Prognostic factors for survival in acute leukemia patients after bone marrow transportation using Semi-Markov multi-state models in Tehran Shariati Hospital

Semi-Markov multi-state models are very important to describe regression and progression in chronic diseases and cancers. Purpose of this research was to determine the prognostic factors for survival after acute leukemia using multi-state models.

In this descriptive longitudinal research, a total of 507 acute leukemia patients (206 acute lymphocyte leukemia, (ALL), 301 acute myeloid leukemia (AML)) at Shariati Hospital, Tehran, Iran were selected. The median of follow up time was 1.5 year. A multi-state model with four states, (state 1: bone marrow transplantation, state 2: chronic graft verses host disease( cGvHD), state 3: platelet recovery, state 4: death, (absorb state) ) were defined. Weibull distribution was considered as transition time between states. Data was analyzed using R Ver 2.13 software.

With increasing sojourn time in state 1, hazard of cGVHD increased, the effects of some covariates were not constant during disease time for example after platelet recovery, death hazard of acute lymphoblastic leukemia patients was 2.15 times of acute myeloid leukemia patients. There was a negative correlation between sojourn time in state 1 and state 2(r = -0.17, P<0.001).

Semi-Markov multi-state models are useful models to accommodate multiple events and time dependent covariates.