European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx Contents lists available at ScienceDirect European Journal of Pharmaceutics and Biopharmaceutics journal homepage: www.elsevier.com/locate/ejpb Research paper Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models Margot Fonteyne a,⇑, Delphine Gildemyn a, Elisabeth Peeters b, Séverine Thérèse F.C. Mortier c, Jurgen Vercruysse b, Krist V. Gernaey d, Chris Vervaet b, Jean Paul Remon b, Ingmar Nopens c, Thomas De Beer a a Laboratory of Pharmaceutical Process Analytical Technology, Ghent University, Ghent, Belgium Laboratory of Pharmaceutical Technology, Ghent University, Ghent, Belgium c BIOMATH, Department of Mathematical Modelling, Statistics and Bioinformatics, Ghent University, Ghent, Belgium d Department of Chemical and Biochemical Engineering, Technical University of Denmark, Lyngby, Denmark b a r t i c l e i n f o Article history: Received 21 December 2013 Accepted in revised form 27 February 2014 Available online xxxx Keywords: NIR spectroscopy Raman spectroscopy Real-time monitoring End-point detection Granule size fractions Fluid bed drying PAT a b s t r a c t Classically, the end point detection during fluid bed drying has been performed using indirect parameters, such as the product temperature or the humidity of the outlet drying air. This paper aims at comparing those classic methods to both in-line moisture and solid-state determination by means of Process Analytical Technology (PAT) tools (Raman and NIR spectroscopy) and a mass balance approach. The six-segmented fluid bed drying system being part of a fully continuous from-powder-to-tablet production line (ConsiGma™-25) was used for this study. A theophylline:lactose:PVP (30:67.5:2.5) blend was chosen as model formulation. For the development of the NIR-based moisture determination model, 15 calibration experiments in the fluid bed dryer were performed. Six test experiments were conducted afterwards, and the product was monitored in-line with NIR and Raman spectroscopy during drying. The results (drying endpoint and residual moisture) obtained via the NIR-based moisture determination model, the classical approach by means of indirect parameters and the mass balance model were then compared. Our conclusion is that the PAT-based method is most suited for use in a production set-up. Secondly, the different size fractions of the dried granules obtained during different experiments (fines, yield and oversized granules) were compared separately, revealing differences in both solid state of theophylline and moisture content between the different granule size fractions. Ó 2014 Elsevier B.V. All rights reserved. 1. Introduction About 80% of the currently available pharmaceuticals are formulated as solid dosage forms, the majority of them being tablets. In many cases tabletting requires granulation of the starting materials (powders) prior to compaction. As a consequence industrial tabletting is still a multi-step process nowadays. In case of wet Abbreviations: API, Active Pharmaceutical Ingredient; DoE, Design of Experiments; KF, Karl Fischer; NCO, Non-contact optic; NIR, Near Infrared; PAT, Process Analytical Technology; PC, Principal Component; PCA, Principal Component Analysis; PLS, Partial Least Squares; PVP, polyvinylpyrrolidone; RMSEP, Root Mean Square Error of Prediction; rpm, rotations per minute; SNV, Standard Normal Variate. ⇑ Corresponding author. Laboratory of Pharmaceutical Process Analytical Technology, Faculty of Pharmaceutical Sciences, Ghent University, Ghent, Belgium. Tel.: +32 9 264 83 55; fax: +32 9 222 82 36. E-mail addresses: Margot.Fonteyne@Ugent.be (M. Fonteyne), Thomas.DeBeer@ Ugent.be (T. De Beer). granulation, a drying step follows the granulation step. Afterwards the dry granules may be mixed with an external phase (i.e. lubricant, disintegrant). The final blend is then fed to the tabletting machine for compaction, which might be followed by coating, before the tablets are blistered and packed. After each step of this production chain the critical (intermediate) product characteristics of random samples of the batch are generally evaluated by means of offline analyses in analytical laboratories. Batches will either proceed to the next processing step or will be rejected in case of failure of these analysis tests. Hence, traditional batch production is a timeconsuming and expensive production method. Partly due to the increasing competition and decreasing profits in the pharmaceutical industry (i.e. generics, smaller pipelines, expiring patents, etc.), innovative manufacturing models are more and more desired in order to make the production processes faster, cheaper, more efficient and hence more competitive. Therefore, continuous production gains increasing interest in the pharmaceutical industry, also in tabletting applications. Recent manufacturing http://dx.doi.org/10.1016/j.ejpb.2014.02.015 0939-6411/Ó 2014 Elsevier B.V. All rights reserved. Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 2 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx technology advances have already shown the advantages of this approach [1–6]. Recently, full continuous from-powder-to-tablet production lines became commercially available [7–11]. The traditionally applied quality assessment approach based on off-line analysis is not applicable in a continuous manufacturing environment, since the process cannot be stopped and immediate and continuous process and product quality information is required. The need for offline tests would counterbalance the advantages of continuous manufacturing. Therefore, continuous real-time quality control should be ensured by means of in-process analysis methods as it is advised in the Food and Drug Administration’s PAT guideline [12]. Conventionally, the end point of a drying process – usually performed in a fluid bed dryer – after wet granulation is determined by means of indirect parameters. The humidity and the temperature of the outgoing drying air, the product temperature and the pressure difference over the fluidized bed can be evaluated in order to control the fluidized bed drying of wet granules [13,14]. These methods give an idea about the water evaporation progress. The drying process is considered as finished when water evaporation is no longer detected. However, the in-line monitoring of residual moisture content during drying using Near Infrared (NIR) spectroscopy is beneficial for two reasons: (i) monitoring the moisture content allows end point detection of the drying process and makes real-time decision making possible, hence preventing over-drying; and (ii) information regarding possible structural changes of the Active Pharmaceutical Ingredient (API) or/and excipients can be obtained. The utility of NIR spectroscopy for the in-line monitoring and moisture assessment during fluid bed and spray drying processes has been demonstrated extensively in the literature [15–23]. Furthermore, NIR spectroscopy has also recently been used for the monitoring of a continuous granulation and drying process [10]. However, in this study the measurements were performed after the drying unit and not during drying. Besides monitoring of the drying process, NIR spectroscopy can also be used for the continuous evaluation of process induced solid-state transformations of both APIs and excipients. Romer et al. [24] monitored the solidstate conversions of erythromycin dehydrate using an in-line NIR spectrometer in a miniaturized fluid bed dryer. Aaltonen et al. [14] used both NIR and Raman spectroscopy to monitor the solid-state changes of theophylline using the same mini-dryer. They linked the in-line obtained spectra to the traditionally monitored fluid bed parameters such as absolute humidity of the outlet air and pressure difference over the fluidized bed. The solid-state changes were quantified in real-time, which is impossible with the traditional indirect parameters. Furthermore, the same micro scale fluid bed dryer was used by Kogermann et al. [25] to quantify the solid-state changes of piroxicam and carbamazepine in-line using Raman spectroscopy. The presented study aims at evaluating Raman and NIR spectroscopy for the in-line monitoring of the drying process and determination of the end point, the residual moisture content and the product solid state during continuous drying in a six-segmented continuous fluid bed drying unit, which is part of a fully continuous from-powder-to-tablet manufacturing line (ConsiGma™-25, GEA Pharma Systems nv., Collette™, Wommelgem, Belgium). Furthermore, data derived from the in-line acquired spectroscopic data are compared with the conclusions obtained from the conventional indirect approach using the logged univariate parameters such as humidity of the outlet air and product temperature. Additionally, the spectroscopic observations are compared with the residual moisture content conclusions that can be derived from a mass balance model, which was recently developed for the six-segmented continuous fluid bed dryer [26]. This mass balance model is based on the physics governing the continuous drying process, hence forcing fundamental process understanding. Mass balance modeling requires the definition of the composition of the physical inand outgoing gas (moisture content) and the liquid and solid streams in the process. It is examined whether feeding the continuously logged process parameters (e.g., humidity and temperature of inlet and outlet air, product temperature, etc.) into this mass balance model allows visualizing the drying process progress and calculating the end point of drying and the corresponding residual moisture content. By comparing these results to the spectroscopic results, the necessity of using spectroscopic monitoring during drying is evaluated and discussed. 2. Materials and methods 2.1. Materials Anhydrous theophylline (Farma-Quimica Sur S.L., Malaga, Spain) (30%, w/w) was used as a model drug and granulated together with lactose monohydrate 200 M (Caldic Belgium NV, Hemiksem, Belgium) as filler. Polyvinylpyrrolidone (Kollidon 30Ò, BASF, Burgbernheim, Germany) was added as a binder to the dry powder mixture in a concentration of 2.5% (w/w). Distilled water was used as granulation liquid. Sodium lauryl sulfate (Fagron, Waregem, Belgium) was added to the granulation liquid (0.5% w/ v) to improve the wettability of the dry powder mixture. 2.2. Continuous twin-screw granulation and fluid bed drying Continuous granulation and drying was performed using the ConsiGma™-25 unit (GEA Pharma Systems nv., Collette™, Wommelgem, Belgium), which consists of three major units: a continuous twin screw high shear granulator, a six-segmented fluid bed dryer and a discharge system. The system has been extensively described elsewhere [7,11]. After discharging, a lubricant can be added and blended into the dried granules, after which the final blend can be compressed using an in-line tabletting machine. One of the assets that ConsiGma™ offers is the continuous logging and storage of numerous process parameters and outcomes in each unit (i.e., temperature granulator barrel, torque on twin screws, weight powder dosing unit, temperature of product in the dryer, etc.). 2.3. NIR spectroscopy A Fourier-Transform NIR spectrometer (Thermo Fisher Scientific, Zellik, Belgium, Nicolet Antaris II near-IR analyzer) equipped with an InGaAs detector, a quartz halogen lamp and a fiber optic contact probe was used. The probe was inserted in cell 5 of the six-segmented fluid bed dryer by means of an in-house developed accessory (Fig. 1). Each spectrum was collected in the 10,000– 4500 cm 1 spectral region with a resolution of 16 cm 1 and was averaged over 16 scans. Spectra were recorded continuously during drying and a spectrum was collected approximately each 10 s. The same fiber optic contact probe and spectrometer settings were used for off-line measurements. Spectra were mean centered and Standard Normal Variate (SNV)-corrected prior to multivariate data analysis. Data collection and data transfer were done using Thermo Fisher Scientific’s Result Software. 2.4. Raman spectroscopy A RamanRxn1 spectrometer (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, US) equipped with an air-cooled CCD detector (back-illuminated deep depletion design) was used. For the in-line Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 3 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx Fig. 1. NIR fiber optic contact probe mounted in cell 5 of the dryer by means of an in-house made accessory. measurements, an immersion optic (IO), connected to an MR Probe (Kaiser Optical Systems, Ann Arbor, Michigan) was inserted in cell 2 of the six-segmented continuous fluid bed dryer by means of an in-house developed accessory. An IO with a working distance of 3 mm was used (i.e. ‘‘long’’). The off-line spectra were collected with a non-contact optic (NCO), being connected to an MR probe (Kaiser Optical Systems, Inc., Ann Arbor, Michigan, US). All spectra were recorded with a resolution of 4 cm 1 and an exposure time of 10 s, using a laser power of 400 mW. In-line spectra were taken continuously during the 20 min of drying. The system was not paused in between two spectra, resulting in a spectrum each 10 s approximately. Spectra were mean centered and SNV-corrected before data analysis in the spectral region from 200 to 1800 cm 1. Data collection and data transfer were automated using the HoloGRAMS™ data collection software, the HoloREACT™ reaction analysis and profiling software and the Matlab software (version 7.1, The MathWorks Inc., Natick, Massachusetts, US). to dry inside the first cell for a user defined drying time after which they are unloaded and discharged to the discharge module. Every cell repeats this cycle of loading, drying and discharging, hence enabling the continuous drying of the continuously produced wet granules. Two of the dryer cells (cell 2 and cell 5) contain an interfacing opening for PAT tools. In-house designed accessories were used to insert the Raman (cell 2) and NIR (cell 5) spectroscopic probes in these interfacing places (Fig. 1). One of the aims of this study is to evaluate the applicability of in-line NIR spectroscopy for the in-process monitoring of the drying progress and for the granule moisture content prediction at the end of drying. To develop an NIR calibration model for residual moisture content, granules with different residual moisture contents were produced. Therefore, wet granules were manufactured using a powder feed rate of 10 kg/h, a screw speed of 950 rpm, 9.94% (w/w) granulation liquid and a granulator barrel temperature of 25 °C. During the consecutive drying step, the inlet airflow and temperature of the drying air were varied as well as the filling and drying time, hence generating granules with different residual moisture contents (Table 1). Immediately after drying, the granules were collected and twenty NIR spectra were measured off-line. Karl Fischer moisture determination was also performed directly after the NIR measurements of each calibration experiment. These data (i.e., the off-line collected NIR spectra (X) and the corresponding Karl Fischer determined residual moisture contents (Y)) were used for the construction of a Partial Least Squares (PLS) model. Afterwards, six test experiments were performed, where NIR and Raman spectra were collected in-line during drying, for both moisture and drug solid-state monitoring. These test experiments were performed using a powder feed rate of 17.5 kg/h, a liquid rate of 9.94% (w/w), a granulator barrel temperature of 25 °C and a screw speed of 950 rpm. The filling time and the drying time were 5 and 20 min respectively. An airflow of 360 m3/h was blown through the bottom plate of the dryer. The test experiments differed regarding the applied drying air temperature varying from 30 °C to 80 °C in steps of 10 °C (Table 1). No other cells, besides 2 and 5 were filled with granules. Directly after each in-line measurement, the moisture content of the dried granules of cell 2 and cell 5 was measured in triplicate by means of Karl Fisher titration. 2.5. In-process moisture and drug solid-state monitoring This study focuses on the continuous fluid bed dryer of the ConsiGma™25-system, which consists of six parallel cells. Wet granules are loaded into the first cell during a user defined time period, after which wet granules are loaded into the second cell, etc. While the second cell is being filled, wet granules continue 2.6. Moisture and drug solid-state variability between the different granule size fractions In order to obtain granules with diverse characteristics (i.e. granule size distribution, residual moisture content, solid state of Table 1 Experiments for the development of the NIR moisture calibration model and the test experiments. Exp Air flow drying air (m3/h) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Wet granules 360 360 360 360 200 360 360 360 360 360 360 360 360 360 Test experiments TX 360 Temperature drying air (°C) Filling time (s) Drying time (s) Moisture content (%) 45 45 45 45 45 45 45 45 45 45 60 60 60 60 75 45 30 30 75 60 60 60 120 300 150 60 300 300 75 45 30 30 75 90 150 380 600 900 900 900 900 1500 11.54 5.17 6.24 5.34 6.05 7.08 6.16 5.75 5.03 3.95 4.26 3.62 3.78 3.53 3.6 X = 30, 40, 50, 60, 70, 80 300 1200 Varied Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 4 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx and methods and Table 1). Twenty NIR spectra were collected per calibration sample, resulting in 20  15 = 300 spectra. A PLS model was developed by regressing these NIR spectra versus the residual moisture contents of the calibration samples, assessed via Karl Fischer titration. Subsequently, six test experiments were performed hereby using in-line NIR and Raman spectroscopic monitoring in order to assess moisture content during drying (using the developed NIR moisture PLS model) as well as the drug solid state, respectively. The optimal number of PLS components for the NIR moisture model was evaluated according to the cumulative fraction of the variation of the y-variable explained (R2[Y]), the fraction of the total variation that can be predicted estimated by cross-validation (Q2) and the Root Mean Square Error of Prediction (RMSEP) for the predictions in the test set. The RMSEP was determined in two ways: (i) comparing the moisture content predicted from the last five collected in-line NIR spectra of each test experiment. Then the average of the five obtained predictions was compared to the actual moisture content determined via Karl Fischer titration immediately after each experiment; (ii) comparing the moisture content predicted from only the last NIR spectrum of each test experiment to the same KF results. Applying the spectral range of 10,000–4500 cm 1 of the 300 calibration spectra resulted in a three PLS-components model, with an R2[X] of 0.936, an R2[Y] of 0.94 and a Q2 of 0.916. Applying this model for the moisture prediction based on the last 5 in-line collected NIR spectra of the 6 test experiments (6  5) resulted in an RMSEP of 0.5659%. When only the last spectrum (6  1) of each test experiment was used, a similar RMSEP of 0.5676% was obtained. The model was optimized by removing calibration runs 0, 7 and 8 (these runs had a high standard deviation on the Karl Fischer measurements, hence creating doubt on the correctness of the reference measurements for these experiments). The R2[X] was 0.994 and both the R2[Y] and the Q2 of this model were 0.962. The final model consisted out of four PLScomponents and an RMSEP (6  5) of 0.5608% and (6  1) 0.5649% was obtained. This model, including the spectral range from 10,000 to 4500 cm 1, but excluding calibration experiment 0, 7 and 8 was used for further calculations (Fig. 2). The six in-line test experiments were performed at six different drying temperatures varying from 30 °C to 80 °C in steps of 10 °C per experiment (Table 1). The end point of the drying cycle could be determined by means of real-time NIR monitoring of the granule’s moisture content. The actual moisture content values, determined with Karl Fischer at the end of the drying cycle of each test experiment are shown in Fig. 2 (red dashed line). The granules are considered as dry when a moisture content of 3.6% is reached, since this is the moisture content of the premix used to produce the granules. The in-line NIR predicted moisture contents during the entire drying process of each test experiment are also plotted in Fig. 2 (blue full lines). During the first five drying minutes of each test experiment, the cell was loaded with granules and the NIR probe was not embedded in granules. The probe was measuring the API), an experimental design (DoE) was performed on the ConsiGma™-25 unit. A 4-factor full factorial design with high and low levels and three centerpoints, resulting in 19 experiments, was performed (Table 2). The temperature of the granulator barrel was varied between 25 °C and 40 °C and powder feed rates of 10 and 25 kg/h were used. The screw speed was kept constant at 950 rpm and the granulation liquid was added at 9.16% (w/w). For the drying unit the drying air temperature was varied between 35 °C and 75 °C and an air flow between 350 and 450 m3/h was applied. Two of the 19 experiments, namely experiment 8 and 12, could practically not be performed, since the produced granules consisted mainly of fines and the high air flow (450 m3/h) blew the fines in the filters, leading to blockage of the filters. The remaining 17 experiments were performed in randomized order. An amount of 1.25 kg was dried in each cell during 10 min and each DoE experiment was done in duplicate. After each DoE experiment, the moisture content and the solidstate of the API were determined for separate sieve fractions: fines (<150 lm), yield (151 lm–1400 lm) and oversized granules (>1400 lm). The granules were separated in these three fractions by means of a sieve tower. For each experiment and its duplicate, 5 Raman spectra were taken from the total granule load and from the three sieve fractions, resulting in 40 spectra per DoE experiment. Furthermore, the residual moisture content was determined for both the total fractions of granules and the three sieve fractions individually by means of Karl Fischer titration (n = 2). 2.7. Karl Fischer moisture determination The residual moisture of samples was determined by volumetric Karl Fischer titration (KF) using a V30 volumetric Karl Fischer titrator (Mettler Toledo, Zaventem, Belgium). Methanol (Hydranal, Sigma Aldrich, Germany) was used as a solvent. Before titration, granules were stirred and dissolved during 5 min. 2.8. Data analysis The analysis of the spectra and the development of the Principal Component Analysis (PCA) and Partial Least Squares (PLS) models was done using the Simca P+ 12.0.1 software (Umetrics AB, Umeå, Sweden). 3. Results and discussion 3.1. In-line moisture calibration model and mass balance approach Fifteen different loads of granules (i.e., calibration set) were produced for the development of the moisture calibration model. The process parameters of these 15 runs were varied to obtain granules with different residual moisture contents (see materials Table 2 Overview of the DoE parameters. Average moisture content of the total granule load and its granule size fractions. Run 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Temperature granulator barrel (°C) Powder feed rate (kg/h) Temperature drying air (°C) Air flow drying air (m3/h) 40 40 40 40 32.5 25 25 25 32.5 32.5 25 25 25 25 40 40 40 40 25 10 75 350 25 75 450 10 35 450 25 35 350 17.5 55 400 25 75 350 10 35 350 10 75 450 17.5 55 400 17.5 55 400 25 35 450 10 35 450 10 75 350 25 35 350 25 35 450 10 75 450 10 35 350 25 75 350 25 75 450 4.00 4.37 4.17 4.24 3.68 4.07 4.06 4.21 6.51 6.74 6.75 5.81 7.34 7.64 7.10 6.07 4.96 5.79 4.85 4.60 3.88 4.05 3.85 3.84 6.39 6.71 6.80 5.99 5.40 5.87 5.40 4.72 4.26 5.25 4.57 4.05 5.76 6.68 6.32 4.76 3.79 3.97 3.95 3.76 5.48 6.39 5.99 4.95 6.46 6.81 6.39 5.52 3.87 4.05 4.03 4.03 5.13 6.21 5.70 4.62 3.85 3.99 4.10 3.99 3.73 4.06 3.85 3.85 Moisture content (w/w%) of Total granule load Oversized granules Yield Fines Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx 5 Fig. 2. Moisture content (%) in function of time (min). Prediction of moisture content according to in-line NIR measurements (full blue line) and residual moisture content as obtained by means of Karl Fischer after each test experiment (red dashed line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) air and the resulting noisy spectra were therefore excluded. As drying proceeds, the moisture content, measured by means of NIR, reaches a steady state for all six test experiments. One can notice certain ‘‘plateaus’’ of the blue lines in the graphs (during the first part of drying), where for different time points exactly the same predicted moisture contents were obtained. The plateaus were caused by fouling of the NIR probe. When wet particles were stuck on the probe window, similar spectra were captured for several time points. This phenomenon can clearly be seen at the start of the cycles with low drying temperatures (T 30 and T 40). For the granule loads that were dry after the drying cycle (T 60, T 70 and T 80), the drying process end points can be clearly detected and are indicated in Fig. 2. Conventionally, the end point of drying is estimated on the basis of a measurement of the humidity of the outlet air or by evaluating the temperature of the drying product. These data are plotted Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 6 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx 10 75 9 70 11:43 65 8 7 60 55 13:38 6 50 5 45 Humidity Drying air oultet (%RH) Product Tempearture (°C) 80 4 40 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Time (min) Fig. 3. Product temperature (yellow line) and humidity of the outlet air (black line) during the test experiments dried at a. 60 °C, b. 70 °C and c. 80 °C. The vertical blue line indicates the end point as determined by the NIR-based moisture determination model. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) in Fig. 3 for the test experiments, which resulted in a dry product (i.e., test experiments which were dried at 60, 70 and 80 °C). The time points when the humidity of the outlet air and the product temperature reach steady state are indicated as well as the end point predicted by the in-line NIR measurements (vertical blue line). Regarding the humidity of the outlet air, no steady state can be observed for the experiment conducted at 60 °C. The humidity of the outlet air reaches a steady state for the experi- ments run at 70 °C and 80 °C, but this is at a later time instant compared to the drying process end point as indicated by the NIR moisture determination model. The product temperature seems to be a more reliable parameter for the drying endpoint determination of the six-segmented fluid bed dryer since each cell has its own temperature sensor installed, and the sensor is really embedded in the product. The determined product temperature steady states are indicated as well in Fig. 3. Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx The evolution of the granules’ residual moisture content during drying can also be calculated using a mass balance, which was recently developed by our group for the six-segmented fluid bed drying system of the ConsiGma™-25 system [26]. Using this mass balance, univariate data (e.g., temperature of inlet and outlet air, product temperature during drying, humidity of inlet and outlet air), which are automatically and continuously logged by the ConsiGma™25-system are processed to calculate the amount of water, which evaporates from the granules during drying. Since Karl Fischer titration measures both free and bound water and since the mass balance model only determines the free water, the moisture content of the granules in the dryer determined via the mass balance needs to be corrected for the bound water fraction. Therefore, the water content of the dry premix – being 3.6% – has to be added to the moisture content values predicted via the mass balance model. The drying curves obtained through the mass balance model are plotted in Fig. 4. At the end of the drying cycle, the granules dried at 60, 70 and 80 °C are considered to be dry according to the model (they reach 0%). This is confirmed by both the off-line Karl Fischer measurements and the NIR-based drying curves (Fig. 2). The mass balance model predicted moisture content values at the end of the drying cycle when drying at 30, 40 and 50 °C are comparable to Karl Fischer findings (6.69%, 5.93% and 4.3% respectively). The typical shape of the drying curves in Fig. 4 can unfortunately not be compared with the curve obtained via NIR predictions (Fig. 2), since it was only possible to collect good spectra after five drying minutes in the used experimental set-up (cfr. supra). When comparing the NIR model and mass balance model based obtained moisture content values after 5 min of drying, mass balance model predictions are approximately 2% (absolute value) higher than the values obtained by means of NIR prediction, when granules are dried at 30, 40 and 50 °C. When granules are dried at 60, 70 an 80 °C the predicted moisture value by the NIR method is comparable to the moisture value calculated via the mass balance model. As an example, in Fig. 5, the moisture contents as predicted by the NIR calibration model and the mass balance model for the experiment performed at 70 °C are presented in an overlay plot. A value of 3.6% (i.e. moisture content of the dry premix) is added to the mass balance model predictions for better comparison. 7 Fig. 5. Experiment at 70 °C: Overlay of the predicted moisture content via the mass balance model (MBM, dotted black line) and the NIR calibration model (NIR, full blue line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Even though the granules are not dry when a temperature of 30 °C or 40 °C is applied a steady state in the drying curves can be observed in Fig. 4. The same steady state can be seen in the NIR prediction curves, which suggests that at lower temperatures after a given time no extra water evaporates from the granules, even though they are not dry yet and still fluidizing. The experiment ran at 50 °C does not show this steady state, but a slow decrease in water moisture content during the last 5 min in both the moisture curves obtained by the NIR model as the mass balance model can be detected. This means the results of the mass balance model and the NIR-based model are in line. 3.2. In-line solid-state monitoring 3.2.1. In-line Raman spectroscopy The theophylline solid-state changes were monitored using inline Raman spectroscopy during the six test experiments (Fig. 6a– f). Some spectral differences can be noted according to the applied temperature of the drying air. To evaluate the solid state of theophylline, one should focus on the spectral region from 1650 to 1750 cm 1, since theophylline monohydrate shows a band at Fig. 4. Drying curves of the six different test experiments as calculated with the mass balance model. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 8 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx (d) 0,5 0,5 0,0 0,0 Intensity Intensity (a) -0,5 -0,5 -1,0 -1,0 -1,5 Raman Shift (1/cm) 1718 1706 1712 1700 1694 1688 1682 1676 Raman Shift (1/cm) (e) (b) Intensity 0,5 Intensity 1670 1664 1658 1652 1640 1646 1718 1712 1706 1700 1694 1688 1682 1676 1670 1664 1658 1652 1646 1640 -1,5 0,0 -0,5 0,5 0,0 -0,5 -1,0 -1,0 -1,5 (f) 0,5 1718 1712 1706 1700 1694 1688 1682 0,5 0,0 0,0 Intensity -0,5 -0,5 -1,0 -1,0 1717 1711 1705 1699 1693 1687 1681 1675 1669 1663 1657 1651 1717 1711 1705 1699 1693 1687 1681 1675 1669 1663 1657 1651 1645 1640 Raman Shift (1/cm) 1645 -1,5 -1,5 1640 Intensity 1676 Raman Shift (1/cm) Raman Shift (1/cm) (c) 1670 1664 1658 1652 1646 1640 1718 1712 1700 1706 1694 1688 1682 1676 1670 1664 1658 1652 1646 1640 -1,5 Raman Shift (1/cm) Fig. 6. In-line Raman spectra collected during the six test experiments. The spectra are colored from light blue to black according to drying time. The first spectrum is light blue while the last collected spectrum is black. Test experiments dried at: a. 30 °C, b. 40 °C, c. 50 °C, d. 60 °C, e. 70 °C, f. 80 °C. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 1687.5 cm 1, whereas a spectrum of anhydrous theophylline shows two peaks, one at 1665 cm 1 and one at 1707 cm 1 [7]. Furthermore, a metastable form of theophylline exists, which shows a typical Raman band at 1692 cm 1 [27]. When granules were dried at 30 °C, no spectral changes (i.e. changes in solid state) can be noticed (Fig. 6a). Only a band at 1687.5 cm 1 can be seen indicating that theophylline remains in the monohydrate form from the start till the end of the drying cycle at 30 °C. In the spectra, collected during drying at 50 °C (Fig. 6c), two extra peaks can be noticed at 1665 cm 1 and 1707 cm 1, indicating that an amount of theophylline is dehydrated to the anhydrous form. This can also be noted, when granules are dried at 40 °C albeit to a much lesser extent (Fig. 6b). The test experiments with drying temperatures of 60 °C and up (Fig. 6d, e and f) show a shift of the Raman band at 1687.5 cm 1–1692 cm 1. This shift denotes a polymorphic change from theophylline monohydrate to metastable theophylline. When comparing Fig. 6d to Fig. 6e and f, the color code of the spectra shows that the higher the temperature of the incoming air is, the faster this conversion takes place. Furthermore, additional peaks at 1665 cm 1 and 1707 cm 1 appear during these test experi- ments, indicating the formation of anhydrous theophylline. This means that the dried granules of these test experiments contain both metastable and anhydrous theophylline. The findings regarding the solid state of theophylline correspond with the detected steady states in the drying curves (Figs. 2 and 4) of the test experiments conducted at 30 and 40 °C. From a certain time point, no more water is evaporating during these experiments, furthermore theophylline stays in the monohydrate form. These findings suggest that, when drying at 30 or 40 °C the theophylline hydrate water will stay bound to the theophylline molecule and hence no extra water will evaporate. For the test experiment conducted at 50 °C a slow conversion from theophylline monohydrate to anhydrous theophylline can be noted. This can be correlated with the slightly decreasing slope in the NIR predicted drying curve (minute 13 till end) and the slope in the mass balance calculated curve (minute 10 till end). The spectra collected during the drying test experiment at 60 °C were analyzed using Principal Component Analysis (PCA). The spectral region from 200 cm 1 to 1800 cm 1 was centered and SNV-corrected. The resulting principal component 1 versus Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 9 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx Fig. 7. PC1 versus PC2 scores plot of the Raman spectra of the test experiment at 60 °C. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) principal component 2 scatter scores plot (Fig. 7) shows the transition from theophylline monohydrate towards a mixture of metastable and anhydrous theophylline. The first principal component represented 84.65% of the spectral variance, the second 8.25%. The spectra taken in the first 12 min only show a band at 1687.5 cm 1 indicating that theophylline remained in the monohydrate state. After 15 min the band has clearly shifted to 1692 cm 1, and bands at 1665 cm 1 and 1707 cm 1 appear as well. From minute 12 to minute 15, one can observe a transition period in the plotted data and the Raman spectra. When drying at 70 °C it took 11m30s to convert theophylline monohydrate to a mixture of metastable and anhydrous theophylline and the transition period started after 8 min. A mixture of metastable theophylline and anhydrous theophylline could be detected after 11 min of drying at 80 °C. A transition period from the eighth till the eleventh minute could be observed. These findings are different compared to the results obtained by others, since some found the metastable theophylline at lower drying temperatures (40 °C) [28] when heating theophylline monohydrate powder in a variable temperature X-ray powder diffractometer. Morris et al. [29] stated that the metastable form of theophylline most likely occurs at low temperatures (< 60 °C) when fluid bed drying is applied after wet granulation. These findings were confirmed when theophylline was wetmassed, followed by drying at 60 °C during 50 min [30]. Airaksinen et al. [31] already found metastable theophylline when drying at a temperature as low as 30 °C and using dry inlet air. When using ambient air for drying the highest relative amount of metastable theophylline was found when drying at 40 °C. The differences in observations compared to this study, where metastable theophylline could only be found at 60 °C, might be due to the fact that all other authors that were mentioned used pure theophylline or pure theophylline granules and did not use any excipients. 3.2.2. In-line NIR spectroscopy When evaluating the in-line recorded NIR spectra, not only the disappearing water bands attract attention. As for Raman 2,0 1,5 Intensity 1,0 0,5 0,0 -0,5 -1,0 -1,5 5901 5908 5916 5924 5931 5939 5947 5955 5962 5970 5978 5985 5993 6001 6009 6016 6024 6032 6039 6047 6055 6063 6070 6078 6086 6093 6101 -2,0 Wavenumber (1/cm) Fig. 8. In-line collected NIR spectra, SNV-corrected in the spectral range from 6100 to 5900 cm 1. The earliest spectrum is light blue changing to black for the last spectrum. Test experiment at 60 °C. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 10 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx spectroscopy, the changes in solid state of the API can also be identified in the NIR spectra. Two extra bands appear during dehydration, which can be attributed to appearance of anhydrous theophylline: one at 5962 cm 1 and one, less obvious at 6009 cm 1 [7]. Near infrared spectroscopy is incapable of differentiating between stable and metastable anhydrous theophylline [14]. In Fig. 8, the spectral range from 5900 to 6100 cm 1 of the in-line collected spectra (T = 60 °C) was selected and was corrected using SNV. PCA was applied on these corrected spectra. Fig. 9 shows the resulting PC 1 versus PC 2 scores plot for this test experiment. The PC 1 represented 95.1% of the variance and PC 2 represented 3.91%. Three clusters can be distinguished in the scores plot: (i) the spectra of the first 12m30s, (ii) the spectra collected between 12m30s and 15m45s (iii) the spectra collected after 15m45s minutes of drying. During the first 12m30s no bands can be seen at 5962 cm 1 and 6009 cm 1, indicating that all theophylline remained in the monohydrate form. After 15m45s, the two extra bands appear clearly, indicating the transformation to anhydrous theophylline. Between 12m30s and 15m45s a mixture of theophylline anhydrate and monohydrate was found. For the test experiment conducted at 70 °C, anhydrous theophylline could be found after 11m15s. When granules were dried at 80 °C, the anhydrous theophylline bands appeared after 9 min. 3.2.3. End-point detection during drying and solid-state monitoring: evaluation of different techniques Four different methods to detect the end point of a drying cycle have been presented: (i) by means of an NIR-based PLS model, (ii) by logging of the humidity of the outlet air, (iii) by logging of the product temperature and (iv) by means of a mass balance model. Secondly, NIR and Raman proved to be able to monitor the Fig. 10. Residual moisture content of the different granule size fractions as determined by means of Karl Fischer titration. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) solid-state changes of the API in real-time during drying. Table 3 shows an overview of the time points (end of drying or solid-state change) as calculated using the different methods. For the indirect methods (ii) and (iii), a clear steady state can be found for the product temperature, but not for the humidity of the outlet drying air. Both the NIR prediction model and the mass balance model indicate the drying process end point sooner than the two indirect methods. The dehydration of theophylline monohydrate was monitored with both NIR and Raman spectroscopy. The time points of 1,5 1,0 0,5 PC 2 145 146 143 144 142 141 140 139 138 137 47 48 56 49 50 61 60 41 42 43 44 45 46 57 58 59 66 52 53 54 55 68 69 67 51 62 63 64 65 77 135 134 136 133 132 131 78 81 80 79 76 71 84 82 85 83 72 70 87 86 75 73 74 88 89 0,0 130 129 127 128 126 124 125 122 123 121 90 120 119 118 117 -0,5 9192 94 93 97 96 95 98 99 100 101 102 103 104 106 105 107 108110 109 -1,0 116 115 114 111112 113 -1,5 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 PC 1 Fig. 9. PCA of In-line collected NIR spectra, SNV-corrected in the spectral range from 6100 to 5900 cm 1. PC1 versus PC 2 scores plot, test experiment at 60 °C. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Table 3 Summary and comparison between the different detected drying process end points and solid-state changes from theophylline monohydrate towards metastable and anhydrous theophylline. Applied Temperature (°C) 60 70 80 Drying endpoint according to Time point of solid-state shift according to NIR PLS model Mass balance model Humidity of the outlet air Product temperature Raman NIR 16:08 11:37 10:00 13:30 11:15 9:00 / 15:46 13:38 16:27 14:49 11:43 15:00 11:30 11:00 15:45 11:15 9:00 Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 11 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx conversion from theophylline monohydrate to anhydrous theophylline, derived from the Raman and NIR spectra respectively, are comparable. It is important to emphasize that both measurements were conducted in different cells of the dryer, and thus a small deviation between the results obtained from both spectroscopic methods is understandable. These time points are also comparable with the drying process end points calculated using the NIR-based PLS model. It can be concluded that the in-line monitoring of moisture content by means of PAT tools and mass balance models is beneficial. The indirect methods seem to overestimate the needed drying time. It should be remarked that, unlike a conventional fluid bed dryer where only one load is dried at a time, the Consigma™25 consists of six parallel drying cells, and that only one humidity sensor is placed on top of the dryer. For this reason, the NIR method is to be preferred over the mass balance approach in a production set-up, since the measured humidity of the outlet air will correspond to the average value of the six cells when all drying cells are filled with product. Secondly, NIR spectroscopy does not only provide information regarding residual moisture, but also regarding the solid state of the API (and excipients). Raman spec- troscopy will give information regarding the solid state, and is able to differentiate between anhydrous theophylline and metastable theophylline. Challenges for a successful integration of in-line NIR and Raman spectroscopy are (i) the avoidance of fouling of the probe; and, (ii) an adequate method to mount the probe in the dryer should be found as well, so that spectra can be collected from the very start of the drying cycle. Furthermore software packages with high performances for the fast collection and processing of in-line data would be beneficial. 3.3. Particle size and polymorphism of the API Differences in polymorphism of excipients depending on the particle size of spray-dried particles have been reported previously [32]. The in-line Raman and NIR spectroscopic measurements give information about the total granule load, but do not provide granule information related to size fractions within that total granule load. This section of the paper focuses on the differences in moisture content and solid state that can be observed between (a) 8 6 4 PC 2 2 0 -2 -4 -6 -8 -10 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 2 4 6 8 10 12 14 16 18 20 22 24 PC 1 (b) 15 10 PC 2 5 0 -5 -10 -15 -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 12 14 16 18 20 22 24 PC 1 Fig. 11. PCA of the Raman spectra of the different granule size fractions: PC 1 versus PC 2 scores plot. Spectra of fines (blue triangles), yield (red dots), oversized granules (turquoise blue squares) and total granule load (black diamonds) a. Experiment 4 and b. Experiment 19. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 12 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx fines (<150 lm), yield (151–1400 lm) and oversized granules (>1400 lm). To obtain granules with diverse characteristics, a 4-factor full factorial design (19 experiments) was performed (see materials and methods, Table 2). Based on the residual moisture content of the granules of the 19 experiments, they can be classified in three groups (Table 2): (i) granules, which are ‘dry’ after drying, with a moisture content of 4% or lower (experiments 1, 2, 6, 13, 16, 18 and 19); (ii) granules with a residual moisture content higher than 5% (experiments 3, 4, 7, 9, 11, 14, 15 and 17); and (iii), granules having a residual moisture content that is in between 4% and 5% (experiments 5 and 10). These three groups, except experiment 9, correspond to the applied drying air temperature. The ones with residual moisture contents higher than 5% were dried at 35 °C, while the granules having residual moisture contents below 4% were dried at 75 °C. The latter group does not show significant differences in residual moisture content between the oversized fraction, the yield and the fines (Fig. 10). The residual moisture content of the granules dried at 55 °C and 35 °C increases with increasing granule size (Fig. 10). The fines fraction has a lower residual moisture content compared to the yield and oversized fraction. This is because fine particles can dry much faster due to a higher surface to volume ratio. Furthermore, water in the core of the granule will reach the surface of the granule faster in small granules compared to larger granules. Since the moisture content of each size fraction is different, there might be a difference in solid state of the API as well. Experiment 4 shows the largest differences in residual moisture content according to granule size. The Raman spectral region from 200 cm 1 to 1800 cm 1 was selected to perform PCA on the 40 collected and SNV preprocessed Raman spectra resulting from experiment 4. PC 1 explained 79.65% of the variance, whereas PC 2 represented 12.01%. Fig. 11a shows the resulting PC 1 versus PC 2 scores plot. The Raman spectra corresponding to the fine fraction are clearly isolated along PC 1 from the spectra corresponding to the other size fractions. The information regarding the solid state of theophylline is reflected in the Raman spectral region from 520 cm 1 to 600 cm 1 [33] and from 1650 cm 1 to 1750 cm 1 [27,34]. The loadings plot of the first principal component (Fig. 12) in these regions show three maxima: at 554 cm 1, 1685 cm 1 and 1707 cm 1, which indicates that the fraction of fines contains anhydrous theophylline. The minima at 573 cm 1 and 1687 cm 1 on the other hand, imply that the other size fractions mainly contain theophylline monohydrate. Similar solidstate conclusions were obtained for experiment 3, 5, 7, 9, 10, 11, 14, 15 and 17 (fine fraction = anhydrate; other fractions = mainly monohydrate). Although the ‘dry’ runs (experiment 1, 2, 6, 13, 16, 18 and 19) do not show differences in residual moisture content between the different granule size fractions, similar solid-state observations were made (Fig. 11b). The PCA of the spectra obtained after experiment 19, resulted in PC 1 representing 61.28% and PC 2 representing 23.84% of the spectral variance, respectively. Again, fines were mostly clustered in the positive part along the PC 1-axis, whereas oversized granules are located in the negative part. The spectra of the yield fraction can be found centrally. The loadings plot of the first principal component of experiment 19 (Fig. 12) shows again three maxima, attributable to theophylline anhydrate, and indicating that the fines consist of anhydrous theophylline. The minimum in the PC 1 loadings plot on the other hand has shifted from 1687 cm 1 to 1692 cm 1, indicating that the oversized granules from the runs dried at higher temperatures contain metastable theophylline. These experiments showed that the overall determined moisture content and the solid-state information derived from the total granule load differ from moisture content and solid state of the different granule size fractions. Fines tend to dry faster and will contain theophylline anhydrate, whereas the oversized granules will consist of theophylline monohydrate or metastable theophylline. 4. Conclusion Raman and NIR spectroscopy were used successfully for the real-time monitoring of a fluid bed drying process. The drying process end point could be predicted by means of an NIR-based PLS model. NIR spectroscopy proved to be superior to conventional end-point determination by means of indirect process parameters or mass balance modeling in a production set-up. NIR and Raman spectroscopy proved to be capable of in-line solid-state monitoring as well. The time point of conversion to the dehydrated state of the API was comparable to the drying process end point. Furthermore care should be taken when evaluating the average moisture content and solid-state values obtained on the total granule load, since significant differences between the different sieve fractions of the granules were observed. Acknowledgements The authors express thanks to Lien Saerens and Anneleen Burggraeve (Ghent University at that time) for their technical assistance. Fund for Scientific Research Flanders (FWO Vlaanderen – aspirantmandaat Séverine Thérèse F.C. Mortier) is gratefully acknowledged. References Fig. 12. Loadings plot of the first PC of the PCA of the Raman spectra of the different granule size fractions for Experiment 4 and 19. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) [1] C. Vervaet, J.P. Remon, Continuous granulation in the pharmaceutical industry, Chem. Eng. Sci. 60 (2005) 3949–3957. [2] G. Betz, P. Junker-Burgin, H. Leuenberger, Batch and continuous processing in the production of pharmaceutical granules, Pharm. Dev. Technol. 8 (2003) 289–297. [3] H. Leuenberger, New trends in the production of pharmaceutical granules: batch versus continuous processing, Eur. J. Pharm. Biopharm. 52 (2001) 289– 296. [4] H. Leuenberger, Scale-up in the 4th dimension in the field of granulation and drying or how to avoid classical scale-up, Powder Technol. 130 (2003) 225– 230. [5] H. Leuenberger, New trends in the production of pharmaceutical granules: the classical batch concept and the problem of scale-up, Eur. J. Pharm. Biopharm. 52 (2001) 279–288. [6] J. Werani, M. Grunberg, C. Ober, H. Leuenberger, Semicontinuous granulation – the process of choice for the production of pharmaceutical granules?, Powder Technol 140 (2004) 163–168. Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015 M. Fonteyne et al. / European Journal of Pharmaceutics and Biopharmaceutics xxx (2014) xxx–xxx [7] 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. [8] M. Fonteyne, S. Soares, J. Vercruysse, E. Peeters, A. Burggraeve, C. Vervaet, J.P. Remon, N. Sandler, T. De Beer, Prediction of quality attributes of continuously produced granules using complementary pat tools, Eur. J. Pharm. Biopharm. 82 (2012) 429–436. [9] 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. [10] L. Chablani, T.M. K, M. Amit, R. Patrick, S.W. C, Inline real-time near-infrared granule moisture measurements of a continuous granulation-drying-milling process, AAPS PharmsciTech 12 (2011) 1050–1055. [11] 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. [12] FDA-Administration, Guidance for Industry – PAT – A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance, in, 2004. [13] K. Wöstheinrich, P.C. Schmidt, Evaluation and validation of a fully instrumented Hüttlin HKC 05-TJ laboratory-scale fluidized bed granulator, Drug Dev. Ind. Pharm. 26 (2000) 621–633. [14] J. Aaltonen, K. Kogermann, C.J. Strachan, J. Rantanen, In-line monitoring of solid-state transitions during fluidisation, Chem. Eng. Sci. 62 (2007) 408–415. [15] P. Frake, D. Greenhalgh, S.M. Grierson, J.M. Hempenstall, D.R. Rudd, Process control and end-point determination of a fluid bed granulation by application of near infra-red spectroscopy, Int. J. Pharm. 151 (1997) 75–80. [16] R.L. Green, G. Thurau, N.C. Pixley, A. Mateos, R.A. Reed, J.P. Higgins, In-line monitoring of moisture content in fluid bed dryers using near-IR spectroscopy with consideration of sampling effects on method accuracy, Anal. Chem. 77 (2005) 4515–4522. [17] A. Hartung, M. Knoell, U. Schmidt, P. Langguth, Role of continuous moisture profile monitoring by inline NIR spectroscopy during fluid bed granulation of an Enalapril formulation, Drug Dev. Ind. Pharm. 37 (2011) 274–280. [18] F.J.S. Nieuwmeyer, M. Damen, A. Gerich, F. Rusmini, K. van der Voort, H. Vromans, Granule characterization during fluid bed drying by development of a near infrared method to determine water content and median granule size, Pharm. Res. 24 (2007) 1854–1861. [19] A. Peinado, J. Hammond, A. Scott, Development, validation and transfer of a Near Infrared method to determine in-line the end point of a fluidised drying process for commercial production batches of an approved oral solid dose pharmaceutical product, J. Pharm. Biomed. Anal. 54 (2011) 13–20. [20] J. Rantanen, S. Lehtola, P. Ramet, J.P. Mannermaa, J. Yliruusi, On-line monitoring of moisture content in an instrumented fluidized bed granulator [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] 13 with a multi-channel NIR moisture sensor, Powder Technol. 99 (1998) 163– 170. J. Rantanen, O. Antikainen, J.P. Mannermaa, J. Yliruusi, Use of the near-infrared reflectance method for measurement of moisture content during granulation, Pharm. Dev. Technol. 5 (2000) 209–217. J. Rantanen, A. Jorgensen, E. Rasanen, P. Luukkonen, S. Airaksinen, J. Raiman, K. Hänninen, O. Antikainen, J. Yliruusi, Process analysis of fluidized bed granulation, AAPS PharmsciTech 2 (2001). article 21. E. Rasanen, J. Rantanen, J.P. Mannermaa, J. Yliruusi, H. Vuorela, Dehydration studies using a novel multichamber microscale fluid bed dryer with in-line near-infrared measurement, J. Pharm. Sci. 92 (2003) 2074–2081. M. Romer, J. Heinamaki, I. Miroshnyk, N. Kivikero, N. Sandler, J. Rantanen, J. Yliruusi, Phase transformation of erythromycin a dihydrate during fluid bed drying, J. Pharm. Sci. 97 (2008) 4020–4029. K. Kogermann, J.A. Zeitler, J. Rantanen, T. Rades, P.F. Taday, M. Pepper, J. Heinamaki, C.J. Strachan, Investigating dehydration from compacts using terahertz pulsed, Raman, and near-infrared spectroscopy, Appl. Spectrosc. 61 (2007) 1265–1274. S.T.F.C. Mortier, K.V. Gernaey, T. De Beer, I. Nopens, Analysing drying unit performance in a continuous pharmaceutical manufacturing line by means of mass – energy balances, Eur. J. Pharm. Biopharm. (2013), http://dx.doi.org/ 10.1016/j.ejpb.2013.12.014 (Epub ahead of print). A. Dukic-Ott, T. De Beer, J.P. Remon, W. Baeyens, P. Foreman, C. Vervaet, Invitro and in-vivo evaluation of enteric-coated starch-based pellets prepared via extrusion/spheronisation, Eur. J. Pharm. Biopharm. 70 (2008) 302– 312. N.V. Phadnis, R. Suryanarayanan, Polymorphism in anhydrous theophylline – implications on the dissolution rate of theophylline tablets, J. Pharm. Sci. 86 (1997) 1256–1263. K.R. Morris, U.J. Griesser, C.J. Eckhardt, J.G. Stowell, Theoretical approaches to physical transformations of active pharmaceutical ingredients during manufacturing processes, Adv. Drug Deliv. Rev. 48 (2001) 91–114. M. Alleso, F. Tian, C. Cornett, J. Rantanen, Towards effective solid form screening, J. Pharm. Sci. 99 (2010) 3711–3718. S. Airaksinen, M. Karjalainen, E. Rasanen, J. Rantanen, J. Yliruusi, Comparison of the effects of two drying methods on polymorphism of theophylline, Int. J. Pharm. 276 (2004) 129–141. Y.Y. Lee, J.X. Wu, M.S. Yang, P.M. Young, F. van den Berg, J. Rantanen, Particle size dependence of polymorphism in spray-dried mannitol, Eur. J. Pharm. Sci. 44 (2011) 41–48. A. Jorgensen, J. Rantanen, M. Karjalainen, L. Khriachtchev, E. Rasanen, J. Yliruusi, Hydrate formation during wet granulation studied by spectroscopic methods and multivariate analysis, Pharm. Res. 19 (2002) 1285–1291. H. Wikstrom, P.J. Marsac, L.S. Taylor, In-line monitoring of hydrate formation during wet granulation using Raman spectroscopy, J. Pharm. Sci. 94 (2005) 209–219. Please cite this article in press as: M. Fonteyne et al., Moisture and drug solid-state monitoring during a continuous drying process using empirical and mass balance models, Eur. J. Pharm. Biopharm. (2014), http://dx.doi.org/10.1016/j.ejpb.2014.02.015