Accepted Manuscript

Impact of fill-level in twin-screw granulation on critical quality attributes of

granules and tablets

Robin Meier, Klaus-Peter Moll, Markus Krumme, Peter Kleinebudde

PII: S0939-6411(16)30579-3
DOI: http://dx.doi.org/10.1016/j.ejpb.2017.02.010
Reference: EJPB 12444

To appear in: European Journal of Pharmaceutics and Biophar-

maceutics

Received Date: 19 September 2016

Please cite this article as: R. Meier, K-P. Moll, M. Krumme, P. Kleinebudde, Impact of fill-level in twin-screw
granulation on critical quality attributes of granules and tablets, European Journal of Pharmaceutics and
Biopharmaceutics (2017), doi: http://dx.doi.org/10.1016/j.ejpb.2017.02.010

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Impact of fill-level in twin-screw granulation on critical quality

attributes of granules and tablets

1 2 2 1,*
Robin Meier , Klaus-Peter Moll , Markus Krumme , Peter Kleinebudde

1
Institute of Pharmaceutics and Biopharmaceutics, Heinrich Heine University, Universitaetsstr. 1,

40225 Duesseldorf, Germany

2
Novartis AG, 4002 Basel, Switzerland

*Peter Kleinebudde, tel.: +492118114220, fax: +492118114251, [email protected]

Abstract

In a previous study a change of the fill-level in the barrel exerted a huge influence on the twin-screw

granulation (TSG) process of a high drug loaded, simplified formulation. The present work investigated

this influence systematically. The specific feed load (SFL) indicating the mass per revolution as

surrogate parameter for the fill-level was applied and the correlation to the real volumetric fill level of

an extruder could be demonstrated by a newly developed method. A design of experiments was

conducted to examine the combined influence of SFL and screw speed on the process and on critical

quality attributes of granules and tablets. The same formulation was granulated at constant liquid level

with the same screw configuration and led to distinctively different results by only changing the fill-level

and the screw speed. The power consumption of the extruder increased at higher SFLs with hardly

any influence of screw speed. At low SFL the median residence time was mainly fill-level dependent

and at higher SFL mainly screw speed dependent. Optimal values for the product characteristics were

found at medium values for the SFL. Granule size distributions shifted from mono-modal and narrow

shape to broader and even bimodal distributions of larger median granule sizes, when exceeding or

falling below a certain fill-level. Deviating from the optimum fill-level, tensile strength of tablets

decreased by about 25 % and disintegration times of tablets increased for more than one third. At low

fill-levels, material accumulation in front of the kneading zone was detected by pressure

1
measurements and was assumed to be responsible for the unfavored product performance. At high

fill-levels, granule consolidation due to higher propensity of contact with the result of higher material

temperature was accounted for inferior product performance. The fill-level was found to be an

important factor in assessment and development of twin-screw granulation processes as it impacted

process and product attributes enormously.

Keywords

continuous manufacturing, disintegration time, fill-level, high drug loads, material pressure, material

temperature, residence time distribution, tableting, twin-screw granulation

2
Abbreviations

API active pharmaceutical ingredient

CPP critical process parameter

CQA critical quality attribute

D diameter of the extruder screws

DOE design of experiments

GSD granule size distribution

IQR interquartile range

L/S-ratio liquid to solid ratio

MRT median residence time

PAT process analytical technology

SFL specific feed load

SS screw speed

TSG twin-screw granulation

x50 median granule size

3
1. Introduction

Triggered by the Food and Drug Administration’s guidance regarding process analytical technologies

(PAT) in 2004 [1], a paradigm change in the pharmaceutical industry towards the continuous

production of pharmaceuticals has commended. A major implication of continuous processing is the

necessity of extensive and real-time knowledge of process and material attributes in the framework of

quality by design. Critical process parameters (CPP) and critical quality attributes (CQA) have to be

defined and monitored as consistently as possible to ensure constant and homogenous product

quality. Furthermore, the traceability of all ingredients of a pharmaceutical formulation throughout the

process up to the point of the finished product has to be secured. One approach to reduce the

complexity of this real-time and traceability analysis in case of continuous twin-screw granulation

(TSG) was the introduction of simplified formulations, which contained beside the active

pharmaceutical ingredient (API) as few excipients as possible. These simplified formulations were

investigated for mechanically different behaving drugs (plastic and brittle) and contained only API and

disintegrant or API, disintegrant and binder [2, 3]. Depending on the applied API, different

disintegrants could be used to ensure fast disintegration and strong tablets. In addition to the reduction

of traceability problems and the complexity of real-time analysis, the formulations get less susceptible

to batch inconformities, which can arise with natural, semi-synthetic and fully synthetic products.

During the development of simplified formulations with the API hydrochlorothiazide and the

disintegrant sodium starch glycolate, some problems encountered [3]: a pulsating output of the

extruder, characterized by oscillating intervals of high output following intervals of low output.

Furthermore, at different time points of the process differently sized granules left the extruder. An

accumulation of material in front of the kneading zone was supposed to be responsible for the

problems. It was hypothesized that the accumulation was cleared as soon as new material produced

enough pressure to convey the material through the kneading zone. The problems could be solved by

an increase of the throughput keeping all other process parameters constant. Alongside with the

process improvements, CQAs of the granules and tablets could be improved. Effectively, by increasing

the throughput, the fill-level of the extruder was elevated.

The volumetric fill-level of an extruder is the volume, the material occupies, in proportion to the

maximum available volume in the extruder barrel. There are plenty of studies published, which dealt

4
indirectly with the fill-level in TSG or in which the fill-level was influenced indirectly [4-11]. However, in

none of these studies, the fill-level was determined or directly and systematically investigated as

independent factor. This is because the volumetric fill-level of an extruder cannot be adjusted readily

due to several unknown variables. A method, applied by the research group of Thompson [12-14],

called screw pullout would come closest to determine the volumetric-fill degree, but thereby still not the

whole material would be captured. But as the volumetric fill-level of an extruder is mainly determined

by the throughput and the screw speed (SS) as input and output determining variables, a preliminary

surrogate parameter for the volumetric fill-level will be employed in this study. Literature about

extrusion provides two rather similar parameters, namely the specific feed load (SFL). Kolter et al. [15]

defined it according to Eq. 1 as gravimetric mass flow divided by the screw speed, which results in a

mass, which is transported per screw revolution, with the dimension of mass [g]. Kohlgrüber [16]

defined it similarly with a further multiplication by the screw diameter and the material density in the

denominator, resulting in a dimensionless number. In the context of this study it was decided to

employ the SFL, according to Kolter et al. [15], because of its simplicity. However, it has to be

mentioned that the SFL is only valid for one applied system, meaning the formulation, the extruder

dimensions and the screw configuration, because in both equations the maximum capable volume of

the extruder is not included.

̇
=

Eq. 1

SFL: specific feed load [g]

̇ : gravimetric throughput [g/min]

f: screw speed [1/min]

The aim of this study was to investigate if the SFL correlates with the volumetric fill-level of an extruder

and if it is justified to use it as surrogate parameter. Therefore, a preferably simple method to measure

the volumetric fill-level of an extruder had to be developed. With this knowledge a design of

experiments (DOE) with the SFL as independent factor and the problematic formulation from the prior

study [3] should be conducted, whereby the influence of the SFL on process parameters and CQAs

5
should be investigated. At last, the hypothesis of the pressure built up through material accumulation

in front of the kneading zone at low SFL should be assessed.

2. Materials and methods

2.1 Materials

The applied formulation consisted of 89.8 % (w/w) hydrochlorothiazide (Unichem Laboratories,

Mumbai, India), 10 % sodium starch glycolate (Explotab, JRS Pharma, Rosenberg, Germany) and

0.2 % colloidal silica (Aerosil 200, Evonik, Darmstadt, Germany). Demineralized water was used as

granulation liquid. During the residence time determinations, the red dye Sicovit Amaranth 85 E 123

(BASF, Ludwigshafen, Germany) was used as tracer substance. For the tableting experiments,

magnesium stearate (Parteck LUB MST, Merck Millipore, Darmstadt, Germany) was used as lubricant.

2.2 Methods

2.2.1 Design of experiments

To investigate the influence of the SFL (low level: 0.15 g, high level: 0.4 g) and the SS (low level:

200 1/min, high level: 500 1/min) on the process and on CQAs of granules and tablets, a central

composite design was planned, to allow the evaluation of linear and quadratic effects, as well as

interactions of the factors. This design consisted of a two level full factorial design with three

executions of the center point experiment, resulting in seven trials. Additionally, four star points were

added as trials, with a star distance of 1.414 (square root of two), from the center point in each

direction of the factors. Three additional trials at different SFLs were added to the experimental plan,

resulting in a total of 14 experiments. The sequence of the trials was randomized, to minimize the

effect of unknown, systematic influences. A complete overview of the experiments and the different

factor settings is given in Table 1. Following response variables were evaluated: volumetric fill-

level [%], median granule size [µm], granulation yield (granules from 180 µm - 1000 µm) [%], amount

of oversized granules (> 1700 µm) [%], disintegration time of tablets [s], tensile strength of tablets

[MPa], power consumption of the extruder [%], median residence time of material inside the extruder
nd
[s], material temperature in front of 2 kneading zone [°C]. The DOE was planned and evaluated with

the software Modde 9.0 (Umetrics, Umeå, Sweden) using multiple linear regression at a significance

6
level of α = 0.05. After fitting to the experimental data, the model for each response was refined via
2
backwards regression to result in the highest possible value for the coefficient of prediction (Q ).

2.2.2 Preparation of powder mixtures

Powder amounts of 5 kg each were blended for 20 min in a laboratory scale blender at 35 1/min (LM

40, L.B. Bohle, Ennigerloh, Germany).

2.2.3 Twin-screw granulation and drying

After mixing, powder was transferred to a gravimetric powder feeder (K-ML KT20, K-Tron, Niederlenz,

Switzerland), which fed the powder at the desired fee-rate, according to the experimental plan.

Granulation liquid of 25 °C was fed through a nozzle of 0.12 mm inner diameter directly on the top of

one screw, in front of the first kneading zone. The applied pump was a micro annular gear pump

(MZP 7205, HNP-Mikrosysteme, Schwerin, Germany), which was monitored and manually controlled

by a Coriolis mass flow meter (Proline Promass 80A, Endress+Hauser, Weil am Rhein, Germany).

The liquid to solid ratio (L/S) was kept constant throughout all experiments at 0.11. Granulation was

performed on a twin-screw extruder (Pharmalab 16, Thermo Fisher Scientific, Karlsruhe, Germany),

which had a screw diameter (D) of 16 mm and a total screw length of 40 x D. The SS was adjusted,

according to the experimental plan. The screw was configured as follows:

2 kneading discs as spacer (1/2 D) - 4 D conveying (Helix of 3/2 L/D) - 8 D conveying (Helix of 1 L/D) -

5 x kneading disc 60° (1 1/4 D) – 8.5 D conveying (Helix of 1 L/D) - 7 x kneading disc 60° (1 3/4 D) –

1 D distributive flow element (Helix of 1 L/D) - 15 D conveying (Helix of 1 L/D) - flat sealing screw.

In front of the 2nd kneading zone, on top of one screw, the product temperature was measured (2075J-

D8-100B, Terwin, Nottinghamshire, England) through a specially designed and in-house built port. For

one high and one low SFL the pressure development was measured. Further details for this are given

in section 3.3. After each experimental run, the barrel and the screws were completely cleaned to

prevent influences of probably occurring material fouling, recently described by Saleh et al. [17].

Process parameters of the extruder, and the feeders were monitored and recorded by an in-house

written Labview 2015 application, utilizing a cDAQ 9174 based hardware system at a sampling rate of

1 Hz (National Instruments, Austin, TX, USA).

Approximately 600 g of granules (based on dry mass) were collected at the extruder outlet and

afterwards manually transferred to a fluid bed dryer (GPCG 1, Glatt, Binzen, Germany). Granules were

7
dried at an inlet temperature of 60 °C with an air flow of 80 m 3/h till a product temperature of 37 °C

was reached.

2.2.4 Median residence time

Residence time distributions were measured inline and evaluated, utilizing the ExtruVis 2 system and

the ExtruVis 3 software (ExtruVis, Riedstadt, Germany). The measurement principle of the probe is

based on the measurement of reflections, which change, when the colour of the granules change. A

red dye was added as a pulse (always 1 % of the actual throughput per second) to the powder inlet of

the extruder and the progression of the reflection signal was recorded at the outlet of the extruder with

a frequency of 4 Hz. The resulting curve over time was smoothed by the software and normalized to

the area under the curve to result in an exit age function E (t). The median residence time (MRT) was

consulted for evaluation.

2.2.5 Determination of the volumetric fill-level

After the production of granules for further evaluation, each experiment of the DOE was repeated, to

determine the volumetric fill-levels. The free screw-volume inside the barrel had to be determined.

Therefore, the empty barrel volume was calculated geometrically and the volume displaced by the

screws was measured by water displacement measurements after Archimedes' principle. To measure

the volume of material inside the barrel for each point of the DOE, the extruder barrel and the screws

were completely emptied and cleaned, before each trial. Then, powder feeding, liquid feeding and the

screws were started at once and the wet granules were collected on a catch scale (CP 4201,

Sartorius, Goettingen, Germany). After 5 min, the experiments were stopped and by the data

recording of the powder and liquid feeder and the mass, caught on the balance, the masses, which

have been put in and transported out of the extruder were known. The difference was the remaining

mass of material inside the extruder. Thus, a gravimetric fill-level was known. To calculate the

volumetric fill-level, freshly prepared wet granules were measured with n = 5 for their helium density

(AccuPyc 1330, Micromeritics, Norcross, GA, USA). From their mass and their density, the occupied

volume inside the extruder could be calculated.

2.2.6 Granule porosity

The granule size fraction from 1.4 mm - 1.6 mm was analyzed for particle density (ρ He), utilizing a

helium pycnometer (AccuPyc 1330, Micromeritics, Norcross, GA, USA). Mercury pycnometry was

applied to analyze granule density (ρHg) in duplicate (Pascal 140, Thermo Fisher Scientific, Milano,

8
Italy). Examined masses of the granules ranged from 0.8 - 1.1 g. The granule density at a mercury

intrusion pressure of 0.1 MPa was consulted, to calculate granule porosities according to Eq. 2.

[%] = − ∙

Eq. 2

εGran: granule porosity [%]

ρHg: granule density [g/cm3]

3
ρHe: particle density [g/cm ]

2.2.7 Granule size distribution

Prior analysis of the size distributions, granules larger than 1700 µm were separated by vibrational

sieving, to gravimetrically determine the amount of oversized granules (> 1700 µm). After that the

fractions were united again, divided into samples of appropriate size by an automated spinning riffler

(Retsch, Hahn, Germany) and analyzed for granule size distribution (GSD) in triplicate via dynamic

image analysis with the Camsizer XT (Retsch, Hahn, Germany). Air dispersion pressure was set to 0.3

bar. The high speed cameras inside the Camsizer measured shadow projections of air dispersed

granules. The Camsizer measured a set of maximum chord lengths in 32 measuring directions and

used the diameter of the smallest maximum chord length for calculating a volume distribution.

Furthermore, the median granule size (x50) and the yield for tableting experiments (180 µm - 1000

µm), were consulted for evaluation.

2.2.8 Tableting

The granule fraction of 180 µm - 1000 µm was blended for 2 min with 0.5 % magnesium stearate in a

Turbula mixer (T2C, Willy A. Bachofen AG, Muttenz, Switzerland) and after a minimum storage time of

72h under controlled conditions (21 °C, 45 % r.h.), transferred to the hopper of a rotary die press

(Pressima MX Eu-B/D, IMA Kilian, Cologne, Germany). Flat-faced tablets of 8 mm diameter were

compressed at a pressure of 100 MPa, a rotation speed of 10 1/min with an aimed tablet weight of

180 mg. Prior analysis all tablets were stored for 72h under controlled conditions (21 °C, 45% r.h.).

2.2.9 Disintegration time, tensile strength and porosity of tablets

Tablets were tested for disintegration time in 700 ml of demineralized water of 37 ± 0.1 °C with n = 6

(DT 2, Sotax, Allschwil, Switzerland).

9
A semi-automatic tablet testing system (Smart Test 50, Sotax, Allschwill, Switzerland) was used to

determine thickness, diameter, weight and breaking force of the tablets (n = 20). The tensile strength

was calculated via Eq. 3 according to Fell and Newton [18]. Porosities of tablets were calculated from

their mass, thickness and diameter (tablet density) and the particle density, according to Eq. 4. The

particle densities were obtained by helium pycnometry, using the granule fractions from

180 - 1000 µm, mixed with magnesium stearate, according section 2.2.7.

∙
=
∙ ∙

Eq. 3

TS: tensile strength [MPa]

F: breaking force of tablet [N]

t: tablet thickness [mm]

d: tablet diameter [mm]

[%] = ( − )∙

Eq. 4

εTab: tablet porosity [%]

3
ρTab: tablet density [g/cm ]

ρHe: particle density of granules [g/cm 3]

10
3. Results and discussion

3.1. Start-up phase and determination of the volumetric fill-level

In Table 1 a complete overview over the factor settings within the different experiments, as well as the

complete raw data are given. The experiments N 12 - N 14 were planned additionally to the central

composite design and included in the DOE, to fill the gap in between the different SFLs. Those trials

were conducted at the zero-level for the factor SS.

Vercruysse et al. [19, 20] showed in different studies the importance of the TSG system to run in

steady state conditions, especially because the extruder power consumption needed time to turn into

a constant region. The studies pointed out that CQAs of the products underlay changes during the

start-up phase. Only few studies dealing with TSG mentioned that a waiting period was exercised,

before collecting samples. Thus, also for this study it was important to wait for steady state conditions,

even if a huge amount of material was wasted, because especially a DOE can dramatically be affected

and become insignificant, when the process conditions are not comparable. Figure 1 exemplifies by

the process data of experiment N 13 a typical start-up phase, which occurred during this study. The

power consumption reached its steady state within a period of approximately 700 s. However, also the

barrel wall (exemplified by barrel zone 7, where the 2nd kneading zone is located) and the material

temperature slowly increased during startup phase and needed even more time to reach steady state.

Consequently, it was important to wait until those parameters were constant. Otherwise, it could occur

that during the start-up phase granule populations with changing properties would have been

collected. The time to reach steady state varied for the different experiments from 430 s - 1400 s. All

granules for further processing within this study were collected after reaching steady state conditions

i.e. after constancy of power consumption and all temperatures. Solely, during the determination of the

volumetric fill level, which was done after the other trials (section 2.2.5), it was not possible to wait until

steady state conditions were reached, because the material was running out.

From theoretical considerations the SFL should behave proportional to the volumetric fill-level. But to

justify the use of SFL as a surrogate parameter for the volumetric fill-level of an extruder, it was

necessary to develop a simple method for the measurement of the volumetric fill-level, because only if

both parameters correlated, the DOE could be evaluated in a reasonable way. The results for the

11
correlation of SFL and volumetric fill-level are depicted in Figure 2a and Table 1. The free volume of
3
the screws inside the extruder barrel was calculated as 177.6 cm . Volumetric fill-levels between 6 %

and 20 % where reached in this study, which indicated that even at high SFL values of this DOE, still a

lot of free space was left inside the extruder. A linear correlation between both parameters was visible

and thereby it was confirmed that the SFL could be used as surrogate parameter for the volumetric fill-

level. Deviations from linearity may be explained by the missing steady state conditions in this part of

the study, by the fact that powder feeder data is not always 100 % reliable [21] and by possible

inaccuracies during the He-density measurements, because water may have been evaporated before

and during measurements. Nevertheless, a simple and reliable method for the measurement of the

volumetric fill-level in TSG for the first time has been developed and as it is independent from extruder

size and geometry it allows comparison of differently scaled extruders. Table 2 contains results from

the statistical evaluation of the DOE and reveals that the model for the volumetric fill-level is

adequately able to describe the experimental data. Coefficient of determination (R 2), coefficient of
2
prediction (Q ), lack of fit (p) and reproducibility showed consistently high values. In each column for

the respective responses, the coefficients ± CI are given, as well as the p-value for the coefficients to

get an idea about significance. For the volumetric fill-level only SFL is a significant coefficient, which is

important, because it shows that the volumetric fill-level is only influenced by the set SFL.

3.2. Influence on critical quality attributes of granules and tablets

Especially the influence of the factors of the DOE on CQAs of produced granules and tablets was of

interest in this study. The GSD is of major importance as it can ensure the robustness towards a

subsequent process e.g. tableting. Mono-modal and narrow size distributions with an appropriate

flowability for example can diminish deviations from the desired tablet weight. Figure 3a depicts the

GSD from four DOE experiments, which differed distinctively. All granules were obtained with a SS of

350 1/min at different SFLs, which are denominated in the figure legend. A huge influence of the SFL

became visible, as the GSDs ranged from mono-modal and narrow distributions in case of a SFL of

0.275 g and 0.06 g up to much broader, respectively almost bimodal distributions with a higher amount

of oversized granules in case of SFLs of 0.35 g and 0.45 g. These findings disagree with results from

Vercruysse et al. [22], who did not see an influence of increasing throughput and thus fill-level, on

resulting GSD. However, this may be due to the overall applied higher throughputs and SS as well as

12
a different formulation. The finding of changing GSD could also be confirmed by the interquartile range

(IQR) of the GSD, which is a measure for the width of the distribution. The IQR for the SFL of 0.275 g

was lowest with 587 µm, whereas it increased up to 816 µm for the SFL of 0.45 g. Thus, an increase

in SFL, at constant SS led to an increase in granule size, which could also be shown by Fonteyne et

al. [23], when the throughput was increased, although a different formulation was used. A reason for

this can be seen in a higher propensity of granule consolidation, when SFL increases, because the

contact points of the wet material increased. The raw data for the median granule size (x 50), which is

shown in Figure 3b underlines the findings from the GSD and confirms a steady granule growth at

SFLs above 0.2, up to extreme big granules with an x 50 of above 1150 µm. Combined with the contour

plot, shown in Figure 3c, an optimum for the x 50 was visible for the lowest third of the SFL for SS up to

approximately 330 1/min. At higher SFLs the SS had almost no influence, which was confirmed by the
2
coefficients, shown in Table 2. Only the SFL and the SFL were significant factors. The model for the

x50 was adequate to describe the experimental data. Solely the lack of fit, which is also a measure for

the model validity, is small with a value of 0.105. However, this low value could be ascribed to the high

reproducibility of 0.988. When the reproducibility is this high, the observational error is estimated by

the program as very low and thus, even small deviations of the experimental data from the model are

ascribed to systematic errors, which leads to a decrease of the lack of fit. Consequently, in such a

case, which occurred a few times more in this study, a low value for the lack of fit does not

automatically impair the quality of the model.

The yield for tabletting experiments and the amount of oversized granules could also be well described

by the model, with the same observation regarding the lack of fit, as already occurred for the x 50.

Figure 4a presents that the yield decreases with SFL values above 0.2 g, which is connected to the

increasing x 50 values at increasing SFL. The contour plot for the yield in Figure 4b confirmed these

findings and revealed an optimal region for the yield, which was located in the same region as for the

x50 in terms of SS and which covered an SFL up to 0.25 g. According to the contour plot and the

coefficients in Table 2, the SS did merely influence the yield at lower SFLs with a reduction at

increasing SS.

Figure 4c/d depicts the raw data and the contour plot for the amount of oversized granules

(> 1700 µm) with a minimum amount of 0.4 % at an SFL of 0.15 %. Above and beyond this value, the

13
amount of oversized granules increased again, indicating that there is also an optimum. Only the
2
coefficients for SFL and SFL (Table 2) had influence on the amount of oversized granules with an

increase at increasing SFL. The reason for that must be the same as for the increasing value of the

x50. A higher SFL led to more contacts and thus consolidation and granule growth. The SS was
2
excluded from the model in terms of maximizing Q . But also the raw data revealed that SS hardly had

any influence on the amount of oversized. Therefore, the contour plot depicted only straight lines,

without the influence of SS.

Besides granule quality, the CQAs of tablets are of major importance for the pharmaceutical industry.

Disintegration time, which is a requirement for dissolution and tensile strength were investigated as

CQAs in this study. The tensile strength should be as high as possible and the disintegration time

should be as low as possible. The model for the tensile strength in terms of R 2 and Q2 (Table 2) is the

weakest one of this study and p is again low, because of the high value for the reproducibility, but the

model is still able to describe the experimental data adequately. The tensile strength values differed

from 1.2 MPa - 1.77 MPa for the different experiments (Figure 5a) and again an optimal region in

terms of SFL became visible. These values were obtained for the same formulation and a constant

L/S-ratio and screw configuration. Merely process parameters during granulation were changed.

According to the contour plot (Figure 5b), the optimum could be found at SFL values around 0.25 g

and high SS above 450 1/min. Too high and too low SFL values led to a decrease of the tensile

strength, whereas an increase of the SS always was beneficial, especially at high SFL. The reason for

that may probably be connected to the granule sizes. Especially smaller particles were often shown to

positively contribute to the tensile strength of a tablet [24-26]. The increase of x 50 above and beneath

an SFL of 0.25 g could therefore lead to lower tensile strength values, as less fine granules were

available.

The raw data for the disintegration time (Figure 5c) revealed that no fast disintegration according to

Ph. Eur. (< 900 s) could be achieved, but this was acceptable as the aim of this study was simply to

examine the quantitative influence of the SFL on specific parameters. The model for the disintegration

time again is capable of explaining the experimental data appropriately. Overall results showed

disintegration times ranging from 1440 s - 2200 s. The raw data featured an optimum at an SFL of

0.15 g and 500 1/min SS. The optimum, which is also recognizable by the contour plot (Figure 5d), lied

14
at lower SFL values, compared to the tensile strength. Higher values for SS again promoted more

favourable results. The influence of the SS got lower at higher SFL.

The relationship between tensile strength and disintegration time is depicted in Figure 6, with the

values for SFL and SS, written beside the data points. A trend of higher tensile strength, combined

with faster disintegration got obvious. This is beneficial for a tablet, but it is also difficult to explain. The

tablet porosities could not serve as explanation for it, because they ranged from 17.2 % - 18.7 %

without significant differences among each other. The granule porosities differed in a wider range from

14 % - 22 %. There is not much literature available to compare the achieved granule porosities, which

furthermore are highly formulation dependent but most of the studies reported higher porosities, as

found in this study [6, 27-30]. Nevertheless, also the granule porosities could not be correlated to

tensile strength or disintegration times to serve as explanation for this behaviour. However, Figure 6

underlines that SFL in a medium range from 0.15 g - 0.275 g, preferably at higher SS yielded in more

favoured tablet properties. On the other hand, very low and high SFL and a medium SFL with low SS

resulted in worse tablet properties, characterized by lower tensile strength and longer disintegration

times.

3.3. Influence on process parameters

The influence of SFL on the power consumption of the extruder is depicted in Figure 7a. The power

consumption was mostly constant up to an SFL of 0.15 g and increased at higher SFL values. The

increase was anticipated, as more mass had to be transported. Figure 7b and Table 2 present the

contour plot and the model data for the power consumption. The model could excellently describe the

experimental data. Increasing the SS led to a decrease of power consumption and increasing the SFL

led to an increase of it, but to a bigger extent. Interestingly, at higher SFLs of above 0.35 g the SS had

almost no influence anymore.

The residence time distribution of granules inside the extruder is of major interest, because it

determines the duration and the magnitude of shear forces, the material is subdued to. Furthermore,

when more than one powder feeder is applied, the residence time can highly affect the mixing inside

the extruder. Usually, increasing SS and increasing material throughput decrease the residence time

of materials in an extruder. Especially for TSG in-line measurements (e.g. NIR/Raman) of the

15
residence time distributions are important, because they usually lie below one minute or even in the

region of seconds [21, 29, 31, 32]. In this study the ExtruVis 2, a probe, which measured the visible

light reflection signal of granules leaving the extruder, was utilized (setup: Figure 8a). The raw signal

of the ExtruVis 2 revealed some scattering (Figure 8b), which could be ascribed to the non-constant

distance of the granules, when leaving the extruder. The MRT was calculated from the E (t) function

and was evaluated in the context of this study. The MRT as a function of the SFL (Figure 8c) revealed

a trend of decreasing MRT with increasing SFL, but there was also a plateau visible at higher values

for the SFL. Because of a positive skewness of the raw data, the log MRT was evaluated statistically.
2 2
The model is able to adequately describe the experimental data, as the values for R and Q (Table 2)

are consistently high. The value for lack of fit is again low, because of a high value for reproducibility.

Furthermore, the good reproducibility revealed that the ExtruVis 2 provided reliable and meaningful

data for the evaluation of the MRT. The contour plot for the MRT indicated that SS had the biggest

influence. Especially at low SFLs, the MRT decreased with increasing SS. At values above 0.25 g the

SFL hardly had any influence on the MRT. Below this value an increase of SFL led to a decrease in

MRT.

In a previous study [3] at low fill-levels, a fluctuating output of granules out of the extruder was

observed, whereupon it was hypothesized that material blockage in front of kneading zones of the

screw could be responsible for this. Also the temperature of the barrel zones increased to a differing

extent. Consequently, it was important to control these variables, too. In order to facilitate this, a new

port for the extruder was built in-house, which enabled the application of pressure and temperature

probes directly on the top of one screw on any position alongside the barrel. The measurement was

done in the shown region (Figure 9), directly in front of the second kneading zone, where the material

accumulation was expected. The temperature was measured for all experiments of the DOE. The

pressure was measured for one experiment of low and high SFL, each. The material temperature,

depending on the SFL, is depicted in Figure 10a. The temperature of the material in front of the

kneading blocks rose up to 78°C, which might be specific for this formulation, due to high plastic

energy dissipation [33] of deformed sodium starch glycolate. In prior studies, using the same system

with different APIs and excipients, no temperature increase was observed. Furthermore, it is obvious

that the SFL did not alone control this temperature, as two trend-curves were recognizable in the

figure. Table 2 confirms the model quality as appropriate to describe the experimental data and also

16
contains the model coefficients. As expected, SFL and SS have a positive influence on the resulting

material temperature, which increased with increasing values of the two factors. This could be

explained for increasing SS with higher shear forces and for increasing SFL with increasing friction

due to higher material density inside the extruder. Also the quadratic effect of SFL and the

combinatory effect of SFL and SS influence the material temperature significantly in positive direction.

Merely the quadratic effect of SS influences the material temperature to a negative direction. In the

coefficient plot (Figure 10b) two regions could be identified. In the region of high SS the SFL showed a

bigger influence as at low SS. Therefore, the SS showed higher influence in the region of high SFL as

in the region of low SFL. The high material temperatures at high SFL could also somehow contribute

to the slower disintegration of tablets produced from these granules, due to higher plastic deformation

at higher temperatures.

The pressure development in front of the second kneading zone was measured to evaluate the

hypothesis of material accumulation in front of the kneading zone, proposed in a previous study [3].

Different time frames were alternating in which either plenty of material or almost no material exited

the extruder. A material accumulation would also result in a built up of pressure and thus would be

measurable. In a recent study, Li et al. [30] performed pressure measurements in the center of both

screws, directly on top of the kneading zone, in order to understand granulation mechanisms within

these screw elements. They found highly frequent pressure fluctuations inside the kneading zone,

which were related to the SS. Differing to that, the progress of pressure in front of the second

kneading zone is shown for the lowest SFL of 0.06 g in Figure 11a. A time frame of 200 s is magnified

to better distinguish the course. Pressures < 0 bar are due to the sensor drift at changing

temperatures, as the zero signal of the sensor had to be calibrated with an empty extruder and thus at

ambient temperature. A periodic increase and decrease of the pressure in front of the kneading zone

is recognizable for this SFL with pressures up to approximately 10 bar. A Fourier transformation was

performed, to confirm the periodic character of the pressure fluctuations (Figure 11b). The four main

recurring frequencies, which occurred in the periodogram, could not be correlated with the SS and

thus had to be connected to the amount of material, which accumulated in front of the kneading zone.

A different behaviour was detected at a high SFL of 0.4 in Figure 11c. With the exception of

background noise, no noteworthy pressure development could be detected for this SFL, which must

17
be due to the pressure of new material, which is constantly high enough to force the material through

the kneading zones.

These findings support the hypothesis of material blockage in front of kneading zones at low SFL and

also have new implications for continuous TSG. Despite the fact that power consumption and material-

and barrel temperatures were in steady state, the pressure measurement revealed that the process

was kinetically fluctuating. Through these fluctuations, at different time intervals, different populations

of granules must have left the extruder, which is inacceptable for a continuous process. Without this

modified PAT tool, this instability could not have been detected. Perhaps it is meaningful, especially at

low SFL, for example in development processes, to implement pressure transducers as additional PAT

tool in TSG to detect and prevent such larval process instabilities.

4. Conclusion

This study for the first time provided an insight into the role of the fill-level in TSG. Experiments were

conducted with a specific, high drug loaded formulation at volumetric fill-levels from 6 % - 20 %

corresponding to SFLs from 0.06 g - 0.45 g. The correlation of the SFL and the volumetric fill-level was

confirmed and thereby it became clear that the SFL could be used as surrogate parameter in process

development for quality by design purposes. A systematic influence of the SFL on various process

parameters and CQAs of granules and tablets was found in the framework of a DOE and in

combination with the SS. At low SFL the SS influenced all parameters. At high SFL the importance of

SS was much lower and it did merely influence the tensile strength, the MRT and the material

temperature. Above a certain power consumption of approximately 30 %, which was exceeded above

SFLs of approximately 0.2 g- 0.3 g, all CQAs were affected negatively: The x50, the amount of

oversized granules and the disintegration time of tablets increased, the granulation yield and the

tensile strength decreased. The data hinted to a region of optimal SFL for this formulation, which was

met at about 0.25 g. Below and above this value, the CQAs of the granules and tablets impaired

again. This got also obvious by different GSDs. For one fixed SS, an SFL of 0.275 g featured an

optimal GSD, which was mono-modal and narrow with an appropriate median granule size, whereas

by higher and lower values the GSD got worsened. An overall optimum of SFL was not found in the

framework of this study but medium SFLs at higher SS resulted in more preferable products.

18
Material accumulation in front of a kneading zone at low SFLs and the intense increase of material

temperature at high SFLs, seemed to be connected to the formation of optimal regions for product

characteristics. A pressure measurement in front of a kneading zone enabled the detection of larval

disturbances in a continuous process, which apparently ran in steady state conditions. Therefore, if

low fill-levels cannot be prohibited, it might be beneficial to apply a pressure sensor to assure stable

process conditions. Further research should be conducted in view of the great influence, the fill-level

can execute on product and process.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgements

The authors would like to thank Mr. Andreas Gryczke for the provision of the ExtruVis system to

measure and evaluate residence times. We also acknowledge Stefan Stich and the fine mechanics of

the faculty of Mathematics and Natural Sciences of the Heinrich Heine University Duesseldorf for the

support in planning and manufacturing of the port for the temperature/pressor sensor. Robin Meier

was funded by Novartis.

19
Figure & table captions

Figure 1: Start-up phase of the granulation process, exemplified by trial N 13; power consumption,
th nd
material temperature and barrel temperature in the 7 extruder zone (2 kneading zone) are depicted.

Figure 2: Correlation of the factor SFL and the response volumetric fill-level.

Figure 3: a) Volume density distributions of granule sizes resulting from different SFLs (numbers in
legend), at 350 1/min, mean ± sd, n=3; b) raw data of median granule size (x 50), depending on SFL,
mean ± sd, n=3; c) contour plot for the x 50 as function of the factors SFL and SS.

Figure 4: a) Raw data of the yield (granules from 180 µm - 1000 µm), depending on SFL, mean ± sd,
n=3; b) contour plot for the yield as function of the factors SFL and SS; c) raw data of the amount of
oversized (granules > 1700 µm), depending on SFL; d) contour plot for the amount of oversized as
function of the factors SFL and SS.

Figure 5: a) Raw data of tensile strength of tablets, depending on SFL, mean ± sd, n = 20; b) contour
plot for the tensile strength as function of the factors SFL and SS; c) raw data of disintegration time of
tablets, depending on SFL, mean ± sd, n = 6; d) contour plot for the disintegration time as function of
the factors SFL and SS.

Figure 6: Relationship between disintegration time (n = 6) and tensile strength (n = 20) of tablets,
mean ± sd; numbers next to data point stand for SFL [g] and SS [1/min].

Figure 7: a) Raw data of power consumption of the extruder, depending on SFL, mean ± sd, n ≥ 300;
b) contour plot for the power consumption of the extruder as function of the factors SFL and SS.

Figure 8: a) Experimental setup at the outlet of the extruder with ExtruVis 2 sensor head pointing on
granules, leaving the extruder; b) example for raw data signal of the ExtruVis 2 sensor and calculated
residence time distribution (E (t)), smoothed and normalized by the area under the curve
(experiment N 8); c) raw data of MRT, depending on SFL; d) contour plot for the MRT as function of
the factors SFL and SS.

Figure 9: In-house built port for pressure and temperature measurements at the top of one screw; in
this setup, pressure was measured directly in front of the 2 nd kneading zone.

Figure 10: a) Raw data of material temperature, depending on SFL; b) contour plot for the material
temperature as function of the factors SFL and SS.
nd
Figure 11: a) Development of material pressure over time on top of one screw in front of the 2
kneading zone for experiment N 1; b) fourier transformed data displaying frequencies of pressure
fluctuations; c) development of material pressure over time for experiment N 4.

Table 1: Raw data of the DOE – factors and response variables.

2
Table 2: Results from the DOE: power of the model (coefficient of determination (R )), coefficient of
2
prediction (Q ), lack of fit (p) and reproducibility; coefficients for factors (SFL, SS, squares of both,
interaction term of both and a constant) to the responses (coefficients ± CI, α = 0.05) and their
respective p-values.

References

[1] FDA-Administration, Guidance for industry – pat – a framework for innovative pharmaceutical
development, manufacturing, and quality assurance, (2004).

[2] R. Meier, M. Thommes, N. Rasenack, M. Krumme, K.P. Moll, P. Kleinebudde, Simplified

formulations with high drug loads for continuous twin-screw granulation, Int. J. Pharm., 496 (2015)
12-23.

20
[3] R. Meier, K.P. Moll, M. Krumme, P. Kleinebudde, Simplified, high drug loaded formulations
containing hydrochlorothiazide for twin-screw granulation, submitted to CHEM-ING-TECH, (2016).

[4] T. Chan Seem, N.A. Rowson, I. Gabbott, M. de Matas, G.K. Reynolds, A. Ingram, Asymmetric
distribution in twin screw granulation, Eur. J. Pharm. Biopharm., 106 (2016) 50-58.

[5] R.M. Dhenge, J.J. Cartwright, D.G. Doughty, M.J. Hounslow, A.D. Salman, Twin screw wet
granulation: Effect of powder feed rate, Adv. Powder Technol., 22 (2011) 162-166.

[6] D. Djuric, P. Kleinebudde, Continuous granulation with a twin-screw extruder: Impact of material
throughput, Pharm. Dev. Technol., 15 (2010) 518-525.

[7] A. Kumar, J. Vercruysse, M. Toiviainen, P.E. Panouillot, M. Juuti, V. Vanhoorne, C. Vervaet, J.P.
Remon, K.V. Gernaey, T. De Beer, I. Nopens, Mixing and transport during pharmaceutical twin-screw
wet granulation: Experimental analysis via chemical imaging, Eur. J. Pharm. Biopharm., 87 (2014)
279-289.

[8] K.T. Lee, A. Ingram, N.A. Rowson, Twin screw wet granulation: The study of a continuous twin
screw granulator using positron emission particle tracking (pept) technique, Eur. J. Pharm.
Biopharm., 81 (2012) 666-673.

[9] M.F. Saleh, R.M. Dhenge, J.J. Cartwright, M.J. Hounslow, A.D. Salman, Twin screw wet
granulation: Binder delivery, Int. J. Pharm., 487 (2015) 124-134.

[10] V. Vanhoorne, B. Bekaert, E. Peeters, T. De Beer, J.P. Remon, C. Vervaet, Improved tabletability
after a polymorphic transition of delta-mannitol during twin screw granulation, Int. J. Pharm., 506
(2016) 13-24.

[11] J. Vercruysse, M. Toiviainen, M. Fonteyne, N. Helkimo, J. Ketolainen, M. Juuti, U. Delaet, I. Van

Assche, J.P. Remon, C. Vervaet, T. De Beer, Visualization and understanding of the granulation liquid
mixing and distribution during continuous twin screw granulation using nir chemical imaging, Eur. J.
Pharm. Biopharm., 86 (2014) 383-392.

[12] H. Li, M.R. Thompson, K.P. O'Donnell, Progression of wet granulation in a twin screw extruder
comparing two binder delivery methods, AIChE J., 61 (2015) 780-791.

[13] Y. Liu, M.R. Thompson, K.P. O'Donnell, Function of upstream and downstream conveying
elements in wet granulation processes within a twin screw extruder, Powder Technol., 284 (2015)
551-559.

[14] M.R. Thompson, K.P. O'Donnell, "Rolling" phenomenon in twin screw granulation with
controlled-release excipients, Drug Dev. Ind. Pharm., 41 (2015) 482-492.
[15] K. Kolter, M. Karl, A. Gryczke, Hot-melt extrusion with basf pharma polymers, 2nd edition ed.,
BASF SE, Pharma Ingredients & Services, 2012.

[16] K. Kohlgrüber, Co-rotating twin-screw extruders - fundamentals, technology, and applications,

in, Hanser Publishers, 2007.

[17] M.F. Saleh, R.M. Dhenge, J.J. Cartwright, M.J. Hounslow, A.D. Salman, Twin screw wet
granulation: Effect of process and formulation variables on powder caking during production, Int. J.
Pharm., 496 (2015) 571-582.

21
[18] J.T. Fell, J.M. Newton, Determination of tablet strength by the diametral-compression test, J.
Pharm. Sci., 59 (1970) 688-691.

[19] J. Vercruysse, U. Delaet, I. Van Assche, P. Cappuyns, F. Arata, G. Caporicci, T. De Beer, J.P.
Remon, C. Vervaet, Stability and repeatability of a continuous twin screw granulation and drying
system, Eur. J. Pharm. Biopharm., 85 (2013) 1031-1038.

[20] J. Vercruysse, E. Peeters, M. Fonteyne, P. Cappuyns, U. Delaet, I. Van Assche, T. De Beer, J.P.
Remon, C. Vervaet, Use of a continuous twin screw granulation and drying system during formulation
development and process optimization, Eur. J. Pharm. Biopharm., 89 (2015) 239-247.

[21] R. Meier, M. Thommes, N. Rasenack, K.P. Moll, M. Krumme, P. Kleinebudde, Granule size
distributions after twin-screw granulation – do not forget the feeding systems, Eur. J. Pharm.
Biopharm., 106 (2016) 59-69.

[22] J. Vercruysse, D. Córdoba Díaz, E. Peeters, M. Fonteyne, U. Delaet, I. Van Assche, T. De Beer, J.P.
Remon, C. Vervaet, Continuous twin screw granulation: Influence of process variables on granule and
tablet quality, Eur. J. Pharm. Biopharm., 82 (2012) 205-211.

[23] M. Fonteyne, J. Vercruysse, D.C. Díaz, D. Gildemyn, C. Vervaet, J.P. Remon, T.D. Beer, Real-time
assessment of critical quality attributes of a continuous granulation process, Pharm. Dev. Technol.,
18 (2013) 85-97.

[24] F. Khan, N. Pilpel, The effect of particle size and moisture on the tensile strength of
microcrystalline cellulose powder, Powder Technol., 48 (1986) 145-150.

[25] A. McKenna, D.F. McCafferty, Effect of particle size on the compaction mechanism and tensile
strength of tablets, J. Pharm. Pharmacol., 34 (1982) 347-351.

[26] J. Mosig, P. Kleinebudde, Critical evaluation of root causes of the reduced compactability after
roll compaction/dry granulation, J. Pharm. Sci., 104 (2015) 1108-1118.

[27] P. Beer, D. Wilson, Z. Huang, M. De Matas, Transfer from high-shear batch to continuous twin
screw wet granulation: A case study in understanding the relationship between process parameters
and product quality attributes, J. Pharm. Sci., 103 (2014) 3075-3082.

[28] D. Djuric, P. Kleinebudde, Impact of screw elements on continuous granulation with a twin-screw
extruder, J. Pharm. Sci., 97 (2008) 4934-4942.

[29] A.S. El Hagrasy, J.R. Hennenkamp, M.D. Burke, J.J. Cartwright, J.D. Litster, Twin screw wet
granulation: Influence of formulation parameters on granule properties and growth behavior,
Powder Technol., 238 (2013) 108-115.

[30] H. Li, M.R. Thompson, K.P. O’Donnell, Understanding wet granulation in the kneading block of
twin screw extruders, Chem. Eng. Sci., 113 (2014) 11-21.

[31] A. Kumar, J. Vercruysse, V. Vanhoorne, M. Toiviainen, P.E. Panouillot, M. Juuti, C. Vervaet, J.P.
Remon, K.V. Gernaey, T.D. Beer, I. Nopens, Conceptual framework for model-based analysis of
residence time distribution in twin-screw granulation, Eur. J. Pharm. Sci., 71 (2015) 25-34.

22
[32] R. Sayin, A.S. El Hagrasy, J.D. Litster, Distributive mixing elements: Towards improved granule
attributes from a twin screw granulation process, Chem. Eng. Sci., 125 (2015) 165-175.

[33] C.C. Gogos, Z. Tadmor, M.H.O. Kim, Melting phenomena and mechanisms in polymer processing
equipment, Adv. Polym. Tech., 17 (1998) 285-305.

23
24
25
26
27
28
29
30
31
32
33
34
Table 1: Raw data of the DOE – factors and response variables.

resulting resulting
volumetric tensile disinte- power
exper- SS powder liquid oversized T (material)
SFL [g] fill-level x50 [µm] yield [%] strength gration consump- MRT [s]
iment [1/min] feed-rate feed-rate [%] [°C]
[%] [MPa] time [s] tion [%]
[g/min] [g/min]
N1 0.15 200 27.0 3.0 7.8 669 72.6 1.2 1.47 1781 29.3 39.2 51.6
N2 0.4 200 72.1 7.9 15.4 999 48.5 14.5 1.25 1966 45.0 34.7 59.9
N3 0.15 500 67.6 7.4 9.1 803 54.6 0.4 1.77 1439 21.6 25.5 61.4
N4 0.4 500 180.2 19.8 19.4 1019 46.8 14.1 1.43 2208 44.7 21.3 79.0
N5 0.09825 350 31.0 3.4 7.0 790 62.0 5.4 1.39 1862 23.2 42.6 56.3
N6 0.45175 350 142.4 15.7 19.5 1167 37.3 25.5 1.35 2174 49.6 22.3 73.8
N7 0.275 138 34.2 3.8 11.9 773 66.6 2.1 1.50 1654 37.9 46.7 51.4
N8 0.275 562 139.3 15.3 12.0 776 61.2 4.4 1.77 1514 27.1 16.9 67.7
N9 0.275 350 86.7 9.5 10.4 750 69.1 2.2 1.65 1552 32.0 25.9 64.1
N 10 0.275 350 86.7 9.5 11.8 738 66.1 3.4 1.67 1667 29.8 27.6 62.8
N 11 0.275 350 86.7 9.5 13.6 718 68.6 2.6 1.70 1529 31.1 26.1 63.1
N 12 0.06 350 18.9 2.1 5.8 736.2 62.6 7 1.28 2046 23.2 53.3 61.0
N 13 0.2 350 63.1 6.9 11.1 668 70.5 2.7 1.60 1538 27.5 26.2 60.2
N 14 0.35 350 110.4 12.1 14.6 932 53.8 9 1.47 1867 39.0 22.9 66.8

35
Table 2: Results from the DOE: power of the model (coefficient of determination (R2)), coefficient of prediction (Q2), lack of fit (p) and reproducibility; coefficients
for factors (SFL, SS, squares of both, interaction term of both and a constant) to the responses (coefficients ± CI, α = 0.05) and their respective p-values.

volumetric tensile power

parameter / oversized disintegration log - MRT T (material)
fill-level x50 [µm] yield [%] strength consump-
coefficient [%] time [s] [log s] [°C]
[%] [MPa] tion [%]
R2 0.942 0.933 0.944 0.975 0.870 0.926 0.988 0.915 0.973
2
Q 0.894 0.853 0.783 0.949 0.637 0.717 0.947 0.718 0.886
p 0.857 0.105 0.183 0.490 0.096 0.490 0.473 0.068 0.134
reproducibility 0.855 0.988 0.976 0.992 0.979 0.916 0.985 0.989 0.992
SFL [g] 4.40 ± 0.79 151.89 ± 30.37 -9.46 ± 2.26 6.92 ± 0.82 -0.07 ± 0.05 173.9 ± 58.80 9.52 ± 0.89 -0.07 ± 0.04 5.79 ± 1.13
p-value 2.20E-07 5.85E-07 1.12E-05 1.19E-09 0.0079 8.95E-05 7.71E-09 0.0012 2.44E-06
SS [1/min] 0.67 ± 0.90 19.78 ± 34.44 -3.42 ± 2.57 - 0.11 ± 0.06 -37.25 ± 66.66 -2.92 ± 1.01 -0.13 ± 0.04 6.49 ± 1.28
p-value 0.1275 0.2294 0.0153 - 0.0016 0.2380 0.0002 5.30E-05 2.63E-06
SFL * SFL [g2] 0.53 ± 0.80 96.56 ± 30.64 -7.94 ± 2.37 5.62 ± 0.83 -0.15 ± 0.05 239.19 ± 59.32 2.74 ± 0.93 0.05 ± 0.04 1.91 ± 1.19
p-value 0.1702 3.62E-05 5.65E-05 1.17E-08 4.22E-05 7.65E-06 0.0001 0.0101 0.0060
SS * SS [1/min2] - - -1.89 ± 2.78 - - - 0.68 ± 1.09 0.02 ± 0.04 -1.73 ± 1.39
p-value - - 0.1569 - - - 0.1875 0.3162 0.0210
SFL * SS [g/min] - - 4.07 ± 3.63 - - 146.00 ± 94.27 1.87 ± 1.42 - 2.32 ± 1.81
p-value - - 0.0321 - - 0.0067 0.0166 - 0.0182
constant 12.19 ± 0.94 762.28 ± 35.82 66.56 ± 3.42 2.91 ± 0.97 1.64 ± 0.06 1593.08 ± 69.35 31.42 ± 1.34 1.40 ± 0.05 62.90 ± 1.71
p-value 5.51E-11 4.20E-13 6.69E-11 3.76E-05 2.67E-14 1.82E-12 1.54E-11 5.96E-13 4.14E-13

36