Physiologically based pharmacokinetic (PBPK) modeling of the oral absorption of 5-flucytosine to support further development of a sustained-release formulation for the treatment of cryptococcal meningoencephalitis
Johanna Eriksson (1), Erik Sj÷gren (1), Jean-Yves Gillon (2), Vishal Goyal (2), Stephen Robinson (2), Henri Caplain (2), Isabela Ribeiro (2), Marylore Chenel (1)
(1) Pharmetheus, Sweden, (2) Drugs for Neglected Diseases initiative (DNDi), Switzerland
5-flucytosine (5FC) is used for the treatment of cryptococcal meningoencephalitis (CM). Due to its short plasma half-life and safety profile, it is currently dosed four times a day. This frequent dosing involves high risks of low compliance with potential consequences for both pharmacodynamic effect and toxicity [1,2]. To address this problem, DNDi is developing a sustained release (SR) formulation to decrease frequency of administration and thus improve the treatment. A model-informed drug development (MIDD) strategy was implemented to inform decisions along the project . In a first step, a PBPK model was developed to support the selection of SR formulations to be studied. From this first work, 3 SR formulation prototypes were selected to be tested for safety and plasma pharmacokinetics (PK) in fasted healthy participants (study 1), along with a commercial immediate-release (IR) tablet. To further support dosing and formulation selection in a subsequent clinical study (study 2: fed study in healthy participants) the legacy PBPK model was updated with the PK data obtained in study 1.
The aim of this analysis was to predict the food effect in healthy participants for the SR and IR formulations to suggest a dosing regimen to move forward with in the study in fed healthy volunteers.
PBPK modeling was performed in PK-Sim v.9.1. A legacy PBPK model for 5-FC had previously been developed based on literature data of an IR formulation . The PK data obtained from study 1 was used to further update and refine the legacy model. The final model included a fraction unbound in plasma set to 97%  and elimination attributed to glomerular filtration . The intestinal permeability was estimated to a high value to describe the fast Tmax seen for solution  and Weibull functions were estimated for the IR and SR formulations to describe the slower absorption for undissolved formulations. In addition, the colonic absorption was decreased to describe the observed PK data.
The model was evaluated by visual inspection of the concentration-time profile and comparison of simulated versus observed PK parameters. To predict the food effect on PK for 5-FU, the gastric emptying time (GET) was prolonged in accordance with the available implementation in PK-Sim. The solubility of 5-FC was not considered to be altered by food intake as it is a drug with high water solubility, and thus not expected to be affected by solubilization by bile salts. A therapeutic interval of Cmax not higher than 100 mg/L and Ctrough between 20-70 mg/L was considered in this analysis.
The refined 5-FC PBPK model could well describe the PK data in study 1, with absolute average fold error (AAFE) values for AUClast and Cmax of 1.13 for both parameters. The SR formulation with the fastest release (formulation D) had the highest bioavailability (62%) of the SR formulations relative to IR tablet in study 1. Prolonged GET enables longer time for absorption and the predicted relative bioavailability for formulation D in fed state increased to 80%, still having the highest bioavailability of the three SR formulations. The estimated dose to be within the therapeutic target concentrations with a bi-daily dosing were approximately 5000 mg in fed state and 6000 in fasted state. According to simulations, a dose of 6000 mg can be given as a first dose regardless of prandial state (if unknown for unconscious patients when admitted to the hospital) without exceeding the therapeutic Cmax concentration. In addition, the simulations do not support a loading dose to be given together with the SR formulation to achieve comparable PK as the IR formulation, since the simulated difference in time to reach same Cmax as IR was only about 20 min.
The SR formulation D was recommended to move forward with in the development due to the beneficial PK compared to the other two SR formulations. In addition, a dose of 5000 mg was suggested by the simulations to achieve therapeutic target concentrations in fed state. The need for loading dose for the SR formulation to be comparable to the IR formulation was not supported by modeling. The risk of exceeding the therapeutic interval, as a consequence of uncertainties in prandial state, was predicted to be low even at a higher starting dose (6000 mg). The presented model will be refined based on data from study 2 and applied for the subsequent step in the MIDD strategy, to inform the study design of Phase II.
 Vermes A, Guchelaar HJ, Dankert J. Flucytosine: a review of its pharmacology, clinical indications, pharmacokinetics, toxicity and drug interactions. The Journal of antimicrobial chemotherapy 2000; 46(2): 171-9.
 Archibald LK, Tuohy MJ, Wilson DA, et al. Antifungal susceptibilities of Cryptococcus neoformans. Emerging infectious diseases 2004; 10(1): 143-5.
 EFPIA MID3 Workgroup, Marshall SF, Burghaus R, Cosson V, Cheung SY, Chenel M, DellaPasqua O, Frey N, Hamrén B, Harnisch L, Ivanow F. Good practices in model‐informed drug discovery and development: practice, application, and documentation. CPT: pharmacometrics & systems pharmacology. 2016 Mar;5(3):93-122.
 Pai MP, Bruce H, Felton LA. Clinical pharmacokinetics of oral controlled-release 5-fluorocytosine. Antimicrobial agents and chemotherapy. 2010 Mar;54(3):1237-41.
 AHFS Drug Information 2010. McEvoy GK, ed. Flucytosine. Bethesda, MD: American Society of Health-System Pharmacists; 2010:567-70
 Cutler RE, Blair AD, Kelly MR. Flucytosine kinetics in subjects with normal and impaired renal function. Clinical Pharmacology & Therapeutics. 1978 Sep;24(3):333-42.