene Fabregas},
journal= {arXiv preprint arXiv:2510.00178},
year = {2025}
}
备注
17 pages, 6 figures (main manuscript). Includes 6 pages of supplementary material containing 2 supplementary figures and 3 supplementary tables. The study employs statistical regression and machine learning (Random Forest) to develop and assess a predictive model