Understanding the effects of formulation and process variables on the printlets quality manufactured by selective laser sintering 3D printing

International Journal of Pharmaceutics 570 (2019) 118651

Contents lists available at ScienceDirect

International Journal of Pharmaceutics journal homepage: www.elsevier.com/locate/ijpharm

Understanding the effects of formulation and process variables on the printlets quality manufactured by selective laser sintering 3D printing

T

Sogra F. Barakh Alia, Eman M. Mohameda,b, Tanil Ozkanc, Mathew A. Kuttolamadomd, ⁎ Mansoor A. Khana, Amir Asadid, Ziyaur Rahmana, a

Irma Lerma Rangel College of Pharmacy, Texas A&M Health Science Center, Texas A&M University, College Station, TX 77843, USA Department of Pharmaceutics, Faculty of Pharmacy, Beni-Suef University, Beni-Suef 62514, Egypt c Dover Precision Components, Woodlands, TX, USA d Dept. of Engineering Technology & Industrial Distribution, College of Engineering, Texas A&M University, College Station, TX 77843, USA b

A R T I C LE I N FO

A B S T R A C T

Keywords: 3D printing Selective laser sintering Box-Behnken Diclofenac sodium Disintegration time Dissolution

The focus of the study was to understand the effects of formulation and process variables on the printlets quality manufactured by selective laser sintering (SLS) 3D printing. The Box-Behnken response surface methodology was used to evaluate effects of individual variables and combinations thereof. The formulation and process variables studied were printing chamber temperature (°C, X1), laser scanning speed (mm/sec, X2) and lactose monohydrate concentration (%, X3). The responses studied were weight of printlets (mg, Y1), hardness (N, Y2), disintegration time (sec, Y3) and dissolved drug fraction in 15 min (%, Y4). The values of Y1, Y2, Y3 and Y4 varied from 170.2–257.0 mg, 5.5–32.0 N, 20–120 s and 64.4–97.5%, respectively. The studied factors showed statistically significant effects on the dependent variables (p < 0.04). The correlation coefficient between empirical and model predicted values for Y1, Y2, Y3 and Y4 were 0.999, 0.992, 0.998 and 0.983, respectively. The model was validated by an independent experiment and actual values of the responses were in close agreement with model predicted values. Fourier transformed infrared spectroscopy indicated no chemical interactions between the components of the formulation during printing process. X-ray powder diffractograms suggested a decrease in crystallinity of the drug and lactose in the printlets. Chemical images indicated uniform distribution of the drug. Scanning electron microscopy and X-ray micro-CT scanning showed a very porous microstructure of the printlets with a porosity of about 37.89%. In conclusion, the SLS method of manufacturing provides a feasible and flexible avenue for fabricating dosage forms with tailored characteristics.

1. Introduction Three-dimensional printing (3DP) is a group of manufacturing processes where a solid part is created layer-by-layer in which each technology variant has different process capabilities and resolutions, the material types it can handle, and final part characteristics. The 3DP methods reported in literature for pharmaceutical applications are fused deposition modeling (FDM), binder jetting, selective laser sintering (SLS), stereolithography (SLA), etc (Rahman et al., 2018; Rahman et al., 2019). Some of the 3DP process aspects are analogous to the pharmaceutical processes. For example, FDM and binder jetting, which resemble hot-melt extrusion and wet granulation processes, respectively. FDM is the most commonly reported method for pharmaceuticals printing. This method requires a filament form of drug and excipients, which is extruded via a hot end nozzle (Nukala et al., 2019;

⁎

Algahtani et al., 2018). SLA requires polymerizable polymer, miscibility of the drug in the monomer solution, and post-printing UV curing (Martinez et al., 2018; Wang et al., 2016). Binder jetting requires very low viscosity liquid binder due to the limitation of print-head, postprinting drying to remove the solvent, and separation of printed dosage forms from the powder bed (Tian et al., 2019; Vithani et al., 2018). The primary requirement of SLS is laser sintering or melting of one or more of the components of the powder formulation, where the laser (usually, a carbon dioxide laser) binds the powder particles together due to fusion/sintering. Compared to FDM, binder jetting and SLA, SLS does not require significant post-processing, polymerizable excipients or a filament form of formulation, and it is a faster, solvent-free printing process while being easier to clean as only the reservoir and printing platform are in contact with the powder. During the SLS printing process, the powder mixture is spread as a thin layer over the building

Corresponding author at: 310 Reynolds Medical Sciences Building, College Station, TX 77843-1114, USA. E-mail address: [email protected] (Z. Rahman).

https://doi.org/10.1016/j.ijpharm.2019.118651 Received 28 May 2019; Received in revised form 23 August 2019; Accepted 27 August 2019 Available online 04 September 2019 0378-5173/ © 2019 Elsevier B.V. All rights reserved.

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

platform with the help of a roller, followed by laser scanning of a selected region of the powder layer. Once the layer is formed, the platform of the SLS printer lowers (usually by less than 0.1 mm depending upon the printer capability), and a new powder layer is spread for the laser to scan and fuse together. This process repeats until the entire object has been printed (Awad et al., 2019; Fina et al., 2018; Shirazi et al., 2015). Most of studies on process and formulation factors affecting the quality of printed dosage forms are reported for FDM processes (Nukala et al., 2019; Algahtani et al., 2018). SLS is commonly used in metal printing (Oyar 2018; Mizoshiri et al., 2018; Revilla-León et al., 2019). Fina et al. (2017) was first to report application of SLS for pharmaceutical application in 2017. However, information on SLS regarding its utility in pharmaceuticals printing is very limited (Awad et al., 2019; Fina et al., 2017; Fina et al., 2018a,b). It is important to understand the interplay of process parameters such as powder bed temperature, laser scanning speed, laser power, etc., and formulation factors such as particle size, density, melting point, etc. on the critical quality attributes of printed dosage forms. The interplay of process and formulation factors can be understood by utilizing design of experiments statistical methods. Design of experiments is one of the quality-by-design approaches utilized in formulation development. This could provide a thorough understanding of process and formulations factors and their relationship with quality attributes of the printlets (Rahman et al., 2014; Krishnaiah et al., 2014). In the current research, diclofenac sodium (DFS) was selected as a model drug, which was fabricated as printlets using SLS. The formulation and process variables affecting the quality of the printlets were understood using the response surface methodology.

Table 1 Variables and experimental matrix. Independent variables

Temperature (X1, °C) Laser scanning speed (X2, m/sec) (mm/sec) Lactose monohydrate concentration (X3, %) Formulation F1 F2 F3 F4 F5 *F6 F7 *F8 *F9 F10 F11 F12 F13 F14 F15

Level Low

Medium

High

115 270 8 X1 (°C) 115 125 125 125 135 125 135 125 125 115 115 135 115 135 125

125 300 10 X2 (mm/sec) 300 330 330 270 330 300 270 300 300 270 300 300 330 300 270

135 330 12 X3 (%) 8 8 12 8 10 10 10 10 10 10 12 12 10 8 12

* - Formulation representing center point of the design.

energy absorption from the laser and aid printability (Fina et al., 2018). Cylindrical printlet models of 10 mm diameter and 3 mm height was designed using Solidworks (Solidworks Inc., USA). The design templates of 3D models were transferred as a stereolithography (.stl) file. The printlets were printed using Sintratec central software version 1.1.13. Powder reservoir (150 × 150 × 150 mm) of the Desktop SLS printer (Sintratec Kit, AG, Brugg, Switzerland) was filled with the blend and moved to the printing build area (150 × 150 × 150 mm) in 0.5 mm thickness by a roller creating a uniform layer. The chamber temperatures varied from 115 to 135 °C during printing as per the design. Layered powder was sintered by a 2.3 W blue diode laser (445 nm) by scanning at 270–330 mm/sec. The process continued as the reservoir moved upward and the printing platform moved downward thus sintering the powder layer by layer into the final shape of the tablet based on the pattern from the .stl file. Upon completion of the printing process, the printer was allowed to cool down, the printlets were removed from the powder bed, and the excess powder was brushed away.

2. Materials and methods 2.1. Materials DFS (Leap Chem, Hangzhou, China), Kollidon® VA 64 (BASF, Germany), Candurin® NXT Ruby Red (Merck, Darmstadt, Germany), lactose monohydrate (LMH, SuperTab® 14SD, DFE Pharma, Paramus, NJ), acetonitrile (ACN), potassium hydroxide and monobasic potassium phosphate (Fisher Scientific, Asheville, NC) were used as received. 2.2. Methods

2.2.3. Dimension, hardness and disintegration time The diameter and thickness of the printlets were measured using a digital Vernier caliper. The hardness of printlets was measured using tablet hardness tester (VK 200, Varian Inc, Cary, NC). Disintegration tests were conducted using USP disintegration (900 ml water medium, Vankel Varian VK-100, NC, USA) and by a petri-dish method; for the latter, in a petri dish containing 20 ml of water at 37 ± 0.5 °C, the printlet was gently placed on the surface of the water and the time for the printlet to completely disintegrate was observed (six printlets for each condition was evaluated).

2.2.1. Experimental design Initial trials were performed to select a range of independent variables. A three-factor, three-level Box-Behnken design was used for optimization of process and formulation variables using JMP 14 software (SAS, Cary, NC, USA). Independent variables studied were surface temperature (X1, °C), laser scanning speed (X2, mm/sec) and LMH concentration (X3, %), and responses measured were weight of printlets (Y1, mg), hardness (Y2, N), disintegration time (DT) (Y3, sec) and percent drug dissolved in 15 min (Y4, %). Fifteen formulations with high, medium and low levels of the selected independent variables were performed as given in Table 1. The design of experiments also utilized the center point of the design in triplicate to demonstrate reproducibility of the data. The statistical analysis of the model is represented in the form of analysis of variance (ANOVA). Optimal process and formulation variables on the dissolution and DT were obtained using a tool response optimizer. An additional experiment was conducted to validate the model.

2.2.4. Dissolution Dissolution of the printlets were performed using USP apparatus 2 (Model 708-DS with 850-DS autosampler, Agilent Technologies, CA, USA). The dissolution was performed in 900 ml of 0.2 M phosphate buffer pH 6.8 at 50 rpm and 37 °C. 1 ml samples were collected at 2, 5, 10, 15, 30, 45 and 60 min, and 20 µL of sample was injected into HPLC system to determine amount of dissolved drug. Dissolution was performed in triplicate. HPLC system (HP 1260 series, Agilent Technologies, Wilmington, DE, US) equipped with a quaternary pump, online degasser, column heater, autosampler and UV/Vis detector. HPLC method for DFS was developed and validated as per ICH guidelines (ICH, 2005). Separation of the drug was achieved on a 4.6 × 150 mm, 5 µm Luna C18 (Phenomenex, Torrence, CA, USA)

2.2.2. Printing process All the formulations were prepared as per Table 1. 200 g of a mixture of drug and excipients (DFS 30%, LMH 8–12% and Kollidon® VA 64 55-59%) were blended using a V-Blender (Model VH-2). 3% of Candurin® NXT Ruby Red was added to the formulations to enhance 2

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

VA 64 (vinylpyrrolidone-vinyl acetate copolymer) (Kollidon® VA 64, 2019) was performed at laser scanning speeds of 250–350 mm/sec. Printlets prepared at low laser scanning speed showed higher weights than those printed at higher laser scanning speed. At low laser scanning speed, that powder particles are exposed to the laser for a longer duration and thus absorbs higher energy during the sintering process, thus lowering the void volume by more/complete melting and fusion of the polymeric component of the powder mixture as compared to the particles sintered at higher laser speeds. This result concurred with previous findings (Fina et al., 2017; Fina et al., 2018a,b). The printlets disintegration time (DT) was < 5 min with high friability/brittleness characteristics. Nevertheless, just by increasing or decreasing the laser scanning speed, it was not feasible to increase drug loading, improve mechanical properties, and decrease DT of the printlets. Further experiments were carried out by adding LMH to the formulation. Addition of LMH to formulation helped in printing mechanically improved printlets, shorten DT (< 2 min) and increase drug loading to 30%. Candurin® NXT (food and pharmaceuticals pigment, silicon dioxide coated with ferric oxide) percentage was also changed in the formulation. Experiments were performed at 1, 2, 3 and 3.5% Candurin® NXT. Low level (1 and 2%) resulted in poorly formed/sintered printlets. On the other hand, a high level of Candurin® NXT produced burnt printlets. These observations can be explained that at a low level, Candurin® NXT amount was not enough in the formulation to absorb the laser energy, and thus led to insufficient melting/sintering. Moreover, high energy was absorbed at the high Candurin® NXT level than required to cause sintering. The printing process was successful at 3% w/w Candurin® NXT with excellent sintering which was suitable for printing the printlets with orally disintegrating release/ immediate release characteristics (Fina et al., 2017; Fina et al., 2018a,b). Chamber/surface temperature effect was studied on the printability of dosage forms. The melting point of Kollidon® VA 64, LMH and DFS is 140, 149.20 and 290 °C, respectively (Kollidon® VA 64. Safety data sheet; α-Monohydrate Phase in Lactose by DSC; Pasquali et al., 2007). Printlets did not form successfully if the chamber temperature was below 115 °C. Additionally, the solidified matrix of powder-mixture was formed in the reservoir if the chamber temperature was more than 135 °C.

column and a C18, 4.6 × 2.5 mm (5 µm packing) Luna C18 guard column (Phenomenex, Torrance, CA, USA). Mobile phase was ACN:phosphate buffer pH 7.0 (30:70, v/v) flowing at 1.0 ml/min. 20 µL sample volume was injected and detected at a wavelength of 280 nm. Data collection and analysis were performed using OpenLab software (Agilent Technologies, Wilmington, DE, US). 2.2.5. Vibration spectroscopies Fourier transform infrared (FTIR) spectra of samples were collected using a modular Nicolet TM iSTM 50 system (Thermo Fisher Scientific, Austin, TX). Data collection parameters for FTIR were: absorbance mode, wavenumber range 400–4000 cm−1, data resolution 8 cm−1 and 100 scans. OMNIC software, version 9.0 (Thermo Fisher Scientific, Austin, TX) was used to capture and analyze the spectra. NIR (near infrared) chemical images of printlets were collected using Via-Spec II Hyperspectral Imaging System. The images were collected from 900 to 2500 nm with SWIR hyperspectral camera (MRC-303–005-02, Middleton Spectral Vision, Middleton, WI). The data acquisition software used was Middleton Spectral Vision (Middleton Spectral Vision, Middleton, WI) and the data analysis software was Prediktera EvinceTM (Prediktera AB, Umea, Sweden) (Barakh Ali et al., 2019). 2.2.6. X-ray powder diffraction XRPD patterns of pure DFS, excipients, blend and the printlets were collected using Bruker D2 Phaser SSD 160 Diffractometer (Bruker AXS, Madison, WI) equipped with the LYNXEYE scintillation detector and Cu Kα radiation (λ = 1.54184 Å) at a voltage of 30 KV and a current of 10 mA. The samples were prepared by evenly spreading the appropriate amount of powder on the sample holder. The mounted samples were scanned over 2θ range of 5–40° at 1 s per step with an increment of 0.0202° and rotated at 15 rpm to get the average diffractogram. The collected data was evaluated using Diffrac.EVA Suite version V4.2.1 and further processed using File Exchange 5.0 (Bruker AXS, Madison, WI). 2.2.7. Scanning electron microscopy Surface and cross-sectional morphology of the printlets were determined by scanning electron microscopy (SEM, JSM-7500F, JEOL, Tokyo, Japan). Samples were coated approximately 5 nm thick with carbon using a sputter coater (Cressington, 208 HR with MTM-20 High Resolution Thickness Controller) under high vacuum (argon gas pressure 0.01 mbar) and high voltage of 40 mV. Morphology was captured at a working distance of 15 mm, an accelerated voltage of 5 KV and an emission current of 20 µÅ.

3.2. Weight (Y1) Based on the experimental design generated, the factor combinations resulted in different responses. The positive and negative signs in the model equation for a dependent variables indicate an increase and a decrease in a dependent variable with an increase in an independent variable value (Chopra et al., 2007; Rahman et al., 2010). Significance of a parameter is determined at p < 0.05. Weight of printlets varied from 170.20 ± 2.2 mg (F11) to 257.0 ± 4.9 mg (F7) for the various factor combinations (Table 1). The following polynomial model can explain the effects of various factors on the weight of printlets (Y1):

2.2.8. X-ray micro computed tomography A high-resolution X-ray micro computed tomography scanner (SkyScan1172, Bruker-microCT, Belgium) was used to spatially visualize the internal structure, and measure the density and porosity of the printlets. The printlet was scanned with a resolution of 2000 × 1048 pixels. 3D imaging was performed by rotating the object through 180° with steps of 0.4° and 4 images were recorded. Image reconstruction was performed using NRecon software (version 1.7.0.4, BrukermicroCT). 3D model rendering and viewing were performed using the associate program CT-Volume (CTVol version 2.3.2.0) software. The collected data was analyzed using the software CT Analyzer (CTan version 1.16.4.1). Different colors were used to indicate the density of the printlets. Closed and open porosity values were calculated using the 3D analysis in the morphometry preview (200 layers were selected at the central part of the printlet as area of interest and analyzed).

Y1 = 185.5 + 22.4X1 − 10.3X2 − 2.0X3 − 9.9X1X2 + 7.5X1X3 + 13.9X2 X3 + 16.5X12 + 10.9X22 + 1.0X32 Except X32 interaction term, all studied factors and interaction terms have statistically significant (p < 0.04) effect on the Y1. Correlation coefficient (r) between empirical and model predicted values was close to unity (0.999) (Fig. 1A). Determination coefficient (R2) value was 0.997 which means 99.7% variation in Y1 can be explained by studied factors. Furthermore, root mean squared error (RMSE-1.96) and residual (−2.19 to 2.19) were low indicating low error in the model (Fig. 1B). Based on coefficient of factors in the model equation, the most important factor affecting Y1 (printlet weight) was chamber temperature. Chamber temperature has positive while laser scanning speed and LMH have negative effect on Y1, respectively. The printlet weight increased with an increase in chamber temperature as well. This was

3. Results and discussion 3.1. Preliminary printability Preliminary printing of formulations containing DFS and Kollidon® 3

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

Fig. 1. A) Empirical and model predicted values of weight of the printlets (Y1 in mg) and B) residual of Y1.

Fig. 2. Response surface profiler showing effects of chamber temperature (X1), laser scanning speed (X2) and lactose monohydrate concentration (X3) on A) printlet weight (Y1 in mg), B) hardness (Y2 in N), C) disintegration time (Y3 in sec) and D) drug dissolved in 15 min (Y4 in %).

printlets formulations F5 and F7 showed the effect of laser scanning speed. The weight of printlets was 214.3 ± 3 and 257.0 ± 4.9 mg, respectively. LMH also produced bulkier printlets and thus reduced value of the Y1. This was related to melting point of LMH and Kollidon® VA 64. Melting point of Kollidon® VA 6 and LMH are 140 and 149.20, respectively (Kollidon® VA 64, 2012; α-Monohydrate Phase in Lactose

related to the sintering/melting process. The extent of sintering/ melting was higher at high chamber temperature which produced dense printlets. Formulations F1 (191.5 ± 1.8 mg) and F14 (221.0 ± 1.0 mg) showed greater effect of chamber temperature on the response Y1. Similarly, sintering/melting was less effective at high laser scanning speed, which produced bulkier/less dense printlets. The 4

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

Fig. 3. A) Empirical and model predicted values of hardness (Y2 in N) and B) residual of Y2.

also depends upon weight of printlets (Fig. 2B). Positive correlation coefficient of 0.80 was observed between weight and hardness of the printlets.

by DSC, 2019). The printlets containing high LMH concentration required high energy (slow laser scanning speed and high chamber temperature) to sinter/melt the powder and produce dense printlets. Printlets F1 (191.5 ± 1.8 mg) and F11 (170.2 ± 2.2 mg) showed the effect of LMH concentration on the Y1 (Fig. 2A).

3.4. Disintegration time (Y3) 3.3. Hardness (Y2)

Disintegration time (DT) varied from 20 ± 1.7 (F13) to 120 ± 5.0 s (F7) depending on the various factor combinations selected for printing. The effect of independent variables on DT can be described by following equation:

The hardness value varied from 5.5 ± 0.3 (F11) to 32.0 ± 2.8 N (F14) for the selected independent factor combinations. The following equation elucidating the effects of various factors on Y2:

Y3 = 29.6 + 27.3X1 − 22.5X2 − 5.6X3 − 3.75X1X2 − 4.0X1X3 + 2.75X2

Y2 = 6.4 + 8X1 − 5.12X2 − 4.25X3 − 2.75X1X2 − 5.0X1X3 + 2.0X2 X3 +

4.55X12

+

4.8X22

+

X3 + 24.6X12 + 11.9X22 + 15.6X32

2.8X32

All the factors (X1, X2, X3) show a statistically significant (p < 0.002) effect on the hardness of printlets. A correlation coefficient of 0.992 was obtained between empirical and model predicted values and residual values varies from −2.12 to 2.12 (Fig. 3). The model can explain 98.4% variation in data as indicated by value of R2 of 0.984. Model has low error as indicated by RMSE value of 1.96. Chamber temperature has a positive influence over Y2. This could be explained by sintering/melting phenomenon. Laser sintering at a high temperature increased the degree of melting/sintered compared to one sintered at a low temperature. Effect of chamber temperature on hardness of printlets can be noticeably seen in the formulations F5 and F7 where laser scanning speed and LMH concentration were kept constant. Sintering at 115 °C produced printlets of 6.5 ± 0.1 N hardness (F1) while sintering at 135 °C produced printlets of 32.0 ± 2.8 N hardness. Laser scanning speed has a generally negative effect on the hardness. At higher laser scanning speed, contact time between laser and powder spot was short which accounted for lesser degree of sintering or melting resulting in poor mechanical strength of printlets. The opposite was true for printlet hardness that were sintered at slow laser scanning speed. The effect of laser scanning speed was clearly observed in F5 and F7 printlets which were sintered at 330 and 270 mm/sec, respectively, while keeping other parameters constant. The hardness of F5 and F7 were 17.0 ± 0.3 and 31.0 ± 1.3 N, respectively. LMH concentration in the formulations has a negative effect on the hardness. However, its effect on hardness was not as significant as X1 and X2 variables (as shown by associated coefficients in the model equation). Mechanical strength was primarily imparted by Kollidon® VA 64. This property can be manipulated by increasing or decreasing the fraction of polymer. The effect of LMH on Y2 was noticeable in F4 and F15. These formulations contained 8 and 12% LMH, respectively. The hardness of F4 and F15 are 25 ± 2.2 N and 15.0 ± 0.9 N, respectively. Hardness

Statistically significant (p < 0.002) effect of independent variables was observed on Y3. The value of ‘r’ was 0.998 between actual and model predicted value of Y3 (Fig. 4A). Model can explain 99.6% variation in the model. RMSE and residual values were 3.21 and −3.12 to 3.13, respectively which indicated higher error in the model of the Y3 compared to other studied responses (Fig. 4B). The Y3 increased with an increase in chamber temperature and decreased with an increase in laser scanning speed and LMH concentration. DT was related to weight of printlets and hardness. The correlation coefficient between Y1 and Y3 and Y2 and Y3 were 0.865 and 0.911, respectively. Denser and mechanical stronger printlets were associated with high DT. Printing at high chamber temperature produced denser and stronger printlets and longer DT. Formulations F1 (45 ± 3 s) and F14 (111 ± 8.9 s) showed effect of chamber temperature on DT. Mechanically weak and bulkier printlets were produced at high laser scanning speed and LMH concentration. Effect of laser scanning speed was represented by printlets F5 (65 ± 4.6 s) and F7 ((120 ± 5.0 s) which were printed at 330 and 270 mm/sec speed, respectively. Similarly, printlets F4 (85 ± 5.0 s) and F15 (72 ± 1.7 s) showed effect of LMH on the Y3 (Fig. 2C) which contained 8 and 12% LMH, respectively. Weight and hardness of the printlets have positive effect of the DT as indicated by ‘r’ of 0.865 and 0.911 between Y1 and Y3, and Y2 and Y3, respectively.

3.5. Dissolution (Y4) The dissolution was more than 75% in 15 min from all the printlet formulations except F4 and F14 (Fig. 5). The dissolution in 15 min (Y4) varied from 64.4 ± 3.16 (F14) to 97.5 ± 1.5% (F6 and F9), respectively (Fig. 5). The Y4 response can be expressed by following polynomial equation: 5

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

Fig. 4. A) Empirical and model predicted values of disintegration time (Y3 in sec) and B) residual of Y3.

and residual values of −2.01 to 2.01 (Fig. 6B). Dissolution was inversely related to weight, hardness and DT. This was indicated by negative correlation coefficient value of −0.539, −0.697 and −0.732 between Y1 and Y4, Y2 and Y4, and Y3 and Y4, respectively. The chamber temperature and laser scanning speed have positive and negative effect on Y4, respectively. These effects are related to Y1, Y2 and Y3. Higher dissolution was observed when printlet weight, hardness and DT were low, and vice-versa. F1 (86.4 ± 2.9%) and F14 (64.4 ± 3.2%) showed

Y4 = 97.6 − 5.38X1 + 2.9X2 + 0.07X3 + 0.45X1X2 + 6.13X1X3 − 4.82X2 X3 −

7.3X12

− 1.65X2 −

13.3X32

Chamber temperature and laser scanning speed have statistically significant (p < 0.02) effect on Y4. The ‘r’ value was 0.983 between actual and model predicted values that was lower than other studied responses (Y1, Y2 and Y3) (Fig. 6A). Variation in Y4 explained by the studied independent variables was 96.6%. The model has RMSE of 2.1

Fig. 5. Dissolution profile of printlet formulations. 6

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

Fig. 6. A) Empirical and model predicted values of drug dissolved in 15 min (Y4 in %) and B) residual of Y4.

at 3447 cm−1, CeH asymmetrical/symmetrical vibrations at 2951 cm−1 and sharp intense peaks of C]C stretch at 1730 cm−1, 1660 cm−1, 1232 cm−1. Additive spectra were obtained for physical mixture (PM) of components encompassing peaks of all the major components (DFS and Kollidon® VA 64). LMH and Candurin® NXT did not show characteristic peaks due to their low concentration in the formulation compared to other components and interference with other components of the formulation. The printlets showed the spectrum similar to PM indicating no chemical interactions during printing process (Fig. 7A). The X-ray diffractogram of DFS exhibited characteristic reflection peaks at 6.62°, 8.51°, 11.19°, 12.25°, 15.17°, 23.46°, 27.07° and 27.89° that indicated crystalline nature of the drug (Fig. 8). LMH displayed characteristic peaks at 16.46° with a doublet and triplet peaks at 20.90° and 19.21°, respectively. Kollidon® VA exhibited a halo diffractogram, characteristics of an amorphous material. Physical mixture of

the effect of chamber temperature on Y4. The effect of laser scanning speed was shown by formulations F5 (85.5 ± 3.8%) and F7 (79.9 ± 4.5%). The printlets F4 (65.7 ± 3.8%) and F15 (92.5 ± 3.4%) represented the effect of LMH on the Y4 (Fig. 2D). 3.6. Analysis of variance analysis Analysis of variance (ANOVA) at 95% confidence level was used to test significance of the model. The model was considered statistically significant at 0.05 probability level, which is represented by “Prob > F”. “Prob > F” for weight of printlets and DT was 0.0001, while it was 0.0006 for hardness and dissolution, implying that it was highly significant for the studied responses, and that independent variables significantly affect the responses. Error in the ANOVA is measured by F-ratio. A larger F-ratio value means that differences between the means are significant and are due to non-random effects, inferring smaller errors in the model. The independent responses could be ranked in increasing order of significance and error in the model based on “Prob > F” and F-ratio values as follows: weight of printlets > DT > dissolution > hardness. 3.7. Model optimization Having examined the effect of independent variables on the responses, the levels of significance of these factors were determined by validating the model using a response optimizer tool and desirability function. Experiments were conducted using theoretically optimized process and formulation variables. The responses were compared with the predicted values in order to validate the model. At the optimized conditions of 124.2 °C chamber temperature, 302.4 mm/sec laser scanning speed with LM concentration of 10%, the experimental value of dissolution was 99.9%, which was very close to the model predicted value. The residual between two values was 1.5%. 3.8. Physical characterization of printlets FTIR spectra of DFS showed asymmetrical and symmetrical stretching vibration bands at 1573 cm−1 and1450 cm−1 due to carboxylate group, respectively. A vibration peak due to secondary amino group also appeared at 3385 cm−1 (Shivakumar et al., 2008; Ramachandran and Ramukutty, 2014). LMH presented a distinct OH stretching vibration at 3521 cm−1. The shape and location of this peak indicated the presence of water molecules within the crystal lattice. It also exhibited the characteristic doublet peaks at 1069 and 1030 cm−1 corresponding to skeletal vibrations of CeC stretch. Kollidon® VA 64 showed a characteristic broad spectrum of OeH stretch, hydroxy group

Fig. 7. FTIR spectra of LM, DFS, Kollidon® VA 64, physical mixture and the printlet. 7

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

Fig. 11. X-ray micro-CT image of the optimized printlets.

Fig. 8. X-ray powder diffractograms of LM, DFS, Kollidon® VA 64, physical mixture and the printlet.

crystalline DFS and LMH into amorphous form. This can be explained by sintering/melting followed by dissolution phenomena. As the laser impinges on the surface of the powder blend, Kollidon® and possibly LMH can melt, thus partially dissolving the drug into the molten matrix. Partial least squares (PLS) was used as a multivariate method for NIR-CI data analysis. A library was generated containing components of the formulations of the printlet. PLS concentration images of the printlets were generated using DFS component. Red and blue pixel in the PLS concentration image indicates low and high concentration of the drug. The images of the printlets were uniform in color, which indicated a uniform distribution of the drug in the printlets. Uniformity was further indicated by low values of kurtosis and skewness (Fig. 9). Similarly, Kollidon® VA 64 showed similar images (not shown) when PLS images was created using Kollidon® VA 64 component of the library. SEM images provided morphological properties of the printlets. The sintering of individual powder particles can be clearly seen in the SEM images (Fig. 10A and B). As the range of laser scanning speed and the temperatures selected were not very extreme (Fina et al., 2017; Fina et al., 2018a,b), a drastic variation in the porosity of particles is not observed between formulations. However, the printlets fabricated at highest laser scanning speed (F13, Fig. 10A) and lowest temperature showed particles were loosely packed and more porous than the

Fig. 9. PLS concentration images of the printlets showing distribution of the drug.

formulation components and printlets showed characteristics peaks of DFS and LMH, which indicated that DFS and LMH retained their crystalline nature. However, there were reduction in intensities of some of the peaks of DFS and LMH which indicated partial conversion of

Fig. 10. SEM images of the printlets. 8

International Journal of Pharmaceutics 570 (2019) 118651

S.F. Barakh Ali, et al.

printlets fabricated at slow laser scanning speed and highest temperature (F7, Fig. 10B), respectively. X-ray micro-CT was performed for the optimized formulation Fopt (model validation formulation) to calculate open and closed porosity (Fig. 11). Open porosity is defined by the existence of void spaces inside the printlet which is in direct contact with the exterior environment whereas it is reverse in the case of closed porosity. When the printlets comes in contact with the dissolution medium, the open pores tend to dissolve first and faster than closed pores. The open and closed porosity of Fopt was 0.11% and 37.78%, respectively, which was similar to previously reported values (Fina et al., 2018).

Fina, F., Goyanes, A., Madla, C.M., Awad, A., Trenfield, S.J., Kuek, J.M., Patel, P., Gaisford, S., Basit, A.W., 2018a. 3D printing of drug-loaded gyroid lattices using selective laser sintering. Int. J. Pharm. 547 (1–2), 44–52. Fina, F., Madla, C.M., Goyanes, A., Zhang, J., Gaisford, S., Basit, A.W., 2018b. Fabricating 3D printed orally disintegrating printlets using selective laser sintering. Int. J. Pharm. 541 (1–2), 101–107. ICH guideline-Validation of analytical procedures: Text and methodology Q2(R1), 2005. Krishnaiah, Y.S., Xu, X., Rahman, Z., Yang, Y., Katragadda, U., Lionberger, R., Peters, J.R., Uhl, K., Khan, M.A., 2014. Development of performance matrix for generic product equivalence of acyclovir topical creams. Int. J. Pharm. 475 (1–2), 110–122. Kollidon® VA 64. Safety data sheet, BASF 2012. Accessed on May 15, 2019. https://www. nwmissouri.edu/naturalsciences/sds/k/Kollidon%20VA%2064.pdf. Kollidon® VA 64. Technical information. Accessed on May 15, 2019. http://www. rumapel.com.ar/pharma_excipientes/ficha_tecnica/Kollidon%20VA%2064.pdf. Martinez, P.R., Goyanes, A., Basit, A.W., Gaisford, S., 2018. Influence of geometry on the drug release profiles of stereolithographic (SLA) 3D-printed tablets. AAPS PharmSciTech. 19 (8), 3355–3361. Mizoshiri, M., Nishitani, K., Hata, S., 2018. Effect of heat accumulation on femtosecond laser reductive sintering of mixed CuO/NiO nanoparticles. Micromachines (Basel) 9 (6), E264. α-Monohydrate phase in lactose by DSC, TA293. Accessed on May 15, 2019. http://www. tainstruments.com/pdf/literature/TA293.pdf. Nukala, P.K., Palekar, S., Patki, M., Patel, K., 2019. Abuse deterrent immediate release egg-shaped tablet (egglets) using 3D printing technology: quality by design to optimize drug release and extraction. AAPS PharmSciTech. 20 (2), 80. Oyar, P., 2018. Laser sintering technology and balling phenomenon. Photomed. Laser Surg. 36 (2), 72–77. Pasquali, I., Bettini, R., Giordano, F., 2007. Thermal behaviour of diclofenac, diclofenac sodium and sodium bicarbonate compositions. J. Therm. Anal. Calorim. 90 (3), 903–907. Rahman, Z., Barakh Ali, S.F., Ozkan, T., Charoo, N.A., Reddy, I.K., Khan, M.A., 2018. Additive manufacturing with 3D printing: progress from bench to bedside. AAPS J. 20 (6). Rahman, Z., Charoo, N.A., Kuttolamadom, M., Asadi, A., Khan, M.A., 2019. Printing of personalized medication using binder jetting 3D printer. In: Faintuch, J., Faintuch, S. (Eds.), Precision Medicine for Investigators, Practitioners and Providers. Elsevier Science (in press). Rahman, Z., Xu, X., Katragadda, U., Krishnaiah, Y.S., Yu, L., Khan, M.A., 2014. Quality by design approach for understanding the critical quality attributes of cyclosporine ophthalmic emulsion. Mol. Pharm. 11 (3), 787–991. Rahman, Z., Zidan, A.S., Habib, M.J., Khan, M.A., 2010. Understanding the quality of protein loaded PLGA nanoparticles variability by Plackett-Burman design. Int. J. Pharm. 389 (1–2), 186–194. Ramachandran, E., Ramukutty, S., 2014. Growth, morphology, spectral and thermal studies of gel grown diclofenac acid crystals. J. Cryst. Growth 389, 78–82. Revilla-León, M., Meyer, M.J., Özcan, M., 2019. literature review of current status and prosthodontic applications. Int. J. Comput. Dent. 22 (1), 55–67. Shirazi, S.F.S., Gharehkhani, S., Mehrali, M., Yarmand, H., Metselaar, H.S.C., Kadri, N.A., Osman, N.A.A., 2015. A review on powder-based additive manufacturing for tissue engineering: selective laser sintering and inkjet 3D printing. Sci. Technol. Adv. Mater. 16 (3), 033502. Shivakumar, H.N., Desai, B.G., Deshmukh, G., 2008. Design and optimization of diclofenac sodium controlled release solid dispersions by response surface methodology. Indian J. Pharm. Sci. 70 (1), 22–30. Tian, P., Yang, F., Yu, L.P., Lin, M.M., Lin, W., Lin, Q.F., Lv, Z.F., Huang, S.Y., Chen, Y.Z., 2019. Applications of excipients in the field of 3D printed pharmaceuticals. Drug Dev. Ind. Pharm. 11, 1–9. Vithani, K., Goyanes, A., Jannin, V., Basit, A.W., Gaisford, S., Boyd, B.J., 2018. An overview of 3D printing technologies for soft materials and potential opportunities for lipid-based drug delivery systems. Pharm. Res. 36 (1), 4. Wang, J., Goyanes, A., Gaisford, S., Basit, A.W., 2016. Stereolithographic (SLA) 3D printing of oral modified-release dosage forms. Int. J. Pharm. 503 (1–2), 207–212.

4. Conclusions This study elucidated the interplay of formulation and process variables on the quality of the printlets. Lactose monohydrate, laser scanning speed and chamber temperature had statistically significant (p < 0.05) effects on the weight of the printlets, disintegration time, hardness and dissolution. The printlets with desired characteristics (good mechanical integrity and high disintegration/dissolution rates) can be printed according to the data of the optimized formulation. Vibration spectra and diffractograms indicated no chemical interactions between the components and maintenance of crystalline nature of the drug. SEM and X-ray micro-CT indicated a porous internal structure of printlets. There was no segregation/demixing of the components during the printing process as suggested by NIR-CI images. These results suggest that SLS 3D printing can provide viable method of printing drug for personalized medications in a pharmacy or hospital settings. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References Algahtani, M.S., Mohammed, A.A., Ahmad, J., 2018. Extrusion-based 3D printing for pharmaceuticals: contemporary research and applications. Curr. Pharm. Des. 24 (42), 4991–5008. Awad, A., Fina, F., Trenfield, S.J., Patel, P., Goyanes, A., Gaisford, S., Basit, A.W., 2019. 3D printed pellets (miniprintlets): a novel, multi-drug, controlled release platform technology. Pharmaceutics 11 (4). Barakh Ali, S.F., Rahman, Z., Dharani, S., Afrooz, H., Khan, M.A., 2019. Chemometric models for quantification of carbamazepine anhydrous and dihydrate forms in the formulation. J. Pharm. Sci. 108 (3), 1211–1219. Chopra, S., Patil, G.V., Motwani, S.K., 2007. Release modulating hydrophilic matrix systems of losartan potassium: optimisation of formulation using statistical experimental design. Eur. J. Pharm. Biopharm. 66, 73–82. Fina, F., Goyanes, A., Gaisford, S., Basit, A.W., 2017. Selective laser sintering (SLS) 3D printing of medicines. Int. J. Pharm. 529 (1–2), 285–293.

9