Digital twin models for predicting venetoclax and azacitidine-induced neutropenia in patients with acute myeloid leukemia

Abstract

Therapeutic toxicity, which can be life-threatening, presents a major challenge in treating patients with acute myeloid leukemia (AML). Medical digital twins, which are virtual representations of patient disease, have the potential to forecast disease progression and simulate potential treatments. Using neutrophil counts and blast percentages, we developed mechanistic models to predict toxicity (neutropenia) in AML patients receiving combination venetoclax and azacitidine treatment. We identified a best-fitting model, though patient-specific accuracy was highly variable. To address this variability, we investigated subsets of patients based on their accordance with model assumptions, and were able to identify features predictive of model fit. In addition, we found that continuous updating over time improves model accuracy. The model evaluated in this study could be further validated in a larger clinical setting and may support a digital twin for decision making in forecasting therapeutic toxicity of venetoclax and azacitidine treatment.