PURPOSE: The clinical indications and economic consequences of human granulocyte colony-stimulating factor (G-CSF) prescription during small-cell lung cancer (SCLC) chemotherapy remain controversial. The aim of this study, based on a Markov model, was to assess the impact of routine G-CSF use in the treatment of SCLC on costs and patient comfort. Markov models allow the modeling SCLC chemotherapy, in which the risk of febrile neutropenia (FN) is continuous over time and may occur more than once. PATIENTS AND METHODS: We used a Markov model to compare three strategies: a chemotherapy dose reduction after FN and nonuse of G-CSF ("never" strategy), secondary use of G-CSF ("CSF if FN" strategy) and primary use of G-CSF ("systematic CSF" strategy). Model baseline probabilities were based on a review of medical records for all patients (n = 39) treated for SCLC in our unit during 1993 (when G-CSF was not used) and on published reductions in the incidence of FN obtained by using G-CSF. Two different types of rewards were used: a cost-utility scale that took into account the costs of FN (CFN) episodes and G-CSF (CCSF) courses; and a comfort-utility scale that took into account the discomfort of FN and G-CSF therapy. Costs were analyzed from the health care payer's perspective and utilities were assessed prospectively in standardized interviews with treated SCLC patients. RESULTS: The never strategy was the least costly ($4,875 [United States] versus $5,816 and $7,690 for CSF if FN and systematic CSF) and gave the highest comfort value (378 U v 365 and 327 for CSF if FN and systematic CSF). Sensitivity analyses showed that the never strategy remains the less costly when the probability of a first FN episode was less than 49%, the probability of FN recurrence was less than 60%, or the CFN was less than $6,077, or the CCSF was greater than $863. In terms of patient comfort, the never strategy was the best choice, except for patients who considered that a course of G-CSF caused little or no discomfort, whether or not it prevented FN. CONCLUSION: Routine use of G-CSF during SCLC chemotherapy is not justified by clinical benefits, improved patient comfort, or economic considerations.