Killing curves and a sepsis model were performed with Streptococcus pneumoniae strains (MICs of penicillin = 0.01, 1, 2 and 4 mg/L) to assess the in vivo effect of in vitro early bactericidal activity. Optimal bactericidal concentration (OBC) was defined as the minimal concentration needed to obtain the maximal bactericidal activity during the sampling time for colony counting in killing curves. Animals were treated with amoxycillin, cefuroxime or cefpodoxime every 8 h for 48 h, with doses ranging from 2.5 to 50 mg/kg. ED100 (minimal antibiotic dose obtaining a 100% survival) was used as efficacy endpoint. Cmax/MIC, AUC/MIC and deltaT >MIC did not accurately predict efficacy against the most resistant strains, deltaT >OBC being the most predictive efficacy parameter indicating the in vivo effect of early bactericidal activity. Lower deltaT >OBC values for amoxycilin vs oral cehalosporins were needed for efficacy. The higher early bactericidal activity of amoxycillin may explain its higher in vivo efficacy.