(W1030-09-51) Early-Phase Form Screening of Solubility-Limited Drugs Via In Vitro Characterization and In Silico Modeling

Purpose: As a drug candidate graduates from drug substance to drug product, the path is rarely linear. Early-phase pharmaceutical drug development is becoming increasingly complex and often the drug substance team is isolated from the drug product team resulting in an inefficient transitioning process and prolonged timelines. Additionally, many active pharmaceutical ingredients (APIs) rely on strategies for enhancing bioavailability through enabled formulations such as micronization, spray-dried dispersions (SDDs), salts/co-crystals, and hot-melt extrusions (HMEs). Each technology requires significant development time, costs, and API quantities. Thus, it is important to bridge the gap between drug substance and drug product via a streamlined early-phase screening approach to enable quick selection of an optimal form/formulation for in-vivo studies while minimizing time, costs, and materials. This study utilizes a combination of in-vitro analytical testing and in-silico physiologically based pharmacokinetic (PBPK) modeling to understand the potential risks and enhancements of multiple enabled forms compared to the free form of the API. Specifically, this screening approach was used to understand the potential enabled form/formulation landscape of erlotinib, a well understood tyrosine kinase inhibitor currently marketed as an HCl salt.

Methods: The API is initially analyzed via standard analytical methods including powder x-ray diffraction (PXRD), thermogravimetric analysis (TGA), differential scanning calorimetry (DSC), polarized light microscopy (PLM), nuclear magnetic resonance (NMR), and amorphous and aqueous solubilities in biorelevant media. This structural and analytical data acts as an input for in-silico PBPK modeling. ADMET Predictor^® and GastroPlus^® software suites from SimulationsPlus^® use Quantitative Structure Activity Relationship (QSAR) and PBPK models to predict risks, sensitivities, and the impact on absorption from formulation changes. Sensitivities identified from in-silico predictions are not guaranteed, but flag potential risks that should be monitored during further in-vitro and in-vivo testing. GastroPlus^® uses the combination of ADMET^® predicted constants (through structural analysis), in-vitro parameters, and baseline model assumptions (Image 1) to predict how absorption is influenced by common parameters. The absorption risk assessment (ARA) coupled with in-vitro solubilities identify if an enabled form will overcome the solubility limitations of the free base. For this study, the HCl salt was synthesized and used as a control. Additionally, attempts to reproduce other salts/co-crystals reported in literature resulted in two more crystalline (salt or co-crystal) forms from glutaric acid and adipic acid¹, of which one is an unreported polymorph. All four forms were compared using both in-silico and in-vitro data.

Results: Analytical characterization of the Erlotinib free base is summarized in Image 2. The API is highly crystalline with a needle-like morphology, and has a moderately high T_m/T_g (melting temperature/glass transition temperature). Image 3 shows graphically how, for a T_m/T_g of ~1.57 and cLogP of 3.5, there is a high propensity for crystallization, instability, and limited dissolution with high drug loading in ASDs  (amorphous solid dispersions) indicating high risk with this formulation. The ARA of the free form also indicates likely sensitivity to dose, dissolution rate/particle size, and gastric pH. In-silico modeling suggests that a salt form is likely to overcome the absorption risks of the free form with low risk. Upon generation of the three salt/ co-crystal forms, in-vitro characterization and subsequent in-silico modeling agrees with the formulation prediction. In particular, the three identified salt forms are predicted to have increased absorption with high doses (tested up to 1000mg) and remove the risk of elevated gastric pH (a result of the fed state and patients on Proton Pump Inhibitors (PPIs)). At pH 4, the percent of the dose absorbed for erlotinib free base drops from ~15% to 0% while the corresponding HCl salt decreases from ~70% to ~60% absorbed. Whereas the glutaric and adipic acid forms are above 70% absorbed up to pH 8. Additionally, at a 100 mg dose, the glutaric and adipic forms reach ~70% of dose absorbed after 3 hours while the HCl salt and free base only reach ~30% and ~10% absorbed at 3 hours respectively.  It is thus possible that either the glutaric or adipic acid salt/co-crystals may outperform the HCl salt in in-vivo studies.

Conclusion: A combination of in-vitro and in-silico analysis suggests that a glutaric or adipic acid salt/co-crystal is likely to exceed the performance of both the free base and the HCl salt. Specifically for oncology patient populations on PPIs, estimated to be ~33%², the glutaric and adipic acid forms overcome the decreased absorption of the HCl salt at pH 4 and above. Application of this screening method prior to drug product initiation for erlotinib would indicate a choice of the glutaric or adipic acid salt/co-crystal for in-vivo studies. In general, this streamlined method of early-phase screening allows for efficient rank ordering of multiple forms using a combination of in-silico modeling and in-vitro characterization of the free form of the API. From this list, enabled formulation/s can be analyzed efficiently, providing lead formulation/s for further development.

References: 1. Sanphui, P., Rajput, L., Gopi, S. P. & Desiraju, G. R. (2016). Acta Cryst. B72, 291-300.
2. Smelick, G.S., Heffron, T.P., Chu, L., Dean, B., West, D.A., DuVall, S.L., Lum, B.L., Budha, N., Holden, S.N., Benet, L.Z., Frymoyer, A., Dresser, M.J., & Ware A.J. (2013) Mol. Pharmaceutics 10 (11), 4055-4062

Image 1. Summary of baseline assumptions the PBPK modeling uses to predict formulation risks, sensitivities, and bioavailability enhancement extent.

Image 2. Summary of analytical characterization for the free base and salt/ co-crystal forms.

Image 3. Graphic representation of how the Tm/Tg and LogP affect the risk of crystallization, instability, and limited dissolution rate in SDD formulations (generated in-house).