In hematological malignancies, thrombocytopenia is a frequent feature of the disease. Anti-leukemic therapies are known to impact thrombocytopenia, which is commonly selected as a dose-limiting toxicity in early clinical trials. However, given the concurrent myelosuppressive effects of the drug and the underlying leukemia, it can be difficult to differentiate whether the dose of the investigational treatment should be decreased to manage thrombocytopenia or, on the contrary, increased in pursuit of higher efficacy. In this work, we evaluated the use of population PK/PD modeling as an approach in such challenging situations, by characterizing the dynamic relationship between drug exposure and blood counts, considering the underlying condition and drug effect. Clinical data from siremadlin, an investigational treatment for acute myeloid leukemia (AML), was used in this analysis. Plasma concentrations of siremadlin and platelet counts in patients with hematological malignancies and solid tumors were used in model development to enable the differentiation of drug effect from disease effect on hematopoiesis. Subsequently, the potential utility of model averaged prediction to predict drug effect in patients following an allogeneic hematopoietic stem cell transplantation, with partially compromised bone marrow function, was also evaluated. The proposed modeling approach provides a feasible methodology that can be used to represent recovering patients with different degrees of hematological malignancies and can be broadly applicable in similar clinical settings to support early decisions in drug development.
Sou et al. (Sun,) studied this question.