Correlation of pharmacokinetic/pharmacodynamic-derived predictions of antibiotic efficacy with clinical outcomes in severely ill patients with Pseudomonas aeruginosa pneumonia.

STUDY OBJECTIVE

To use pharmacokinetic/pharmacodynamic (PK/PD) modeling to correlate predicted antibiotic efficacy with actual clinical outcomes in patients with serious infections.

DESIGN

Retrospective cohort analysis.

SETTING

University medical center.

PATIENTS

A total of 182 adult intensive care patients with Pseudomonas aeruginosa pneumonia during a 5-year period from 2000 to 2004.

MEASUREMENTS AND MAIN RESULTS

The primary study end point was correlation of predicted antibiotic efficacy as determined by PK/PD modeling with actual clinical outcomes in individual patients. PK/PD analyses were conducted by determination of PD indexes using calculated patient-specific PK parameters and known pathogen minimum inhibitory concentrations, and by determination of predicted PD target attainment by using Monte Carlo simulation. Patients achieving PD targets were predicted to have clinically responded to therapy; patients not achieving PD targets were predicted to have failed therapy. A total of 128 patients (70%) apparently achieved desired PD targets; however, PK/PD modeling correctly predicted actual clinical outcome in only 47% of patients (86 of 182) with sensitivity of 49% and specificity of 43%. Percentages of patients apparently achieving PD targets were similar among those experiencing clinical response or clinical failure (67% vs 74%, respectively; p=0.344). Predicted achievement of PD targets was significantly associated only with reduction in intensive care unit and hospital lengths of stay. Achievement of PD targets was not significantly associated with clinical response by univariate or multivariate analysis, but factors related to severity of illness were significantly associated with clinical response.

CONCLUSION

PK/PD modeling did not accurately predict clinical or microbiologic success in patients with P. aeruginosa pneumonia. This study highlights the difficulties in applying PK/PD modeling at the level of the individual patient due to extreme PK variability and issues such as severity of illness. Antibiotic dosing based on sound PK/PD principles is strongly advocated, but additional studies are needed to confirm the role of PK/PD modeling in optimizing outcomes of patients with serious bacterial infections.