Ye Xiaofei, Fu Zheng, Wang Hainan, Du Wenmin, Wang Rui, Sun Yalin, Gao Qingbin, He Jia
Department of Health Statistics, Second Military Medical University, Shanghai, China.
Pharmacoepidemiol Drug Saf. 2009 Feb;18(2):154-8. doi: 10.1002/pds.1695.
We developed a computerized system for signal detection in spontaneous reporting system (SRS) of Shanghai. Data acquisition, data mining could be carried out automatically and the process of data preprocessing and cleaning could be facilitated. This system was expected to detect signals from SRS after drug licensing with minimum patient exposure.
This system was developed by Microsoft visual basic (VB) 6.0. Data preprocessing, data cleaning, and data mining were based upon visual basic for application (VBA) in Microsoft Excel 2003. Database of drug generic name and adverse drug reaction (ADR) standard dictionary were set up initially for data cleaning and coding. Algorithms including reporting odds ratio (ROR), proportional reporting ratio (PRR), measure used by the Medicines and Healthcare Products Regulatory Agency (MHRA), Bayesian confidence propagation neural network (BCPNN) were employed in this system. Crude ADR reports submitted to Shanghai ADR SRS from December 2003 to April 2007 were used as a material in this study to test the feasibility and flexibility of this system.
Thirty two thousand seven hundred and fourty six crude ADR reports were acquired from the SRS automatically. Two thousand one hundred and fourty seven drug generic name and 621 ADR name were kept in the database after data preprocessing and cleaning. A total of 1430, 1419, 868 and 997 possible drug-ADR signals were generated by ROR, PRR, BCPNN and MHRA, respectively.
The results indicate that this computerized system is a flexible one that can help to detect possible drug-ADR signals intelligently in SRS of Shanghai now. It is a promising system for post-marketing surveillance on both chemical medicine and Chinese traditional medicine.
我们开发了一个用于上海药品不良反应自发报告系统（SRS）信号检测的计算机化系统。该系统可自动进行数据采集和数据挖掘，并有助于数据预处理和清理过程。预期该系统能够在药物获批上市后，以最少的患者暴露量从SRS中检测出信号。
本系统由微软Visual Basic（VB）6.0开发。数据预处理、数据清理和数据挖掘基于微软Excel 2003中的应用程序可视化Basic（VBA）。最初建立了药品通用名数据库和药品不良反应（ADR）标准字典，用于数据清理和编码。本系统采用了包括报告比值比（ROR）、比例报告比值（PRR）、英国药品和健康产品管理局（MHRA）使用的方法、贝叶斯置信传播神经网络（BCPNN）等算法。以2003年12月至2007年4月提交至上海药品不良反应SRS的原始ADR报告作为本研究的材料，以测试该系统的可行性和灵活性。
从SRS中自动获取了32746份原始ADR报告。经过数据预处理和清理后，数据库中保留了2147个药品通用名和621个ADR名称。分别通过ROR、PRR、BCPNN和MHRA生成了总共1430、1419、868和997个可能的药品-ADR信号。
结果表明，该计算机化系统是一个灵活的系统，目前能够帮助在上海的SRS中智能检测可能的药品-ADR信号。它是一个对化学药品和中药进行上市后监测的有前景的系统。
