Model-based approach to early predict prolonged high grade neutropenia in carboplatin-treated patients and guide G-CSF prophylactic treatment
Résumé
Purpose Neutropenia is a major dose-limiting side effect of chemotherapy and is closely related to febrile neutropenia which mainly occurs during the first cycle. Our objectives were to establish model-based decision rules from early absolute neutrophil counts (ANC) to anticipate prolonged high grade neutropenia at cycle 1 and to prevent it through delayed granulocyte colony stimulating factor (G-CSF) administration in carboplatin-treated patients. Methods The decision rules were built from Monte Carlo simulations performed with a previously published semi-mechanistic model describing ANC time-course in carboplatin-treated patients with or without concomitant G-CSF therapy. Results ANC measured at day 0 (D0, baseline), D4 and D5 were good predictors of prolonged high grade neutropenia at cycle 1. Pegfilgrastim administration on D5 was as effective as the conventional pegfilgrastim administration on D1 but none avoided prolonged high grade neutropenia in all patients. Additional decision rules were thus derived, using the same ANC combination, to identify patients for whom G-CSF was beneficial. All decision rules showed good performances (sensitivity/specificity). Conclusion We propose an innovative approach to guide oncologist in their clinical practice. The next step is to perform prospective studies to implement, validate and possibly refine the proposed decision rules.