(M0930-03-21) Bioequivalence Prediction Based on In Vitro Flux Assay through the Example of Aripiprazole

Purpose: One of the biggest challenges of generic formulation development is to ensure appropriate performance during in vivo human correlation studies. In silico models supported by various in vitro assays are utilized to yield predictions, but the success rate of bioequivalence studies is still far from 100%. Current mathematical models for the estimation of the fraction absorbed (F_a%) of orally administered drugs base predictions on physicochemical properties. This study introduces a new mathematical model, based on the GUT framework[1,2], which predicts F_a% versus time profiles using in vitro flux profiles measured from formulations as input. The purpose of this study was to predict the in vivo performance of the solution drug product and compare it to the original solid dosage form using in vitro flux measurement as input for in silico modeling. Aripiprazole (ARI) has been selected as the model compound, and both the solid formulation (Abilify, Sample 1) and the oral solution (Sample 2) were studied. The presence of UV active preservatives in the solution made the direct UV analysis of the solution sample challenging. The obtained flux and the predicted F_a data were compared to in vivo human data published in the public assessment reports.

Methods: Small volume flux assay on µFlux- The formulations were tested using MicroFLUX™ apparatus. Concentration in both chambers was monitored using in situ fiber optic dip probes connected to the Rainbow instrument (Pion Inc.). A PVDF membrane impregnated with 25µL n-dodecane was applied to form a lipophilic barrier between the donor and the acceptor chamber. Blank channel - During the blank channel measurement, the absorption of the two parabens (without the active ingredient) was determined. This measurement served as a reference and was later taken into account when conducting the flux measurements of the solution. Multi-component regression - The method is based on a modified classical least squares (CLS) technique, which determines the contribution coefficient of the known spectra (the standard spectra of each component) by minimizing the difference between the calculated spectrum and the measured mixture spectrum. HPLC method - Chromatographic experiments were conducted on an Agilent 1100 HPLC system with a UV detector. Measurements were performed on a Kinetex column (50x4.6mm, 2.6μm) at 40°C, with a flow rate of 0.9 mL/min. 10 μL samples were injected using a mobile phase comprising water:ACN:acetic acid (70:30:0.1 v/v%) in isocratic mode. Fraction absorbed prediction - Predictions of in-vivo F_a% were generated using the fraction absorbed calculation of the Pion Predictor^TM software package.

Results: MicroFLUX assays were performed with both the solid and the solution formulations. While measurements for the tablet formulation were straightforward (tablet flux = 0.417±0.04 µg/cm²*min), the excipients in the solution formulation convoluted the UV concentration measurement. Two approaches have been tested to overcome this issue, and the average flux value for the Abilify solution, after blank channel correction, was determined to be 0.535±0.09 µg/cm²*min, while using linear regression the value was 0.521±0.07 µg/cm²*min. A statistical two-sample t-test was performed, indicating that there is no significant difference between the values obtained from the two different UV data evaluations. To further validate the effectiveness of both blank correction and linear regression, the final concentration on the acceptor side was measured by HPLC for two independent samples (Table 1). The results obtained from these comparisons demonstrated a small standard deviation (maximum 6%) during the flux evaluation. This indicates that both blank correction and linear regression are suitable methods for data evaluation. Predictor SW was used to predict the in vivo fraction absorbed % for both formulations. Published in vivo data indicates the solid dosage form provided a 87% fraction absorbed, while the solution outperformed the solid by ~20% based on the bioequivalence study, suggesting a fraction absorbed ratio ~100%. Predictor provided a slight overestimation for the solid dosage form (93%) and slight underprediction for the solution formulation (~94%), but the prediction accuracy is within the +/- 15% range which is considered an accurate prediction.<

Conclusion: The investigation focused on evaluating the suitability of blank correction and linear regression as evaluation methods for in vitro flux measurement of a multi-component drug formulation. Statistical analysis indicated no significant difference between the flux values obtained from these two evaluation approaches, which was also confirmed by HPLC. Fa% prediction by the predictor software provided an accurate estimation of the in vivo absorption ratio, even though the small difference reported in the bioequivalence study could not be resolved by this estimation.

References: 1. K. Sugano, Possible reduction of effective thickness of intestinal unstirred water layer by particle drifting effect, International Journal of Pharmaceutics 387 (2010) 103–109.
2. Sugano K. Biopharmaceutics Modelling and Simulations p266-321 (Aug 2012)

Table 1: Comparison of endpoint concentration determined by the 2 UV-based methods and HPLC.