Reevaluation of current susceptibility breakpoints for Gram-negative rods based on pharmacodynamic assessment

Although pharmacodynamic (PD) modeling is now being considered for decision support for susceptibility breakpoint determination against Gram-negative bacteria, these PD-derived breakpoints should be verified using a clinically applicable population of organisms. In this analysis, a 5000-patient Monte Carlo simulation was used to determine PD breakpoints, the highest 2-fold MIC in which the probability of bactericidal target attainment (PTA) remained > or = 90%. Percent susceptibilities for 639 Pseudomonas aeruginosa, 103 Acinetobacter baumannii, 705 Escherichia coli, and 418 Klebsiella spp. collected during the 2004 Meropenem Yearly Susceptibility Test Information Collection surveillance study were then defined according to the PD-derived breakpoint (%S(PD)) and compared with the current Clinical Laboratory Standards Institute (CLSI)-defined breakpoints (%S(CLSI)). %S(PD) and %S(CLSI) were compared with the bactericidal PTA for each pathogen population to determine the degree of agreement. Resulting PD breakpoints were drug and dose dependent; moreover, values were commonly 2 to 4 MIC dilutions lower than CLSI breakpoints. Overall, %S(PD) more closely agreed with the PTA for the tested beta-lactam and fluoroquinolone dosing regimens. In contrast, %S(CLSI) overestimated PTA for many dosing regimens, especially against Pseudomonas: piperacillin/tazobactam 4.5 g qid (+9.7%), ciprofloxacin 0.4 g bid (+13.7%) and 0.4 g tid (+9.3%), and levofloxacin 0.5 g every 24 h (+22.4%) and 0.75 g every 24 h (+9.9%). Differences were most pronounced against the nonfermenting Gram-negative bacteria and were not observed among the Enterobacteriaceae. As a result, a new method of breakpoint classification is proposed, which is dosing regimen and pathogen specific, and is designed to denote isolates as susceptible only if target bactericidal exposures are achievable with the dosing regimen selected.