ranulation is an important step in pharmaceutical soliddosage form processing. The flow and compressioncharacteristics of a formulation are improved throughgranulation. Also, the content uniformity of the for-

mulation is maintained after blending by the agglomeration ofsmaller particles to form larger particles through granulation(1).

A common method of pharmaceutical granulation is topspray granulation, where the powder is fluidized in a fluidbed dryer and liquid binder solution is sprayed onto theproduct layer from the top counter-currently to the fluidizinggas. After spraying the liquid into the formulation and form-ing the granule the product must be dried to the propermoisture level. If the granules are over-dried the action of thefluid bed can cause the fracture of granules creating undesir-able fines and can damage the formulation due to hydrationchanges in some actives and excipients (2). If the granules arenot dry enough the product will not flow properly and cancake and cause problems with subsequent processing, includ-ing product sticking to the faces of the tablet press punchesand problems with product stability during storage. Samplestypically are withdrawn from the fluid bed with a thief dur-ing processing and analyzed off-line in a laboratory for mois-

ture content. Commonly there is a delay before analysisresults are available to the operator that causes processingdecisions, like end-point determination, to be made withoutoptimal product moisture information. Top spray granula-tion end point is often based on time or product temperatureand not moisture content.

Near-infrared (NIR) spectroscopy is a rapid non-destruc-tive technique often used for in-process analysis of moisturein the manufacturing environment (3). Real-time measure-ments can be made with no sample prep and the data can beanalyzed and stored automatically. NIR fits in well with theProcess Analytical Technology (PAT) initiative as developedby FDA (4–7). One of the elements of the PAT initiative is touse in-line analysis to increase process understanding andcontrol to verify product quality and release it for subsequentprocessing without delay (8). Using NIR the process can bemonitored for low levels of residual moisture and alcoholsand other process constituents to yield better process controland end-point determination (9).

ExperimentalAll NIR spectra contained in this study were collected using aFOSS NIRSystems XDS Process Analyzer and Vision software.

In-line Process Analysis of ResidualMoisture in a Fluid BedGranulator–Dryer Using NIRSpectroscopyThe authors describe the in-line moisture measurement of a pharmaceutical granulation oflactose, microcrystalline cellulose and crospovidone in a fluid bed granulator–dryer usingtop sprayed granulating liquid. A near-infrared (NIR) prediction model was developed formoisture on spectra collected during a calibration run. Subsequent granulations were ana-lyzed for moisture content real-time throughout the granulation and drying process usingthe NIR process instrument.Robert A. Mattes, Denise E. Root, and Andrew P. Birkmire

G

The Role of Spectroscopy inThe Role of Spectroscopy in Supplement to

ProcessAnalytical

Technologies

ProcessAnalytical

Technologies

Each spectrum consists of 16 co-addedscans of sample and reference in theNIR range of 800–2100 nm. Theprocess instrument is a rugged designthat can be equipped for explosion-proof environments. It has a fiber opticprobe that can be inserted into aprocess vessel at a location remote fromthe instrument.

The process instrument was set upand allowed to equilibrate to tempera-ture. A probe of novel design, specifi-cally for the fluid bed application, wasinserted into a Niro MP 2/3 PrecisionGranulator at a 45° angle to the centralaxis of the product container as seen inFigures 1 and 2. Note the collection

“spoon” and purge vents located on theprobe tip. After each NIR spectrumwas collected, the software sent a “datacomplete” signal that energized an airpurge exiting through the ports in theprobe and cleared the “spoon” for anew sample.

A charge of lactose (Pharmatose200M, DMV), microcrystalline cellu-lose and (Avicel PH 101, FMC), andcrospovidone (Polyplasdone XL 10,ISP) was prepared by Niro and loadedinto the product container. The prod-uct was fluidized for 5 min to blendand dry the mixture to homogeneity.An aqueous solution of 15% polyvinylpyrrolidone (Plasdone K29/32) was

Figure 1. The 1-in. NIR probe is inserted in the port at a 45° angle in the product container of the fluid bed dryer.

Figure 2. The novel design fluid bed NIR probe inserted into the product container of the fluid bed dryer. Note the collection “spoon” and purge vents.

Figure 3. Raw spectra taken in-process in the fluid bed dryer.

0.7975

0.7206

0.6438

0.5670

0.4901

0.4133

0.3365

0.2596

0.1828

0.1059

0.0291

900 1100 1300 1500 1700 1900 2100

Wavelength (nm)

Ab

sorb

ance

Figure 4. Second derivative of spectra taken in-process in the fluid bed dryer.

0.0438

0.0268

0.0101

�0.0067

�0.0235

�0.0402

�0.0570

�0.0738

�0.0905

�0.1073

�0.1241

900 1100 1300 1500 1700 1900 2100

Wavelength (nm)

Inte

nsi

ty

Figure 5. An analytical wavelength region used for moisture analysis.

0.0405

0.0353

0.0301

0.0249

0.0196

0.0144

0.0092

0.0040

�0.0013

�0.0065

�0.0117

1842 1851 1860 1863 1873 1887 1896 1905 1914 1923 1932

Wavelength (nm)

Inte

nsi

ty

9.6 9.2 8.8 8.4 8.0 7.6 7.2 6.8 6.46.05.65.24.84.44.0

4.0 4.6 5.2 5.8 6.4 7.0 7.6 8.2 8.8 9.4 10.0

LOD moisture (%)

NIR

pre

dic

tio

n m

ois

ture

(%

)

Figure 6. A PLS model was developed with an R2 value of 0.9896 and a SEC of 0.2171.

added by top spray at 1.5-bar atomiza-tion pressure. The fluidizing airflowand liquid spray rate were increasedtwice during the batch as the granulesformed. NIR spectra were collectedevery 40 s during the blending opera-tion and samples for loss on drying(LOD) analysis were withdrawn atapproximately 5-min intervals. A 2.0-gsample was analyzed for LOD at 160 °Cfor 15 min using the Mettler Toledo(Columbus, OH) HR73 halogen mois-ture analyzer. When the water–bindersolution pump was stopped the dryingprocess began. The drying operationwas uniform and gradual over a period

of 15 min.

Results and DiscussionFigure 3 shows the raw spectra of thedryer samples. Water absorbs stronglyin the NIR around 1400 nm and 1900nm as evidenced by the peaks in thoseregions. Figure 4 shows the secondderivative of the same spectra. Thesecond derivative math treatment isused commonly in NIR spectroscopyto minimize baseline offset caused byscattering and enhance absorbancepeaks (10). The second derivativespectral peaks appear inverted withrespect to the raw spectra (11). Figure5 shows an enlargement of a spectralregion that was used to model themoisture in the samples. A two-factor

partial least squares (PLS) regressionmodel was developed with spectrafrom a calibration run and loss-on-drying (LOD) reference values (seeTable I). The second derivative inten-sity over the range 900–2100 nm wasused to develop a prediction modelwith an R2 value of 0.9896 and a standard error of calibration (SEC) of 0.2171. See Figure 6 for a plot ofNIR predicted versus LOD % mois-ture. Although the prediction modelperformed well, it would be more ro-bust with more calibration samplesincluded.

Figure 7 shows a typical routineanalysis output trend chart. Routineanalysis methods can be developed inthe software to include qualitativeand quantitative analysis methodsand custom output graphics for real-time visual monitoring as well aselectronic process control.

Figure 8 shows moisture predic-tions for three top spray granulationsusing 1.5-bar atomization pressureplotted on the same graph. Table IIshows comparative data of NIR ver-sus LOD for the same three granula-tions. The LOD values demonstrated

Figure 7. Process trend chart from routine analysis for granulation/drying operation. This typical software routine analysis output shows the predicted moisture trend during granulation and drying.

13 12 11 10

9 8 7 6 5 4 3

09:37 13:37 17:37 21:37 25:37 29:37 33:37 37:37 41:37 45:37 49:37 53:37 57:37 00:43

Time

Qu

anti

tati

ve r

esu

lt

Figure 8. Three NIR moisture prediction runs monitoring the fluid bed granulator/dryer.

14

12

10

8

6

4

2

0

0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 45.00 50.00

Time (min)

Mo

istu

re (

%)

Series1Series2Series3

Figure 9. NIR predicted versus LOD values of validation set. The standard error of prediction is 0.4232.

10.0

9.5

9.0

8.5

8.0

7.5

7.0

6.5

6.0

5.5

5.05.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0

LOD moisture (%)

NIR

pre

dic

tio

n m

ois

ture

(%

)

Table I. Calibration sample set data.Sample # NIR Prediction % LOD % Residual %

0019 5.80 5.7 0.100039 7.26 7.46 -0.200059 9.83 9.64 0.190081 7.88 8.04 -0.160089 6.19 6.36 -0.170095 5.51 5.3 0.210103 4.72 4.69 0.03

from 0.03 % to 0.58 % moisturerepeatability error with higher errorat the higher moisture levels. Therealso are errors due to LOD samplingand the representative capability ofthe thief system.

Figure 9 shows the NIR predictedmoisture versus LOD value. The stan-dard error of prediction is 0.4232%.The LOD standard error was estimated

to be 0.33% moisture and there was adelay of more than a week betweensample collection and analysis thataccounts for deviations. The modelaccuracy would be improved with KarlFischer reference data analyzed in amore timely manner.

The endpoint determination can bemade when the moisture level asymp-totically approaches a lower limit dur-ing the drying cycle. As seen in Figure8, the change in moisture reaches aminimum when the product is dry. Theoperator is aided in making the deci-sion to end the drying operation beforethe product is damaged or degraded.The delay caused by waiting for labresults before the product can bereleased for subsequent processing canbe minimized or eliminated. Outputfrom the NIR computer could be usedby the fluid bed dryer’s programmablelogic controller (PLC) for closed loop

process control decisions.

ConclusionThe NIR process instrument demon-strated the ability to predict the mois-ture content of a pharmaceutical granu-lation of lactose, microcrystallinecellulose and crospovidone being driedafter wet granulation in a fluid beddryer. Endpoint determination can be

made when the moisture level asymp-totically approaches a lower limit dur-ing the drying cycle. This trial alsodemonstrated the ability of the novelfluid bed probe to measure a fluidizedsample for residual moisture. Althoughthe prediction model performed well, itwould be more robust with more cali-bration samples included. The modelaccuracy would be improved with KarlFischer reference data analyzed in atimely manner. The correct NIR probemust be placed in the product containerin a manner that provides sufficientsample contact with the probe tip win-dow. Correct probe design and properplacement in process equipment is ofhigh importance for success of NIR im-plementations. Future work will evalu-ate the ability to monitor other residualgranulating liquids and constituent lev-els using the same instrument/probeconfiguration.

References1. A.G. Rogers, “Granulation and Drying

Principles”, “Hands-on” PostgraduateCourse in Tablet Technology, Univ.Tenn., Memphis (2003).

2. S.M. Maggard, D. E. Root, and M. Duell, J. Process Analytical Chemistry 7(1)(2002).

3. K.A. Bakeev, Spectroscopy 19(1) (2004).4. M.L. Balboni, Pharm. Tech. 27(10)

(2003).5. US FDA Draft Guidance “PAT – A Frame-

work for Innovative PharmaceuticalManufacturing and Quality Assurance,”August 2003,http://www.fda.gov/cder/OPS/PAT.htm.

6. H. Forcinio, Spectroscopy 18(9) 16–24(2003).

7. R.C. Lyon, E.H. Jefferson, C.D. Ellison, L.F. Buhse, J.A. Spencer, M.M. Nasr, and A.S. Hussain, Am. Pharm. Rev.6(3) (2003).

8. A.M. Afnan, J. Process Analytical Technology 1(1) (2004).

9. R.A. Mattes, R. Schroeder, V. Dhopesh-warkar, R. Kowal, and W. Randolph,

“Monitoring Granulation Drying UsingNear-Infrared Spectroscopy for In SituAnalysis of Residual Moisture andMethanol”, Pharmaceutical Technology,Process Analytical Technology Supple-ment, September 2004.

10. T.C. O’Haver and T. Begley, Anal. Chem.53, 1876–1878 (1981).

11. H. Mark and J. Workman Jr., Spec-troscopy 18(4) (2003). ■

Robert A. Mattes is an instrumenta-tion scientist and Denise E. Root ismarketing and process instrument managerat FOSS NIRSystems, Inc. (Laurel, MD).Andrew P. Birkmire is a processengineer at Niro Pharma Systems, Inc.(Columbia, MD).Address correspondence to: rmattes@foss-nirsystems.com.

Table II. NIR versus LOD values for three top spray granulations.Batch Elapsed Time NIR LOD Residual

(min.)9/14#3 12.00 7.04 6.92 0.129/14#3 18.66 8.18 7.98 0.209/14#3 27.33 9.56 9.55 0.019/14#3 34.66 6.27 6.06 0.219/14#4 14.66 7.13 6.77 0.369/14#4 22.66 9.33 8.56 0.779/14#4 28.66 9.71 9.46 0.259/14#4 37.99 6.27 6.07 0.209/16#2 9.33 8.22 7.34 0.889/16#2 18.66 8.53 8.25 0.289/16#2 22.66 8.69 9.3 -0.619/16#2 36.66 5.92 5.92 0.00

ONE OF THE ELEMENTS OF THE PAT INITIATIVE IS TO USE IN-LINE ANALYSIS TO

INCREASE PROCESS UNDERSTANDING AND CONTROL TO VERIFY PRODUCT QUALITY AND

RELEASE IT FOR SUBSEQUENT PROCESSING WITHOUT DELAY.

© Reprinted from SPECTROSCOPY, January 2005 Printed in U.S.A.Copyright Notice Copyright by Advanstar Communications Inc. Advanstar Communications Inc. retains all rights to this article. This article may only be viewed or printed (1) for personal use. User may not

actively save any text or graphics/photos to local hard drives or duplicate this article in whole or in part, in any medium. Advanstar Communications Inc. home page is located at http://www.advanstar.com.