Objective: We developed a model to predict the pharmacodynamic exposure of antibiotics against bacteria commonly implicated in nosocomial bloodstream infections to determine which dosage regimens would provide the greatest likelihood of obtaining a bactericidal effect.
Methods: Pharmacodynamic exposures were simulated for 5000 subjects receiving standard doses of ceftazidime, cefepime, piperacillin/tazobactam, meropenem, imipenem, or ciprofloxacin. Exposures were indexed to the MICs of bacteria weighted by their prevalence in causing nosocomial bloodstream infections, derived from 2002 SENTRY data. Enterococci were excluded. MIC data were derived from the 2003 Meropenem Yearly Surveillance Test Information Collection resistance study. The probabilities of achieving bactericidal exposures (ie, target attainment) for each antibiotic regimen were compared. The effect of increasing prevalence of methicillin-resistant Staphylococcus aureus (MRSA) on attainment of bactericidal targets was tested.
Results: All dosage regimens except ciprofloxacin and ceftazidime 1 g q8h achieved >90% likelihood of bactericidal exposure. The rank order of target attainment was as follows: imipenem 500 mg q6h, 100.0%; imipenem 1 g q8h, 99.9%; cefepime 2 g q12h, 99.4%; meropenem 1 g q8h, 98.4%; cefepime 1 g q12h, 98.2%; piperacillin/tazobactam 3.375 g q6h, 97.9%; piperacillin/tazobactam 4.5 gq8h, 95.0%; ceftazidime 2 g q8h, 94.2%; ceftazidime 1 g q8h, 71.7%; ciprofloxacin 400 mg q8h, 63.3%; and ciprofloxacin 400 mg q12h,63.0%. Target attainments dropped to <90% for all agents when MRSA was modeled at > or =10% prevalence.
Conclusions: The results of this model analysis suggest that standard doses of the carbapenems, piperacillin/tazobactam, and cefepime, and higher doses of ceftazidime, may provide optimal likelihood of achieving bactericidal exposure against pathogens implicated in nosocomial bloodstream infections, excluding MRSA and enterococci. When MRSA rates are > or =10%, therapy with an antibiotic that has activity against this phenotype should be empirically initiated.