Predicting Chemotherapy-Induced Neutropenia and Granulocyte Colony-Stimulating Factor Response Using Model-Based In Vitro to Clinical Translation

Abstract

The ability to predict the incidence of chemotherapy-induced neutropenia in early drug development can inform risk monitoring and mitigation strategies, as well as decisions on advancing compounds to clinical trials. In this report, a physiological model of granulopoiesis that incorporates the drug's mechanism of action on cell cycle proliferation of bone marrow progenitor cells was extended to include the action of the cytotoxic agents paclitaxel, carboplatin, doxorubicin, and gemcitabine. In vitro bone marrow studies were conducted with each compound, and results were used to determine the model's drug effect parameters. Population simulations were performed to predict the absolute neutrophil count (ANC) and incidence of neutropenia for each compound, which were compared to results reported in the literature. In addition, using the single agent in vitro study results, the model was able to predict ANC time course in response to paclitaxel plus carboplatin in combination, which compared favorably to the results reported in a phase 1 clinical trial of 46 patients (r² = 0.70). Model simulations were used to compare the relative risk (RR) of neutropenia in patients with high baseline ANCs for five chemotherapeutic regimens: doxorubicin (RR = 0.59), paclitaxel plus carboplatin combination (RR = 0.079), carboplatin (RR = 0.047), paclitaxel (RR = 0.031), and gemcitabine (RR = 0.013). Finally, the model was applied to quantify the reduced incidence of neutropenia with coadministration of pegfilgrastim or filgrastim, for both paclitaxel and the combination of paclitaxel plus carboplatin. The model provides a framework for predicting clinical neutropenia using in vitro bone marrow studies of anticancer agents that may guide drug development decisions.

Keywords: (Peg)filgrastim; Carboplatin; Doxorubicin; Gemcitabine; Paclitaxel.

Fig. 1

Diagram of the physiological model of granulopoiesis. The dashed boxes represent the three major model subsystems: bone marrow; blood margination; G-CSF kinetics and regulation. The dotted lines indicate G-CSF’s action on proliferation, maturation, mobilization, and margination.

Fig. 2

In vitro bone marrow toxicity assay results and model fits. a) Paclitaxel and b) gemcitabine cell counts relative to DMSO at end of exposure versus exposure concentration. c) Carboplatin cell counts relative to DMSO at end of exposure versus concentration. d) Carboplatin recovery ratio versus time from the recovery experiment. e) Doxorubicin cell counts relative to DMSO at end of exposure versus concentration. f) Doxorubicin recovery ratio versus time from the recovery experiment. Symbols - measurements; lines - model predictions.

Fig. 3

a) Physiological model predicted paclitaxel ANC time course simulations (left panel) and frequency of grade 2, 3, 4 neutropenia quantified across four baseline ANCs (BANC: 10⁹ cells/L) (right panel).b) Corresponding paclitaxel results obtained from implementing the model of Friberg et al. as reported in (26). c) Physiological model predicted carboplatin ANC time course simulations (left panel) and frequency of grade 2, 3, 4 neutropenia quantified across four baseline ANCs (BANC: 10⁹ cells/L) (right panel). d) Corresponding carboplatin results obtained from implementing the model as reported in (23). Solid lines - median; dashed lines - 25^th and 75^th percentiles.

Fig. 4

Physiological model ANC time course simulations (gemcitabine a) and doxorubicin b)) (left panels) and frequency of grade 2, 3, 4 neutropenia quantified across four baseline ANCs (BANC: 10⁹ cells/L) (gemcitabine a) and doxorubicin b)) (right panel). c) Model predicted ANC time profiles following 75 mg/m² doxorubicin. Symbols - literature-reported median values; solid line - model predicted median, dashed lines - model predicted 25^th and 75^th percentiles; shaded area - prediction interval; arrows - time when drug administered.

Fig. 5

Model predictions versus observations from a phase I clinical trial of paclitaxel plus carboplatin combination therapy. a) Model predicted (line – median; shaded area - PI) and measured (symbols) ANC-time profiles in 12 selected patients from four BANC groups (three per group), representing BANC ranges less than 3.3×10⁹ (first row), between 3.3–4.2×10⁹ (second row), between 4.2–5.9×10⁹ (third row), and above 5.9×10⁹ cells/L (bottom row). b) Model predicted versus measures ANC values in all subjects. Symbols - measurements; solid line - line of identity; dashed line - regression line.

Fig. 6

Model predictions of responses following G-CSF therapies. a) ANC time profiles of median BANC and its prediction interval following 400 mg paclitaxel (left panel), with 6 mg pegfilgrastim (middle panel), and 5 μg/kg/day filgrastim (right panel). b) ANC time profiles of median BANC and its prediction interval following 400 mg paclitaxel plus 880 mg carboplatin combination (b, left panel), together with 6 mg pegfilgrastim (middle panel), and with 5 μg/kg/day filgrastim (right panel).