To describe and investigate the covariate effects of cerebrospinal fluid (CSF) amikacin pharmacokinetics in neonates, CSF samples were prospectively collected from neonates in whom amikacin had been initiated before a diagnostic lumbar puncture was performed. CSF analysis (amikacin concentration, white blood count [WBC], glucose content, and protein concentration) and amikacin therapeutic drug monitoring results (peak and trough concentrations) in serum were recorded. Correlations (Spearman rank) between the CSF amikacin concentration and the CSF WBC and glucose and protein concentration were investigated. There were 44 CSF amikacin concentrations and 83 serum samples available from 43 neonates (mean postmenstrual age, 36 weeks [range, 26 to 41 weeks]; mean weight, 2.43 kg [range, 0.87 to 3.86 kg]). The median time interval between initiation of amikacin administration and CSF sampling was 25 h (range, 2.5 to 93.7 h). The median amikacin concentration in the CSF was 1.08 mg/liter (range, 0.34 to 2.65 mg/liter), and the mean trough and peak amikacin concentrations in serum were 3.8 +/- 2.5 mg/liter and 35.7 +/- 5.9 mg/liter, respectively. A correlation between CSF amikacin and CSF protein contents (P < 0.01, r = 0.41, 95% confidence interval = 0.13 to 0.63) but not between CSF WBC and CSF glucose was documented. A two-compartment (central and CSF) linear disposition model was used to estimate population pharmacokinetics. The half time for equilibration (T(eq)) between serum and CSF compartments was used as a measure of blood-brain barrier permeability. The T(eq) was 7.58 h (coefficient of variation [CV] = 49.1%) with a partition coefficient of 0.103 (CV = 26.4%). There was no relationship between the T(eq) and CSF WBC, CSF glucose content, or CSF protein content.