Background: Survival analysis is a collection of statistical methods for inference on time-to-event data. If several causes of failure occur and the occurrence of one event precludes the occurrence of the other events, the situation is known as competing risks. Since the competing risks violate the fundamental assumption of independent censoring, specific methods for inference are needed.
Objectives: The aim of this paper is to recall the competing risks model and statistical methods for nonparametric analysis, and to illustrate the competing risks methods on a real data set of 118 Chronic Myeloid Leukemia (CML) patients from the Clinic of Haemato-oncology of the University Hospital in Olomouc.
Methods: The overall survival probability and risk factors of two types of failure (death due to CML and death from other causes) are assessed. Predicted probabilities of the two types of failure with stratification based on the risk factors (Sokal score, haematological response to treatment) are shown.
Results: Outcomes of the specific methods designed for the competing risks analysis are compared with the outcomes of the standard survival analysis methods. The effect of the Sokal score classification is found ambiguous. While the score should identify high- and low-risk CML patients, it seems to be predictive only for the failure due to other causes than CML.
Conclusions: The importance of careful censoring and the need of using proper methods of analyses of competing risks data is shown. The use of the Sokal score for classification of the CML patients should be considered more thoroughly.