Background Linezolid (LNZ) is used to treat cerebral infections in critically ill patients. However, adaptation of dosing regimens should be considered (1,2). In vitro Hollow-Fiber Infection Model (HFIM) are used to test dosing regimens but are generally performed in rich medium such as Mueller-Hinton broth (MHB) which can lead to difficult in vivo extrapolations. The aim of this study was to compare the antibacterial efficacy of LNZ in artificial cerebrospinal fluid (aCSF) and MHB on meticillin-susceptible Staphylococcus aureus (MSSA) in a HFIM. Methods HFIM (n=2) were performed, both in aCSF and MHB, on a MSSA strain (ATCC29213) at an inoculum of 106 CFU/mL for 96h. CSF concentrations, obtained from PK simulations performed with the population PK model of LNZ in intensive care unit infected patients (PHRCN 16-0501), were reproduced in HFIM for three dosing regimens: 600 mg q12h, 900 mg q12h, 900 mg q8h. MIC (n=2) were performed on ATCC29213 before and after each HFIM in MHB and aCSF. Results Results are presented on Fig. 1. In MHB, an exponential growth was observed reaching a plateau at 1010 CFU/mL from 6h in the absence of LNZ. Bacterial counts were globally stable at 106 CFU/mL when LNZ is administered at 900mg q8h but for 600 and 900 mg q12h, a regrowth was observed respectively from 6h and 72h, reaching 1010 CFU/mL and 108 CFU/mL at 96h. On the contrary, without LNZ in aCSF, bacterial counts remained constant until 72h and a slight decrease was observed thereafter reaching 105 CFU/mL at 96h. The same trend was observed with the three dosing regimens tested. MIC was the same (4 ± 2 mg/L) before and after LNZ exposure. Conclusion Bacterial growth and LNZ efficacy observed in HFIM were different in aCSF compared to MHB. These results are in accordance with those obtained in TKC performed in human CSF (3), suggesting to use human-like medium in in vitro experiments for closer clinical extrapolations.