• 作者：Pablo Martí ; nez-Camblor^a ; ^b ; ^{pmcamblor@hotmail.com} ; Nerea Larrañ ; aga^a ; ^b ; Cristina Sarasqueta^a ; ^c ; Marí ; a José ; Mitxelena^d ; Mikel Basterretxea^a ; ^b
• 关键词：Aná ; lisis de supervivencia ; Supervivencia relativa ; Riesgos competitivos ; Modelo de decrementos mú ; ltiples ; Cá ; ncer de mama
• 刊名：Gaceta Sanitaria
• 出版年：2009
• 期刊代码：pca166_02139111
• 类别：med
• 出版时间：November-December 2009
• 卷：23
• 期：6
• 页码：554-557
• 文件大小：155 K

Objective

To analyze time of disease-free survival and relative survival in women diagnosed with breast cancer in the province of Gipuzkoa within the context of competing risks by assessing differences between the direct use of the Kaplan-Meier estimator and the multiple decrement method on the one hand, and relative survival on the other.

Methods

All registered breast cancer cases in Gipuzkoa in 1995 and 1996 with stages other than stage IV were included. An 8-year follow-up for recurrence and a 10-year follow-up for survival were performed. Time of disease-free survival was studied by the multiple decrement model. Observed survival and survival corrected by the expected mortality in the population (relative survival) were also studied.

Results

Estimation of the probability of recurrence at 8 years with the multiple decrement method was 8.8%lower than that obtained with the Kaplan-Meier method. The difference between the observed and relative survival rates at 10 years was 10.8%. Both results show how, in this case, the Kaplan-Meier estimator overestimates both the probability of recurrence and that of mortality from the disease.

Conclusions

Two issues are often overlooked when performing survival analyses: firstly, because of the lack of independence between survival time and censoring time, the results obtained by the Kaplan-Meier estimator are uninterpretable; secondly, it is an incontrovertible fact that one way or another, everyone causes failures. In this approach, survival analyses must take into account the probability of failure in the general population of reference. The results obtained in this study show that superficial use of the Kaplan Meier estimator overestimates both the probability of recurrence and that of mortality caused by the disease.