Application of Multi-State Model in Analyzing of Breast Cancer Data

Abstract

Background: The multistate model is used generally to fit the longitudinal data. This model can determine the natural trend of disease progress in different states of treatment, recuperate, metastasis and finally death. We aimed to use multistate models in order to analyzing breast cancer (BC) data. Study design: A historical cohort study. Methods: In this historical cohort study, 573 women with BC were studied. These patients were referred to Isfahan Sayed-o-Shohada Hospital during 1999-2006 and followed up to Apr 2017. The corresponding provided data were gathered by Isfahan Cancer Prevention Center. Then data analyzed by multistate models in R 3.4.1 software. Results: The mean and standard deviation of women age were 47.19 +/- 10.77 years. The transition probability from state of first treatment to recuperate state was 71, to metastasis state 2 and to death was 16. The sojourn time in different states of disease was 2.39 yr for first treatment, 6.93 yr for recuperate and 0.16 yr for death. Conclusion: This model is able to predict the transition probabilities in different state of disease, so its results are useful for clinical researches. In addition, with transition probabilities and also survival mean in each state in hand, the physicians will be able to suggest suitable treatment plans for patients.