DLS was employed to determine the average polydispersity index (PDI), particle size, and zeta potential of the samples, using a zetasizer (Malvern Instruments, UK) at a temperature of 25 °C. A 100-fold dilution of all samples was prepared using deionized water and then injected into a disposable cuvette. The zeta potential was measured for both the optimized formulation and CBD-loaded NLC. All measurements were carried out in triplicate (n = 3) [50].

The technique used for determining the entrapment efficiency (EE) and drug loading (DL) was based on ultrafiltration/centrifugation [31]. To achieve this, CBD-NLCs (0.5 ml) was introduced into Amicon^Ⓡ (50-KD cut-off) ultrafiltration devices and centrifuged at 3400 rpm for 30 min. The NLCs held on the filter were washed three times to eliminate any free drug, and the HPLC method described above (the “HPLC method for quantification of CBD” section) was used to determine the quantity of CBD in the filtered pool (free drug). Total amount of CBD was determined by first diluting the NLCs (50 µL) in simulated salivary fluid (X 20 dilution) and analysing using HPLC. Equations (1) and (2) were used to calculate the EE (%) and DL (%) respectively.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{EE}\%=\frac{\text{drug amount(initial)}-\text{drug amount(free)}}{drug\;amount(total)}\times 100$$\end{document}EE%=drug amount(initial)-drug amount(free)drugamount(total)×1002\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{DL}\%=\frac{\text{drug amount(total)}-\text{drug amount(free)}}{lipid\;amount(total)}\times 100$$\end{document}DL%=drug amount(total)-drug amount(free)lipidamount(total)×100

The optimization of CBD-loaded NLCs involved the utilization of numerical optimization and the desirability function approach. The main aim was to obtain NLCs with the smallest possible particle size and PDI. To determine the optimal values for the independent variables, the desirability function method was employed. This approach entailed evaluating the desirability index for each response variable and then combining all response variables into a single desirability function that ranged from 0 to 1, indicating the ideal values of the independent parameters [51].

Smoothness and homogeneity were assessed for the printed films, followed by the characterization of physicochemical properties and release kinetics. The thickness and weight of the films were determined after drying them at room temperature for 48 h using a digital micrometer and weighing balance, as outlined by Bala et al. [52].

The particle size recovery from the 3D printed film was evaluated in triplicate using zetasizer (Malvern Instruments, UK) at a temperature of 25 °C. Each 3D printed film (20*20 mm²) containing CBD was dispersed in 10 mL of deionized water, under constant stirring until its complete disintegration. Subsequently, the samples were filtered using 0.22 µm syringe filter and diluted 100-fold before injecting into a disposable cuvette for particle size analysis.

A texture analyzer (Stable Micro Systems, Godalming Surrey, UK) was used to evaluate the elongation at break and tensile strength (TS). The films were pulled apart, at the loading length of 200 mm, until breakage occurs by moving the upper clamp at a rate of 1 mm/s. The lower clamp remains stationary. The mucoadhesion was determined using a texture analyser as previously described by our group [27]. Briefly, the porcine mucosa was first mounted on the platform and the film was attached to a probe using a double-adhesive tape. The probe was then lowered at 0.5 mm/s and allowed to maintain contact with mucosa for 2 min. Lastly, the probe was withdrawn at 1 mm/s and the maximum force applied to completely detach the film from the buccal tissue (Fmax) was recorded. All the measurements were done in triplicate.

FTIR-attenuated total reflectance spectra of the CBD, Precirol^® ATO 5, lipid mix (Precirol^® ATO 5 + Caprylic oil), HEC, physical mixture (CBD, Precirol^® ATO 5, lipid mix and HEC), blank and CBD loaded film were obtained using FTIR spectrometer (Bruker, Massachusetts, USA). The spectra were recorded at room temperature in a range of 4000 to 450 cm⁻¹ in transmittance mode using 4 scans per analysis at a resolution of 4.0 cm⁻¹. A small portion of the films or powder was placed on ATR diamond crystal followed by application of force with the use of the clamp to ensure adequate contact of the sample with the crystal.

DSC measurements of CBD, Precirol^® ATO 5, lipid mix (Precirol^® ATO 5 + Caprylic oil), HEC, physical mixture (CBD, Precirol^® ATO 5, lipid mix and HEC), blank and CBD loaded film were taken in Discovery DSC 2920 (TA Instruments (New Castle, USA) calibrated with an indium standard. Samples weighing 4.0 ± 0.5 mg were put in aluminum pans followed by recording of thermal profiles by heating the samples from 25 to 250 °C at a rate of 10 °C/min while continuously flowing nitrogen gas.

The morphology of the films and pure drug was evaluated using a Zeiss Merlin Field-Emission Dispersive X-Ray Spectroscopy (Jena, Germany) operating at an accelerating voltage range of 2–5 kV, after sputter-coating with platinum.

Thickness of the film was determined by measuring five locations (four corners and one center) using a digital micrometer (ID-S1012, Mitutoyo, Japan) as described by Bala et al. [53]. Dry weight of the film was determined by randomly cutting four pieces (0.64 cm²) and weighing them using a digital balance.

The surface pH of each film (n = 3) was measured by adding a drop of MilliQ water to the surface and measuring with a pH meter (Orion Star A121, Thermo Scientific, USA) [52].

The folding endurance was assessed by continually folding each film (n = 3) at the same spot until breakage and recording the total number of folds.

To determine the drug loading, films (20*20 mm²) (n = 3) were placed in a Falcon tube containing a hydro-alcoholic solution (10 mL, 50:50 v/v) maintained at 37 °C for 1 h. The solution was then centrifuged at 3000 rpm for 5 min, filtered, and analysed using HPLC.

The method used to determine the in vitro release of CBD from the buccal film was similar to the one reported by our research group earlier [54]. The films (n = 3) were placed in a Falcon tube with 10 mL of simulated salivary fluid (SSF) and kept in a shaking water bath (Julabo SW22, Germany) at 37 ± 0.5 °C while being stirred at 50 rpm. At fixed time intervals of 10, 20, 30, 45, 60, 90, 120, 180, 240 and 360 min, 1 mL aliquots of the sample were withdrawn and an equal volume of fresh SSF was replaced. HPLC was used to analyse the drug content in the withdrawn samples after filtering the samples with 0.45 µM syringe filters.

Several mathematical models were fitted to the drug release data obtained from in vitro release studies in the simulated salivary medium using a DD solver add-in in Microsoft excel [55] (Supplementary Table S1). The adjusted R2, the Root Mean Square Error (RMSE), and the Akaike Information Criterion (AIC) were used to evaluate the goodness of fit.

A convolution approach was used to predict the in vivo performance of the film, as described by our previous report. A convolve function in R programming language was used to perform the convolution [54].

The size of nanoparticles has been shown to influence the optimal interaction with buccal mucosa [56]. The observed particle size ranges from 12.17 nm (SA12) to 84.91 nm (SA11) as shown in Table 3. Table 3

The composition and the measured responses of NLCs

The ANOVA test was conducted to assess the impact of independent variables on the particle size of CBD-NLCs, and the quadratic model demonstrated a high level of significance with a narrow gap between predicted R² (0.9396) and adjusted R² (0.9865), and adequate precision (36.6857) (Table 4). The lack of fit was not significant (p > 0.05). Except for the interaction of lipid concentration and sonication (p > 0.05), all three independent variables and their interactions had a significant effect on the particle size of the NLC formulation (Fig. 4). Factor X₁ (solid lipid amount), X₃ (sonication time), X₁₂, X₂₂, and X₃₂ had a positive effect on PS, whereas X₂ (surfactant concentration) had a negative effect. The final equation, which is a combination of coded factors, confirms the result. Table 4

Regression analysis of the dependent variables using the best fitting model

The positive effect of the total lipid on the size of the nanoparticles could be due to the increase in the viscosity of the formulation that in turn reduces the effectiveness of particle-breaking (sonication) processes [57]. The increased particle size with the increase in the amount of total lipid could also be attributed to other reasons such as aggregation between lipid particles, increased chances of collision as well as inadequate surfactant amount to cover the lipid particle’s surface [58]. The finding agrees with the work of Kim et al. where NLCs of imiquimod with higher lipid concentration resulted in larger particle size. A similar effect of lipid amount on particle size was illustrated by Jain and colleagues [59] showing that the mean particle size of the prepared NLCs was significantly affected by the amount of total lipid. A counter effect of surfactant concentration on particle size was observed, where increasing the surfactant concentration produced a smaller particle size. High surfactant concentration was illustrated to decrease surface tension, thereby stabilising the surface during homogenisation and preventing particle agglomeration which in turn leads to the production of smaller particle size [60, 61]. Kim et al. [58] and Taha et al. [62] also reported similar findings. Sonication time showed biphasic responses indicating optimal sonication time was required to produce particles of the desired size as illustrated in Fig. 3b and c. The size of particles size decreases as the ultrasonic power increases however, excessive ultrasonic power promotes excessive growth of particles [63, 64].

The PDI of the prepared NLCs varied from 0.099(SA 13) to 0.298(SA 15) as shown in Table 3. PDI gives information about the uniformity of the prepared nanoparticles. PDI values range from 0 to 1, 0 representing a perfectly homogenous system whereas 1 indicates a highly polydisperse system [65]. The low PDI values (< 0.3) confirm the uniform distribution of particle size in the NLCs formulation. Values closer to 0 assure the homogeneity of the formulations. The ANOVA test using a quadratic model revealed that the model was insignificant (p = 0.1484) with adequate precision of 5.9882. From the equation, there is a trend towards increased PDI with increasing total lipid amount and decreased PDI with increasing surfactant concentration and sonication time (Fig. 5).Fig. 5

3D response surface plots illustrating the impact of independent factors on polydispersity index (PDI)

A quantitative comparison between predicted and actual values for Y1 and Y2 is illustrated by linear correlation plots with R² of 0.9941 and 0.5535 respectively (Figs. 6A and 5B). Moreover, the reliability of dependent variables was tested using a residual plot between the run number and the residuals in Fig. 6a and b, respectively. All the data points lay within a 95% confidence interval as illustrated by the vertical spread of the studentized residuals from bottom-to-top, implying that (Fig. 7A, B).Fig. 6

Correlation plot between actual and predicted a particle size (PS) and b polydispersity index (PDI)

From the screening of mucoadhesive polymers including PVA, HPMC, and HEC either alone or in combination, HEC-based gel resulted in smooth and flexible printed films upon physical examination and visual inspection (Fig. 9). HEC-based gel was also reported to exhibit excellent printability and mucoadhesion [71]. A plain film with a 100% infill pattern was printed showed no sign of drug crystallisation.Fig. 9

3D printed film (20*20 mm²) a wet film and b dry film

The thickness of the film was 0.284 ± 0.009 mm, which lies in the ideal thickness range for buccal films [72]. Suitable thickness aids comfortable application of the film and determines the quantity of drugs [73]. The average weight of the film (0.64 cm²) was 0.14 ± 0.008 g. The weight variability was low as illustrated by low standard variation.

Recovery of particle size of nanocarriers after disintegration of the 3D printed film in water was determined using zetasizer. The particle size of the redispersed system was 183.7(PDI = 0.3). The result revealed an increment in particle size of the redispersed system compared to the CBD-NLCs before being dispersed in the polymer solution. This could be due to protective layer formed around the lipid nanoparticle by polymers. Freitas and colleagues illustrated that carbohydrates can form a thick protective layer around the lipid nanoparticles which protects them against the mechanical stress and heat stress during spray drying. They showed that higher concentration of carbohydrate such as mannitol resulted in an increased particle size upon redispersion in water [74].

The mechanical properties of the films were evaluated to ensure handling without breaking. Tensile strength refers to the traction that can be applied before the film breaks while elongation helps to assess the brittleness of the films [43]. Percent elongation and tensile strength of the drug-loaded 3D printed films were determined using a texture analyzer. The tensile strength and percent elongation were 0.67 ± 0.04 MPa and 9.2 ± 1.5% (Fig. 10). The tensile strength and elongation break of blank film was 1.26 ± 0.43 MPa and 4.4 ± 0.7% respectively. The introduction of CBD, which is lipophilic in nature, might have contributed to the decrease strength and improved elongation.Fig. 10

Tensile strength and Elongation break (%) of CBD-NLCs 3D printed films as box plots (n = 3)

The result of mucoadhesion test using porcine buccal mucosa revealed that both blank and CBD loaded films have mucoadhesion capacity. The calculated DFmax for blank film and CBD loaded was 0.16 ± 0.03 N and 0.14 ± 0.02 N respectively. The slightly lower mucoadhesion of the CBD loaded film could be related to the oily nature of CBD. Previous studies have shown that HEC has strong mucoadhesive property [75].

The pH of the printed film was 5.9 ± 0.06. Oral films should have a neutral pH or close to a pH value of 7. This is important to avoid irritation to the oral mucosa. Of note, it might also affect the dispersion, taste as well as the release of the drugs [52, 73].

The drug content for the printed film (20*20 mm²) was determined after dissolving the films in a 10 mL of hydro-alcoholic solution (50:50 v/v) maintained at 37 ^OC for 1 h. The sample was centrifuged as described above ("Drug loading") and supernatant was collected and analysed by HPLC ("NLCs preparation and optimization" section). The achieved drug content was 0.4 ± 0.03 mg for the 20*20 mm². The dose can easily be tailored to cater for individual patient’s requirements by changing thickness and size of the film.

The study assessed potential interactions between CBD and the components of a film using infrared analysis (Fig. 11). The CBD spectrum displayed distinct bands, with the highest at 3519.30 cm⁻¹ and 3406.56 cm⁻¹ due to O–H stretching, bands in the range of 3100–2600 cm⁻¹ caused by symmetric and asymmetric C–H stretching, two bands at 1622 and 1581 related to C = C stretching, and bands at 1442 cm⁻¹ (C–H bending) and 1213 cm⁻¹ (C–O stretching) [76]. In HEC, a characteristic peak for the stretching vibrations of saturated C–H was observed at 3371 cm⁻¹ and 2888 cm⁻¹, while the band at 1061 cm⁻¹ was due to the stretching vibration of ether (C–O) [77]. The results of the FTIR analysis on the physical mixture of the film components showed peaks corresponding to Precirol^® ATO 5. The peaks of CBD were possibly concealed by the peaks of the polymer and Precirol^® ATO 5. The FT-IR spectra of the blank and drug-loaded film were similar, indicating that CBD was successfully incorporated into the film polymers without any interaction. Of note, this technique is not robust enough to prove encapsulation due to overlap of peaks.Fig. 11

FTIR spectra of blank film, drug-loaded film, and their components

The crystallinity of CBD could affect both drug encapsulation and its release from the film. The DSC was evaluated for the film and NLCs components separately as illustrated in Fig. 12. A melting point peak was observed at 66.9 °C for pure CBD as reported in the literature [78] (Fig. 12a, b). No melting peak of CBD was observed for both CBD-NLCs and CBD film showing that either CBD has dissolved or in amorphous state. Similar finding was reported by Morakul and collegues [79]. Furthermore, the endothermic peak of Precirol^® ATO 5 was observed at 57.2 °C in lipid mix and pure Precirol^® ATO 5 [31, 80]. The peak of the lipid mix was less sharp than the solid lipid (Precirol^® ATO 5) which could be due to reduced crystallinity when melted with liquid oil (Fig. 13b). The glass transition temperature for pure HEC was observed around 143 °C (Fig. 13a) consistent with previous reports [81, 82]. No peak of CBD was observed in drug-loaded film as well as CBD-NLCs confirming the drug was not in crystalline state anymore.Fig. 12

DSC thermograms of CBD film and CBD-NLCs and their components

Results from the SEM showed that pure CBD appeared as an irregular shaped crystals similar to a previous report [83]. The surface morphology of both blank films (Fig. 13a) and drug-loaded (Fig. 13b) was smooth indicating the drug was evenly distributed through the system. Nevertheless, the surface of CBD film was smoother compared to the blank which might be due to the surface being packed with tiny particles of the CBD.

The release of the drug from the films was evaluated in a falcon tube filled with 10 mL artificial saliva adjusted to 37 °C for 6 h, as illustrated by our previous work [54]. The result of the in vitro release of CBD film is shown in Fig. 14. The release profile showed a slow and sustained release of CBD from the film (84. 11 ± 7.02% in 6 h). Similar release pattern was reported for Dexibuprofen-Loaded Nanostructured Lipid Carriers [84]. The release of drug particles on the surface of the NLCs during the first hours might have contributed to the relatively faster drug release in the initial phase [85, 86]. Several factors including production temperature, type and concentration of the emulsifier, the production techniques, and partition coefficient of drugs have been shown to affect drug release from the NLCs. Furthermore, the composition of the dosage form including polymer degradation and diffusion of the drug from the matrix governs drug release [87, 88]. HEC which is a controlled-release polymer might have also contributed to the slow and sustained release of the drug [89].Fig. 14

Cannabidiol (CBD) release profiles from CBD-NLC film (n = 3)

The mechanism of drug release from the buccal films was determined by fitting the release data into several release kinetics models including zero-order, first-order, Weibull, Hixson-Crowell, Korsmeyer-Peppas, and Higuchi models. The release kinetics parameters and regression coefficients were calculated, and the Weibull model was the best fit for the data, with an adjusted R2 value of 0.9984 (Fig. 14). This model is an empirical and generalized form of the exponential function and often used to describe drug release from nanoparticles [90, 91]. The Weibull model is expressed as follows.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F= 100{\left.\left(1- e-\frac{(t-Ti}{\boldsymbol{\alpha }}\right.\right)}^{\beta }$$\end{document}F=1001-e-(t-Tiαβwhere F represents the fraction (%) of drug released in time t, Ti represents the lag time before the start of the dissolution or release process, which is usually near zero, β is the shape parameter that characterizes the curve, and α is the scale parameter that defines the time scale of the process [55].

The drug release from the printed film has a β of 0.925, indicating a combined mechanism of Case II transport and Fickian diffusion. Values of β less than 0.75 indicate Fickian diffusion, while values between 0.75 and 1 illustrate a combined mechanism, and values higher than 1 indicate a complex release mechanism [92]. From the result, the release of CBD from the 3D printed film follows a combination of Case II transport and Fickian diffusion (Fig. 15).Fig. 15

A release data of 3D printed CBD film a and release data fitted to Weibull model (b)