Population pharmacokinetics and CSF penetration of flucytosine in adults with HIV-associated cryptococcal meningoencephalitis

Abstract

Background: There are limited data describing clinical flucytosine pharmacokinetics (PK). The variability of flucytosine partitioning into the CNS is not known. We described the interindividual variability in flucytosine PK in patients with HIV-associated cryptococcal meningoencephalitis. In addition, we quantified the extent and variability of CSF partitioning of flucytosine.

Methods: A PK study was conducted in 64 patients with confirmed HIV-associated cryptococcal meningoencephalitis in Blantyre, Malawi. A four-compartment PK model was developed, and Monte Carlo simulations were performed with flucytosine administered at different doses and in different schedules.

Results: The estimated mean apparent volume of the central compartment was 17.50 (SD 9.99) L; mean apparent clearance was 5.88 (SD 3.35) L/h; mean apparent volume of the CNS compartment was 41.73 (SD 13.66) L. From the Bayesian posterior estimates, AUC24 values at steady state (144-168 h) with doses of 25 mg/kg q6h were median (IQR) 890.38 (603.81-1213.70) mg.h/L in plasma and 595.66 (425.69-776.64) mg.h/L in CSF. The ratio of CSF:plasma AUC24 was 0.69 (IQR 0.58-0.82).

Conclusions: This study revealed significant interindividual variability in flucytosine PK in plasma and CSF in patients with HIV-associated cryptococcal meningoencephalitis. The population PK model is a first critical step for revised flucytosine regimens that maximize fungal killing and minimize toxicity and the emergence of resistance.

Figure 1.

Scatterplots of observed versus predicted values in plasma and CSF, for the chosen population PK model. Goodness-of-fit plots of observed versus mean predicted flucytosine concentrations for the chosen population PK model after the Bayesian step. Upper plots show the fit of the mean individual-level posterior PK estimates in plasma and CSF. Lower plots show the fit of the population-level PK estimates. Circles represent observed-predicted data points. Dashed lines represent the line of identity. conc, concentration; Obs, observed flucytosine concentration; Pred, predicted flucytosine concentration. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 2.

Visual predictive check. Simulations were performed to predict flucytosine concentrations in patients administered the standard regimen of 25 mg/kg q6h. Median (black line), 5th percentile and 95th percentile (grey lines) flucytosine concentrations were plotted per hour over 168 h. Observed flucytosine concentrations from our cohort were overlayed (crosses). Overall, 76% of plasma concentrations and 86% of CSF concentrations fell within the predicted 5th and 95th percentiles. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 3.

Simulated AUC_144–168 in plasma and CSF resulting from various flucytosine dosing regimens, based on the non-parametric prior distribution from the final PK model. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.