Survival analysis and prognosis model construction of elderly patients with acute promyelocyticleukemia: a retrospective study based on SEER database.

Abstract

The survival and death cause in elderly acute promyelocytic leukemia (APL) patients were analyzed and a prognosis model was constructed. Retrospectively, the medical data of elderly APL patients (N = 1723, from year 2000 to 2020) were gained from surveillance, epidemiology, and end results (SEER) database, and patients were randomly divided into train set and test set 1 (7:3). Test set 2 was composed of 17 APL patients from our hospital. Single factor and multi-factor analyses were performed by COX regression analysis. Death causes were analyzed among dead APL patients. Prognosis prediction model was constructed and verified. Single factor analysis results showed that age, marital status, income, year of diagnosis, months from diagnosis to treatment and chemotherapy were highly correlated with the death of APL patients. Multi-factor analysis results indicated that age, year of diagnosis and chemotherapy could independently serve as predictors to the death of APL patients. More patient succumbed to APL relative to other causes. The prognosis model had a higher diagnostic value for 0.5-year (AUC = 0.797) overall survival in train set, and for 2-year (AUC = 0.784) overall survival in test set 1. 1-year survival prediction in train set and test set 1 indicated the stability of the model. Age, year of diagnosis and chemotherapy are the independent risk factors, and APL is the leading cause of death in elderly APL patients. The prognosis model has a good clinical prediction value for elderly APL patients.