Development of a model-based optimal dosage regimen design scheme to minimize paclitaxel and cisplatin induced myelosuppression in non-small cell lung cancer

Abstract

Lung cancer is usually discovered after the disease is progressed and chemotherapy plays a major role in the treatment. According to the chemotherapy guideline, a combination therapy with two or more drugs is recommended for advanced stage non-small cell lung cancer (NSCLC) patients. However, a combination therapy can increase not only the effectiveness for the treatment but also the chance of adverse drug reaction (ADR) due to its toxic effect. Myelosuppression is one of the most frequently occurred ADRs during chemotherapy in cancer patients, which can lead to the reduced efficacy of chemotherapy by dose reduction and susceptibility to infection due to reduced neutrophil count. Myelosuppression is known as dose limited toxicity. However, patients treated with a chemotherapy show variable responses in the currently used BSA-based dosing system. The aim of this study is to develop a model-based optimal dosing scheme to minimize chemotherapy-induced myelosuppression in NSCLC. Analysis data were retrieved respectively from clinical data retrieve system (CDRS) and electrical medical records (EMR). Included patients were those who were newly diagnosed as stage IIIB or IV NSCLC between January 2009 and December 2013 in Severance hospital, treated with paclitaxel 175 mg/m2 3hr infusion in day 1 and cisplatin 75 mg/m2 in day 2 as the first line therapy and aged between 18 and 85 yrs old. Patients whose primary tumor was removed by surgery or who were treated with concurrent chemoradiation were excluded. The analysis variable was absolute neutrophil count (ANC) which represents the myelocyte function. A semi-mechanistic model was used to describe ANC change with time during chemotherapy. A kinetic-pharmacodynamic (K-PD) model incorporating a virtual compartment was used to describe the kinetics of paclitaxel, cisplatin and granulocyte colony stimulating factor (G-CSF) as blood concentration data for these substances were not available. Neutrophil production was described by a single compartment representing proliferative cells, 3 transit compartments representing neutrophil maturation, and a single compartment representing circulating observed blood neutrophils, where neutrophil production was influenced by negative feedback from blood neutrophil count and was assumed to be reduced by chemotherapy. The final structural model was described as a semi-mechanistic model where combined drug effect for paclitaxel and cisplatin was described by response surface model, and G-CSF effect by an ordinary Emax model. The model shows that IR50,d, virtual dose rate of paclitaxel and cisplatin corresponding to 50% of maximum drug inhibition, was reduced by 38% in women, and KDEp, paclitaxel elimination rate from the virtual compartment, was reduced by 85% in patients with DM, which was consistent with the trend observed in raw data. Nevertheless, due to the lack in physiological basis for the relation between these covariates and ANC, more stuides inclusing external validatation would be needed to confirm this finding. Model evaluation based on the precision of parameter estimates, goodness of fit plots and visual predictive check suggested that the proposed model is reasonable and parameter values were estimated with good precision. Clinically, the proposed model can be used to predict the time to nadir ANC when G-CSF treatment should be involved and to predict how to change dose regimen when it occurs, thereby avoiding serious risk in the immune system caused by severe neutropenia.