089: The Utilization of Pharmacokinetic/Pharmacodynamic (PK/PD) Model-based Analysis of Myelosuppression in Clinical Trials: Analysis from Adavosertib Clinical Studies

Statement of Purpose: Drug-induced hematological toxicities are often challenges during clinical development of novel oncologic agents. Adavosertib is a first-in-class, highly potent inhibitor of WEE1 kinase. Studies were initiated to investigate the potential role of adavosertib in treating uterine serous carcinoma, ovarian cancer and advanced solid tumors. Neutropenia and sepsis cases have been reported in monotherapy studies. By utilizing longitudinal changes in absolute neutrophil counts (ANC) from collected hematological labs, we were able to mechanistically capture the relationship between adavosertib pharmacokinetics and neutropenia, and further identified patient-level factors that may aggravate risk for neutropenia during treatment.

Description of Methods & Materials: Adavosertib clinical data were pooled from three early-phase monotherapy clinical studies. Longitudinal changes in ANC were characterized using a mechanistic model which mimicked the proliferation of progenitor cells in the bone marrow, while encompassing patient-specific time-course adavosertib plasma concentration as drug effect. ANCs measured following the administration of granulocyte colony-stimulating factors were excluded from analysis. A total of 170 patients (1792 ANC measurements) and physiologically plausible covariates were analyzed using the Friberg model (Monolix^® v2021R1). Age, baseline creatinine clearance (BCrCL), baseline ANC, and use of dexamethasone were screened as potential predictors of patients’ ANC. The final model was validated using statistical tests and diagnostic visualizations to compare model-predicted with observed ANC. Lastly, a Monte Carlo simulation of 500 patients per dose level of 300 mg, 250 mg and 200 mg QD at a regimen of 5 days on and 2 days off was performed to predict the percentage of grade ≥ 3 neutropenia across different dose levels.

Data & Results: The final model described measured ANC without significant bias. Patients’ BCrCL was a predictor of neutropenia and the model indicated that BCrCL impacts the maturation of the progenitor cells in the bone marrow during adavosertib monotherapy. The overall predicted grade ≥ 3 neutropenia aligned with the observed values (model-predicted vs observed, 11.1% vs 11.2%). More specifically, within 4 cycles at 300 mg dose, simulated patients with BCrCL ≥ 50 mL/min have grade ≥ 3 neutropenia and grade 4 neutropenia of 9.6% (observed: 9.3%; n=8/86) and 3.8% (observed: 3.5%; n=3/86), respectively. Based on Monte Carlo simulation, the predicted probability of grade ≥ 3 or grade 4 neutropenia at 200 and 250 mg QD dosing is lower than 300 mg QD dosing (Table 1). The decrease in predicted probability of developing Grade ≥3 neutropenia is 29.2% when dose is reduced from 300 to 250 mg.

Interpretation, Conclusion or Significance: This model-based analyses identified the risk factors for severe neutropenia and supported the recommendation of reducing the RP2D monotherapy dose from 300 mg QD 5 days on/2 days to 250 mg QD 5 days on/2 days for further evaluation. This modeling approach served useful to optimize adavosertib dosing regimens and can be used in early phase studies to mitigate hematological toxicities of different modalities and combinations.

Disclosures: Jincheng Yang, Shankar Lanke, Kowser Miah, Damilola Olabode, Dana Ghiorghiu, Srinivas Bachina, Diansong Zhou, Alex Phipps, and Karthick Vishwanathan are salaried and shareholders of AstraZeneca.