Materials

Celecoxib (98.0 ~ 102.0% purity) was purchased from Shandong Zhishang Chem Co., Ltd (Shandong Province, China). Hydroxypropyl methylcellulose (HPMC) of grades 4, 100, 4000, 15000, and 100000 were donated by Shin-Etsu (Shin-Etsu, Japan), while hydroxypropyl methylcellulose acetate succinate (HPMCAS) was kindly offered by Seppic (Seppic SA, France). Soluplus ® was acquired from BASF (Rheinland-Pfalz, Germany), polyvinyl pyrrolidone K30 (PVP K30) was purchased from Fluka (Missouri, USA), polyvinyl alcohol (PVA) and polyethylene glycol 400 (PEG 400) were purchased from Sigma-Aldrich (Missouri, USA). Water was ultrapurified using a Milli-Q water apparatus (Millipore®, USA) and filtered through a 0.22 μm nylon filter before use. All other reagents and solvents were from analytical or high-performance liquid chromatography grade.

MethodsQuality target product profile and initial risk assessment

The Quality Target Product Profile (QTPP) is a fundamental element of the QbD approach, wherein the essential attributes a drug product should attain are prospectively summarized, envisioning the quality features intended to be reached, with a focus on ensuring the safety and efficacy of the drug product [18, 19]. A risk estimation matrix (REM) was constructed to identify and prioritize potential high-risk process parameters that could influence the critical quality attributes.

High-throughput screening of polymers

The study aimed to streamline high throughput screening assays involving five distinct polymers, including various grades and celecoxib (CXB), at a concentration of 1% (w/V) in sample volumes of 10 mL. The drug and polymers were dissolved in different solvents. The preferred solvent was ethanol due to its ability to solubilize CXB and the polymers. A mixture of ethanol and dichloromethane (2:1) was used when the polymers were not soluble in ethanol, which was the case for all HPMC and PVA molecular weights. The solutions were dispersed into the 24-well plates (n = 3), evaporating the solvent. Subsequent to solvent evaporation, the solutions were incubated at 37 °C for 2h with a dissolution media composed of 1.5 mL of HCl 0.01 N + PEG 400 (70:30, % V/V) to represent the gastric fluid pH in the fasted state. The solubilization capacity of the carriers for CXB was determined using a high-performance liquid chromatography (HPLC) method previously validated according to the FDA and ICH recommendations [20].

3k full-factorial design

A three-level full factorial design with two variables, 32, was used to study the effect of formulation variables on critical quality attributes (CQAs) of solid dispersions. Two factors were considered, each at low, intermediate, and high levels, numerically expressed as -1, 0, and +1, respectively. A central point is included to understand the model curvature in the response function and investigate a quadratic relationship between the response and each factor [21]. The optimal conditions were selected based on the quadratic polynomial function represented in

$$\beginY\,&=\,_\,+\, __\,+\, __\,+\, ___\,\\&+\, __}^\,+\, __}^\end$$

(1)

where Y is the measured response, \(_\) is the response in the absence of effects, \(_\) and \(_\) are the coefficients of the respective independent variables, \(_\) is the interaction coefficient between the two factors, and \(_\) and \(_\) are the quadratic coefficients obtained from the observed experimental values of Y, \(_\) and \(_\) are the terms representing the two factors and \(_\) and \(_\) represent the quadratic terms [20, 22].

The two independent variables, displayed in Table 1, were the total solid content and the ratio of Drug:Polymer 1:Polymer 2. As responses, particle size, polydispersity index, and assay were considered. The experimental design was solved using the JMP Statistical Analysis Software, JMP Pro version 17.0.0. Both the Student t-test and ANOVA were conducted to assess the significance of the regression model terms and evaluate the model fit.

Table 1 Experimental design independent variables and respective codificationAnalytical centrifugation

The LUMiSizer equipment (LUM GmbH, Berlin, Germany) allowed a predictive assessment of the formulations’ physical stability behavior. This multisample analytical centrifuge enables the measurement of the intensity of transmitted near-infrared light as a function of time and position through the entire sample extent. The transmission profiles provide information relative to the kinetics of separation processes, allowing sedimentation and/or creaming rates to be determined. The instability index, a quantitative parameter, also acknowledges the evaluation of formulation stability and ranges from 0 to 1. A value of 0 suggests formulation stability with no changes under test conditions, whereas a value of 1 indicates greater instability of the formulations, revealing different instability phenomena represented in the transmission profiles [23, 24]. The stability of the formulations was analyzed after 1h30 of centrifugation at an acceleration of 4000 rpm and 25 °C. The determination of stability parameters was performed using the SEPView® software.

Preparation of dispersions by antisolvent co-precipitation

As represented in Table 2, three different formulations were prepared using the antisolvent precipitation technique. When required, according to the formulation composition, the solvent (S) organic phase containing CXB and HPMCAS-HG was prepared by dissolving the components in ethanol due to their insolubility in water. The remaining polymers were dissolved in ultrapurified water, constituting the antisolvent (AS) aqueous phase.

Table 2 Pre-formulation compositionsHigh-shear homogenization

Dispersions by antisolvent precipitation were obtained by introducing the antisolvent aqueous phase in the solvent organic phase directly under a high-speed stirrer (Ultra-Turrax X1020; Ystral GmbH, Dottingen, Germany) at 11000 rpm for 15 min. Subsequently, the suspension formulation was diluted with ultrapurified water to analyze particle size and polydispersity index.

High-pressure homogenization

After being submitted to high-shear homogenization, the dispersions were further processed in a high-pressure homogenizer (HPH) (Emulsiflex®-C3, Avestin, Inc., Ottawa, Canada) at 25ºC and a pressure of 1000 bar for 15 min. The obtained formulation was subsequently diluted with ultrapurified water to analyze particle size and polydispersity index.

Spray drying

Amorphous solid dispersions obtained by spray drying were manufactured using a Mini Spray Drier B-290 (BÜCHI Labortechnik AG), equipped with a two-fluid nozzle (∅ 0.7 mm, stainless steel). Temperature conditions of the spray drier were an inlet temperature of 140 °C and an outlet temperature of 45 ± 5 °C [25]. The peristaltic pump was set at 30% (approximately 9 mL/min), and the aspirator rate was set at 100% (approximately 35 m3/h). Automatic nozzle cleaning was used throughout the whole experiment.

Microfluidics-on-a-chip

Microfluidics-on-a-chip was here considered as an alternative technique. Briefly, the organic and aqueous phases described in Table 2 were respectively pumped into two reservoirs connected to the inlet channels of a pressure-driven flow control system Elveflow OB1 MK3+ (Elvesys, Paris, France) at various flow rates to obtain S:AS ratios of 1:1, 1:2, 1:4, 1:8 and 1:10. The rapid mixing of solvent and antisolvent streams occurred in a Y-junction herringbone microfluidics microchip (200 μm depth, 188 μm thickness). The obtained dispersions were subsequently diluted with ultrapurified water to analyze particle size and polydispersity index. Before and after each formulation development, pure solvents were run through the equipment to clean the system and prevent contamination thoroughly. The pressure was systematically monitored to guarantee the same conditions for every formulation.

Analytical methods for characterization of amorphous solid dispersions

Characterizing the properties of ASDs is crucial in all stages of product development. This allows us to understand molecular interactions and rationally select the formulation composition and processing methods, obtaining high-quality products [26].

Particle size analysis

Evaluation of particle size is of utmost importance since it has been demonstrated that it is inversely proportional to the dissolution rate [27]. Dynamic light scattering and laser diffraction were considered depending on the sample size range.

Dynamic light scattering

Dynamic light scattering (DLS) is a well-established technique for determining the size of particles in the submicron (< 1µm) region. Alongside particle size, the polydispersity index is a parameter that should be considered when defining the particle size distribution. The formulations’ particle size was determined using a Zetasizer Nano ZS, Malvern (Malvern, Worcestershire, UK) set at a detection angle of 173° and a temperature of 25 °C. All samples were diluted 100 times in ultrapurified water and analyzed in triplicate.

Laser diffraction

The laser diffraction measures volume-weighted particle size distributions, and the parameters are reported based upon percentiles, of which Dv10, Dv50, and Dv90 are the most commonly portrayed. The span value is an additional parameter that indicates the size distribution width. It is calculated according to

$$\text\,=\, \frac_ \,-\,\text_}_}$$

(2)

Laser diffraction was performed using the MasterSizer 3000 equipped with the Aero S accessory (Malvern, Worcestershire, UK). The sample was placed inside the Aero S, and the equipment performed background evaluation and system alignment. An obscuration range of 0.1 – 10% was used. The real and imaginary refractive indices were set to 1.61 and 0.01, respectively. Six measurements were carried out for each sample.

Differential scanning calorimetry

Differential scanning calorimetry (DSC) was performed in a DSC-204 F1 Phoenix differential scanning calorimeter (Netzsch, Germany) to assess the thermal properties of pure compounds and amorphous solid dispersions. Samples of 2 mg were weighted and placed in aluminum pans that were hermetically sealed. Each sample was heated at a heating rate of 10 °C/min from 25 to 200 °C, with a nitrogen purge of 20 mL/min (Table 3).

Table 3 Melting temperature of active pharmaceutical ingredient and glass transition temperatures of the polymersAttenuated total reflectance Fourier-transform infrared

Attenuated total reflectance infrared spectroscopy (ATR-FTIR) spectra were obtained using a FT-IR/FT-NIR spectrometer (Spectrum 400, PerkinElmer, MA, USA) equipped with a crystal diamond ATR sampling accessory. Pure compounds and samples were placed on top of the ATR device and measured using 32 scans for each spectrum, recorded in the 4000–650 cm-1 range with a 2 cm-1 resolution.

X-ray powder diffraction

X-ray powder diffraction (XRPD) diffractograms were obtained using a Rigaku MiniFlex 600 diffractometer (Rigaku, Japan) to assess the amorphization state of the samples and crystallinity of pure compounds. Samples were placed in zero background holders, and spin was used to avoid preferential orientation. In all measurements, Cu Kα radiation (λ = 1.541862 Å) was used.

Scanning electron microscopy

Optimized formulations were analyzed by scanning electron microscopy (SEM). A JSM 6010LV/6010LA, Jeol (Tokyo, Japan) was used to determine the morphology of the samples. Prior to analysis, the powder was spread on a double-sided carbon tape mounted onto an aluminum stud. The analysis was conducted at acceleration voltages of 3kV.

In vitro dissolution testsIntrinsic dissolution test

The intrinsic dissolution rate (IDR) is stated as the dissolution rate of a pure active compound following compaction under the condition of constant surface area [28]. The equivalent of 50 mg of CXB in each sample was accurately weighed. The powder mixtures were compressed using a hydraulic press (Speca Press., UK) at 1 ton for 1 min. Pure CXB compacts were used for reference. After compression, the discs were assembled in the shafts of the dissolution apparatus. The dissolution medium used was 250 mL of pharmacopoeial phosphate buffer solution (PBS) at pH 6.8 with 30% (V/V) PEG 400, at a temperature of 37 ± 1 °C and the intrinsic dissolution test was conducted for 5 h at a rotation speed of 250 rpm. Sink conditions were maintained throughout the experiment. 1 mL of samples were collected at predefined time intervals (0.25, 0.5, 0.75, 1, 1.5, 2, 2.5, 3, 3.5, 4, and 5 h) from the dissolution vessels and replenished with the same volume of fresh medium. All samples were filtered through 0.22 µm pore-sized membrane filters and analyzed by HPLC (see Section “High-performance liquid chromatography determination of celecoxib”). To calculate the IDR, the cumulative amount of drug per unit area of the compact must be plotted against time. Linear regression is then performed, and the intrinsic dissolution rate is determined from the slope of the regression line, expressed in terms of the dissolved mass of a substance per time per exposed area [29, 30]. If the graph obtained presents curvature, only the initial linear region must be considered for the calculation of this parameter [31]. Enhancement ratios (ER) were also considered and calculated according to

$$\text\,=\,}_}/}_}$$

(3)

Dissolution studies

In vitro dissolution testing simulates and predicts the in vivo performance of oral dosage forms in the gastrointestinal tract [32]. Tablets containing the test product were prepared by combining the ASD powder, lactose, and magnesium stearate and were obtained by direct compression in a hydraulic press (Speca Press., UK) at 5 tons, presenting a total weight of 342 mg (10 mg CXB). Tablets containing pure API, lactose, and magnesium stearate were prepared in the same conditions as a reference, weighing ~369 mg (equivalent to 10 mg of CXB). Each excipient was previously sieved (180 µm).

Drug release behavior of amorphous solid dispersions in tablets containing 10 mg of CXB was performed using USP Apparatus II (paddle apparatus) [33] with a rotation speed of 75 rpm. The dissolution medium used was 250 mL of pharmacopoeial phosphate buffer solution at pH 6.8 with 30% (V/V) PEG 400 at a temperature of 37 ± 1 °C to simulate the conditions of intestinal fluid without enzymes. The test was conducted for 8 h, and sink conditions were maintained throughout the whole experiment. 1 mL of samples were collected at predefined time intervals (0.25, 0.5, 0.75, 1, 1.5, 2, 3, 4, 6, and 8 h) from the dissolution vessels and replenished with the same volume of fresh medium. All samples were filtered through 0.22 µm pore-sized membrane filters and analyzed by HPLC (see “High-performance liquid chromatography determination of celecoxib” section). Dissolution profiles were achieved by plotting the percentage of drug release against time, calculated according to

$$\mathrm\mathrm\,\times\,\text$$

(4)

Ex vivo permeability studies in Ussing chambersAnimals and tissue preparation

Adult RNU rats aged between 10–12 weeks and weighing 200–250 g were sourced from Charles River Laboratories (Lyon, France). They were housed in local animal facilities under controlled environmental conditions, including a 12-h light/dark cycle, a temperature maintained at 20 ± 2 °C, and a relative humidity of 50 ± 5%. All animal experiments were conducted in strict compliance with international regulations of the European Union (European Directive 2010/63/EU regarding the protection of laboratory animals used for scientific purposes) and were in accordance with the Portuguese law on animal welfare (Decreto-Lei 113/2013). The experimental protocols were thoroughly reviewed and approved by the Portuguese National Authority for Animal Health, Phytosanitation, and Food Safety (DGAV—Direção-Geral de Alimentação e Veterinária, Lisbon, Portugal), with the project reference "0421/000/000/2023". Effort was made to minimize both the number of animals utilized and any potential suffering they might endure.

Rat intestinal tissue was chosen as it has been widely utilized in intestinal drug permeability studies [34,35,36]. Additionally, mouse models have also been employed for similar purposes [37]. Fasted rats were sacrificed by decapitation, and the entire small intestine was rapidly removed and flushed with a PBS solution of pH 7.4. To minimize the tissue damage during the preparation, the intestine was allowed to cool for approximately 10 min. Afterward, jejunum segments with approximately 2 cm were cut along the mesenteric border and carefully placed in vertical Ussing chambers (Harvard Apparatus Inc., Holliston, MA, U.S.A.) with 0.64 cm2 of exposed area. Each side of the tissue was bathed with 1.5 mL of PBS solution with a pH of 7.4 and pH 6.8 for the acceptor (serosal membrane) and donor (mucosal membrane) compartments, respectively. Thereafter, chambers were screwed tightly and kept at 37 °C for the entirety of the experiment.

Ex vivo permeation experiment

Ex vivo permeation experiments were conducted following a 20 min equilibration period. 1.5 mL of drug solution composed of PBS pH 6.8 at a concentration of 10 mg/mL was inserted in the donor compartment. The serosal membrane was kept with PBS pH 7.4 with PEG 400, 70:30 (% V/V). Samples (200 µL) were collected from the acceptor compartment at 0.25, 0.5, 1, 2, and 3 h and analyzed by HPLC (see Section “High-performance liquid chromatography determination of celecoxib”). After each sample collection, 200 µL of fresh PBS 7.4 with PEG 400 medium, maintained at 37ºC, was replaced in the acceptor compartment.

To evaluate tissue integrity during the study, a stock solution (37.5 mg/mL) of fluorescein disodium salt in PBS pH 6.8 was previously prepared, and 20 µL was added in each donor compartment to obtain a final concentration of 0.05% (w/V) in the chamber. Samples of 100 µL were collected from the acceptor compartment to a microplate at the time points previously mentioned, and the same volume of fresh medium was replaced. The collected samples were analyzed by fluorometry (λexcitation: 485 nm; λemission: 515 nm) in a Synergy HT Multi-Mode Microplate Reader (BioTeK, UK). Standard calibration curves were prepared for fluorescein to proceed to its quantification. The apparent permeability (Papp) of CXB was calculated as

where Q (µmol) is the total amount of drug that permeated to the receiver compartment throughout the incubation time, C (µmol/mL) is the initial drug concentration in the donor compartment, A (cm2) is the diffusion area of the Ussing chamber, and t (s) is the duration of the experiment.

Membrane viability studies

To assess the viability of cells within the jejunum membrane following the permeability assay, the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) method was employed. After the permeability assay, rat intestinal membranes were incubated with 500 μL of MTT solution (5 mg/mL in PBS at pH 7.4) at 37 °C under stirring for 3h. As a control, rat intestinal membranes in PBS underwent a parallel incubation during the same period. Post-incubation, MTT was removed, and the tissue was rinsed in acidified isopropanol (1 μL concentrated hydrochloric acid per 1 mL isopropanol) to solubilize formazan crystals. Formazan absorbance was then measured at 570 nm and 620 nm using a Synergy HT Multi-Mode Microplate Reader (BioTeK, UK). Cell viability (%) was subsequently calculated and compared to the untreated control.

High-performance liquid chromatography determination of celecoxib

The quantification of CXB was performed using a previously validated HPLC method [20]. A Shimadzu LC-2010HT apparatus equipped with a quaternary pump (LC-20AD), an auto-sampler unit (SIL-20AHT), a CTO-10AS oven, and a SPD-M2OA detector was used. The analytical column used was Kinetex® EVO C18, with a 5 mm particle size, 4.6 mm internal diameter, and 150 mm length, operated at an oven temperature of 35ºC. The mobile phase was composed of a mixture of 2% (V/V) glacial acetic acid:acetonitrile (50:50, % V/V) and was eluted at an isocratic flow rate of 1.2 mL/min. A run time of 10 min was established, and CXB was eluted at 7.5 min. The UV-Vis detection was carried out at 250 nm, and an injection volume of 10 μL was used for all samples. The results were further processed using the Shimadzu LC-solution version 1.12 software.

Statistical analysis

Statistical significance was evaluated using Student's t-test. A value of p < 0.05 was considered significant. This analysis was performed with GraphPad Prism PRISM 8.3.0 (GraphPad Software, San Diego, CA, USA).