untitled ISSN 2 Screenin determin Ahmed Sha Ahmed Em 1 Analytical Chem 2 School of Pharm 3 Analytical Chem * Corresponding Tel.: +2.02.23760 ARTICLE INF DOI: 10.5155/eu Received: 03 July Received in revis Accepted: 05 Au Published online Printed: 30 Sept KEYWORDS Moisture conten NIR spectroscop Fluid bed granul PLS regression m Plackett‐Burman Granulation pro 1. Introducti Fluid bed producing gr to a fluidized wetted and c nules. Similar objectives of compression minimize seg applied wet industry, exh the multistag and drying a simplifies the transfer losse ced by FBG w FBG is a com variables that an understan variables is Plackett‐Burm 2153‐2249 (Prin ng the flui nation of awky Abouz mad El Gindy mistry Department, macy, University of mistry Department, author at: Analytic 023. Fax: +2.02.247 FORMATION urjchem.8.3.265-2 y 2017 sed form: 31 July 2 gust 2017 e: 30 September 20 tember 2017   nt py lation model n design cess variables ion d granulation is anules by spra powder. The p collide with eac r to all other we the FBG are to properties, inc gregation, and r granulation t hibiting some si ge wet granulat are all carried e process [2]. es, and time [3] were finer, mor mplex process t can influence nding of the inf necessary for man (PB) design Eu t) / ISSN 2153‐2 ht Europ id bed gra pharmac zaid 1,2,*, Ma y 1, Stephen Faculty of Pharma Maryland, Baltimo Faculty of Pharma cal Chemistry Depa 772038. E‐mail add 272.1608 2017 017 s a wet granul aying solution ( articles in the p ch other to ad et granulation t o improve the f crease the dens reduce the dust technique in t ignificant advan ion methods. T out in single Besides, it sav . Furthermore, re flowing and because there the granule pr fluence of the g r controlling n is a widely us uropean Journal Europe 2257 (Online)  2 ttp://dx.doi.org/10 pean Jo Journal web anulation ceutical gr aissa Yacoub Hoag 2 and A acy, Misr Internatio ore, Maryland, 2120 acy, Cairo University artment, Faculty of dress: shawky0225@ ABSTRACT The fluid bed g complex proce Therefore, an necessary for c role in determ Plackett‐Burma process variab squares (PLS) the influence o with six factors results reveale that have stron with PLS was u PLS model w chemometric c latent variable were compared Cite this: Eur. J lation techniqu (binder solution path of the spra here and form techniques, the flow characteri sity, uniform bl [1]. FBG is a w the pharmaceu ntages compar The mixing, spra e equipment, w ves on labor c the granules pr more homogen e are many pro roperties. There granulation pro the process [ sed screening d of Chemistry 8 ( ean Journal of Ch 2017 Atlanta Pub 0.5155/eurjchem ournal bpage: www. process v ranules by b Salem 3, Em Ahmed Ibra onal University, Cai 01, USA ty, Cairo, 11562, Egy f Pharmacy, Misr In @gmail.com (A.S. A granulation (FBG ess because ma understanding controlling the p mining the outc an design for scr les on granules regression to pr of the process v s, two levels and ed that the atom ng influence on used to determi was fitted and criteria. The root s (LVs). The pr d with those of t J. Chem. 2017, 8( ue for n) on ay get m gra‐ main istics, ends, widely utical ed to aying which costs, rodu‐ nous. ocess efore, ocess [4‐6]. esign for t prod man trial I size whe that mois dete leve or u Ther gran N stru and mois the radi high dipo (3) (2017) 265‐2 hemistry lishing House LLC m.8.3.265-272.16 of Che eurjchem.com variables y NIR Spe man Saad Elz ahim 2 iro, 11562, Egypt gypt ternational Univer Abouzaid). G) is a wet granu any process var of the influen process. The mo ome of the bat reening of proce properties and redict the moist variables on the d three replicate mizing pressure a the granules pro ine the moisture its predictive t mean square e roposed NIR me he reference LO (3), 265‐272 the identificatio duct quality [7 ny factors can b s [8]. In FBG, the end is reached. T ere the moistur determines sture content ermining the o ls are not mon nder wetting o refore, it is cri nules through th Near Infrared ctive technique can be used sture content d samples are ir ation is absorb her vibrational s ole moment of a 272 C ‐ All rights rese 08 emistry m and mois ectroscop zanfaly 3, rsity, Cairo, 11562, E ulation technique riables can influ nce of the gran oisture content tch. The purpos ess variables in the use of NIR ture content of t granules prope es at the center and the airflow operties. The NI e content of gran performance w error of predicti ethod was valid D method using on of “main fact 7]. The advanta be screened wit d‐point is reach The granule gr re content of g granule densi of granules a outcome of the itored or contr f the powder be itical to predic he granulation p Spectroscopy e that requires in‐line, at‐lin determination. rradiated with bed by the mo state. Only vibr a molecule can a rved ‐ Printed in y sture cont y Egypt. e for producing uence the gran nulation proces of granule also se of this work FBG, study the i spectroscopy an the granules. In erties, Plackett‐B point (15 runs) rate are the pr IR spectroscopy nule in the FBG was evaluated on (RMSEP) wa dated and the re a paired t‐test. tors” that cause age of the PB th a relatively f hed when the t rowth is a com granule is a cri ity and granu also play a cr e batch, and w rolled, this coul ed resulting in ct the moisture process in FBG. (NIRS) is a minimal sampl ne, on‐line and In NIR spectro NIR radiation. olecules bringi ations resulting absorb NIR rad the USA tent granules. It is a ule properties. ss variables is plays a critical k was to apply influence of the nd partial least order to study Burman design was used. The ocess variables y in conjunction . The proposed by traditional s 4.15% with 2 esults obtained e variability in design is that few number of target granule mplex process itical attribute ule size. The ritical role in when moisture ld lead to over batch failures. e levels in the . rapid, nonde‐ le preparation, d off‐line for oscopy [9‐11], . Some of this ing them to a g in changes in iation. 266 Table 1. The ma Batch no P 1 + 2 − 3 + 4 − 5 0 6 + 7 − 8 − 9 − 10 + 11 0 12 + 13 0 14 − 15 + NIRS can properties (e affects the ab OH stretch reported on m NIRS in a FBG wave‐lengths blend in the channel NIR spectroscopy particle size endpoint. Ha granulation o NIR spectrosc container. Ob control of a fl sensor. The variables in F influence on granules (Fig determine th validate the guidelines. 2. Experimen 2.1. Material Acetamin was obtained assay range 5919), Pharm New Jersey, 6387), Avice TN08819980 USA. Magnes source was o USA. Polyvin 6PT0) was ob atrix of Plackett‐Bu Pattern Spr Ato pre +−+++− 9 −−+−++ 5 +−−−+− 9 −−−+−− 5 000000 7 ++−−−+ 9 −+++−− 5 −+−−+− 5 −−+−−+ 5 +−−+−+ 9 000000 7 ++++++ 9 000000 7 −+−+++ 5 +++−−− 9 Fi n be used for .g. moisture co bsorption intens region for wa monitoring the G. Rantanen et a s to determine instrumented moisture sens y to simultaneou in a fluid be artung et al. of an Enalapril copy with the N bregon et al. [1 luidized bed dr aim of this s FBG using Plack the pharma‐ce gure 1). Besid he moisture co developed m ntal ls and reagents nophen (Lot # d from Sigma A was 99.5%. La matose® 200M USA. Microcry el®102 and 0) were gifted b ium stearate (L obtained from nylpyrrolidone, btained from BA Abo urman design for s raying phase omization essure (PSI) igure 1. Fluid bed g r the determin ntent). It is we sity near 1935 ater [12]. Sev moisture conte al. [13] used th the moisture fluid bed gran sor. Findlay et usly monitor m ed granulator [15] monitor l formulation b NIR probe insta 16] developed ryer with an inl tudy was to s kett‐Burman de eutical properti es, to develop ontent of granu method in acc s MKBQ8028V) ldrich Company actose monohy was obtained ystalline cellulo croscarmellose by FMC Biopoly Lot # L06615), MACRON Chem Kollidone® K ASF, New York, uzaid et al. / Eur creening the proce Inlet temp. (°C) 50 50 50 50 60 70 70 70 50 50 60 70 60 70 70 granulation; the in nation of chem ll known that w nm, the well‐kn veral authors ent of granules u ree to four diffe content of po nulator using m t al. [14] used moisture conten and determin red the fluid by means of in alled in the pro a model predi ine NIR as moi screen the pro sign and study ies of the prod p a NIR metho ule in the FBG cordance with ), active ingre y, Missouri, US ydrate (Lot # from DFE Pha ose (Lot # P2 e sodium (Lo ymer, Pennsylv , Hyqual® vege micals Pennsylv K30 (Lot # G USA. ropean Journal of ess variables in flu Air flow rate (SCFM) 12 12 8 8 10 8 12 8 12 8 10 12 10 8 12 nfluence of the pro mical water nown have using erent wder multi‐ d NIR t and ne its bed n‐line oduct ictive sture ocess their duced od to G and ICH dient A. Its 1021 arma, 1382 ot # vania, etable vania, 1097 2.2. C coef usin appl the gran and The zatio spra dryi chan Tabl 2.3. T acet mon w:w) (0.5% solu dien befo Penn proc proc size cons wate 10 m the g temp appa of Chemistry 8 (3) id bed granulation Drying p e Inlet tem (°C) 70 50 50 70 60 50 70 50 50 70 60 70 60 70 50 cess variable on th Experimental Construction o fficients and st ng JMP® softwa lied for screenin influence of th nules. Plackett‐ three replicate granulation pr on pressure, in aying phase and ng time during nged according le 1. Granulation se The granulati aminophen (3 nohydrate (43.5 ), croscarmello %, w:w) and P tion (15%, w:v nt‐excipients we ore mixing in 8 nsylvania, USA cess was per cessor, Model 0 was 500g. Th sisting of 15% er was added d mL/min. After e granules were perature (accor aratus. ) (2017) 265‐272 n. phase mp. Air flo (SCFM 12 12 12 8 10 8 8 12 8 8 10 12 10 12 8 he granule propert design of the experime tatistical param are from SAS. P ng the process v he process vari Burman design es at the cente rocess variable let air tempera d the inlet air te the drying pha g to Plackett‐B et‐up ion formulati 32%, w:w) as 5%, w:w), micr ose sodium (2% PVP (Kollidone v). Appropriate ere weighted an 8 qt V‐blender A) at 30 rpm formed using 002 (Fluid Air® he required am (w:v) solution uring granulati ending the spr dried for varia rding to the PB 2 w rate M) Dr (m 15 25 15 15 20 25 15 15 25 25 20 25 20 25 15 ties. ental design, co meters have be Plackett‐Burma variables in FB ables on the p n with six facto er point (15 ru es investigated ature, airflow ra emperature, air ase. The Process Burman design, on studied active ingred rocrystalline ce %, w:w), magne e® K30) in wa e quantities of nd sieved throu r (Patterson Ke for 5 min. Th a Magnaflow Inc., Illinois, US mount of gran of PVP (Kollid ion at a constan aying of the bi able time period B design matrix rying time min) 5 5 5 5 0 5 5 5 5 5 0 5 0 5 5 omputation of een performed an design was G and to study harmaceutical ors, two levels uns) was used. were: atomi‐ ate during the rflow rate and s variables are , as shown in consisted of dient, lactose ellulose (22%, esium stearate ater as binder f active ingre‐ ugh #18 mesh elly Company, he granulation w® fluid bed SA). The batch nulation liquid done® K30) in nt spray rate of inder solution, ds at different x) in the same Abouzaid et al. / European Journal of Chemistry 8 (3) (2017) 265‐272 267 2.4. Granules characterization The properties evaluated for the granules produced during an experiment were; granule mean diameter, particle size distribution, bulk density (untapped density), tapped density, Hausner ratio, Carr’s index (compressibility index), and moisture content of the granules [6]. The samples powders were manually collected from the sampling port of fluid bed granulator at regular intervals (2 min) during the granulation process. The sample size was 5 g divided into two portions; the first portion for moisture content determination and the second for particle size measurements. 2.4.1. Granule mean diameter and particle size distribution The granule mean diameter and particle size distribution were measured by laser diffraction using the Malvern Mastersizer X/S (Malvern Inc., Worcestershire, UK) and Fraunhofer model analysis routine. The dry powder feeder was operated at an air pressure of 20 psi and a sample size of 3 g. The granule mean diameter was determined by measuring the D[4,3] which are the particle sizes at the 40th and 30th of the cumulative undersize distribution [17]. The particle size distribution is performed by determination the span according to the following equation: (1) Here, D (10), D (50) and D (90) are the particle sizes at the 10th, 50th and 90th percentiles of the cumulative undersize distribution, respectively. 2.4.2. Bulk density, tapped density, Hausner ratio and Compressibility index (Carr’s index) Granules were analyzed for bulk density, tapped density, Hausner ratio and Compressibility index, all these determinations were performed according to the USP method <616> [18] for bulk density, tapped density and USP method <1174> [18] for Hausner ratio and Compressibility index. Bulk density and tapped density were determined using JEL Stampf®Volumeter Model STAV 2003 (Ludwigshafen, Germany). Hausner ratio (HR) is the ratio of the tapped density to its initial bulk density (2) A lower HR value (< 1.25) is generally an indication of good flow in accordance with USP method <1174> [18]. The compressibility Index (CI) was calculated using the values of bulk and tapped density according to the following equation: 100 (3) A lower CI % value (< 20) is generally an indication of good flow in accordance with USP method <1174> [18]. 2.4.3. Moisture content of the granules The moisture content of samples was determined by loss on drying (LOD). To measure LOD, about 2 g of sample was evenly spread on the pan of the moisture analyzer (Mettler Toledo, Model HB43) and the sample weight loss was determined at 105 °C. 2.5. NIR equipment The Metrohm NIRS XDS Rapid Content Analyzer (RCA) was used for offline NIR reflectance measurements. Samples were placed in sealed glass vials and scanned in reflectance mode over a wavelength range of 400 to 2500 nm with data collected every 0.5 nm. 2.6. Software and data analysis Data handling, Principal component analysis (PCA) and Partial least squares (PLS) routine work were done using SOLO®8.0 (Eigenvector Research Inc., Washington, USA). PLS model was applied to the NIR spectra. In order to build PLS model, the raw data was preprocessed using one or a combination of two preprocessing methods [19]. Two types of data preprocessing, namely mean centering (MC) and auto‐ scaling (AS) were used in this study. The root mean square error of prediction (RMSEP) and number of latent variables (LVs) were used to evaluate the performance PLS models [20,21]. 2.7. Method validation The proposed PLS model for the NIR spectroscopy for determination the moisture content of granules was validated in accordance with ICH guidelines [22]. The method linearity, specificity, accuracy and precision (repeatability) are measured for the proposed method [23]. In addition to, the traditional chemometric criteria are calculated to evaluate the predictive ability of the developed PLS models to predict the moisture content of granules [24]. These criteria are regression coefficient (r2), the root mean squared error of cross‐validation (RMSECV) and of prediction (RMSEP) for external validation set, not involved in the calibration set. 3. Results and discussions 3.1. Analysis of the influence of the process variables on the granules properties The Plackett‐Burman design was applied for screening the process variables in fluid bed granulation and to study the influence of process variables on granules physical properties (granule mean diameter and span), granules flow properties (Hausner ratio and Carr’s index) and the moisture content of the granules. The most important six process variables were investigated by Plackett‐Burman design; three variables in the spraying phase (atomization pressure, airflow rate and inlet temperature) and three variables in the drying phase (airflow rate, inlet temp and drying time). Results obtained are represented in Table 2. The regression analysis table (Table 3 and 4) was used to show the effect of the process variables on the properties of granules, it shows the contrasts (regression coefficients) of each variable, t‐Ratio values and p‐values to assess the significant of each variable. The t‐Ratio values are calculated as Contrast/ PSE, where PSE is Pseudo Standard Error. p‐ Values are obtained by the t‐test to assess the significant of each variable. p‐Values more than 0.1 indicate that the variables are not significant. P‐values from 0.05 to 0.10 indicate that the variables are weakly significant. While p‐ values less than 0.05 indicate that the variables are strongly significant. 3.1.1. Analysis of the influence of the process variables on the granule mean diameter and span The airflow rate in the drying phase (p‐value = 0.074) has a weak significant influence on granule mean diameter, as shown in Table 3a. The airflow rate in the drying phase (p‐ value = 0.099), and the interaction between the airflow rate in the spraying phase and inlet temperature in the drying phase (p‐value = 0.085) have a weak significant influence on span, as shown in Table 3b. 268 Abouzaid et al. / European Journal of Chemistry 8 (3) (2017) 265‐272 Table 2. Matrix of the results after screening by Plackett‐Burman design. Batch no Granule mean diameter (μm) Span (μm/μm) Moisture content (%) Bulk density (g/mL) Tapped density (g/mL) Hausner ratio Carr's index (%) 1 159.129 1.405 1.59 0.46 0.57 1.24 19 2 207.922 1.263 2.81 0.47 0.67 1.43 30 3 346.023 1.533 2.74 0.43 0.61 1.42 42 4 157.786 1.738 1.60 0.45 0.60 1.33 25 5 148.236 1.427 2.60 0.46 0.58 1.35 26 6 89.700 1.780 1.98 0.47 0.56 1.19 16 7 168.800 1.812 2.27 0.49 0.56 1.33 25 8 214.690 1.316 3.00 0.45 0.63 1.40 29 9 141.281 1.935 2.53 0.50 0.63 1.26 21 10 158.714 1.446 2.46 0.43 0.53 1.23 19 11 143.521 1.701 2.27 0.46 0.61 1.33 25 12 399.215 2.169 1.18 0.53 0.71 1.34 25 13 158.841 1.688 1.50 0.48 0.60 1.25 20 14 235.093 1.212 2.53 0.44 0.65 1.47 32 15 237.844 1.430 1.94 0.84 0.68 1.42 29 Table 3. The influence of process variables on granule mean diameter (a), span (b), bulk density (c), and tapped density (d). Term a Contrast t‐Ratio Individual p‐Value Screening for granule mean diameter (a) X1 19.7559 0.83 0.3731 X2 13.0055 0.55 0.6169 X3 8.3618 0.35 0.7454 X4 3.0766 0.13 0.8981 X5 45.3137 1.91 0.0735 b X6 ‐3.9017 ‐0.16 0.8705 X5*X5 23.7935 1.00 0.2934 X5*X1 21.7830 0.92 0.3310 X5*X2 18.7009 0.79 0.4011 X1*X2 7.6242 0.32 0.7667 X5*X3 ‐26.0708 ‐1.10 0.2550 X1*X3 ‐36.9951 ‐1.56 0.1269 Null 14 2.1077 0.09 0.9303 Null 15 1.9408 0.08 0.9343 Screening span (b) X1 0.036299 0.67 0.5081 X2 0.029740 0.55 0.6139 X3 0.073716 1.35 0.1687 X4 0.039131 0.72 0.4498 X5 ‐0.092648 ‐1.70 0.0989 b X6 0.042560 0.78 0.4060 X5*X5 ‐0.007500 ‐0.14 0.9002 X5*X3 0.016833 0.31 0.7759 X5*X6 ‐0.030619 ‐0.56 0.6045 X3*X6 0.183212 3.36 0.0135 c X5*X4 0.025538 0.47 0.6659 X3*X4 0.098347 1.81 0.0852 b Null 14 ‐0.056079 ‐1.03 0.2781 Null 15 ‐0.006406 ‐0.12 0.9138 Screening for bulk density (c) X1 0.026833 0.92 0.3323 X2 0.035777 1.23 0.2157 X3 0.046212 1.59 0.1239 X4 ‐0.026833 ‐0.92 0.3323 X5 ‐0.029814 ‐1.03 0.2865 X6 ‐0.020870 ‐0.72 0.4588 X3*X3 0.012000 0.41 0.7039 X3*X2 0.015333 0.53 0.6296 X3*X5 ‐0.005715 ‐0.20 0.8537 X2*X5 0.014491 0.50 0.6475 X3*X1 0.047329 1.63 0.1168 X2*X1 ‐0.017889 ‐0.62 0.5710 Null 14 ‐0.004201 ‐0.14 0.8945 Null 15 0.000354 0.01 0.9924 Screening for tapped density (d) X1 ‐0.005963 ‐0.36 0.7404 X2 0.013416 0.80 0.3974 X3 0.017889 1.07 0.2687 X4 ‐0.011926 ‐0.71 0.4583 X5 0.020870 1.24 0.2029 X6 0.007454 0.44 0.6835 X5*X5 0.008000 0.48 0.6609 X5*X3 ‐0.001333 ‐0.08 0.9402 X5*X2 0.017419 1.04 0.2815 X3*X2 0.016102 0.96 0.3154 X5*X4 0.019720 1.18 0.2254 X3*X4 ‐0.010435 ‐0.62 0.5587 Null 14 ‐0.003181 ‐0.19 0.8568 Null 15 0.004582 0.27 0.7971 a X1, atomizing pressure; X2, inlet air temperature (spraying); X3, air flow rate (spraying); X4, inlet air temperature (drying); X5, air flow rate (drying); X6, drying time. b Weak significant variable. c Strong significant variable. Abouzaid et al. / European Journal of Chemistry 8 (3) (2017) 265‐272 269 Table 4. The influence of process variables on Hausner ratio (a), Carr’s index (b) and moisture content of granule (c). Term a Contrast t‐Ratio Individual p‐Value Screening for Hausner ratio (a) X1 ‐0.028324 ‐1.34 0.1660 X2 0.017889 0.85 0.3561 X3 ‐0.001491 ‐0.07 0.9471 X4 ‐0.013416 ‐0.64 0.5446 X5 0.040249 1.91 0.0683 b X6 ‐0.016398 ‐0.78 0.3982 X5*X5 0.011333 0.54 0.6202 X5*X1 ‐0.014667 ‐0.70 0.4645 X5*X2 0.002994 0.14 0.8948 X1*X2 0.010811 0.51 0.6361 X5*X6 0.050428 2.39 0.0336 c X1*X6 ‐0.009690 ‐0.46 0.6735 Null 14 ‐0.011517 ‐0.55 0.6139 Null 15 ‐0.015514 ‐0.74 0.4305 Screening for Carr’s index (b) X1 ‐0.89443 ‐0.55 0.6168 X2 0 0 1.0000 X3 ‐1.04350 ‐0.64 0.5490 X4 ‐1.63978 ‐1.00 0.2888 X5 3.13050 1.91 0.0717 b X6 ‐1.93793 ‐1.18 0.2197 X5*X5 0.93333 0.57 0.6007 X5*X6 2.80000 1.71 0.0986 b X5*X4 ‐0.51711 ‐0.32 0.7735 X6*X4 1.95519 1.19 0.2159 X5*X3 ‐2.50729 ‐1.53 0.1278 X6*X3 0.67082 0.41 0.7100 Null 14 0.26950 0.16 0.8791 Null 15 1.14243 0.70 0.4729 Screening for moisture content (c) X1 ‐0.212426 ‐2.29 0.0430 c X2 ‐0.061865 ‐0.67 0.5120 X3 ‐0.148326 ‐1.60 0.1192 X4 ‐0.251185 ‐2.71 0.0280 c X5 0.079753 0.86 0.3701 X6 0.026087 0.28 0.7956 X4*X4 0.038333 0.41 0.7056 X4*X1 0.042333 0.46 0.6789 X4*X3 ‐0.021229 ‐0.23 0.8318 X1*X3 ‐0.275687 ‐2.97 0.0226 c X4*X5 0.111734 1.20 0.2197 X1*X5 0.007379 0.08 0.9392 Null 14 0.171531 2.22 0.0468 c Null 15 ‐0.114287 0.12 0.9112 a X1, atomizing pressure; X2, inlet air temperature (spraying); X3, air flow rate (spraying); X4, inlet air temperature (drying); X5, air flow rate (drying); X6, drying time. b Weak significant variable. c Strong significant variable. The interaction between airflow rate in the spraying phase with the drying time (p‐value = 0.014) has a strong significant influence on span. The increase of airflow rate in the spraying phase with atomizing pressure allows us to obtain fine granules (small granule mean diameter) with more broadly granules dispersed (large span). This could be explained by increasing the atomization pressure decreases the moisture content of granules, which leads to a decrease in the granule size, thus obtaining granules with small mean diameter and more broadly dispersed granules (large span) [25]. 3.1.2. Analysis of the influence of the process variables on the bulk and tapped density The process variables were found to have a little influence on the bulk and tapped density which is not significant, as shown in Tables 3c and 3d. The increase of the airflow rate in the spraying phase leads to an increase in density (both bulk and tapped). This could be due to that increasing the airflow rate in the spraying phase leads to denser granules by the spatial configuration of the obtained granules. 3.1.3. Analysis of the influence of the process variables on the granule flow properties; Hausner ratio and Carr’s index To characterize granules flow properties, the Hausner ratio and compressibility index (Carr’s index) were deter‐ mined. The airflow rate in drying phase (p‐value = 0.07) has a weak significant influence and the interaction between airflow rate in drying phase and drying time has a strong significant influence (p‐value = 0.03) on the Hausner ratio, as shown in Table 4a. The airflow rate in drying phase (p‐value = 0.07) and the interaction between airflow rate in drying phase and drying time (p‐value = 0.1) have a weak significant influence on the compressibility index, as shown in Table 4b. The increase of airflow rate (drying phase) in the same time with increasing drying time leads to an increase of both Hausner ratio and Carr’s index, resulting in poorer compressibility and flowability of the granules. Controversy, The increase of atomization pressure leads to granules with better compressibility and flowability properties. Generally, the adhesion force and gravity force are significant forces that act directly on the granules during packing [26]. Therefore, the increase of airflow rate in the drying phase with the drying time leads to increase the granule attrition, resulting in a decrease the granule size and consequently, the flowability decreases. As the granule size decreases, the influence of the gravity force becomes smaller than the adhesion force, and consequently the flowability decreases [26]. 270 3.1.4. Analys the moisture As can be value = 0.04) 0.03) and th phase with significant in the moisture the atomizati inlet air tem explained by decrease the moisture con temperature ration rate, re granules. Nu factors (proce 0.05) has a s the Null 14 p This may be inlet air. 3.2. Moisture 3.2.1. NIR spe The raw process (spra spectral featu shown in Fig region around group. Spectr first overtone 1970 nm, the increases dur drying phase, Figure 2. (a) Ra wavelength ran combination ban 3.2.2. Model c Spectra re validation sa is of the influe e content of the e seen in Table , inlet air temp he interaction b atomization p fluence on the content of the ion pressure, a mperature in t y increasing th e droplet size, ntent of granule (drying phase esulting in a de ll 14 (uncontr ess variables) w significant influ ositively affects due to the rela e content deter ectra NIR spectra aying and dryin ures correspon gure 2a. Peaks d 1450 nm rela ra regions from e from ‐CH, ‐CH e peak correspo ring the sprayin , as shown in Fi aw NIR spectra of nge (1870‐1970 n nd in water. calibration egions from 18 amples were a Abo nce of the proc e granules e 4c, the atomi perature in dryi between airflo pressure (p‐va e moisture cont granules is ne airflow rate in the drying pha he atomization , resulting in es. Besides, incr ) leads to incr ecrease in the m rolled factor) w were exhausted uence on the m s the moisture ative humidity rmination by N obtained duri ng phase) were nding to the m corresponding ates to the first 1600‐1800 nm 2, and ‐CH3 and onds to combina ng phase and d igure 2b. f the calibration sa nm) was selected 870‐1970 nm of nalyzed using uzaid et al. / Eur cess variables o ization pressur ing phase (p‐va w rate in spra alue = 0.02) tent of granule egatively affecte spraying phase ase. This coul n pressure lead a decrease in reasing the inle rease in the ev moisture conte was added afte d. Null 14 (p‐va moisture conten content of gran of the uncontr NIR Spectroscop ng the granul analyzed to ide moisture conten g to the wavele overtone of the m corresponds t d in the region 1 ation band of w decreases durin amples. (b) The sp d, corresponding t f the calibration PCA where 2 ropean Journal of on re (p‐ alue = aying have es. As ed by e and ld be ds to n the et air vapo‐ ent of er all alue = nt. As nules. rolled py lation entify nt, as ength e ‐OH to the 1870‐ water; ng the pectral to OH n and PC’s expl capt show nm Ther 1970 usin obta prim LOD calib the prep in e obta the n LVs) Table mode PLS m Raw AS MC * * Sele ** Sel Figur and v 1935 T mois of Chemistry 8 (3) lained 100% v turing 99.90% wn in Figure 3a which resembl refore, the spe 0 nm to build t ng NIR spectra ained from fifte mary laboratory D). The sampl bration model d raw data wa processing meth xperimental da ained by applyin number of late ), as shown in T e 5. The performa el. models * LV data 5 3 * 2 ected wavelength ( lected pre‐process re 3. (a) PC1 versu validation samples nm and demonstr The selected P sture levels in t ) (2017) 265‐272 variability in t , and 0.10% a. PC 1 loading les the water p ectral range w the PLS model. a acquired fro een granulation y values (the les were insp development. I as preprocesse hods to remove ata. The select ng the mean ce nt variables us Table 5. ance of different te RMSEC (%) 3.55 3.97 4.30 (1870‐1970 nm). sing method. us PC2 score plot fo s. (b) Loadings pl rating the specificit PLS model was the fluid bed gra 2 he data with variability, re plot shows ma peak, as shown was selected be PLS models w m 100 calibra n batches with c moisture cont pected for ou In order to bui ed using diffe e the irrelevant ted PLS model entering (MC) a sed by 60% (fro ested PLS models RMSECV (%) 3.93 4.19 4.46 or the NIR spectra lot of PC1 showin ty of PC1 to water s constructed t anulator. PC1 and PC2 espectively as axima at 1935 n in Figure 3b. etween 1870‐ ere developed ation samples corresponding ent (%) from utliers before ild PLS model, erent spectral t spectral part l was the one as it decreased om 5 LVs to 2 for NIR moisture RMSEP (%) 2.50 3.17 4.15 (a) (b) of the calibration ng the maxima at peak. to predict the Figure 4 LOD% values and RMSECV Figure 4. Regre the measured by 3.2.3. Model v To valida data set (25 predicted wh regression co The linea by establishi content (%) p the reference samples, as sh values are rep method was d extreme sam which was 1.1 NIR method i Figure 5. Regre the measured by The speci in the presen to be prese establishing t principal sour 99.90% varia nm) which is nation band f To establi t‐test for ind moisture con determined b carried out f confirmed th two methods than the ttab (2 The preci is the relation s and the calibra values of 0.958 ession plot betwee y LOD in the calibr validation te the model, th samples) not u hich resulted in oefficient (r2) w arity of the pro ing the predic predicted by th e LOD method hown in Figure presented in Fi determined by mples (lower an 18 (%) and 10.5 s only valid wit ession plot betwee y LOD in the valida ificity is the abi nce of other com nt. The specif the loading plo rce of variation ability) shows m s correspondin for water, as sho ish the accuracy dependent sam ntent (%) predi by the referenc for the twenty e absence of s s; as the texp (1 2.06) at the 95% ision of the pro Abouzaid et al nship between ation model res 8, 4.30%, and 4. en the moisture (% ation set using PLS he moisture co used in the cal n RMSEP value as 0.988. oposed method ction plot betw he model and th d for the twe e 5. The interce igure 5. The ran the LOD metho nd upper) in 59 (%), respect thin this range. en the moisture (% ation set using PLS ility to identify mponents whic ficity was dem t. The loading p n observed in th maximum at the ng to the chara own in Figure 3 y of the propos mples was perfo cted by the NIR e method (LOD ‐five validation ignificant differ .23) for the NI % confidence le oposed method l. / European Jou NIR predicted sulting in R2, RM 46%, respectiv %) predicted by NI S model. ntent of an ext ibration model e of 4.15% and d was demonstr ween the moi hose determine enty five valid ept, the slope an nge of the prop od (%) values o the calibration tively. The prop %) predicted by NI model. the analyte (w ch may be expe monstrated by plot of the PC1 he spectra, captu e wavelength ( cteristic OH co 3b. sed method, a p ormed between R method and t D). The analysis n samples. The rences between IR method was evel. was determine urnal of Chemistr d and MSEC, vely. IR and ternal l was d the rated sture ed by ation nd R2 posed of the n set, posed IR and water) ected y the 1 (the uring 1935 ombi‐ aired n the those s was e test n the s less ed by mea sam anal relat med mois expo ding dete succ 4. Co T Burm stud prop that proc prop airfl allow diam and Carr incre with nega spra The a NI the F NIR pred metr 4.15 NIR with meth meth been succ gran Ackn T Phar supp Mary tech Refe [1]. [2]. [3]. [4]. [5]. [6]. [7]. [8]. [9]. [10]. ry 8 (3) (2017) 2 asuring the rep ples. Repeatab lysis three tim tive standard diate precision sture content o osure time due g atmosphere. ermination the cessfully validat onclusion The first object man design for dy the influen perties. The res the atomizing cess variables t perties. The re ow rate in the ws us to obt meter) with mo better compr r's index and ease of airflow h poor flowabili atively affected aying phase and second objectiv R method to de FBG. For this pu was applied. Th dictive perform ric criteria. The 5% with 2 laten spectroscopy h ICH guidelines hod were com hod using a p n observed. Th cessfully applied nules in FBG. nowledgemen The authors w rmacy, Misr I port as well a yland. Special t hnical assistance erences Appelgren, C. Dr Burggraeve, A.; Eur. J. Pharm. Bi Aulton, M. E. Livingstone, Spa Bouffard, J.; Dum 335(1), 54‐62. Tomuta, I.; Alec 35(9), 1072‐108 Djuris, J.; Medar World J. 2012, 2 Martens, H.; Na Chichester, UK, 1 Tabasi, S. H.; Fah Sci. 2008, 97(9), Siesler, H. W.; spectroscopy: p Sons, 2008. Williams, P.; No and food indust 65‐272 peatability for bility was det es for each sa deviation (RS was not eval of granules may to absorption Therefore, the e moisture ted in accordan tive of this stud r screening of p nce of proce sults obtained f g pressure an that have stron esults also ind e spraying pha tain fine gran ore broadly gra ressibility and Hausner ratio w rate in the dry ity. The moistu by the atomiza d inlet air temp ve of this study etermine the m urpose, a partia he proposed PL mance was eval e root mean sq t variables. The was successfu s. The results o mpared with th paired t‐test. N herefore, the d for determina nts would like to a nternational U as the School thanks to Dr. St e and interest. rug Dev. Ind. Pharm Monteyne, T.; Ve opharm. 2013, 83( The science of ain, 2002. mont, H.; Bertrand u, C.; Rus, L. L.; Le 81. revic, D.; Krstic, M. 2012, 1‐10. aes, T. Multivariat 1989. hmy, R.; Bensley, D , 4052‐4066. Ozaki, Y.; Kawat principles, instrum orris, K. Near‐infra tries. 1987: Ameri r the twenty‐fi termined by r mple on the s SD%) was 2.5 luated in this y increase with of water from developed NI content of g nce with ICH gui dy was to apply process variabl ess variables from the PB de nd the airflow ng influence on dicated that th ase with atomi nules (small g anules disperse flowability pr o). On the oth ying phase lead ure content of t ation pressure, perature in the y was to develo moisture conten al least squares LS model was fi luated by tradi quare error of p e proposed PLS ully validated i btained by the hose of the r No significant d proposed NIR ation the moist acknowledge t University for of Pharmacy, tephen Hoag fo m. 1985, 11(2‐3), 7 rvaet, C.; Remon, (1), 2‐15. f dosage form d d, F.; Legros, R. Int eucuta, S. E. Dev. In .; Vasiljevic, I.; Mas te Calibration, Joh D.; O'Brien, C.; Hoa ta, S.; Heise, H. M ments, application ared technology in can Association of 271 ive validation repeating the ame day. The 0. The inter‐ study as the increasing the the surround‐ R method for granules was idelines. y the Plackett‐ es in FBG and on granules esign revealed rate are the n the granules he increase of izing pressure granule mean d (large span) roperties (low her hand, the ds to granules the granules is airflow rate in drying phase. p and validate t of granule in s model for the itted and their itional chemo‐ prediction was S model for the in accordance proposed NIR eference LOD difference has method was ture content of the Faculty of the financial University of or his support, 25‐741. J. P.; De Beer, T. design, Churchill t. J. Pharm. 2007, nd. Pharm. 2009, sic, I.; Ibric, S. Sci. hn Wiley & Sons. ag, S. W. J. Pharm. M. Near‐infrared ns, John Wiley & n the agricultural f Cereal Chemists 272 Abouzaid et al. / European Journal of Chemistry 8 (3) (2017) 265‐272 Inc., 1987. [11]. Osborne, B. G.; Fearn, T.; Hindle, P. H. Practical NIR spectroscopy with applications in food and beverage analysis, Longman Scientific and Technical, 1993. [12]. Frake, P.; Greenhalgh, D.; Grierson, S. M.; Hempenstall, J. M.; Rudd, D. R. Int. J. Pharm. 1997, 151(1), 75‐80. [13]. Rantanen, J.; Lehtola, S.; Rämet, P.; Mannermaa, J. P.; Yliruusi, J. Powder Technol. 1998, 99(2), 163‐170. [14]. Findlay, P. W.; Peck, G. R.; Morris, K. R. J. Pharm. Sci. 2005, 94(3), 604‐ 612. [15]. Hartung, A.; Knoell, M.; Schmidt, U.; Langguth, P. Drug Dev. Ind. Pharm. 2011, 37(3), 274‐280. [16]. Obregon, L.; Quinones, L.; Velazquez, C. Control Eng. Pract. 2013, 21(4), 509‐517. [17]. Kona, R.; Qu, H.; Mattes, R.; Jancsik, B.; Fahmy, R. M.; Hoag, S. W. Int. J. Pharm. 2013, 452(1), 63‐72. [18]. USP 30 ‐ NF 25. United States Pharmacopeial Convention, Rockville, MD, 2007. [19]. Shawky, A.; Ibrahim, A.; Salem, M. Y.; El Gindy, A. E. Am. Chem. Sci. J. 2014, 4(1), 24‐37. [20]. Szostak, R.; Mazurek, S. J. Mol. Struct. 2004, 704(1), 235‐245. [21]. Mazurek, S.; Szostak, R. J. Pharmaceut. Biomed. 2006, 40(5), 1235‐ 1242. [22]. International Conference on Harmonisation (ICH) of technical Requirements for Registration of Pharmaceuticals for Human Use, Validation of Analytical Procedures, Geneva, 2005. [23]. Peinado, A.; Hammond, J.; Scott, A. J. Pharmaceut. Biomed. 2011, 54(1), 13‐20. [24]. De Bleye, C.; Chavez, P. F.; Mantanus, J.; Marini, R.; Hubert, P.; Rozet, E.; Ziemons, E. J. Pharmaceut. Biomed. 2012, 69, 125‐132. [25]. Hu, X.; Cunningham, J.; Winstead, D. J. Pharm. Sci. 2008, 97(4), 1564‐ 1577. [26]. Otsuka, T.; Iwao, Y.; Miyagishima, A. Int. J. Pharm. 2011, 409(1), 81‐ 88.