Xiang Y, Jin F, Sun L
Shanghai Cancer Institute.
Zhonghua Zhong Liu Za Zhi. 1996 May;18(3):181-5.
There are some limitations or disadvantages of statistical methods traditionally used in descriptive epidemiology of cancer. It can not handle the true relationships of several variables under srudy, and the effectiveness of a variable may often be confounded by other variables. This paper describes two kinds of multivariable regression models frequently used in descriptive epidemiology of cancer, such as the age-period-cohort (APC) model for the analysis of cancer incidence or mortality rate and the relative survival (RSR) model for the analysis of cancer survival rate. Detailed statistical methods, model fitting, parameter estimation, etc., are presented and two examples are used for illustration using data sets of oesophageal and stomach cancers diagnosed in urban Shanghai. The advantages of multivariable regression models are able to adjust effectiveness of confounder factors, and give estimations and evaluations of adjusted relative risks for the population.
传统上用于癌症描述性流行病学的统计方法存在一些局限性或缺点。它无法处理研究中的几个变量之间的真实关系，一个变量的有效性常常会被其他变量混淆。本文描述了癌症描述性流行病学中常用的两种多变量回归模型，如用于分析癌症发病率或死亡率的年龄-时期-队列（APC）模型以及用于分析癌症生存率的相对生存（RSR）模型。文中给出了详细的统计方法、模型拟合、参数估计等内容，并以上海市区诊断的食管癌和胃癌数据集为例进行说明。多变量回归模型的优点是能够调整混杂因素的效应，并对人群的调整后相对风险进行估计和评估。
