现时法统计模型在全人群肿瘤登记生存分析中的应用

Application of model-based period analysis in estimating survival rates for population-based cancer registration data

目的系统介绍现时法统计模型的基本原理及其在全人群肿瘤登记生存数据分析中的应用。方法首先介绍常规的现时法,并在此基础上对现时法统计模型的基本原理进行了详细描述,最后以实例说明如何在R软件中应用现时法统计模型进行相对生存率的计算和趋势检验。结果利用上海市区男、女性胃癌1972—1986年生存随访数据作为实例,分别采用队列法、完全法、现时法和现时法统计模型估计1977—1981年胃癌5年相对生存率,并对相对生存率在1978—1980年、1981—1983年和1984—1986年3个观察时期间的趋势变化进行统计学检验。生存率估计结果显示,4种方法估计的5年相对生存率分别为:男性12.77%、13.76%、14.27%和16.41%;女性11.51%、12.04%、12.35%和14.74%。现时法统计模型估算的结果与后来实际观测到的胃癌5年相对生存率(男性:16.46%,女性:14.23%)较接近。趋势检验结果提示,胃癌5年相对生存率在3个观察时期间的趋势变化有统计学意义。结论现时法统计模型可在保证精确度的前提下,提高生存率估计的时效性,还可用于评估相关协变量对生存率的影响。国内肿瘤登记处在分析全人群肿瘤生存数据时,可以根据实际情况加以应用和验证。

Objective To systematically introduce the theory and computational methods of the model-based period analysis in estimating survival rates for the population-based cancer registration survival data.Methods The conventional period analysis was firstly introduced,and then the model-based period analysis was described in detail.In addition,two examples were given to illustrate how the model-based period analysis may be used in the estimation and trend test of relative survival rates in R software.Results The survival data of male and female gastric cancer in urban Shanghai from 1972 to 1986 were used in the empirical examples.The 5-year relative survival rates of gastric cancer in urban Shanghai during 1977 to 1981 were estimated by cohort method,complete method,period method and model-based period analysis.A trend test was conducted for the relative survival rates over the periods of 1978-1980,1981-1983 and 1984-1986.The estimates of 5-year relative survival rates for the above methods were 12.77%,13.76%,14.27%and 16.41%for men,and 11.51%,12.04%,12.35%and 14.74%for women,respectively.The results of the model-based period analysis were much closer to what we observed later(16.46%for men and 14.23%for women).The trend test indicated that the improvement of the relative survival rates of gastric cancer were statistically significant during the three periods.Conclusion The model-based period analysis improves the up-to-dateness of survival estimation on the premise of precision.It can also be used to evaluate the effect of relevant covariates on the survival rates of cancer patients.The method could be applied and validated in the analysis of population-based cancer registration survival data.