Model-based Recursive Partitioning for Survival of Iranian Female Breast Cancer Patients: Comparing with Parametric Survival Models

BACKGROUND: Precise diagnosis of disease risk factors via efficient statistical models is the primary step for reducing the heavy costs of breast cancer, as one of the most highly prevalent cancer throughout the world. Therefore, the aim of this study was to present a recently introduced statistical model in order to assess its proficiency for model fitting. METHODS: The information of 1465 eligible Iranian women with breast cancer was used for this retrospective cohort study. The statistical performances of exponential, Weibull, Log-logistic and Lognormal, as the most proper parametric survival models, were evaluated and compared with ‘Model-based Recursive Partitioning’ in order to survey their capability of more relevant risk factor detection. RESULTS: ‘Model-based Recursive Partitioning’ recognized the largest number of significant affective risk factors, whereas, all four parametric models agreed and unable to detect the effectiveness of ‘Progesterone Receptor’ as an indicator; ‘Log-Normal-based Recursive Partitioning’ could provide the paramount fit. CONCLUSION: The superiority of ‘Model-based Recursive Partitioning’ was ascertained; not only by its excellent fitness but also by its susceptibility for classification of individuals to homogeneous severity levels and its impressive visual intuition potentiality.