Effect of Process and Formulation Variables on Quality, Stability, and Taste of Lamivudine Printlets Manufactured by Binder Jetting 3D Printing Method

Abstract

This study investigated effect of formulation and process variables on the quality attributes of lamivudine printlets manufactured by binder jetting 3D printing method. The variables studied through fractional design of experiment were number of solvent sprays, delay time, powder layer thickness, drying temperature and hydroxypropyl methyl percentage. The printlets were assessed for hardness, disintegration time (DT), and dissolution and chemical information by X-ray powder diffraction (XRPD) and Fourier transformed spectroscopy, near infrared hyperspectroscopy, and taste evaluation by electronic tongue. The printlet hardness and DT ranged from 2.5 to 21.1 N, and 2.3 to 317 s, respectively. The dissolution profile met USP specifications of>85% drug dissolved in 30 min. XRPD indicated partial amorphization of the drug in the printlets and hyperspectroscopy indicated uniform distribution of the components in the formulation. Stability data indicated no significant change in quality of the printlets in terms of dissolution and assay. Excipients used in the formulation reduced the bitterness of the drug to some degree. In summary, binder jetting can be used to print personalized medication with quality control characteristics.

Introduction

According to the World Health Organization (WHO), there were approximately 1.4 million children between the ages of 0 and 14 reported infected with the Human Immunodeficiency Virus (HIV) by the end of 2023, with 120,000 new infections recorded. Moreover, it is estimated that 76,000 children under the age of 15 succumbed to death due to AIDS-related ailments at the end of 2023 [1]. UNICEF reported that in 2023, only 57% of these children were receiving essential antiretroviral therapy (ART) [2]. It’s crucial for caregivers and healthcare providers to work together to ensure that children receiving antiviral therapy are appropriately monitored and supported throughout the treatment process [3]. Additionally, guidelines and recommendations for the use of antiretroviral medications in children should be periodically updated by consulting the latest medical literature. One of the challenges in pediatric HIV treatment is the lack of appropriate formulations in terms of palatability, ease of administration, convenience of administration frequency, stability during storage, and dose flexibility [4].

Lamivudine (LMV) is an antiretroviral medication used primarily in the treatment of HIV as single agent or in combination with other drugs. It belongs to a class of drugs called nucleoside reverse transcriptase inhibitors [5]. Use of LMV in children is well-established, and it is included in the recommended ART regimens for pediatric patients with HIV [6]. Liquid solution formulation of LMV is clinically available which is convenient to administer with dose flexibility capability. Moreover, combination ART of LMV with abacavir, stavudine, tenofovir, or zidovudine, etc. is more effective than single ART to control HIV symptoms [7]. Combination formulations of LMV with other antiretroviral medication is available. However, combination dosage form available is tablet which lack dose flexibility capability. LMV is generally well-tolerated, but like any medication, can have side effects. Some of adverse events related to inappropriate doses such as overdose or, underdose [8, 9]. Therefore, correct dosing is paramount for optimum clinical outcome.

Pediatric patient requires correct dose, which is calculated based on age and body weight, that cannot be provided by clinically available fixed dose formulations [10]. There are various techniques by which flexible dosage forms of a drug can be formulated, and thus, a personalized treatment is given according to need of a patient [11]. 3D printing technologies offer an alternative to current practice of compounding with quality control and stability tested features. The printed dosage unit contain a known amount of drug per dose unit that can be modified according to patients’ needs, and further customized with features like dispersibility or chewability, thus, providing better stability, dosing flexibility and convenience [12]. It’s a facet of personalized medicine, where treatments can take into consideration more factors such as sex, genetics, and clinical traits in addition to age and body weight to achieve superior outcomes with minimal adverse effects. With its ability to fabricate intricate structures precisely, 3D printing shows promise not only in drug delivery systems, but also in crafting personalized medical devices and implants [13].

Binder jetting is one of 3D printing techniques to print the dosage forms. Literature reported binder jetting in designing personalized dosage forms, multi-drug combinations, and controlled release printlets [14]. The printing starts with a powder bed made of polymers, excipients, and drug. Layer by layer, the powder blend is spread, and a liquid binder is selectively applied based on the dosage form’s 3D model [15]. The liquid binds the particles together to form printlets. Post-printing, printed dosage forms undergo drying at a specific temperature to remove sprayed solvent followed by sieving. Properties of solvent, binder, excipients, and polymer used is binder jetting process impacts the success of the process and the properties of the finished product [15]. Investigators developed various dosage forms e.g., fast orodispersible, oral disintegrating, fast disintegrating, complex, and fast dissolving tablets, etc. of acetaminophen, indomethacin, warfarin sodium using binder jetting process and excipients such as microcrystalline cellulose, povidone, lactose, mannitol [16, 17]. The objective of the present research was to understand interactions of formulation and process variables of binder jetting process on the critical quality of LMV dosage form that can be applied to combination dosage used in pediatric population and also demonstrate dose-flexibility. Second objective was to develop non-destructive multivariate model for quantification of drug in the printed dosage forms and taste assessment using electronic tongue [18, 19].

Materials and Methods

Materials

LMV was purchased from AK Scientific, Ahren Ave, CA. Hydroxypropyl methylcellulose (HPMC, Viva Pharm, Rosenberg, Germany), lactose monohydrate (LMH, Supertab-14SD, DFE Pharm, Goch, Germany), methanol (MeOH), glacial acetic acid, and acetonitrile (ACN, Fisher Scientific, Asheville, NC) were used as received.

Methods

Design of Experiments

Screening design was selected to understand the effect of formulation and process variables on printlet quality attributes and stability. Independent variables studied were number of spray (X₁ 2–4), HPMC (X₂, 10–30%,), layer thickness (X₃, 0.32–0.40 mm), delay time in printing each layer (X₄. 0–10 s), and drying temperature (X₅, 40–60°C). Fractional factorial screening design was created using JMP 15 software (SAS, Cary, NC). Responses measured were hardness (Y₁, disintegration time (DT, Y₂), and dissolution (Y₃) (Table I).

Table I.

Independent Variables and Experiment Matrix of the Printlet Formulations

Formulations	Number of sprays (X1)	HPMC (X2, %)	Layer thickness (X3, mm)	Delay time (X4, sec)	Drying temperature (X5, °C)
F1	2	10	0.32	10	60
F2	4	30	0.32	0	60
F3	2	10	0.40	0	60
F4	2	30	0.40	10	40
F5	2	30	0.32	0	40
F6	4	10	0.40	0	40
F7	3	20	0.36	5	50
F8	4	30	0.40	10	60
F9	4	10	0.32	10	40

Manufacturing with Dose Flexibility Printlets

The printlet formulations contained LMV—50%, HPMC – 10–30%, and LMH – 20–40%. The effect of particle size was eliminated by passing the components through ASTM sieve #60 followed by blending in V-blender for 5 min. Batches of 80 (10 × 8) printlets of different sizes were prepared. The size of printlets included 8 mm diameter × 3 mm height, 6 mm diameter × 3 mm height, and 4 mm diameter × 3 mm height [20]. The printlets were prepared by varying number of sprays 2–4, layer thickness 0.32–0.40 mm, delay time in printing each layer from 0–10 s, and drying temperature 40–60°C. The printing process involved spreading of reservoir powder as a layer in the printing area followed by spraying of solvent (water) at the selected region of layer of powder as per the design. Powder layering and solvent spraying were repeated till the desired printlets were printed. The printlets along with support powder were dried in oven to remove the sprayed solvent and sieved to separate the printlets from the support powder. The printlets were characterized for weight, diameter, and thickness. USP disintegration apparatus (Vankel Varian VK-100, NC) was used to determine disintegration time (DT, n=6) using water as a medium at 37°C [21].

Hardness

The texture analyzer (TA.XT Plus, Stable Micro Systems, Surrey, UK) was used to measure the mechanical strength of the printlets. Mechanical strength test was performed on printlets with radial direction, compression mode, compression distance 4 mm, 2 mm/s pretest speed, 1 mm/s test speed, 10 mm/s post-test speed, target mode-distance, 2 cm, auto trigger, and 1 N trigger force. The experiments were performed in six replicates [22].

Assay

The drug content in the printlets were determined by content uniformity method. The printlet was placed in a 100 mL volumetric flask, followed by addition of 50 mL water and sonication for 30 min. The volumetric flask was filled up to 100 mL with water, filtered through 0.45 μm filters (Thermo Scientific, Waltham, MA) and diluted with mobile phase. Samples were analyzed for LMV by validated high performance liquid chromatography (HPLC) method. Experiment was performed in five replicates.

Dissolution

Dissolution of the printlets was performed in a USP apparatus 2 (Model 708-DS with 850-DS autosampler, Agilent Technologies, Santa Clara, CA) using minipaddle and 200 ml mini-vessel at 50 rpm and 37°C. Dissolution medium used was 100 mL water, and sample (0.5 mL) was collected at 5, 10, 15, 20, and 30 min. The dissolved drug was determined using the validated HPLC method.

Fourier-Transform Infrared Spectroscopy

Modular Nicolet^™ iS^™ 50 system (Thermo Fisher Scientific, Austin, TX) equipped with ATR accessory was used to collect FTIR spectra of the samples. The sample (5–10 mg) was placed on the diamond crystal and pressed with the attached arm to remove entrapped air in the sample. The data collection parameters were absorbance mode, wavelength range of 400–4000 cm⁻¹, resolution of 4 cm⁻¹ and 100 scans. OMNIC software, version 9.0 (Thermo Fisher Scientific, Austin, TX) was used to capture and analyze the spectra [23].

X-ray Powder Diffraction

X-ray Powder Diffraction (XRPD) patterns were obtained using a Bruker D2 Phaser SSD 160 Diffractometer (Broker AXS, Madison, WI). Approximately 400 mg of the powdered sample was used to fill the sample holder cavity and pressed to reduce preferential crystal orientation. The samples were scanned across a 2θ range of 5–10° with a step size of 0.0202°, spending 1 s per step (for a total of 3000 steps). To collect average diffractogram, the samples were rotated at a speed of 15 rpm. Data collection and analysis were done with Diffrac.EVA Suite version V4.2.1, and processing was completed using File Exchange 5.0 (Bruker AXS, Madison, WI) [24].

Differential Scanning Calorimetry

Thermal analysis of the samples was performed using a Q2000 differential scanning calorimeter (DSC) (TA Instruments, New Castle, DE). Approximately 2.5 mg of the sample was accurately weighed and sealed in standard aluminum pans. The thermal scans were conducted up to 300°C at a heating rate of 10°C/min. An inert atmosphere was maintained by purging nitrogen gas at a flow rate of 50 mL/min and a pressure of 20 psi throughout the analysis.

Near Infrared Hyperspectral Imaging

Calibration samples were prepared as physical mixtures containing known concentrations of LMV, combined with different amounts of two excipients HPMC and LMH. The LMV concentration was gradually increased from 0 to 100% while changing the levels of HPMC and LMH accordingly. Each sample was placed on a metal holder and positioned on the movable stage of the Via-Spec II system. A halogen line light source was used to illuminate the samples for image collection. The spectral camera can capture images at a resolution of 384×288 pixels, with each pixel measuring 24×24 μm. This camera is a cryogenically cooled mercury-cadmium-telluride sensor and operates through a camera link interface, controlled by the Via-Spec SWIR control systems. The instrument uses a push-broom setup to create a chemical image by moving the sample under an imaging spectrograph, allowing for simultaneous spectral measurements across adjacent spatial points. The lens-to-sample distance was adjusted for optimal focus in reflectance mode, ensuring clearer images. Data were collected with an integration time of 7.005 ms, a frame rate of 75 MHz, and a scan-axis speed of 0.2 in/sec. Reference images, both pure white and dark, were captured before sample measurement. Middleton Spectral Vision software was used for data acquisition, while data analysis was done using Prediktera Evince^™ (Prediktera AB, Umea, Sweden) [25].

Taste Assessment—Electronic tongue

Washing solutions for sensors were prepared based on their surface charge. For negatively charged sensors, washing solution consisted of 10% v/v 1 M HCl in 60% v/v alcohol in water. For positively charged sensors, the washing solution consisted of 7.46 g of KCl, 300 ml ethanol, and 10 ml 1 M KOH, and water was added to a liter. A standard solution was used both for cleaning and as a reference solution (Vr). It was prepared by dissolving 45 mg of tartaric acid and 2.24 g of KCl in water and made up the volume up to a liter. HPMC, LMH and printlets (F1 and F2) samples were prepared by dissolving in artificial saliva (supplementary data) [26] to simulate oral environment followed by filtration through 0.45 μm membrane. It was assumed that a printlet will be dispersed in 5 ml medium considering patient in use condition. 8 printlets were dissolved in 40 ml medium (LMV 25 mg/40 ml), since the minimum volume requirement for the electronic tongue test instrument is 40 ml. Samples of HPMC of 40 mg (low) and 120 mg/40 ml (high) medium that represent 10 and 30% polymer used in the prinlet formulations, and LMH 80 mg (low) and 160 mg/40 ml (high) medium that represent 10 and 30%, and 20 and 40% sugar used in the prinlet formulations, and LMV 25 mg/40 ml used as a positive control. Placebo of F1 (HPMC 40 mg/40 ml and LMH 160 mg/40 ml medium) and F2 (HPMC 120 mg/40 ml and LMH 80 mg/40 ml medium) formulations were prepared to delineate taste masked by the excipients in the formulations. Taste of the formulation was assessed using the TS-5000Z taste sensing system (Insent Intelligent Sensor Technologies, Inc, Atsugi City, Japan). This electronic tongue can accommodate up to eight lipid membrane sensors along with four corresponding reference electrodes. The system operates on a potentiometric principle, with sensor outputs record electrical potential in millivolt (mV) values. Prior to measurement, a routine sensor check was performed by the software to ensure that all sensors were operating within their acceptable mV range. Each sample was analyzed four times using a set of four electronic taste sensors: two positively charged sensors namely C00 (measures initial bitterness, sensor 1) and AE1 (measures initial astringency, sensor 2), and two negatively charged sensors namely BT0 (sensor 3) and ANO (sensor 4), both measure aftertaste bitterness. A complete measurement cycle included the following steps: measurement of a reference solution (Vr), measurement of the sample solution (V_s), a brief cleaning step (2—3 s) followed by another measurement of the reference solution to assess aftertaste. The aftertaste was evaluated by calculating the change in membrane potential resulting from the adsorption of sample components onto the lipid membrane after cleaning [27].

Stability Study

Patient in-use stability of the printlets was performed by packaging F1, F7 and F8 printlets in pharmacy vials and stored at 30°C/75% RH for a month. The printlets were monitored for assay, dissolution, and physicochemical characteristics (XRPD and FTIR).

High-Performance Liquid Chromatography

The HPLC system (HP 1260 series, Agilent Technologies, Wilmington, DE) consisted of a quaternary pump, autosampler, and UV as detector. A mixture of Ammonium acetate buffer (5 mM) pH 6.0 and acetonitrile (95:5 v/v) was used as a mobile phase. The mobile phase was pumped at 1 mL/min through reversed phase stationary phase, Eclipse, XDB 5 μm, C8 column 150×4.6 mm (Agilent, Santa Clara, CA). The sample (20 μL) was injected through system, and LMV was detected at 270 nm. Data were collected and analyzed using OpenLab software (Agilent Technologies, Wilmington, DE).

Results and Discussion

Preliminary Data

Formulation and process variables were selected based on preliminary data and literature. Using water as the binding solution due to technical limitation of the printer and considering the physicochemical properties of LMV, various polymers were explored such as HPMC, hydroxypropyl cellulose (HPC), starch, and polyvinyl pyrrolidone (PVP). Through one-factor-at-a-time trials, LMH combined with HPMC was the most effective excipients combination for the printlets. HPC and PVP proved unsuitable, as the printlets crumbled quickly after drying. Drying temperatures were also tested ranging from 30 to 60°C, however the optimal temperature range was found to be between 40 and 60°C. Temperatures below 40°C led to prolong drying times and incomplete drying of the printlets. This temperature range is commonly used in pharmaceutical manufacturing [28]. Furthermore, temperatures above 60°C could degrade drug and produce impurities in the printlets. Concentration of HPMC was also critical: amount lower than 10% compromised the printlets’ mechanical strength, causing them to break after removal from the oven, while amount above 30% extended disintegration time. Thus, HPMC range of 10–30% was selected for the screening study.

Effect of Independent Variables on Dependent Responses

Hardness (Y₁ of the printlets varied from 2.5 ± 0.1 (F3) to 21.1 ± 0.5 (F2) N (Fig. 1A). Effect of variables on the hardness can be explained by the following mathematical equation:

$${\text{Y}}_1=7.96+2.70{\text{X}}_1+3.68{\text{X}}_2-0.48{\text{X}}_3-0.13{\text{X}}_4+2.2X_5-1.50{\text{X}}_1{\text{X}}_3-2.42{\text{X}}_1{\text{X}}_4$$

Fig. 1.

A Hardness, and B disintegration time of the printlet formulations

A good correlation of 0.998 was obtained between actual and model predicted values. The model can explain 95.85% variation in data. Among all the studied variables, number of sprays and percentage of HPMC had statistically significant effect on Y₁ (p<0.05) [29]. An increase in the hardness with an increase in values of these variables were observed. This could be observed by comparing hardness of formulations F1 and F2 containing 10% HPMC and manufactured with 2 and 4 number of sprays, respectively. The hardness of F1 and F2 were 5.1 and 21.2 N, respectively. Similarly, HPMC percentage increased the hardness of the formulations as demonstrated by F2 and F9 formulations manufactured with 4 number of sprays but contained 30 and 10% HPMC in their composition, respectively [30]. Hardness decreased from 21.2 (F2) to 4.9 N (F9) with a decrease in HPMC concentration in the composition. Other investigated independent variables had statistically insignificant (p>0.05) effect on Y₁. On the other hand, powder layer thickness had inverse relationship with the hardness [31] as thicken layer decreased traverse of the solvent across the deposited layer. Thus, decreased the effective fusion with the preceding layer. Increasing delay time produced printlets of higher mechanical strength compared to the one produced at shorter delay time. Printing with longer delay time allowed the blend to absorb binder and swell, thus producing printlets of higher mechanical strength. This was demonstrated by F8 and F5 formulations with 12.1 and 5.4 N hardness, respectively. F8 and F5 formulations were printed with delay in layering time of 10 and 0 s, respectively, while keeping the layer thickness and number of sprays same in both the formulations [32]. Increasing temperature of drying increased hardness of the printlets which could be explained by faster drying that limit the lateral spread of binder [33].

The printlet formulations meet the requirement of orally disintegrating tablets in terms of DT of≤ 30 s except F2 and F8. Printlets showed DT varied from 2.3 ± 0.5 (F3) to 317 ± 10 s (F2) (Fig. 1B). The following mathematical equation can explain effect of the studied variables on the DT (Y₂):

$${\text{Y}}_2=63.85+59.59{\text{X}}_1+57.5{\text{X}}_2-20.83{\text{X}}_3-19.17{\text{X}}_4+56.41{\text{X}}_5-19.17{\text{X}}_1{\text{X}}_3-19.17{\text{X}}_1{\text{X}}_4$$

Model exhibited a good correlation of 0.999 between empirical and model predicted values. Further, the model could explain 97.02% variability in the data. Among the studied variables number of sprays, drying temp, and HPMC percentage had statistically significant impact (p < 0.05) on the DT [34]. Increasing HPMC percentage, drying temperature, or number of sprays prolong disintegration of the printlets. The effect on DT was related to hardness as indicated by positive correlation of 0.99 between these two responses. The effect of number of sprays was represented by F1 and F2 with 317 and 8 s DT, respectively. Similarly, F2 and F9 were examples of the printlets showing the effect of HPMC percentage on the DT with their disintegration in 317.3 and 12.0 s, respectively. Effect of drying temperature on DT was related to HPMC percentage, layer thickness and number of sprays. As the values of these parameters increased, DT increased [35].

The assay limit for LMV tablets, according to the USP monograph, is between 90.0 and 110.0%. The printing process did not affect the drug content in the printlets that ranged from 100.2 ± 1.6 to 105.2 ± 2.3%. According to the United States Pharmacopeia, the dissolution specification for LMV tablets is 85% in 30 min. The printlets achieved complete dissolution in 15 min, with a range of 98.8 ± 0.9 to 104.2 ± 1.2% in 30 min. Process and formulation variables had no notable impact on dissolution between 20 and 30 min. To understand the impact of formulation and process variables on dissolution (Y₃), the dissolution in 10 min was chosen as a dependent response [36]. Dissolution in 10 min varied from 62.5 ± 1.8 (F2) to 103.5 ± 1.0% (F4) (Fig. 2) and can be described by the following mathematical equation:

$${\text{Y}}_3=101.52-1.2{\text{X}}_1-0.14{\text{X}}_2+0.04{\text{X}}_3+1.11{\text{X}}_4-0.19{\text{X}}_5-0.67{\text{X}}_1{\text{X}}_3+0.06{\text{X}}_1{\text{X}}_4$$

Fig. 2.

Dissolution profiles of the printlet formulations

The relationship between empirical and model calculated values of the Y₃ was good as indicated by correlation coefficient of 0.999, and explainable data variation by the model was 94.0%. Number of sprays and delay time had statistically significant impact on the dissolution (p<0.05) among the studied variables. Dissolution was inversely impacted by the hardness and DT. Mechanically strong printlets had longer DT that would subsequently dissolve less amount of drug in a specified time. This was indicated by negative correlation between dissolution and DT, and dissolution and hardness of −0.073 and −0.67, respectively [37]. The effect of number sprays was exemplified by F2 and F9 with 62.5 ± 1.8 and 95.9 ± 2.7% dissolution in 10 min, respectively. F1 and F2 formulations demonstrated the impact of number of sprays on the dissolution. Similarly, F2 and F8 represented examples of formulations that showed the effect of delay time. F2 and F8 were printed with 0 and 10 s delay time between the spray while keeping other parameters same in these formulations.

The model’s error is demonstrated by the root mean-squared error (RMSE) and residuals. The RMSE for hardness, DT, and dissolution were 3.28, 4.96, and 1.26, respectively. Residuals represent the difference between experimental values and model-predicted values. The residual ranges −3.1 to 0.4, −4.7 to 5.9, and −1.2 to 0.1 for hardness, DT, and dissolution were, respectively. These data, along with correlation coefficient values and the explained variation, indicate a good fit and low error in the models [38],

Developed predictive model of various responses can be utilized in rapid screening and selection of range of independent variables (e.g., layer thickness, binder content, and drug to excipient ratio, time, and temperature for drying) to achieve the desired response without the need for extensive experimental trials. This not only accelerated formulation development but also reduced material wastage and overall development costs. In the later stages, particularly during process optimization and scale-up, the model can predict the impact of minor variations in process parameters on printlet quality attributes ensuring consistent product quality with mechanical integrity.

Fourier Transform Infrared

LMV spectrum showed absorption band at 1650 cm⁻¹ due to vibration of carbonyl group (C = O) of cytidine nucleus (Fig. 3). In addition to the carbonyl peak, the spectrum showed absorption bands at 3328 and 3200 cm⁻¹ due to stretching vibration of primary amino (NH₂) group of pyrimidine ring and hydroxyl (OH) group of 1,3-oxathiolane ring, respectively [39]. Furthermore, the spectrum included absorption bands at 1286 and 1160 cm⁻¹ due to asymmetrical and symmetrical stretching vibrations of the C-O-C (ether) bond of oxathiolane ring. The physical mixture (pre-print powder formulations) of the drug and excipients of the formulations showed additive spectra. The distinctive peaks of the drug did not interfere by excipients. Similarly, the printlet formulations showed the spectra that were similar to physical mixture. No differences were observed in spectra of the formulations printed with different processing parameters. It was expected that drug crystallinity may change during printing due to drug solubility in water. It appeared that FTIR technique cannot differentiate between amorphous and crystalline form of the drug [40].

Fig. 3.

Fourier transformed infrared spectra of lamivudine, placebo, physical mixture and printlet formulations

X-ray Powder Diffraction

The LMV showed characteristic reflection bands at 2θ value of 12.3, 13.5, 14.4, 14.9, 15.9, 17.0, 17.7, 20.7, 21.5, 23.8, 24.5, 25.0, and 26.7° that indicated its crystalline nature. Some of the reflection bands of the drug in the physical mixture of drug with excipients were interfered by lactose while HPMC is amorphous in nature. The non-interfered reflection bands of the drug were 13.5, 14.4,14.9,17.7, 20.7, 21.5, 24.5, 25.0, and 26.7°. The printlet diffractograms were similar to the physical mixture albeit with some differences [41]. The major difference in the diffractograms of physical mixture and printlets were intensity of reflection bands of the drug. Reduction in intensity of reflection peaks of the drug in the printlets were observed compared to the physical mixtures. Reduction in reflection bands intensity of prominent at 20.7, 21.5, and 25° were observed. This was due to drug dissolution in the solvent. Furthermore, reduction in the intensity of lactose also observed but was not as significant as the drug was. For examples, reduction in drug and lactose intensity was 40–50% and 5–10% in F2 printlet formulation, respectively. Furthermore, reduction in reflection bands of the drug was also dependent of printing process parameters especially number of the spray. Reduction in intensity of reflection band at 20.7° were 28.5 and 50.1% in Fl and F2, respectively, which were manufactured with 2 and 4 number of sprays (Fig. 4) [42].

Fig. 4.

X-ray powder diffraction of lamivudine. physical mixture and printlet formulations

Differential Scanning Calorimetry

Thermogram of the LMV exhibited a distinct endothermic melting peak at 178.07°C (Fig. 10) with a ΔH value of 124.7 J/g, confirming its crystalline nature and aligning with previously reported literature data. Among the excipients used in the printlet formulations, lactose showed a characteristic crystalline endothermic peak near 150°C. The physical mixtures of formulations Fl and F2 displayed sharp endothermic peaks corresponding to both LMV and lactose. However, the heat of fusion (ΔH) values for LMV decreased to 58.5 J/g and 57.9 J/g for Fl and F2, respectively, representing almost 53% reduction in crystallinity compared to the drug. This reduction may be attributed due to dilution with lactose and HPMC that constituted 50% mass of the matrix. The ΔH values of lactose in a physical mixture was 48.7 J/g and 41.0 J/g for Fl and F2 respectively. Thermograms of Fl and F2 formulations showed melting endothermic peaks of the drug and lactose similar to corresponding physical mixture. However, a decrease in intensity of endothermic peak of the drug and lactose was observed in the printlets due to dissolution of the drug and lactose while printing the printlets. The DSC data aligned with XRPD; however, crystallinity values/pattern did not match with thermogram data. This may be due to sample non-homogeneity and small sample size used in DSC compared to diffractogram. ΔH values of LMV for Fl and F2 were 45.77 and 45.69 J/g, respectively, that correspond to approximately 22% reduction in crystallinity compared to the physical mixtures. Similarly, ΔH values of lactose for Fl and F2 were 32.7 and 30.1 J/g, respectively, that correspond to approximately 30% reduction in crystallinity compared to the physical mixtures.

Fig. 10.

DSC of lamivudine. physical mixture and printlet formulations

Particle Size Analysis

Not all the powder is printed into the printlets as the powder provides support to the printlets. The printed printlets were dried along with the unprinted powder. The powder properties such as particle size, assay, crystallinity etc. may change due to powder exposure to printing and drying processes. It is important to understand the powder properties before it can be recycled or mixed with fresh powder before printing it as it may change quality attributes of the printed dosage forms. The particle sizes of the powders before and after printing displayed statistically insignificant differences (p>0.05) [43], The particle size (D₉₀) distribution of Fl, F7 and F8 formulations were compared due to different composition and process variables. Minor differences in D₉₀ were observed between pre- and post-printed powder which could be attributed to compositional differences, printing, and drying processes. D₉₀ of pre- and post-printed powder of F1,F7 andF8 were 118.6±2.4 and 118.2±3.0, 113.9±2.1 and 111.3 ±3.7, and 108.2 ±3.6 and 109.0 ±0.2 pm, respectively. These results suggest that the printing process did not statistically significantly affect the particle size distribution of the powders after exposure to the printing process (p>0.05) [17],

FTIR spectra of pre- and post-printed powder (not shown) did not show significant differences that indicated no chemical transformation in powder after exposure to printing and drying processes. Similarly, no significant in qualitative diffractograms were observed between pre- and post-printed powder. However, quantitative changes were observed in terms of reflection peak intensity, especially 20.7° reflection band, which was related to the drug. Furthermore, a decrease in intensity was higher in the formulations printed with greater number of sprays. For example, decrease in intensity 20.7° was 5.7 and 12.3% in Fl and F2, respectively, which were printed at 2 and 4 sprays, respectively. This was due to drug solubilization due to lateral movement of the solvent beyond printed region (Fig. 5) [44].

Fig. 5.

X-ray powder diffraction of preprinitng and postprinting powder formulations of printlet

Near Infrared Hyperspectral Imaging

The distribution of LMV in the formulation components within the sample was visualized using NIR hyperspectral imaging. To minimize noise and correct baseline variations, data was preprocessed with standard normal variate (SNV) method. In this study, principal component analysis (PCA) concentration images were generated using three components in the analysis since formulation contained three components. The resulting concentration images revealed a relatively uniform distribution of pixel colors, suggesting an even dispersion of the formulation components in the matrix. Furthermore, uniform distribution of pixels in the printlets indicated that the manufacturing process did not significantly impact the distribution of the drug (Fig. 6) [45], Moreover, uniformity of components distribution was measured by minimum, and maximum values of pixel intensity, standard deviation, and skew, which measure asymmetry of the distribution. The minimum and maximum intensity of the of pixel intensity varied from −1.9 to 4.0. Standard deviation measures distribution of the data and its values were narrow as indicated by values of standard deviation ranging from 0.9–1.21. Skew measures asymmetry of the distribution and value below considered acceptable and skew value ranged from 0.63 to 1.95.

Fig. 6.

NIR hyperspectral chemical images of the printlet formulations

Partial least squared regression was performed on the PCA data set for development of quantification model. LMV, HPMC and LMH were used as predictors and samples representing composition of the drug were used calibrators. In the calibration, pixels color was shifting from blue to red color as the concentration of drug increased in the calibrators. Further, pixels color distribution in the printlet formulation concurred with calibrator as both contained 50% drug in their composition. The PLS model demonstrated excellent linearity with drug concentration in actual and model predicted value as indicated by R² value of 0.999 (Fig. 7). Furthermore, the model exhibited low error as suggested root mean square error of 0.8239. The developed model was applied to the printlets data for drug quantification. The model predicted value of the drug in the printlets varied from 99.1 ± 3.3 (F5) to 105.0 ± 6.6% (F7) of nominal value of drug content in the printlets. Thus, all the printlets met USP pharmacopoeia limit of the assay. On comparison, assay values by HPLC method varied from 100.2 ± 1.8 to 105.2 ± 1.0% of the nominal values. This showed concurrence of the results between two methods. However, standard deviation was higher in the case of NIR hyperspectral method [46].

Fig. 7.

A NIR hyperspectral chemical images showing distribution of LMV in the calibration samples and B calibration curve of the data

Taste Assessment—Electronic Tongue

The readings obtained from blank artificial saliva were subtracted from the sample measurements to eliminate any background interference. Among the used sensors, BT0 (bitter sensor 3) was most sensitive to bitter taste of the drug. BT0 is a negatively charged sensor, and its readings are interpreted such that more negative values of voltage potential correspond to greater bitterness, while values of voltage potential that are less negative or shift toward the positive indicates a reduction in bitterness intensity [47], LMV alone produced a value of −3.33 mV voltage potential reflecting bitterness of the drug that concurred with the literature. On the other hand, HPMC and LMH exhibited positive values of electrical potential that indicated their potential to reduce bitterness. Increasing concentration of either HPMC or LMH resulted in an increased positive value of voltage potential [48]. Moreover, LMH performance in terms of test reducing bitterness would be better than HPMC as indicated by higher value of the voltage potential. This was further evidenced by value of the voltage potential when combined both HPMC and LMH mixture. Placebo formulation of Fl and F2 printlets exhibited positive values of voltage potential which were almost eight folds of the drug. However, voltage potential of the printlets formulation were significantly lower than corresponding placebo and difference between them statistically significant (p<0.05). On further comparison between printlet formulation and the drug, statistically significant difference was observed (p<0.05) [49]. However, difference was less between F1 formulation and drug compared to between F2 formulation and drug which means excipients composition of F2 would be more effective in reducing bitter taste of the drug. The value of voltage potential was positive 1.8-fold of the drug value. Similarly, voltage potential value was negative and 0.6-fold of the drug value. This could be explained by compositional difference between placebos F1 and F2. The placebos F1 and F2 contained 10% HPMC and 40% LMH, and 30% HPMC and 20% LMH, respectively. HPMC and lactose work by different mechanism to reduce bitter taste. LMH has sweet taste while HPMC work by coating the drug molecule to reduce the degree of the bitterness (Fig. 8) [27, 50].

Fig. 8.

Electronic tongue data of the printlets

Stability

No significant changes in the surface morphology of the printlets were detected after exposure. FTIR showed no significant changes in the spectra of the printlets before and after exposure to stability condition (Fig. 3). Similarly, XRPD diffractogram of stability condition exposed printlets showed insignificant change in pattern and intensity of major reflection peaks (Fig. 4). Contrastingly, DSC data showed about 29% decreased in crystallinity based on the values of heat of fusion of the drug after storage. Heat of fusion of F1 and F2 were 58.5 and 57.9 J/g before storage, respectively. Values of ΔH of F1 and F2 decreased to 40.91 and 39.83 J/g, respectively. Similarly, a decrease in crystallinity of 23–31 % based on ΔH values were observed for lactose after exposure to storage condition. ΔH values of lactose for F1 and F2 decreased from 32.7 and 30.1 J/g to 22.54 and 23.05 J/g after storage to patient in-use condition, respectively. Decrease in crystallinity of the drug and lactose could be explained by moisture mediated dissolution of these components present in the formulations.

Dissolution results indicated that exposed printlets showed insignificant changes in the dissolution profiles when compared to before stability condition exposed printlets. The formulations still met USP specification of dissolution of 85% in 30 min. Amount of drug dissolved in 30 min in F1 and F2 formulation before and after stability exposure were 103.1 ± 1.2 and 100.6±0.1%, and 99.4 ± 1.5 and 105.1 ± 2.3, respectively (Figs. 9 and 10).

Fig. 9.

Dissolution profiles of the printlet formulations before and after storage at 30°C/75% RH

Conclusion

The study demonstrated feasibility of fabricating lamivudine oral printlets using binder jetting 3D printing technology. Process and formulation could impact the quality of the printlets. Among the studied variables, number of sprays and percentage of HPMC had statistically significant effect on studied (p<0.05). The printlets were mechanically strong that can withstand mechanical stress during transportation and usage period. However, the printlets were porous as indicated by quick disintegration of <30 s that entail quick dispersion in water or administration without a sip of water. Diffractometry and thermograms indicated partial amorphization of the drug during printing process. However, solid phase transformation of drug has no significant impact on the quality as drug is highly soluble. Hyperspectral image indicated uniform distribution of the components with no segregation or aggregation during printing process. Moreover, non-destructive quantification data was in concurrence with the wet chemistry method. Short term stability indicated no significant change in quality and meeting the pharmacopeial quality parameters. The excipients used could be able to reduce bitter taste of the drug to a certain degree. However, artificial sweetener and flavoring may be required to effectively mask the bitter taste of the drug.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1208/s12249-025-03289-z.