Accepted Manuscript Title: Towards a real time release approach for manufacturing tablets using NIR spectroscopy Author: Aude Pestieau Fabrice Krier Gr´egory Thoorens Ana¨ıs Dupont Pierre-Franc¸ois Chavez Eric Ziemons Philippe Hubert Brigitte Evrard PII: DOI: Reference:
S0731-7085(14)00232-5 http://dx.doi.org/doi:10.1016/j.jpba.2014.05.002 PBA 9565
To appear in:
Journal of Pharmaceutical and Biomedical Analysis
Received date: Revised date: Accepted date:
24-2-2014 28-4-2014 1-5-2014
Please cite this article as: A. Pestieau, F. Krier, G. Thoorens, A. Dupont, P.-F. Chavez, E. Ziemons, P. Hubert, B. Evrard, Towards a real time release approach for manufacturing tablets using NIR spectroscopy, Journal of Pharmaceutical and Biomedical Analysis (2014), http://dx.doi.org/10.1016/j.jpba.2014.05.002 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.
1
Towards a real time release approach for manufacturing tablets using NIR spectroscopy
2 Aude Pestieau1, Fabrice Krier1, Grégory Thoorens1, Anaïs Dupont1, Pierre-François Chavez2,
4
Eric Ziemons2, Philippe Hubert2, Brigitte Evrard1
ip t
3
5 6
1
7
of Liège, 4000 Liège, Belgium
8
2
9
Liège, 4000 Liège, Belgium
10
Corresponding author
cr
Laboratory of Pharmaceutical Technology, Department of Pharmacy, C.I.R.M., University
an
us
Laboratory of Analytical Chemistry, Department of Pharmacy, C.I.R.M., University of
Aude Pestieau
12
[email protected] 13
+3243664306
14
CHU, Tour 4, 2nd floor, Laboratory of Pharmaceutical Technology, Department of Pharmacy,
15
University of Liège, Avenue de l'Hôpital, 1, 4000 Liège, Belgium
17
d
te
Ac ce p
16
M
11
Highlights
18
‐
We used NIR Spectroscopy to control in‐line blends uniformity.
19
‐
We used NIR Spectroscopy to evaluate the conformity of paracetamol tablets.
20
‐
Mixing time of blend could be reduced because NIR showed when it was homogenous.
21
‐
Tablet NIR analyses successfully allowed the prediction of their conformity.
22
‐
The benefit of NIR methods is significant for reducing batch release time.
23 1 Page 1 of 23
24
25
1. Introduction Conventional pharmaceutical manufacturing is generally accomplished using batch
27
processing with laboratory testing conducted on collected samples to evaluate quality. This
28
conventional approach has been successful in providing quality drug products to the public.
29
However, quality controls on process materials and on finished products are time-consuming
30
and often require a lot of sample preparation steps and laboratory work. For a few years,
31
significant opportunities have existed for improving pharmaceutical development,
32
manufacturing and quality assurance through innovation in product and process development,
33
process analysis, and process control. For example, the publication of the Process Analytical
34
Technology (PAT) initiative [1] by the Food and Drug Administration (FDA)has increased
35
the interest for PAT tools in the pharmaceutical industry. One of the principles described in
36
the guidance document is Real Time Release (RTR) which can be defined as the ability to
37
evaluate and ensure the acceptable quality of in-process and/or final product based on process
38
data [2]. The combined process measurements and other test data gathered during the
39
manufacturing process can serve as the basis for RTR of the final product and would
40
demonstrate that each batch conforms to established regulatory quality attributes. According
41
to the European Medicines Agency (EMA), “Real Time Release Testing (RTRT) will, in
42
general, comprise a combination of process controls which may utilise PAT tools e.g. NIR
43
and Raman spectroscopy (usually in combination with multivariate analysis), together with
44
the control of relevant material attributes.”[3]. For example, a well-known application of this
45
concept is the use of NIRS to control the blend uniformity with an in-line measurement. This
46
type of application is very useful in the context of RTR of the product by reducing the batch
47
release time [4]. Moreover, the blend homogeneity control is a crucial step to ensure active
Ac ce p
te
d
M
an
us
cr
ip t
26
2 Page 2 of 23
content uniformity in the end product, especially in a direct compression process, because it is
49
the only step before compression. Indeed, the present manufacturing process time compared
50
to the time spent on quality testing after manufacturing is very low, so RTR offers some
51
significant benefits to the manufacturer [5].
ip t
48
52
To perform this implementation, the use of analytical techniques able to provide accurate results in a simple and a rapid way is required. Due to its non-destructive nature and its
54
immediate (real time) delivery of results without sample preparation, NIRS is a suitable
55
analytical technique for this goal. This is probably why the European Pharmacopeia has
56
decided to add a monograph on this technique in the latest published version (version 8.0.)
57
[6].
us
an
However, after the development of the NIRS method, another major step was validation
M
58
cr
53
enabling its routine use. Validation is based on guidelines of the International Conference on
60
Harmonisation (ICH) Q2 (R1) [7] and is a crucial and mandatory step in the lifecycle of an
61
analytical method. This step is needed to prove that the spectroscopic analytical method is
62
suitable for its intended use and consequently, to show the reliability of the results obtained
63
within well-defined limits. It integrates all the useful required validation criteria such as
64
accuracy, trueness, precision, limits of quantification, range and linearity. The approach based
65
on the accuracy profile makes the visual and reliable representation of the future
66
performances of the analytical method possible and thus, enables better risk management [8].
67
In this manner, it also complies with the ICH Q9 regulatory documents [9].
te
Ac ce p
68
d
59
In the pharmaceutical industry, NIRS combined with chemometrics have already been
69
applied to acquire information on a particular quality attribute or on multiple predicates.
70
There are many examples in the literature of acquiring information on a particular quality
71
attribute such as the active content [5, 8], the blend homogeneity [10], the coating level [11,
72
12], the moisture content [13], the polymorphic transformations [14], the particle size of
3 Page 3 of 23
powders [15], the flow property of powder[16], the tablet mechanical strength [17] and the
74
dissolution profiles [18].However, for all these applications, only few analytical methods have
75
been validated. In fact, the validation step was only performed on the active content
76
determination in tablets [5] or in pellets [8] and on the moisture content determination in
77
pellets [13]. A number of successful combinations of two or three qualitative and/or
78
quantitative simultaneous determinations are already described. Simultaneous determination
79
of content uniformity and tablet hardness [19] or blend uniformity, content uniformity and
80
coating thickness [4] or tablet hardness, content uniformity and dissolution test [20] on intact
81
tablets can be found. However in these studies, only the NIR method for the determination of
82
the API content was validated. Regarding oral dosage forms, the literature also shows the
83
paracetamol determination in low-dose pharmaceutical syrup [21]or in matrix carrier
84
produced by hot-melt extrusion [22]. Of course, this technique can be applied to other dosage
85
forms that are not intended for oral administration such as the in-line monitoring of the
86
implant manufacturing process [23] and the determination of API and preservatives in a gel
87
[24].
89 90
cr
us
an
M
d
te
In the present study, tablets for oral administration were selected. At first NIRS has been
Ac ce p
88
ip t
73
used to control the blend uniformity. It was an in-line measurement. Following the manufacture of tablets, we attempted to develop and validate NIR methods
91
that can replace the conventional techniques usually used to test tablets following their
92
manufacture, ie. the European Pharmacopeia tests for content uniformity, hardness,
93
disintegration time and friability. To the best of our knowledge, no literature describes the use
94
of NIR to replace four Pharmacopeia tests simultaneously.
4 Page 4 of 23
95
2. Materials and methods 2.1. Chemicals
97
Paracetamol (CompapTM PVP3) was provided by Covidien (Mallinckrodt, USA).
ip t
96
Microcrystalline cellulose (Avicel®PH-102) was provided by FMCBioPolymer (Cork,
99
Ireland). Sodium starch glycolate (Glycolys®) was supplied by Roquette (Lestrem, France).
100
cr
98
Magnesium stearate was obtained from Fagron (Waregem, Belgium).
All solvents used in the reference methods were of analytical grade. Methanol was
102
purchased from J.T. Baker® (Deventer, Netherlands). Water was purified by a Millipore®
103
system (18.2 MΩ/cm resistivity, Milli-Q) before filtration through a 0.22 µm Millipore
104
Millipak® – 40 disposable filter units (Millipore Corporation, USA).
M
an
us
101
2.2. Tablets manufacturing
106
The tablet formulation was developed for direct compression and based on paracetamol as
107
API. This API was previously wet granulated with 3% of polyvinylpyrrolidone and therefore
108
easily compressible. The excipients that have been added to this Paracetamol CompapTM
109
PVP3 are a binder (microcrystalline cellulose), a super-disintegrant (sodium starch glycolate)
110
and a lubricant (magnesium stearate). This formulation represents a typical formulation for
111
the manufacture of tablets by direct compression and provides optimum flow properties.
te
Ac ce p
112
d
105
Paracetamol tablets were manufactured by direct compression with an eccentric press
113
(AC27, GEA-Courtoy, Halle, Belgium). Blends were mixed in a high shear mixer Gral-
114
10®(GEA-Colette, Wommelgem, Belgium). The process bowl has a volume of 10 l. The
115
blend load was 2.4 kg. The mixing was performed for 400 s without magnesium stearate at
116
200 rpm. The lubricant was then added and mixed during 1 min. Flat face bevel edge tablets
117
were obtained using round punches with a diameter of 10 mm. Targeted tablet weight was
118
fixed at 350 mg. 5 Page 5 of 23
On the one hand, three different active pharmaceutical ingredient concentrations were
120
manufactured: 80, 100 and 120 % of a predetermined dosage (242.5 mg of paracetamol) at a
121
compaction pressure of 45 kg/cm2. To obtain these different concentrations, the weight of the
122
API was increased or decreased relative to the target concentration. As can be seen in Table 1,
123
the quantity of microcrystalline cellulose and sodium starch glycolate were adapted to
124
maintain a tablet weight of 350 mg while the quantity of magnesium stearate remained
125
constant (1 % w/w). Indeed, during an industrial production of tablets, uniformity of mass is a
126
parameter which is very often controlled. In this way a de-mixing of paracetamol and
127
excipients during tableting should be simulated.
130
cr
us
an
129
On the other hand, two additional compaction pressures were used for tablets with an API concentration of 100 %: 25 and 65 kg/cm2.
M
128
ip t
119
For the calibration and external validation, one batch of tablets was manufactured each day during three days (D1, D2 and D3) for each concentration (C80, C100, C120) and for each
132
compaction pressure (CP25, CP45, CP65).One batch consisted of approximately 6,000 tablets.
te
d
131
2.3. FT-NIR equipment
134
Intact tablets were analyzed by transmission mode with a multipurpose analyzer Fourier
Ac ce p
133
135
transform near infrared spectrometer (MPA, Bruker Optics, Ettlingen, Germany). The spectra
136
were collected with the Opus software 6.5 (MPA, Bruker Optics, Ettlingen, Germany). Each
137
spectrum was an average of 32 scans and the resolution was 8 cm-1 over the range from 3600
138
to 14000 cm-1.For in-line monitoring of the blending process, the NIR spectrometer was
139
equipped with a NIR reflectance probe (Series 400 Diffuse Reflectance probe, Precision
140
Sensing Devices, Massachusetts, USA) interfaced with the mixing bowl. Each spectrum was
141
an average of 4 scans and the resolution was 16 cm-1 over the range from 12500 to 4000 cm-1.
142
The time required for a NIR measurement was 1.7 sec and the time interval between measures
143
was 1 sec. 6 Page 6 of 23
144
2.4. Reference methods
145
A conventional evaluation blend uniformity method was performed as a reference [25].
ip t
146
2.4.1. Powder blend uniformity
This method involved blending for a pre-determined length of time (400 s), stopping the
148
blender, and manually withdrawing a powder blend sample representative of a unit dose from
149
the bin with a single compartment end-sampling thief probe. Ten locations were selected from
150
two depths along the axis of the bowl. The samples were then analysed using a USP high
151
performance liquid chromatography (HPLC) method.
us
This HPLC method uses a GraceSmart® steel column 25 cm long x 4,6 mm ID RP 18 with
an
152
cr
147
particles of 5 µm, a mobile phase consisting of methanol-water (3:1, v/v) flowing at 1.5
154
ml/min, an injected volume of 10 µl, a chromatographic run time of 6 min and a detection
155
wavelength of 243 nm. The HPLC apparatus used was an Agilent 1100® (Santa Clara, United
156
State).
d
M
153
Before injection, each sample was weighed, dissolved in 20.0 ml of mobile phase, and
158
sonicated for 5 min. An aliquot of 0.5 ml was then diluted to 200.0 ml with the mobile phase.
159
This solution was filtered through a 0.22 µm polyvinylidene fluoride (PDVF)filter. A standard
160
was prepared from pure paracetamol in the same dilution solvent.
Ac ce p
161
te
157
We considered samples to be uniform if the drug concentration of each individual sample
162
was within 10% of the average concentration and the relative standard deviation (RSD) was
163
less than 5%.
164
2.4.2. Content uniformity
165
The same USP HPLC method as described above was used as a reference method to
166
determine tablet content uniformity. Each tablet was weighed, dissolved in 100 ml of mobile
167
phase, sonicated for 15 min and diluted to 250 ml with the same solvent. An aliquot of 1 ml 7 Page 7 of 23
168
was then diluted to 100 ml with the mobile phase. This solution was filtered through a 0.22
169
µm PDVF filter before injection.
2.4.3. Tablet hardness
171
Tablet hardness was determined using a Sotax HT1®(Allschwil, Switzerland). This
ip t
170
apparatus consists of two jaws facing each other, one of which moves towards the other. The
173
flat surfaces of the jaws are perpendicular to the direction of movement and the tablet was
174
placed horizontally between the jaws [26].
us
cr
172
2.4.4. Disintegration time
176
For the determination of tablet disintegration time, a Sotax DT3® (Allschwil,
177
Switzerland) was used. This apparatus consists of a basket-rack assembly, a 1 liter, low-form
178
beaker for the immersion fluid, a thermostatic arrangement for heating the fluid at 37 ± 2 °C
179
and a device for raising and lowering the basket in the immersion fluid at a constant
180
frequency rate. The basket-rack assembly moves vertically along its axis. One dosage unit
181
was placed in one of the 6 tubes of the basket and time where the dosage unit has
182
disintegrated completely was noted. This method corresponds to the reference [27].
184
M
d
te
Ac ce p
183
an
175
2.4.5. Tablet friability
For tablet friability, the method used corresponds to the reference [28]with a
185
Friabilitator Sotax F1/2® (Allschwil, Switzerland). This equipment consists of a drum made of
186
transparent synthetic polymer. The tablets are tumbled at each rotation of the drum by a
187
curved projection that extends from the middle of the drum to the outer wall. The drum is
188
attached to the horizontal axis of a device that rotates 100 times in 4 minutes (25 ± 1 rpm).
189
Thus, at each rotation the tablets roll or slide and fall onto the drum wall or onto each other. A
190
sample of whole tablets corresponding as near as possible to 6.5 g (19 tablets) was tested.
8 Page 8 of 23
These tablets were carefully dusted and then weighed prior to testing. At the end of the test,
192
the tablets were removed from the drum, dusted and then reweighed. If tablets that were
193
obviously cracked, cleaved, or broken were present in the tablet sample after tumbling, the
194
sample failed the test. A maximum loss of mass less than 1.0 % is considered acceptable for
195
most products.
ip t
191
2.5. Development of calibration models
us
197
cr
196
2.5.1. Powder blend uniformity
199
For in-line monitoring, the mixing bowl was perforated to allow the introduction of a
an
198
NIR fibre probe. This equipment was widely described by Bodson in [29].
201
The conformity test was used to follow the mixing kinetics [29]. The conformity test is an
202
easy method to test the deviation of measured NIR spectra. To set limits, samples of the final
203
product are needed as reference spectra which belong to at least one batch or one production
204
cycle. These reference spectra vary within the accepted range of specifications. The NIR
205
spectra of these samples reflect the different sample variations and give a confidence band in
206
the spectral range. To pass the conformity test, the spectrum of a new sample has to be within
207
this confidence band at each wavelength. The conformity tests were computed with Opus
208
software 6.5(Bruker Optics, Ettlingen, Germany) and the Conformity Index (CI) was
209
calculated using the following equation:
d
te
Ac ce p
210
M
200
Conformity Index = (Āref,i - Asamp,i) / σref,i[30]
211
Āref,i : Average of absorbance values of reference spectra at wavenumber i.
212
Asamp,i : Absorbance value of test sample spectrum at wavenumber i.
213
σref,i : Standard deviation of absorbance values of reference spectra at
214
wavenumber i. 9 Page 9 of 23
215
The last ten spectra of each blend were selected as reference spectra.
216
The selected algorithm added up all y-values above the CI limit and divided this sum by the
217
total number of data points within the frequency ranges selected [30].
219
2.5.2. Pharmacopeia tests
ip t
218
To build the NIR model for content uniformity, twenty-seven tablets were analysed.
Indeed, as mentioned above three levels of concentrations (C80, C100, C120) were produced
221
each day during three days (D1, D2D3) at a compaction pressure of 45 kg/cm2(CP45)and 3
222
tablets were measured per production (3 repetitions). To build the NIR models for tablet
223
hardness and disintegration time, twenty-seven tablets were also analysed. One level of
224
concentration (C100) was compressed each day during three days (D1, D2 D3) at three
225
compaction pressures (CP25, CP45, CP65) and 3 tablets were measured per production (3
226
repetitions).Theseproductionswerealso used for tabletfriability testing,butfor this model19
227
tabletswere analyzedper production. So, hundred seventy one tablets were analysed to build it.
228
For each validated model the same number of tablets as used for calibration was produced
230
us
an
M
d
te
and measured for the external validation.
Ac ce p
229
cr
220
In a first step, a PLS regression model was built for each test using calibration samples.
231
Pre-treatments for offset baseline correction, sample normalization and variable centering
232
were performed as Standard Normal Variate (SNV), Multiplicative Scatter Correction (MSC),
233
Mean Centering, derivatives, detrend and smoothing. Cross-validation based on random
234
subsets was carried out to select the optimal number of PLS factors for API concentrations,
235
tablet hardness and disintegration time tests. For tablet friability, a qualitative approach was
236
adopted by a KNN analysis, whichallowed us to detect non-conforming tablets.
237
All data were mean centred and the number of latent variables and of each PLS model was
238
selected based on Root Mean Square Error of Cross-Validation (RMSECV) versus latent
239
variable plots. 10 Page 10 of 23
240
The best models were chosen based on their RMSEP.
241
PLS models were built using PLS_Toolbox 7.0.3 (Wenatchee, WA, USA) running on Matlab R2013a (The Mathworks, Natick, MA, USA).
243
3. Results and discussion
ip t
242
3.1. Powder blend uniformity
245
To confirm blend uniformity of the final blends, the reference method described above
us
cr
244
was used. The spectral areas from 9033 to 8277 cm-1and from 7212 to 4242 cm-1,
247
corresponding to absorption bands of paracetamol, were selected to compute the conformity
248
tests. SNV was used as pre-processing. As the conformity curve represents the spectral
249
variability during the blending step, the blend can be assessed as homogeneous when the
250
curve reaches a stable level. In Figures 1, 2 and 3, green points are the reference spectra
251
corresponding to the end of the mixing process when the blend is expected to be
252
homogeneous, while blue points are sample spectra of the blend from the beginning to the end
253
of the mixing process. As can also be seen in these figures, the 80%, 100% and 120% blends
254
seem to be homogeneous before the end of the mixing time. This homogeneity has been
255
confirmed by thieved sampling after the blending process and HPLC analysis with the
256
reference method described above. The results obtained with this technique are shown in
257
Table 2. As can be seen in this table, the means of sample content uniformity were close to
258
the target value and the coefficient of variation was less than 2.5 % for all blends. These
259
results show that blends are effectively homogenous after 400 s of mixing. Since all the
260
conformity curves reach the final homogeneity state before the end of the process, it can be
261
concluded that the blends are homogeneous before the 400 s and therefore, this time could be
262
reduced.
Ac ce p
te
d
M
an
246
11 Page 11 of 23
3.2. NIR models for Pharmacopeia tests
264
As can be seen in Table 4, a spectral range was manually selected for the calibration of
265
each NIR model. In a general way, as seen in Figure 4, the region above 7500 cm-1(or 9000
266
cm-1) was selected because below this wavenumber, the detector signal becomes noisy.
267
ip t
263
This table also shows the spectral pre-treatment used, the number of PLS factors selected and the R2, RMSEC and RMSEPobtained for each test. In fact, the validation step using the
269
accuracy profile approach was performed for the content uniformity method and for the tablet
270
hardness method. Due to the uncertainty of the reference method for disintegration time
271
(subjective visual method), the validation step was based on the values of RMSEC and
272
RMSEP. Finally, regarding tablet friability, a qualitative approach was performed in order to
273
detect non-conforming tablets.
276 277
us
an
M
d
Table 4 shows that three PLS factors were selected for the NIR model with regard to the
te
275
3.2.1. Content uniformity
lowest value of RMSEC. It was at 0.415for all three factors. This NIR method was validated and the accuracy and risk profiles were evaluated. The
Ac ce p
274
cr
268
278
acceptance limits were set at ±10% in accordance with USP recommendations [31].As it can
279
be seen on Figure 5, the method was successfully validated for the entire dosage interval from
280
80% to 120%. Furthermore, the relative bias was very close to 0 % for each concentration
281
level and the distribution of the random errors was quite low too. In fact we can say that this
282
method presents excellent trueness and precision.
283 284
3.2.2. Tablet hardness As a reminder, three compaction pressures were used to build this model. In fact, these
285
compaction pressures were selected according to the mechanical constraints of the eccentric
286
press and the formulation. Indeed, a compaction pressure below 25 kg/cm2 gave weak tablets, 12 Page 12 of 23
which were difficult to handle and had very low hardness values. Conversely, a compaction
288
pressure above 65 kg/cm2 would produce excessive mechanical stress to the tablet press.
289
Regarding the third compaction pressure, it was the median value between these two
290
extremes, that is to say 45 kg/cm2. However, this compaction pressure provides the
291
appropriate hardness to this type of tablet for oral use (80 N – 120 N). Indeed, a lower
292
hardness would lead to tablets that are too friable and would not withstand the handling
293
required for their packaging. Conversely, a higher hardness would reduce disintegration time
294
and may decrease the bioavailability and the effectiveness of the drug.
cr
us
We can see in Figure 6 that the acceptance limits were set at ± 25% for this validation
an
295
ip t
287
method. The β-expectation tolerance limits are included in these acceptance limits for
297
hardness values between 60 N and 160 N. As explained above, these hardness values are in
298
agreement with the expected values for this type of tablet. It is reasonable to admit that tablets
299
having a hardness of less than 60 N or greater than 160 N will be considered as non-
300
conforming. In this case, it is no longer necessary to try to accurately quantify their hardness.
301
It can be concluded that, during routine use, this method will provide results with adequate
302
accuracy. The risk of finding future results outside the ±25% acceptance limit is below 5%,
303
which is the chosen maximum risk level (α-risk: 5%).
d
te
Ac ce p
304
M
296
3.2.3. Disintegration time
305
The PLS model chosen for tablet disintegration time required three factors. The NIR
306
model was tested by the calibration and the validation sets. Figure 7 shows NIR predictions
307
versus reference method results for these two sets. The calibration set is represented by the
308
red triangles and the validation set by the black points. The values of the RMSEC and the
309
RMSEP were 1.67 and 4.18 respectively. These results indicate the robustness and the global
310
accuracy of the NIR model. It is important to keep in mind that the reference method is a
311
visual subjective method. Indeed, the disintegration end-point is left to the judgment of the 13 Page 13 of 23
operator and is defined as “that state in which any residue of the unit, except fragments of
313
insoluble coating or capsule shell, remaining on the screen of the test apparatus or adhering
314
to the lower surface of the discs”[27]. Thus, it is difficult to obtain accurate and robust
315
quantitative results with this technique. However, the conformity of uncoated tablets for the
316
disintegration test is achieved when disintegration time is less than 15 min. In our case, all
317
tested tablets used to build the model disintegrated before one minute of testing was complete.
318
Thus, one can reasonably assume that the NIR model will easily detect non-conforming
319
tablets. To test this hypothesis, it would be necessary to have tablets with poor disintegration.
320
This seems to be difficult to obtain with the current formulation. Indeed, compaction pressure
321
could not be increased without exceeding the maximum mechanical stress of the tablet press
322
and tablets were uncoated.
M
an
us
cr
ip t
312
3.2.4. Tablet friability
324
For this test, according to [28], a sample of whole tablets corresponding as near as
d
323
possible to 6.5 g have to be tested. This engenders the analysis of 19 tablets in a single
326
friability test. Under these conditions, it seems to be difficult to do a quantitative analysis. The
327
classification for this test was based on KNN, a qualitative approach. To account for scaling
328
effects and offset effects, a MSC was applied as spectra pre-treatment.
Ac ce p
329
te
325
Figure 7 shows the results of the multivariate pattern recognition method KNN. With
330
this approach, it can be seen that the NIR method was able to classify tablets on two levels
331
(levels 1 and 2 in Fig.8). After comparison of these results with those of the reference method,
332
it appeared that these two levels correspond to samples that failed the friability test (in red)
333
and samples that passed it (in green). Thus this qualitative approach allows us to determine
334
whether or not tablets conform in terms of friability.
14 Page 14 of 23
335 336
4. Conclusions During this study, we showed that NIRS is an interesting tool as a RTR system. First, we developed a real-time, non-invasive, in-line qualitative blend uniformity method. The
338
homogeneity has been successfully confirmed by thieved sampling after the blending process
339
(400 s) and HPLC analyses. Thanks to the NIRS method it was shown that all the conformity
340
curves reach the final homogeneity state before the end of the process. In fact, with the use of
341
NIR for this application, an overview of the blend homogeneity over time is given, whereas
342
by the reference method, only specific results can be obtained. In our case, NIR results
343
showed that the blends are homogenous before the 400 s and therefore, this time could be
344
reduced.
cr
us
an
After that, we successfully developed and validated four off-line NIRS methods intended
M
345
ip t
337
to replace the conventional Pharmacopeia tests for content uniformity, tablet hardness,
347
disintegration time and tablet friability. Taking into account that the quality control of
348
finished products is time-consuming, this nondestructive method offers significant
349
advantages, especially in terms of batch release time. Thanks to the rapidity and the ease of
350
use of this method once developed, NIRS can be considered as a quality control technique and
351
a cost effective solution.
te
Ac ce p
352
d
346
353
Abbreviations
354
NIRS
Near-Infrared Spectroscopy
355
API
Active Pharmaceutical Ingredient
356
PLS
Partial Least Squares
357
RMSEC
Root Mean Squared Error of Calibration
358
RMSEP
Root Mean Squared Error of Prediction 15 Page 15 of 23
KNN
K-Nearest-Neighbors
360
PAT
Process Analytical Technology
361
FDA
Food and Drug Administration
362
RTR
Real Time Release
363
EMA
European Medicines Agency
364
RTRT
Real Time Release Testing
365
ICH
International Conference of Harmonisation
366
USP
United State Pharmacopeia
367
HPLC
High Performance Liquid Chromatography
368
PVDF
PolyVinyliDene Fluoride
369
RSD
Relative Standard Deviation
370
CI
Conformity Index
371
SNV
Standard Normal Variate
372
MSC
Multiplicative Scatter Correction
373
RMSECV
Root Mean Squared Error of Cross-Validation
375
cr us
an
M
d
te
Ac ce p
374
ip t
359
Acknowledgements
376
Thanks to SMB Technology, FMCBioPolymer and Roquette for supplying us with some
377
excipients.
378 379 380 381 382 383 384 385 386
References [1] FDA, Guidance for Industry PAT - A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance, (2004). [2] E.T.S. Skibsted, J.A. Westerhuis, A.K. Smilde, D.T. Witte, Journal of Pharmaceutical and Biomedical Analysis, Examples of NIR based real time release in tablet manufacturing, 43 (2007) 1297-1305. [3] EMA, Guideline on Real Time Release Testing, (2012). 16 Page 16 of 23
te
d
M
an
us
cr
ip t
[4] J.J. Moes, M.M. Ruijken, E. Gout, H.W. Frijlink, M.I. Ugwoke, International Journal of Pharmaceutics, Application of process analytical technology in tablet process development using NIR spectroscopy: Blend uniformity, content uniformity and coating thickness measurements, 357 (2008) 108-118. [5] C. Bodson, E. Rozet, E. Ziemons, B. Evrard, P. Hubert, L. Delattre, Journal of Pharmaceutical and Biomedical Analysis, Validation of manufacturing process of Diltiazem HCl tablets by NIR spectrophotometry (NIRS), 45 (2007) 356-361. [6] Monograph 2.2.40. Near-Infrared Spectroscopy, in: Eur.Ph. 8.0., 2014. [7] ICH, Topic Q2 (R1): Validation of Analytical Procedures: Text and Methodology, International Conference on Harmonization of Technical Requirements for registration of Pharmaceuticals for Human Use, (2005). [8] J. Mantanus, E. Ziémons, E. Rozet, B. Streel, R. Klinkenberg, B. Evrard, J. Rantanen, P. Hubert, Talanta, Building the quality into pellet manufacturing environment – Feasibility study and validation of an in-line quantitative near infrared (NIR) method, 83 (2010) 305-311. [9] ICH, Topic 9 : Quality risk management, International Conference on Harmonisation of Technical requirements for registration of pharmaceuticals for human use, (2005). [10] Y. Sulub, B. Wabuyele, P. Gargiulo, J. Pazdan, J. Cheney, J. Berry, A. Gupta, R. Shah, H. Wu, M. Khan, Journal of Pharmaceutical and Biomedical Analysis, Real-time on-line blend uniformity monitoring using near-infrared reflectance spectrometry: A noninvasive offline calibration approach, 49 (2009) 48-54. [11] M.-J. Lee, D.-Y. Seo, H.-E. Lee, I.-C. Wang, W.-S. Kim, M.-Y. Jeong, G.J. Choi, International Journal of Pharmaceutics, In line NIR quantification of film thickness on pharmaceutical pellets during a fluid bed coating process, 403 (2011) 66-72. [12] C.V. Möltgen, T. Puchert, J.C. Menezes, D. Lochmann, G. Reich, Talanta, A novel inline NIR spectroscopy application for the monitoring of tablet film coating in an industrial scale process, 92 (2012) 26-37. [13] J. Mantanus, E. Ziémons, P. Lebrun, E. Rozet, R. Klinkenberg, B. Streel, B. Evrard, P. Hubert, Analytica Chimica Acta, Moisture content determination of pharmaceutical pellets by near infrared spectroscopy: Method development and validation, 642 (2009) 186-192. [14] M. Blanco, M. Alcalá, J.M. González, E. Torras, Analytica Chimica Acta, Near infrared spectroscopy in the study of polymorphic transformations, 567 (2006) 262-268. [15] M. Blanco, A. Peguero, Talanta, An expeditious method for determining particle size distribution by near infrared spectroscopy: Comparison of PLS2 and ANN models, 77 (2008) 647-651. [16] X. He, X. Han, N. Ladyzhynsky, R. Deanne, Powder Technology, Assessing powder segregation potential by near infrared (NIR) spectroscopy and correlating segregation tendency to tabletting performance, 236 (2013) 85-99. [17] J.D. Kirsch, J.K. Drennen, Journal of Pharmaceutical and Biomedical Analysis, Nondestructive tablet hardness testing by near-infrared spectroscopy: a new and robust spectral best-fit algorithm, 19 (1999) 351-362. [18] Y. Hattori, M. Otsuka, Vibrational Spectroscopy, NIR spectroscopic study of the dissolution process in pharmaceutical tablets, 57 (2011) 275-281. [19] M. Blanco, M. Alcalá, Analytica Chimica Acta, Content uniformity and tablet hardness testing of intact pharmaceutical tablets by near infrared spectroscopy: A contribution to process analytical technologies, 557 (2006) 353-359. [20] M. Blanco, M. Alcalá, J.M. González, E. Torras, Journal of Pharmaceutical Sciences, A process analytical technology approach based on near infrared spectroscopy: Tablet hardness, content uniformity, and dissolution test measurements of intact tablets, 95 (2006) 2137-2144.
Ac ce p
387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
17 Page 17 of 23
te
d
M
an
us
cr
ip t
[21] E. Ziémons, J. Mantanus, P. Lebrun, E. Rozet, B. Evrard, P. Hubert, Journal of Pharmaceutical and Biomedical Analysis, Acetaminophen determination in low-dose pharmaceutical syrup by NIR spectroscopy, 53 (2010) 510-516. [22] P.R. Wahl, D. Treffer, S. Mohr, E. Roblegg, G. Koscher, J.G. Khinast, International Journal of Pharmaceutics, Inline monitoring and a PAT strategy for pharmaceutical hot melt extrusion, 455 (2013) 159-168. [23] F. Krier, J. Mantanus, P.-Y. Sacré, P.-F. Chavez, J. Thiry, A. Pestieau, E. Rozet, E. Ziemons, P. Hubert, B. Evrard, International Journal of Pharmaceutics, PAT tools for the control of co-extrusion implants manufacturing process, 458 (2013) 15-24. [24] M. Blanco, M. Alcalá, M. Bautista, European Journal of Pharmaceutical Sciences, Pharmaceutical gel analysis by NIR spectroscopy: Determination of the active principle and low concentration of preservatives, 33 (2008) 409-414. [25] F.J. Muzzio, C.L. Goodridge, A. Alexander, P. Arratia, H. Yang, O. Sudah, G. Mergen, International Journal of Pharmaceutics, Sampling and characterization of pharmaceutical powders and granular blends, 250 (2003) 51-64. [26] Monograph 2.9.8. Resistance to crushing of tablets, in: Eur.Ph. 8.0., 2014. [27] Monograph 2.9.1. Desintegration of tablets ans capsules, in: Eur.Ph. 8.0., 2014. [28] Monograph 2.9.7. Friability of uncoated tablets, in: Eur.Ph. 8.0., 2014. [29] C. Bodson, W. Dewé, P. Hubert, L. Delattre, Journal of Pharmaceutical and Biomedical Analysis, Comparison of FT-NIR transmission and UV–vis spectrophotometry to follow the mixing kinetics and to assay low-dose tablets containing riboflavin, 41 (2006) 783-790. [30] J. Mantanus, E. Rozet, K. Van Butsele, C. De Bleye, A. Ceccato, B. Evrard, P. Hubert, E. Ziémons, Analytica Chimica Acta, Near infrared and Raman spectroscopy as Process Analytical Technology tools for the manufacturing of silicone-based drug reservoirs, 699 (2011) 96-106. [31] Acetominophen tablets monograph, in: USP 32, 2010.
Ac ce p
435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
18 Page 18 of 23
8°80 % % 200 mg (57%) 120 mg (34%) 26,5 mg (8%) 3.5 mg (1%) 350 mg
Paracetamol + 3% PVP Microcrystalline cellulose Sodium starch glycolate Magnesium Stearate Total
100 %% 250 mg (71%) 79 mg (23%) 17.5 mg (5%) 3.5 mg (1%) 350 mg
Table 1 : Tablets compositions for the 3 API concentrations.
466
80 % 120 %
2.17 2.07 1.92 1.64 2.26
2.14 2.06 1.86 2.01 1.84
M
Disintegration time (s)
Tablet friability KNN
9200-1200
9120-12000
7615-12000 MSC
te
9000-12000
nd
SNV
2 derivative (17pt)
Detrend + Smoothing (15 pt)
3
4
3
0.995 0.415 1.958
0.972 3.87 12.80
0.820 1.67 4.18
Ac ce p
Model Spectral range selected (cm-1) Spectral pretreatment Number of PLS factors R2 val RMSEC RMSEP
Tablet hardness (N) PLS
d
Content uniformity (%)
471
Coefficient of variation (%)
Table 2 : Results of blends content uniformity with the reference method.
468
469 470
Standard deviation (%)
us
A B C
100 %
Mean of sample content uniformity (%) 101.54 100.40 103.32 81.46 123.36
an
Blends
467
120 % % 300 mg (86%) 38.1 mg (11%) 8.4 mg (2%) 3.5 mg (1%) 350 mg
ip t
465
Tables
cr
463 464
N/A
R2 : determination coefficient ; N/A : Not applicable ; all data were mean-centered Table 3 : Conventional criteria of the NIR models.
19 Page 19 of 23
Figure captions
us
cr
ip t
493
Figure 1. Results of real time analysis of the 80% blend. Green points are the reference corresponding to the end of the mixing process when the blend is expected to be homogenous. Blue points are sample spectra of the blend from the beginning to the end of the mixing process.
498 499 500 501
Figure 2. Results of real time analysis of the 100% blend. Green points are the reference corresponding to the end of the mixing process when the blend is expected to be homogenous. Blue points are sample spectra of the blend from the beginning to the end of the mixing process.
Ac ce p
te
d
M
an
494 495 496 497
21 Page 20 of 23
ip t cr
Figure 3. Results of real time analysis of the 120% blend. Green points are the reference corresponding to the end of the mixing process when the blend is expected to be homogenous. Blue points are sample spectra of the blend from the beginning to the end of the mixing process.
us
502 503 504 505
509 510 511 512
Ac ce p
507 508
te
d
M
an
506
Figure 4. Pure API (in green) and tablet (in blue) NIR spectra.
Figure 5. Accuracy profile based on the validation results of the NIR model for the API content determination. The plain line is the relative bias, the dashed lies are the β-expectations tolerance limits (β=95%) and the dotted lines represent the acceptance limits (±10%).
22 Page 21 of 23
ip t cr
Figure 6. Accuracy profile based on the validation results of the NIR model for the tablet hardness determination. The plain line is the relative bias, the dashed lines are the β-expectations tolerance limits (β=95%) and the dotted lines represent the acceptance limits (±25%).
517 518 519
Figure 7. Disintegration time (s) predicted by the NIR model versus the disintegration time measuredby the reference method results for calibration (in red) and validation (in black) sets.
520 521
Figure 8. Results of KNN analysis for tablet friability.
Ac ce p
te
d
M
an
us
513 514 515 516
522 23 Page 22 of 23
Ac
Blenduniformity
ce pt
ed
M an
us
cr
ip t
*Graphical Abstract
Content uniformity
Tablet friability
Disintegration time
Tablet hardness
Real time release approach with:
NIR Spectroscopy
Page 23 of 23