University of Michigan Ann Arbor, Michigan, United States
Background: Over 60% of cancer patients receiving paclitaxel develop chemotherapy-induced peripheral neuropathy (CIPN), which impairs functional ability and quality of life. With no effective treatments, ~25% of patients require treatment alterations including delaying, decreasing, or discontinuing treatment, which can compromise treatment efficacy. Likewise, patients without CIPN may tolerate higher doses, enhancing efficacy. This study aims to develop a pharmacokinetic-pharmacodynamic (PK-PD) model of CIPN to inform an optimal personalized dosing strategy to minimize toxicity and maximize efficacy Methods: UMCC 2014.002 (NCT02338115) enrolled 60 breast cancer patients receiving 80 mg/m2 1-hour paclitaxel infusion weekly for 12 weeks. Paclitaxel plasma concentrations were measured at the end of and 16-26 hours after the first infusion. CIPN data were collected weekly via the CIPN20 questionnaire. Paclitaxel PK profiles were simulated using an optimized two-compartment PK model (Hertz et al. BJCP 2022). The CIPN PD model was built using the simulated PK profiles and corresponding CIPN20 sensory subscale scores (CIPN8) in MONOLIX. Sequential PK-PD modeling was used by fixing individual PK parameters and estimating PD parameters. External validation was attempted by comparing the simulated rates of CIPN from our model with the actual CIPN incidence from two dosing regimens (80 mg/m2 and 100 mg/m2) tested in a Phase III clinical trial (Seidman et al. JCO 2008) Results: The best CIPN model was a hybrid model with a turnover and effect compartment structure, incorporating a threshold in the effective compartment. CIPN8 predictions captured the trends of observed CIPN8 scores well (R2=0.82) (Figure 1). Setting CIPN8>8 as the definition of CIPN, the proportion of patients that were falsely predicted to have (Type I error) or not have (Type II error) CIPN at each time point were below 4% and 8%, respectively. Similar rates of CIPN were seen in our simulations compared with the actual incidence from the clinical trial (80 mg/m2: 17% simulated, 21% actual, 100 mg/m2: 35% simulated, 30% actual) Conclusion: We developed a paclitaxel PK-PD model predictive of CIPN. Simulations from this model will aid the design of personalized dosing strategies that can be tested as interventions to optimize paclitaxel treatment outcomes in patients with cancer
