Model‐based approach to identify predictors of paclitaxel‐induced myelosuppression in “real‐world” administration

Abstract

Abstract Taxanes are currently the most frequently used chemotherapeutic agents in cancer care, where real‐world use has focused on minimizing adverse events and standardizing the delivery. Myelosuppression is a well‐characterized, adverse pharmacodynamic effect of taxanes. Electronic health records (EHRs) comprise data collected during routine clinical care that include patients with heterogeneous demographic, clinical, and treatment characteristics. Application of pharmacokinetic/pharmacodynamic (PK/PD) modeling to EHR data promises new insights on the real‐world use of taxanes and strategies to improve therapeutic outcomes especially for populations who are typically excluded from clinical trials, including the elderly. This investigation: (i) leveraged previously published PK/PD models developed with clinical trial data and addressed challenges to fit EHR data, and (ii) evaluated predictors of paclitaxel‐induced myelosuppression. Relevant EHR data were collected from patients treated with paclitaxel‐containing chemotherapy at Inova Schar Cancer Institute between 2015 and 2019 (n = 405). Published PK models were used to simulate mean individual exposures of paclitaxel and carboplatin, which were linearly linked to absolute neutrophil count (ANC) using a published semiphysiologic myelosuppression model. Elderly patients (≥70 years) constituted 21.2% of the dataset and 2274 ANC measurements were included in the analysis. The PD parameters were estimated and matched previously reported values. The baseline ANC and chemotherapy regimen were significant predictors of paclitaxel‐induced myelosuppression. The nadir ANC and use of supportive treatments, such as growth factors and antimicrobials, were consistent across age quantiles suggesting age had no effect on paclitaxel‐induced myelosuppression. In conclusion, EHR data could complement clinical trial data in answering key therapeutic questions.