BACKGROUND: In some survival studies, several events are taken into consideration. If the events are independent then the ordinary methods such as Kaplan-Meier, Cox or parametric models can be used. If one of the events dependently (informatively) censors the other, the results are biased. The present study was designed to assess the risk factors for recurrence of patients with gastric cancer in the presence of informative censoring using parametric models with a semi-competing risk approach. MATERIALS AND METHODS: In a retrospective study, 408 cases of gastric cancer were selected from the patients referred to the Tehran Cancer Institute from March 2003 to March 2007. Gender, age at diagnosis, distant metastasis, tumor size, histology type, tumor grade, pathologic stage, tumor site, and type of treatment were studied as prognostic factors and used in the models. Parametric models such as Weibull, exponential, log-logistic were used with informative right censoring using Akaike Information Criteria (AIC) as criteria to compare models. The data were analyzed using R statistical software. A p-value of less than 0.05 was considered as statistically significant. RESULTS: Based on Akaike information criteria (AIC), the Weibull model best fitted to data. The effect of tumor size and pathologic stage were significant on recurrence in both univariate and multivariate analyses. Tumor site and tumor grade were significant only in univariate analysis. CONCLUSIONS: The results showed that semi-competing risk methods perform well in determining risk factors for disease recurrence.
BACKGROUND: In some survival studies, several events are taken into consideration. If the events are independent then the ordinary methods such as Kaplan-Meier, Cox or parametric models can be used. If one of the events dependently (informatively) censors the other, the results are biased. The present study was designed to assess the risk factors for recurrence of patients with gastric cancer in the presence of informative censoring using parametric models with a semi-competing risk approach. MATERIALS AND METHODS: In a retrospective study, 408 cases of gastric cancer were selected from the patients referred to the Tehran Cancer Institute from March 2003 to March 2007. Gender, age at diagnosis, distant metastasis, tumor size, histology type, tumor grade, pathologic stage, tumor site, and type of treatment were studied as prognostic factors and used in the models. Parametric models such as Weibull, exponential, log-logistic were used with informative right censoring using Akaike Information Criteria (AIC) as criteria to compare models. The data were analyzed using R statistical software. A p-value of less than 0.05 was considered as statistically significant. RESULTS: Based on Akaike information criteria (AIC), the Weibull model best fitted to data. The effect of tumor size and pathologic stage were significant on recurrence in both univariate and multivariate analyses. Tumor site and tumor grade were significant only in univariate analysis. CONCLUSIONS: The results showed that semi-competing risk methods perform well in determining risk factors for disease recurrence.