The workshop included 10 breakout sessions to provide opportunities to discuss current challenges and limitations, including emerging approaches and techniques for complex formulations. Table I lists breakout session topics and main questions addressed. Overall, topics that repeatedly rose across sessions were the following: (1) meaning and assessment of “dissolved drug,” particularly of poorly water soluble drug in colloidal environments (e.g., fed conditions, ASDs); (2) potential limitations of a test that requires sink conditions for a poorly water soluble drug; (3) non-compendial methods (e.g., two-stage or multi-stage method, dissolution/permeation methods); (4) non-compendial conditions (e.g., apex vessels, non-sink conditions); and (5) potential benefit of having both a QC method for batch release and a biopredictive/biorelevant method for biowaiver or bridging scenarios. An identified obstacle to non-compendial methods is the uncertainty of global regulatory acceptance of such methods.

In vitro Approaches to Interpret/Predict Food Effects

This breakout was facilitated by Martin Brandl, Annette Bauer-Brandl, and Kimberly Raines. The main question was the following: What in vitro dissolution or dissolution/permeation methods can anticipate positive, negative, or a lack of food effects?

In vitro methods to assess the effects of food often aim to provide insights for formulation optimization, typically with the aim to minimize food effects or to create robust formulations, the performance of which is independent of food. Several in vitro and in silico tools are being explored to predict the direction and extent of food effects. However, the complexity of different interactions within the GIT and the inter-individual variability in vivo makes a reliable prediction of food effect challenging. Furthermore, food intake changes physiologic factors such as luminal hydrodynamics, splanchnic blood-flow, and induction of metabolic enzymes, which are difficult to capture in vitro.

Thus, simple bio-predictive tools typically consider drug/bile-salt interactions and range from simple equilibrium solubility studies in media mimicking the bile-salt and lipid composition of the intestinal fluid in different prandial states to “biorelevant” in vitro dissolution testing, to combined dissolution-permeation models base on tissue, cells, or cell-free systems. Approaches may use real-time analytics and biopharmaceutics modeling-simulation approaches.

Food effects should be assessed by comparing the dissolution of the drug product in two media. A higher drug release under fed state–simulated intestinal fluid (FeSSIF) conditions over fasted state–simulated intestinal fluid (FaSSIF) conditions indicates a higher drug dissolution in the media with the higher concentration of bile salts that are released in the intestine in fed sate. However, when determining food effects in vitro, multiple factors need to be investigated beyond dissolution and solubility, namely permeability, and the interplay between dissolution and permeation, as well as metabolism.

One example discussed was the investigation of a cyclodextrin (Cyd)-formulation of itraconazole which revealed moderately negative food effects based on the AUC ratio under fed and fasted conditions. The solubility of itraconazole in phosphate buffer solution, and in different concentrations of FaSSIF and FeSSIF media (which represent different concentration of bile salts) provided an insight into the interaction between itraconazole and bile salt micelles. The presence of bile salts resulted in the displacement of itraconazole from Cyd complex. This finding could serve as a starting point in understanding the need for additional types of dissolution studies, considering the interplay between Cyd, the drug, and bile salts. Two-stage dissolution/permeation methods and combined dissolution/permeation studies provided a greater understanding of the food effects. Such studies should be aimed at utilizing high-throughput techniques and also at developing a roadmap for in vitro studies.

Currently, FDA recommends an in vivo clinical food effect study for all extended-release oral products using the to-be-marketed formulation irrespective of the solubility and permeability of drug substance. Although there are no standardized in vitro tools nor an FDA requirement for assessing food effects during the drug approval process, such studies, would be assessed on a case-by-case basis, and would be beneficial in obtaining an insight into these effects prior to submission. An understanding of the food effects could potentially impact clinical trials, and early knowledge would guide formulation development through the clinical stages.

Ionizable Drugs or Excipients: Buffer Capacity Considerations

This breakout was facilitated by Rohit Jaini and Parnali Chatterjee. Four questions were addressed. This breakout session was focused on buffer capacity considerations while developing dissolution methods for drug products that contain ionizable drugs (especially weak acids and weak bases that would be considered Biopharmaceutics Classification System (BCS) Class IIa/IIb drug substances) or ionizable excipients (e.g., polymers, pH modifiers, solubilizers). The discussion was structured into four phases, (i) understanding current practices in the industry and academia when developing dissolution methods for ionizable compounds; (ii) gauging industry stance on the importance of buffer capacity on dissolution, delving into buffer concentration and buffer type; (iii) influence on developing QC dissolution method versus biorelevant dissolution method; and (iv) leveraging in silico tools to predict in vitro dissolution of drug products with ionizable molecules or excipients.

Firstly, at what stage during the product development process are dissolution media modified to take into consideration ionizable drugs and excipients? The general sentiment, concurrent with prior literature on in vitro dissolution of ionizable compounds, was that method development is highly compound specific and drug product specific. The group noted that, in most cases, a single dissolution medium under sink conditions is used for dissolution testing of development formulations of weak acids or weak bases. For weak acids or weak bases, dissolution is influenced by a variety of factors including dose/solubility ratio, stability of the drug substance in the dissolution medium, intrinsic solubility, acid/basic nature of the drug, media pH, pKa of the molecule, particle size distribution, buffer species and concentrations, and hydrodynamics. Solid-state properties (e.g., thermodynamic stability) of the weak acids/weak bases are an additional criterion that can play a pertinent role in the selection of a suitable dissolution medium. At early stages of product development, the thermodynamically stable form is not known. Through form screening studies, the thermodynamically stable form is identified. At the same time, solid-form conversion is monitored using dissolution testing so that it is of less concern during later stages of the product development. If however, solid-form conversion does occur and dissolution testing cannot discern the two forms, the method development will then have to be initiated from the beginning.

Consequently, this makes developing biorelevant or bio-predictive dissolution methods a formidable challenge, more specifically for ionizable drugs or dosage forms containing ionizable excipients. It indeed is essential to understand the key limiting factors controlling drug dissolution when developing dissolution methods.

Industry representatives provided anecdotal instances where a certain zwitterionic molecule exhibited slower in vitro dissolution while exhibiting good in vivo absorption. This was largely due to common ion effect with a specific ionic species present in compendial buffer media, but not encountered by the molecule in vivo. Consequently, dissolution was performed in an alternative buffer system to have more representative dissolution of the molecule. It was suggested to comprehensively evaluate common ion effect, not just with sodium or chloride ions, but all ionic species that the molecule may encounter in vitro and in vivo. The discussion was then focused on the buffer capacity consideration when developing dissolution methods for ionizable molecules.

Secondly, what different buffer systems are investigated for ionizable drugs and excipients? Are buffer systems used for dissolution testing part of quality control dissolution method or are they used exclusively for product development/research activities? Two schools of thought were evident. One side presented that the influence of buffer capacity may be overplayed and may not be as limiting a factor as generally expected. This view held that, despite lower buffer capacity, physiologically, there is excessive amounts of continuous secretion of bicarbonate buffer in the body that maintain a bulk pH of 6.8, unlike under in vitro conditions that work with a fixed volume of media. Moreover, the implication was that the in vivo buffer system makes up for lower buffer capacity in its sheer volume. An alternative view presented that buffer capacity is indeed essential for compounds where dissolution is highly sensitive to changes in surface pH in the unstirred boundary layer around dissolving particles. This is a function of pKa and intrinsic solubility of the molecule or excipient. It would be pertinent to use a buffer type or buffer concentration to replicate the lower in vivo buffer capacity when developing a more biorelevant dissolution method to aid formulation selection and drug product design. Other physiologically relevant buffer systems were discussed including fasted state–simulated gastric fluid (FaSGF), FaSSIF, and FeSSIF that are used for formulation selection and for physiologically based biopharmaceutics modeling (PBBM). However, these systems suffer from stability issues (~1 week), are generally expensive, and not scalable. Taking both arguments into consideration, the trade-off between reducing buffer capacity and maintaining the bulk pH (as is well regulated in vivo) is crucial to consider. Additionally, the group advised considering buffer capacity considerations for enteric coated dosage forms and for molecules with high precipitation tendencies.

Thirdly, the group was then asked whether considerations of buffer capacity were influential in developing QC methods. The group appeared unanimous in its opinion of keeping the QC dissolution methods as close to the conventional compendial methods whenever possible to facilitate ease of testing at manufacturing facilities. It is possible for some molecules that the QC method and biorelevant dissolution method might end up being identical. Nevertheless, the preference seemed to be to use a more complex biorelevant method primarily to aid drug product design and formulation selection. Lastly, gaps in currently available in silico tools to predict in vitro dissolution were discussed.

Fourthly, what are the major gaps in developing/leveraging in silico tools to predict performance of ionizable drugs and drug products with ionizable excipients? The group considered this to be the ‘holy grail’ of dissolution. Research groups are striving toward developing more mechanistic in silico dissolution models with a focus on capturing species effects (e.g., pH, pKa, bile mixed micelles, surfactants). Lacking a full consideration of these interactions in a PBBM model limits scientists’ capability in mechanistically describing the drug absorption, distribution, metabolism, and elimination. As these in silico tools are new and need to be validated, the consensus was to encourage greater collaborations between industries and academia to facilitate data sharing to improve and validate in silico tools. The more accurate the in silico predictions of in vitro dissolution are, the easier it will be for method development and translatability to predict in vivo performance.

Non-Compendial Testing for ASDs from Industry and Regulatory Perspective

This breakout was facilitated by Lynne Taylor, Andre Hermans, and Rajesh Savkur. The main question was the following: For ASDs, what compendial and non-USP dissolution methods are most useful, and what are the challenges? It was acknowledged that a common, fundamental challenge to assessing ASD dissolution is the lack of clarity about what it means for a poorly water-soluble drug to be dissolved from an ASD; even in favorable conditions, ASDs result in significant amounts of drug being “dissolved” as colloidal species, rather than true molecular solutions. Correspondingly, given limitations to the current understanding of drug dissolution from ASDs, there is no universal in vitro “biopredictive/biorelevant” dissolution method to guide ASD formulation optimization. Three techniques that were discussed, with an eye on measuring or assessing “dissolved” drug, were the filtering method (i.e., separate dissolved from undissolved drug), dissolution/permeation system, and centrifugation to isolate polymer and drug in undissolved particles.

A discussion point initiated around how to define sink conditions for QC methods. Sink conditions are not well defined for ASDs, including whether to consider crystalline or amorphous solubility should be considered a basis for sink conditions. It was noted that sink conditions are not required by regulatory agency and that procedures to measure amorphous solubility are not uniformly well defined, since amorphous solubility depends upon a phase separation process which is time-dependent and can change based on composition. Similarly, a single value for crystalline solubility can be challenging, since the presence of excipients (e.g., polymers) may impact measured solubility. For example, surfactant either from formulation or dissolution medium can interact with drug, polymer, or other excipients to modulate the amount of molecularly dissolved drug. The use of buffers as a dissolution media without surfactant may reduce this complexity.

Given these complexities, more standardized approaches to assess ASD formulations, in terms of measuring dissolution or solubility versus time for ASD (e.g., how much excess drug; what time points) would be beneficial. However, identifying standardized approaches for ASDs would be challenging at this time. It should also be taken into consideration that ASDs are typically different from neat amorphous drug, since the polymer (and surfactant) impact the measured solubility. Amorphous solubility or “kinetic solubility” implies not a true equilibrium, but rather a snapshot in time. While a regulatory perspective encouraged the examination of the concentration profile versus time, it was recognized that during such an experiment that different species will be present (e.g., nanoprecipitate, drug-rich domains formed following Liquid-Liquid Phase Separation), in part depending on quantities of drug, polymer, surfactant, and their interactions (e.g., drug-polymer interactions). In practice, even when drug from ASD is already considered to be in solution, drug concentration determination from filtered samples versus centrifuged samples will typically yield different results. Additionally, drug can crystallize from supersaturation in samples subjected to filtration at high pressure.

Two common, but opposing, observations of ASD dissolution are fast drug release, as well as very slow disintegration/dissolution due to gelling. In many cases, fast release has been observed regardless of formulation changes, potentially limiting the utility of the test conditions. Also, depending on the polymer and drug load, gelling can cause very slow dissolution with unknown in vivo relevance.

In the context of early formulation development, an important dissolution test capability is in vivo sensitivity to impact of polymer type, drug load, and process parameters in in vivo performance. A dissolution test employing a single dissolution compartment will often over-estimate the degree of drug precipitation. A useful approach may be application of a range of non-compendial methods (e.g., a single dissolution compartment, a two-stage or multi-stage method, and a dissolution/permeation system) to select formulations for subsequent development in human studies. When the ASD composition is finalized, the focus moves to identifying a QC dissolution method.

Discussions about early formulation development also noted that different dissolution approaches can be applied to address differing issues or phenomena, such as spring effect, or parachute/precipitation kinetics. Non-sink methods offer opportunity to screen formulations using a simple method. Relatedly, early formulation development studies may use smaller drug amounts, such that larger dissolution systems are only used when a more final dosage form is available.

Discussions about later formulation development acknowledged two potential dissolution methods: a QC method for batch release and a biopredictive/biorelevant method for biowaiver or bridging scenarios. It was acknowledged that non-compendial approaches have potential to enrich modeling, including PBBM, to improve in vivo prediction. However, such biopredictive/biorelevant methods will need to be effective in adding value beyond the QC test and robustness in terms of being transferable across site locations. Some contract development and manufacturing organizations (CDMOs) may have limited ability to apply non-compendial approaches to formulation development. Concerns whether a manufacturing site can repeat what has been observed in the research laboratory setting, using a more complex non-compendial method were raised.

A final discussion concerned potential for non-compendial methods to be acceptable to regulatory authorities. Feedback indicated flexibility and willingness of regulators to consider non-compendial methods, with appropriate scientific justification (e.g., test results are concordant with pharmacokinetic exposure). However, it was acknowledged that harmonization uncertainties tend to force sponsors to a common QC approach. Also, a more complex non-compendial methods can be expected to present greater challenges for method transfer to multiple sites throughout the world.

Drug Dissolution from Amorphous Solid Dispersions

This breakout was facilitated by Dana Moseson and Debasis Ghosh. The main question was the following: What basic and applied laboratory methods provide the best insights into drug dissolution from amorphous solid dispersions? Focus was on using dissolution testing as a method to assess performance failures in amorphous solid dispersions, specifically with respect to crystalline content. As an amorphous solid dispersion provides a bioavailability benefit over its crystalline counterpart due to its higher solubility, any crystalline content present in the formulation reduces the solubility advantage or serves as a substrate for crystal growth and further supersaturation loss. Crystallinity within an amorphous solid dispersion may result from nucleation and growth pathways (i.e., during stability storage) or incomplete crystalline-to-amorphous phase transformation (i.e., during hot melt extrusion) (32). Key attributes of crystals formed by these pathways when compared to bulk crystalline material include small crystallite size, high surface area, high surface energy, and possibly a different polymorphic form. Additionally, these sorts of endogenous crystals are encased within an amorphous polymeric matrix, which may prevent extensive crystal growth depending on its properties as well as environmental conditions.

In vitro dissolution testing to study the impact of crystallinity on amorphous solid dispersion supersaturation profiles may have two main purposes. First, the method may seek to quantitate crystallinity. Second, the method may serve to predict the potential bioavailability implications of crystallinity within an amorphous solid dispersion.

Detecting crystallinity in a QC dissolution method is possible only under very few circumstances, due to inherent dissolution test variability, selection of appropriate sink/non-sink conditions, media selection, and the propensity of the drug for crystal growth under selected conditions (33). Dissolution test design can be done by introducing crystallinity created in one of several ways: (i) Directly adding additional crystalline content (spiking experiment) (34); (ii) Stress the ASD formulation through high temperature/humidity exposure with the intent of crystallizing some or all of the amorphous drug (35); and (iii) Creating crystallinity within the ASD by altering the manufacturing process. For spray drying, this could be done by modifying the spray solvent system with an anti-solvent (such as water). For hot melt extrusion, this could be done by reducing the processing temperature or residence time (36).

While method 1 (spiking) appears to be the most straightforward, it may not accurately mimic crystal properties which may form in the ASD formulation in terms of crystal particle size/surface area, surface energy, and polymorphic form. Designing experiments that mimic different crystal properties to study their impact on supersaturation profiles is challenging (37). There is a lack of practical tools to assess the surface area of the crystalline particles when inside the ASD matrix. Microscopy-based methods, such as polarized light microscopy, transmission electron microscopy, and micro computed tomography may be used as non-quantitative surrogate methods to identify crystal attributes (depending on length scale) within an amorphous polymer matrix, but no current bulk property techniques are available. Solid state methods such as powder X-ray diffraction (PXRD) or other spectroscopic techniques are suitable methods for quantifying crystallinity on a mass basis, but are limited by dilution, crystal attributes, and method parameters (32).

Practical difficulties with spiking studies were highlighted. Since crystal growth occurs on a surface area basis, addition of bulk crystals on a mass basis underestimates the impact on supersaturation and crystal growth. Additionally, whether additional crystalline API material is added to the full quantity of ASD (e.g., 10% crystalline API + 100% amorphous API in the ASD) or it is replaced (10% crystalline API + 90% amorphous API in the ASD), the driving force for dissolution and crystal growth is changed. Even the use of micronized spiked crystals does not provide for an unequivocal improvement in dissolution test design over bulk crystals. Two conflicting examples can be given. In a study by Moseson et al., the use of bulk crystals followed a concentration dependent solubility advantage decrease. However, in samples containing residual crystals remaining from the hot melt extruded manufacturing process, greater supersaturation loss was detected than expected based on the quantification of crystals on a mass basis from a PXRD method (36). In a study by Hermans et al., the use of bulk versus micronized crystals were used in spiking studies. When 10% bulk crystalline material was used, the dissolution method was able to detect and provide an approximate quantification of crystalline content (34). However, when micronized crystals were used, rather than over-estimating crystalline content (as a surface area-based crystal growth hypothesis would suggest), no concentration difference was detected between this experiment and the 100% amorphous sample, speculated to be due to the greater solubility found in high surface energy small particle size crystals.

Crystallinity as a general critical quality attribute for ASD formulations was also discussed. A risk-based approach should be used when determining the risk of crystallinity occurring within the ASD drug product. For example, not all amorphous drugs are likely to crystallize based on their high glass transition temperature, or when drug-polymer interactions persist during storage. Orthogonal tools should be used to characterize, detect, and/or quantify crystallinity within the amorphous solid dispersion drug product. Regulators recommend that drug manufacturers/sponsors perform a risk assessment and investigate the impact of crystallinity in dissolution and in vivo performance. Sponsors should thoroughly characterize their drug product performance and drug manufacturing processes and justify their control strategy.

Drug Dissolution from Nano-Formulations

This breakout was facilitated by David Curran and Anitha Govada and proceeded via six questions (Table I). Firstly, what are key technical challenges presented by nanoparticle formulations? There is a terminological discrepancy with “nanoformulations” between filing institutions, which describe formulations with particles between 1 and 1000 nm as nanoformulations. A size range of approximately 1–100 nm is commonly used in various working definitions or descriptions regarding nanotechnology proposed by the regulatory and scientific community. However, FDA does not have an established regulatory definition, and considers any material or end product with at least one external dimension, or an internal or surface structure, in the nanoscale range (approximately 1–100 nm), or material/end product which exhibit dimension-dependent properties or phenomena up to 1 µm (38, 39).

Differentiating if drug species are solubilized or not is challenging. Filtration of nanoparticles on the benchtop using Anotop 20 nm syringe filters can present issues with backpressure and filter clogging, and they are not compatible with automated dissolution systems. A proposed solution is to characterize particle size distribution (PSD) with dynamic light scattering (DLS) prior to filtration or analytical characterization, as the smallest particles in the distribution may dissolve before sampling; therefore, filtration at the 20 nm level may not be necessary in all cases. However, DLS is generally low-throughput and does not integrate easily into automated dissolution systems. Characterizing the PSD of nanoparticles in the formulation is challenging, since the size distribution may be dynamic during processing and dissolution characterization.

Secondly, what CBAs affect the in vitro drug release from nano-formulations? Nanoformulation PSD impacts its in vitro dissolution, especially when the API is hydrophobic. In the discussed case of fenofibrate, transitioning from conventional to micronized to nanosized forms improved bioavailability due to decreased PSD. While specific surface area (SSA) is a better indicator, practical challenges limit its use and PSD is commonly used as a CBA.

Permeation was also suggested as a CBA, since ultimately in vivo performance is determined by permeation rather than dissolution. However, it was argued that nano-formulated APIs are typically BCS class II, and are generally not permeation-limited, motivating dissolution as a more important CBA.

The process history of the nanoformulation was also offered as an important CBA. For example, wet milling and drying of nanoformulations can create an amorphous layer that may necessitate additional surfactant in the dissolution test. For nanoformulations produced from antisolvent precipitation, the identity of the antisolvent should also be considered a CBA.

Thirdly, what are the key considerations or strategies for achieving a clinically relevant dissolution method for nano-formulations? Conventional dissolution sample timings may not be fast enough for nanoformulations. For nanoformulations with rapid disintegration, a dissolution method must be able to sample on the order of 1 min, suggesting in situ analytics may be necessary.

A continuous, in situ method such as UV fiberoptics may be useful for characterizing dissolution, but also may be distorted due to effects from particle aggregation and non-sink conditions. To account for aggregation, academicians suggested using the zero-intercept method to detect aggregation or derivative spectra to compensate for scatter by nanoparticles.

Fourthly, what are potential benefits of methods that add a permeation component, such as dialysis, microdialysis, or dissolution-permeation? Many suggested that dialysis sampling times are too slow for nanoformulation assessment. However, others suggested that two-stage dialysis can sustain supersaturation. Additionally, regulators commented that microdialysis can obtain measurement time-scales around 2–3 min.

Dissolution-permeation was promoted by some since it may be more predictive and can maintain supersaturation without precipitation. However, regulators again suggested that bioavailability of drugs likely to be nanonized (e.g., BCS Class II) is not permeation-limited. Continuous microdialysis was also proposed as a potential method. However, challenges in executing this method with biorelevant media and artificially enhanced dissolution from micelle formation were reported. Reverse dialysis and the dispersion releaser were suggested as potential non-standard methods.

There was an open debate about whether any dissolution method utilized for nanoformulations should be under sink or non-sink conditions. Another open question was whether permeation enhancers should be considered in any of these methods. It was concluded that for orally-delivered peptides, permeation enhancers become important in dissolution-permeation characterization, but these therapeutics are hydrophilic and unlikely to benefit from being formulated in a nanoformulation.

Fifthly, what options exist for automating dissolution methods? All participants expressed interest in an FDA-developed automated compartment model based on tangential flow filtration (TFF). However, this promising method is still in development. Successful use of Agilent’s NanoDis System, which uses USP Apparatus 1 and 2, was reported. Challenges with adsorptive loss onto fibers were discussed, which may limit the effectiveness of this tool with precious samples, depending on method and molecule attributes. Methods based on in situ analytics with fiber optics were favorably discussed, since these methods do not require sample removal. Regulators noted that for some drug products, this may also be only the method listed in the FDA dissolution method database. However, reviewers still must monitor and validate these automated methods before being approved as an analytical method.

Sixthly, for nanocrystalline oral dosage forms, can disintegration be used as a proxy QC method instead of dissolution? Using disintegration as a proxy QC method is promising for formulations where disintegration is rate-limiting rather than dissolution. However, this approach requires substantial upfront investment from filers to establish a correlation between dissolution and disintegration for a given product.

Drug Dissolution from Lipid-Based Formulations

This breakout was facilitated by Anette Müllertz and Leah Falade and proceeded via two questions.

It was identified that the four most common types of lipid-based formulations for poorly soluble drugs were type 1 composed mainly of triglyceride, type 2 composed of oil/triglyceride + lipophilic surfactant, type 3 self-(nano)emulsifying drug delivery systems (S(N)EDDS) with lipids and high amount of hydrophilic surfactant, and type 4 containing hydrophilic surfactant and co-solvent. Type 4 formulations were the most widely used among the discussion group, mainly because these often provide the highest drug solubility (load). The goal of any given development of a lipid-based formulation is to obtain an isotropic lipid formulation (non-phase separating), to increase drug load and prevent precipitation during dispersion in the GI fluids. Predictive in vitro methods for this formulation include dispersion tests in water, and also in vitro digestion models, simulating the environment the lipid formulation is subjected to in the GIT. In type 4 formulations, a small amount of ethanol can be added to increase drug loading as high as possible, and here in vitro digestion is not relevant.

Nano-emulsion forming lipid formulations (type 3), such as S(N)EDDS, have been formulated for both oral and injectable dosage forms. S(N)EDDS have also been designed for oral proteins/peptides delivery, and here it is important to include medium chain lipids, that can act as permeation enhancers. In addition, the need to depend on bile salts for emulsification of lipid systems can be eliminated by addition of surfactants in the formulations. This further ensures uniform particle size of the emulsion droplets and consistent lipid digestion in vivo and could potentially eliminate one variable caused by food effect.

Challenges include drug precipitation, interactions with bile salts, the presence of high content of co-solvents, and developing lipid systems with high enough drug loading and less surfactants/co-surfactants. For drugs dissolved in lipid-based formulations, rupture tests of the capsule, dispersion tests for batch uniformity, and later disintegration tests have also been proposed as valuable QC control tests. In addition, droplet size tests have been utilized. It was mentioned that droplet size tests can also be adequate for QC purposes specifically for an emulsion-based drug product, but a dissolution/dispersion method may be necessary for scale up and post-approval change (SUPAC) guidance purposes.

Formulation challenges include the presence of too much co-solvent, which can cause the drug substance to precipitate upon dispersion if drug solubility in the lipid formulation is due to co-solvent presence; co-solvent will partition into the aqueous phase during dispersion, thereby reducing drug solubility in the emulsion droplets. Such challenges are current issues to enhance the potential of lipid-based formulations as a prominent delivery strategy.

Drug Dissolution from Co-Crystals

This breakout was facilitated by Abu Serajuddin and Alaadin Alayoubi. The main question was the following: What are the current roles of co-crystals in drug development? Co-crystals are defined as crystalline materials composed of two or more different molecules, typically the API and co-crystal formers (coformers) in the same crystal lattice, according to the FDA guidance “Regulatory Classification of Pharmaceutical Co-Crystals” definition (40). Unlike salts, where the API forms an ionic complex with an acid or base in the same crystal structure, co-crystals rely on weaker intermolecular interactions such as hydrogen bonds, π bonds, and van der Waals forces. One advantage of co-crystals over salts is that they are not restricted by the pKa of the API. While salts usually require a pKa difference greater than two between the API and the counterion for successful formation, co-crystals can be formed based on these weaker interactions.

Co-crystals, similar to salts, are useful for improving the physicochemical properties of APIs, to enhance solubility and dissolution rate, which can increase systemic exposure. Additionally, co-crystals can mitigate the negative effects of high pH on solubility and dissolution for basic drugs. The dissolution behavior of co-crystals under different pH conditions was discussed by briefly reviewing highlights of two papers on ketoconazole co-crystals published in the literature. Like salts, co-crystals demonstrated microenvironmental pH effects in increasing dissolution rates under certain pH conditions (41, 42). For example, at pH 5.0, ketoconazole co-crystals formed with fumaric, succinic, and adipic acids demonstrated a parachute effect (supersaturation) and enhanced dissolution over the free base. The dissolution rates, however, decrease with further increase in pH to pH 6.5. These findings hint that co-crystals can reduce the impact of food effect, since the gastric pH increases with food intake, which may decrease dissolution rate of the free base; increased dissolution rates of co-crystals observed at relatively high pH may minimize such effects.

One interesting point was whether a co-crystal may be considered a NCE. The FDA does not consider a co-crystal as NCE but rather analogous to a new polymorph of the API given that dissociation of the API from its co-crystal form occurs before reaching the site of pharmacological activity. However, some indicated that if salts are considered NCE, co-crystals should also be viewed as NCE because they possess different physical and chemical properties (e.g., melting point, solubility) that are more favorable over the free base or acid form of API. Dissolution testing of co-crystals requires considering the dose at a given pH, and if a parachute effect is observed, alternative approaches may need to be explored. Overall, the participants felt that cocrystals may be treated similar to salts for the purpose of developing dissolution methodologies. While co-crystals are generally not believed to have an effect on the intrinsic permeability property of the drug substance, further studies are needed to confirm it. However, co-crystals can help achieve supersaturation, which could result in a higher drug permeation rate. Some co-crystals may not dissolve in a 1:1 ratio, and the use of dissolution-permeation models may be helpful for their development.

A new perspective was also discussed regarding the use of co-crystals as a tool to reduce the formation of nitrosamines (N-nirosodimethylamines; NDMAs) in drug products. Some nitrosamines may be carcinogenic and may form when vulnerable amines (secondary, tertiary, and quaternary) react with nitrosating agents (N2O3 and NO-). Vulnerable amines may be a constituent part of the drug molecule, and nitrosating agents may be found in some excipients. Therefore, NDMAs may be formed in drug products. Recently, various research papers have shown that using antioxidants such as ascorbic acid as excipients may inhibit NDMA formations or even reduce existing NDMA levels in drug products via redox reaction. Co-crystal formation may play a role in this area, particularly by exploring co-crystals with antioxidants possessing potent nitrite scavenging properties.

Lastly, it was acknowledged that successful co-crystal formation is drug-dependent and can be challenging due to its reliance on weak intermolecular forces. Over the years, the interest in co-crystal development has diminished, possibly because more straightforward alternatives such as salt formation and amorphous solid dispersion (ASD) have emerged to address undesirable physicochemical properties of drug substances. Furthermore, the formation of co-crystals requires additional efforts to study the human safety of coformers. Nonetheless, co-crystals remain a useful tool for drug delivery and may be valuable in mitigating NDMA formation.

Non-Compendial Methods

This breakout was facilitated by Kerstin Schaefer and Hansong Chen and proceeded via three questions (Table I). Firstly, what is the definition of non-compendial methods? There was a consensus that non-compendial methods are not listed in the USP. In addition, modified compendial apparatuses like the use of mini or peak vessels or non-compendial conditions used with compendial apparatuses (e.g., uncommon stirring speeds, high surfactant concentrations) can be considered non-compendial methods. Hence, non-compendial methods use specialized equipment or modified USP setups as well as non-compendial media and settings.

Secondly, under what conditions has any particular non-compendial method been helpful, and why? Non-compendial methods are usually not pursued for QC dissolution testing, as they often use complex setups and media. These setups aim to help in formulation development, such as use of biorelevant media to bridge between in vitro dissolution and in vivo data. Ideally, non-compendial methods can help build simulation models and reduce or eliminate animal testing.

Thirdly, are such non-compendial methods complementary or potential replacements for compendial methods? Ideally, every QC dissolution method should be biopredictive and help in establishing in vitro–in vivo correlation (IVIVC) or IVIVR. However, in reality, this is often not possible. Methods used for QC testing have to be robust and deliver reproducible results. It might not be possible and feasible to apply all criteria required in QC dissolution methods to non-compendial methods. Setups like the tiny-TIM system are highly complex and require a set of specialized media. Hence, currently compendial and non-compendial approaches are used for different purposes. QC dissolution testing is a product-specific quality test, whereas non-compendial methods can help to better understand underlying mechanisms. Currently, some pharmaceutical companies have used non-compendial methods in authority interactions. However, very limited to no feedback was received back from authorities. Both industry and regulatory authorities are open to using non-compendial methods. It was discussed that a database containing (all) case studies would be helpful due to the novelty and lack of expertise in industry and at regulatory authorities in the use of data generated with non-compendial methods e.g., in submission documents. A future goal should be to devise a mechanism to generate and share a database.

Non-USP Methods versus Regulatory Methods: Biopharmaceutic Risk Assessment

This breakout was facilitated by Yi Gao and Tapash Ghosh. The USP methods (i.e., USP apparatuses 1 to 6) and non-USP methods (i.e., non-USP standardized apparatuses or methods that are not yet well defined) were reviewed. It was concluded that developing and filing a non-USP method has not become a well-established practice. For example, a QC method that combines dissolution and permeability concepts is a current research topic, may result in clinically relevant approaches for conducting dissolution and setting dissolution specifications of certain supersaturating formulations, but has not been developed and implemented in QC. Discussions proceeded via four questions.

Firstly, how do you initiate developing a dissolution method for your proposed product? For 505(b)(1) vs 505(b)(2) vs ANDA products? For NDA products, the dissolution method is often developed from scratch. For Abbreviated New Drug Application (ANDA) products, the previously known method (i.e., found in the database or regulatory filing documents) was the basis for starting the method development.

Secondly, what leads you to pursue or not pursue a non-USP method as your ultimate regulatory method? Except for one sponsor applying the apex vessel method (e.g., peak vessels) during developing commercial dissolution testing for one drug product, no other company in the audience has developed a non-USP method. The most common non-compendial methods that have been used are those with apex vessels to overcome coning issues. Although the variability due to coning was reduced, there is still a lot of variability from vessel to vessel due to apex vessels not being standardized. Significant differences existed among the apex rising angles and heights manufactured by different vendors. USP is actively working on standardizing the dimensions of the apex vessels, like other compendial apparatuses with specific dimensions and configurations. Once completed, USP will publish their proposal for apex vessels in the Pharmacopeial Forum which will remain open for public comments for 90 days. There was a consensus and interest in moving to make apex vessels compendial. Before the availability of USP apex vessels, coning may be addressed by adjusting agitation speed. Though paddle speeds of 50 and 75 rpm are recommended for USP apparatus 2, sometimes it becomes necessary to apply higher speeds. In those cases, it is important to provide data to justify the proposed higher speed, and especially important to demonstrate discriminating ability.

Thirdly, have you filed a non-USP method and gained approval from regulatory agencies worldwide?

The above sponsor did not file the apex vessel method due to the concerns of vessel QC problems, as discussed above.

Fourthly, what can be the potential risk/benefits and technical challenges in adopting and transferring a non-compendial method to commercial testing sites, especially for a non-NME? Since no one in the audience had filed and gained approval of a non-USP method, a thorough discussion on this question was not carried out, beyond the above comments about apex vessels. However, general risks were discussed. If non-compendial approaches are pursued in dissolution method development, the conditions should be well characterized, and all specifications need to be provided for review. These details are essential to ensure that data can be replicated, and have the ability to transfer the method to other sites.

Approaches to, and risks concerning, dissolution method development and selection were discussed. One discussion centered around the FDA and USP Dissolution Method Databases. These databases give a general outline of the methods previously used for similar drug molecule and dosage forms. Once a method is included in the FDA dissolution methods database, if subsequent generics or other sponsors deviate from the method, the company is encouraged to petition the USP to include the conditions in the USP database, and the specific USP Dissolution Test utilized will be listed in the product labels. It was noted that dissolution methods in the FDA and USP databases are provided as a suggested starting point to assist industry in method development. However, the dissolution specifications (i.e., method and acceptance criteria) are product specific. FDA encourages an applicant to exercise due diligence before they take an unusual measure, like increasing paddle speed to 200 rpm. The goal of FDA is to work alongside industry to ensure that the Prescription Drug User Fee Act (PDUFA) and Generic Drug User Fee Act (GDUFA) timelines are met, and that proposed dissolution conditions and methods are justified (e.g., exhibit discriminating ability) to support medications are safe and effective. The FDA is open to receiving data from studies in which a safe space could be defined. This concept of a safe space can be built based on established IVIVC, IVIVR, and/or PBPK models and can be used to extend the approved dissolution acceptance criterion.

Another discussion concerned non-USP methods used for specialized products such as chewing gums, buccal tablets, and orally disintegrating tablets. There is a European apparatus developed for medicated chewing gum; however, the equipment is very expensive and not readily available. Buccal tablets are designed to stick to the cheek pouch and therefore release the drug from one side. Given this mechanism, it was debated whether it is scientifically meaningful to subject such a product to dissolution vessel-based testing to quantify drug release. It was discussed that disintegration is a test that can be used for QC and therefore replace dissolution studies if conditions are qualified in the International Conference on Harmonization (ICH) guidance (e.g., orally disintegrating tablets that disintegrate quickly).

Real-Time Release Testing to Replace in vitro Dissolution

This breakout was facilitated by Hanlin Li and Haritha Mandula. Real-time release testing (RTRT) is the “ability to evaluate and ensure the quality of in-process and/or final drug product based on process data, which typically includes a valid combination of material attributes and process controls” (43). RTRT, if utilized, becomes a powerful tool in implementing Control Strategy and allows for increased manufacturing flexibility and efficiency, enhanced process understanding for real time corrective actions or segregations (in instances of continuous manufacturing), and added assurance of product quality. With respect to replacing in vitro dissolution, RTRT needs to be coupled with a predictive model. This leads us to the following questions: (1) What are the best RTRT practices to establish confidence in the model, (2) What are the challenges to implementing dissolution RTRT, and (3) What future improvements are needed?

The discussion first centered around the importance of the QC dissolution method. The QC method is the foundation for RTRT development. A clinically relevant QC dissolution method is critical for RTRT success. The confidence in the predictive dissolution model is limited by the robustness and discriminatory ability of the QC dissolution method. There was debate about the challenge to develop a clinically relevant QC method. Many times, the QC method is solely designed for discriminating capabilities with respect to manufacturing process; however, the manufacturing changes may not be relevant to in vivo situation. One reasoned that manufacturing design spaces are often so tightly controlled, that it may be near impossible to manufacture a non-bioequivalent drug product batch within the confines of those controls. In that case, the factors found relevant to in vitro dissolution performance should still be included in the RTRT dissolution model to support formulation robustness and manufacturing process.

Nevertheless, to truly harness the flexibility of the safe space for which a drug product is still expected to be ‘bioequivalent,’ one needs to have a firm understanding of the RTRT design space. A very tightly controlled manufacturing process can consistently produce a drug product that meets the target quality profile. However, it may not provide much flexibility for manufacturing operations or regulatory specifications. Clinical relevance of Critical Quality Attributes (CQAs), critical process parameters (CPPs), and critical material attributes (CMAs) is crucial for establishing the flexibility. For dissolution RTRT, an understanding of the dissolution process is required to ascribe relevant factors in the raw materials and process parameters to changes in the in vitro, and thus in vivo, dissolution. There is also a need to understand the level of interaction with CMAs and CPPs. Proper design of experiments (DoE) and risk assessment is paramount to the holistic approach for RTRT. The RTRT dissolution model can be used to predict a single time point or the entire dissolution curve. In the case of full dissolution profile prediction, it typically involves a partial least squares (PLS) model, predicting the dissolution rate “Z,” followed by using Noyes-Whitney or Weibull equation to predict the dissolution curve.

Open conversation is needed between regulators and industry to successfully implement a dissolution model for RTRT. While regulators assess the risk (i.e., what is the risk of relying on RTRT to predict dissolution for this drug product), the industry (or sponsor) is knowledgeable about their drug product and therefore could share the critical aspects and explain how the process knowledge is built into the dissolution RTRT. To reduce the risk from the pharmaceutical development perspective, it is important to test the limits of the model with respect to the extremes of the material attributes and operating ranges of the process parameters. The dissolution model should be challenged with an external data set and the model’s ability to detect non-conforming batches also needs to be demonstrated. It was emphasized that appropriate model maintenance (e.g., via regulatory commitment of annual parallel testing and internal QC that triggers model updates) is also required for the longevity of dissolution RTRT.

Finally, to encourage broader use of dissolution RTRT, publications of successful case studies are needed to help identify a structured approach as well as to move RTRT into prediction of more complex formulations. This brought the discussion full circle as it was determined that the biggest challenge for expanding the use of RTRT as a surrogate for in vitro dissolution is the need for an appropriate QC dissolution method on which to build the predictive model.