使用主成分分析的药品不良反应风险信号挖掘方法探索

目的：探索新的药品不良反应信号的提取、验证方法。方法：对收集到的头孢呋辛不良反应数据，使用Excel数据透视表，按时间统计出不同时间段发生的不良反应例次。通过主成分分析对数据进行降维，根据头孢呋辛各不良反应降维后的主成分得分，在第一、二主成分坐标平面上绘制头孢呋辛不良反应分布图，通过此分布图观察药品不良反应发生与药品使用的相关性，发现新的药品不良反应，并以传统药品不良反应监测方法优势比计算进行验证。结果：头孢呋辛的不良反应分布图中，距离“总计”点较近的分别为“皮疹”、“瘙痒”、“恶心”、“心悸”等，提示头孢呋辛的使用与这些不良反应的出现高度相关。其中，“心悸”未在说明书上载明，是头孢呋辛新的不良反应。头孢呋辛发生“心悸”风险相对其他药品优势比为2.005。结论：主成分分析能够直观鉴别出药品不良反应风险信号。

Exploration of principal component analysis for discovering the signal of adverse drug reactions

OBJECTIVE To explore a new method to extract and verify the signal of adverse drug reactions.METHODS According to the collected data of adverse reactions of cefuroxime, cases of adverse reactions happened in different periods were counted by using the Excel pivottable report. Dimensions of data were reduced by principal component analysis to count scores of principal components and draw a distribution map of adverse reactions of cefuroxime in the first and second principal component analyses on the coordinate plane. Adverse drug reactions were discovered by observing the association between adverse drug reactions and drug use, then the reactions were verified through the traditional method of odds ratio calculation.RESULTS On the distribution map of adverse reactions of cefuroxime, the points closer to the ‘total’ points were respectively represented as ‘rash’, ‘itching’, ‘disgusting’ and ‘palpitation’, etc. These prompted that the adverse reactions were highly related to the use of cefuroxime. Palpitation not stated in the instructions was a new adverse reaction of cefuroxime. The odds ratio for ‘palpitation’ was 2.005 compared to other drugs.CONCLUSION The principal component analysis can directly identify the risk signal of adverse drug reactions.