EXECUTIVE SUMMARY SPONSORED CONTENT Structured Formulation Development of Poorly Soluble Drugs Guided by a Virtual Formulation Assistant New digital tools can aid solubilization approach selection and structured formulation development for poorly soluble oral drugs INTRODUCTION Formulation of a poorly soluble active pharmaceutical ingredient (API) for oral administration presents several challenges, particularly during early development. The suitability of various formulations must be evaluated based on limitations of solubility and bioavailability, appropriate excipients and manufacturing techniques, and prototype formulations must be screened for drug load, stability, and potential in vivo performance. This process can be costly and time consuming. This paper reviews several systems and parameters that can facilitate this structured formulation development process and shows how they are implemented in BASF’s ZoomLab digital application, with particular attention to how these tools can identify potential solubility issues and guide formulators in the rational selection of a suitable drug formulation approach. PHARMACEUTICAL CLASSIFICATION AND BIOAVAILABILITY Two major classification systems can be used to identify potential limitations in oral bioavailability of a new API. The Biopharmaceutics Classification System (BCS) was established in 1995 by Amidon et al. (1), who based their system on the correlation of drug dissolution and intestinal permeability with the rate and extent of drug absorption. The oral bioavailability of a drug was estimated based on its solubility in 250 mL of aqueous buffer solution, as well as the fraction absorbed from the intestinal lumen. The BCS has particular applicability for regulatory purposes; however, due to its simplicity, it has also been used for drug development despite being overly discriminative for numerous formulation candidates. Butler and Dressman (2) adapted the BCS for drug development in their Developability Classification System, which is based on drug solubility in 500 mL fasted statesimulated intestinal fluid (FaSSIF), as well as effective intestinal permeability. Class I drugs under either system are expected to have high bioavailability due to their high solubility and high permeability, while Class IV drugs are expected to have low Martin Hofsäss Development, Pharma Solutions BASF Sponsored by EXECUTIVE SUMMARY SPONSORED CONTENT bioavailability due to both low solubility and low permeability. Class II drugs, with low solubility and high permeability, and in some cases Class III drugs, with high solubility but low permeability, require appropriate excipients to ensure adequate oral bioavailability. Class II subclasses in the DCS differentiate between drugs with limited solubility (Class IIb) and those with dissolution rate-limited bioavailability (Class IIa). ZoomLab can calculate the DCS class for an API based on input parameters including target product dosage strength, FaSSIF solubility, and effective intestinal permeability. Intestinal permeability input can be based on actual values, either entered manually for a new API or loaded from the drug database for existing APIs, or can be calculated based on molecular weight, hydrogen bond donor/acceptor counts, partition coefficient, and topological polar surface area (TPSA) data entered manually or imported from PubChem. For example, the antifungal drug, itraconazole, is classified by ZoomLab as DCS Class IIb, with a calculated solubility limited absorbable dose of 1 mg and a maximum d90 particle size distribution cutoff-diameter of 4 microns for an adequate dissolution rate. In addition to the DCS classification of a given API, ZoomLab also provides general formulation advice based on the properties of the API. In this example, ZoomLab recommends using solubilization approaches for itraconazole such as addition of surfactants or use of lipid-based or amorphous solid dispersion formulations, since conventional formulation would likely not result in sufficient bioavailability. SOLUBILIZATION APPROACH SELECTION Once the limitations of bioavailability have been identified for an API, the next step in formulation is to evaluate suitable solubilization approaches. There are many possible approaches for poorly soluble drugs, including amorphous solid dispersions (ASD), lipid-based formulations (LBF), complexation with cyclodextrin, or colloidal delivery systems such as nanoparticles. The most frequently used approaches are ASD and LBF. Many parameters have been described in the pharmaceutical literature that can be used to evaluate the probability of success for ASD versus LBF given an API’s structural and chemical properties. While specific cutoff values have been defined for some parameters, such as a molecular weight of >300 Da for good glass forming ability related to ASD formulations or a melting point below 150° C for LBF, other favorable chemical parameters have been more vaguely defined. We have collected molecular property data for APIs that have been successfully formulated in ASD versus LBF formulations to examine which previously described parameters were discriminative (FIGURE 1) and can be use in a classification tool. Surprisingly, melting point and partition coefficient did not differ significantly between ASD and LBF candidate drugs. While good lipid solubility could be observed for APIs with low melting points, more than half of the APIs in successfully developed LBF had melting points higher than the proposed cutoff value of 150° C. However, other parameters were more distinctly distributed: ASD formulations tended to have higher molecular weight and larger topological polar surface area (TPSA) compared to LBF formulations (FIGURE 1). Estimated solubility in soybean oil and Hansen solubility parameters (HSP) also differed significantly between APIs formulated as ASD versus LBF (data not shown). Based on our findings, we have developed two classification models for choosing between ASD and LBF techniques (FIGURE 2). The first is a decision tree that uses recursive EXECUTIVE SUMMARY SPONSORED CONTENT EXECUTIVE SUMMARY SPONSORED CONTENT good and the classification models indicate high similarity of physical, chemical, and computed parameters with other APIs typically formulated as an ASD. Based on the ratio of melting point to glass transition temperature of itraconazole, ZoomLab recommends a moderate drug load of 35–50% by weight to prevent accelerated recrystallization. ZoomLab can also query its database for API with similar properties and provide supplementary information regarding similar APIs in successfully formulated drug products, as well as additional information on potentially suitable manufacturing techniques. In the case of itraconazole, there is a direct match to the database and hot melt extrusion is suggested due to its good glass forming ability and melting point below 180° C. SOLUBILITY PARAMETERS AND EXCIPIENT SELECTION partitioning (left panel) based on HSP, molecular weight, solubility, and molecular structure. The second is a logistic regression model that calculates a probability score based on the aforementioned parameters (example using TPSA in the right panel). In these models, the properties of an API are compared to the previously collected data and probability of success for either formulation class is assigned based on similarity of the API to formulated drugs in the training data set. These probability scores are used to help ZoomLab users make informed decisions regarding their solubilization approach. The Decision Support for Solubilization tool then evaluates various formulation strategies and makes a recommendation to the user for a solubilization approach based on experimentally determined properties such as melting point and glass transition temperature of the API, as well as calculated structural and chemical properties. These can be entered either manually or may be imported from PubChem and include parameters such as molecular weight, hydrogen bonds, rotatable bonds, TPSA, and HSP parameters. Using the previous example API, itraconazole, ZoomLab suggests that its estimated solubility in lipids of only 0.3 mg/g is insufficient to achieve a reasonable drug load as an LBF. However, its glass-forming ability is expected to be After a solubility-enabling approach has been chosen, the next step is the selection of excipients that allow high drug loads and can provide good stability for the API in the formulation. Hansen solubility parameters (HSP) (3) are widely used to screen for polymers and solvents that are suitable for a given API. The Hansen solubility parameter theory states that two compounds with similar intermolecular cohesive energies from dispersion, dipolar, and hydrogen bond interactions are expected to be miscible or even mutually soluble (3). Solubility parameters for an API and an excipient with an HSP distance less than 3–5 as calculated using the Bagley method (4) are expected to yield good miscibility. Experimental determination of solubility parameters is laborious and time-consuming. However, HSP may be estimated based on a drug's molecular structure by identifying specific functional groups on a molecule and assigning parameter values according to their contribution to the total cohesive energy. This process is automated in ZoomLab, using the group contribution methods of Van Krevelen (5), Just (6), and/or Hoy (7). The simplified molecular-input line-entry system (SMILES) string of the API can be input manually or imported from PubChem to generate the molecular structure and the EXECUTIVE SUMMARY corresponding parameter values of each functional group are assigned to the total calculated HSP. Depending on the molecular size and types of functional groups, some group contribution methods may not be applicable. For example, the Hoy method cannot be applied to intraconazole due to its molecular size, but ZoomLab used the Van Krevelen and Just methods to compute its total solubility parameter at 25.68 and 22.84, respectively. The calculated solubility parameters can then be used to find the most suitable polymers and solvents for an API. The integrated ZoomLab excipient database provides a searchable database of all excipients currently included in ZoomLab, from both BASF and non-BASF sources. Relevant excipients for a given formulation problem can be searched based on the excipient’s application purpose. In upcoming ZoomLab features, excipient properties such as HSP can be automatically screened to find the most suitable excipient for a given API. In the case of the example API, itraconazole, the most promising polymer candidates are identified as cellulose derivatives, copovidone, and povidone (HSP distances <2), SPONSORED CONTENT followed by polyethylene glycols (HSP distance 2.8) based on literature HSP data for the polymers and HSP calculated with ZoomLab for the active ingredient. Taken together, the results from ZoomLab and HSP comparison suggest that a potential prototype formulation of itraconazole for further experimental characterization in vitro would be an ASD based on cellulose or povidone derivatives with an intraconazole load of 35–50% by weight. CASE STUDY: FENOFIBRATE A second example is oral formulation of 200 mg fenofibrate, which like itraconazole shows limited solubility in FaSSIF. However, fenofibrate is further characterized by a lower molecular weight and thus a lower melting point, as well as a lower TSPA, when compared to itraconazole. Febofibrate was also classified by ZoomLab as a DCS Class IIb compound; however, unlike itraconazole, a larger dose of 42 mg was predicted to be soluble in intestinal fluid, and with micronization below 31 microns (FIGURE 3), a certain amount of fenofibrate would even be expected to be absorbed from the intestine in the absence of further solubilization. However, EXECUTIVE SUMMARY solubilization approaches would likely aid in maximizing oral bioavailability, so potential approaches were suggested by ZoomLab. Fenofibrate has a lower melting point and glass transition compared to itraconazole, and the contribution of dipolar interaction forces to the HSP is also markedly lower. No nitrogen atoms are present in the molecular structure, which leads to a higher expected solubility in lipids. These properties were reflected in the results of the ZoomLab Solubilization Decision Support tool (FIGURE 4). Estimated solubility in soybean oil was relatively high at 50 mg/g. While glass forming ability was expected to be good, only a limited number of APIs formulated as ASDs shared similar molecular properties when compared with fenofibrate. Therefore, ZoomLab recommended formulation in a lipid-based system. This was borne out by the list of similar drug products, where fenofibrate was found in at least two distinct LBF formulations, but an alternative formulation as an ASD was also present (FIGURE 4). The Solubility Parameter Calculation SPONSORED CONTENT tool was used to calculate HSP for fenofibrate. The most promising excipient candidate based on the HSP distance approach was polylactic acid (HSP distance 0.3), but many excipients typically used in LBF such as polyethylene glycols, pegylated emulsifiers, and poloxamers were also expected to be suitable (HSP distance ≤1.5). These HSP-based excipient recommendations were consistent with excipients used in marketed formulations of fenofibrate. SUMMARY Pharmaceutical classification systems and parameters have been defined which facilitate structured development of oral formulations for poorly soluble APIs. Digital solutions such as ZoomLab integrate these tools using scientific models to provide customized formulations advice related to selection of solubility-enabling formulations and compatible excipients for poorly soluble APIs based on their molecular properties, without the need to disclose proprietary information such as the chemical structure of the API. EXECUTIVE SUMMARY SPONSORED CONTENT REFERENCES 1. Amidon, G.L.; Lennernas, H.; Shah, V.P.; Crison J.R. A Theoretical Basis for a Biopharmaceutic Drug Classification: The Correlation Of In Vitro Drug Product Dissolution and In Vivo Bioavailability. Pharm Res. 1995 12 (3), 413–420. DOI: 10.1023/a:1016212804288 2. Butler, J.M.; Dressman, J.B. The Developability Classification System: Application of Biopharmaceutics Concepts to Formulation Development. J Pharm Sci. 2010 99 (12), 4940-4954. DOI: 10.1002/ jps.22217 3. Hansen, C.M. The Three Dimensional Solubility Parameter and Solvent Diffusion Coefficient: Their Importance in Surface Coating Formulation. PhD Dissertation, Danish Technical Press, Copenhagen, 1967. 4. Bagley, E.; Nelson, T.; Scigliano, J. Three-Dimensional Solubility Parameters and Their Relationship to Internal Pressure Measurements in Polar and Hydrogen Bonding Solvents. J Paint Tech. 1971 43 (555), 35-42. 5. Van Krevelen, D.W.; Te Nijenhuis, K. Cohesive Properties and Solubility. In Properties of Polymers, 4th ed; Elsevier, 2009; pp 189-227. 6. Just, S.: Sievert, F.; Thommes, M.; Breitkreutz, J. Improved Group Contribution Parameter Set for the Application of Solubility Parameters to Melt Extrusion. Eur J Pharm BioPharm. 2013 85 (3, Part B), 1191-1199. DOI: 10.1016/j.ejpb.2013.04.006 7. Hoy, K.L. Solubility Parameter as a Design Parameter for Water Borne Polymers and Coatings. J Coat Fabr. 1989 19 (1), 53-67. DOI: 10.1177/152808378901900106 Phantip/adobe.stock.com