Introduction

First applied for the fabrication of pharmaceutical products in 19961, research activities on 3D-printing have rapidly grown in the pharmaceutical field over the past decade2,3. Benefits of 3D-printing that are beyond the capabilities of traditionally applied manufacturing strategies are highly versatile including (i) the creation of complex geometrical structures (e.g., Eder et al.4), (ii) the corresponding provision of sophisticated controlled drug release patterns (e.g., Obeid et al.5), and (iii) patient-centric product design that considers preferences, limitations, and lifestyles of certain populations (e.g., pediatrics6). More recently, 3D-printing was shown to aid early-stage drug development through simplification of formulation development and the facilitation of early clinical trials7. Although various 3D-printing technologies are available, fused filament fabrication (FFF) – often referred to as fused deposition modeling (FDM™) – is the most commonly explored technique, given the high number of publications2. FFF provides several benefits for pharmaceutical products including inexpensive equipment and the absence of (organic) solvents which are commonly used in alternative 3D-printing methods such as semi-solid extrusion8,9.

Still, FFF is associated with several challenges due to high printing temperatures and the requirement of filaments with a perfectly spherical cross-section, a precise target diameter, as well as suitable mechanical and rheological properties10– all eventually limiting the material options. In other words, the requirements of FFF regarding material, process, and machine-specific parameters may be contradictory to the inviolable requirements of pharmaceutical products. Already the first step, which is the fabrication of a filament via hot-melt extrusion (HME), may be challenging due to (i) rheological properties making HME challenging11, (ii) poor powder flowability resulting in inconsistent material feeding into the extruder12, (iii) irregular filament shape and diameter due to die swell13and/or filament deformation upon cooling14or upon storage15, and/or high filament brittleness16,17,18impeding filament feeding into the 3D-printer. Besides these material-related challenges, a key limitation of filament fabrication is the comparatively high material need for the HME process. Typically, several hundred grams are required, given that process setup activities are inevitable for fabricating a high-quality filament14. Only a few publications have demonstrated the processing of smaller amounts of material (< 10 g) into filaments using small-scale extrusion setups19,20, which again multiplies the amount of required material due to process setup necessities. These high material amounts limit the applicability of FFF, especially during early-stage product development, where often only small material amounts are available for the exploration of different formulation strategies and printing feasibility tests.

In an attempt to considerably reduce the efforts imposed by HME, we present an alternative melt-based technology for filament fabrication that requires only three grams of material. This is especially useful for small-scale feasibility studies as well as material and formulation screening for FFF. We aim to produce filaments from materials that are not easily processible via HME but might exhibit material properties advantageous for 3D-printing, thereby expanding the material portfolio for FFF. Additionally, we demonstrate the utilization of this technology for manufacturing filaments containing a highly-potent model active pharmaceutical ingredient (API; i.e., progesterone). Highly elastic ethylene vinyl acetate (EVA) polymers with vinyl acetate (VA) content of 28 wt% (EVA28) and extremely low melt viscosities were processed in the present case. Low melt viscosities impose hurdles to HME processes using conventional process setups but are highly beneficial for FFF as they facilitate a constant melt flow from the nozzle. The filaments served to 3D-print subcutaneous implants with the outer geometry resembling the marketed EVA28-based product Nexplanon®. For the first time, EVA28 was printed using small nozzle sizes of 250 μm to create complex infill geometries of these small cylinders (i.e., 2 mm diameter and 40 mm length), which is the basis for subsequent personalization strategies.

Materials and methods

Materials

Ethylene-vinyl acetate polymers (EVAs) with a vinyl acetate (VA) content of 28 wt% and a melt flow rate (MFR; determined according to ASTM D 1238 at 190 °C and 2.16 kg load) of 150 g/10 min (Ateva 2830 A; EVA28/150) and of 400 g/10 min (Ateva 2842 A; EVA28/400) were kindly donated as powders by Celanese Corp. (Irving, TX, USA). EVA with a VA content of 28 wt% and an MFR of 25 g/10 min (Greenflex XS70; EVA28/25; Eni Versalis, San Donato, Milanese, Italy) that was previously used for filament fabrication via HME21 served as a reference. Micronized progesterone (P4) was used as received from Euro OTC & Audor Pharma GmbH (Bönen, Germany). As the first layer adhesive aid during 3D-printing, the black rubber mat P-surface 141 (PP-Print GmbH, Bayreuth, Germany) was applied.

Material characterization

The thermal behavior of the EVAs, P4, binary EVA/P4 powder blends, and P4-loaded filaments was assessed via differential scanning calorimetry (DSC) using the 204 F1 Phoenix (Netzsch GmbH, Selb, Germany) equipped with an automated sampling unit and an external cooling device. Samples (4–7 mg) were accurately weighed and placed into aluminum pans that were subsequently closed with pierced lids via cold welding. The pure polymer samples were heated from − 60 to 195 °C and the P4 containing samples from 20 °C to 195 °C with a heating rate of 10 K min−1 using nitrogen (flow rate 250 mL min−1) as purge gas. Every sample was investigated in triplicates, and the thermograms were evaluated using the Netzsch Proteus Analysis software. The total crystallinity of the EVAs was determined from the thermograms based on the enthalpy of fusion of a perfect polyethylene crystal (i.e., 277.1 J g-122).

The viscosities of the EVA polymers and EVA/P4 blends (16 wt% P4) were determined via the rheometer Physica MCR 302 (Anton Paar GmbH, Graz, Austria). To obtain a homogeneous polymer sample, disks of 25 mm diameter and 1.1 mm thickness were produced via vacuum compression molding (VCM, MeltPrep GmbH, Graz, Austria) at temperatures of 85 °C and a heating time of 6 min for the EVAs. For the P4-loaded samples, a powder blend of EVA and P4 (16 wt%) was cryo-milled (SPEX SamplePrep 6875 Freezer/Mill®, SamplePrep, Metuchen, USA; using Liquid Nitrogen as cooling agent) and processed into homogeneous disks via VCM at a temperature of 140 °C (EVA28/400 and EVA28/150) or at 180 °C (EVA28/25) for 5 min. The rheology measurements were performed in parallel plate geometry at a gap distance of 1 mm. Each sample’s viscosity was determined in its linear viscoelastic region at a fixed frequency of 100 rad s−1 and temperatures ranging from 80 to 195 °C with a heating rate of 2 °C min−1. Each drug-free and drug-loaded EVA grade was investigated in triplicates.

Filament fabrication via hot melt extrusion

As a reference manufacturing method for filament fabrication, a standard horizontal hot melt extrusion setup, including the twin-screw extruder ZE 9 HMI (Three-Tec GmbH, Seon, Switzerland) with three heating zones and co-rotating kneading screws (20 L/D) was used. The polymer powder was used as received and fed to the extruder using the micro feeder DDW-MDO-MT-1-HYD (Brabender GmbH & Co. KG, Duisburg, Germany). A conveyor belt (GES – 25, Dorner GmbH, Jülich, Germany) combined with a pressurized air-cooling tunnel was used for haul-off, and the filament diameter was determined inline using a triple-axis laser diameter measurement device (USYS2000, Zumbach Electronic AG, Orpund, Switzerland).

Novel filament fabrication technique

Filaments were fabricated using the VCM Filament Molder (MeltPrep GmbH, Graz, Austria), which is a novel tool based on the VCM technology. It consists of a specially designed mold, which goes through the processing steps on the VCM platform:

In the first step, cylindrical polymer and polymer/P4 disks were produced via melting and compaction of the polymer inside a VCM Disc Tool D25, as described previously23. For manufacturing of the drug-loaded filaments, the VCM disks were produced from powder blends of EVA and P4 (16 w% P4 loading), which were previously cryomilled (SPEX SamplePrep 6875 Freezer/Mill®, SamplePrep, Metuchen, USA) to ensure homogeneity of the mixtures. The chamber was filled with 3 g of sample and closed with a piston, which is connected to a vacuum unit. The used process parameters (i.e., temperature and time) for the disk preparation are described in Table 1.

Table 1 Parameters for the VCM-disk and filament production.
Full size table

In the second step, the polymer disk undergoes processing (Fig. 1) to form the filament: The base plate of the VCM Filament Molder has a spiral channel to accommodate a PTFE tube with the inner diameter of that of the filament. The PTFE tube passes through the tube clamp into the reservoir chamber of the molder, which holds the premanufactured polymer disks. The reservoir is lined with PTFE foils in a similar fashion to the VCM Disc Tools, but the lower front foils contain a puncture hole to allow flow from the reservoir into the tube. Finally, the polymer disk is placed into the VCM Filament Molder chamber. Before being placed on the heating unit, the VCM Filament Molder is connected to the vacuum unit, and all the air is removed. The sample is then heated for a specific temperature and time, during which the polymer melt flows from the reservoir chamber into the PTFE tubing. Similar to the VCM disk production, the process is driven by uniform pressure provided by a movable lid on top of the VCM Filament Molder.

Fig. 1
figure 1

VCM filament molder explosion view.

Full size image

The processing temperature (90–195 °C) and time (3–30 min) were optimized with the aim of fully filling the PTFE tube with polymer (Table 1). Thereafter, the chamber was cooled under vacuum, and finally, the tube was removed by cutting it open along the length using a customized tool, ensuring no filament damage. In the current setup, the produced filaments exhibit a maximum length of 100 cm, dictated by the tube length. Importantly, the VCM disk mass must slightly exceed the filament mass to ensure proper tube filling. The remaining disk can be re-used as several lower-weight disks can be stacked vertically to yield an overall disk weight of 3 g.

Filament characterization

The filament diameter and ovality were determined offline using the three-axis laser measurement head ODOC 13TRIO and the corresponding processing unit USYS 200 (Zumbach Electronic AG, Orpund, Switzerland). For each EVA grade (drug-free and P4-loaded), three filaments were evaluated, and the average diameter and ovality were determined.

The filaments’ tensile properties were determined to assess their printability. Therefore, the Zwick Z001 universal testing machine (Zwick GmbH & Co. KG, Germany) with a 1 kN load cell at a gauge length of 50 mm was used. The Young’s modulus was evaluated between 0.05% and 5% strain with the strain rate set at 50 mm·min−1 until filament rupture, as this represents a realistic pulling speed during 3D-printing (assuming a printing speed of around 15 mm·s−1). Each sample was evaluated in five independent measurements.

To evaluate the API content over the length of the filaments, the filaments were cut into five equally sized pieces (about 0.5 g each). The pieces were cryomilled individually using a Retsch cryomill (Retsch Technology GmbH, Haan, Germany), and 100 mg of the milled samples were treated with 30 ml Methanol/THF 8:2 as the extraction medium. The samples were incubated at 55 °C at 300 rpm in an orbital shaker, and after 3 h, the samples were diluted with water to 100 ml total volume and filtered using a 0.2 μm MCE filter. The P4 content in the samples was determined via HPLC (Waters Acquity® H class system; Waters GmbH, Eschborn, Germany) utilizing a Merck Purospher STAR RP-18 end-capped column (2 μm, 2.1 mm X 100 mm; Merck KgaA, Darmstadt, Germany) at a column temperature of 50 °C. As a mobile phase, ethanol/water 11/9 v/v was used at a flow rate of 0.25 ml·min−1. An injection volume of 1 µl served for the P4 content determination at a wavelength of 244 nm. The linearity of the method was proven from 5 ppm to 200 ppm (R2 = 0.999).

3D-printing of implants

Cylindrical implants of the Nexplanon® geometry, i.e., 2 mm diameter and 40 mm length with geometry details provided elsewhere24, were constructed using the software AutoCAD, sliced in the software Simplify3D, and fabricated using a filament-based 3D-printer customized for flexible materials (HAGE3D medmex, HAGE3D, Graz, Austria)21. The 3D-printing parameters were investigated based on our previous work exploring the printing of EVA28/25 filaments21 together with the thermal, rheological, and mechanical properties of EVA28/150 and EVA28/400. The maximum temperature was 195 °C (considering the degradation temperature of semi-crystalline EVAs25 and progesterone26) and was lowered by 5 °C increments until the optimum temperature was reached. Nozzle sizes of 0.4 (commonly used), 0.3, 0.25, and 0.2 mm were tested, aiming at using the lowest nozzle size possible to allow complex infill patterns that are essential for a subsequent personalization of drug release. A print speed of 10 mm s−1 was applied and increased by increments of 10 mm s−1 until the highest speed possible was reached. The layer height was kept constant at 30% of the nozzle size (i.e., 75 μm for a nozzle size of 250 μm), and cooling was applied using the integrated fan at 50% power from the third layer onwards. To improve first-layer adhesion, a black rubber mat was used, and the build platform temperature was kept at ambient temperature. The 3D-printing parameters were determined using a rectilinear infill at 100% infill density. Thereafter, different infill patterns (honeycomb, wiggle) and infill densities (25 and 50%) were printed. The print process was optimized for each EVA grade based on the drug-free filaments and adapted where necessary (i.e., temperature and print speed) for the drug-loaded filaments.

Implant characterization via µCT

µCT measurements visualizing the implants’ internal structure were performed on the RX Solutions Easytom 160 (RX Solutions, Chavanod, France) equipped with the 160 kV Hamamatsu Nano focus tube, a LaB6 filament, and a 1920 pixel × 1536 pixel Varian flat panel detector. A voltage of 90 kV, a current of 63 µA, an integration time of 167 ms, and 1088 projections were used, employing stack mode. A voxel size of (8 μm)³ was achieved. Volume data evaluation was performed using a filtered back projection algorithm using the xact64 reconstruction software (RX Solutions, Chavanod, France). The image processing and data evaluation were performed using the VGStudio Max 2023.2 (Volume Graphics) software.

Results and discussion

Polymer characterization

As expected, the MFR did not affect the thermograms. Hence, the thermograms of EVA28/150 and EVA28/400 were similar to those of EVA28/25 (details are provided in Eder et al.21) (Fig. 2a). All EVAs exhibited a glass transition around − 30 °C and a broad melting event (30–85 °C) with two peaks around 50 and 70 °C. The lower temperature peak occurred due to the melting of imperfect smaller crystallites, whereas the higher temperature peak was due to the melting of more organized polyethylene chains27,28. As previously reported for EVA2829,30, the first peak was more pronounced. Compared to EVA28/150 and EVA28/400, the second peak of EVA28/25 is slightly more pronounced and shifted toward a higher temperature, which can be attributed to differences in the production process, given the fact that this grade was provided by a different supplier. The degree of crystallinity slightly decreased with increasing MFR and was 18 ± 0.5%21, 17 ± 0.4%, and 16 ± 0.1% for EVA28/25, EVA28/150 and EVA28/400, respectively. DSC measurement of filaments after the incorporation of P4 leads to an additional thermal event at around 120 °C and 129 °C. The two distinct peaks can be attributed to the melting peaks of progesterone’s monotropically related crystalline forms II (119.7–122.0 °C) and I (126.2–129.0 °C), respectively, with the crystalline form I exhibiting higher thermodynamic stability than the metastable form II. Both crystalline forms are listed in the European and United States Pharmacopeia30,31,32. The melting peaks of both crystal structures are overlapping, and the signal shows an onset ranging from 100 to 115 °C suggesting the presence of additional lower melting polymorphic forms (i.e., III, IV, and V) of P430,32,33. Furthermore, the incorporation of P4 leads to shifting of the EVA melting peaks toward slightly lower temperatures and lower enthalpy of melting (e.g., from 41.5 ± 1.5 J·g−1 for EVA28/400 to 26.4 ± 1.2 J·g−1for EVA28/400 + P4). This is due to the disruption of the EVA polymer’s crystalline lattice after embedding P4 crystals30,34.

Fig. 2
figure 2

Thermal and rheological analysis of raw materials: (a) DSC thermograms of EVA28 with different MFRs (25, 150 and 400) with and without P4; (b) representative rheology curves of the respective formulations.

Full size image

The main purpose of the rheological characterization was two-fold: (i) to rationally select the 3D-printing temperatures, and (ii) to correlate the polymers’ viscosity to the preferred processing parameters of the VCM Filament Molder, i.e., time and temperature, which has not yet been reported. As expected, the viscosities decrease with increasing temperature and MFR (Fig. 2b). The highest viscosities in the considered temperature range for EVA28/150 and EVA28/400 were 762 and 426 Pa s (both at 80 °C), respectively, which is clearly below the range preferable for HME (i.e., between 1 000 and 10 000 Pa s11,35). By contrast, EVA28/25 showed considerably higher viscosities with the highest value being 1 970 Pa s (at 80 °C) making it easily processable via HME. For 3D-printing, a viscosity of 100 Pa s is favorable36, which cannot be met for EVA28/25 in the tested temperature range. Nevertheless, we previously showed that using our customized 3D-printer, EVA28/25 can be processed at 195 °C corresponding to a viscosity of 235 Pa s with a minimum nozzle diameter of 0.8 mm21. Lower nozzle diameters lead to an elevated nozzle counter-pressure; a challenge that is typically addressed through the reduction of the melt viscosity (e.g., through an increase of the printing temperature)37. Generally, an increase in the printing temperature poses the risk of degradation of the EVA polymer and relevant APIs. Compared to EVA28/25, EVA28/150 and EVA28/400 exhibit overall much lower melt-viscosities at 195 °C (251 Pa s, 57 Pa s, and 21 Pa s, respectively) suggesting that FFF of these grades can be performed using nozzle diameters smaller than 0.8 mm. In all cases, the addition of P4 leads to a reduction in melt viscosity which is in agreement with previous studies (Fig. 2b)21. This effect is beneficial in the case of 3D-printing, as the same melt viscosities can be achieved at lower temperatures. This allows the reduction of the print temperature and reduces the thermal stress on the API and polymer.

Filament fabrication via the VCM filament molder

The choice of VCM Filament Molder parameters demonstrates a trade-off of process time versus process temperature. Since the filling of the PTFE tube follows the Hagen-Poiseuille law, the melt viscosity has a linear influence on the volumetric flow rate and the filling time. The use of a low processing temperature leads to a high melt viscosity meaning a high processing time and vice versa. In the present case, 195 °C, which is just below P4’s degradation temperature of 200 °C26, represents the maximum operating temperature. As a proof of concept, for the drug-free filaments, both cases of process optimization were studied: one to yield the lowest processing temperature (i.e., the highest reasonable processing time of 30 min) and the other to achieve the lowest processing time (i.e., the highest feasible processing temperature of 195 °C) (Table 1). Reducing the processing time while keeping the process temperature constant at the lowest value possible resulted in non-fully filled tubes – in other words, not one continuous filament was obtained but several shorter filament pieces or pieces with vacuoles were obtained (Fig. 3a). Vacuoles form if shrinkage during cooling cannot be compensated by compaction pressure. In the case of a partially filled tube, a pressure gradient within the channel results from the flow generating friction. Once the tube is full, isobaric conditions are achieved, and shrinkage can be compensated, yielding a vacuole-free filament. For the drug-loaded filaments, the process was optimized to yield low processing times while achieving a fully filled tube. To this end, the temperature was kept constant at 195 °C, and the processing time was adjusted for each grade to achieve full filling of the tubes. The inspection window to the last section of the PTFE tube and the movement of the high-pressure lid are helpful in finding the right processing times.

Fig. 3
figure 3

Filaments manufactured using the VCM filament molder: (a) filament with vacuoles, (b) drug-free filaments from EVA28/25, EVA28/150, and EVA28/400, and (c) drug-loaded filaments (16 w% P4) from EVA 28/25, EVA28/150 and EVA28/400 with optimized VCM filament molder process settings.

Full size image

For EVA28/150, drug-free filaments with a smooth surface and without any vacuoles – considered high-quality filaments – were successfully manufactured via the VCM Filament Molder at a minimum temperature of 120 °C (corresponding to a melt viscosity of 250 Pa s) and a heating time of 30 min (Fig. 3b). Logically, a further reduction in processing temperature is possible, however, at the expense of higher processing times. On the contrary, increasing the temperature to a maximum of 195 °C leads to a reduction in process time by 70% (i.e., to 8:45 min). For EVA28/400, the lowest temperature was 90 °C (corresponding to a melt viscosity of 285 Pa s) while keeping the process time constant at 30 min (Fig. 3b). It is worth highlighting that the melting of this polymer is only finished at 80 °C (Fig. 2a), which does, in general, not allow for considerably lower processing temperatures. Similarly, the increase of the processing temperature to 195 °C resulted in a 85% decreased processing time (i.e., to 04:30 min). For EVA28/25, which is used as a reference, the lowest temperature, which produced high-quality filaments, was 195 °C (corresponding to a viscosity of 235 Pa s), yielding a processing time of 30 min. In this case, increasing the process temperature to decrease the process time was not possible due to the risk of degradation.

Based on the results generated on the EVA polymers, all P4-loaded filaments were manufactured to minimize the processing times. Therefore, a processing temperature of 195 °C was chosen (Fig. 3c). Compared to the corresponding drug-free filaments, the processing time was substantially reduced (i.e., to 3 min, 4:15 min, and 15 min) for EVA28/400, EVA28/150, and EVA28/25, respectively. This can be attributed to the decrease in melt viscosity upon the addition of P4, leading to faster filling of the VCM Filament Molder tube.

Overall, these findings suggest that a broad range of viscosity profiles can be processed through the adaption of the process time. Using only 3 g of powder, a ready-to-use filament can be fabricated within processing times as low as 3 min. In an attempt to reduce the potential risk of thermal degradation, processing times increased to 30 min when using temperatures that correspond to a melt viscosity of around 250 Pa s, which is still a reasonably low processing time. In a nutshell, applying high temperatures decreases the melt viscosity and, correspondingly, the process times but may increase the risk of thermal degradation, which is especially crucial in the case of thermosensitive APIs. Vice versa, the temperature and corresponding degradation risk may be decreased at the expense of increased process times. Thereby, the VCM Filament Molder offers the great advantage that the processing temperature is not restricted by the manufacturing technology – as it is the case for HME – but can be adapted toward the requirements of the formulation and the 3D-printing process. Typically, the requirements for rheology profiles for HME and 3D-printing are fundamentally different. As a result, either filament manufacturing via HME is challenging due to low material viscosity (resulting in e.g., sticking of the filament to the conveyor belt or high ovality) or high printing temperatures are necessary. As presented here, low-viscosity EVA grades, which are hardly processible via HME using conventional setups but are beneficial for 3D-printing due to lower printing temperatures and/or smaller nozzle sizes, can be easily fabricated into high-quality filaments.

Hot melt extrusion, typically used for filament fabrication was used as a reference process to compare the quality of the filaments produced via the VCM Filament Molder to a standard process. The drug-free EVA grades were processed in a set-up well-established for filament production4,21,24,38. The processing temperature was set to 80 °C for both EVA28/400 and EVA28/150, representing the lowest possible temperature guaranteeing complete melting of the polymer (Fig. 2a). Despite their low viscosities at 80 °C (i.e., 700 Pa s and 425 Pa s for EVA28/150 and EVA28/400, respectively) being clearly below values recommended for HME, the strands exhibit no susceptibility to sticking to the conveyor belt after leaving the extruder die. This is due to the high melt stability of semicrystalline EVA polymers, which is attributed to their long-chain branching.

Filament characterization

As expected from the VCM filament molder’s tube geometry, the diameter and ovality of all molded filaments were close to the target values of 1.75 mm and 0 mm (Table 2), respectively. The PTFE tubing confined in the spiral channel combined with the applied pressure allow for high-diameter consistency. In contrast, despite achieving the target diameter of 1.75 mm, the hot melt extruded strands showed significantly higher ovalities since extrudate deformations were induced due to a structural collapse of the strand on the conveyor belt caused by the material’s low viscosity and low melting onset of around 35 °C (Fig. 2). For the used 3D-printer, the lowest possible filament diameter to ensure the grip of the transport wheels is 1.5 mm, while 1.85 mm represents the maximum feasible diameter that can be fed through the feeding tube. Consequently, the filament ovality must not exceed 0.35 mm as the diameter at the thinnest point might be below 1.5 mm, preventing the transport wheels from gripping the filament effectively while simultaneously exceeding 1.85 mm at the thickest point. Both extruded EVA28/400 and EVA28/150 locally exceed ovalities of 0.35 mm, excluding them from further 3D-printing studies and filament characterization (i.e., tensile tests). Naturally, the ovality should be as low as possible (ideally 0 mm) since high ovality can result in inconsistent material flow and, in cases of extreme ovality, cause the filament to get stuck in the feeding tube of the 3D-printer.

Table 2 Relevant properties of the filaments manufactured via the VCM Filament Molder.
Full size table

For all tested filaments, the strain at break was sufficiently high to allow spooling. The extruded filament of EVA28/25 (i.e., the only extruded EVA filament having proper diameter and sphericity) is shown for comparison reasons with the strain at break and the Young’s modulus being affected by the manufacturing technique. The significantly lower melt viscosity during filament production using the VCM Filament Molder (i.e., around 250 Pa s), compared to HME (i.e., around 1 000 Pa s) promotes increased polymer chain mobility. This results in enhanced polymer chain alignment as the polymer melt flows into the mold39. The chain alignment promotes higher deformation before reaching fracture, increasing the strain at break. Additionally, during HME, the filament undergoes comparatively rapid cooling21compared to slow cooling within the metal mold over the course of several minutes in the VCM Filament Molder process. This extended cooling time is likely to result in the formation of larger crystalline spherulites40, which contributes to a higher stiffness and, consequently, a higher modulus of the filament fabricated via the VCM Filament Molder.

Comparing the different EVA grade filaments made using the VCM Filament Molder, the strain at break was similar for EVA28/25 and EVA28/150, whereas it was considerably lower for EVA28/400. The strain at break is a function of both molecular weight and processing history/molecular orientation. Thus, identifying the key parameters causing different values may be challenging. In general, the strains at break values are very high for all samples, suggesting there is very little orientation in the filaments. The Young’s modulus increased with decreasing MFR since a reduction in zero viscosity implies a reduction in mean molecular weight, which decreases the Young’s modulus41. A lower Young’s modulus indicates that the counter-pressure in the nozzle may not be overcome, and consequently, filament jams and buckling are more pronounced. However, the lower Young’s modulus may be counterbalanced by the reduction in melt viscosity or lower material flow rates.

Since the solubility of P4 in EVA28 is around 5 wt% after thermal treatment30, the majority of the P4 in the filaments was present as crystals forming a solid crystalline suspension. Thus, the drug-loaded filaments exhibit higher stiffness (i.e., higher Young’s modulus) and reduced strain at break compared to their drug-free counterparts. This is anticipated to aid the 3D-printing process, as during 3D-printing, the filament serves as a piston, pushing the molten material in the hot end out of the print nozzle. Higher filament stiffness improves the piston effect for the same melt viscosity, as the filament is less prone to buckling.

Molded filament content uniformity

In HME, shear forces and mixing through the extruder screws play an important role in the homogenization of API and polymer – even though frequently, a pre-blend is fed into the extruder. By contrast, the VCM Filament Molder does not involve significant mechanical mixing. Therefore, proper blending before filament fabrication is crucial in ensuring homogeneous API distribution throughout the entire filament. In other words, the API content must not vary along the filament, as this would result in printed products with different API content. The visual appearance of the produced filament is uniform indicating a homogeneous solid dispersion. The results of the content determination demonstrate that cryomilling EVA together with P4 powder results in a uniform P4 content along the filament (i.e., low standard deviations among different filament segments). In any case, the average P4 content was close to the anticipated loading of 16 wt% over the entire length of the filaments (Table 3).

Table 3 Content uniformity of progesterone-loaded molded filaments (n = 5, sampling mass 0.5 g each).
Full size table

3D-printing of implants

Highly elastic polymers such as EVA28 have been used in several marketed pharmaceutical products (e.g., subcutaneous implants (Nexplanon®), intravaginal rings (Nuvaring® and generic products)) for decades due to its biocompatibility and flexibility avoiding tissue irritation42. While printing using a 0.8 mm nozzle has recently been demonstrated as effective for large vaginal inserts21, challenges arise for printing smaller dosage forms, limiting the possibility of advanced internal structures or defined internal porosities24. Nevertheless, smaller nozzle diameters are associated with a higher counterpressure which may lead to buckling of highly flexible EVA filaments as the filament acts as a piston, pushing the polymer melt through the print nozzle. At a given nozzle geometry, the nozzle counterpressure can effectively be reduced through (i) an increase in the filament stiffness (e.g., by using EVA with a lower VA content resulting in higher crystallinity43), (ii) a decrease in the melt viscosity, or (iii) a decrease in the material flow rate. Stiffer EVA grades with a lower VA content may concomitantly reduce the flexibility of the printed device, having a negative impact on the wearing comfort. Notably, the melt viscosity can be effectively decreased through an increase in print temperature which, however, holds the risk of thermal degradation24. Alternatively, the reduction in the polymer chain length, commonly represented by an increased MFR, leads to a lower melt viscosity without changing the processing temperature. In fact, EVA28 is commercially available with MFRs ranging from 3 to 850 g/10 min44. Material flow rates are typically decreased through a decrease of the print speed or a decrease of the print layer height, which results in high process times. This work follows the strategy of applying EVA grades with high MFRs, which neither negatively affect the flexibility of the printed device nor yield excessively high 3D-printing process times.

All optimized 3D-printing parameters elaborated are summarized in Table 4. The print process was initially established for the drug-free implants and then adapted where necessary (i.e., print temperature, print speed, and cooling) for the P4-loaded filaments, accounting for the lower melt viscosity upon the addition of P4. It is evident that both the decrease in nozzle diameter and altered material properties require a thorough adaption of print parameters to maintain flawless print results, which is described in more detail in the following paragraphs.

Table 4 3D-printing parameters for the different EVA grades.
Full size table

The theoretical internal implant structure, according to the slicer software, is depicted in Fig. 4a. Owing to its low melt viscosity, a nozzle size as low as 250 μm was applied to flawlessly print both EVA28/400 and EVA28/150 despite the filaments’ low Young’s moduli (Fig. 4b). This enabled the creation of layers as low as 69.1 ± 7.3 μm in height (determined from the µCT images), which means that per cross-section, a total of 24 layers (compared to 9 layers for a 0.8 mm nozzle) are deposited on top of each other. This clearly provides sufficient space for internal structure design to tailor the implant’s infill and, hence, the subsequent drug release. Figure 4c shows µCT images of the internal implant structure using an infill density of 25% and a rectilinear infill pattern. The images show flawless strand deposition and intentional porosities identical to the theoretical slicer image (Fig. 4a). The lowest nozzle temperatures yielding proper strand deposition were rather high, i.e., 195 °C for EVA28/150 and 185 °C for EVA28/400. These temperatures correspond to low viscosities of 57 and 27 Pa s, respectively. These viscosities are comparatively low for 3D-printing applications, however, they are needed to overcome the increased nozzle counterpressure caused by the small nozzle size in combination with the high filament flexibility.

Fig. 4
figure 4

3D-printed implants: (a) slicer image: theoretical internal implant structure 25% rectilinear infill using a 250 μm print nozzle; (b) macroscopic image of implants (25% rectilinear infill); (c) µCT of implants’ internal structure (25% rectilinear infill).

Full size image

A print speed of 10 mm s−1 was used for EVA28/150 and could be increased up to 30 mm s−1 for EVA28/400, thereby decreasing the overall maximum print time (i.e., for 100% infill) of one implant from 7 min to 3 min. The reduced melt viscosity of EVA28/400 compared to EVA28/150 reduces the nozzle counterpressure enabling elevated print speeds of up to 30 mm s−1. Due to the high flexibility of both filaments, the filaments exhibit a poor piston effect resulting in buckling at higher print speeds (i.e., higher than 10 and 30 mm s−1 for EVA28/150 and EVA28/400, respectively). For both polymers, it was of utmost importance to print the first layer as well as the perimeter at only 50% speed to guarantee proper strand deposition, being the basis for a smooth cylindrical outer shape (Fig. 4b). In case the first layer and the perimeter were printed at 100% speed – which is the common way to print – poor deposition of the outer perimeter was observed around the edges of the implant, leaving behind gaps and a rough non-cylindrical outer shape. For a subcutaneous implant, this is unacceptable for two reasons. First, the implant must fit into the needle of an applicator for insertion into the tissue45. Second, non-smooth, irregular surfaces may be associated with local tissue irritation46 and may affect the drug release rates, which are typically a function of the device’s free surface for EVA-based systems30.

The incorporation of P4 resulted in a lower melt-viscosity (Fig. 2b) enabling the reduction in print temperature from 185 °C to 145 °C for EVA28/400 and from 195 °C to 185 °C for EVA28/150. In addition to reducing print temperature, the print speed of EVA28/150 could be increased from 10 mm s−1 to 20 mm s−1. Notably, print temperature and print speed are typically cohesive, so an increase in print temperature typically enables an increase in print speed. This was demonstrated in the case of EVA28/150 + P4: While printing at a temperature of 195 °C was successful at a print speed of up to 30 mm s−1, reducing the print temperature to 185 °C required a simultaneous reduction in print speed to a maximum of 20 mm s−1 for a successful print job. Independent of the EVA grade, the cooling was increased to 100% of fan speed upon incorporation of P4 to guarantee geometrical integrity and to avoid merging of the complex inner structures. For this reason, the build platform was not heated, while proper object adhesion was guaranteed through an adhesion-enhancing rubber mat. The decrease in nozzle diameter and related increased nozzle counterpressure resulted in underfeeding and visible unintended gaps between the deposited strands. Consequently, the material flow rate was increased to 1.3 for EVA28/150 and EVA28/400.

Despite its high popularity in pharmaceutical research, detailed 3D-printing process descriptions are rare, and process development is rather limited to the application of standard settings. As such, testing of different nozzle sizes has been neglected. Instead, the most common nozzle size of 0.4 mm is usually applied47. Occasionally, larger nozzle sizes are used (e.g., for tablet printing from polyvinyl alcohol5). Most publications subjecting 3D-printing of EVA focus on EVA grades with a VA content lower than 28%. Typically, only ranges in layer height between 0.148,49 and 0.25 mm50,51 are reported, in some cases without specification of the used nozzle size48,49. To the best of the author’s knowledge, there is only one work addressing high-resolution FFF of implants with complex internal structures, which was also provided by our group24 and builds the basis for personalized biodegradable subcutaneous implants. The present work aims at developing a similar strategy for non-biodegradable materials that have been used in marketed products for several years.

It should be noted that the incorporation of a drug typically modifies the filament properties and correspondingly21,50, adaptions of the 3D-printing process are required. Depending on (i) the drug’s thermal stability, (ii) its interactions with the polymeric carrier, and (iii) the anticipated drug loading, the entire FFF process chain may necessitate manifold adaptations. Here, the benefits of the VCM Filament Molder, which does not have any strong requirements regarding the formulation’s rheological properties, become evident: (i) polymers can be selected considering maximum process temperatures applicable to thermolabile drugs (e.g., by selecting an EVA grade with a high MFR), (ii) filament fabrication is expected to be neither limited by drugs that dissolve in the polymer (i.e., show plasticizing effect) nor by drugs that remain crystalline (i.e., show an anti-plasticizing effect) and hence, (iii) a broad range of drug loadings is anticipated to be feasible. These aspects raise the hypothesis that the VCM Filament Molder enables the processing of any kind of formulation and the corresponding fabrication of filaments with highly varying properties. Furthermore, this process is beneficial when handling highly potent APIs with safety concerns (e.g., hormones), as the filament processing can take place entirely within an isolated glovebox. From a cleaning perspective, the VCM Filament Molder also benefits from disposable tubing, which limits potential contamination of the device with hazardous APIs to a minimum.

The use of FFF for 3D-printing EVAs with a high VA content was recently constrained by the lack of high-quality 3D-printing filaments with suitable mechanical and rheological properties49. Consequently, EVA has predominantly been processed via direct powder or pellet extrusion 3D-printing52,53. However, direct powder extrusion presents significant challenges when working with hazardous APIs as it increases the risk of contamination and spillage during the print process. Additionally, issues such as electrostatic charging and poor powder flowability can lead to inhomogeneous material flow and poor print results54. While such feeding problems can be mitigated through direct pellet extrusion 3D-printing, this approach requires a preliminary extrusion and pelletization process that demands large quantities of both API and polymer, making it less efficient and more resource-intensive55. The filament molder as an alternative to HME resolves recent challenges in the extrusion process (i.e., processing of high MFI EVA grades while maintaining high diameter stability and low ovality). This advancement makes these polymer grades accessible to FFF. Compared to direct powder or pellet extrusion 3D-printing, FFF in combination with the filament molder in any case enables homogenous feeding, tremendously reduces the demand for raw material, and allows fast screening of formulations without an elaborated extrusion and pelletization process.

Conclusion

Historically, the choice of polymers for FFF has been restricted to polymers that can also be processed into filaments via HME. In this work, this historic assumption is proven wrong, presenting an alternative filament fabrication method, the so-called VCM Filament Molder, which can be operated almost independently of the material’s melt rheology. As a result, the FFF material portfolio is tremendously extended to materials that are highly relevant to the pharmaceutical industry but were previously not accessible to FFF due to unsuitable rheology profiles for either HME or FFF. Furthermore, the VCM Filament Molder process offers the potential for resource-efficient process and formulation development due to the low material demand of < 4 g highlighting its great potential during early-stage product development.

In this work, low-viscosity EVA28 grades (with and without the model API progesterone) that cannot easily be processed via HME using conventional setups were fabricated into high-quality filaments. The filaments were subsequently used to 3D-print subcutaneous implants in high-resolution, unprecedented for this pharmaceutically highly relevant material. The produced filaments meet or even exceed the quality attributes of filaments produced using a standard extrusion setup (i.e., high diameter stability, low ovality, high content uniformity). This builds the basis to efficiently design 3D-printing processes for EVA28-based formulations to fabricate drug-eluting subcutaneous implants with customized complex internal structures.