Background: Fluconazole is a potent antifungal agent used for both prophylactic and systemic treatment of candidiasis. However, we do not have sufficient pharmacokinetic information for fluconazole in neonates. Current fluconazole dosing guidance relies on pharmacokinetic data from 12 critically ill preterm (26-29 wk) infants sampled on 1^st 14 days of life (Saxen et al., 1993).

Objective: The objectives of this study were to (1) Construct a population PK model of fluconazole drug disposition in preterm neonates from retrospective data (Saxen et al., 1993). (2) Perform clinical trial simulations incorporating parameter uncertainty for the design and evaluation of a prospective multi-center, fluconazole study in infants, with subsequent sensitivity analysis. (3) Perform simulations to evaluate the effectiveness of current dosing guidelines using different pharmacodynamic measures like time above MIC, 24h AUC/MIC ratio, C_max/MIC ratio and steady state trough levels.

Methods: Retrospective population PK analysis was performed using nonlinear mixed effect modeling (NONMEM). The model was validated using another dataset of slightly older infants (Wenzl et al., 1998). Predictive performance of the model was assessed by predictive check, and evaluating precision and bias for all parameters from the 500 bootstrapped datasets. Eight different fluconazole dosing regimens (6-12mg/kg, q24-72 hr) were explored via simulation (n = 500) in 48 patients (stratified on GA and PNA into six different groups) based on the final population PK model. Weight (Wt), Gestational age (GA), and postnatal age (PNA) were obtained from a previous trial in infants and NICHD growth chart calculator (http://neonatal.rti.org). MIC distributions (0.5-32 mg/l) from two different species of yeast, Candida albicans and Candida parapsilosis were used. Simulations incorporated posterior probability (uncertainty) distributions across all parameters; fixed-effect parameters from a multivariate normal distribution; and inter-subject and residual variance parameters from Inverse-Wishart distributions (as implemented in R/NONMEM toolbox for Simulation from posterior probability by Metrum Institute, Tariffville, CT, USA). A global sensitivity analysis was performed by examining precision and bias estimates for each of the model parameters.

Results: Fluconazole pharmacokinetics was best described by a two compartment model, with all parameters scaled by body weight and CL was modeled as a function of PNA (power-model). Population mean (%CV) parameter estimates for CL, V1, Q, and V2 were 0.012 (37%) L/hr, 0.621 (21%) L, 2.1 L/hr and 0.783 L. Simulation results from 500 trials showed good precision and minimal bias for CL and V1. Adult guidelines for invasive candidiasis recommend > 400mg/day, a dose that yields steady state trough levels ~12 mg/L and C_max >6mg/L. Simulation results showed that higher doses than the traditional dosing [6mg/kg q72 hr (<14d), q48 hr (14-28d), qd (>24d)] might be needed to reach the targeted metrics based on adult doing guidelines. A 10 mg/kg dose [(q72 hr (<14d), q48 hr (14-28d), qd (>24d)] or a 12 mg loading dose (followed by 6mg maintenance dose qd) might help achieve the target PD metric.

Conclusion: Population PK analysis of the fluconazole dose-exposure relationship in preterm neonates support the PK discrimination of patients by PNA and Weight. Simulations derived from this model contest current dosing practices in neonates based on the validity of target exposure metrics. Our ongoing prospective study will refine the proposed fluconazole population PK model over a broader demographic range, and hopefully optimize the treatment guidelines and help prevent the emergence of resistance.

References:
1. Saxen H, et al. Clin Pharmacol Ther 1993;54(3):269-277
2. Wenzl TG, et al. Eur J Pediatr 1998;157(8):661-2