Influence of culture site-specific MIC distributions on the pharmacokinetic and pharmacodynamic properties of piperacillin/tazobactam and piperacillin: A data analysis

Background: Investigators who perform pharmacokinetic/pharmacodynamic (PK-PD) modeling with Monte Carlo simulation have historically not stratified microbiological data by culture site. This lack of stratification might be problematic if susceptibility patterns differ among sites and might lead to differences in PK-PD. Objective: This study compared the PK-PD of 2 antimicrobial regimens against 5 gram-negative bacterial species form 3 culture sites. Methods: This data analysis was performed at the Department of Pharmacology, The University of Texas Health Science Center, San Antonio, Texas. Blood, pulmonary (ie, bronchial, endotracheal, lung, respiratory, sputum, and tracheal secretions), and wound distributions of MICs were extracted from the 2002 Intensive Care Unit Surveillance System database. Bacteria included Acinetobacter baumannii, Enterobacter cloacae, Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa. The PK properties of piperacillin/tazobactam (3.375 g every 4 hours) and piperacillin (3 g every 4 hours) were obtained from studies in healthy volunteers. Monte Carlo simulation was used in 10,000 patients for each antimicrobial-bacteria-culture site combination. The cumulative fraction of response (CFR) for a free percentage time above the MIC of ≥50% was determined for each combination, and a clinically significant difference was defined a priori as ≥10%. Results: Data from 2408 pulmonary, 490 blood, and 242 wound isolates were included. For piperacillin/tazobactam, the CFR varied <10% by culture site in all 5 bacterial species. Site-specific differences were noted in MIC₅₀ for piperacillin versus E cloacae and E coli and MIC₉₀ for piperacillin/tazobactam versus K pneumoniae and P aeruginosa. Likewise, for piperacillin, the CFR was similar among the 3 culture sites for P aeruginosa. However, the CFR for piperacillin varied by ≥10% for A baumannii (blood > wound), E cloacae (pulmonary > blood), E coli (pulmonary and blood > wound), and K pneumoniae (wound > blood). Conclusions: The PK-PD models based on PK properties found in healthy humans and site-specific MIC distributions in this study suggest that for piperacillin, culture-site differences subsequently resulted in CFR differences that exceeded a predetermined level of clinical significance. Furthermore, these data suggest that traditional reporting strategies for microbiological data (ie, MIC₅₀ and MIC₉₀) might fail to adequately characterize the MIC population.