The Application of the Design of Experiments and Artificial Neural Networks in the Development of a Fast and Straightforward HPLC-UV Method for Fluconazole Determination in Hemato-Oncologic Pediatric Patients and Its Adaptation to Therapeutic Drug Monitoring.

This study aimed to develop an optimized and wide concentration range HPLC-UV method for fluconazole (FLU) determination and its adaptation for pharmacokinetics (PK) studies in the pediatric population. The following parameters of chromatographic separation were optimized: retention time, tailing factor, peak height, and the sample preconditioning parameter, such as recovery. The optimization process involved the use of a central composite design (CCD) and Box-Behnken design (BBD) in the design of experiments (DoE) approach and a multilayer perceptron (MLP) for artificial neural network (ANN) application. Statistical and PK analyses were performed using Statistica and PKanalix. The acetonitrile (ACN) concentration revealed the most significant factor influencing the retention time, tailing factor, and peak height of FLU and the internal standard. For recovery, the extracting agent's volume was the most significant factor. In most cases, the analysis conducted with the DoE and ANN indicated the same factors in a similar order regarding their impact on the analyzed variables. The optimization process allowed for achieving a wide range of determined concentrations (0.5-100 mg/L) and ~100% recovery. The developed method enabled PK analysis of 12 samples from three pediatric patients, proving its clinical usability. The estimated median clearance (CL) and volume of distribution (Vd) were 1.01 L/h and 18.64 L, respectively. DoE and ANNs are promising and useful tools in the optimization of sample preconditioning as well as the HPLC separation procedure. The investigated fluconazole determination method meets the European Medicines Agency (EMA) validation objectives and might be used in pediatric and adult PK studies.