一种基于互信息的药品不良反应信号检测方法*

南京大学学报(自然科学版) ›› 2010, Vol. 46 ›› Issue (6): 705–712.

一种基于互信息的药品不良反应信号检测方法*

魏建香 1 ** , 孙越泓 2 , 朱云霞 1 , 徐厚明 3 , 孙? 俊 3 , 帅友良 1

出版日期:2015-04-03 发布日期:2015-04-03
作者简介: ( 1. 南京人口管理干部学院信息科学系, 南京, 210042; 2. 南京师范大学数学科学学院, 南京, 210097; 3. 江苏省药品不良反应监测中心, 南京, 210002)
基金资助:
国家社科基金( 09CT Q022) , 江苏省六大人才高峰!第六批项目( 09- E ?016)

A method for signal detection of adverse drug reactions based on mutual information

Wei J ian X iang 1, Sun Yue H ong 2 , Zhu Yun Xia 1 , X u H ou Ming 3 , Sun J un 3 , Shuai You?Liang 1

Online:2015-04-03 Published:2015-04-03
About author: ( 1. Department of Information Science, Nanjing College for Population Program Management, Nanjing,
210042, China; 2. School of M athematical Sciences, Nanjing Normal University, Nanjing, 210097,
China; 3. Jiangsu Center for Adverse Drug Reactions Monitoring, Nanjing, 210002, China)

摘要/Abstract

摘要： 提出一种新的、适合我国数据特点的方法, 即基于互信息的信号检测方法. 引起目标不良反应的药物可能是目标药物, 也可能是其它药物, 将目标药物和目标不良反应看作两个随机发生的事件集.
通过计算两个事件之间的互信息值来度量目标不良反应与目标药物之间的关联强度. 如果互信息值达到预先设定的标准, 则可疑的信号产生. 设计了基于 4 格表的互信息计算公式, 提出了基于互信息的信
号检测方法. 用该方法对江苏省 2008 年药品不良反应数据库进行信号检测, 其中包含 11 591 种药品与不良反应组合. 可疑信号的最小检测标准为: 互信息 MI ?0 ? 000 06 且 a-3. 检测出符合信号检测标准
的药品- 不良反应组合 820 个, 与英国药品和保健产品管理局(M HRA) 方法比较: 相同的信号 732 个, 灵敏度达到 0 89, 特异度为 0.9, 约登指数为 0 ? 79. 实验结果表明基于互信息的药品不良反应信号检测方法是可靠的和有效的.

Abstract: As drug is crucial for life safety of the people, it is of great importance how to find the potential danger signal of the marketable drugs on time and effectively. At present, there are many methods for signal detection of
adverse drug reactions ( ADR), but no international gold standard is available. Results from application of various detection methods in our database differ to a great extent. T hus, a new method based on mutual information suited
to the characteristics of our data is proposed in this paper. Mutual information is a powerful statistical tool and is useful to measure the correlation between two random events. As the target ADR can be caused by the target drug
and the other drugs, we suppose the target drug and the target ADR are two random event sets. The value of mutual information between the target ADR and drugs can reflect the strength of their correlation. If the value
achieves the predetermined criteria, the suspicious signal is produced. We design the mutual information formulabased on Two -by two contingency table and bring forward a method for signal generation based on mutual
information called MI. The method is applied to the ADR database of Jiangsu Province of China in 2008, which includes 11, 591 drug ?ADR combinations. T he minimum signal criteria is a-3 and MI .0.000 06. T he method
identifies about 820 drug-ADR combinations. In comparison with the results of M edicines and Healthcare Products Regulatory Agency ( MHRA) method, the number of same signal detected is 732 and the sensitivity of MI is 0.89;
the specificity, 0 9 and Youden index, 0 79. T he experimental results show the method is reliable and effective.

魏建香 1 ** , 孙越泓 2 , 朱云霞 1 , 徐厚明 3 , 孙? 俊 3 , 帅友良 1
. 一种基于互信息的药品不良反应信号检测方法*
[J]. 南京大学学报(自然科学版), 2010, 46(6): 705–712.

Wei J ian X iang 1, Sun Yue H ong 2 , Zhu Yun Xia 1 , X u H ou Ming 3 , Sun J un 3 , Shuai You?Liang 1
. A method for signal detection of adverse drug reactions based on mutual information
[J]. Journal of Nanjing University(Natural Sciences), 2010, 46(6): 705–712.

参考文献

[ 1 ] Finney D J. Systemic signaling of adverse reactions to drugs. Methods of Information in Medicine, 1974, 13: 1~ 10.
[ 2 ] Evans S J W, Waller P C, Davis S. Use of proportional reporting ratios ( PRRs) for signal generation from spontaneous adverse drug reaction reports. Pharmacoepidemiology and Drug Safety, 2001, 10( 6) : 483~ 486.
[ 3 ] Wilson A M, Thabane L, Holbrook A. Application of data mining techniques in pharmacovigilance. Journal of Clinical Pharmacology, 2003, 57(2): 127~ 134.
[ 4 ] Van Puijenbroek E, Diemont W, van Groothest K. Application of quantitative signal detection in the Dutch Spontaneous Reporting System for adverse drug reactions. Drug Safety, 2003, 26 (5) : 293~ 301.
[ 5 ] Hauben M, Zhou X F. Quantitative methods in pharmacovigilance: focus on signal detection. Drug Safety, 2003, 26(3): 159~ 186.
[ 6 ] Bate A. Bayesian confidence propagation neural network. Drug Safety, 2007, 30 ( 7 ): 623~ 625.
[ 7 ] Bate A, Lindquist M, Edwards I R, et al. A data mining approach for signal detection and a nalysis. Drug Safety, 2002, 25(6) : 393~ 397.
[ 8 ] Bate A, Lindquist M, Edwards I R, et al. A bayesian neural network method for adverse drug reaction signal generation. European Journal of Clinical Pharmacology, 1998, 54: 315~ 321.
[ 9 ] Bate A. Bayesian confidence propagation neural network. Drug Safety, 2007, 30( 7) : 623~ 625.
[ 10] DuMouchel W. Bayesian data mining in large frequency tables, with an application to the FDA spontaneous reporting system. American Statistician, 1999, 53: 177~ 189.
[ 11] He L, Liu Y, Cheng N N, et al. An improved algorithm for ADR signal detection and its application to Shanghai spontaneous ADR reporting system. Chinese Journal of Clinical Pharmacy, 2008, 17( 6) : 348~ 352.
( 何乐, 刘媛, 程能能等. 一种 ADR 信号检测的改进算法及其在上海市 ADR 自发呈报数据库上的应用. 中国临床药学杂志, 2008, 17( 6) : 348~ 352) .
[ 12] Zhang S H, Wang D Y, Wang Y, et al. Use of proportional reporting ratios ( PRRs) for signals detection from spontaneous reporting database of adverse drug reactions. Chinese Journal of Pharmacoepidemiology, 2006, 15( 2) : 77~ 79.
( 章少华, 王大猷, 王越等. 报告率比例失衡信号检测的实证研究. 药物流行病学杂志, 2006, 15( 2) : 77~ 79) .
[ 13] Ye X F, Wang H N, Chen W, etal. Application of data mining algorithms in pharmacovigilance. Chinese Journal of Pharmacovigilance, 2008, 5
( 1) : 36~ 40. (叶小飞, 王海南, 陈? 文. 数据挖掘在药物警戒中的应用. 中国药物警戒, 2008, 5 ( 1) : 36~ 40) .
[ 14] Li C J. Researches on theory and application of adverse drug reaction signal detection. The Fourth M ilitary Medical University, 2008. ( 李婵娟. 药品不良反应信号检测方法理论及应用研究. 第四军医大学, 2008).
[ 15] Maes F, Collignon A. Multimodality image registration by maximization of mutual information. IEEE T ransactions on Medical Image, 1997, 16( 2) : 187~ 198.
[ 16] Rigau J, Feixas M, Sbert M, et al. Medical image segmentation based on mutual information maximization. Proceedings of MICCAI, 2004: 135~ 142.
[ 17] Liu Z Q, Pan K G, Wang T C. Matching design of encoder and frequency transition function in different frequency hopping system. Journal of Nanjing University ( Natural Sciences), 2009,
45(4): 508~ 516. ( 刘智泉, 潘克刚, 王庭昌. 差分跳频系统编码器与频率转移函数的匹配设计. 南京大学学报( 自然科学) 2009, 45( 4): 508~ 516).
[ 18] Battiti R. Using mutual information for selecting features in supervised neural net learning. IEEE Transactions on Neural Networks, 1994, 5: 537~ 550.
[ 19] Yang H H, Amari S. Adaptive on line learning algorithms for blind separation-maximum entropy and minimum mutual information. Neural Computation, 1999, 7: 1457~ 1482.
[ 20] Wang Z F, Wang Z H, Fu B. A local scoringsearch algorithm for restricted Bayesian networks structure learning. Journal of Nanjing University ( Natural Sciences), 2009, 45( 5):
656~ 664. ( 王中锋, 王志海, 付彬. 一种局部打分搜索型限制性贝叶斯网络结构学习算法. 南京大学学报( 自然科学) , 2009, 45( 5) : 656~ 664) .

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed