File(s) under embargo
Reason: The journal papers that will come out of this thesis are yet to be published.
until file(s) become available
TOWARDS PERSONALIZING HYDROXYUREA TREATMENT OF SICKLE CELL DISEASE
This work addresses sickle cell disease (SCD), a hereditary disorder caused by a single gene mutation in the beta-globin gene that produces abnormal hemoglobin and makes red blood cells sickle-shaped. Currently, SCD affects approximately 100,000 Americans and millions of people worldwide. Hydroxyurea (HU) is the most widely used disease-modifying drug for SCD, requiring daily oral doses for individuals with SCD. The main challenges associated with HU treatment are substantial interindividual variability in pharmacokinetic (PK) and pharmacodynamic (PD) profiles, cytotoxicity, and non-adherence.
HU gets cleared from the body within 24 hours, but the drug-related effects manifest on a timescale of days and take months to stabilize. Because of the short lifespan of HU inside the body, existing models have only captured HU trajectory inside the plasma over 24 hours; the relationship between daily drug dosage and the long-term effects of HU was not taken into account. In this work, the HU biomarkers trajectory on a timescale ranging approximately from 1-9 years was modeled for pediatric participants with SCD. In addition, the effect of skipping drug intake on the biomarkers trajectory was investigated to study how different patterns of non-adherence can result in different physiological profiles.
A PK model that captures the temporal changes in the HU concentration in the plasma was developed. The model performance was satisfactory for the clinical PK parameters calculated. The PK model was simulated every day with the given dose as input. The average drug concentration was computed for each day and plugged into the PD models, where drug efficacy alongside drug side effects was studied. For estimating efficacy, the effect of HU on biomarkers - fetal hemoglobin (HbF) and mean cell volume (MCV) was modeled. For HbF, the HbF activation by HU was modeled through an intermediate that directly activates the HbF. For MCV, the erythropoiesis process was modeled to examine the effects of HU on the formation of red blood cells (RBC) and its manifestation in the MCV. The HbF and MCV model performed well for both adherent as well as non-adherent participants, subject to the condition that the dosing profile contains the non-adherent information. Further, the effect of HU on white blood cells (WBC) is a manifestation of its effect on the early precursor cells. Therefore, for capturing myelosuppression, a model for the leukopoiesis process was implemented, which describes the formation of WBC in the blood circulation and how HU affects cells in different stages. For participants showing myelosuppression, the model was able to mimic their response.
It was observed that for many participants, the HbF and MCV indicated non-adherence; however, the dosing data did not contain the non-adherence information. The non-adherence in the model was incorporated using a probabilistic algorithm which led to improved model fits. In addition, to see how different forms of non-adherence affect HbF and MCV profiles, non-adherence was imposed in the model. Missing a dose once in a few days over an extended period of time was less harmful when compared to missing a dose continuously for the equivalent number of days. In summary, mathematical models were developed to simulate HU response in participants with SCD and quantify non-adherence, which can eventually help clinicians differentiate treatment inefficacy from non-adherence.
Although HU has been beneficial in improving the life expectancy of individuals with SCD and reducing sickle cell-related complications, there are challenges associated with the management of the drug. The modeling approach presented in this thesis is a key step towards understanding the long-term effects of HU on the patients’ physiology within a shorter timeframe as compared to the clinical studies that can require years of monitoring. The models developed here can be helpful in not only predicting patients’ PK-PD trajectory but also in understanding why some patients respond well while others do not and how the treatment benefit can be maximized for poor responders.