Background: Midazolam (MZ) is commonly used in critically ill neurosurgical patients. Neuro-penetration of MZ and its metabolite, 1-hydroxy-midazolam (1-OH-MZ), is not well characterized. Objective: This study evaluated correlations between serum and cerebrospinal fluid (CSF) concentrations of MZ and 1-OH-MZ and assessed implications on patient sedation. Methods: Adults in the neurosurgical intensive care unit (ICU) with external ventricular drains receiving MZ via continuous infusion were prospectively studied. Serum and CSF samples were obtained 12-24 h and 72-96 h after initiation, and concentrations were determined in duplicate by high-performance liquid chromatography with tandem mass spectrometry. Bivariate correlation analyses used Pearson coefficient. Results: A total of 31 serum and CSF samples were obtained from 18 subjects. At sampling, mean MZ infusion rate was 3.9 ± 4.4 mg/h, and previous 12-h cumulative dose was 51.4 ± 78.2 mg. Mean concentrations of MZ and 1-OH-MZ in serum and CSF were similar between timepoints. Similarly, ratios of 1-OH-MZ to MZ in serum and CSF remained stable over time. Serum MZ (126.2 ± 89.3 ng/mL) showed moderate correlation (r2 = 0.68, P < 0.001) with serum 1-OH-MZ (17.7 ± 17.6 ng/mL) but not CSF MZ (3.9 ± 2.5 ng/mL; r2 = 0.24, P = 0.005) or CSF 1-OH-MZ (2.5 ± 0.6 ng/mL; r2 = 0.47, P = 0.30). CSF MZ did not correlate with CSF 1-OH-MZ (r2 = 0.003, P < 0.001). Mean serum ratio of 1-OH-MZ to MZ (0.14 ± 0.2 ng/mL) did not correlate with CSF ratio (1.06 ± 0.83 ng/mL; r2 = 0.06, P = 0.19). Concentrations and ratios were unrelated to MZ infusion rate or 12-h cumulative dose. Sedation was weakly correlated with CSF 1-OH-MZ, but not with serum MZ, serum 1-OH-MZ, or CSF MZ. Conclusion and Relevance: Continuous infusions of MZ result in measurable concentrations of MZ and 1-OH-MZ in CSF; however, CSF concentrations of MZ and 1-OH-MZ poorly represent serum concentrations or dosages. Accumulation of MZ and 1-OH-MZ in serum or CSF over time was not evident. Concentrations of MZ and 1-OH-MZ do not predict sedation levels, reinforcing that pharmacodynamic assessments are warranted.