Postinfusion data obtained from 17 patients with cystic fibrosis participating in two clinical trials were used to develop population models for ceftazidime pharmacokinetics during continuous infusion. Determinant (D)-optimal sampling strategy (OSS) was used to evaluate the benefits of merging four maximally informative sampling times with population modeling. Full and sparse D-optimal sampling data sets were analyzed with the nonparametric expectation maximization (NPEM) algorithm and compared with the model obtained by the traditional standard two-stage approach. Individual pharmacokinetic parameter estimates were calculated by weighted nonlinear least-squares regression and by maximum a posteriori probability Bayesian estimator. Individual parameter estimates obtained with four D-optimally timed serum samples (OSS4) showed excellent correlation with parameter estimates obtained by using full data sets. The parameters of interest, clearance and volume of distribution, showed excellent agreement (R2 = 0.89 and R2 = 0.86). The ceftazidime population models were described as two-compartment kslope models, relating elimination constants to renal function. The NPEM-OSS4 model was described by the equations kel = 0.06516+ (0.00708.CLCR) and V1 = 0.1773 +/- 0.0406 liter/kg where CLCR is creatinine clearance in milliliters per minute per 1.73 m2, V1 is the volume of distribution of the central compartment, and kel is the elimination rate constant. Predictive performance evaluation for 31 patients with data which were not part of the model data sets showed that the NPEM-ALL model performed best, with significantly better precision than that of the standard two-stage model (P < 0.001). Predictions with the NPEM-OSS4 model were as precise as those with the NPEM-ALL model but slightly biased (-2.2 mg/liter; P < 0.01). D-optimal monitoring strategies coupled with population modeling results in useful and cost-effective population models and will be of advantage in clinical practice, as it allows pharmacokinetic-pharmacodynamic modeling with sparse data, thus describing the relationship between ceftazidime exposure and response in the treatment of acute exacerbations in patients with cystic fibrosis.