常见结局事件的前瞻性研究中修正Poisson回归模型的应用

目的 介绍应用修正poisson回归模型计算常见结局事件的前瞻性研究中暴露因素的调整相对危险度的精确区间估计值.方法 应用稳健误差方差估计法（sandwich variance estimator）来校正相对危险度（RR）的估计方差,并通过SAS程序中GENMOD过程的REPEATED语句实现修正poisson回归.此外,采用不同的统计方法对5个虚拟的研究数据进行了分析比较.结果 以分层的Mantel-Haenszel法为标准参照,修正poisson回归对aRR点和区间估计均较为理想,普通poisson回归的aRR区间估计偏于保守.而logistic回归得到的aOR值明显偏离真实的RR值.结论 修正poisson回归模型适合于处理常见结局事件的前瞻性研究资料.

Modified Poisson Regression Model for Data of Prospective Studies with Common Outcomes

Objective To explore the application of modified poisson regression to estimating confidence interval of the adjusted relative risk in prospective studies with common outcomes.Methods Sandwich error estimation (a robust error variance procedure) was used to rectify the error for the estimated relative risk,and this was implemented by the SAS PROC GENMOD procedure with the REPEATED statement.Additionally,for the purpose of comparison,several statistical methods were employed to analyze five hypothetical data sets.Results Taking the results from Mantel-Haenszel method as standard reference,the point and interval estimators of adjusted relative risks obtained from modified poisson regression were similar to those,while the ordinary poisson regression produced very conservative confidence interval estimators.Furthermore,the adjusted odds ratios derived from logistic regression drastically overestimated or underestimated the true relative risks.Conclusion Modified poisson regression approach is an appropriate alternative for estimating relative risk consistently and efficiently in prospective studies with common outcomes.