3.2.1. Correlation of Material Pressure and Diameter Variations
Filament diameter homogeneity is an important quality attribute to ensure printability and dosing accuracy. It was reported that variations higher than 1.75 ± 0.05 mm are undesired [22
]. Prasad et al. [9
] investigated the HME process for HPMC-based formulation to find an operating window where the screw speed, temperature, die diameter and drug load was assessed related to material pressure, torque and filament diameter. The organoleptic appearance was considered additionally. However, the impact of the different settings on material pressure and diameter variations along with CQAs of printed dosage forms was not assessed. Korte and Quodbach [23
] systematically evaluated the effect of PFR, screw speed and conveyer belt speed on the filament diameter and also on the mass of printed dosage forms. A significant effect of PFR and the conveyer belt speed on the mean diameter of filaments was found. The variation of diameter and its impact on the mass variation of printed dosage forms was not investigated.
Preliminary extrusion experiments showed that fluctuations of the recorded material pressure are reflected in diameter variations of the filaments, as exemplarily depicted in Figure 4
. This has already been demonstrated in the work of Korte [23
]. Pressure fluctuations occur because of material accumulations in front of KB and the die. This leads to pulsatile melt conveyance, as a minimum amount/pressure is necessary to force material through the KBs [24
]. Consequently, the material amount which passes the die varies and, with it, the diameter. To what extent these fluctuations occur is highly dependent on the viscoelastic behavior of the used material and die swell [25
]. In this study, an impact of inhomogeneous powder feeding could be excluded (data not shown). Pressure fluctuations were observed to depend on process parameters. Accordingly, a rational process design was applied to determine the influence of PFR and screw speed on diameter variations. Three different PFRs at a screw speed of 40 rpm and five different screw speeds at a PFR of 5 g/min were tested (Table 1
3.2.3. Influence of Screw Speed
Regarding the influence of the screw speed, different results were obtained (Figure 6
A,B). An increasing screw speed resulted in lower material pressure because of a decreased barrel filling degree and the potential shear thinning of used polymers. With increasing screw speed, the IQR of the material pressure was increased. This had a distinct influence on diameter variations. Diameter fluctuations were significantly higher (IQR = 0.027–0.082 mm), especially above 40 rpm (p
< 0.01), although the mean value was similar in all cases (Table 2
). A likely reason is that at higher screw speeds the barrel filling degree is lower, resulting in more pulsatile melt conveyance through the KB and die. A faster removal of lower amounts of material takes place and the material held up in front of the KBs until enough material accumulated to build up the necessary pressure to push the melt along the barrel. At lower screw speeds (20–40 rpm) and thus higher barrel filling degrees, the diameter variations were lower (IQR = 0.027–0.044 mm), as enough material was present to convey the formulation more homogenously with less fluctuations. Filaments produced with 20 rpm and 5 g/min showed the lowest diameter variations.
Compared to commercial filament, the self-extruded filament is less homogeneous. The PLA filament has a very low IQR of 0.005 mm (1.756 ± 0.004 mm, CoVdiameter
= 0.21%; Figure 6
B), whereas the variation of self-produced filament showed a 5-fold higher IQR of 0.027 mm (1.782 ± 0.019 mm, CoVdiameter
It has to be considered that commercial filaments are usually produced with single screw extruders with a rather low mixing capacity because it is not required to incorporate materials with a uniform distribution. Consequently, pulsatile material transport and resulting diameter variations are minimal, as depicted in Figure 6
C, where the PLA filament diameter is plotted as function of time. In pharmaceutical processing, an extensive, dispersive and distributive mixing is necessary at controlled conditions. This is only provided by twin screw extruders (TSE) [26
]. Additionally, process ability is highly dependent on material characteristics, like rheological properties, molecular weight, particle size distribution, etc. Thus, diameter variations can be caused by the material itself but also by the selected extruder type.
A homogenous material throughput along the barrel is expected to be decisive for diameter homogeneity. With decreasing screw speed the barrel filling degree is increased, resulting in the abovementioned lower filament diameter fluctuations. The data from different PFR indicates that the fill level of the barrel was high enough in all cases to obtain homogeneous filaments (Section 3.2.2
). In the following, it should be investigated if lower filling degrees with different PFR/screw speed combinations affect diameter homogeneity.
3.2.4. Influence of SFL on Diameter Homogeneity
To investigate the influence of the barrel filing degree, the SFL was calculated according to Kohlgrüber and Wiedmann (Equation (1)) to enable a comparison with other formulations and (potentially) equipment.
In Figure 6
B, the SFLs for the respective applied screw speed are listed. The data indicates that if the SFL is low (<0.03), the material transport inside the barrel is more pulsatile, leading to larger diameter variations. If the SFL is higher (>0.03), diameter variations decrease distinctly, likely because of a more homogeneous melt conveyance, independent from the set PFR–screw speed combination. It should be investigated whether the same diameter quality can be achieved by keeping the SFL constant at different settings. In this study, the highest applied SFL was 0.059 and the lowest one 0.02. Both SFLs were investigated at two different PFRs (Table 1
). Four different setting combinations were applied in total. Therefore, the PFR was set to a low level (LL) with 5 g/min and to a high level (HL) with 10 g/min. The respective predefined SFLs were reached by applying different screw speeds (20–120 rpm). The IQR was used to evaluate variations.
In Figure 7
A,B, the results for the investigation of the SFL are displayed. At a low SFL of 0.02 the pressure data for both applied setting combinations was comparable (mean = 9.7 bar and IQR = 1.4 bar at 10 g/min and mean = 10.0 bar and IQR = 1.9 bar at 5 g/min; Figure 7
A). For the high SFL of 0.059, IQRs were also similar (1.3 and 1.2 bar). Even though the barrel filling was the same, the mean material pressure deviated (14.9 vs. 19.6 bar). This was not expected and might be caused possibly by a faulty calibration of the pressure probe at different temperatures. Although the pressure IQRs for the different SFLs did not differ a lot, the influence on diameter fluctuations was distinct (IQR = 0.082 mm at a low SFL and IQR = 0.027 and 0.035 mm at a high SFL). A limitation of the conducted experiments is that the evaluation of the absolute values of the material pressure might be not valid. Figure 4
shows that the fluctuations of pressure data depended on the set screw speed. At 60 rpm, pressure data appeared as a non-fluctuating line as sampling took place at the same screw position (sampling rate 1 Hz). In contrast, at 40 rpm, a signal oscillation was observed, which was not caused by fluctuating material transport but sampling at different screw positions. With the applied screw speed of 40 rpm, only 2/3 revolution per second occurred. Consequently, the initial position was only reached after every third second, resulting in fluctuating data.
Furthermore, the device-related low sampling rate (1 Hz) was probably not sufficient to obtain complete pressure information. The signal to be sampled (material pressure) contained frequency components that were higher than half the sampling frequency, which leads to an artifact, known as Alias effect in signal analysis or computer graphics [27
]. This discriminated the output signal, resulting in incorrect amplitudes and apparently lower frequencies. The pressure minima and maxima are then underestimated [28
]. The obtained material pressure trends still convey important information about variations, as shown in Figure 4
, but absolute values cannot be discussed with high confidence due to the aforementioned factors.
However, the stated hypothesis of the SFL as a key parameter for diameter variations could be confirmed (Figure 7
B). For the first time, it could be demonstrated that the same diameter quality was achieved at a given SFL, independently of the PFR and screw speed combination. This is true at least in the investigated operation window.
At higher SFLs and increased barrel filling degrees, filaments with less fluctuations in diameter (IQR = 0.035 and 0.027 mm) compared to lower SFL (IQR = 0.082 mm both) were obtained. The higher the SFL, the more even the material transport is, resulting in a uniform filament diameter independent from the extrusion parameter combination.
For the SFL of 0.02 the mean diameter was slightly higher (1.821 ± 0.062 mm) when 10 g/min were fed compared to the batch where 5 g/min was used as the PFR (1.777 ± 0.064 mm), although the same haul-off speed was applied. This was likely due to the combination of an increased shear stress at the higher PFR with the additional high screw speed of 120 rpm, resulting in higher die swell and thus a larger diameter [21
]. The hypothesis was further supported by plotting the diameter IQR as a function of SFL (Figure 7
C). A material dependent threshold of 0.03 as SFL was found. Below this value, a sharp increase in diameter variation was observed. This explains why no influence of the PFR was found in Section 3.2.2
, as the SFL was ≥0.03 for all combinations. For the SFL of 0.059 the lowest diameter variations were found. To underline this assumption, further analysis, including other formulations and extruder systems, is desirable.
The presented results are promising with regard to throughput-upscale studies to obtain consistent filament quality. Furthermore, these findings are beneficial for formulation development since process optimization and scale-up might be accelerated. Although experiments were performed for only one formulation, the results are likely to be applicable to other formulations.
3.2.5. Impact on Critical Quality Attributes of Filaments and 3D-Printed Dosage Forms
In the last part of the study, the impact of diameter variations on the CQAs printability and uniformity of mass of printed dosage forms was analyzed. Self-produced filaments (T1, T5–T8; screw speeds = 20, 30, 40, 50 and 60 rpm) with different extents of diameter variation were used. These filaments were chosen since it was found that with increasing screw speed and, consequently, lower barrel filling degrees, diameter variations increased (3.2.3). Therefore, a pronounced impact on CQAs of filaments and printed dosage forms was expected. The mechanical resilience of filaments was defined as the main factor for conveyance inside the print head and thus printability [4
]. Therefore, the influence of diameter variations on the YM and distance at break as measures for the mechanical resilience was determined in a tensile test and 3PBT according to Korte and Quodbach [4
(left) revealed that the YM was hardly influenced. Although there was a significant difference between the YM of filaments produced at 20 rpm and 60 rpm (p
<< 0.01), the effect was considered to be of little practical relevance, as the differences are small (Table 3
) and do not influence printability in any way. The impact of screw speed (or SFL) on the mean value of the distance at break was also negligible (Table 3
). However, the variation of the distance at break differed distinctly.
Thin sections were less resilient and tended to break faster compared to thicker sections. This was particularly visible when the CoVdistance at break
was calculated (Figure 8
, right). Higher screw speeds increased the CoVdistance break
from 3% to 20%. This has to be considered, especially when brittle formulations are used. In this case, the reliability of printing is likely to be reduced due to frequently occurring filament breakage inside the print head.
, the mean filament diameter is used for the calculation of the feed rate in the G-code. Therefore, it was expected that diameter variations would influence the mass variation of printed dosage forms, as diameter deviations are not considered. The influence of diameter fluctuations due to different screw speeds in extrusion on the mass variation of 20 printed test geometries (3 × 15 × 7.5 mm, Figure 2
) was tested.
Additionally, the obtained printed objects were assessed according to Ph.Eur. 2.9.5 “Uniformity of mass of single dose preparations”. In Figure 9
(left), the results of the mean mass of printed geometries from filaments produced at different screw speeds are depicted. The obtained masses of dosage forms printed from the self-extruded formulations were similar, with a slight trend to higher masses from filaments produced at higher screw speeds. Test objects printed from filaments produced with 20 and 30 rpm (T5 and T6) were printed on different days. Due to a necessary reassembly of the print head, a different gap width between the conveyer gears might have been set, leading to different feeding properties and deviating mean masses.
(right) reveals that a higher screw speed causes higher mass variation of printed objects. Screw speeds above 40 rpm led to CoVmass
between 6% and 7%, which was not acceptable to ensure content uniformity of printed DDS. According to Ph.Eur. 2.9.5, only one mass may deviate by more than 5% from the mean value. Only dosage forms produced from filaments manufactured at 20 and 30 rpm (T5 and T6; SFL = 0.059 and 0.04) passed the test (Table 3
). The other filaments (T1, T7 and T8) did not comply with requirements of the Ph.Eur. For filament formulation T8 (60 rpm), eight dosage forms deviated more than 5%, three of them more than 10%. Filaments T1, T5 and T6 (CoVdiameter
< 2%) resulted in mass variations of printed objects lower than 4% (20–40 rpm, Table 2
of printed objects from filaments produced with screw speeds of 20 and 30 rpm were similar, even though diameter variations were lower for the filament produced at 20 rpm (T5, CoVdiameter
= 1.07% vs. 1.76%). An optimization of the print settings might enable an improvement of the quality of printed dosage forms and reveal differences even between low CoVdiameter
]. The results of printed objects from commercial PLA filaments revealed that the printer worked at a high precision, enabling the production of even higher quality dosage forms. The study demonstrates that low diameter variations are necessary to obtain drug products with sufficient quality. Compared to objects printed from PLA filaments with a mass variation of approx. 0.5%, the lowest deviation in this study, 3.3% (T5, 20 rpm), is higher. Still, objects printed from T5 were in accordance with the quality requirements of the Ph.Eur.