A novel pharmacology modeling approach to monitor acute lymphoblastic leukemia treatment

By Tarun Sai LomteJul 26 2023Reviewed by Benedette Cuffari, M.Sc.

A recent study published in Scientific Reports developed pharmacokinetic (PK) pharmacodynamic (PD) models of maintenance therapy for acute lymphoblastic leukemia (ALL).

Study: Pharmacokinetic–pharmacodynamic modeling of maintenance therapy for childhood acute lymphoblastic leukemia. Image Credit: jittawit21 / Shutterstock.com

How is ALL treatment monitored?

ALL is a common pediatric malignancy characterized by the proliferation of malignant lymphoblasts with the displacement of normal hemopoiesis. Combination chemotherapy is the primary treatment for ALL, followed by maintenance therapy with 6-mercaptopurine (6MP) and low-dose methotrexate (MTX).

Both 6MP and MTX doses are modified based on blood counts, including absolute neutrophil counts (ANC). PK and PD during maintenance therapy have yet to be elucidated.

Mathematical modeling to predict the effects of 6MP and MTX is a promising approach to individualizing treatment schedules. This is particularly important given the variations in white blood cell (WBC) count among patients.

The use of mathematical models to personalize treatment in leukemia has been increasing, with numerous models reported for ALL. Nevertheless, there remains a lack of models adjusting PK to individual patient measures.

About the study

In the present study, researchers performed PKPD modeling of the effects of 6MP and low-dose MTX on ANC. Patients with childhood ALL were treated with a combination of oral 6MP, prednisone, low-dose MTX, intrathecal MTX, and intravenous vincristine and high-dose MTX.

Observations included regular measures of ANC, as well as erythrocyte concentrations of MTX (E-MTX) and thioguanine nucleotides (E-TGN). PK models were generated for 6MP and MTX to predict E-TGN and E-MTX levels, which served as the input for the effect function modeling the declining renewal rate of proliferating cells in response to E-MTX or E-TGN increases.

The full PKPD model was then constructed to predict ANC according to 6MP and MTX doses. This was generated by combining the two PK models with a previously reported myelosuppression model.

The PD and PK aspects of the model were linked by a linear effect function modeling the effect of E-MTX and E-TGN on the proliferating rate. Different effect functions were tested based on combined E-MTX and E-TGN levels or E-TGN alone.

Study findings

Twenty PK models were compared for MTX to identify a suitable model. One model with linear kinetics and another with Michaelis-Menten kinetics were the best candidates.

Cross-validation of the two models revealed that one model that was constructed using linear kinetics produced more robust results. This model was better, even with fewer observations, and worked particularly well for the low-dose MTX regimen. 

There were no differences between the models when comparing individual patient trajectories. The goodness-of-fit plot of the model with linear kinetics showed good agreement between observed and predicted values of E-MTX. Similarly, two distinct model variants emerged as potential candidates for 6MP pharmacokinetics based on Michaelis-Menten and linear kinetics.

Although the linear kinetics model was robust, even with fewer observations, the study findings suggest that Michaelis-Menten kinetics were essential for E-TGN dynamics. Furthermore, goodness-of-fit plots between observed and predicted values preferred the model with Michaelis-Menten kinetics.

Nevertheless, the reduction and significant relative standard error of one fixed effect parameter prompted the inclusion of the linear kinetics 6MP PK model in PKPD modeling for comparison.

The best-performing PKPD model was the variant with Michaelis-Menten kinetics in the 6MP PK model and the effect function of E-TGN alone. This model was similar to a previously described PKPD model but exhibited distinct differences.

The reduction in the fixed effect parameter in the PKPD model was not as immense as in the 6MP PK model. The goodness-of-fit plot of the PKPD model revealed a good agreement between observed and predicted values of E-TGN and E-MTX.

Conclusions

The best MTX PK model with linear kinetics provided good predictions in the low-dose MTX regimen and was more robust than the model that utilized Michaelis-Menten kinetics. For 6MP, Michalis-Menten kinetics were crucial to model dynamics adequately. The final PKPD model was constructed using the Michaelis-Menten kinetics of the 6MP PK model and relied only on E-TGN for the effect function.

Although the model replicated the dynamics of ALL maintenance therapy, its ability to reach the maxima and minima of ANC was limited as compared to models of acute myeloid leukemia. Overall, the parameter estimation results were similar to a prior model; however, the current model had a better fit for observations even with less data.