Producing In-Situ Nanoparticles of Griseofulvin using Supercritical Antisolvent Methodology

Poor aqueous solubility of drug candidates is a major challenge for the pharmaceutical scientists involved in drug development. Particle size reduction to nano scale appears as an effective and versatile option for solubility improvement. Unlike the traditional methods used for the particle size reduction, supercritical fluid (SCF) processing techniques offer advantages ranging from superior particle size control to clean processing. Amongst all of the SCF based techniques, supercritical antisolvent (SAS) processing is of particular interest because most pharmaceuticals, including the model drug for this study-griseofulvin, are insoluble in supercritical carbon dioxide (scCO2), and SAS is one of the technique that can effectively process such compounds. Additionally, SAS is the only technique amongst SCF based technologies that has been successfully applied at an industrial scale. There are number of factors in effect during SAS processing. These factors can be grouped into two main categories; formulation related, and process related. In order to design a robust SAS process, it is extremely important to understand the impact of all of these variables on the desirable SAS product attributes, such as particle morphology, particle size, particle size distribution, and % yield of the process. Although several researchers have studied these variables, there is widespread disagreement amongst them. Hence, the goal of the studies shown in this dissertation is to address these gaps in the literature by carrying out a screening design of experiment (DOE), where 7 factors were studied, at 2 levels each, for their impact on particle size, particle size distribution, and process yield. A 2 fractional factorial design of 16 experiments, plus 3 center point runs, for a total of 19 experiments, was performed. The factors that impacted the particle size the most were the nozzle diameter, temperature, and spray rate of liquid, in the order of decreasing importance. In case of particle size distribution, nozzle diameter, spray rate of liquid, drug concentration, pressure, and polymer concentration played significant roles. The yield was affected by polymer concentration, pressure, and the drug concentration. Additionally, we were able to find optimum processing and formulation variables, which would consistently deliver product of high yield (~90%), small particle size (d50 of ~ 0.4 μm), and narrow particle size distribution. Further, we prepared and compared the physical and physicochemical characteristics of griseofulvin-polymer composite particles produced via three different methods: (1) supercritical antisolvent (SAS) process, (2) spray-drying process, and (3) the conventional solvent evaporation process. The polymers used were Kollidon® VA64, HPMCAS-LF, and Eudragit® EPO. Particle properties were analyzed using scanning electron microscopy, powder X-ray powder (PXRD), differential scanning calorimetry (DSC), and Fourier transform infra red (FTIR). Particle size and particle size distribution measurements were made using Malvern laser diffractometer. The dissolution behavior of pure API and solid dispersions were compared. Amorphous solid dispersions of spherical shapes were obtained, independent of the type of polymer used, when spray drying process was used. FTIR spectra indicated the formation of hydrogen bonding between the drug and polymers, during spray drying process. Whereas, the drug remained in its crystalline form when the processing method was SAS or conventional solvent evaporation, and there was no hydrogen bonding for these formulations. The griseofulvin particles used as unprocessed starting material had a mean diameter of approximately 12 μm with a size distribution range between 5-20 μm. With the spray drying or SAS process, and using any of the three hydrophilic polymers, in-situ nanoparticles with the mean particle size of 0.3 to 0.5 μm were obtained. These nanoparticles were associated with improved dissolution performance compared with unprocessed crystalline griseofulvin. In conclusion the physicochemical properties and dissolution of crystalline griseofulvin could be improved by physical modification such as particle size reduction using SAS process, and generation of amorphous state using spray-drying

. SCF technologies used for particle formation …………………………… . 21 Table 5. Material and process variables in SCF based processes that affect formation of particles with desired properties ………………………………………………… 22 Manuscript 2      In drug delivery, application of nanotechnology is commonly referred to as Nano Drug Delivery Systems (NDDS). In this article, commercially available nanosized drugs, their dosage forms and proprietors, as well as the methods used for preparation like milling, high pressure homogenization, vacuum deposition, and high temperature evaporation were listed. Unlike the traditional methods used for the particle size reduction, supercritical fluid-processing techniques offer advantages ranging from superior particle size control to clean processing.
The primary focus of this review article is the use of supercritical CO 2 based technologies for small particle generation. Particles that have the smooth surfaces, small particle size and distribution and free flowing can be obtained more, via few SCF techniques. In almost all techniques, the process variables involved may be of thermodynamic and aerodynamic nature and the result of the design of the particle collection environment.

Rapid Expansion of Supercritical Solutions (RESS), Supercritical Anti Solvent (SAS)
and Particles from Gas Saturated Solutions (PGSS) are three groups of processes which lead to the production of fine and monodisperse powders. Few of them may also control crystal polymorphism. Among the aforementioned processes, RESS involves dissolving a drug in a supercritical fluid (SCF) and passing it through an appropriate nozzle. Rapid depressurization of this solution causes an extremely rapid nucleation of the product. This process has been known for a long time but its application is limited. Carbon dioxide, which is the only supercritical fluid that is preferentially used in pharmaceutical processes, is not a good solvent for many Today, nanonization may help drug substances to be actively targeted to the site of action, or directly taken by the cells and for delivery of the genes. However, their application for solubility enhancement is the most frequently applied reason.
Reduction of the particle size of the poorly soluble drugs to nano scale increases their dissolution rate, saturation solubility, and in turn, the oral bioavailability; therefore, nanonization is becoming a very popular process enabling the use of such drugs [1][2][3][4].
The effect of nanonization on solubility improvement of a drug substance can be explained by modified Noyes Whitney equation [5] which demonstrates the relationship among the rate of dissolution and solubility as well as the surface area and particle size of a given drug substance. On the other hand, Kelvin equation [6,7] helps us to understand the physics behind the particle size effect on saturation solubility when the particle size is reduced to nano scale. The Kelvin equation may be written as; Eq. (1) where p is the actual vapor pressure, p 0 is the saturated vapor pressure, γ is the surface tension, V m is the molar volume, R is the universal gas constant, r is the radius of the droplet, and T is temperature.
Kelvin equation mathematically describes that, the vapor pressure of lipid droplets in a gas phase (aerosol) increases with an increase in the curvature of the surface of the dispersed phase which is realized by the particle size reduction of such a system. The vapor pressure is equivalent to the dissolution pressure. In the state of saturation solubility, there is equilibrium between the molecules dissolving and molecules recrystallizing. This equilibrium can be shifted when the particle size is reduced, which causes an increase in the dissolution pressure by increasing the saturation solubility.

INDUSTRY
Since 1990s, nanotechnology continues to gain popularity not only in the health care industry but in many other industries such as information and communication, energy production, food and agriculture, aerospace, construction etc. The term nanotechnology is used for any system of producing structures or devices in the nanometer range (1 nm=one thousand millionth of a meter, 10 −9 m). Generally, size range of the nanoparticles fall between 1-100 nanometers. In some drug delivery systems, this range may exceed the nanosizes, while it is still being considered as nanoparticles. Application of nanotechnology in health care arena is by far the most promising and beneficial for drug delivery, improved therapies, in vivo imaging, in vitro diagnostics and for production of biomaterials and active implants. In drug delivery application of nanotechnology is commonly referred to as Nano Drug Delivery Systems (NDDS).
Examples of commercially available NDDS are given in Table 1  The makers and the users of NDDS claim that, nanonization reduces the side effects of the drugs; improve efficacy and therapeutic effectiveness in disease stages that currently cannot be treated with conventional drugs. Interest of the pharmaceutical industry in NDDS is increasing worldwide [8][9]. There are several companies ranging from start up to a large corporations currently working in the field of NDDS, across the world.

SCALE
A major challenge in reducing the particle size of solids to nano scales is the accurate control of the size and shape, which in turn is directly linked with the nano materials processing method. There are two approaches generally used to reduce particles to nanosizes: bottom-up and top-down methods. In the bottom-up approach a colloidal solution of drug is prepared and the solvent is evaporated obtaining controlled rearrangement of single atoms and molecules into larger nanostructures. Supercritical fluid based technologies utilize this approach. The top-down approach reduces the particle size of drug particles using media milling or high pressure homogenization, or similar alternative methods. Table 2 provides an overview of industry leaders in each of these technologies.

Many academic interests
This article focuses on supercritical fluid use as a method of choice for producing nanoparticles.

SUPERCRITICAL FLUIDS (SCF)
"Supercritical" is a state of a substance above its critical temperature (TC) and critical pressure (PC). A substance in its supercritical state is defined as Supercritical Fluid (SCF). The critical point represents the highest temperature and pressure at which the substance can exist as a vapor and liquid in equilibrium.
In the supercritical stage, there is no phase boundary between the gas phase and the liquid phase. In short, it can behave as if it is a liquid or a gas, but is actually neither.
The properties of SCF are in between that of gas and liquid. The densities of a substance in its supercritical state are either the same or close to that of same substance in its liquid state. This property allows SCF to enhance solubility of poorly soluble drugs more than the gaseous state could. On the other hand, the diffusivity and viscosity of SCF are close to that of gas; which allow rapid mass transfer or penetration of SCF into materials than that of the liquid states.
SCFs are highly compressible, particularly near the critical point, and their density and thus the solvation power can be carefully controlled by small changes in temperature and/or pressure [15,16]

Solubility of pharmaceutical compounds in supercritical carbon dioxide
Even though the mechanism of particle formation in different SCF based technologies is different, they all rely on either the solubility or insolubility of the solute (drug &/or polymer) in scCO2. Hence, determining the solubility of a solute of interest in scCO2 is important in choosing the SCF technology. This can be done experimentally or via theoretical estimation. The solubility of a pure solid component (2) in a supercritical fluid (1) can be expressed as a function of the operating pressure P and temperature T, as described in Equation (2) Eq. (2) where y 2 is the equilibrium mole fraction of the solid component (2) in the supercritical fluid phase, P 2 SAT is the saturated vapor pressure of the solid component By controlling the pressure and the temperature, the density and solvent strength of SCF can be altered to simulate organic solvents ranging from chloroform to methylene chloride to hexane. scCO2 is a relatively non polar solvent, therefore if a compound is soluble in hexane, it should also dissolve in scCO2. It is also possible to modify the solvation power of a particular SCF by incorporating a small amount of volatile cosolvent, like acetone or ethanol.
When scCO2 is used as an anti-solvent, the key to particle production is generally the super saturation of the solution of materials via the counter-diffusion of scCO2 and the solvent. The insolubility of the solute in scCO2 influences the degree of this super saturation.

PHARMACEUTUCAL INDUSTRY
SCF based technologies are extremely flexible. Some applications are already at industrial capacity, whereas others remain under development. They include particle formation, extraction of trace amounts of organic solvent from the drug substances, impregnation of drug into polymer, coating, and reactive systems such as hydrogenation, biomass gasification, and supercritical water oxidation. Table 3 lists some of the well documented pharmaceutical applications of SCF technologies.
However, this review article focuses on its use for nanoparticle production.

Pharmaceutical & Biomedical Applications Reference
Particle formation/particle size reduction (micron and nano scale) [17][18][19][20][21][22][23] Residual organic solvent stripping (for drying purposes) [24][25][26][27] Impregnation of drug into polymer to prepare solid dispersion [28][29][30] Encapsulation, coating [31][32][33] Polymer processing (eg. Extrusion) [34][35][36][37][38][39][40][41][42] Liposome involved in these processes may lead to thermal and chemical degradation of some of the drugs and ingredients. Therefore, a method that particulates a drug substance in a cGMP compliant manner, which produces controllable particle properties, and requires minimal downstream processing is definitely the most suitable method for manufacturing a wide range of therapeutic agents [16,38].  difference amongst these processes is the role a SCF, whether it acts as a solvent or an anti-solvent or as a solute, in formation of the particles. These technologies are further modified based on the particle growth mechanisms and their collection environment. There are many more modified processes which are not described in this review article as they have not created a wide interest in the production of drugs or drug products.

INFLUENCE OF OPERATING PARAMETERS
In almost all of the SCF processes ,the morphology (crystalline, amorphous, or both), the particle size, shape and distribution of the resulting product depend on factors like the properties of the material [67, 68], the process variables i.e. the thermodynamic and aerodynamic factors [69][70][71][72][73][74][75][76][77][78][79], and to the design of the particle collection environment [22]. Among the thermodynamic factors temperature, pressure conditions, rate of addition of one component to another and, phase and composition changes during the expansion can be listed. The aerodynamic factors include impact distance of the jet against a surface, nozzle geometry and mechanical shear that a particle undergoes. Table 5 summarizes these functional variables, which affect the properties of the finished product in almost all SCF processes. between a solvent and a dense gas antisolvent at conditions that are supercritical with respect to the solvent-antisolvent mixture. The extent of droplet swelling or shrinking as a function of temperature and pressure was correlated to the difference in density and diffusivity between the solvent-rich and antisolvent-rich regions. The difference created in the droplet behavior at subcritical and supercritical conditions and their implications for particle production was also studied by the same researchers [77].
They concluded that supercritical conditions results in faster mass transfer suggesting that in the presence of a solute, in supercritical operations causes a higher degree of droplet super-saturation, resulting in higher nucleation rates and smaller particles. They argued that spherical particles are the result of droplet drying after effective atomization of the liquid solution. At a pressure greater than critical pressure, the particle formation mechanism is in competition with the surface tension vanishing followed by the formation of a gas plume. When the latter mechanism prevails, atomization is not obtained and the nanoparticles are produced by precipitation from the fluid phase. The same group further demonstrated the proposed mechanism via in situ laser scattering techniques. They were able to study the location of first particle precipitation inside the vessel, the partial densities of all species in the system, the super-saturation, and overall mixture composition [79].
Reverchon et al. [78], also tabled a large number of published work and listed drugs that were obtained in micron sizes with SCF, at different processing conditions. They sought relationships among the solute concentration, the vessel temperature, the pressure inside the vessel, and molar fraction of CO2 (X CO2 ). Upon studying the data available, they have concluded that the X CO2 larger than approximately 0.95 -0.97 is the most useful ratio for small particle production. At this condition, the binary mixture of CO2 and liquid organic solvent achieves the mixture critical point (MCP).
The effects of solute concentration, the pressure and the temperature on the particle size and morphology were difficult to generalize, because data available was contradictory.
Although each study has emphasized, the effects of one or two variables that are the most critical in obtaining small particles, due to the complexity of the procedure, there is not a published work which takes into account, expansion in the nozzle and in the jet, along with nucleation, growth, and agglomeration in addition to the factors mentioned above.   In any of the SCF based particle formation techniques, preserving the particle characteristics such as morphology and size is very critical; therefore, the particle collection is an important task. Particle agglomeration is a common problem found during RESS process which is worsened if residual amount of co-solvent remained in the processed material. Researchers have used different particle collection environments in order to overcome the agglomeration problems which will be presented in the following section of this article.

Advantages and limitations of RESS
RESS is an attractive method since it is a single step process which requires minimum to no organic solvent, and can be implemented relatively easily at least at a small scale. During the rapid expansion, the solute experiences both the pressure and temperature quenches simultaneously that enhance the precipitation process considerably.
However, RESS process is only applicable to those solutes which exhibit good solubility in scCO2. Majority of the poorly soluble active pharmaceutical ingredients (APIs) have high molecular weights and polar bonds, and are excellent candidates for nanoparticle preparation. Unfortunately, many of them have low to negligible solubility in scCO2 at moderate temperatures (less than 60 °C) and pressures (less than 300 bar). Co-solvents, such as methanol, may be added to carbon dioxide to enhance solubility of the drug. However, such applications will alter environmentally safe nature of the RESS process. The need for removal of residual organic solvents will further increase the cost and complexity of the process.

Summary of RESS applications for production of spherical nanoparticles
Over the last 2 decades, RESS technology has been successful in the production of There are several variations of the basic RESS process which were designed to minimize the agglomeration problems (Table 5). Amongst these, Rapid Expansion of a Supercritical Solution into a Liquid Solvent (RESOLV) [22] and Rapid Expansion from Supercritical to Aqueous Solution (RESAS) [84,85] are the most notable ones.
In RESOLV, the traditional RESS is modified by expanding the supercritical solution into a liquid solvent instead of ambient air. Pathak et al. [22] demonstrated that RESOLV technique can successfully produce individual and spherical particles of naproxen and ibuprofen in nanoscales when they expanded the product into aqueous solution containing PVP as a polymer. Without PVP, the particles obtained were nanosized, but agglomerated and non-spherical. Mechanistically, the liquid at the receiving end of the rapid expansion in RESOLV, suppresses the particle growth in the expansion jet, thus allows nano sized and round particle production.
In In order to overcome the low solubility of polar drugs in the SCF solvent, Thakur and Gupta [86] proposed a modified RESS process that used a solid cosolvent (RESS-SC). They tested this process for nanoparticulating phenytoin by the use of menthol, utilized as the solid cosolvent. In the conventional RESS process, each particle is surrounded by the same drug particles in the expansion zone that results in coagulation and formation of larger particles, as illustrated in Fig. 3 (a). In the RESS-SC process, the drug particles were surrounded by particles of solid cosolvent like menthol . Hence, the particle growth could be minimized in the expansion zone resulting in smaller nanoparticles as schematically illustrated in Fig. 3 (b). Phenytoin particles surrounded by menthol, avoided surface to surface interaction with other phenytoin particles. The cosolvent (menthol) could be easily removed by sublimation using a lyophilizer following the particle recovery from the expansion chamber. and isopropanol. In order that the particle precipitation occurs, the solute must be virtually insoluble in carbon dioxide while the organic solvent must be completely miscible with carbon dioxide at the precipitation temperature and pressure.
Collection of the precipitated particles in the antisolvent processes is carried out in the same vessel where solvent extraction takes place. The particles are collected on the filters, located at the bottom of the vessel. Additionally, a drying cycle is performed at the end by passing a generous amount of SCF to remove any unextracted solvent. Minor modifications of this basic principle are applied in GAS and SAS. GAS is favored by some researchers as it is a slow process and allows the growth of the particles in a controllable manner. However, it has rarely been successfully scaled up to an industrial magnitude. Fig. 4 (b) provides the schematic representation of the SAS process. Unlike GAS, this technique utilizes gas in its supercritical stage as an antisolvent for the solute. In addition, the mechanism involved is different than that employed in the GAS process.

Mechanism of particle formation in GAS, SAS/ASES
The solute is first dissolved in a liquid solvent and then this solution is sprayed using a nozzle into a chamber which contains the supercritical fluid (antisolvent). The supercritical fluid dissolves in the liquid solvent droplets followed by a large volumetric expansion by reducing the solvent power of the liquid. As a consequence, the super saturation of the liquid mixture increases causing formation of small and uniform particles. Unlike GAS, this technology has produced favorable results during scale up to industrial capacity [87]. As mentioned before, the ASES (aerosol, solvent extraction system) is the same as SAS in principle.

Applications of GAS/SAS/ASES
Kalogiannis et al.
[88] used SAS technology to produce amorphous nanoparticles of amoxicillin within the range of 500 nm to 800 nm when they have used DMSO as the organic solvent. When they partially replaced dimethyl sulfoxide ( DMSO) with EtOH and MeOH, the particle size was reduced to the range of 350 nm. Reverchon et al.
[89] used semi-continuous SAS technique to produce rifampicin micro-and nanoparticles with controlled particle size (PS) and particle size distribution (PSD). SAS experiments were performed using different liquid solvents. When they used DMSO and 40 °C operating temperature, they obtained amorphous nanoparticles with mean diameters ranging from 400 nm to 1000 nm. HPLC analysis showed no degradation during supercritical processing. They also observed that, when the liquid concentration was increased, the mean PS increased and the PSD was widened.
Lee et al. [29] demonstrated that ASES could be a promising technique not only to reduce the particle size, but also to prepare amorphous solid dispersion of itraconazole with a hydrophilic polymer, HPMC 2910. The particle size of solid dispersion prepared, ranged from 100 to 500 nm. Authors verified that itraconazole was molecularly dispersed in HPMC 2910 in an amorphous form.
Production of round nanoparticles with SAS and modified SAS methods is doable.
Cefnidir was nanoparticulated to 150 nm size by Park et al. [90] by using methanol as the solvent. They calculated the Intrinsic Dissolution Rate (IDR) of the processed and unprocessed particles and found out that cefidinir nanoparticles prepared with SAS had 9.42-9.94 times more dissolution.
Sanganwar et al [91] and Chattopadhyay and Gupta [92] modified the conventional SAS process in order to minimize the agglomeration of drug particles. Sanganwar et al [91] produced microparticles of a poorly water-soluble nevirapine using SAS method. The drug was simultaneously deposited on the surface of lactose and microcrystalline cellulose respectively in a single step, to reduce drug-drug particle aggregation. Another technique named as "Supercritical Antisolvent-Drug Excipient mixing (SAS-DEM)", was used to minimize particle aggregation. In this method, the drug, dissolved in dichloromethane, was precipitated in the scCO2 vessel, which contained suspended excipient particles. The SAS-DEM treatment was effective to minimize particle aggregation but has not interfered with the crystallinity or physicochemical properties of nevirapine. The drug/excipient mixture obtained, had a significantly faster dissolution rate compared to SAS processed drug microparticles alone or of its physical mixtures prepared with the same excipients. Chattopadhyay and Gupta [92] also modified the conventional SAS technique and included a step where the solution jet was deflected by a surface, vibrating at an ultrasonic frequency, that atomized the jet into small micro droplets. This technique which is called SAS-EM (enhanced mass transfer) produced griseofulvin nanoparticles of 130 nm size.
Many more modifications of conventional SAS technique to overcome the challenges faced in SAS were reported. One such modification is called Supercritical-Assisted Atomization (SAA). SAA technique that was reported by Reverchon [93] was used to produce micro-and nanoparticles of several pharmaceuticals with controlled size and distribution. In this technique, controlled quantities of scCO2 was mixed with solutions containing a solid solute and the entire ternary solution is subsequently atomized through a nozzle. The technique successfully micronized some superconductor, ceramic, and catalyst precursors as well as several pharmaceutical compounds, such as carbamazepine, ampicillin trihydrate, triclabenzadol, and dexamethasone. Liquid solvents used to form the starting solution were methanol, water and acetone. The author explained the mechanism of the t SAA process as, formation of primary small droplets by atomization of the liquid in the thin wall nozzle in Step 1. In Step 2, due to the extremely rapid release of CO2 from inside of the primary droplets, the droplets formed broke up (decompressive atomization), by forming smaller secondary droplets. Creation of primary and secondary droplets eventually resulted in formation of submicron drug particles.
Another modification of SAS was proposed by Rodrigues et al. [94] and is termed

Atomization of Supercritical Antisolvent Induced Suspensions (ASAIS).
Mechanistically, ASAIS is similar to SAA where a small volume supercritical antisolvent is dissolved in line, with the liquid solvent before the liquid atomization for the solvent extraction step. Mixing of scCO2 in a small volume immediately before the nozzle orifice, leads formation of conditions such that causes the precipitation of the solute and the suspension formed in this way, is then spray-dried for solvent separation. The process was successfully demonstrated to produce submicron particles of theophylline using tetrahydrofuran as the organic solvent. The authors argued that compared to other similar particle-production techniques, this approach allowed a more efficient control of the antisolvent process and reduced the volume of the high-pressure precipitator by several orders of magnitude.

FLUIDS (SEDS)
Mechanism of particle formation SEDS process produced particles of salmeterol xinafoate with a polymer matrix [95].
Two separate solutions of the active substance and the polymer (hydroxypropylcellulose) which was dissolved in acetone, were prepared, and cointroduced with supercritical CO2 in a precipitator, using a three-passage nozzle.
Analysis made, confirmed inclusion of the drug into the polymer matrix. In this manner, Ghaderi et al. [96] produced spherical microparticles of hydrocortisone entrapped within the biodegradable polymer poly( D,L -lactide-co-glycolide) ( DL -PLG) by using a combination of supercritical N 2 and CO2, at 130 bar pressure and 38 0 C temperature . The use of N 2 simultaneously with CO2 improved the homogeneity of mixing and led to a more efficient integration of the polymer and the drug.
Chen et al. [97] produced nanoparticles of puerarin and microencapsulated them with poly(L-lactide) (PLLA) by using a modified SEDS process. The modification included an "injector" which injected nanoparticles of puerarin inside the polymer solution in dichloromethane . The Puerarin nanoparticles obtained, exhibited good spherical shapes, smooth surfaces and a narrow particle size distribution with a mean particle size of 188 nm. After microencapsulation, the Puerarin-PLLA microparticles had a mean size of 675 nm. A drug load of 23.6% and an encapsulation efficiency of 39.4% was obtained. Data obtained clearly demonstrated that this process is a promising technique to prepare a drug-polymer carrier for a drug delivery system.

OPERATIONS WHERE SCF ACTS AS SOLUTE
Particles from Gas Saturated Solution (PGSS): Fig. 6 provides schematic illustration of PGSS equipment set up. As discussed earlier, many drug substances are either polar or have high molecular weights. It is difficult to dissolve these compounds in CO2, which is a non polar solvent, even in a supercritical state without the aid of a cosolvent. On the other hand, scCO2 has the ability to diffuse into organic compounds, such as polymers. When scCO2 diffuses into the polymer, it lowers the melting point and decreases its viscosity. These characteristics are made use of in PGSS process.
In the PGSS operations, the physical mixture of the drug and the polymer is first Sencar-Bozic et al. [98] made composite microparticles of nifedipine and poly ethylene glycol, (PEG 4000), using the PGSS process. They showed that the solid dispersions had increased dissolution rates of Nifedipine. Similar results were reported for the anti-angina drug felodipine by Kerc et al. [19]. Rodrigues et al. [99] prepared the micro particles of theophylline with hydrogenated palm oil (HPO) by the PGSS process. Particle size obtained was about 3.0 μm in diameter. Spherical morphology with a regular surface was obtained at higher expansion pressures. This technology has successfully enlarged to an industrial scale [100]. The PGSS process has similar advantages to those of the RESS processes. It can be performed without using an organic solvents. It usually requires lower pressures and gas consumption than the RESS processes . One problem typically associated with the conventional PGSS process is separation of the ingredients as they pass across the pressure drop. This phenomenon can result in segregation of the components.
Particles of the drug and the polymer are formed separately, but the polymer microspheres containing the drug could not be obtained PGSS also has been modified to overcome the agglomeration and non uniform particle size distribution problems.
Skekunov et al. [101] proposed a technique to overcome the segregation problem, by using two separate mixing chambers in the equipment set up. In the first chamber, the drug and the polymer are mixed to homogeneity, allowing them to melt in scCO2.
This melt was then passed from first chamber to the second one where it was mixed with more SCF, causing further reduction in the viscosity of the melt. The mixture was finally sprayed and via further expansion occurred, the uniform micro-particles of the polymer-drug mix were obtained.
Another successful modification was proposed by Hu et al. [102]  This is an advantageous technique compared to other drying processes such as spraydrying. It allows drying of the solutions with a reduced thermal degradation or contamination of the solid substance, because the process is carried out in a closed system inertized with CO2. The only instrumental part of the process operating at the high temperature is the static mixer. This process was successfully used by Martin et al [103] to micronize polyethylene glycol from aqueous solutions, by producing spherical PEG particles with average particle sizes of 10 microns and residual water content below 1 %.

Carbon dioxide assisted nebulization with a Bubble Dryer® (CAN-BD)
This is also an example of a process where CO2 is used for drying purposes. CAN-BD process can dry and micronize pharmaceuticals that are especially used in the pulmonary drug delivery. In this process, the drug is first dissolved in water or an alcohol (or both), and is mixed very well with near-critical or supercritical CO2 by pumping both fluids through a low volume tee to generate microbubbles and microdroplets. These microbubbles are then decompressed into a low temperature drying chamber, where the aerosol plume dries in seconds. Similar to PGSS drying, there is less decomposition of thermally labile drugs. In this method, there is no need for a high pressure vessel and the particles obtained are generally 1-3 microns in size [104].

CONCLUSION
The temperature, and the nozzle geometry on the particle size, shape, surface properties and distribution. As we continue to improve our fundamental understanding of the SCF chemistry, we can reliably scale up more of the SCF processes to obtain free flowing drug particles of nano sizes.
Throughout the years, various modifications made on the original SCF process, improved the properties of the particles obtained and small spheres with smooth surfaces, and narrow particle size distribution were produced. However, scaling up several of SCF processes is still difficult. It is is disappointing to realize that although the literature has numerous articles that have mechanistically investigated the particle formation process, commercializing this technology for the drug delivery applications has not been very successful.
In spite of the disadvantages mentioned, and the questions remained unanswered to obtain more reliable and repeatable applications, the SCF technology appears to be an exciting tool to process nanoparticulated drugs of the future. It has matured greatly over the last 10 years and has a great potential of becoming a key drug delivery technology in the near future.

ABSTRACT
Particle size reduction to micro and nano scales using supercritical antisolvent (SAS) methodology is an effective and versatile option for solubility and bioavailability improvement of poorly water-soluble drugs. However, there are several factors that influence the particle morphology, particle size, and particle size distribution when SAS methodology is applied to produce nano particles. Hence, a successful application of SAS technology to drug particle production requires a careful evaluation of these factors. A fractional factorial 2  screening design of experiments is applied to supercritical antisolvent precipitation of griseofulvin using carbon dioxide (CO 2 ) as an anti solvent, and acetone as solvent. The design of experiment (DOE) proposed is useful for identifying the key factors involved in the SAS process in just a few runs at an early stage of experimentation. Seven factors were studied at two levels each. Mean particle size (PS), particle size distribution (PSD), and % yield of the SAS process were chosen as responses to evaluate the process performance.
Statistical analysis of the results from DOE study identified the nozzle diameter and spray rate of organic solvent as two most significant factors affecting PS and PSD.
Temperature of the precipitation vessel only impacted the particle size, whereas, the pressure, drug concentration and polymer concentration affected PSD. Lastly, the yield of the SAS process was impacted by drug concentration, polymer concentration, and the pressure condition inside the precipitation vessel. We were able to rank order these factors in terms of their overall impact on all three responses.
Based upon the outcomes of this study, an optimum and robust SAS process was developed. Operating at 45 0 C, 80 bar pressure, 20 mg/ml drug concentration, 5 mg/ml polymer concentration, solvent spray rate of 2 ml/min, CO 2 addition rate of 40 g/min, and using nozzle diameter of 150 µm, in-situ nanoparticles d 50 (volume based mean particle size of 50th percentile) of 0.362 µm were produced. Scanning Electron Microscopy revealed that coprecipitates of drug and polymer were fluffy and fibrous in nature. DSC analysis as well as PXRD revealed that the co-precipitates were in crystalline form. FTIR study of the products confirmed that there was no interaction between drug and the polymer. Lastly, formulations obtained with the SAS process had significantly improved rate of dissolution compared to that of the physical mixture of as-is drug and the polymer.

INTRODUCTION
Technologies that can effectively control particle formation are of utmost importance in the pharmaceutical industry. A reduction in particle size of poorly water soluble drug to the ultra-fine state such as nano scale, increases the surface area, results an increase in the dissolution rate, saturation solubility, and in turn, the bioavailability. "Supercritical" is a state of a substance above its critical temperature (T C ) and critical pressure (P C ). A substance in its supercritical state is defined as a supercritical fluid (SCF). Carbon dioxide is the most preferred SCF for the processing of pharmaceuticals, because it has low critical temperature (31.2°C) and pressure (73.8 bar or 7.4MPa), and is nonflammable, nontoxic and environmentally safe. It is highly compressible and its density and thus the solvation power can be carefully controlled by small changes in temperature and/or pressure 5,6 . The solvent strength of supercritical CO 2 can be altered to simulate organic solvents ranging from chloroform to methylene chloride to hexane.
Primarily there are two categories of SCF based particle formation techniques; using SCF as solvent or as an anti-solvent. Example of SCF use as the solvent is Rapid Thermodynamic properties 9-10 such as the temperature, pressure, phase composition, rate of addition of one component to another; and aerodynamic properties 11 such as nozzle geometry and impact distance of jet against a surface are the most important process variables that impact the outcome of SAS precipitation. In order to design a robust SAS process, it is extremely important to understand the impact of all of these variables, on the desirable SAS product attributes, such as particle morphology, particle size, particle size distribution, and % yield of the process.
Although several researchers [12][13][14][15][16] have studied these variables, there is widespread disagreement amongst them. For example, Guha et al., 12 using Cholesterol & Poly (L-lactic acid) in dichloromethane (DCM) found that increase in pressure leads to smaller particle size. Similarly, Reverchon et al., 13 working with SAS and using griseofulvin, tetracycline, and amoxicillin in dimethylsulfoxide (DMSO), and DCM also concluded that, increase in pressure leads smaller particle size. However, Lee et al., 14 who carried out SAS precipitation of itraconazole and HPMC2910 in DCM-Ethanol mixture, found that an increase in pressure produced larger particles.
Randolph et al., 15 as well, using SAS, found that an increase in pressure produced larger particles of Poly (L-lactic acid) from methylene chloride. Whereas, when Uzun et al., 16 carried out SAS processing using methanol on cefuroxime axetil in the presence of PVPK30, concluded that the change in pressure did not affect the particle size significantly.
One of the reason of such wide-spread disagreements amongst the researchers, in defining the variables that impact particle properties during SAS process, is that We have used conventional SAS methodology to reduce the particle size of GF.
However, our attempts were unsuccessful to produce nanoparticles of GF, which prompted us to explore a co-precipitation with a polymer. A polymer, which can potentially act as a crystal growth inhibitor would be added to the formulation. Such an approach can be beneficial from two perspectives: firstly, it would prevent the uncontrolled crystal growth, and secondly, the polymer could act as a stabilizer to prevent the aggregation of formed micro particles.

Selection of polymer
There are various factors that affect the selection of a polymer in SAS processing, such as: a suitable drug-polymer interaction, solubility of a polymer in water, and in organic solvents, global regulatory acceptance, stabilizing ability of a polymer in an aqueous environment, and ease of processibility. We identified three hydrophilic polymers, having different molecular structure, and ionic properties; namely Kollidon® VA64, HPMCAS, and Eudragit EPO®. EUDRAGIT ® EPO is a cationic copolymer based on dimethylaminoethyl methacrylate, butyl methacrylate, and methyl methacrylate. It is also practically insoluble in water, and has a pH dependent solubility in buffered media.
We wanted to develop a product that may be relevant clinically, and hence the pharmaceutical polymer selected must be recognized by the United States FDA as GRAS (generally regarded as safe) and must have prior precedence of being used in an oral dosage form. The polymer should be well tolerated, not known to have any toxicity, and must have been used for decades in the pharmaceutical industry for various applications. Finally, the selected polymer should be freely soluble in the organic solvent chosen for SAS processing, and in water. where practically all formulation and processing factors will be included and we believe that such a study will be more reliable in identifying the key variables involved. Secondly, we will follow the changes in the particle size, particle size distribution, and percent yield of product during the process, as the specific responses for measuring the process performance. Lastly, we will apply the optimum operating variables and produce coprecipitates of GF which, when placed in an aqueous medium would yield in-situ micro and nanoparticles of GF having a narrow particle size distribution, while obtaining the highest yield and improved solubility.

Materials
The model drug griseofulvin (GF) was purchased as micronized API (lot # 115H1180 ) through jet milling process, from Ria International (East Hanover, NJ). has dissolution rate limited absorption 24 . Thus, particle size reduction to nanosize is highly desirable. Hence, GF serves as a good model drug for demonstrating the utility of SAS for improving drug solubility and product performance by the formation of nanoparticles.  (from left to right)

Figure 3: Schematic diagram of Tharr SAS apparatus
An SAS apparatus (model: SAS 50, Thar Technologies Co., USA) was used to generate GF-polymer co-precipitates. Figure 3 shows the schematic diagram of Tharr SAS system using supercritical carbon dioxide (SCCO 2 ) as an anti-solvent. The SAS 50 system is made up of the following components: two high pressure pumps, one for the CO 2 and the other for the organic solvent; a stainless steel particle collection vessel (0.5 L volume, 54 mm internal diameter and 218 mm internal height) consisting of the main body, the frit, and electric heating jacket ; and an automated back pressure regulator (ABPR) of high precision. Firstly, the CO 2 coming from tank passes through low pressure heat exchanger (HE1) and was cooled down with a cooling bath operating at 4 °C to assure liquid state in the pump. The liquefied CO 2 was then pumped into particle collection vessel using a high-pressure pump through the spray nozzle. CO 2 was heated using another heat exchanger (HE2) before entering the precipitation vessel. Stainless steel orifice nozzles of 100 μ, 150 μ, and 200 μ were used, depending on a particular experiment, as laid out in Table 5. The selection of organic solvent between acetone and DMSO was based on the quality of GF crystals obtained and ease of solvent removal from product. Acetone was preferred as the organic solvent. The flow-rate of CO 2, and acetone were adjusted via using computer software. The pressure in particle collection vessel was controlled using an automatic back-pressure regulator (ABPR), and a temperature controller regulates the amount of heat being applied to the heating jacket.
At the beginning of the experiment, acetone was sprayed for 10 minutes to establish the steady state condition. After that, the solvent pump was used to spray the solution of drug and polymer into the precipitation vessel through the spray nozzle. Finally, the powder was collected from the inside of the precipitator for further characterization.

Testing reproducibility of SAS process
In order to find out the variability of the processing method, two sets of experiments were repeated three times each. In these experiments, 150 ml of acetone solution was sprayed in the SAS vessel, and Kollidon VA64 was used as the polymer. All experiments were carried out at 2 ml/min acetone solution spray rate, 20 g/min CO 2 rate, and nozzle diameter was kept at 100 µm. Drug concentration was tested at 15 and 20 mg/mL, polymer concentration at 20% and 70%, temperature at 35 and 45 0 C, and pressure at 80 and 85 bar. Particle size and particle size distribution measurements were performed using Malvern light diffraction method, as described in later sections. Table 2 summarizes the results, and shows that the yield of the process was 81 + 3.6 % for formulation with 70% polymer, and 88 + 3 % for formulation with 20% polymer; whereas d 50 values were 239 + 31 nm, and 370 + 3 nm for two different formulations. Lastly, the span values, measure of particle size distribution, were 5.377 + 0.47, and 2.534 + 0.03. We believe that the higher variability in particle size and particle size distribution for the first set of experiments in

Evaluation of solubility of drug, and miscibility of organic solvents in scCO 2
The RESS 50 (Tharr Technologies Co., USA) system was used for the preliminary solubility and miscibility evaluations. The system, schematically shown in Appendix A, is made up of the following components: a high pressure CO 2 pump; a high pressure agitator; and a stainless steel extraction vessel with a sapphire view cell carved in the middle of vessel. The extraction vessel consists of the main body, an electric heating jacket, and an automated back pressure regulator (ABPR) of high precision.
In this apparatus, firstly, the CO 2 coming from tank passed through low pressure heat exchanger (HE1) and was cooled down with a cooling bath operating at 4 °C to assure liquid state in the pump. The liquefied CO 2 was then pumped into extraction vessel using a high-pressure pump. The CO 2 was heated using another heat exchanger (HE2) before entering the extraction vessel. The pressure in extraction vessel was controlled using an automatic back-pressure regulator, ABPR. A temperature controller provides pre-specified heat to the heating jacket. Agitator mounted on top of the extraction vessel provides mixing to the contents of the vessel.
At the beginning of the experiment, an untreated as-is GF or raw organic solvent (acetone or DMSO) or mixture of GF and organic solvent, was placed in the extraction vessel. Agitator was turned on to gently mix the contents of the vessel.
After that, the CO 2 pump was turned on to fill the extraction vessel. Temperature was gradually raised from 35 to 100 0 C, and the pressure gradually raised from 80 to 300 bar. During this time, visual observations were made through the view cell.

Characterization of particles 2.2.4.1 Particle size and particle size distribution
Particle size (PS) and particle size distribution (PSD) were measured using a The results, shown in Figure 4, provide an assurance that measurements made on SAS formulations will be accurate, within the instrumental limitations.  In order to verify the reproducibility of Malvern particle size measurements, one of the SAS formulation sample was run 10 times. The average value for d 50 was 371.5 + 10.7 nm, with a relative standard deviation (RSD), also called coefficient of variation, of 2.88%. The results demonstrate that the reproducibility of measurements is acceptable.
Samples obtained with SAS were sonicated prior to particle size measurement. Samples were exposed for 0, 0.5, 1, 3 and 5 minutes of sonication. Results showed that there was no further reduction in sample particle size after 0.5 minutes sonication. Hence, 1 minute sonication was chosen as the standard sonication, which was sufficient to dissolve the polymer and to obtain sufficient de-agglomeration.

Surface appearance
The surface morphological analysis of the sample was performed using an ultra-high

Description of the statistical method 2.2.5.1 Design of Experiment (DOE)
Determination of statistically significant variables of the SAS process was made by the use of DOE approach. Analysis of the data from DOE identifies the statistical importance of different factors on the measured responses, and the methods are described in basic textbooks 27 . We utilized MODDE® software (UMETRICS, version 9.0.0) for designing the fractional factorial study and for performing analysis on the data obtained.
Using a conventional approach, to study k different factors, each having only two levels, the minimum number of experimental runs needed is 2 × 2 × ... × 2 = 2 k . We

Factor and level selection
Seven factors that were identified and used for this screening DOE study were temperature (T) , and pressure (P) of the precipitator; drug concentration (C d ), and polymer concentration (C p ); organic solvent (F liq ) and antisolvent (F CO2 ) flow rates; and nozzle diameter, d n . The two levels for each of the seven factors selected are shown in Table 3. Values were chosen based on the preliminary experiments conducted with SAS while taking into account limitations of our equipment.  Drug concentration below 10 mg/mL resulted in no precipitation and hence was chosen as the lower limit. The upper limit of 20 mg/mL was established based on solubility limit of drug in acetone, which is 25 mg/mL. Polymer concentration limit of 20-70% was selected to see the effect of polymer on drug precipitation mechanism over a wide range of polymer concentration.
Flow rate limits for acetone solution and supercritical carbon dioxide were established to obtain both a wide range of acetone mass fraction inside the precipitator vessel and a wide range of CO 2 /acetone flow ratio. In all of these experimental conditions, CO 2 mole fraction was always maintained at 0.91 or higher.
Maintaining high CO 2 fraction is important to obtain optimum effectiveness of CO 2 as an anti-solvent, as reported by other researchers as well 9 .
It was noticed during the preliminary experiments that the low level acetone flow rate of 1 mL/min in combination with 200 μ diametered nozzle did not atomize the solution, and no precipitation could be obtained. As a result, a higher flow rate (2 -3 ml/min) was selected. We were limited to 100 µm diametered nozzle as the smallest nozzle diameter, because it was the smallest-diametered nozzle available in the market for this instrument. The upper limit of nozzle diameter was setup to 200 µm.

Response identification
Three responses which were d 50 (particle diameter of 50 th percentile of distribution), span (the measure of particle size distribution), and % yield, were selected. The yield in percentage is calculated as the ratio of the solute (drug + polymer) processed and the amount collected following SAS processing. Particles are collected on top of 0.22 µm filter paper in the bottom of vessel, as well as scrapped from vessel walls. When a sticky mass is obtained around the nozzle apart from the bulk powder, this portion which was not usable was discarded and not considered in the yield calculation.

2 (7-3) Fractional factorial design generation & data interpretation
In 2    To interpret the influence of various factors on the measured responses, coefficient plots will be generated. The coefficient plot displays the coefficients, when the selected response is changed from low to high value, with the confidence interval as error bars. The Table 4 shows that out of 7 variables, three of them were confounded, or aliased. showed that both acetone and DMSO have excellent miscibility with pressurized CO2, and scCO 2 can act as an anti-solvent for GF.

Production of GF particles with SAS
Before initiating co-precipitation experiments, GF crystals were obtained without any polymer, to observe the changes occurring in the drug morphology and particle size.
To define the working domain of SAS process, which was necessary to set up design of experiment; temperature (T) , pressure (P), solution flow rate (F liq ), CO 2 flow rate (F CO2 ), and concentration of GF in solvent(C d ), were tested in the preliminary experiment. Using acetone and DMSO we varied the operating parameters in the range of : T = 45-90°C, P = 80-300 bar, F liq = 1 -4 ml/min, F CO2 =10x to 30x F liq , and C d = 20% -80% of saturation solubility in organic solvent. Nozzle diameter (d n ) was kept constant at 100µm.
No precipitation was observed at drug concentration lower than 40% saturation solubility in organic solvent, or pressures greater than 200 bar, due to increasing solubility of GF in solvent-CO 2 mixtures. In case of acetone, 40% saturation solubility is 10 mg/ml, and in DMSO it is 24 mg/ml. Long fibrous and needle like product was obtained at C d ≥ 40% of saturation solubility in organic solvent , and 75 ≤ P < 200 bar. Table 5 summarizes the variables tested in preliminary experiments. Untreated GF particles ( as is GF) comprises of irregular shaped particles (Fig. 6) which have a mean particle size of 14 μ and size distribution of 5-21 μ, as measured by Malvern instrument.
The product obtained by supercritical processing has much larger needle shape, 500 µm to few mm in length ( Figure 6). The products obtained with acetone and DMSO were different in size, and uniformity. DMSO produced thicker, longer, non-uniform crystals (Figure 7-a), whereas acetone based GF crystals were thinner, shorter in length, and lot more uniform (Figure 7-b).

Polymer effect on particle properties of GF
Based on the selection criteria discussed in the introduction section, Kollidon® VA64 (KVa64) was selected as the polymer for coprecipitation. When drug (GF) and polymer (KVa64) were coprecipitated, the size of the crystals changed significantly. Some polymers like HPMC 36 , polyethylene glycol 36 , PVP 22 , Poly (sebacic anhydride) 23 , are known to inhibit particle growth of the drugs that are treated with, by adsorbing on the surface of drug particles. Surface adsorption of KVa64 on GF drug crystals would have prevented the nuclei coalescence, and hence smaller particles were produced. Additionally, many polymers can act like a stabilizer and minimize the particle agglomeration.
SAS processed material is analyzed for particle size, by dispersing it in an aqueous media. Malvern instrument detected bimodal distributions of particle size through out this study, regardless of the processing conditions. In Figure 8, an example of particle size distribution of the SAS coprecipitates, measured by Malvern instrument is shown. The appearance of the same material when investigated under SEM, is shown in Figure 9. While performing an SEM analysis, a dry powder obtained from SAS processing is used, and it is difficult to de-agglomerate . The size of intact drugpolymer mixture was approximately 5 to 10 µm.
The size of the products obtained provides some information on the location of the particle precipitation. The bimodal distribution suggest the occurrence of the precipitation at multiple times and locations along the length of the precipitation vessel. The particles that are formed as soon as they come out of the nozzle would be smaller as compared to particles that are formed at the bottom of the vessel. Similar bimodal distribution was reported by Jarmer et al., 23 when they precipitated griseofulvin in the presence of polymer Poly (sebacic anhydride), using a variation of SAS process called PCA (Particles from Compressed Antisolvent). Their particle sizes ranged from 0.5 µm to 100 µm, with one peak observed around 1 µm, and second peak observed around 30 µm. When Varughese et al., 37 precipitated indomethacin from dichloromethane using SAS technology; they also obtained bimodal particle size distribution, ranging from 0.1 µm to 100 µm.  The zeta potential values obtained ranged from -31.1 mV to -35.5 mV. However the zeta potential values were not used as one of the responses during the statistical analysis.

Statistical analysis of data from DOE
The raw data was then analyzed using Modde® statistical software. Figures 10 -13 provide the outcome of analysis in a graphical manner. * particle size, and zeta potential data could not be obtained since the yield were too low, + outlier

Summary of Fit Plot
Summary of fit plot, shown in Figure 10, displays the 4 key values for each response.
These values are; R Square (R 2 ) , Q Square (Q 2 ) , Model Validity, and Reproducibility. There are many causes of LOF which result in poor "Model Validity". Statistical outliers as well as non-normally distributed responses can cause LOF. However, when there is true LOF, both R 2 and Q 2 will be small. If there is a good R 2 and Q 2 (> 0.5) and a reproducibility value is close to 1.0, the lack of fit is probably artificial.
We ran a normal probability plot of residuals to find the statistical outliers. The normal probability plot of residuals displays the residuals (standardized or studentized) on a double log scale. This plot helps to detect statistical outliers and assess normality of the residuals. If the residuals are random and normally distributed, the normal probability plot of the residuals has all the points lying on a straight line between -4 and + 4 studentized or standardized standard deviation. When we plotted the normal probability plot of residuals, shown in Figure 11, we found that experiment number 6 was an outlier for particle size distribution. The outlier was not removed from the calculation. Finally, due to the reasons explained here, the low values for model validity in case of particle size distribution (0.1) , was not cause for concern.

Figure 11: Normal Probability Plot of Residuals
When the reproducibility bar is 1.0, the pure error is 0. This means that under the same conditions the values of the response are identical. If the reproducibility is below 0.5, it implies that there is a large pure error and poor control of the experimental set up (the noise level is high). The validity of such models cannot be assessed. This also results in poor R 2 and Q 2. . In our experiments, the values for reproducibility were almost 1.0, for d 50 , % yield, and PSD.

Identification of statistically significant variables of SAS process
The coefficient plot ( Figure 12) displays the statistically significant factors that are affecting d 50 (Figure 12-a), % yield ( Figure 12-b), and particle size distribution ( Figure 12-c). The data obtained is centered and scaled in this plot. The scaling of the data allows comparable coefficients. The size of the coefficient bar represents the degree of change in a response when a factor varies from low value to high value, while the other factors are kept at their averages. The coefficient is considered significant (different from the noise), when the 95% confidence intervals (shown as bars in Figure 11 a, b, and c) do not cross zero.

Figure 12: Coefficient Plot
For the particle size (d 50 ) response, three significant factors identified were; temperature, spray rate of liquid, and nozzle diameter. The nozzle diameter was the most important factor impacting the particle size as it is the largest coefficient for d 50 response. When nozzle diameter increased from 100 to 200 µm, the particle size (d 50 ) was also increased.
Smaller nozzle diameter produced finer droplet and more pronounced atomization of the liquid spray, leading to smaller particles. The spray rate of liquid had an opposite effect on d 50 , compared to nozzle diameter. Increased spraying rate can produce better atomization of the liquid spray, as more drug is available, super saturation will be reached quicker, leading to smaller particle sizes. Similarly, higher temperature increases the miscibility of acetone with scCO 2 ; at the same time higher temperature of 55 0 C may cause evaporation of acetone. Both of these phenomena ultimately result in quicker removal of acetone, and hence smaller particles are produced with increase in temperature. Additionally, higher temperature results in lower viscosity of polymer solution, which enhances mass transfer and more efficient removal of solvent, resulting in quicker precipitation and smaller particle size. The order of importance of the seven factors affecting d 50 can be summarized as follows ; For d 50 : d n (nozzle diameter) > T(Temperature) > F liq (spray rate of liquid) > P(pressure), C d (drug conc), C p (pol.conc), F CO2 (rate of CO2) In Figure 12-b, similar analysis is shown for % yield. It can be seen that the polymer concentration has highest influence on % yield. The increase in % of polymer in the formulation causes reduction in the yield. The cause may be the presence of excessive polymer which leads to precipitation in the nozzle, leading to low yield. Additionally, as stated earlier, higher polymer concentration causes increase in solution viscosity which leads to entanglement of polymer chains and delay in jet break up. This delay in jet break up makes solvent removal and mass transfer difficult, causing drug extraction instead of precipitation, causing low yield. Increase in the pressure also leads to a significant reduction in yield. This is understandable as increased pressure increases the solubility of drug in scCO 2 , leading to less precipitation of the drug.
Drug concentration has an opposite effect on % yield. The yield increases with increasing drug concentration. This is due to availability of more drug in the liquid droplets causing quicker super saturation which prevents extraction of drug from the precipitation vessel.
The order of importance of the factors affecting % Yield can be presented as follows ; For % Yield : C p (pol.conc) > P(pressure) > C d (drug conc) > T(Temperature), F liq (spray rate of liquid) , F CO2 (rate of CO2), d n (nozzle diameter) Finally, a similar analysis conducted for PSD shown in Figure 12-c, revealed that liquid spray rate had the most pronounced influence on PSD. Additionally, there were four more factors that also had significant impact on the particle size. These were pressure, drug concentration, polymer concentration, and the nozzle diameter.
The explanation of this finding is as follows: Consider that there are two opposing effects of increased liquid spray rate: reduction in particle size and increase in particle size distribution. When the flow rate of liquid is increased, the precipitation vessel contains larger quantity of solute (both drug and polymer). Increased amounts of solute may reduce anti-solvent effect of CO 2 , and growth as well as coalescence of nuclei may dominate the process which will cause broader particle size distribution.
Within the given operating conditions of our experiments, both of these phenomena are effective at any given time. These competing mechanisms may also explain the bi-modal distribution of the particles.
An increase in pressure, and drug concentration causing smaller PSD is explainable using the same arguments made in previous sections. An increase in polymer concentration provides more availability of polymer for adsorption on drug nuclei, and hence more uniform drug crystals are produced, leading to smaller PSD. An increase in the nozzle diameter creating narrower PSD was unexpected. It could be that increased nozzle diameter, uniformly produced significantly larger particles, and hence the PSD was narrow. The order of importance of the factors for PSD can be summarized as follows ; For PSD : F liq (spray rate of liquid) > d n (nozzle diameter) > C d (drug conc) > P(pressure) > C p (pol.conc) > T(Temperature), F CO2 (rate of CO2),

Optimization using Sweet Spot Analysis
Finally, based on the information gathered from the outcomes of statistical analysis, we carried out a Sweet Spot analysis which allowed us to predict the values for formulation and process variables, resulting in responses that we desire. We wanted to produce in-situ nano particles, hence we chose d 50 range of 0 to 900 nm. A desirable yield of 70 to 100% would be efficient for the process. Lastly, a narrow PSD (span value of 0 to 3.0) would allow more predictable dissolution behavior.
While performing the sweet spot analysis, drug concentration was kept constant at 20 mg/ml as that was important to obtain high yield. Polymer concentration was fixed at 20%, in order to formulate a product with minimum amount of polymer, and CO 2 addition rate was fixed at 40 g/min which was 20 times the spray rate of liquid.
It can be seen from Sweet Spot analysis shown in Figure 13, that, in order to achieve the responses within the desired range described above, we could choose different set of values for the variables.  Table 7, and they concur with predictions of sweet spot analysis.

CONCLUSIONS
A conventional SAS process yields GF crystals which are needle shaped, and several mm long. Using the coprecipitation approach in SAS processing, one can successfully produce in-situ nanoparticles of GF having d 50 value of approximately 0.4 µm. The fractional factorial design 2  was applied for screening of large number of variables, allowing identification of statistically significant factors all within 19 experimental run. The factors that impacted the particle size the most, were the nozzle diameter, temperature, and spray rate of liquid, in the order of decreasing importance.
In case of particle size distribution, nozzle diameter, spray rate of liquid, drug concentration, pressure, and polymer concentration played significant roles. Whereas, the yield was affected by polymer concentration, pressure, and the drug concentration. Additionally, we were able to find optimum processing and formulation variables, which would consistently deliver product of high yield, small particle size, and narrow particle size distribution.
Optimized SAS formulation of GF was crystalline in morphology, regardless of changing formulation and processing conditions. There was no evidence of any size reduction and generation of amorphous state using spray-drying process. The results also demonstrate that the crystalline nature of griseofulvin particles depends on the method of production.

INTRODUCTION
The enhancement of solubility and oral bioavailability of poorly water soluble drugs remain as the most challenging aspects of drug development. Amongst several approaches developed to improve the solubility of poorly soluble compounds, preparation of solid dispersion is a formulation strategy that is widely utilized by pharmaceutical scientists 1-2 . Solid dispersions contain one or more active ingredients in an inert carrier or matrix at solid state which can be prepared by melting, solvent or melting solvent method 2 . Many drug products commercially available in the world market 3 contain poorly soluble active pharmaceutical ingredients, and are safe and effective when they are used as solid dispersions for the solubility enhancement.
Besides solid dispersions, particle size reduction to micro and nano scale also appears as an effective and versatile option for solubility improvement [4][5][6][7] . Other approaches include formation of complexes 8 , chemical modification to pro drug or salt formation 9 , and lipid based drug delivery systems 10 .
Various methods are cited in the literature for preparation of solid dispersion 3 . All of these methods involve mixing of drug with a matrix, preferably at a molecular level.
These approaches can be broadly classified into two main categories; 1) Fusion method, 2) Solvent evaporation method. Fusion method, also called melt method, requires high processing temperature, at which many active pharmaceutical ingredients may undergo degradation. Hot melt extrusion is an example of fusion method. Solvent evaporation method although requires milder processing conditions, has a drawback of utilizing excessive organic solvents and difficulty in removing trace amounts of these toxic solvents left in the processed product. There are three sub-categories of solvent evaporation method: conventional solvent evaporation, supercritical fluid based technologies [11][12] , and spray drying [13][14] .
Solvent evaporation in its most simple and conventional form is carried out in a rotary evaporator under a reduced pressure and at an elevated temperature. The typical problem encountered in the conventional solvent evaporation process is the removal of the solvent from the mass of solids to an acceptable level quickly during the process.
This is because the mass becomes more and more viscous during the "drying", which prevents further evaporation of the residual solvent [15][16] . process air: 80 ± 2 kg/hr; atomization air flow: 6 kg/hr; and feed rate: 20 ± 2 g/min.
These two examples show that irrespective of the processing conditions, the shape of the SD material was spherical.
However, the spray drying process fails to produce high bulk density product. This is not desirable, as it would require further downstream processing for the densification of the material. Additionally, low product recovery and dust collection issues increases the cost of drying, and high initial investment is required compared to other types of Additionally, SAS is the only technique amongst SCF based technologies, that has been successfully applied at an industrial scale 26 .
Particle precipitation mechanisms are somewhat similar in SAS and in spray-drying.
In both methods, a solution is sprayed through a nozzle and is allowed to atomize under elevated temperature in a one-step process. The major difference is that SAS processing is carried out at an elevated pressure conditions, as opposed to an atmospheric condition used in spray drying. In an earlier study, we reported the various processing and formulation factors that affect the reduction of particle size of griseofulvin. The goal of this study is to compare and evaluate the physiochemical and dissolution properties of the products, produced by three methods mentioned here; SAS, spray drying, and conventional solvent evaporation. Griseofulvin

Materials
The model drug Griseofulvin

Supercritical antisolvent (SAS) Process
An SAS apparatus (model: SAS 50, Thar Technologies Co., USA) was used to generate GF-polymer co-precipitates. The schematic diagram of contained Eudragit® EPO as the polymer matrix.

Solvent Evaporation Method (SE)
As conventional for solvent evaporation method, the drug (20 mg/mL) and carrier polymer (

Preparation of the physical mixture
The physical mixture of griseofulvin with various polymers were prepared as a control, in 80:20 (drug:polymer) w/w ratio, by simple blending in a vial for 30 minutes.

Microscopy
A scanning electron microscope (SEM), model TM-1000 manufactured by Hitachi (Hitachi High-Technologies Europe GmbH, Germany) was used to examine the particle size and morphology. The magnifications were altered in order to get clear images. The samples were fixed by mutual conductive adhesive tape on aluminum stubs. In addition to SEM, an ultra-high resolution digital microscope , model VHX 600, manufactured by Keyence (KEYENCE America, Elmwood Park, NJ) was used to obtain 3-D images.

Laser Diffraction Particle Size Analysis
Particle size (PS) and particle size distribution (PSD) were measured using a

Equation 1
Sample measurements were done in two different dispersing media; n-hexane and phosphate buffer (pH 6.8) . Measurements done using n-hexane provides particle size of intact drug-polymer coprecipitates, as neither dissolves in n-hexane. Whereas, all polymers are completely soluble in phosphate buffer (pH 6.8), and hence that dispersing media provides measurement of in-situ particles of GF alone.
Prior to performing particle size measurements of processed samples, the accuracy and reproducibility of Malvern laser light diffraction method was challenged using polystyrene latex microspheres of known diameter. The choice of dispersing media did not affect the accuracy, or reproducibility of measurements. The details of these experiments can be found in manuscript II.

Powder X-Ray Diffraction (PXRD)
The morphological characteristics of the substance was determined using Powder X-Ray Diffraction (PXRD). PXRD was performed using Bruker D8 Advance Powder X-Ray Diffractometer (Brukler Corporation, Madison, WI). Samples were analyzed using a Cu (ƛ=1.54) K α radiation. The X-ray pattern was collected in the 2θ range of 1 to 40 0 in the step scan mode (scan speed 0.27 0 /sec and step size 0.0045°). PXRD depicts sharp peaks for crystalline substances and disappearance of these peaks indicates a transformation of a crystalline substance into an amorphous form.

Fourier Transformed Infrared Spectroscopy (FTIR)
FTIR spectra were collected on a Nicolet 6700 from Thermo scientific (Thermo Fisher Scientific Inc., Pittsburgh, PA) . Powders were measured directly using the smart orbit accessory. Spectra were collected from 400 -4000 cm-1 using 64 scans at a resolution of 4 cm-1. Spectra were analyzed using the Omnic software (v.7.2).

Thermal Analysis
Differential Scanning Calorimetry (DSC) was performed by using a DSC Q 2000 ® differential scanning calorimeter (TA Instruments, New Castle, Delaware).
Calibrations were performed prior to analysis using pure samples of indium and zinc. Melting point (T m ) and glass transition temperature (Tg) values were determined by the Pyris software.

Dissolution Rate and Intrinsic Dissolution Rate (IDR)
The dissolution rate and IDR of griseofulvin samples were measured in Distek® Dissolution Apparatus (Distek, North Brunswick, NJ) equipped with UD-lite® fiber optic measurement capability. Solid dispersion samples from SD, SE, and SAS processes were compressed into 100 mg tablets using a flat faced ¼" round tooling, under carver press. Each tablet contained equivalent of 7 mg of drug, polymer, and lactose was used as filler. The physical mixtures of drug and a polymer were also compressed into tablet having the same ratio of drug to polymer, and the filler.
Dissolution media was pH 6.8 phosphate buffer (0.05M) which was considered to be a simulated intestinal fluid (SIF). The dissolution analysis was performed in 500 ml SIF at UV wavelength of 295 nm using USP dissolution apparatus type II, at 37 0 C, 50 rpm.
In conventional dissolution studies using tablet or powder, factors such as rate of wetting, effect of particle size and hence specific surface area, disintegration, clumping etc., can affect the rate of dissolution. Whereas, intrinsic dissolution rate (IDR), using Wood's apparatus, provides a constant surface area, and permits a constant hydrodynamic system and in general avoids many of the problems associated with powders or tablet.
For Intrinsic dissolution studies, a Distek stationary disk system was used to prepare the compact pellets and perform the test. Approximately 200 mg of solid dispersion sample from different processing methods, was compressed with the aid of a benchtop Carver press (Carver, Inc., Wabash, IN, USA) at 4000 psi with a dwell time of 10 seconds to form a compact pellet of 0.5 cm 2 exposed surface area. Assemblies, each composed of the pellet, die, gasket, and a polypropylene plastic cap, were immersed with the pellet side up, into the bottom of flat-bottom dissolution vessels containing 500 mL of SIF at 37°C. The USP Apparatus II paddle was positioned 1 inch above the assembly and rotated at 50 rpm.

Microscopy
The microscopy pictures showed that the micronized drug purchased had irregularlyshaped, crystalline structures , and were approximately 5 -20 µm in dimensions ( Figure 3). The solid dispersion samples obtained by the SAS process, were fibrous, and needle-shaped, independent of the type of polymer used (Figure 4). The length of the intact needles was approximately 1 -10 µm, and the thickness was less than 1 µm.
In contrast, spray-dried micro-particles were spherical ( Figure 5) in shape for all three polymers used. Lastly, the morphology of crystalline particles obtained via conventional solvent evaporation process was plate like and irregular shaped ( Figure   6). As it was experienced earlier by many workers 17-18 who studied several drugs of different origin, and very different polymers, our drug, griseofulvin, and selected polymers (Kollidon® VA64, HPMCAS-LF, Eudragit® EPO) also formed spherical amorphous particles, when spray drying process was applied. The mechanism of particle formation during spray drying might explain this roundness and smoothness.
At first, there is formation of small or micro sized droplets at the end of nozzle. These droplets meet a stream of hot air and they lose their moisture very rapidly while still suspended in the drying air. The dry powder is then separated from the moist air by centrifugal action in a cyclone separator . This cyclonic movement of particles during the drying process prior to reaching the final collector is most likely giving the particles the spherical shape and smoothness [33][34] .
Crystalline, needle shaped morphology of GF particles processed via SAS technique has been observed by other researchers as well. Foster et al., 35 and Reverchon et al., 36 found that GF by itself (without any polymers) tend to precipitate out as crystalline, long needles of several hundred microns long when SAS process was used. In our study, we applied the coprecipitation approach. Even though the coprecipitates of GF and all three polymers were fibrous and needle shaped, the length of the needle became significantly shorter (d 50 : 2.3 to 3.7 µm), compared to GF precipitated alone (500 µm to 1mm through SAS process (Figure 7). Obviously, the presence of polymer used is affecting the crystallization of GF resulting in modified morphologies.
The reduced size may be the result of adsorption of the polymer onto fast growing drug crystal faces. Researchers like Jarmer et al., 37 also studied the precipitation of GF in presence of polymer Poly (sebacic anhydride) using a modified SAS process called PCA (particles from compressed antisolvent). They found that while the particle size of GF, when processed alone was around 500 µm, it reduced to 1-100 µm, when processed with Poly (sebacic anhydride).

Particle size (PS) and particle size distribution (PSD) measurement
PS and PSD of SAS, and spray dried formulations were measured by laser diffraction method, as obtained. Samples obtained with solvent evaporation method were collected in the form of films by scraping the walls of glass flask, and hence the samples were crushed using mortar and pestle prior to particle size measurement, in phosphate buffer (pH 6.8) or in n-hexane.
The particle size measurements of the SAS formulations for each polymer, using two different dispersing media, are summarized in Table 1, and Table 2. Using n-hexane as the dispersing media, volume weighted mean diameter (D (4,3) ) of all intact samples ranged between 2.9 µm to 5.0 µm, whereas d 50 value ranged between 2.3 µm to 3.7 µm . When the particle size measurement was done using phosphate buffer (pH 6.8) as the dispersion media, an in-situ nanoparticles of GF had volume weighted mean diameter (D (4,3) ) of 0.5 µm, and the d 50 value was approximately 0.4 µm. The particle size measurements of the spray dried formulations for each polymer, was also carried out in the same manner, and results are summarized in Tables 1 and 2.
Using n-hexane as the dispersing media, volume weighted mean diameter (D (4,3) ) of intact SD solid dispersions ranged between 2.5 µm to 15.9 µm, and the d 50 value for all intact SD formulations ranged from 1.9 µm to 12.9 µm. Once the polymer is dissolved from the spray dried amorphous solid dispersion, the drug crystallizes out in the form of nano particles. We were able to verify the crystalline nature of GF by observing the birefringence under polarized light microscopy ( Figure 8). As shown in Table 2, when phosphate buffer (pH 6.8) was used as the dispersion media, the volume weighted mean diameter (D (4,3) ) of in situ nano particles ranged between 0.6 µm to 0.9 µm, whereas the corresponding d 50 value for all spray dried formulations was approximately 0.3 µm. formulation of GF dispersed in phosphate buffer (pH 6.8) Lastly, the particle size measurements of the solvent evaporation method formulations, carried out in the same manner, are also reported in Table 1 and Table 2. The volume weighted mean diameter (D (4,3) ) of all intact samples ranged between 332.4 µm to 377.5 µm, whereas d 50 value ranged between 271.2 µm to 290.7 µm . As mentioned before, solvent evaporation method samples were collected in the form of film, and were gently crushed prior to sample measurement. Hence, these results which are largely different from SAS, and SD process, is explainable. When the particle size measurement was done using phosphate buffer (pH 6.8) as the dispersion media, the in-situ particles of GF had volume weighted mean diameter (D (4,3) ) of 24.7 to 82.2 µm, and the d 50 value was 18.3 to 28.1 µm. As opposed to SAS and SD process, which produced in-situ nanoparticles of GF, solvent evaporation process resulted in extremely large particles of GF in-situ.

Analysis of Variance (ANOVA)
ANOVA was performed on d 50 values to see if the type of polymer used (Kollidon® VA64, HPMCAS-LF, Eudragit® EPO), or the kind of process applied (SAS vs spray drying) had any significant impact on the particle size of intact drug polymer mixture.
Results as summarized in Table 3, show that F value of polymers, and that of process (SAS and SD) is larger than their F critical values, hence null hypothesis is rejected, and there is statistically significant difference in particle size amongst three different polymers, and between SD and SAS processes. The d 50 values from solvent evaporation method formulation were not used for ANOVA test, as it was evident without any doubt that, those particles were significantly larger than the SAS or SD process particles.
It is also evident from the Table 3 that amongst the three polymers tested, ionic polymers (HPMCAS-LF and Eudragit EPO), produced smaller particles in SAS and spray drying process, as compared to non-ionic polymer (Kollidon VA64). Ionic polymers provide low interfacial tension, and tend to produce smaller droplet upon exiting from the nozzle (in both SAS and spray drying process), and hence produced the smallest particles, compare to Kollidon VA64 which have no ionizable hydrophilic group.

PXRD
The micronized GF obtained from supplier, is crystalline with well-defined peaks in PXRD ( Figure 9). PXRD of GF solid dispersion samples prepared by the solvent evaporation and SAS methodology show that the material morphology remained crystalline. PXRD patterns of untreated GF matches that of all samples prepared using solvent evaporation and SAS method, independent of type of polymer used. In addition, PXRD patterns of the physical mixtures of GF and polymer samples were similar to that of untreated, GF, indicating the drug remained in crystalline form in the physical mixtures. In contrast, spray drying yielded an amorphous solid dispersion as was evident from the absence of peaks in the powder XRD scans.
The crystal structure is considered to be highly ordered structure, repeating itself in three dimensions. However, in practice, there are always imperfections in the crystal lattice such as point defects (e.g. vacancies, impurity defects etc), line defects (e.g edge dislocation) and plan defects (e.g grain boundaries) 38 . The level of such imperfections are likely to increase many fold during rapid drying process such as spray drying. Hence, spray drying process produced amorphous morphology.
During SAS processing, there was multicomponent system of drug, polymer and solvent. The presence of drug and polymer could have affected the solubility of acetone in scCO 2 . Reverchon et al., 36 hypothesized that when there is incomplete miscibility between the organic solvent and the scCO 2 , the particle formation takes longer duration, and particles are formed at the bottom of the vessel only when the organic phase reaches the super-saturation. This process takes longer duration (as compared to spray drying) and hence that may have allowed preferred packing of the molecules into its most stable form, the crystalline form. This is why our SAS process yielded crystalline morphology, and spray drying process produced amorphous material. Figure 9. PXRD of untreated as-is GF, SAS formulations, spray dried formulation, and solvent evaporation method formulations.

Thermal Analysis
DSC curves of micronized GF obtained from supplier, SAS coprecipitate, solvent evaporated solid dispersion, and spray-dried samples to corroborate the amorphicity and/or crystallinity of GF in the solid dispersion, are shown in Figure 10.

FTIR Analysis
FTIR is an effective technique in detecting presence of interaction in drug-carrier solid dispersions. The appearance or disappearance of peaks and/or the shift of their positions are often an indication of interactions such as hydrogen bonding. 39 To examine the possibility of hydrogen bond formation, an FTIR study was undertaken.
As shown in Figure 11(b), GF has two characteristic peaks; the first peak (1,704 cm −1 ) corresponds to the stretching of carbonyl group of the benzofuran, and the second peak (1,658 cm −1 ) corresponds to the stretching of the carbonyl group of cyclohexene.
These FTIR spectra of GF were in agreement with published work of Nair et al., 40 .
The FTIR results showed that there is a broadening of the GF carbonyl peak at 1,704 cm −1, and slight shift in 1,658 cm −1 peak, in the binary solid dispersion of spray dried formulations with all polymers (Figure 12). The broadened peaks indicated that the drug has formed hydrogen bonds, resulting in the shift of the peak. The broadening also refers to the distribution of free and bound carbonyl groups of GF. FTIR spectrum from SAS and solvent evaporation processes showed no discernible differences.

Dissolution Study
The drug dissolution behaviors of SAS formulations, spray-dried formulations, and solvent evaporation method formulations, were compared to untreated GF in physical mixture with a water soluble polymer (Kollidon VA64) . It can be seen in the Figures   13 and 14, that enhancement of GF dissolution rate was achieved in SAS and spray dried formulations. Dissolution curves of spray dried and SAS coprecipitates showed the steep initial slope and the dissolution rate was more than 8 times higher than raw drug after 100 minutes of dissolution.
In regards to the physicochemical properties studied in this work, similar mechanisms seemed to govern the dissolution of GF , independent of the type of polymer used in the solid dispersion. The dissolution profiles of products prepared by spray drying, and SAS enhanced the dissolution of GF to all most the same extent; whereas solvent evaporation method formulation did not improve the rate of dissolution of GF.
GF in spray dried solid dispersions was amorphous, however in SAS formulation it was crystalline. Prior to starting the dissolution, our expectation was that spray dried formulation would be better in rate of dissolution compared to SAS formulations. It was surprising to find that there was apparently no difference in rate of dissolution between SAS and spray dried formulations. This is because the improvement in dissolution was due to reduction in particle size. SAS formulation when exposed to phosphate buffer produces nano crystalline GF in situ. Similarly, amorphous GF solid dispersion from spray drying process undergoes fast re-crystallization when exposed to aqueous environment (as confirmed in polarized light microscopy), and leads to formation of nano crystalline material in situ, identical to SAS formulation. Thus, it is the small particle size (of SAS and spray dried processes) with large surface area, which facilitate the disintegration of the solid dispersions, and provides faster rate of dissolution. It was confirmed by Malvern laser light diffractometer, that particle size of in situ GF was identical in SAS and spray drying process.

Intrinsic dissolution rate (IDR)
Apparent IDR of formulations manufactured by three different methods, and three different polymers, were calculated by measuring the slope of concentration (µg/mL) vs time (min) profile; and are compared to that of physical mixture of GF with KollidonVA64. The IDR of pure GF in physical mixture with polymer is 0.0038 µg/cm 2 /min. Table 4 shows the IDR of various formulations prepared by SAS, spray drying, and solvent evaporation processing. The IDR of SAS formulations ranged between 0.0058 µg/cm 2 /min to 0.0065 µg/cm 2 /min, which was an improvement of 53% to 71%. The IDR of spray dried formulations ranged between 0.0063 µg/cm 2 /min to 0.0068 µg/cm 2 /min, which was an improvement of 66% to 79%.
Whereas, the IDR of solvent evaporation process formulations ranged between 0.0036 µg/cm 2 /min to 0.0039 µg/cm 2 /min, and showed almost no improvement. for their potential to produce rapidly dissolving dosage form of griseofulvin, a BCS class II API, with hydrophilic polymers. The properties of the product, and key processing attributes are summarized in Table 5. Table 5. Comparison of SAS, spray drying, and solvent evaporation methods, for preparation of GF coprecipitates It was shown that the properties of GF-polymer coprecipitates are influenced by the processing methods. Overall, SAS is the most desirable process, as it had reasonably high product yield. The product obtained was crystalline, and was subjected to the lowest temperature during processing. Due to crystalline morphology and exposure to milder processing conditions, it is expected to have the least stability complications.
On the whole, it is a promising approach to produce crystalline solid dispersions in a few processing steps. However, the complexity of the process must be weighed against its benefits.
From downstream dosage form processing perspective, spray dried products being spherical in shape, would be desirable compounds as they would have good powder flow characteristics. The amorphous GF was molecularly dispersed in the spray dried solid dispersions of each polymer through hydrogen bonding. However, due to low polymer content (20%), the drug re-crystallizes rapidly in an aqueous environment.
Therefore, it is assumed that the drug would undergo rapid recrystallization during storage. In spite of several prediction methods available to measure reversion to crystallinity, it is still a difficult task to determine accurately and to control. Hence, we conclude that the spray drying process is not desirable for this formulation.
We expect almost the same cost of operation for SAS and SD processes. Conventional solvent evaporation process was the least desirable one even though it was the least expensive, as it did not improve the rate of dissolution of GF.   30 50 Three phase system of acetone, liquid CO 2 , and vapors of CO 2 . Acetone level rises due to CO 2 being absorbed into it.
25 ml DMSO + CO 2 30 50 Three phase system of DMSO, liquid CO 2 , and vapors of CO 2 . DMSO level rises due to CO 2 being absorbed into it.

Verification of in-solubility of GF in acetone-CO 2 & DMSO-CO 2 system
At the beginning of the experiment, 100ml of acetone solution containing GF (25 mg/mL) was placed in the extraction vessel. Agitator was turned on to gently mix the acetone solution. When looked through the view cell, a clear acetone solution is seen, and agitator can be seen. After that, the CO 2 pump was turned on to fill the extraction vessel. Temperature was gradually raised from 35 0 C to 60 0 C, and pressure was gradually raised from 0 to 150 bar, and observations were made through the view cell.
The solubility or insolubility of GF in CO 2 + acetone was judged by the visual appearance of cloudiness. We found that at 40 0 C and 100 bar, the vessel was extremely cloudy, and agitator could not be seen, providing evidence that scCO 2 acts as an anti-solvent for GF. Experiments were then repeated by using DMSO as the organic solvent, at a GF concentration of 60 mg/ml. The results are summarized in the Gioannis et al., 40 conducted quantitative solubility determination of GF in presence of acetone. The solubility measurements for the GF-CO 2 -acetone system were performed at 39 0 C at 60 and 100 bar, and at 53 0 C and 100 bar. They found that there was dramatic reduction in solubility of GF in acetone with increase in mole fraction of CO 2 . At a CO 2 mole fraction of 0.9, pressure between 60 -100 bar, and temperature between 39 0 C to 53 0 C, the solubility of GF in the binary system was approximately 0.0005 mol/mol . Vessel extremely cloudy, indicating drug is almost insoluble at this condition. Agitator bar not visible due to extreme cloudiness.
Raw GF is crystalline in nature with well-defined peaks. PXRD of GF coprecipitates after SAS processing shows that the material morphology remained crystalline, independent of changes in processing and/or formulation variable. As shown in Figure   E.2, the PXRD patterns of the optimized formulation produced from SAS were super imposable to the spectra of drug from the supplier. These findings were consistent with the results obtained by DSC.
In our study, there was multicomponent system comprising of drug, polymer and organic solvent. The presence of drug and polymer may have shifted the phase equilibrium and affected the solubility of acetone in scCO2, and hence the efficiency of SAS process to remove the solvent from the feed is reduced. This long duration likely allowed preferred packing of the molecules into its most stable form, the crystalline form.   The zeta potential value is an important particle characteristic as it can influence both particle stability as well as particle mucoadhesion. The electrostatic repulsion between particles with the same electric charge prevents the aggregation of the spheres 41 .
Hence, more pronounced zeta potential values either positive or negative, can stabilize particle suspension. When SAS solid dispersion of drug and polymer is added to water, the polymer dissolves, leaving behind a suspension of drug particles. The zeta potential measurements were done to understand the characteristics of drug suspension. A value lower than 30mV indicates that there is aren't enough charges on the particles to keep them in a stable, non-aggregated state. On the other hand, a large zeta potential value (>30mV) could support our argument that during particle size measurement we are measuring individual particles. The zeta potential values for our formulations (N1 to N19, and OP1) ranged from -31.1mV to -35.5mV, as shown in Table 6. BET surface area measurements were done for untreated GF from supplier and for optimized formulation of SAS coprecipitates (OP1). BET surface area value of 5.2457 m 2 /g for SAS formulation was not significantly different from a value of 5.2095 m 2 /g for untreated Griseofulvin. However, it should be noted that untreated GF is a micronized material. Secondly, SAS coprecipitates of drug and polymer together are not that much different in particle size compared to untreated GF. It is only in the in-situ conditions when the polymer is removed, the drug particle size is nano sized. Even though the BET surface area of SAS drug is similar to untreated drug from supplier, the SAS drug is expected to have better solubility and rate of dissolution.

E.5 Dissolution and intrinsic dissolution rate
Dissolution studies were performed by USP Dissolution Apparatus Type II , paddle method using Distek® dissolution apparatus (Distek, Inc., North Brunswick, NJ). The apparatus was equipped with UD-lite® fiber optic measurement capability. Solid Samples obtained with SAS processing were compressed into 100 mg tablets using a flat faced ¼" round tooling, under carver press. Each SAS formulation tablet