Population pharmacokinetics of vancomycin in serum and cerebrospinal fluid: variability in cerebrospinal fluid concentrations despite targeted serum exposure.

Vancomycin is a key component of empirical therapy for central nervous system (CNS) infections, yet dosing is guided by systemic pharmacokinetic/pharmacodynamic (PK/PD) targets that may not reflect drug exposure at the site of infection. Cerebrospinal fluid (CSF) penetration is variable, and the relationship between serum and CSF concentrations remains poorly defined. This study aimed to develop a population pharmacokinetic model simultaneously describing vancomycin concentrations in serum and CSF using real-world therapeutic drug monitoring (TDM) data and to characterize the relationship between systemic and CSF exposure. A retrospective observational study was conducted in adult neurosurgical patients (n = 64) with secondary meningitis. Vancomycin concentrations in serum (n = 133) and CSF (n = 100) were analyzed using nonlinear mixed-effects modeling. Monte Carlo simulations were performed to evaluate exposure across dosing regimens and renal function strata. Additional simulations under targeted serum exposure (AUC24 ~500 mg·h/L) with minimized interindividual variability in clearance were used to isolate variability in CSF exposure. A two-compartment model with a CSF compartment adequately described the data. Clearance was significantly associated with the estimated glomerular filtration rate. The median CSF-to-serum AUC ratio was 0.29, with substantial interindividual variability. Simulations demonstrated that, despite consistent serum exposure, CSF concentrations remained highly variable, with median (5th-95th percentile) average CSF concentrations of 6.2 (3.2-10.4) mg/L. The probability of achieving higher CSF concentration thresholds declined markedly with increasing targets. Vancomycin CSF exposure exhibits substantial variability that is not reliably predicted by serum pharmacokinetics, even under standardized systemic exposure. These findings highlight limitations of serum-guided dosing for CNS infections and support the need for compartment-specific dosing strategies and PK/PD targets.