社区老年人群阿尔茨海默病的早期诊断和预测模型研究

Alzheimer Disease (AD) is a major public health issues in an aging society, there is no ideal clinical drug and treatment at present, so predicting and early diagnosis of the disease is an urgent problem to be resolved. Scholars at home and abroad made a searching study on the early diagnosis and risk prediction of the disease by using mathematics model, but the research work of AD are apparently slight. Just in these development, The topic conducted in urban community, combining systematic review, nested case-control study, laboratory research with Larger follow-up study, screening remarkably reliable and accepted AD related risk factors (including biologic genetic, environment, behavior and life style of individual) as parameters, establishing three mathematical models which are Risk Score, Artificial Neural Network, Competing risk model for predicting and early diagnosis of AD, and evaluating various models under different items, such as the area under ROC curves, the precision, the coincidence rate, the predictive value and decision analysis, etc. Finally, the model with the needs of the community was selected, which is high diagnostic value, simple and economic. This tool could provide important information and application technology, settings to target prevention and intervention strategies toward the community's high-risk individuals, as well as the early diagnosis and treatment of AD to change the present passive situation of the prevention and control.

阿尔茨海默病（AD）是老龄化社会的重大公共卫生问题，该病目前尚无理想的临床治疗药物和方法，因此预测和早期诊断该病是迫切需要解决的问题。国内外学者利用数学模型在疾病预测和早期诊断中进行了一些探索，但对AD的研究成果甚微。鉴于此，本课题选择城市社区为研究现场，将文献系统评价、巢式病例对照研究、实验室研究和大样本的人群随访研究相结合，筛选出AD发病可信度高、公认度好的风险因子（包括生物遗传、环境、个人生活行为方式）为参数，建立风险评分（RS）、人工神经网络（ANN）、竞争风险（CRM）数学模型，用于AD发病的预测和早期诊断；并通过ROC曲线下面积、精度、符合率、预测值等指标对不同模型进行评价，结合决策分析最终确定诊断价值高、简便经济、适宜社区的模型。为社区综合防控和临床早期治疗提供重要信息和应用技术，以改变目前AD防治的被动局面。

阿尔茨海默病（AD）是老龄化社会的重大公共卫生问题，该病目前尚无理想的临床治疗药物和方法，因此预测和早期诊断该病是迫切需要解决的问题。本研究选择城市社区为研究现场，应用巢式病例对照研究、实验室研究和大样本的人群随访研究相结合，筛选出年龄、MMSE、Aβ42、Aβ42:Aβ40、ADL、尿AD7c-NTP等AD发病相关的风险因子为参数，建立了风险评分（RS）、反向传播算法人工神经网络模型（BP–ANN）、竞争风险（CR）三种数学模型，用于AD发病的预测和早期诊断。所建RS模型ROC曲线下面积（AUC）为0.838，灵敏度为80.8%，特异度为75.7%；BP-ANN模型的预测效果及精度均优于RS模型，其准确度为82.1%，AUC为0.987，灵敏度为96.5%，特异度为94.0%，约登指数为0.905。与RS模型和ANN模型相比较，CR模型的优势在于可以衡量时间变动的因素，且能准确预测终点事件概率。本研究所建CR模型的Pearson拟合优度检验统计量为76.028，自由度的上下限值分别是89和56，P值的上下限均大于0.05，具有较高的校准度，预测结果可靠，适用于社区人群AD筛查。这3种模型的建立和应用，为社区AD的综合防控和临床早期治疗提供了重要信息和应用技术，对改变目前AD防治的被动局面有着积极的现实意义。

内容获取失败，请点击重试