3.3 Study Characteristics
The included studies were published from 2005 to 2019. The total quantity of subjects in included articles between 12 and 573. Most PPK analysis was performed in patients with epilepsy [24, 26-33], and only Antunes et al. [34] enrolled healthy subjects. One study was conducted only in infants [25], seven studies were conducted in children [24, 25, 27, 28, 30-32], two studies included both children and adults [26, 33], and two studies only enrolled adults [29, 34]. Subjects in all the included studies were administered OXC orally, with doses ranging from 75 to 1800 mg/d. In most studies, sparse sampling was employed, with 1 to 4 samples collected per individual. Detailed characteristics of all included studies were listed inTable 1 .
The PPK analyses were all conducted using population modelling software, including NONMEM (Icon, Dublin, Ireland) [24-26, 29-31, 34, 35], Phoenix NLME (Certara L.P. Pharsight, St. Louis, MO, USA) [26–28], and Monolix (Lixoft, Antony, France) [32]. The most commonly used algorithm was first-order conditional estimation with the η-ε interaction.
Most studies described the PK of MHD as a one-compartment model with first-order absorption and elimination. Rodrigues et al. [32] described the PK of OXC and its active metabolite MHD with a two-compartment model with first-order absorption and elimination. Antunes et al. [34] described the PK of OXC and its active metabolite enantiomers using a two-compartment model with transit absorption and first-order elimination.
All studies used either internal or external evaluation. The commonly method of external evaluation is the goodness-of-fit plot, while the second popular method was the visual predictive check. Although a normalised prediction distribution error (NPDE) has better properties [36], it is not commonly used [29, 30, 32, 34]. Four studies performed an external evaluation using an independent dataset, all of which showed acceptable predictability [28, 30, 31, 33].