Determination of meropenem penetration into the lung from Sparse data.

I read with interest the report by Lodise and colleagues on the penetration of meropenem into the epithelial lining fluid (ELF) of 17 patients with ventilator-associated pneumonia (VAP) (6). Briefly, they report a median penetration rate of meropenem into ELF of 25.4%, but Monte Carlo simulations predict an enormous variability (25th to 75th percentiles, 9.0 to 70.14%; 10th to 90th percentiles, 3.7 to 178%), which would leave many patients with insufficient drug exposure at the site of infection and thus possibly a poor clinical outcome. Like the authors themselves, I am somewhat surprised by the huge magnitude of variability of ELF penetration reported in this study, which exceeds by far the variability in the cited studies on the ELF penetration of piperacillin (median, 42%; interquartile range [IQR], 32 to 63.5%), ertapenem (median, 32%; IQR, 28 to 46%), ceftazidime (mean standard deviation [SD], 20.6 8.9%), and cefepime (mean concentration SD, 13.5 3.3 mg/liter in plasma and 14.1 2.8 mg/liter in ELF), all reported by Boselli et al. in patients with VAP or severe nosocomial pneumonia (1–4). Whereas it is of course possible that ELF penetration of meropenem is actually more variable, I suggest an explanation based on the macropharmacokinetic properties and the employed sampling strategy. In the cases of piperacillin, ceftazidime, and cefepime, steady-state concentrations (CSS) were determined during continuous infusion. With this method, it is impossible to describe the time course of ELF penetration but very simple to determine the area under the concentration-time curve (AUC steady-state concentration time) and thus the total drug exposure. As for ertapenem, the half-life of about 8 h may still permit the assumption that simultaneous concentrations in serum and ELF are close to equilibrium and representative of total penetration. In the present study, meropenem was given as an intermittent infusion (0.5 or 3 h) every 8 h. Given its short half-life (1 to 2 h), equilibrium between ELF and plasma is unlikely at any given time point, and only comparison of AUCs, as described by the authors, is valid. However, due to the practical limitations of ELF sampling, only one specimen was obtained per patient, and the full concentration-time course and AUC in ELF was estimated from this single concentration. Despite the usefulness of contemporary modeling software, I feel that this overstretches the information content of the actual raw data, which basically are 17 paired plasma and ELF samples from a multicenter study, apparently designed for different primary endpoints. This approach is likely to introduce much methodological uncertainty, adding to the already important error due to the analytical problems involved in the determination of ELF concentration (5). Therefore, I consider it likely that the variability derived from the Monte Carlo simulation overestimates the true interindividual variability, but this is no purely academic question. The answer could decide whether or not meropenem monotherapy for VAP is a rationale choice, and this calls for a more robust data basis from custom-tailored clinical trials. Emphasis should be put on an optimized dosing and sampling strategy, which may include continuous infusion, or analytical techniques allowing for serial measurements, e.g., intrapulmonary microdialysis. REFERENCES