Product Quality Research for Developing and Assessing Regulatory Submissions for Generic Cyclosporine Ophthalmic Emulsions

Approval of the first generic 0.05% cyclosporine ophthalmic emulsion (COE) in the U.S. represents a milestone achievement of the science and research program in the U.S. Food and Drug Administration’s Center for Drug Evaluation and Research (CDER). COE is a locally acting complex drug product indicated to increase tear production in patients whose production is presumed to be suppressed due to ocular inflammation associated with keratoconjunctivitis sicca. The path to approval required overcoming numerous scientific challenges to determining therapeutic equivalence to the reference listed drug. Researchers in CDER’s Office of Pharmaceutical Quality and Office of Generic Drugs developed a quality by design approach to understand the effects of process and formulation variables on the product’s critical quality attributes, including globule size distribution (GSD), turbidity, viscosity, zeta potential, surface tension, and osmolality. CDER researchers explored multiple techniques to perform physicochemical characterization and analyze the GSD including laser diffraction, nanoparticle tracking analysis, cryogenic transmission electron microscopy, dynamic light scattering, asymmetric field flow fractionation, and two-dimensional diffusion ordered spectroscopy nuclear magnetic resonance. Biphasic models to study drug transfer kinetics demonstrated that COEs with qualitative and quantitative sameness and comparable GSDs, analyzed using earth mover’s distance, can be therapeutic equivalents. This body of research facilitated the review and approval of the first U.S. generic COE. In addition, the methods and fundamental understanding developed from this research may support the development and assessment of other complex generics. The approval of a generic COE should improve the availability of this complex drug product to U.S. patients.


Introduction
The first US generic equivalent to Restasis® 0.05% cyclosporine ophthalmic emulsion (COE), a complex drug product, was approved by the US Food and Drug Administration's (FDA) Center for Drug Evaluation and Research (CDER) in February 2022 (1). COE is indicated to increase tear production in patients whose tear production is presumed to be suppressed due to ocular inflammation associated with keratoconjunctivitis sicca, or dry eye syndrome (2). Generic COE should help to make a drug product with the same safety, efficacy, and quality as the reference-listed drug (RLD) available to the public at a lower cost. This generic approval is a milestone in in vitro characterization enabled by the science and research in CDER's Office of Pharmaceutical Quality and Office of Generic Drugs as well as other contributors within FDA (3).
The Generic Drug User Fee Amendments (GDUFA), initially passed in 2012 and renewed in 2017, in part fund research and other activities to support the development of generic drugs (4). One such activity is the issuance of product-specific guidance (PSG), science-based non-binding recommendations for manufacturers submitting an abbreviated new drug application (ANDA) for a particular generic product. As stated in the GDUFA II Commitment letter, FDA strives to issue guidance for complex products as soon as scientific recommendations are available. Complex drug products can have complex formulations, active ingredients, routes of delivery, dosage forms, and/or drugdevice combinations. COE has both a complex formulation, as an emulsion of water and oil phases, and a complex route of delivery, as locally acting eye drops. Often, regulators' research is critical to developing a scientific understanding to support regulatory communications such as PSGs. With certain exceptions, generic drugs for ophthalmic administration marketed in the USA must be formulated qualitatively and quantitatively the same (Q1/Q2) as the RLD (5). In addition to considering the ingredients of the finished dosage form, for some complex products, it is known that variation in the manufacturing process can cause significant differences in critical quality attributes (CQAs) such as oil globule size, which may affect drug performance (6,7). In such cases, information and scientific investigation to improve the understanding of the product's CQAs can help support the development and regulatory assessment of high-quality, affordable generic drugs.
Establishing bioequivalence to the RLD ensures that there is no significant difference in the rate and extent to which the active pharmaceutical ingredient (API) becomes available at the site of action (8). For many systemically acting drug products, this can be established through pharmacokinetic blood studies; however, locally acting topical drugs, such as COE, do not become available at the site of action via the bloodstream. Thus, for these products, the Federal Food, Drug, and Cosmetic Act's implementing regulations provide that data from an approach other than in vivo studies may be acceptable to establish bioequivalence if, as is the case for COE, FDA deems that approach adequate (9,10). This provides a basis for the possibility of manufacturers submitting a locally acting drug ANDA using comparative clinical endpoint studies or in vitro methods to demonstrate bioequivalence to the RLD, as was the case for COE ( Fig. 1) (6). In vitro studies can provide unique insight into the role that formulation variables have on API distribution within the finished dosage form or upon drug release following administration.
Drug release from a product to the release medium, such as the tear film, can depend on several CQAs, including distribution of API in the phases of the product, which the PSG for COE calls drug distribution but this paper refers to as API distribution (13). COE consists of an oil phase and an aqueous phase, but excipients such as surfactants Fig. 1 a Diffusion of API between phases (C oil is API concentration in the oil phase, C water is API concentration in the aqueous phase, and C micelle is API concentration in the micellar phase) affects drug distribution in the intact drug product (11). b Phases of drug product administered to the eye break up in the tear film so API can reach the mucus layer (12) 1 3 also lead to a micellar phase (Fig. 1a) (6,7). While API is soluble in the oil phase, it is only sparingly soluble in the aqueous phase, which complicates API distribution within the formula (13)(14)(15). Once API distribution is understood, the impact of process variables on drug release must also be delineated. In vitro release testing (IVRT) methods for complex formulations require an extensive understanding of the formulation and manufacturing process variables that affect CQAs and drug release. FDA research into COE characterization informed the development of the PSG for COE, released in 2013 and revised in 2016, which includes in vivo and in vitro testing options to demonstrate bioequivalence (9,16,17).
FDA research surrounding COE included developing a quality-by-design approach to discriminate the effect of process and formulation variables on CQAs and drug release (7). Methods to characterize the size and shape of oil globules, such as laser diffraction (LD), nanoparticle tracking analysis (NTA), and cryogenic transmission electron microscopy (cryo-TEM), were used to validate more common and established dynamic light scattering (DLS) techniques (18). A novel GSD profile comparison approach based on the earth mover's distance was developed to compare COE samples given the observed multimodal GSD (19). Other orthogonal methods related to asymmetric field flow fractionation (AF4) and two-dimensional diffusion-ordered spectroscopy nuclear magnetic resonance (2D DOSY-NMR) were developed to better understand features of the COE formulation (20,21). CDER researchers also developed kinetic API distribution models and investigated IVRT methods for COE (13)(14)(15). This review summarizes the FDA research that aided the demonstration of therapeutic equivalence and enabled the first US generic COE approval in February 2022. This research should continue to bring value for the development and approval of other similar complex drug products.

Critical Quality Attributes
When COE drops are topically administered to the eye, the components diffuse through the lipid and aqueous layers of the tear film to deliver cyclosporine to receptors in the mucus layer (Fig. 1b) (12). The critical quality attributes (CQAs) of COE are the physical, chemical, biological, or microbiological properties that ensure the desired product quality (22). For the in vitro BE option, the current draft PSG recommends the measurement of six parameters (GSD, viscosity, pH, zeta potential, osmolality, and surface tension) in addition to formulation considerations and other product properties (23). GSD is clinically important because it directly relates to drug release, and larger globules may be ejected upon blinking (7). Zeta potential and viscosity influence globule coalescence. Viscosity, osmolality, and pH may contribute to irritation or discomfort, and surface tension can disrupt the tear film (6).
In one study, CDER researchers used a quality-by-design approach to study the effects of formulation and process variables on GSD, turbidity, zeta potential, viscosity, osmolality, surface tension, contact angle, pH, and drug diffusion (7). The COE formulation combined the oil phase (with cyclosporine) and aqueous phase (with emulsifiers and stabilizing agents) using a homogenizer to mix at controlled temperatures to form a coarse emulsion (Fig. 2). Subsequently, the coarse emulsion was exposed to high shear force in a microfluidizer (24), producing a nanoemulsion with a much smaller GSD consisting of oil globules coated in a layer of surfactant molecules, an aqueous phase, and a micellar phase ( Fig. 1a) (11). Notably, formulation and process variables were found to significantly affect five CQAs: particle (globule) size distribution, turbidity (directly related to particle size distribution), viscosity, zeta potential, and osmolality. Surface tension, pH, contact angle (related to surface tension), and drug diffusion were not found to be affected by formulation and process variables. Table I describes the instrumental techniques FDA scientists explored to measure each CQA (20).
FDA scientists sought to understand not only which process and formulation variables affect drug performance, but also the nature in which changes in these variables impact CQAs. Table II summarizes how the five CQAs were affected by process and formulation variables. For example, microfluidizer pressure and the number of cycles ( Fig. 2) impacted all five CQAs; when these variables increased, osmolality increased due to water loss (temperature affects osmolality similarly), viscosity decreased due to greater exposure to shear force, and zeta potential increased as a consequence of smaller globule size. Particle size distribution and turbidity decreased when: (i) surfactant was added directly to the oil phase and/or viscosity adjuster was added to the aqueous phase prior to mixing, (ii) the pH was adjusted after emulsifying the mixture rather than before, (iii) the phase volume ratio was lowered (e.g., 1:5 vs. 1:10 oil-to-water ratio), and (iv) microfluidizer pressure or the number of cycles increased (7,20). The establishment of these empirical relationships was foundational to understanding the relationship between possible COE manufacturing processes and product CQAs.

Globule Size Distribution Analysis
Process and formulation variables have been shown to affect GSD, which is known to have a strong relationship with emulsion drug performance (6,7). Cryo-TEM showed differences in GSD between formulations produced under different conditions, highlighting the importance of physicochemical analysis even for formulations with Q1/Q2 sameness ( Fig. 3) (18). Since the GSD of COE spanned approximately 30-300 nm in globule diameter with micelles being sub-20 nm, complementary analytical techniques were developed to characterize COE across the size range, as summarized in Table I (20,21). Cryo-TEM and NTA were each sensitive within this size range, but neither detected particles well across the entire range. Although both cryo-TEM and NTA directly counted particles, even in polydisperse samples, they were lower throughput than other techniques such as DLS. They also required highly specialized operations, making them challenging to use in a quality control environment. DLS could be used to assess the entire GSD of COE; LD was less reliable. Collectively, these techniques provided a thorough understanding of COE's GSD, with DLS emerging as a versatile candidate for regular physicochemical characterization due to its robustness and sensitivity.
The polydisperse nature of COE necessitated that even established sizing techniques, such as DLS, required validation for this product (6). DLS analysis requires an understanding of the sample's viscosity due to its effect on Brownian motion (18). There was concern that DLS may overestimate particle size in some COE formulations because COE exhibits non-Newtonian behavior when formulated with certain viscosity-enhancing polymers. However, dilution effectively reduced this non-Newtonian behavior, facilitating accurate DLS analysis. While darkfield imaging supported the hypothesis that diluted samples may show less agglomeration of particles than the undiluted samples show (Fig. 4) (25), there was a risk that samples could become physically unstable, resulting in an inability to analyze COE as it exists in the drug product. Cryo-TEM and NTA were used to investigate the effects of dilution and demonstrated that the morphology and particle size distribution of globules remained consistent up to 1:50,000 dilution in a dispersion medium of ultrapure water (18). This observation supported the use of COE sample dilution for GSD analysis.
FDA scientists also developed new AF4 and 2D DOSY-NMR techniques to characterize GSD. AF4 measures the behavior of particles which flow, through the diluent purified water, at rates determined by the size and related Fig. 2 Cyclosporine ophthalmic emulsion preparation method. Oil and aqueous phase were combined by a homogenizer to form a primary emulsion. Additional aqueous phase was incorporated to form the coarse emulsion, which was fed to a microfluidizer that imposes a specified pressure on the sample. The sample then flowed to the inter-action chamber where shear forces were applied to the emulsion, generating smaller oil globules. A recirculation tube allowed for a specified number of microfluidizer cycles before the sample exited as the manufactured emulsion. Temperature was controlled throughout the process (7,24) physicochemical properties while 2D DOSY-NMR allowed for measurement of globule size and polydispersity without sample dilution, both with higher resolution than DLS (20,21). AF4 methods were developed using the RLD and compared with existing DLS and cryo-TEM methods. CQA responses to differences in manufacturing processes were   Key: red down arrow indicates a decrease in the variable or a decreased CQA value; blue up arrow indicates an increase in the variable or an increased CQA value; a plus sign indicates addition of a previously absent variable (rather than increase or decrease, as in concentration) of the variable detected by AF4 with improved resolution over DLS alone, which may merit further use of AF4 for COE and other complex formula development. 2D DOSY-NMR was found to be consistent with AF4 results. Although 2D DOSY-NMR and AF4 are less common than techniques such as DLS, they may be valuable for future complex product development, including complex formulations with larger oil globules than COE, as shown by validation of 2D DOSY-NMR for a propofol parenteral emulsion. Once a GSD for any emulsion sample is obtained, appropriate statistical methods are needed to compare it to the RLD. Common GSD histogram descriptors, such as 10th, 50th, and 90th percentile (i.e., D10, D50, and D90), which have historically been used to compare monodisperse populations, are not accurate for polydisperse GSDs (23). To better describe COE, FDA scientists developed a GSD profile comparison approach using earth mover's distance (EMD) (19). This enabled a more direct comparison of the GSD profiles of the RLD and of COE samples produced with different formulas or manufacturing processes. With DLS validated by FDA scientists for accurate GSD measurement of COE, and the novel equivalence approach validated for GSD comparison, FDA recommended these methods for assessing bioequivalence in the COE PSG.

Drug Transfer Kinetics Modeling
For an ophthalmic drug to be bioequivalent to the RLD, there must be no significant difference in the rate and extent of drug release at the site of action (26). The kinetics of drug release from the product to the eye (Fig. 1b) are determined by API distribution in the finished dosage form (Fig. 1a), complicated by cyclosporine's solubilization in both oil and water (7). Although drug release is a separate phenomenon from API distribution, both are governed by the net diffusion of API from the oil phase (11). API distribution in COE reaches equilibrium under stable storage conditions but changes rapidly upon dilution, such as when applied to the eye or during IVRT, which disrupts the equilibrium and drives a release of API from the oil phase. It is important to evaluate whether two products have the same API distribution before they are administered so that drug release, and therefore, performance may in turn be clinically equivalent (6). To understand API distribution within COE, FDA researchers: (i) developed kinetic models of cyclosporine diffusion between phases, (ii) related API distribution to GSD, and (iii) sought to connect API distribution to drug release for the development of IVRT methods. The extent of API diffusion between product phases is usually studied with a shake-flask method; however, shaking may introduce confounding variables (14). Similarly, a dialysis membrane setup can be insensitive to formulation and process variables if the membrane is a rate-limiting component of the drug release measurement (15). Therefore, kinetic investigations of COE required new methods capable of measuring rates of diffusion in addition to equilibrium concentrations without reliance on a rate-limiting membrane.
FDA researchers developed a biphasic kinetic approach using an apparent partition coefficient (log P app ) to measure the ratio of drug concentrations between oil and aqueous phases, monitored in real-time by an in situ UV fiber optic probe at the interface between phases (11,14). A biphasic model is appropriate for COE because diffusion between the micellar and aqueous phases is fast. Thus, the diffusion between the oil and aqueous phase is rate-determining in API distribution. This kinetic model was sensitive to formulation variables, measuring responses to changes in COE excipients; e.g., the presence or increase of polysorbate 80  (20). Images are cropped and carbomer was each found to increase net diffusion from oil to aqueous phase whereas glycerin slightly decreased diffusion. These quantitative diffusion trends provided valuable insight into API distribution in COE.
Directly relating GSD to API distribution in COE was accomplished by examining the relationship between surfactant (polysorbate 80) concentration in the aqueous phase and net diffusion of API to the aqueous phase (11). When a test sample has Q1/Q2 sameness as the RLD but larger globules, the resulting smaller total surface area of the oil phase may need less surfactant to stabilize the interface, resulting in a greater concentration of surfactant in the aqueous phase (forming micelles) and greater amount of API in the aqueous phase. CDER researchers studied the effect of interfacial area (the area of the interface between oil and aqueous phases) on net diffusion and found that increased surface area did not change the average extent of API distribution but did generate faster bidirectional diffusion. The resulting model related globule surface area, surfactant concentration, and other attributes of a COE sample with its API distribution, enabling informed assessment of bioequivalence based on GSD and Q1/Q2 sameness.
FDA researchers applied their kinetic models to investigate drug release under various release conditions of IVRT or clinical application (15). Surfactants or co-surfactants, when added to the release medium for IVRT studies, also increase the rate and extent of oil-to-aqueous cyclosporine diffusion. Interestingly, the increased temperature decreased diffusion, opposite to typical solubility trends but consistent with cyclosporine's unique solubility behavior. These observations about release conditions should aid IVRT development for complex formulations.
The PSG does not describe a specific IVRT method for COE, which is challenged by the rate-limiting problem posed by a dialysis membrane as well as variables in the eye such as rapid drainage and ongoing dilution (6,11,14). While robust IVRT methods using biorelevent conditions that discriminate the effect of process variability can give insight into the in vivo performance of COE, a quality control test (such as measuring the GSD) also helps ensure a consistent release pattern. It has been shown that GSD in COE is a predictor of API distribution and therefore of drug release. Accordingly, the PSG recommends data from IVRT, which are generally correlated to GSD and API distribution, to support the bioequivalence of COE products with Q1/Q2 sameness to the RLD and comparable GSD (23). FDA scientists developed an early theoretical biphasic model to correlate physicochemical characteristics with tear film breakup time, a clinical endpoint in the treatment of dry eye, and validated the model against histological data from a rabbit model scaled to human tissue weights (12). This theoretical model, another perspective on bioequivalence, also depended on GSD. FDA's body of research demonstrated that COE products with comparable physicochemical characteristics to the RLD, particularly GSD, will have comparable drug release and performance. a Undiluted DLS demonstrating variable particle size peaks from six measurements of the sample, b 10 × dilution DLS of the same formula demonstrating more consistency in six measurements of the sample. Dilution of COE decreased globule size measurement variability by DLS (18). Darkfield microscopy images of COE at 100 × magnification (scale bar indicating 10 μm), showing: c undiluted sample showing agglomerates and d 10 × diluted sample of the same formula, demonstrating the particle separation associated with more consistent particle size measurements by DLS upon COE sample dilution

Conclusions and Future Benefits
The public availability of generic drug products is an important consideration in CDER's mission. GDUFA-funded research led by CDER's Office of Pharmaceutical Quality and Office of Generic Drugs helped build the scientific foundation that supported FDA's ability to approve generics for complex drug products like COE. Investigation of CQA responses to changes in formulation and process variables, instrumental analysis, drug diffusion, and drug release informed the in vitro characterization described in the PSG for COE. The approval of the first US generic COE product marks a significant advancement in the application of this in vitro characterization approach.
The research outlined here helped to establish confidence in techniques used to demonstrate the bioequivalence of a proposed generic COE to the RLD. By studying the CQAs of COE (Table II), FDA was able to establish relationships between some CQAs, such as that found between GSD and API distribution (7). Informed by FDA research, the current PSG for COE recommends including GSD, viscosity, pH, zeta potential, osmolality, surface tension, and API distribution data in an ANDA submission (23). FDA thoroughly investigated GSD, the most challenging CQA to measure, indicating that DLS was an appropriate analytical method for characterizing COE. The PSG recommends analyzing both undiluted and serial diluted samples in order to measure the GSD more accurately (since dilution mitigates agglomeration and size overestimation) as well as characterize the baseline, unaltered product (18,23). Statistical equivalence analysis of the resulting GSD histograms should use methods well suited for multi-modal distributions in polydisperse samples, such as the EMD-based equivalence analysis, a technique which FDA adapted to assess COE distribution (19,23). FDA also developed new instrumental techniques to measure GSD by AF4 and 2D DOSY-NMR, providing insight into COE and holding potential for future study of complex formulations (20,21). FDA researchers developed both theoretical and kinetic biphasic methods to study API distribution and performance (11,14). The PSG advises that applicants can submit comparative IVRT data; FDA investigated IVRT for COE (15) and found that such data would generally be effective to identify a COE formulation that differs significantly from the RLD. FDA's kinetic model demonstrated that a generic COE with Q1/Q2 sameness and comparable GSD to the RLD is most likely to be equivalent in drug performance (11), making the approval of this generic COE possible based on in vitro characterization.
Several scientific achievements in better characterizing and evaluating COE also supported the recent approval of a generic difluprednate ophthalmic emulsion (27) and may translate to further complex drug product development and assessment. The documented relationships between formulation and manufacturing processes and CQAs (see Table II) can serve as reference points for applications of future COE generic drug products and as an example of how to study other complex formulations. GSD characterization by DLS might be validated to support the establishment of bioequivalence for other drugs, including emulsions and suspensions, for which GSD is a CQA. The kinetic biphasic method might measure the rate and extent of diffusion between phases in other emulsions and relate their GSDs to API distribution. AF4 and 2D DOSY-NMR methods might be developed and tailored to orthogonally measure other polydisperse formulations (e.g., liposomal products, colloidal iron complexes). This body of research should continue to positively impact the development of complex generics in the global landscape (28).
This article shares the breadth of research conducted by CDER to benefit FDA stakeholders, including the pharmaceutical industry and US patients, as well as streamline regulatory processes. Additional research into general and specific complex drug products is ongoing, including the development of adaptive perfusion and pulsatile microdialysis methods for IVRT (29,30) and ultra-centrifugation with bench-top NMR methods to measure drug phase partitioning (31). Product quality research like this will help FDA continue to support the structured assessment of incoming ANDAs (32), conduct postmarket testing (33), prepare for advanced manufacturing technologies (34,35), and assure that safe, effective, and high-quality generic medicines are available to the US public.
Funding This project was supported in part by an appointment to the Research Fellowship Program at the Office of Pharmaceutical Quality, Center for Drug Evaluation and Research, US Food and Drug Administration, administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the US Department of Energy and FDA. This article reflects the views of the authors and should not be construed to represent FDA's views or policies.

Conflict of Interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.