The pharmacokinetic/pharmacodynamic (PK/PD) indices that define the therapeutic effect of the beta-lactam ceftazidime in a rat model of Klebsiella pneumoniae lung infection were investigated in relation to treatment duration and treatment endpoint. Treatment was started 24 h after infection with dosing regimens of 3.1 up to 1,600 mg/kg of body weight/day and dosing every 6, 12, or 24 h. When animals were treated for a relatively short period of 48 h, the duration of time that unbound plasma ceftazidime levels exceeded the MIC of the antibiotic for the infecting strain was the index that best correlated with therapeutic efficacy in terms of significant bacterial killing in the infected lung (microbiological effect). The maximum effect was reached when plasma ceftazidime levels were above the MIC for 60 to 70% of the dosing interval. However, when the treatment duration was extended to a relatively long period of 18 days instead of 48 h and animal survival rate instead of microbiological efficacy was taken as the endpoint, the fAUC/MIC ratio (where AUC is the area under the concentration-time curve) was the PK/PD index that best correlated with therapeutic efficacy. The PK/PD indices that effect 50% survival of rats for the fAUC/MIC ratios were 18.0 (95% confidence interval [95% CI], 16.3 to 19.9), 20.2 (95% CI, 13.8 to 29.4), and 27.9 (95% CI, 21.3 to 36.5) for the schedules of administration of every 6, 12, and 24 h, respectively. The fAUC/MIC needed for 100% survival was >100. We conclude that the PK/PD index that best correlates with outcome is dependent on the duration of treatment and/or the parameter of outcome. The effect of long-term treatment should be studied more extensively in other models of infection.