The analysis was performed using pooled PK data from a total of 14 single dose PK studies comparing nilotinib capsules and/or tablets given at varied dose levels and under varied food conditions. Single doses of nilotinib using the tablet formulation ranged from 50 to 300 mg and for the capsule formulation the examined doses ranged from 50 to 400 mg. The nilotinib concentration data was utilized for PK evaluation of absorption and elimination profiles and incorporated into the PopPK model (Supplementary information, Table 1). The novel (Danziten™, nilotinib tablet) formulation differs in that it is a tablet instead of a capsule, utilizes another soluble salt form of the active moiety (tartrate instead of hydrochloride monohydrate used in the capsule), and employs a modified composition and process for bioavailability enhancement. The analysis focused on doses of 142 mg (71 mg × 2) and 190 mg (95 mg × 2) delivered via tablets and 300 mg (150 mg × 2) and 400 mg (200 mg × 2) delivered via capsules.

In addition, to characterize the impact of dose administration conditions on the PK of nilotinib exposure, the comparative PK time vs. concentration profiles of nilotinib tablets and capsules were assessed under various single dose administration conditions in Phase one in vivo clinical investigations (ZenRise, Hyderabad, India). The dose ranges evaluated ranged from 50 to 300 mg for nilotinib tablets and from 50 to 400 mg for capsules.

The same data from the 14 single dose studies, was subsequently leveraged to develop a PopPK model and support the prediction of the steady-state PK profiles for nilotinib. Simulations were conducted for both formulations, under different dose levels and food conditions.

Pharmacokinetic evaluation was performed by GK Analytics, Hyderabad, India using SAS® for windows (SAS Institute Inc., USA) version 9.4 and included Cmax, AUC0-t, and AUC0-inf for both nilotinib tablets (test) and capsules (reference). Bioequivalence analysis for nilotinib exposures between test and reference formulations was based on the 90% confidence intervals (CI) for the ratio of the geometric least-squares means for the analysis of variance (ANOVA) of the ln-transformed PK parameters of Cmax, AUC0-t, and AUC0-inf falling within the accepted bioequivalence range of 80 to 125% (SAS version 9.4; SAS Institute Inc., USA).

The PopPK analysis (Fig. 1) was conducted to fulfill the following objectives: (1) characterize the PK of nilotinib from 14 in vivo clinical studies in healthy subjects and evaluate the impact of potential covariates on nilotinib exposure; (2) predict the steady-state PK obtained following dosing of nilotinib tablets and simulate the effect of dosing under fasted and varied non-fasting conditions on the peak-to-trough and overall exposure of nilotinib; and (3) evaluate the steady-state BE of the novel nilotinib tablets compared to capsules under fasted conditions.

Overview of the PopPK model development and PK simulation. A: new nilotinib tablets created; B: evaluation of formulation PK and safety in vivo Phase 1 clinical investigations in healthy subjects; C: formal analysis for bioequivalence of nilotinib tablets to capsules; D: Incorporate nilotinib tablet data into new PopPK model; E: leverage existing data and PopPK model [16] on nilotinib capsules, F: refine model via testing, validation steps, and retesting; G: run model simulations and generate predictions on steady-state PK under various conditions and effects of covariates

Nonlinear mixed-effects modeling software (NONMEM® version 7.4.3), a software package for nonlinear mixed-effects analysis (ICON Development Solutions, Dublin, UK), was used for population PK modeling. The first-order conditional estimation method of NONMEM with interaction (FOCE INTER) was used for PK model development. The published PopPK model by Larson et al. [16] was considered as a reference point for initiation of the analysis, but no data from it were included. Alternative absorption and disposition model structures were evaluated to refine the model based on the available clinical data from the 14 studies. Once a base structural model was identified, individual estimates of PK parameters of interest were evaluated for relationships with potential covariates, including body weight, age, sex, creatinine clearance, total bilirubin, alanine aminotransferase, aspartate aminotransferase, nilotinib tablet/capsule formulation, formulation lot number, and nilotinib dose (Supplemental information, Tables 2 and 3). Covariates that demonstrated a statistically significant (P < 0.01) effect were examined using a stepwise modeling approach using a forward addition (P < 0.01) and backward elimination (P < 0.001) methodology. Statistically significant covariates were retained in the final model. Standard goodness-of-fit assessments, visual predictive checks, and non-parametric bootstrap analyses were used to evaluate the performance, robustness and predictive ability of the final model (details of model development are provided in the Supplemental information, pages 1–7).

Pharmacokinetics were evaluated under four conditions: (1) ‘fasted’: nilotinib capsules or tablets were administered after an overnight fast of ≥ 10 h; (2) ‘fed’: After an overnight fast of ≥ 10 h, subjects consumed a high-fat high-calorie breakfast (800–1000 kcal, non-vegetarian, about 50% fat content) 30 min prior to dosing and completed within 30 min; (3) ‘modified fasting with a low-fat meal’: Fasting for ≥ 10 h overnight and administration of study drug 2 h after the start consumption of a low-fat meal (light fat high calorie breakfast, 400–500 kcal, about 25% fat content); and (4) ‘modified fasting with a high-fat meal’: Fasting for ≥ 10 h overnight and administration of study drug 2 h after the start consumption of a high-fat meal. Subjects did not consume additional food until 4 h post-dose under all prandial conditions.

The final PopPK model was used to predict concentration–time profiles for nilotinib at steady state following BID dosing under varied fasted and non-fasted conditions. Steady-state was achieved simulating up to 10 days of dosing; fed conditions as well as mealtimes were assumed to remain consistent throughout the simulated dosing period according to the specified scenario. The simulations were conducted in R (version 4.2.0) and included both fixed and random effects (inter-individual as well as residual variability). A dataset of 50 healthy subjects (50% male/50% female) was randomly selected from the analysis dataset and formed the basis for simulations. From these individual profiles, AUCss, Cmaxss, and Cminss were calculated over a 0 to 12-h interval following the last dose received in the 10-day dosing period. Exposure parameters were then used for formal BE testing, performed with an appropriate ANOVA model on the log-transformed individual estimates for the PK parameters. A total of 100 study replicates were simulated to establish the probability of BE between the two formulations under fasted condition in a steady-state BE study setting.