1398. β-Lactam Probability of Target Attainment (PTA) and Penetration into Epithelial Lining Fluid (ELF) Based on Multiple Bronchoalveolar Lavage (BAL) Sampling Time Points in a Swine Pneumonia Model

BACKGROUND: Defining ELF concentrations is desired for antibiotics developed for pneumonia. For ethical reasons, BAL sampling in humans is routinely done at a single time point, thereby creating ambiguity in the precise ELF profile. It is unknown if additional sampling of the ELF would lead to more accurate estimates of exposure. The swine pneumonia model was used to characterize the full ELF profiles (5-BAL) of two β-lactams for comparison with models employing 1-BAL (1B) and 2-BAL (2B) sampling time points only. METHODS: Sixteen ventilated swine were infected with Pseudomonas aeruginosa to establish pneumonia and then treated for 72 hours with ceftolozane/tazobactam (C/T) 50 mg/kg q8h (n = 8) or piperacillin/tazobactam (TZP) 200 mg/kg q8h (n = 8). Plasma and BAL concentrations were measured in each swine at 1, 2, 4, 6, and 8 hours after the first dose. Urea correction was used to calculate ELF values. Ceftolozane and piperacillin plasma and ELF data were fitted to a two compartment model using the nonparametric adaptive grid program in Pmetrics. Hypothetical models were refitted after randomly selecting either 1B or 2B sampling time points from each swine. A 5,000 subject Monte-Carlo simulation was performed for each model to define PTA (60% free time above the MIC) and ELF penetration [area under the curve in ELF (AUC(ELF)) vs. free AUC(plasma)]. The KS-test was used to analyze distribution differences, reporting maximum vertical deviation (D) as percent difference; D < 20% was defined as negligible. RESULTS: Thirty-two C/T and 34 TZP plasma samples and 29 and 32 BAL samples were available for the full model, respectively; 1B and 2B sampling models used eight and 16 BAL samples. All models adequately fitted the data. C/T PTA at 4 mg/L was 94.8, 96.1, and 98.0%, for the full, 1B and 2B models. TZP PTA at 16 mg/L was 55.8, 46.8, and 46.7%, respectively. C/T median [interquartile range] penetration differences were negligible between the full (65% [25––109]) and 1B (72% [45––125], D = 15%) or 2B models (62% [32––111], D = 6%). TZP penetration differences were also minimal between the full (32% [9––67]) and 1B (17% [5––49], D = 18%) or 2B models (27% [9––44], D = 15%). CONCLUSION: These data suggest that antibiotic ELF models constructed from a single BAL time point result in similar exposure estimates to full ELF profiles. DISCLOSURES: G. Li Bassi, MSD: Grant Investigator, Grant recipient. A. Torres, MSD: Grant Investigator, Grant recipient.