ASH 2024 | Machine learning derived three-parameter prognostic model for survival in patients with BPDCN
Shai Shimony, MD, Dana-Farber Cancer Institute, Boston, MA, comments on the development of a machine learning-derived three-parameter prognostic model for survival in patients with blastic plasmacytoid dendritic cell neoplasm (BPDCN). Dr Shimony explains that the model, which was generated from a comprehensive dataset of 66 patients, differentiates prognosis into three groups: a favorable group with patients aged 50 or less, an intermediate group with patients over 50 and specific clinical characteristics, and an adverse risk group with patients who do not fit into the other two categories. Dr Shimony notes that the model has been generated using a machine learning process called recursive partitioning and that further verification in a validation cohort is necessary. This interview took place at the 66th ASH Annual Meeting and Exposition, held in San Diego, CA.
These works are owned by Magdalen Medical Publishing (MMP) and are protected by copyright laws and treaties around the world. All rights are reserved.
Transcript (AI-generated)
What is BPDCN? BPDCN is a very rare disease that arises from the plasmacytoid dendritic cells and can affect every organ in the body, but usually affects the skin, the bone marrow, the central nervous system and lymph nodes around the body. The optimal therapy for BPDCN is really unknown with ALL-based therapy, AML-based therapy and tagraxofusp as optional therapies...
What is BPDCN? BPDCN is a very rare disease that arises from the plasmacytoid dendritic cells and can affect every organ in the body, but usually affects the skin, the bone marrow, the central nervous system and lymph nodes around the body. The optimal therapy for BPDCN is really unknown with ALL-based therapy, AML-based therapy and tagraxofusp as optional therapies. Transplant is the optimal goal if we can get there, if patients are eligible, and to try and achieve long term remission. The knowledge about any prognostic factors in BPDCN is really scarce. Thus, we wanted to generate a prognostic model that can help us define risk categories in BPDCN.
Using a comprehensive data set of 66 patients with BPDCN with a median age of 68, we utilized a machine learning process called recursive partitioning, and we generated a three-parameter prognostic model that can differentiate the prognosis into three groups. One, which is favorable, was composed of patients aged 50 or less. The intermediate group was composed of patients who were over the age of 50 and had either overt bone marrow disease, which is defined as more than five percent of BPDCN cells in the marrow and absence of signaling mutations, or patients with no overt bone marrow disease involvement. And finally, the adverse risk group with a median overall survival of less than 10 months was composed of all patients that did not fit into the favorable or intermediate criteria. Thus, using a machine learning process and a comprehensive data set that included therapeutic, molecular and clinical characteristics, we were able to generate this prognostic model. Like any prognostic model, it should be verified in a validation cohort and that’s a work that is already in progress.
This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.
Read more...
