单纯被动随访评估人群肿瘤生存率偏倚分析

Analysis on bias in cancer survival estimation from population-based passive follow up study

目的 评估单纯被动随访在人群肿瘤生存率研究中对5年生存率估计造成的偏倚.方法 从沈阳、鞍山和本溪3个城市2000-2002年肿瘤登记报告数据库中随机抽取21种主要肿瘤8 334例,通过与三市2000-2005年全死因数据库联动进行被动随访,然后再通过医院病案和公安局人口信息库核对,电话、入户、社区与工作单位回访进行主动随访.生存期起始时间是首次确诊日期,根据被动和主动随访结果对患者2005-12-31的生存状态进行不同的假设,采用寿命表法计算观察生存率.结果 单纯被动随访共确认2 572例（30.9％）5年内死亡,主动随访确认2 142例（25.7％）死亡.单纯被动随访将只能确认总死亡病例的54.6％,对21种癌瘤的5年生存率高估9.0％～52.6％,高估幅度各癌间存在很大差异,高估率与生存率呈显著负相关（r=-0.474 7,P＜0.01）.单纯被动随访对甲状腺癌、乳腺癌和子宫体癌等恶性度较低癌症的生存率高估10％～20％,对肝癌、肺癌、胃癌和胰腺癌等恶性度较高癌症的生存率高估30％～50％.结论 对于目前我国肿瘤发病报告和死亡报告人口标识信息填写和（或）录入准确性较差的地区,单纯采用被动随访将严重高估肿瘤的5年生存率,需采用主动与被动随访结合的方式进行人群肿瘤生存率研究.

OBJECTIVE To assess the bias in estimations of cancer patients＇ 5-year survival rate from population- based passive follow-up study. METHODS Totally 8 334 records of 21 primary malignant tumors patients form Sheny- ang,Anshan and Benxi city were randomly extracted from cancer registry database of 2000--2002. Passive follow-up was conducted among the patients selected and the records of the patients were linked to the death registry databases of 2000-- 2005 of the three cities. Active follow-ups were also conducted to confirm vital status of the patients by checking hospital medical records and census register information of public security bureaus,visiting or telephone-calling to family members or relatives and contacts in community or workplace of the patients. Survival period was calculated from the date of initial diagnosis of the tumor to December 31,2005. Observed survival rate was estimated by life-table method based on the re- sults of both passive and active follow-up. RESULTS The passive follow-up confirmed the deaths of 2 572（30. 9%） cancer patients and the active follow-up ascertained the deaths of an additional 2 142（25.7%） patients till the end of the follow-up period. The passive follow-up only confirmed 54.6 % of the total deaths,with the overestimated 5-year observed survival rates from 9.0% to 52.6 % for the patients and the extent of overestimation was negatively correlated with the observed cancer survival rate significantly （r= -0. 474 7 ,P〈0. 01）. Based on the data of passive follow-up,the 5-year ob- served survival rate was overestimated from 10% to 20% for the patients suffering from malignant tumors with longer survival such as thyroid,breast,and uterine cancer,and from 30% to 50v for the tumors with shorter survival such as liv- er, lung, stomach and pancreas cancer. CONCLUSIONS Lack of completeness and accuracy in personal information in cancer and death registry databases could result in obvious overestimations in 5-year observed survival rate using passive follow-up method,which suggests the necessity to use both passive and active follow-up method in the population-based cancer survival study.