OBJECTIVES : Breast cancer (BC) today is the leading oncologic pathology in Russia. The use of new treatments is associated with high healthcare costs. The epidemiological forecast and cost planning are not possible without the building of the mathematical model of BC.

METHODS : The systematic literature review was performed by searching databases (PubMed). The database search query was as follows: breast AND cancer AND markov AND model. Inclusion criteria were: breast cancer population, Markov modeling, real world patient data used for model building. Exclusion criteria: modeling methodology, models using simulation cohort data, BC screening models, health-economic models, tumor growth models. From 547 initially got publications 227 were excluded after article title reading, 280 were excluded after abstract reading and 20 were excluded after full text reading. Twenty articles were included in the final analysis. Not included publications could be divided into groups: pharmacoeconomic model of particular drug, BC screening models, model of tumor growth, models of ВС imaging (US, MRI).

RESULTS : We noted a pronounced heterogeneity of the finally included publications. In 7 of 20 publications we did not find the information about the patient’s sample size. The number of patients included into the models vary from 1 to 10 000. The number of Markov cycle states vary from 3 to 8, the most frequent states used in the models are: health, progressed, local progression, death. Only in one publication were defined clinical states: chemotherapy, hormone therapy and radiation therapy. The analysis showed that non-homogeneous models better describe the natural process of BC.

CONCLUSIONS : BC mathematical model should be based on the patient data from national registers, the time horizon should be not less than 5 years, it should be based on Markov modeling and be non-homogenous. The model has to differentiate several tumor types and disease stage.