(394c) Personalized Treatment of Sickle Cell Disease Using Hydroxyurea Pharmacokinetic-Pharmacodynamic Modeling

Sickle Cell Disease (SCD) is the most common inherited blood disorder, affecting approximately 100,000 people in the United States. Hydroxyurea (HU) is the only FDA approved disease-modifying drug for SCD. The current treatment method involves daily oral dosing of hydroxyurea over a patient’s lifetime. There are many challenges associated with HU treatment. Firstly, due to different genetic and phenotypic makeups, there are wide interpatient variability in pharmacokinetic-pharmacodynamic (PK-PD) profiles which leads to variability in maximum tolerated dose (MTD). Therefore, HU dosing and escalation requires frequent visits to the hospital and usually takes 6-9 months to reach the MTD. Secondly, there is a need for a better biomarker to indicate treatment efficacy. The biomarkers currently in use are fetal hemoglobin (HbF) and mean cell volume (MCV) of red blood cells, both of which increase with HU treatment. But it is found that they take about 3 months or longer to reach a steady-state value, thus implying a delay in the clinical decision making of whether the dose should be increased or not. Lastly, non-adherence presents a major challenge. It is very difficult to differentiate between treatment inefficacy and non-adherence, which can result in underdosing, or overdosing. Due to these challenges, there is a lack of optimal treatment for individual patients. A mathematical model will be useful in addressing all the foregoing challenges through timely and personalized prediction of optimal dose, which is addressed in this work via a PK-PD model.

The PK model consists of three compartments: gastrointestinal tract, plasma, and liver where the drug gets metabolized to nitric oxide and its derivatives. This model describes the temporal distribution of HU and its metabolite in the compartments. The PD model is described by HU toxicity and efficacy models. To capture HU toxicity as a function of blood cells count, a population balance model is formulated with cell age and volume as the internal coordinates. This model also predicts MCV of red blood cells and reticulocytes which is used to calculate treatment efficacy. To quantify HU efficacy as a function of HbF, a signaling model is developed that describes HU-induced HbF activation through the NO-cGMP pathway. By integrating these models, an individual patient’s response (PK-PD variable dynamics) is predicted, which is used to compute the optimal personalized dose. This dose is calculated by maximizing efficacy (HbF, MCV level) and minimizing toxicity (blood cells count). This model is also being used to predict patient adherence. Using a probabilistic distribution for non-adherence, the effect of missing the dose on PK-PD trajectory is analyzed and the expected optimal dose is computed for non-adherent patients. In conclusion, an integrated PK-PD model is developed which can predict optimal HU dose by capturing the variant responses of SCD patients.