Model-Based Optimal AML Consolidation Treatment

AML consolidation therapy: timing matters

PURPOSE

Infections due to severe neutropenia are the most common therapy-associated causes of mortality in patients with acute myeloid leukemia (AML). New strategies to lessen the severity and duration of neutropenia are needed.

METHODS

Cytarabine is commonly used for AML consolidation therapy; we compared high- and intermediate-dose cytarabine administration on days 1, 2, and 3 (AC-123) versus days 1, 3, and 5 (AC-135) in consolidation therapy of AML. Recently, clinical trials demonstrated that high-dose AC-123 resulted in a shortened white blood cell (WBC) recovery time compared with high-dose AC-135. Our main hypothesis is that this is also the case for different cytarabine dosage, granulocyte colony-stimulating factor (G-CSF) administration, and cycle lengths. We analyzed 334 treatment schedules on virtual cohorts of digital twins.

RESULTS

Comparison of 32,565 simulated consolidation cycles resulted in a reduction in the WBC recovery time for AC-123 in 99.6% of the considered cycles (median reduction 3.5 days) without an increase in the number of leukemic blasts (lower value in 94.2% of all cycles), compared to AC-135.

CONCLUSION

Our numerical study supports the use of AC-123 plus G-CSF as standard conventional AML consolidation therapy to reduce the risk for life-threatening infectious complications.

Pharmacokinetic-pharmacodynamic modeling of maintenance therapy for childhood acute lymphoblastic leukemia

In the treatment of childhood acute lymphoblastic leukemia (ALL), current protocols combine initial high-dose multiagent chemotherapy with prolonged oral therapy with 6-mercaptopurine (6MP) and low-dose methotrexate (MTX) maintenance therapy. Decades of research on ALL treatment have resulted in survival rates of approximately 90%. However, dose-response relationships vary widely between patients and insight into the influencing factors, that would allow for improved personalized treatment management, is insufficient. We use a detailed data set with measurements of thioguanine nucleotides and MTX in red blood cells and absolute neutrophil count (ANC) to develop pharmacokinetic models for 6MP and MTX, as well as a pharmacokinetic-pharmacodynamic (PKPD) model capable of predicting individual ANC levels and thus contributing to the development of personalized treatment strategies. Here, we show that integrating metabolite measurements in red blood cells into the full PKPD model improves results when less data is available, but that model predictions are comparable to those of a fixed pharmacokinetic model when data availability is not limited, providing further evidence of the quality of existing models. With this comprehensive model development leading to dynamics similar to simpler models, we validate the suitability of this model structure and provide a foundation for further exploration of maintenance therapy strategies through simulation and optimization.

Expert-enhanced machine learning for cardiac arrhythmia classification

We propose a new method for the classification task of distinguishing atrial fibrillation (AFib) from regular atrial tachycardias including atrial flutter (AFlu) based on a surface electrocardiogram (ECG). Recently, many approaches for an automatic classification of cardiac arrhythmia were proposed and to our knowledge none of them can distinguish between these two. We discuss reasons why deep learning may not yield satisfactory results for this task. We generate new and clinically interpretable features using mathematical optimization for subsequent use within a machine learning (ML) model. These features are generated from the same input data by solving an additional regression problem with complicated combinatorial substructures. The resultant can be seen as a novel machine learning model that incorporates expert knowledge on the pathophysiology of atrial flutter. Our approach achieves an unprecedented accuracy of 82.84% and an area under the receiver operating characteristic (ROC) curve of 0.9, which classifies as "excellent" according to the classification indicator of diagnostic tests. One additional advantage of our approach is the inherent interpretability of the classification results. Our features give insight into a possibly occurring multilevel atrioventricular blocking mechanism, which may improve treatment decisions beyond the classification itself. Our research ideally complements existing textbook cardiac arrhythmia classification methods, which cannot provide a classification for the important case of AFib↔AFlu. The main contribution is the successful use of a novel mathematical model for multilevel atrioventricular block and optimization-driven inverse simulation to enhance machine learning for classification of the arguably most difficult cases in cardiac arrhythmia. A tailored Branch-and-Bound algorithm was implemented for the domain knowledge part, while standard algorithms such as Adam could be used for training.