Upload
niknenad
View
8
Download
0
Embed Size (px)
DESCRIPTION
NIR PAT
Citation preview
Robert Mattes* is an applications scientist at FOSSNIRSystems, Inc., 7703 Montpelier Rd., Laurel, MD 20723tel. 301.680.7251, fax 301.236.0134,[email protected], and Denise Root is a marketing manager atFOSS NIRSystems. Om Anand, Maria Gerald Rajan,Namrata R. Trivedi, and Wen Qu are graduate students,and Yingxu Peng, PhD, and Yichun Sun, PhD, arepostdoctorate fellows, Department of Pharmaceutical Sciences,University of Tennessee, Memphis, TN.
*To whom all correspondence should be addressed.
Submitted: Nov. 11, 2006. Accepted: Jan. 17, 2007.
Key words: content uniformity, near-infrared, process analytical technology
Near-Infrared Assay and Content Uniformity of TabletsRobert Mattes, Denise Root, Om Anand, Maria Gerald Rajan, Namrata R. Trivedi, Wen Qu, Yingxu Peng, Yichun Sun
Near-infrared (NIR) assay and content uniformity
of tablets provide fast, accurate means of
monitoring tablet production that are in step with
FDA’s process analytical technology initiative. The
authors discuss the process for testing a newly
released NIR tablet analyzer to determine
instrument precision and accuracy using
chlorpheniramine maleate tablets. The data show
promising results that could relieve laboratory
workload of high-performance liquid
chromatography analysis and bring analysis closer
to real time for process monitoring.
ear-infrared spectroscopy (NIRS) is an analyticaltechnique based on absorption measured in thenear-infrared region of the electromagnetic spec-trum that is between the visible and the
mid-infrared (IR). The fundamental absorption bands offunctional groups occur in the mid-IR and are very strong.Usually, potassium bromide pellets, mulls, or dilutions arerequired to bring the absorbances within the linear range ofthe mid-IR detector. The overtone absorptions of thesefundamental bands occur in the NIR spectral region andallow direct measurement without sample preparationbecause of the relative weakness of absorption. The OH,CH, NH, and SH bonds have the strongest overtoneabsorbances in the NIR region (1).
There is considerable interest in the ability to test solid-dosage form samples more frequently than the 10 per batchspecified by the US Pharmacopeia monograph on contentuniformity. Interest has increased in using NIR for tabletassay and content-uniformity testing because of concerns ofthe European Union for better statistically based samplingand the US Food and Drug Administration’s initiative onprocess analytical technology (PAT) for better under-standing and monitoring of production (2). NIR can beused as a rapid at-line analysis method to obtain processingfeedback in near real time during a tableting campaign.Transmission NIRS through the tablet has been preferred toreflectance NIRS because of heterogeneity within tablets (3,4). The reflection NIRS technique may be used for coatinganalysis, but for bulk tablet analysis, the transmission NIRStechnique may yield more consistent results.
Laboratory methods for tablet assay and content unifor-mity are usually time-consuming because they routinely aredone by high-performance liquid chromatography (HPLC),which requires lengthy calibration runs, the mixing ofbuffers, and the procurement and disposal of volatilesolvents. Analyzing 10 tablets for content uniformity maytake hours, and the results may not be available to tablet-press operators or for batch release for many days or evenweeks after the tablets are compressed. Statistical processcontrol (SPC) techniques can be applied while measuringthe tablets with NIR in real time during tableting so thatassay and content-uniformity problems can be detectedbefore they go beyond acceptable limits.
N
Figure 1: Examples of tablets being scanned in transmissionusing the novel near-infrared spectrometer. In the insert on theleft is a tablet tray, and to the right is the standards traycontaining the wavelength standard (traceable to NIST SRM-2035) and photometric standards.
ANALYTICAL METHODS
256361:3-column 6/15/07 2:50 PM Page 170
ANALYTICAL METHODS
ExperimentalFive batches of tablets (0.25-in. diameter and 100-mgweight) with 0 mg (placebo tablets), 0.1 mg, 0.5 mg, 1.0 mg,and 2.0 mg of chlorpheniramine maleate (CPM) per tabletwere formulated and compressed on a tablet press (HT-AP18 SS-U/I rotary tablet press, Elizabeth Hata International,Inc., North Huntingdon, PA).
The NIR instrument used in the study was XDSMasterLab (FOSS NIRSystems, Laurel, MD), which wascapable of automatically measuring multiple tablets afterthey were positioned in a special tray (see Figure 1). In theinsert, a tablet tray is to the left, and to the right is the stan-dards tray containing the wavelength standard (traceable toNIST SRM-2035) for photometric standards. The universaltablet tray used for this study had 20 positions for four
different tablet sizes and five positions for the 0.25-in.diameter tablets under test. The tray was loaded twice toscan all 10 tablets. The 10 tablets were scanned in less than5 min, taking a reference spectrum before scanning each setof five tablets. Spectra were collected in the transmissionmode from 800 nm to 1650 nm with 0.5-nm data intervals,and 32 scans were coadded to produce a single spectrum.HPLC analysis was run on each individual calibration andvalidation tablet after spectra were collected with the NIRinstrument. The HPLC reference values and the NIRspectra were used to develop the regression model.
DiscussionFigure 2 shows the raw NIR spectra from the calibration setand a spectrum of pure CPM in green. The pure CPM
5.4000
5.1100
4.8200
4.5300
4.2400
3.9500
3.6600
3.3700
3.0800
2.7900
2.5000
800 846 893 939 986 1032 1079 1125 1171 1218 1264 1311 1357 1404
Ab
sorb
ance
Wavelength (nm)
Figure 2: Raw spectra of calibration samples. The green spectrumis of pure chlorpheniramine maleate in reflectance to show wherethe absorption occurs at 1138 nm.
0.1725
0.1407
0.1088
0.0770
0.0452
0.0133
-0.0185
-0.0503
-0.0822
-0.1140
-0.1459
800 850 900 950 1000 1050 1100 1150 1200 1250 1300 1400 1450 1500
Inte
nsi
ty
Wavelength (nm)
1550 1600 1650
Figure 3: Second derivative mathematical treatment of calibrationspectra. The absorption band at 1138 nm fans out from theconcentration of chlorpheniramine maleate.
8.8377
8.4304
8.0232
7.6159
7.2086
6.8014
6.3941
5.9869
5.5796
5.1724
4.7651
1123 1128 1133 1138 1143 1148 1153
Ab
sorb
ance
X 1
0-5
Wavelength (nm)
Placebo
0.1 mg
0.5 mg
1.0 mg
2.0 mg
Figure 4: Shows the fanning out of the analytical region of thespectrum where the chlorpheniramine maleate has a strongabsorption band at 1138 nm. Blue is placebo, red is 0.1 mg, brownis 0.5 mg, dark blue is 1.0 mg, and pink is 2.0 mg.
2.9501
2.2068
1.4636
0.7204
-0.0228
-0.7660
-1.5093
-2.2525
-2.9957
-3.7369
-4.4822
1103 1124 1145 1166 1187 1206 1229 1250 1271 1292 1313 1334 1355 1376
Wavelength (nm)
Ab
sorb
ance
X 1
0-5
Figure 5: PLS factor loadings around the 1138 nm absorption forchlorpheniramine maleate and thickness-correction regions.Thickness correction was applied between 1250 nm and 1350 nm.
256361:3-column 6/15/07 2:50 PM Page 171
spectrum was scanned in reflectance and multiplied by ascaling factor to superimpose it over the transmission cali-bration spectra. By taking the second derivative of thespectra, as shown in Figure 3, the baseline was normalizedand the spectral features were enhanced so that the fanningout of the analytical region for CPM was observed at 1138nm. Figure 4 shows the expanded analytical band demon-strating the linear response from 0.1 mg to 2.0 mg CPM.Smoothing was done on the derivative with a segment of 10and a gap of zero. A thickness correction was applied as amath pretreatment to correct for tablet thickness anddensity variance over the region of 1250–1350 nm. The rawspectral variance of the calibration set is large over thisregion, and the pure CPM has an absorption minimum asseen in Figure 2. Thickness correction is a normalizationfunction offered in Vision software (Foss NIRSystems, Inc.)
as a spectral mathematical pretreatment used to correct forpath length variance. The integral correction factor wascalculated over the range of 1250–1350 nm with a unity-scaling factor.
Integral correction factor =
0.1510
0.1214
0.0918
0.0622
0.0326
0.0030
-0.0266
-0.0562
-0.0858
-1154
-0.1450
1111 1132 1153 1174 1195 1216 1237 1258 1279 1300 1321 1342 1363 1384
Wavelength (nm)
Inte
nsi
ty
Figure 6: PLS factor weights in the analytical region.
10.000
1.000
0.100
0.010
0.0011 2 3 4 5 6 7 8 9 10
log
(PRE
SS)
Wavelength (nm)
PRESS, using 6 Factors
Table II: Repeatability results for 0.1 mg and 0.5 mgtablets of chlorpheniramine maleate (CPM).
Tablet innumberone trayposition
NIRPredictedmg CPM
HPLC mg CPM
ResidualAverage precisionand average biasfor all 5 tablets
0.1 mg 0.103Individual tabletdata not shown
Tablet 1a 0.099 �0.004
Tablet 1a 0.102 �0.001
Tablet 1a 0.107 0.004
Tablet 1a 0.104 0.001
Tablet 1a 0.109 0.006
Tablet 1a 0.107 0.004
Tablet 1a 0.114 0.011
Tablet 1a 0.107 0.004
Tablet 1a 0.101 �0.002
Tablet 1a 0.114 0.011
Precision 0.0051 0.0039
Bias 0.0034 0.00180.5 mgtablet
0.512
Tablet 1a 0.513 0.001
Tablet 1a 0.526 0.014
Tablet 1a 0.527 0.015
Tablet 1a 0.534 0.022
Tablet 1a 0.521 0.009
Tablet 1a 0.511 �0.001
Tablet 1a 0.529 0.017
Tablet 1a 0.528 0.016
Tablet 1a 0.52 0.512 0.008
Tablet 1a 0.517 0.512 0.005
Precision 0.007442 0.0055
Bias 0.0106 0.0057
*NIR is near-infrared; HPLC is high-performance liquid chromatography.
Table I: Prediction equation statistics.
Number offactors
R2 SEC PRESS F-statistic Xval
6 0.9998 0.0119 0.0095 24093 0.0148
*R2 is the coefficient of determination. SEC is standard error ofcalibration; PRESS is predicted residual error sum of squares; and Xvalis the standard error of cross validation using the leave-one-out method.
Figure 7: Predicted residual error sum of squares (PRESS) plot forfor factors used in the model. Where PRESS reaches a minimum isusually considered the maximum number of factors. The PRESSusing six factors was 0.0095.
256361:3-column 6/15/07 2:50 PM Page 172
The correction factor equals the 0.5-nm increment at which the data werecollected times the sum of the y-axisvalues (S�i) and the y-axis value plus 1(S�i+1), divided by 2. Then, the originalspectrum was divided by this correc-tion factor throughout the analyticalregion of 1120–1380 nm.
Partial least squares (PLS) regres-sion was used to develop theprediction model. PLS uses principalcomponent analysis and is a variationof principal component regression(PCR). The correct number of prin-cipal components or factors wasdetermined by the Vision softwaresupplied with the instrument by deter-mining where the predicted residualerror sum of squares (PRESS) reachesa minimum (5).
Figures 5 and 6 are plots of the PLSfactor loadings and weights around the1138-nm absorption band for CPM.The loadings and weights appearspectra-like and are not noisy, indi-cating good modeling attributes forthe factors chosen. Figure 7 is a plot ofthe PRESS leading to a model witheight factors. The model chosen usedonly six of these factors, trading
decreased error for robustness (6). ThePRESS for six factors was 0.0095. Theresulting model had a multiple corre-lation coefficient (R2) value of 0.9998and a standard error of calibration(SEC) of 0.0119. The one-left-outcross-validation demonstrates goodpredictability with a standard error ofcross-validation of 0.0148.
Table I contains the model statisticsfor the CPM prediction equation. TableII is the repeatability results for fivetablets measured 10 times each ofnominal 0.1 mg CPM and 10 tablets of0.5 mg CPM. The same tablet placed inthe same tray position was scanned 10times. Data for tablet tray position(number 1) are shown, and only thecombined statistical results are shownfor the other four tablets from eachdosage level. The average precision forthe nominal 0.1-mg level was 0.0039.The precision for the nominal 0.5-mglevel was 0.0055. The bias was 0.0018for the lower-level CPM and 0.0057 forthe higher level. Table III contains theresults from scanning the ten 0.1 mgCPM tablets for content uniformity.Table IV contains the results from scan-ning the ten 0.5-mg CPM tablets for
content uniformity. The Vision soft-ware has a convenient routine analysismethod for calculating content unifor-mity automatically. Figures 8 and 9 areX control charts for the 0.1-mg and 0.5-mg CPM content uniformity tests.These charts are for SPC, plotting targetlabel claim and �15% control limits.The HPLC results showed that some ofthe nominal 0.5-mg CPM tablets wereas high as 0.53 mg CPM.
Figure 10 shows the NIR predictedCPM concentrations versus the HPLCresults for each tablet in the calibrationset. One tablet was left out of the cali-bration set at each level forprediction-model validation. Figure 11shows the NIR predictions of the vali-dation set versus the HPLC results forCPM on each tablet.
Better precision and accuracy can beachieved with a training set designedwith smaller increments around thetarget label claim. Tablets from on-linepress processing can be scanned andsent to the laboratory for HPLC analysisand selected for calibration samples tocover the range using a few extra pilot-batch samples needed to extend therange to �15% of label claim.
0.115
0.100
0.085
X Control chart
LCL
UCL
��
Sample number
mg
Figure 8: Statistical process control chart of 0.1-mg chlorpheniramine maleate content uniformity. The X control chart plots the target labelclaim with � 15% control limits. UCL is the upper control limit, and LCL is the lower control limit.
Figure 9: Statistical process control chart of 0.5-mg chlorpheniramine maleate content uniformity. The X control chart plots the target labelclaim with �15% control limits. UCL is upper control limit, and LCL is lower control limit.
0.575
0.500
0.425
X Control chart
LCL
UCL
��
Sample number
mg
256361:3-column 6/15/07 2:50 PM Page 173
Table III: Content uniformity test results for ten 0.1-mgchlorpheniramine maleate tablets.
Test level 1
Sample Target Test result*Percentage of
targetPass or fail
1 0.100 0.091 91.491 Pass
2 0.100 0.099 98.655 Pass
3 0.100 0.103 103.421 Pass
4 0.100 0.101 101.487 Pass
5 0.100 0.100 99.782 Pass
6 0.100 0.099 98.910 Pass
7 0.100 0.101 101.139 Pass
8 0.100 0.100 99.987 Pass
9 0.100 0.100 99.649 Pass
10 0.100 0.102 101.616 Pass
* Relative standard deviation is 3.2%.
Table IV: Content uniformity test results for ten 0.5-mgchlorpheniramine maleate (CPM) tablets.
Test level 1
Sample Target Test result*Percentage of
targetPass or fail
1 0.500 0.523 104.642 Pass
2 0.500 0.520 104.094 Pass
3 0.500 0.530 106.097 Pass
4 0.500 0.525 104.947 Pass
5 0.500 0.529 105.804 Pass
6 0.500 0.526 105.106 Pass
7 0.500 0.525 104.950 Pass
8 0.500 0.529 105.877 Pass
9 0.500 0.524 104.717 Pass
10 0.500 0.531 106.248 Pass
* Relative standard deviation is 0.7%.
ConclusionNear-infrared assay and content uniformity of tabletsprovides a fast, accurate means of monitoring tablets forproduction that is in step with the US Food and DrugAdministration’s process analytical technology initiative.The data show promising results that could relieve labora-tory workload of high-performance liquid chromatographyanalysis and bring analysis closer to “real time” for processmonitoring. Ten tablets can be analyzed in �5 min. Thesoftware provided with the instrument is used for datacollection and developing prediction models. The softwarealso provides dedicated routine analysis methods forcontent-uniformity analysis yielding results in percent labelclaim and percent relative standard deviation as well aspass–fail indication.
The average repeatability result for five different tabletsmeasured 10 times of nominal 0.1-mg chlorpheniraminemaleate was 0.0039 with a bias of 0.0018. The repeatabilityresult for 5 tablets of 0.5 mg chlorpheniramine maleate was
0.0055 with a bias of 0.0057. Better precision and accuracycan be achieved with a training set designed with smallerincrements around the target label claim.
References1. Handbook of Near-Infrared Analysis, D.A. Burns and E.W.
Ciuczak, Eds.(Marcel Dekker, Inc., New York, NY, 2001).2. US Food and Drug Administration, Guidance for Industry: PAT—
A Framework for Innovative Pharmaceutical Development,Manufacturing, and Quality Assurance, (FDA,Rockville, MD, Sept.2004).
3. P.J. Larkin, E. Fruhling, and C. Longfellow, “Comparison ofFourier Transform (FT) and Grating Based NIR Spectrometersfor Content Uniformity of Pharmaceutical Solid Dosage Forms,”Am. Pharm. Review. 9 (6), 102–109 (2006).
4. Q. Ji et al., “Rapid Content Uniformity Determination of Low-Dose TCH 346 Tablets by NIR,” Am. Pharm. Review 9 (5), 20–26(2006).
5. R. Kramer, Chemometric Techniques for Quantitative Analysis,(Marcel Dekker, Inc., New York, NY, 1998).
6. K.R. Beebe, R.J. Pell, and M.B. Seasholtz, Chemometrics: A Prac-tical Guide, (John Wiley & Sons, Hoboken, NJ, 1998). PT
3.0
2.6
2.2
1.8
1.4
1.0
0.6
0.2
-0.2
-0.6
-1.0
-1.0 -0.6 -0.2 0.2 0.6 1.0 1.4 1.8 2.2 2.6 3.0
HPLC CPM mg
NIR
Pre
dic
ted
CPM
mg
Calibration Set: NIR Predicted vs Lab Data
Figure 10 (Left): Calibration set. Predictedversus high-performance liquidchromatography (HPLC). CPM ischlorpheniramine maleate.
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
-0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-1.0 -0.7 -0.4 -0.1 0.2 0.5 0.8 1.1 1.4 1.7 2.0
HPLC CPM mg
NIR
Pre
dic
ted
CPM
mg
Validation Set: NIR Predicted vs Lab Data
Figure 11 (Right): Validation set. One tabletwas left out of the calibration set at eachlevel for production-model validation. CPM is chlorpheniramine maleate.
© Reprinted from PHARMACEUTICAL TECHNOLOGY, April 2007 Printed in U.S.A.
256361:3-column 6/15/07 2:51 PM Page 174
7703 Montpelier RoadLaurel, MD 20723
U. S. A.
T: 301.680.9600F: 301.236.0134
256361:3-column 6/15/07 2:51 PM Page 175