Abstract Introduction Levofloxacin and rifampicin are the preferred treatment for prosthetic joint infection (PJI) caused by Staphylococcus aureus, especially when managed with implant retention (DAIR). However, a significant variability of success has been reported, which could be related to intrinsic characteristics of the microorganism. Our aim was to evaluate the variability in the anti-biofilm response to levofloxacin and rifampicin in a clinical collection of S. aureus. Material and methods Eleven levofloxacin- and rifampicin-susceptible S. aureus isolates causing PJI managed with DAIR were included. Levofloxacin, rifampicin and levofloxacin + rifampicin were tested in an in vitro static biofilm model in microtitre plates, where 48 h biofilms were challenged with antimicrobials during 24 h. Additionally, two genetically similar strains were tested in the CDC Biofilm Reactor, where 48 h biofilms were treated during 56 h. Antimicrobial activity was assessed by viable biofilm-embedded cells recount, and by crystal violet staining. Results All antimicrobial regimens showed significant anti-biofilm activity, but a notable scattering in the response was observed across all strains (inter-strain coefficient of variation for levofloxacin, rifampicin and levofloxacin + rifampicin of 22.8%, 35.8% and 34.5%, respectively). This variability was tempered with the combination regimen when tested in the biofilm reactor. No correlation was observed between the minimal biofilm eradicative concentration and the antimicrobial activity. Recurrent S. aureus isolates exhibited higher biofilm-forming ability compared with strains from resolved infections (7.6 log10 cfu/cm2±0.50 versus 9.0 log10 cfu±0.07). Conclusions Significant variability may be expected in response to levofloxacin and rifampicin among biofilm-embedded S. aureus. A response in the lower range, together with other factors of bad prognosis, could be responsible of treatment failure.