Poisson回归模型和负二项回归模型在评价药物安全性中的应用

Application of Poisson Regression Model and Negative Binomial Regression Model in Drug Safety Evaluation

目的:应用Poisson回归模型和负二项回归模型评价药物的安全性,分析不良事件发生的影响因素。方法:对某上市降糖药的前瞻性临床研究进行实例分析,以处理、性别、年龄及其他基线情况为影响因素,以胃肠道不良事件的发生次数为反应变量,拟合Poisson回归模型及负二项回归模型。结果:Lagrange乘子检验统计量为7.701 5(P=0.005 5),资料存在过离散;负二项回归模型偏差411.146 8,低于Poisson回归模型的偏差546.893 1,负二项回归模型较优。使用负二项回归模型拟合资料,模型有统计学意义,"性别"进入回归模型。结论:负二项回归模型应用于药物的安全性评价,可以解决传统药物安全性评价中仅考虑不良事件发生例数而未考虑例次的不足,能进一步分析影响不良事件发生频次的影响因素,有益于更全面深入地评价药物安全性。

OBJECTIVE： To apply Poisson regression model and negative binomial regression model for the drug safety evalua- tion, so as to analyze influential factors of adverse events. METHODS： By analyzing prospective clinical trial of marketed hypoglyce- mic agents, Poisson regression model and negative binomial regression model were fitted, using treatment, gender, age and baseline as influential factors and the number of gastrointestinal adverse events as response variable. RESULTS： The statistics of Lagrange test was 7.701 5（P=0.005 5）, and data dispersion had existed. The bias of negative binomial regression model was 411.146 8, which was lower than 546.893 1 of Poisson regression model; the negative binomial regression mode/ was better than Poisson model. There was statistical significance of negative binomial regression model and gender factor was included in the regression model. CONCLUSIONS： The application of negative binomial regression model in drug safety evaluation which analyzes not only the num- ber of adverse events but also the frequency of them, and further analyzes the influential factors of adverse events, contributes to the more comprehensive evaluation of drug safety.