摘要:
目的:挖掘治疗2型糖尿病的用药规律,为临床治疗提供参考.方法:收集我院2012年1月-2017年6月807例2型糖尿病患者的临床数据资料,借助Gephi 0.9.1软件构建药物配伍复杂网络及各子网络的可视化图谱;运用MATLAB R2014a软件进行主题模型挖掘,并从节点距离、度中心性两个方面进行复杂网络、子网络和社团重叠结构分析,筛选出治疗2型糖尿病及其并发症的核心药物.结果与结论:本研究构建的药物配伍复杂网络共有节点119个,边24 412条,最大边权重为378,使用频次较高的药物包括门冬胰岛素注射液、氟伐他汀钠缓释片、盐酸二甲双胍片、阿卡波糖片等.该网络的平均距离为1.89,最大距离为4;89.90%的药物的距离为1或2,各药物间的联系较为紧密;门冬胰岛素注射液(0.914)、盐酸二甲双胍片(0.887)、伏格列波糖片(0.866)是该网络中的核心药物.共挖掘典型主题4个,分别为周围神经病变、外周血管病变、脂代谢异常、微细血管病变.子网络分析结果显示,各典型主题均以降糖治疗为基础,并辅以神经营养、降压、调脂以及改善视网膜病变的药物等.通过复杂网络分析所得的用药规律与2型糖尿病相关指南及临床实际相符,该方法在临床数据挖掘中具有一定意义.%OBJECTIVE:To excavate the drug use rules of type 2 diabetes,and to provide reference for clinical treatment. METHODS:Clinical data of 807 patients with type 2 diabetes were collected from our hospital during Jan. 2012-Jun. 2017. Complex network and subnetworks for drug compatibility and their visual chromatograms were established by Gephi 0.9.1 software. MATLAB R2014a software was used for topic model mining. The complex network,subnetworks and the overlapping structures of community were analyzed in respects of node distance and degree centrality. Core drugs for type 2 diabetes were screened. RESULTS & CONCLUSIONS:The drug complication complex network established in the study included 119 nodes and 24 412 edges with maximum edge weight of 378. Drugs with high frequency included Insulin aspart injection,Fluvastatin sodium extended release tablets,Metformin hydrochloride tablets,Acarbose tablets,etc. Average distance of network was 1.89,and maximum distance was 4. The distance of 89.90% drugs was 1 or 2,and the links between the various drugs were close. Insulin aspart injection(0.914),Metformin hydrochloride tablets(0.887)and Voglibose tablets(0.866)were core drugs of the network. Totally 4 typical topics were excavated, including peripheral neuropathy, peripheral vascular disease, abnormal lipid metabolism and microangiopathy. Results of subnetwork analysis showed that typical topics were based on hypoglycemic therapy and supplemented by neurotrophic drugs,antihypertensive drugs,lipid regulating drugs and drugs for improvement of retinopathy. Complex network analysis showed that drug use regularity was in line with related guideline of type 2 diabetes and clinical practice. The method has practical significance of data mining in clinic.