Multi-state modeling for the evaluation of prognostic and predictive biomarkers in acute myeloid leukemia

• Applicants: Axel Benner; Professor Dr. Jan Beyersmann; Professor Dr. Hartmut Döhner
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 514653984

Acute myeloid leukemia (AML) is a malignant disease characterized by (a) multiple, predominantly somatically acquired gene mutations affecting distinct functional groups of genes, (b) a complex clonal architecture, and (c) the evolution of genetic alterations during disease progression. To better understand this complex disease, advanced statistical methods are required. The usual approach to developing prognostic models for AML is to restrict the statistical analysis of the course of the disease to simple time to event endpoints, such as overall survival, or composite event-time endpoints, such as event-free survival. However, these simplified endpoints cannot adequately describe the complex disease trajectory of AML. To more accurately represent the complexity of AML and its underlying processes, and to develop a better understanding of this disease, certain statistical methods ("multistate models") can be used to mathematically model a more nuanced representation of disease states and progressions. In this project, the aim is to sequentially decompose these composite endpoints into their individual components, e.g., recurrence or death, in order to investigate them in a more differentiated manner. In addition, the objective is to identify biological markers ("biomarkers") that can be used to more precisely predict the course of the disease or treatment efficacy. Besides clinical factors (e.g., age, type of AML, or leukocyte count), biological factors to be investigated include in particular molecular markers such as gene mutations. Specifically, the project has the following goals: • Extend existing methods for multistate modeling to decompose composite endpoints, taking into account transplantation and molecular data. • Improve methods for predicting the disease course of individual patients. • Develop methods for the identification of patient subgroups with respect to therapeutic efficacy. The resulting statistical methods will be applied to existing data sets from AML clinical trials and the BiO registry trial. Thus, the analyses will provide both an improved understanding of the disease course of AML and a more precise differentiation of the efficacy of current treatment regimens based on specific patient or disease characteristics. The knowledge gained in this context can form the basis for adapting the treatment individually to the characteristics and needs of the patients and thus making the decision for a treatment strategy more efficient. In addition, it will be possible to apply the developed statistical tools to other diseases characterized by similarly complex disease processes.

• DFG Programme: Research Grants