Yaşam Çözümlemesinde İyileşme Modelleri
Survival analysis is generally defined as a group of statistical methods for analyzing the data where the outcome variable is the time until the occurrence of an event of interest. Various statistical models differ than classical models have been used in survival analysis. The Cox regression model, which is commonly used in survival analysis, assumes that every subject in the study will eventually experience the event of interest. However, this assumption may not be true when there are a lot patients in the population who are cured from their diseases and never experienced the event of interest. Especially with the improvements at cancer therapies in the recent years, lots of patients could be survivors of their diseases and because of this using “cure models” to analyze this cured fraction has become important. In this study, cure models were examined theoretically and were applied to the glioma and stomach cancer data. First, to decide which type of cure models should be used, proportional hazards assumption was examined for both glioma and stomach cancer data sets. The assumption of proportional hazards was valid for the glioma data, whereas it is not valid for stomach cancer data. Cox cure model was considered for the glioma data set. When logit link function was selected for this Cox cure model, none of the covariates were found to be statistically significant for the cured part while malignancy grade and age were statistically significant covariates for uncured part. When cloglog link function was used, malignancy grade and age covariates were statistically significant for both cured and uncured parts. For stomach cancer data, accelerated failure time model is considered due to the violation of proportional hazards assumption. Cloglog link function was selected for this model, it was seen that metastasis covariate was statistically significant for cured and uncured parts. The results which were obtained from these cure models and the ones that were obtained from classical survival models were examined and interpreted.