PyDarwin is an open-source Python package integrating machine learning with NONMEM for automated model selection; however, its utility in model-informed precision dosing (MIPD) remains unvalidated. This study aims to evaluate pyDarwin for population pharmacokinetics (PopPK) model development in MIPD compared with expert-developed models, using paediatric tobramycin as a representative case. The model-building dataset included 450 plasma concentrations from 63 children with cystic fibrosis (CF). One group manually developed a model using stepwise covariate selection, while another applied pyDarwin within a predefined 27-dimensional model space, using a fitness function that incorporated a penalty based on 24-h area under the curve (AUC24h). External validation was performed using 1011 concentrations from 144 children, including evaluation using all therapeutic drug monitoring (TDM) data and clinically relevant MIPD scenarios (a priori [AP], Bayesian 1, Bayesian 2). Both models included weight-based scaling, a maturation function, and estimated glomerular filtration rate as covariates on clearance, yielding comparable clearance estimates. The pyDarwin-selected model was a two-compartment model with fixed allometric scaling exponents, whereas the manual model was a one-compartment model with estimated exponents. Both models showed comparable probability of target attainment for steady-state AUC24h. In external validation using all TDM data, the pyDarwin-selected model demonstrated better predictive performance for CF patients than manual model (prediction-based MPE -14.6
Patient-centricity is increasingly recognized as an important objective in pharmaceutical development. However, despite growing emphasis on incorporating patient perspectives into medicine design, limited guidance exists on how patient-centric insights can be systematically translated into administration route and formulation selection during early development. This white paper discusses the challenge associated with integrating patient-centric considerations into formulation-related decision-making and presents the authors’ perspective on how these challenges may be addressed through a structured framework. To support this objective, the Patient-Centric Formulation Compass (PCFC) is proposed as a framework that links patient and disease understanding with administration route selection, formulation assessment, developability considerations, and downstream development planning. The framework consists of four sequential stages: 1) assessment of the target population and disease, 2) identification of suitable administration routes and formulation options, 3) evaluation of drug candidate compatibility and optimization opportunities, and 4) consolidation of the resulting knowledge into a transparent development summary. Rather than introducing new patient-centric principles, PCFC aims to operationalize and connect existing concepts within a structured and repeatable decision-support approach that may facilitate multidisciplinary discussion and formulation strategy development. The framework is proposed as a reference model that organizations may adapt to their own context and expertise. Its utility and impact should be evaluated through broader application across diverse development programs and therapeutic areas.
Translating pharmacokinetics (PK) and pharmacodynamics (PD) from preclinical to clinical is important in successful development of drugs for central nervous system (CNS) diseases. TAK-418 is a clinical candidate of lysine-specific demethylase 1 (LSD1) inhibitor with a unique mechanism of action; it covalently binds to flavin adenine dinucleotide (FAD) to form a compact adduct formylated FAD (F-FAD), thereby minimizing hematological side effect risks. In this study, we established a liquid chromatography-tandem mass spectrometry (LC/MS/MS)-based approach for evaluating F-FAD as a target engagement marker for TAK-418 to advance PK/PD translation and clinical dose setting. In rats, LC/MS/MS enabled detection of TAK-418-induced F-FAD formation in the brain with chromatographic separation from native FAD. F-FAD levels were positively associated with LSD1 protein abundance across brain regions and increased with increasing TAK-418 dose. LSD1 occupancy calculated from F-FAD concentrations was correlated with the inhibition level of LSD1 enzyme activity in rat hippocampus. This mechanism-based approach provides a link between covalent adduct formation, target engagement, and enzyme inhibition in the brain. In addition, similar point estimates of the half-maximal effective concentration (EC50) for F-FAD formation in rat and human hippocampal homogenates support the potential utility of F-FAD measurements for translational assessment of LSD1 target engagement. These findings provide useful information for evaluating the translational relationship between TAK-418 exposure and LSD1 target engagement and may support future PK/PD modeling and clinical dose selection for TAK-418. Moreover, our methodology for measuring covalent adducts would be applicable for multiple CNS drug candidates with covalent binding property.
A conservative BCS class III classification was recently proposed for amlodipine based on inconclusive permeability evidence. However, the scarce in vivo data suggest that amlodipine is slowly, though completely, absorbed. This means that the very rapid dissolution criterion associated with a BCS class III classification may be overly restrictive for immediate-release amlodipine products. The aim of this manuscript was to develop and validate a physiologically based biopharmaceutics model (PBBM) to assess the risk of bioequivalence failure associated with dissolution. Given that amlodipine is a lipophilic weak base, the absorption model incorporated a surrogate representation of intracellular lysosomal trapping through the FuEnt parameter, resulting in an adequate description of oral pharmacokinetics. Dissolution input included in vitro profiles obtained in phosphate buffer pH 6.8 at 50 rpm for different multisource products, fitted to either the Johnson model or the Z-factor approach. Although several products did not meet biowaiver dissolution criteria, virtual bioequivalence simulations conducted with the validated PBBM consistently predicted bioequivalence across tested formulations. Overall, these findings suggest that dissolution differences would have limited impact on the in vivo performance of amlodipine and support a low dissolution-related bioequivalence risk for immediate-release amlodipine tablets.
Nilotinib is reported to have a reduced maximal drug concentration (Cmax) in patients with hepatic impairment (HI). Given this compound is poorly soluble, highly lipophilic and has a positive food effect, reduced absorption in HI may be due to impaired micelle-mediated solubility resulting from reduced intestinal bile salts. PBPK modelling can be used to test this hypothesis as well as being a diagnostic tool to elucidate other potential mechanisms not captured by the base model. In this study, a mechanistic PBPK model for nilotinib was developed and its performance was evaluated in HI patients. A ‘learn-and-confirm’ approach was taken, using clinical food effect data to validate the model’s mechanistic sensitivity to bile salt-mediated solubility. The final model recovered clinical studies well, with area-under-the-curve (AUC) and Cmax being within 0.8 – 1.25-fold for fasted studies and 1.5-fold for fed/fasted ratios. AUC and Cmax were well recovered in mild HI (within 0.8 – 1.25-fold) and reasonably well recovered in moderate HI (within twofold) but significantly overpredicted in severe HI (> twofold). Bile salt concentrations were reduced across gastrointestinal segments for moderate and severe HI in line with in vitro data. Following twofold and 2.5-fold reductions in bile salts in moderate and severe HI, respectively, sensitivity was minor. Given uncertainty around the fugut, this was reduced to better recover the Cmax (within two-fold). PBPK modelling can be used as a diagnostic tool to evaluate impaired mechanisms in HI. PBPK modelling of nilotinib in cirrhosis. A physiologically-based pharmacokinetic (PBPK) model for nilotinib with mechanistic absorption has been validated in healthy subjects using food effect studies. Model performance was then evaluated in hepatic impairment. Where mispredictions were observed, intestinal bile salts were reduced inline with physiological observations. Alternatively, increased gut binding was identified as a potential source for reduced absorption in these patients. PBPK modelling has been proposed as a diagnostic tool for understanding absorption impairment mechanisms in hepatic impairment. Image adapted from Servier Medical Art ( https://smart.servier.com/ ), licensed under CC BY 4.0.
