- Email: [email protected]
Prediction of Tablet Hardness Based on Near Infrared Spectra of Raw Mixed Powders by Chemometrics
Prediction of Tablet Hardness Based on Near Infrared Spectra of Raw Mixed Powders by Chemometrics MAKOTO OTSUKA,1 IKURO YAMANE2 1
Research Institute of Pharmaceutical Sciences, Faculty of Pharmacy, Musashino University, Shinmachi, Nishi-Tokyo 202-8585, Japan 2
Department of Pharmaceutical Technology, Kobe Pharmaceutical University, Motoyama-Kitamachi, Higashi-Nada, Kobe 658-8558, Japan
Received 19 January 2005; revised 4 September 2005; accepted 12 September 2005 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20514
ABSTRACT: The purpose of this research is to elucidate the effect of lubricant mixing on tablet hardness by near-infrared (NIR) chemometrics as a basic study of process analytical technology. Formulation cellulose (F-C) consisted of sulpyrine (SP), microcrystalline cellulose (MC), and magnesium stearate (MgSt). Formulation lactose/starch (F-L) consisted of SP bulk drug powder, spray-dried lactose (SL), corn starch (CS), and MgSt. First, F-L and F-C without MgSt were mixed in a twin-shell mixer for 60 min. MgSt was added to the mixed powder, and was mixed for various mixing times, after which the mixed powders were compressed by 8-mm diameter punch and die. NIR spectra of raw mixed powders of F-L and F-C were taken using a reflection type of Fourier transform NIR spectra spectrometer, and chemometric analysis was performed using principal component regression (PCR). The tablet hardnesses of F-L and F-C decreased with increasing mixing time. All NIR spectra of the mixed powders of F-L and F-C fluctuated depending on mixing time. In order to predict tablet hardness before tablet compression, NIR spectra of F-L and F-C mixed powders were analyzed and evaluated for hardness by PCR. The minimum standard error of cross-validation values could be realized by using five- and six-principal component models, respectively. In the cases of F-L and F-C, the relationships between the actual and predicted tablet hardnesses showed straight lines, respectively. In the regression vectors of F-L and FC, the peaks related to hydrogen groups of SP, CS, and MC appeared as positive peaks. In contrast, the peaks related to hydrocarbon due to MgSt appeared as negative peaks in the regression vectors. The calibration models to evaluate the tablet hardness were obtained based on NIR spectra of raw mixed powders by PCR. This approach to predicting tablet hardness prior to compression could be used as a routine test to indicate the quality of the final product without spending time and energy to produce samples of questionable quality. ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95:1425–1433, 2006
Keywords: near-infrared spectroscopy; chemometrics; principal component analysis; lubricant effect; tablet hardness; physical characterization; solid state; mixing; tableting; preformulation
INTRODUCTION Process analytical tools allowing for at-or on-line measurement of numerous process variables are Correspondence to: Makoto Otsuka (Telephone and Fax: 81424-68-8658; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 95, 1425–1433 (2006) ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association
now well established in the food product, petroleum chemistry, and other industries. Additionally, on-line, real-time analysis as a tool to monitor and control manufacturing processes is becoming increasingly accepted in the pharmaceutical industry.1–3 On-line analysis holds the promise of reducing or eliminating reworked batches, increasing manufacturing efficiency,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1425
1426
OTSUKA AND YAMANE
decreasing the burden of finished product testing, and ensuring product quality throughout the manufacturing process. Using the near-infrared (NIR) spectroscopic method, spectra can be measured directly on the surface of nondestructed samples without any pretreatment.4 Consequently, NIR spectroscopy is fast becoming an important technique in process analytical methods of the pharmaceuticals industry. Additionally, chemometrics provides an ideal method of extracting quantitative information from samples through NIR spectroscopy5 spectra of multi-component samples. Chemometrical methods such as multiple linear regression (MLR), principal component analysis/principal component regression (PCA/PCR), and partial least squares regression (PLS) are commonly used in all kinds of industry.5 Chemometrical NIR spectroscopic methods have been applied to determine drug contents,6 drug stability,7 polymorphic contents of pharmaceuticals,8–12 and particle size of powders in the pharmaceutical industry.13–16 Mixing is an important process to produce high quality pharmaceuticals since incorrect lubricant mixing might cause incomplete dissolution and/or a decrease in tablet hardness. This is because particles that are coated with magnesium stearate (MgSt) become smooth and less wettable during the process. Since lubricant is a necessary additive for tablet compression, many formulation scientists have studied lubricants and their addition procedure in tabletting.17–21 Recently, Morisseau and Rhodes applied NIR spectroscopy to the evaluation of tablet hardness.22 They introduced chemometrical NIR spectroscopic methods as a nondestructive alternative to conventional tablet hardness testing. Kirsch et al.23 applied several chemometric algorithms to evaluate tablet hardness based on NIR spectroscopic data, and concluded that PLS was the best-fit algorithm. Ebube et al.24 reported the effect of lubricant concentration on tablet hardness by NIR. The fluctuation of avicel tablet hardness was caused by lubricant concentration in the table formulation and could be nondestructively measured by the NIR method. Chen et al.25 reported prediction of drug content and tablet hardness by using an artificial neural network and NIR spectroscopy. They reported that both drug content and tablet hardness had linear calibration models by the combination of artificial neural network and NIR method. These references indicated that the NIR method was able to measure tablet hardness prompt and
nondestructive based on the spectra of the tablets, but not based on raw powder materials. The purpose of this research was to elucidate the effect of lubricant mixing on tablet hardness and to control the tablet hardness, as a basic study of process analytical technology (PAT). A tablet hardness evaluation method is developed based on NIR spectra of raw mixed powders before compression by using chemometrical analysis.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
MATERIALS AND METHODS Materials Bulk powder of sulpyrine (SP) of Japanese Pharmacopoeia (JP) XIV grade was obtained from Ken-Ei Pharm. Co., Osaka, Japan. Microcrystalline cellulose (MC) was supplied by Asahikasei Ind. Co. Ltd. Tokyo, Japan (Avicel, PH102, Lot No. 2012). Spray-dried lactose (SL) (Pharmatose DCL 11, DMV, Veghel, Holland) and potato starch (PS) (Matsuya Chem. Co., Tokyo, Japan) were used as the diluent and disintegrator, respectively. MgSt was provided by NOF Corporation, Tokyo, Japan. All other chemicals were of analytical grade. The formulations for the tablets produced by direct compression are shown in Table 1. Formulations of the Direct Compression Cellulose-Based Formulation (F-C) F-C consisted of 20 g of SP bulk drug powder, 178 g of MC as excipients, and 2 g of MgSt as a lubricant. Bulk drug powder and MC were mixed in the twin-shell mixer (Tokujyu Ind. Co., Model V-1, inside volume 2 L) at 28 rpm for 60 min. MgSt was added, and was mixed in a twin-shell mixer at 28 rpm for 5, 10, 15, 20, 30, 45, or 60 min. Table 1.
Tablet Formulations
Formulation
Content (%)
Content (g)
9.9 63 27 0.1
20 126 54 2
10 89.9 0.1
20 178 2
Formulation L Sulpyrine Spray-dried Lactose Corn starch Magnesium stearate Formulation C Sulpyrine MCC Magnesium stearate
PREDICTION OF TABLET HARDNESS
Lactose/Starch-Based Formulation (F-L) F-L consisted of 20 g of SP bulk drug powder, 126 g of SL, 54 g of corn starch (CS), and 2 g of MgSt. The bulk drug powder, SL and CS were mixed in the twin-shell mixer at 28 rpm for 60 min. MgSt was added, and mixing continued at 28 rpm for 2, 5, 7, 10, 15, 20, 25, or 30 min. Compression Process The mixed powders were compressed by a compression/tension tester (Autograph, model IS5000, Shimadzu Co., Kyoto, Japan) at 25 18C. An 8-mm diameter punch and die with flat surfaces were used to compress 200 mg of a sample at 1000 kg/cm2 (maximum upper punch pressure) and at a speed of 25 mm/min. Sample tablets were ejected from the punch and die at 30 mm/min. Tablet Hardness Test The tablet was compressed at 5 mm/min by a hardness tester (Toyama Co., Osaka, Japan). The tablet hardness was measured four times.
1427
Each NIR spectrum including n spectral data can be seen as a point in an n-dimensional space. In multivariate data analysis, PCA/PCR of a spectral data matrix X is a basic tool. PCA/PCR decomposes X into a score matrix T times a loading matrix P plus a residual matrix E (Eq. (1))5 X ¼ t1 p01 þ t2 p02 þ þ E ¼ TP0 þ E
ð1Þ
This decomposition is particularly useful for converting X to a few information plots (score plots and loading plots) and for modeling the systematic structure in X. Statistical Analysis The Student’s t-test was used to determine the significance of differences level and a p-value of 0.05 was considered significant.
RESULTS AND DISCUSSION Relationship between Tablet Hardness and Mixing Time on Powder Blending
FT-NIR spectra were taken using a NIR spectrometer (InfraProverTM, BRAN þ LUEBBE Co., Norderstedt, Germany). Briefly, a fiber-optic probe was inserted into the sample powder (2 g) in a 20 mL-glass bottle. Five scans per sample part measurement were recorded in the wavelength range of 1000 to 2222 nm. A ceramic (Coor’s Standard) reference scan was taken for each set of samples for calibration purposes. Six FT-NIR spectra were recorded on six different parts of the sample powder with the NIR spectrometer, respectively. Totals of 48 and 42 FT-NIR spectra were recorded for eight and seven calibration samples of F-L and F-C, respectively. For F-L and F-C, 40 and 35 spectral data sets were selected for calibration (calibration set) to predict individual pharmaceutical properties and eight and seven spectra were used for prediction of calibration (prediction set), respectively, and chemometric analysis was performed using the PCR program associated with the Pirouette software (InfoMetrix Co., Woodenville, WA). The best conditions were determined to minimize the standard error of cross-validation (SEV).
Tablet hardness is the result of the hydrogenbonding network between inter-surfaces of excipients particles, such as lactose, starch, and cellulose, and the network is formed by compression force during tabletting. On the other hand, the roles of lubricants in the formulations is to reduce friction between the die and punches, to allow smooth tabletting without troubles such as capping, lamination, and sticking. However, overmixing or excessive dose of lubricant can reduce the mechanical strength of tablets and delay the disintegration time and dissolution rate of tablets,17–21 since excipients particles are coated with a hydrophobic layer consisting of lubricants by mechanical stress. Therefore, it is very important to produce tablets without tablet hardness fluctuations related to lubricants. Figure 1 shows the relationship between tablet hardness and mixing time on powder blending for F-C and F-L. The tablet hardnesses of F-L and F-C decreased with increasing mixing time. The hardness of F-L decreased to 53% (about 1.9 kg) after 30 min mixing, and that of F-C decreased to 35% (about 4.8 kg) after 60 min. Scanning electromicroscopic (SEM) observations of the sample powders (data no shown) indicated that fine micro-particles of MgSt adhered to the surface of excipients and bulk powders after mixing. In a previous investigation, the surfaces of MCC particles were covered
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
Fourier Transform Near Infrared (FT-NIR) Spectroscopy
1428
OTSUKA AND YAMANE
Figure 2. NIR spectra of raw material powders. (a) SL, CS, and MgSt, (b) SP, MC, and MgSt.
Figure 1. Effect of mixing time on tablet hardness of F-L (a) and F-C (b).
with fine MgSt particles after mixing.26 Therefore, the tablet hardness decreasing with increasing mixing time indicated that the coverage of excipients by fine lubricant particles induced a reduction of the particle bonding between excipients such as lactose in F-L and MCC in F-C as reported previously.17–21 NIR Spectra of Sample Powders after Mixing
SL and CS, because the amounts of SL and CS were 63% and 27%. However, the base lines of the spectra for F-L and F-C increased with increasing mixing time, respectively. As described above, fine lubricant particles adhered on the surface of excipient powder particles during the mixing process, indicating that the base line shift reflected the physical condition of particles in the mixed powders. Prediction of Tablet Hardness by NIR Chemometrics
Figure 2 shows spectra of SP, MC, SL, CS, and MgSt. All spectra of raw powders show significant differences from each other. The NIR absorption peaks of SP, MC, SL, CS, and MgSt were as assigned in Table 2. Figure 3a and b shows NIR spectral changes of F-L and F-C during mixing. All NIR spectra of the mixed powders of F-C were similar to that of MC spectrum, since 89% of the total amount was MC in F-C. The NIR spectra of the mixed powders of F-L were similar to those of
Since chemometrics can be used to decompose raw data profiles and helps us to understand the significant contributions of some specific process variables toward explaining the variability in raw data profiles, the quantitative relationships between objective parameters, such as the pharmaceutical properties of granules and loadings of the principle components, in NIR spectra were obtained from the NIR spectral results as reported previously.27 In the present study, in order to predict tablet hardness before tabletting,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
PREDICTION OF TABLET HARDNESS
1429
Table 2. NIR Wavelength of Drug and Excipients Wavelength (nm) SP 2148 1946 1690 1673 MCC 2104 2078 1929 1489 SL 2089 1933 1593 1538 CS 2104 1933 1593 1470 MgSt 1956 1766 1729 1408
Assain
Functional Group
HC –CH st þ C –C st OH st þ OH def CH st 1st overtone CH st 1st overtone
Benzen ring H2 O CH3 Benzen ring
2*OH def þ 2*OH st OH st þ OH def OH st þ OH def OH st 1st overtone
OH OH H2 O OH
2*OH def þ 2*OH st OH st þ OH def OH st 1st overtone OH st 1st overtone
OH H2 O OH OH
2*OH def þ 2*OH st OH st þ OH def OH st 1st overtone OH st 1st overtone
OH H2 O OH OH
C– O st 2nd overtone CH st 1st overtone CH st 1st overtone 2*CH st þ CH def
C– O CH2 CH2 CH2
NIR spectra of F-L and F-C mixed powders were measured, and the tablet hardness was evaluated based on the spectra of raw powder materials by PCR. The calculated results are shown in Table 3. The SEV values of F-L and F-C decreased with an increase in the number of principal component factors, but were almost constant after factors 5 and 6. The SEV decrease of F-C was more significant than that of F-L. The minimum SEV values could thus be realized by using five- and six-principal component models for the analysis in NIR spectra. Figure 4a and b shows the relationships between actual and predicted tablet hardnesses of F-L and F-C by the PCR method. The relationships between the actual and predicted tablet hardness of F-L and F-C show straight lines with slopes of 0.834 and 0.846, Y-intercepts of 0.475 and 1.292, and correlation coefficient constants of 0.9041 and 0.922, respectively; the predicted values were reproducible and had small SDs. As described in the Introduction, many preformulation scientists have applied chemometric algorithms to evaluate tablet hardDOI 10.1002/jps
Figure 3. Mixing effect of lubricant on tablet hardness of F-L (a) and F-C (b).
ness based on NIR spectroscopic data, and the tablet hardness had linear calibration models by the combination of chemometrics and NIR methods. However, there is no report of tablet Table 3. Variance and SEV for Chemometrics Analysis F-L
PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9
F-C
Variance (%)
SEV
Variance (%)
SEV
92.780 6.293 0.597 0.134 0.064 0.024 0.020 0.008 0.007
0.652 0.604 0.448 0.359 0.277 0.281 0.286 0.294 0.253
96.710 3.018 0.088 0.069 0.028 0.016 0.012 0.007 0.005
3.399 3.215 1.671 1.575 1.620 1.289 1.303 1.310 1.271
SEV is standard error of cross-variability. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1430
OTSUKA AND YAMANE
Figure 5. The loadings vectors of the calibration models of F-L (a) and F-C (b) by PCR.
Figure 5 shows the loading vectors corresponding to the principal component (PC). In the case of
F-L, the loading vector of PC1 was a plateau around 0.05, but there were broad negative peaks at 2083 and 1960 nm, attributable to a mixture of excipient powder. The loading vector of PC2 was similar with that of PC1, but there was no baseline shift. In PC3, the positive peaks at 2088 and 1538 nm might be attributable to SL, and the negative peaks at 1956 and 1408 nm might be attributable to MgSt. In the case of F-C, the loading vector of PC1 was a plateau around 0.05, but there were broad negative peaks at 2083 and 1961 nm. The loading vector of PC2 was similar with that of PC1, but there was no baseline shift. In PC3, the positive peaks at 2142, 2115, 2083, and 1947 nm might be attributable to MC, and the negative peaks at 1886 and 1408 nm might be attributable to MgSt of PC3. It is well known that spectral background shift is related to particle size of samples.13–16 We reported that the level of spectral background is related to light scattering of small particles, and applied to the phenomenon to quantitative particle size measurement.28 The sample powder SEM observations (data not shown) indicated that fine micro particles of MgSt adhered on the surface of excipients and bulk powders. Therefore, the above results suggested that the loadings vector of PC1 was attributable to aggregation of fine MgSt
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
Figure 4. The calibration models of F-L (a) and F-C (b) by PCR.
hardness being predicted based on NIR spectra of raw powder material as in this report.22–25 The present results indicated that it was possible to predict tablet hardness from the NIR spectral data of the raw powder materials by nondestructive and noncontact analytical method. Scientific Background of Calibration Models of F-L and F-C to Evaluate Tablet Hardness
PREDICTION OF TABLET HARDNESS
particles on the surface of excipients, meaning that fine particles were formed during the mixing process. On the other hand, the peaks at around 2080 and 1950 nm in both PC2 and PC3 might be related to the OH group of excipients. The relationships of F-L and F-C between the scores of each principal component and tablet hardness are shown in Table 4. The PC-3 and PC-4 for F-L and the PC-2 for F-C showed slightly linear relationships, and their multiple correlation coefficients (R2) were 0.5837, 0.6417, and 0.7171, respectively, but the other PCs did not show linear relationships. This result indicated that individual components could not be directly linked with tablet hardness since tablet hardness involved many complicated factors such as characteristics related to surface, porosity, pore structures, and tablet geometrical factors etc. Since the calibration models to evaluate tablet hardness of F-L and F-C consisted of the score and loadings vector of each component following Eq. (1), the scores of each principal component must have some relationship with tablet hardness. The objective function y^u , tablet hardness could be calculated from original spectral data and regression vector, b.29 y^u ¼ b xu
1431
carbonyl and CH2 groups in MgSt. In the case of FC, the positive peaks at 2083 and 1938 nm, and the negative peaks are at 1876 and 1410 nm. The peaks at 2083 and 1938 nm might be due to the hydroxyl group in MC, and those at 1876 and 1410 nm might be carbonyl and CH2 groups in MgSt. In the formulations of F-L and FC, the peaks related with hydrogen groups at 1527, 1938, and 2142 nm in the regression vectors appeared as positive peaks, meaning that the intensity of the hydrogen groups was proportional to the value of tablet hardness. This result indicated that the degree of tablet hardness reflected the hydrophilic particle bonding between SL, CS, or MC, because the particle bonding was based on intermolecular hydrogen bonding of hydroxyl groups on the surface of excipients particles. Therefore, NIR spectroscopy could measure the tablet hardness based on molecular level change of excipients during the compression process. In general, addition of MgSt as a lubricant in a tablet formulation reduces friction between the punch and die, and reduces sticking, but excess addition of lubricant and overmixing reduce tablet hardness and disintegration time.17–21 In the present study, the peaks related with hydrocarbon of MgSt in the regression vectors appeared as negative peaks, meaning that peak intensity of MgSt reduced with increasing of tablet hardness. These results indicated that the mixing process induced MgSt particles distribution and coating on the surface of excipient particles by mechanochemical stress during the mixing process.20 Therefore, the surface coat layer might inhibit particle bonding between excipients. The coverage of the raw material powder by lubricants also affected the compression behavior, and made a different geometrical tablet structure. However, NIR spectra extracted physical and chemical information on the tablets, and the loading vectors of the calibration model might thus represent
ð2Þ
where b is the regression vector, and xu, and y^u are the test sample data vector and its predicted property, respectively. Therefore, the regression vector indicates important factors for tablet hardness in the NIR spectra of F-L and F-C. Figure 6a and b shows the regression vectors of the calibration models of F-L and F-C, respectively. In the case of F-L, the positive peaks are at 2142, 1938, and 1527 nm and the negative peaks are at 1876 and 1410 nm. The peaks at 2142, 1938, and 1527 might be due to the hydroxyl group in SL or CS, and those at 1876 and 1410 nm might be
Table 4. Relationship between Tablet Hardness and Score of PC F-L
PC1 PC2 PC3 PC4 PC5 PC6
F-C
S
Y
R2
S
Y
R2
0.1004 0.0345 0.0180 0.0054 0.0028 —
0.2787 0.0984 0.0510 0.0153 0.0080 —
0.0870 0.1665 0.5837 0.6417 0.1347 —
0.0147 0.0018 0.0059 0.0003 0.0001 0.0002
0.1264 0.0154 0.0518 0.0021 0.0006 0.0017
0.0380 0.7171 0.2316 0.0403 0.0069 0.0658
S is a slope of the plot, Y is an intercept of the Y-axis, R2 is coefficient of determination. DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1432
OTSUKA AND YAMANE
REFERENCES
This work was supported in part by a Grant-inAid for Scientific Research from the Ministry of Education, Science, and Culture, Japan.
1. Reich G. 2005. Near-infrared spectroscopy and imaging: Basic principles and pharmaceutical applications. Advance Drug Delivery Reviews 57: 1109–1143. 2. Process Analytical Technology (PAT) Initiative, U.S. Food and Drug Administration. Center for Drug Evaluation and Research Home Page ( http:// www.fda.gov/cder/OPS/PAT.htm ) 3. Higgins JP, Arrivo SM, Thurau G, Green RL, Bowen W, Lange A. Templeton AC, Thomas DL, Reed RA. 2003. Spectroscopic approach for on-line monitoring of particle size during the processing of pharmaceutical nanoparticles, Anal Chem 75: 1777–1785. 4. Siesler HW, Ozaki Y, Kawano S, Heise HM. 2001. Near-Infrared Spectroscopy: Principles, Instruments, Applications. Weinheim: Wiley-VCH. 5. Martents H, Næs T. 1989. Multivariate Calibration. New York: John Wiley & Sons. 6. Buchanan BR, Baxter MA, Chen TS, Qin XZ, Robinson PA. 1996. Use of near-infrared spectroscopy to evaluate an active in a film coated tablet. Pharm Res 13:616–621. 7. Drennen JK, Lodder RA. 1990. Nondestructive near-infrared analysis of intact tablets for determination of degradation products. J Pharm Sci 79: 622–627. 8. Norris T, Aldridge PK, Sekulic SS. 1997. Determination of end-points for polymorph conversions of crystalline organic compounds using on-line near-infrared spectroscopy. The Analyst 122:549– 552. 9. Saver RW, Meulman PA, Bowerman DK, Havens JL. 1998. Factor analysis of infrared spectra for solid-state forms of delavirdine mesylate. Int J Pharm 167:105–120. 10. Patl AD, Luner PE, Kemper MS. 2001. Quantitative analysis of polymorphs in binary and multicomponent powder mixtures by near-infrared reflectance spectroscopy. Int J Pharm 206:63–74. 11. Otsuka M, Kato F, Matsuda Y. 2000. comparative evaluation of the degree of indomethacin crystallinity by chemoinfometrical Fourier-transformed near-infrared spectroscopy and conventional powder X-ray diffractometry. AAPS Pharmsci 2: Article 9. 12. Otsuka M, Kato F, Matsuda Y. 2001. Determination of indomethacin polymorphic contents by chemometric near-infrared spectroscopy and conventional powder X-ray diffractometry. Analyst 126:1578–1582. 13. Bull CR. 1991. Compensation for particle size effects in near infrared reflectance. Analyst 116:781–786. 14. Aucott LS, Garthwaite PH, Paul H, Buckland S. 1988. Transformations to reduce the effect of particle size in near-infrared spectra. Analyst 113:1849–1854.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
Figure 6. The regression vectors of the calibration models of F-L (a) and F-C (b) by PCR.
physical, chemical, and intermolecular information about the tablets.
CONCLUSION Calibration models to evaluate tablet hardness were obtained based on NIR spectra of raw mixed powders by using PCR. It was suggested that the tablet hardness could quickly and nondestructively be predicted before compression based on NIR spectra of raw mixed powders. The NIR chemometrical method is expected to provide a rapid quantitative analysis of pharmaceutical properties, as characterized by the simple, nondestructive and highly sensitive nature of the method. It is possible to apply this method to evaluate pharmaceutical properties in the industry for PAT.
ACKNOWLEDGMENTS
PREDICTION OF TABLET HARDNESS
1433
15. Norris KH, Williams PC. 1984. Optimization of mathematical treatments of raw near-infrared signal in the measurement of protein in hard red spring wheat, I. Influence of particle size. Cereal Chem 61:158–165. 16. Franke P, Gill I, Luscombe CN, Rudd DR, Waterhouse J, Jayasooriya A. 1998. Near-infrared mass median particle size determination of lactose monohydrate. evaluating several chemometric approaches. Analyst 123:2043–2046. 17. Chowhan ZT, Chi LH. 1986. Drug-excipient interactions resulting from powder mixing IV: Role of lubricants and their effect on in-vitro dissolution. J Pharm Sci 75:542–545. 18. Chowhan ZT, Chi LH. 1986. Drug-excipient interactions resulting from powder mixing V: Role of sodium lauryl sulfate. Int J Pharm 60:61–78. 19. Gruber P, Glasel VI, Kingelholler W, Liske Th. 1991. Direct lubrication of tablet tools, a contribution to the optimization of tablet manufacture. Drugs made in Germany 34:24–30. 20. Otsuka M, Gao J, Matsuda Y. 1993. Effects of mixer and mixing time on the pharmaceutical properties of theophylline tablets containing various kinds of lactose as diluents. Drug Dev Ind Pharm 19:333– 348. 21. Kikuta J, Kitamori N. 1994. Effect of mixing time on the lubricating properties of magnesium stearate and the final characteristics of the compressed tablet. Drug Dev Ind Pharm 20:343–355. 22. Morisseau KM, Rhodes CT. 1997. Near-infrared spectroscopy as a nondestructive alternative to conventional tablet hardness testing. Pharm Res 14:108–111.
23. Kirsch JD, Drennen JK. 1999. Nondestructive tablet hardness testing by near-infrared spectroscopy: A new and robust spectral best-fit algorism. J Pharm Biomed Analysis 19:351–362. 24. Ebube NK, Thosar SS, Roberts RA, Kemper MS, Rubnovitz R, Martin DL, Reier GE, Wheatley TA, Shukala AJ. 1999. Application of near-infrared spectroscopy for nondestructive analysis of Avicel powders and tablets. Pharm Deve Tech 4: 19–26. 25. Chen Y, Thosar SS, Forbess RA, Kemper MS, Rubinovitz RL, Shukla AJ. 2001. Prediction of drug content and hardness of intact tablets using artificial neural network and near-infrared spectroscopy. Drug Dev Ind Pharm 27:623–631. 26. Otsuka M, Yamane I, Matsuda Y. 2004. Effects of lubricant mixing on compression properties of various kinds of direct compression excipients and physical properties of the tablets. Advanced Powder Tech 15:477–493. 27. Otsuka M, Mouri Y, Matsuda Y. 2003. Chemometric evaluation of pharmaceutical properties of antipyrine granules by near-infrared spectroscopy. AAPS PharmSciTech 4: Article 42, 334–340, (http:// www.pharmascitech.org). 28. Otsuka M. 2004. Comparative particle size determination of phenacetin bulk powder by using Kubelka-Munk theory and principle component regression analysis based on near-infrared spectroscopy. Powder Tech 141:244–250. 29. Prouette manual. Multivariate data analysis for Windows 95/98 and NT, Version 2.6, Infometrix, Inc., Woodinville, WA, USA, (http://www.infometrix. com)
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
Research Institute of Pharmaceutical Sciences, Faculty of Pharmacy, Musashino University, Shinmachi, Nishi-Tokyo 202-8585, Japan 2
Department of Pharmaceutical Technology, Kobe Pharmaceutical University, Motoyama-Kitamachi, Higashi-Nada, Kobe 658-8558, Japan
Received 19 January 2005; revised 4 September 2005; accepted 12 September 2005 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20514
ABSTRACT: The purpose of this research is to elucidate the effect of lubricant mixing on tablet hardness by near-infrared (NIR) chemometrics as a basic study of process analytical technology. Formulation cellulose (F-C) consisted of sulpyrine (SP), microcrystalline cellulose (MC), and magnesium stearate (MgSt). Formulation lactose/starch (F-L) consisted of SP bulk drug powder, spray-dried lactose (SL), corn starch (CS), and MgSt. First, F-L and F-C without MgSt were mixed in a twin-shell mixer for 60 min. MgSt was added to the mixed powder, and was mixed for various mixing times, after which the mixed powders were compressed by 8-mm diameter punch and die. NIR spectra of raw mixed powders of F-L and F-C were taken using a reflection type of Fourier transform NIR spectra spectrometer, and chemometric analysis was performed using principal component regression (PCR). The tablet hardnesses of F-L and F-C decreased with increasing mixing time. All NIR spectra of the mixed powders of F-L and F-C fluctuated depending on mixing time. In order to predict tablet hardness before tablet compression, NIR spectra of F-L and F-C mixed powders were analyzed and evaluated for hardness by PCR. The minimum standard error of cross-validation values could be realized by using five- and six-principal component models, respectively. In the cases of F-L and F-C, the relationships between the actual and predicted tablet hardnesses showed straight lines, respectively. In the regression vectors of F-L and FC, the peaks related to hydrogen groups of SP, CS, and MC appeared as positive peaks. In contrast, the peaks related to hydrocarbon due to MgSt appeared as negative peaks in the regression vectors. The calibration models to evaluate the tablet hardness were obtained based on NIR spectra of raw mixed powders by PCR. This approach to predicting tablet hardness prior to compression could be used as a routine test to indicate the quality of the final product without spending time and energy to produce samples of questionable quality. ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95:1425–1433, 2006
Keywords: near-infrared spectroscopy; chemometrics; principal component analysis; lubricant effect; tablet hardness; physical characterization; solid state; mixing; tableting; preformulation
INTRODUCTION Process analytical tools allowing for at-or on-line measurement of numerous process variables are Correspondence to: Makoto Otsuka (Telephone and Fax: 81424-68-8658; E-mail: [email protected]) Journal of Pharmaceutical Sciences, Vol. 95, 1425–1433 (2006) ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association
now well established in the food product, petroleum chemistry, and other industries. Additionally, on-line, real-time analysis as a tool to monitor and control manufacturing processes is becoming increasingly accepted in the pharmaceutical industry.1–3 On-line analysis holds the promise of reducing or eliminating reworked batches, increasing manufacturing efficiency,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1425
1426
OTSUKA AND YAMANE
decreasing the burden of finished product testing, and ensuring product quality throughout the manufacturing process. Using the near-infrared (NIR) spectroscopic method, spectra can be measured directly on the surface of nondestructed samples without any pretreatment.4 Consequently, NIR spectroscopy is fast becoming an important technique in process analytical methods of the pharmaceuticals industry. Additionally, chemometrics provides an ideal method of extracting quantitative information from samples through NIR spectroscopy5 spectra of multi-component samples. Chemometrical methods such as multiple linear regression (MLR), principal component analysis/principal component regression (PCA/PCR), and partial least squares regression (PLS) are commonly used in all kinds of industry.5 Chemometrical NIR spectroscopic methods have been applied to determine drug contents,6 drug stability,7 polymorphic contents of pharmaceuticals,8–12 and particle size of powders in the pharmaceutical industry.13–16 Mixing is an important process to produce high quality pharmaceuticals since incorrect lubricant mixing might cause incomplete dissolution and/or a decrease in tablet hardness. This is because particles that are coated with magnesium stearate (MgSt) become smooth and less wettable during the process. Since lubricant is a necessary additive for tablet compression, many formulation scientists have studied lubricants and their addition procedure in tabletting.17–21 Recently, Morisseau and Rhodes applied NIR spectroscopy to the evaluation of tablet hardness.22 They introduced chemometrical NIR spectroscopic methods as a nondestructive alternative to conventional tablet hardness testing. Kirsch et al.23 applied several chemometric algorithms to evaluate tablet hardness based on NIR spectroscopic data, and concluded that PLS was the best-fit algorithm. Ebube et al.24 reported the effect of lubricant concentration on tablet hardness by NIR. The fluctuation of avicel tablet hardness was caused by lubricant concentration in the table formulation and could be nondestructively measured by the NIR method. Chen et al.25 reported prediction of drug content and tablet hardness by using an artificial neural network and NIR spectroscopy. They reported that both drug content and tablet hardness had linear calibration models by the combination of artificial neural network and NIR method. These references indicated that the NIR method was able to measure tablet hardness prompt and
nondestructive based on the spectra of the tablets, but not based on raw powder materials. The purpose of this research was to elucidate the effect of lubricant mixing on tablet hardness and to control the tablet hardness, as a basic study of process analytical technology (PAT). A tablet hardness evaluation method is developed based on NIR spectra of raw mixed powders before compression by using chemometrical analysis.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
MATERIALS AND METHODS Materials Bulk powder of sulpyrine (SP) of Japanese Pharmacopoeia (JP) XIV grade was obtained from Ken-Ei Pharm. Co., Osaka, Japan. Microcrystalline cellulose (MC) was supplied by Asahikasei Ind. Co. Ltd. Tokyo, Japan (Avicel, PH102, Lot No. 2012). Spray-dried lactose (SL) (Pharmatose DCL 11, DMV, Veghel, Holland) and potato starch (PS) (Matsuya Chem. Co., Tokyo, Japan) were used as the diluent and disintegrator, respectively. MgSt was provided by NOF Corporation, Tokyo, Japan. All other chemicals were of analytical grade. The formulations for the tablets produced by direct compression are shown in Table 1. Formulations of the Direct Compression Cellulose-Based Formulation (F-C) F-C consisted of 20 g of SP bulk drug powder, 178 g of MC as excipients, and 2 g of MgSt as a lubricant. Bulk drug powder and MC were mixed in the twin-shell mixer (Tokujyu Ind. Co., Model V-1, inside volume 2 L) at 28 rpm for 60 min. MgSt was added, and was mixed in a twin-shell mixer at 28 rpm for 5, 10, 15, 20, 30, 45, or 60 min. Table 1.
Tablet Formulations
Formulation
Content (%)
Content (g)
9.9 63 27 0.1
20 126 54 2
10 89.9 0.1
20 178 2
Formulation L Sulpyrine Spray-dried Lactose Corn starch Magnesium stearate Formulation C Sulpyrine MCC Magnesium stearate
PREDICTION OF TABLET HARDNESS
Lactose/Starch-Based Formulation (F-L) F-L consisted of 20 g of SP bulk drug powder, 126 g of SL, 54 g of corn starch (CS), and 2 g of MgSt. The bulk drug powder, SL and CS were mixed in the twin-shell mixer at 28 rpm for 60 min. MgSt was added, and mixing continued at 28 rpm for 2, 5, 7, 10, 15, 20, 25, or 30 min. Compression Process The mixed powders were compressed by a compression/tension tester (Autograph, model IS5000, Shimadzu Co., Kyoto, Japan) at 25 18C. An 8-mm diameter punch and die with flat surfaces were used to compress 200 mg of a sample at 1000 kg/cm2 (maximum upper punch pressure) and at a speed of 25 mm/min. Sample tablets were ejected from the punch and die at 30 mm/min. Tablet Hardness Test The tablet was compressed at 5 mm/min by a hardness tester (Toyama Co., Osaka, Japan). The tablet hardness was measured four times.
1427
Each NIR spectrum including n spectral data can be seen as a point in an n-dimensional space. In multivariate data analysis, PCA/PCR of a spectral data matrix X is a basic tool. PCA/PCR decomposes X into a score matrix T times a loading matrix P plus a residual matrix E (Eq. (1))5 X ¼ t1 p01 þ t2 p02 þ þ E ¼ TP0 þ E
ð1Þ
This decomposition is particularly useful for converting X to a few information plots (score plots and loading plots) and for modeling the systematic structure in X. Statistical Analysis The Student’s t-test was used to determine the significance of differences level and a p-value of 0.05 was considered significant.
RESULTS AND DISCUSSION Relationship between Tablet Hardness and Mixing Time on Powder Blending
FT-NIR spectra were taken using a NIR spectrometer (InfraProverTM, BRAN þ LUEBBE Co., Norderstedt, Germany). Briefly, a fiber-optic probe was inserted into the sample powder (2 g) in a 20 mL-glass bottle. Five scans per sample part measurement were recorded in the wavelength range of 1000 to 2222 nm. A ceramic (Coor’s Standard) reference scan was taken for each set of samples for calibration purposes. Six FT-NIR spectra were recorded on six different parts of the sample powder with the NIR spectrometer, respectively. Totals of 48 and 42 FT-NIR spectra were recorded for eight and seven calibration samples of F-L and F-C, respectively. For F-L and F-C, 40 and 35 spectral data sets were selected for calibration (calibration set) to predict individual pharmaceutical properties and eight and seven spectra were used for prediction of calibration (prediction set), respectively, and chemometric analysis was performed using the PCR program associated with the Pirouette software (InfoMetrix Co., Woodenville, WA). The best conditions were determined to minimize the standard error of cross-validation (SEV).
Tablet hardness is the result of the hydrogenbonding network between inter-surfaces of excipients particles, such as lactose, starch, and cellulose, and the network is formed by compression force during tabletting. On the other hand, the roles of lubricants in the formulations is to reduce friction between the die and punches, to allow smooth tabletting without troubles such as capping, lamination, and sticking. However, overmixing or excessive dose of lubricant can reduce the mechanical strength of tablets and delay the disintegration time and dissolution rate of tablets,17–21 since excipients particles are coated with a hydrophobic layer consisting of lubricants by mechanical stress. Therefore, it is very important to produce tablets without tablet hardness fluctuations related to lubricants. Figure 1 shows the relationship between tablet hardness and mixing time on powder blending for F-C and F-L. The tablet hardnesses of F-L and F-C decreased with increasing mixing time. The hardness of F-L decreased to 53% (about 1.9 kg) after 30 min mixing, and that of F-C decreased to 35% (about 4.8 kg) after 60 min. Scanning electromicroscopic (SEM) observations of the sample powders (data no shown) indicated that fine micro-particles of MgSt adhered to the surface of excipients and bulk powders after mixing. In a previous investigation, the surfaces of MCC particles were covered
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
Fourier Transform Near Infrared (FT-NIR) Spectroscopy
1428
OTSUKA AND YAMANE
Figure 2. NIR spectra of raw material powders. (a) SL, CS, and MgSt, (b) SP, MC, and MgSt.
Figure 1. Effect of mixing time on tablet hardness of F-L (a) and F-C (b).
with fine MgSt particles after mixing.26 Therefore, the tablet hardness decreasing with increasing mixing time indicated that the coverage of excipients by fine lubricant particles induced a reduction of the particle bonding between excipients such as lactose in F-L and MCC in F-C as reported previously.17–21 NIR Spectra of Sample Powders after Mixing
SL and CS, because the amounts of SL and CS were 63% and 27%. However, the base lines of the spectra for F-L and F-C increased with increasing mixing time, respectively. As described above, fine lubricant particles adhered on the surface of excipient powder particles during the mixing process, indicating that the base line shift reflected the physical condition of particles in the mixed powders. Prediction of Tablet Hardness by NIR Chemometrics
Figure 2 shows spectra of SP, MC, SL, CS, and MgSt. All spectra of raw powders show significant differences from each other. The NIR absorption peaks of SP, MC, SL, CS, and MgSt were as assigned in Table 2. Figure 3a and b shows NIR spectral changes of F-L and F-C during mixing. All NIR spectra of the mixed powders of F-C were similar to that of MC spectrum, since 89% of the total amount was MC in F-C. The NIR spectra of the mixed powders of F-L were similar to those of
Since chemometrics can be used to decompose raw data profiles and helps us to understand the significant contributions of some specific process variables toward explaining the variability in raw data profiles, the quantitative relationships between objective parameters, such as the pharmaceutical properties of granules and loadings of the principle components, in NIR spectra were obtained from the NIR spectral results as reported previously.27 In the present study, in order to predict tablet hardness before tabletting,
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
PREDICTION OF TABLET HARDNESS
1429
Table 2. NIR Wavelength of Drug and Excipients Wavelength (nm) SP 2148 1946 1690 1673 MCC 2104 2078 1929 1489 SL 2089 1933 1593 1538 CS 2104 1933 1593 1470 MgSt 1956 1766 1729 1408
Assain
Functional Group
HC –CH st þ C –C st OH st þ OH def CH st 1st overtone CH st 1st overtone
Benzen ring H2 O CH3 Benzen ring
2*OH def þ 2*OH st OH st þ OH def OH st þ OH def OH st 1st overtone
OH OH H2 O OH
2*OH def þ 2*OH st OH st þ OH def OH st 1st overtone OH st 1st overtone
OH H2 O OH OH
2*OH def þ 2*OH st OH st þ OH def OH st 1st overtone OH st 1st overtone
OH H2 O OH OH
C– O st 2nd overtone CH st 1st overtone CH st 1st overtone 2*CH st þ CH def
C– O CH2 CH2 CH2
NIR spectra of F-L and F-C mixed powders were measured, and the tablet hardness was evaluated based on the spectra of raw powder materials by PCR. The calculated results are shown in Table 3. The SEV values of F-L and F-C decreased with an increase in the number of principal component factors, but were almost constant after factors 5 and 6. The SEV decrease of F-C was more significant than that of F-L. The minimum SEV values could thus be realized by using five- and six-principal component models for the analysis in NIR spectra. Figure 4a and b shows the relationships between actual and predicted tablet hardnesses of F-L and F-C by the PCR method. The relationships between the actual and predicted tablet hardness of F-L and F-C show straight lines with slopes of 0.834 and 0.846, Y-intercepts of 0.475 and 1.292, and correlation coefficient constants of 0.9041 and 0.922, respectively; the predicted values were reproducible and had small SDs. As described in the Introduction, many preformulation scientists have applied chemometric algorithms to evaluate tablet hardDOI 10.1002/jps
Figure 3. Mixing effect of lubricant on tablet hardness of F-L (a) and F-C (b).
ness based on NIR spectroscopic data, and the tablet hardness had linear calibration models by the combination of chemometrics and NIR methods. However, there is no report of tablet Table 3. Variance and SEV for Chemometrics Analysis F-L
PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9
F-C
Variance (%)
SEV
Variance (%)
SEV
92.780 6.293 0.597 0.134 0.064 0.024 0.020 0.008 0.007
0.652 0.604 0.448 0.359 0.277 0.281 0.286 0.294 0.253
96.710 3.018 0.088 0.069 0.028 0.016 0.012 0.007 0.005
3.399 3.215 1.671 1.575 1.620 1.289 1.303 1.310 1.271
SEV is standard error of cross-variability. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1430
OTSUKA AND YAMANE
Figure 5. The loadings vectors of the calibration models of F-L (a) and F-C (b) by PCR.
Figure 5 shows the loading vectors corresponding to the principal component (PC). In the case of
F-L, the loading vector of PC1 was a plateau around 0.05, but there were broad negative peaks at 2083 and 1960 nm, attributable to a mixture of excipient powder. The loading vector of PC2 was similar with that of PC1, but there was no baseline shift. In PC3, the positive peaks at 2088 and 1538 nm might be attributable to SL, and the negative peaks at 1956 and 1408 nm might be attributable to MgSt. In the case of F-C, the loading vector of PC1 was a plateau around 0.05, but there were broad negative peaks at 2083 and 1961 nm. The loading vector of PC2 was similar with that of PC1, but there was no baseline shift. In PC3, the positive peaks at 2142, 2115, 2083, and 1947 nm might be attributable to MC, and the negative peaks at 1886 and 1408 nm might be attributable to MgSt of PC3. It is well known that spectral background shift is related to particle size of samples.13–16 We reported that the level of spectral background is related to light scattering of small particles, and applied to the phenomenon to quantitative particle size measurement.28 The sample powder SEM observations (data not shown) indicated that fine micro particles of MgSt adhered on the surface of excipients and bulk powders. Therefore, the above results suggested that the loadings vector of PC1 was attributable to aggregation of fine MgSt
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
Figure 4. The calibration models of F-L (a) and F-C (b) by PCR.
hardness being predicted based on NIR spectra of raw powder material as in this report.22–25 The present results indicated that it was possible to predict tablet hardness from the NIR spectral data of the raw powder materials by nondestructive and noncontact analytical method. Scientific Background of Calibration Models of F-L and F-C to Evaluate Tablet Hardness
PREDICTION OF TABLET HARDNESS
particles on the surface of excipients, meaning that fine particles were formed during the mixing process. On the other hand, the peaks at around 2080 and 1950 nm in both PC2 and PC3 might be related to the OH group of excipients. The relationships of F-L and F-C between the scores of each principal component and tablet hardness are shown in Table 4. The PC-3 and PC-4 for F-L and the PC-2 for F-C showed slightly linear relationships, and their multiple correlation coefficients (R2) were 0.5837, 0.6417, and 0.7171, respectively, but the other PCs did not show linear relationships. This result indicated that individual components could not be directly linked with tablet hardness since tablet hardness involved many complicated factors such as characteristics related to surface, porosity, pore structures, and tablet geometrical factors etc. Since the calibration models to evaluate tablet hardness of F-L and F-C consisted of the score and loadings vector of each component following Eq. (1), the scores of each principal component must have some relationship with tablet hardness. The objective function y^u , tablet hardness could be calculated from original spectral data and regression vector, b.29 y^u ¼ b xu
1431
carbonyl and CH2 groups in MgSt. In the case of FC, the positive peaks at 2083 and 1938 nm, and the negative peaks are at 1876 and 1410 nm. The peaks at 2083 and 1938 nm might be due to the hydroxyl group in MC, and those at 1876 and 1410 nm might be carbonyl and CH2 groups in MgSt. In the formulations of F-L and FC, the peaks related with hydrogen groups at 1527, 1938, and 2142 nm in the regression vectors appeared as positive peaks, meaning that the intensity of the hydrogen groups was proportional to the value of tablet hardness. This result indicated that the degree of tablet hardness reflected the hydrophilic particle bonding between SL, CS, or MC, because the particle bonding was based on intermolecular hydrogen bonding of hydroxyl groups on the surface of excipients particles. Therefore, NIR spectroscopy could measure the tablet hardness based on molecular level change of excipients during the compression process. In general, addition of MgSt as a lubricant in a tablet formulation reduces friction between the punch and die, and reduces sticking, but excess addition of lubricant and overmixing reduce tablet hardness and disintegration time.17–21 In the present study, the peaks related with hydrocarbon of MgSt in the regression vectors appeared as negative peaks, meaning that peak intensity of MgSt reduced with increasing of tablet hardness. These results indicated that the mixing process induced MgSt particles distribution and coating on the surface of excipient particles by mechanochemical stress during the mixing process.20 Therefore, the surface coat layer might inhibit particle bonding between excipients. The coverage of the raw material powder by lubricants also affected the compression behavior, and made a different geometrical tablet structure. However, NIR spectra extracted physical and chemical information on the tablets, and the loading vectors of the calibration model might thus represent
ð2Þ
where b is the regression vector, and xu, and y^u are the test sample data vector and its predicted property, respectively. Therefore, the regression vector indicates important factors for tablet hardness in the NIR spectra of F-L and F-C. Figure 6a and b shows the regression vectors of the calibration models of F-L and F-C, respectively. In the case of F-L, the positive peaks are at 2142, 1938, and 1527 nm and the negative peaks are at 1876 and 1410 nm. The peaks at 2142, 1938, and 1527 might be due to the hydroxyl group in SL or CS, and those at 1876 and 1410 nm might be
Table 4. Relationship between Tablet Hardness and Score of PC F-L
PC1 PC2 PC3 PC4 PC5 PC6
F-C
S
Y
R2
S
Y
R2
0.1004 0.0345 0.0180 0.0054 0.0028 —
0.2787 0.0984 0.0510 0.0153 0.0080 —
0.0870 0.1665 0.5837 0.6417 0.1347 —
0.0147 0.0018 0.0059 0.0003 0.0001 0.0002
0.1264 0.0154 0.0518 0.0021 0.0006 0.0017
0.0380 0.7171 0.2316 0.0403 0.0069 0.0658
S is a slope of the plot, Y is an intercept of the Y-axis, R2 is coefficient of determination. DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
1432
OTSUKA AND YAMANE
REFERENCES
This work was supported in part by a Grant-inAid for Scientific Research from the Ministry of Education, Science, and Culture, Japan.
1. Reich G. 2005. Near-infrared spectroscopy and imaging: Basic principles and pharmaceutical applications. Advance Drug Delivery Reviews 57: 1109–1143. 2. Process Analytical Technology (PAT) Initiative, U.S. Food and Drug Administration. Center for Drug Evaluation and Research Home Page ( http:// www.fda.gov/cder/OPS/PAT.htm ) 3. Higgins JP, Arrivo SM, Thurau G, Green RL, Bowen W, Lange A. Templeton AC, Thomas DL, Reed RA. 2003. Spectroscopic approach for on-line monitoring of particle size during the processing of pharmaceutical nanoparticles, Anal Chem 75: 1777–1785. 4. Siesler HW, Ozaki Y, Kawano S, Heise HM. 2001. Near-Infrared Spectroscopy: Principles, Instruments, Applications. Weinheim: Wiley-VCH. 5. Martents H, Næs T. 1989. Multivariate Calibration. New York: John Wiley & Sons. 6. Buchanan BR, Baxter MA, Chen TS, Qin XZ, Robinson PA. 1996. Use of near-infrared spectroscopy to evaluate an active in a film coated tablet. Pharm Res 13:616–621. 7. Drennen JK, Lodder RA. 1990. Nondestructive near-infrared analysis of intact tablets for determination of degradation products. J Pharm Sci 79: 622–627. 8. Norris T, Aldridge PK, Sekulic SS. 1997. Determination of end-points for polymorph conversions of crystalline organic compounds using on-line near-infrared spectroscopy. The Analyst 122:549– 552. 9. Saver RW, Meulman PA, Bowerman DK, Havens JL. 1998. Factor analysis of infrared spectra for solid-state forms of delavirdine mesylate. Int J Pharm 167:105–120. 10. Patl AD, Luner PE, Kemper MS. 2001. Quantitative analysis of polymorphs in binary and multicomponent powder mixtures by near-infrared reflectance spectroscopy. Int J Pharm 206:63–74. 11. Otsuka M, Kato F, Matsuda Y. 2000. comparative evaluation of the degree of indomethacin crystallinity by chemoinfometrical Fourier-transformed near-infrared spectroscopy and conventional powder X-ray diffractometry. AAPS Pharmsci 2: Article 9. 12. Otsuka M, Kato F, Matsuda Y. 2001. Determination of indomethacin polymorphic contents by chemometric near-infrared spectroscopy and conventional powder X-ray diffractometry. Analyst 126:1578–1582. 13. Bull CR. 1991. Compensation for particle size effects in near infrared reflectance. Analyst 116:781–786. 14. Aucott LS, Garthwaite PH, Paul H, Buckland S. 1988. Transformations to reduce the effect of particle size in near-infrared spectra. Analyst 113:1849–1854.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006
DOI 10.1002/jps
Figure 6. The regression vectors of the calibration models of F-L (a) and F-C (b) by PCR.
physical, chemical, and intermolecular information about the tablets.
CONCLUSION Calibration models to evaluate tablet hardness were obtained based on NIR spectra of raw mixed powders by using PCR. It was suggested that the tablet hardness could quickly and nondestructively be predicted before compression based on NIR spectra of raw mixed powders. The NIR chemometrical method is expected to provide a rapid quantitative analysis of pharmaceutical properties, as characterized by the simple, nondestructive and highly sensitive nature of the method. It is possible to apply this method to evaluate pharmaceutical properties in the industry for PAT.
ACKNOWLEDGMENTS
PREDICTION OF TABLET HARDNESS
1433
15. Norris KH, Williams PC. 1984. Optimization of mathematical treatments of raw near-infrared signal in the measurement of protein in hard red spring wheat, I. Influence of particle size. Cereal Chem 61:158–165. 16. Franke P, Gill I, Luscombe CN, Rudd DR, Waterhouse J, Jayasooriya A. 1998. Near-infrared mass median particle size determination of lactose monohydrate. evaluating several chemometric approaches. Analyst 123:2043–2046. 17. Chowhan ZT, Chi LH. 1986. Drug-excipient interactions resulting from powder mixing IV: Role of lubricants and their effect on in-vitro dissolution. J Pharm Sci 75:542–545. 18. Chowhan ZT, Chi LH. 1986. Drug-excipient interactions resulting from powder mixing V: Role of sodium lauryl sulfate. Int J Pharm 60:61–78. 19. Gruber P, Glasel VI, Kingelholler W, Liske Th. 1991. Direct lubrication of tablet tools, a contribution to the optimization of tablet manufacture. Drugs made in Germany 34:24–30. 20. Otsuka M, Gao J, Matsuda Y. 1993. Effects of mixer and mixing time on the pharmaceutical properties of theophylline tablets containing various kinds of lactose as diluents. Drug Dev Ind Pharm 19:333– 348. 21. Kikuta J, Kitamori N. 1994. Effect of mixing time on the lubricating properties of magnesium stearate and the final characteristics of the compressed tablet. Drug Dev Ind Pharm 20:343–355. 22. Morisseau KM, Rhodes CT. 1997. Near-infrared spectroscopy as a nondestructive alternative to conventional tablet hardness testing. Pharm Res 14:108–111.
23. Kirsch JD, Drennen JK. 1999. Nondestructive tablet hardness testing by near-infrared spectroscopy: A new and robust spectral best-fit algorism. J Pharm Biomed Analysis 19:351–362. 24. Ebube NK, Thosar SS, Roberts RA, Kemper MS, Rubnovitz R, Martin DL, Reier GE, Wheatley TA, Shukala AJ. 1999. Application of near-infrared spectroscopy for nondestructive analysis of Avicel powders and tablets. Pharm Deve Tech 4: 19–26. 25. Chen Y, Thosar SS, Forbess RA, Kemper MS, Rubinovitz RL, Shukla AJ. 2001. Prediction of drug content and hardness of intact tablets using artificial neural network and near-infrared spectroscopy. Drug Dev Ind Pharm 27:623–631. 26. Otsuka M, Yamane I, Matsuda Y. 2004. Effects of lubricant mixing on compression properties of various kinds of direct compression excipients and physical properties of the tablets. Advanced Powder Tech 15:477–493. 27. Otsuka M, Mouri Y, Matsuda Y. 2003. Chemometric evaluation of pharmaceutical properties of antipyrine granules by near-infrared spectroscopy. AAPS PharmSciTech 4: Article 42, 334–340, (http:// www.pharmascitech.org). 28. Otsuka M. 2004. Comparative particle size determination of phenacetin bulk powder by using Kubelka-Munk theory and principle component regression analysis based on near-infrared spectroscopy. Powder Tech 141:244–250. 29. Prouette manual. Multivariate data analysis for Windows 95/98 and NT, Version 2.6, Infometrix, Inc., Woodinville, WA, USA, (http://www.infometrix. com)
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 7, JULY 2006