- Email: [email protected]
High-throughput prediction of tablet weight and trimethoprim content of compound sulfamethoxazole tablets for controlling the uniformity of dosage units by NIR
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
High-throughput prediction of tablet weight and trimethoprim content of compound sulfamethoxazole tablets for controlling the uniformity of dosage units by NIR Yanhong Dong, Juan Li, Xiaoxiao Zhong, Liya Cao, Yang Luo, Qi Fan ⁎ School of Pharmacy, Chongqing Medical University, Chongqing 400016, China
a r t i c l e
i n f o
Article history: Received 15 November 2015 Received in revised form 15 January 2016 Accepted 20 January 2016 Available online 22 January 2016 Keywords: Uniformity of dosage units Compound sulfamethoxazole tablets Tablet weight Trimethoprim content Near infrared spectroscopy Partial least squares regression
a b s t r a c t This paper establishes a novel method to simultaneously predict the tablet weight (TW) and trimethoprim (TMP) content of compound sulfamethoxazole tablets (SMZCO) by near infrared (NIR) spectroscopy with partial least squares (PLS) regression for controlling the uniformity of dosage units (UODU). The NIR spectra for 257 samples were measured using the optimized parameter values and pretreated using the optimized chemometric techniques. After the outliers were ignored, two PLS models for predicting TW and TMP content were respectively established by using the selected spectral sub-ranges and the reference values. The TW model reaches the correlation coefficient of calibration (Rc) 0.9543 and the TMP content model has the Rc 0.9205. The experimental results indicate that this strategy expands the NIR application in controlling UODU, especially in the highthroughput and rapid analysis of TWs and contents of the compound pharmaceutical tablets, and may be an important complement to the common NIR on-line analytical method for pharmaceutical tablets. © 2016 Elsevier B.V. All rights reserved.
1. Introduction The tablet is the most common dosage form as the finished pharmaceutical products (FPPs). In the quality control (QC) of orally administered pharmaceutical tablets, the uniformity of dosage units (UODU) is a vitally important test item stated in United States Pharmacopeia (USP), European Pharmacopoeia (EP), Japanese Pharmacopoeia (JP) and Chinese Pharmacopoeia (CP) [1–4] because it is related to the safety and effectivity of dosage units. The compound sulfamethoxazole tablets (SMZCO), containing 400.0 mg of sulfamethoxazole (SMZ) and 80.0 mg of trimethoprim (TMP) in each tablet, are prescribed for the treatment of infections, such as urinary tract infection, pneumocystis carinii pneumonia and Stenotrophomonas maltophilia osteomyelitis [5,6]. The UODU of SMZCO is demonstrated by weight variation (WV) for SMZ (its ratio is more than 25% by weight) and by content uniformity (CU) for TMP (its ratio is less than 25% by weight) in USP and CP [1,4]. The WV and CU are calculated separately on the basis of tablet weight (TW) and TMP content. The classical measurement method for TW is weighing method [1–4], and one for TMP content is high performance liquid chromatography (HPLC) [5,7]. However, weighing method is low-throughput and HPLC is sample-destructive, environment polluting and time consuming.
⁎ Corresponding author. E-mail address: [email protected] (Q. Fan).
http://dx.doi.org/10.1016/j.saa.2016.01.030 1386-1425/© 2016 Elsevier B.V. All rights reserved.
Near infrared (NIR) spectroscopy with chemometric techniques can extract simultaneously multiple chemical and physical information of an analyte, which contains X–H (X = C, O, N). Because it is highthroughput, nondestructive, pollution-free, rapid, easy-to-use and available for on-line process control, NIR spectroscopy has been extensively used for the quantitative analysis of multiple chemical and physical characteristics of pharmaceuticals, for example, content [8–10], impurity [11], coating thickness [12], hardness [13], strength [11], disintegration time [14], and dissolution [15]. In this work, the feasibility of high-throughput and rapid prediction of TW and TMP content of SMZCO for controlling UODU is evaluated by NIR spectroscopy with partial least squares (PLS) regression principally based on the reasons below. Firstly, the relationship between TW and SMZ content can be described by Eq. (1) [1]. TW ¼ CSMZ TW=A
ð1Þ
where TW is the individual weight of the tablet tested, CSMZ is the individual estimated SMZ content of the tablet tested, TW is the mean of TWs, and A is the SMZ assay. Secondly, the molecular structures of both SMZ and TMP include many C–H and N–H groups, as seen in Fig. 1. In a word, both TW and TMP content could be predicted by NIR spectroscopy. This work may expand the NIR application in controlling UODU, especially in the high-throughput prediction of TWs and contents of the compound pharmaceutical tablets. And the established
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
79
Fig. 1. Chemical structural formulas of SMZ (a) and TMP (b).
high-throughput and rapid strategy may be an important complement to the common NIR on-line analytical method for pharmaceutical tablets. 2. Materials and methods 2.1. Instruments The NIR spectra were measured by FT-NIR analyzer (Antaris II, Thermo Fisher Scientific, USA), which was furnished with integrating sphere and sample holder and was controlled by the software package RESULT 3.0. The chemometric calculations were executed using the software TQ Analyst 8.0. The reference values of TW and TMP content were determined respectively by analytical balance (AB204-E, Mettler Toledo, Switzerland) and HPLC system (Agilent 1100, Agilent, USA). 2.2. Reagents and samples The reference material of TMP (batch number: 100031–200304) was purchased from the Chinese National Institutes for Food and Drug Control (Beijing, China). The active pharmaceutical ingredients (APIs) of SMZ (batch number: SMXB01312; assay: ≥99.0%) and TMP (batch number: A-10111204022; assay: ≥99.0%) were obtained respectively from Suzhou Sanray Pharmaceutical Co., Ltd. (Suzhou, China) and Shouguang Fukang Pharmaceutical Co., Ltd. (Shouguang, China). Acetonitrile was HPLC grade, other chemicals were analytical grade, and experimental water was purified water. In this experiment, 257 samples of SMZCO (400.0 mg of SMZ and 80.0 mg of TMP in each tablet) from 13 batches were produced by four Chinese manufacturers. The appearances of all samples are consistent: white cylindrical flat-faced tablets coded with “SMZ” and “CO” on the scored side. 2.3. Reference methods 2.3.1. Determination of TW The TW reference value for each tablet was determined by using an analytical balance [1–4]. 2.3.2. Determination of TMP content The reference value of TMP content in each tablet was determined by HPLC and calculated using the external standard method. The separation was performed on an Agilent ZORBAX SB-C18 column (150 mm × 4.6 mm, 5 μm), whose temperature was maintained at 30 °C, by using acetonitrile–water–triethylamine (200:799:1, v/v/v; pH 5.9) as the mobile phase at a flow rate of 1.0 mL/min. The TMP content was detected at 240 nm. And the injection volume was 10 μL [5]. 2.4. Measurement of NIR spectra Each tablet was put on the circle detection window of integrating sphere module of FT-NIR analyzer and the scored side of tablet was
upward. The sample holder was adjusted to ensure the position of each tablet concentric with the detection window. A Fourier transform near infrared diffuse reflectance spectrum (NIRDRS) for each tablet was acquired in the range of 10,000–4000 cm−1 with same resolution and same number of scans, which were selected from the resolutions 4, 8 and 16 cm−1 and the numbers of scans 32 and 64 on the basis of small spectral variation and short measuring time. The ambient temperature was about 25 °C. The background spectrum was determined under the same conditions to eliminate the interferences from H2O and CO2 in air. The API of SMZ and the API of TMP were respectively poured into two same empty circular glass vials, which diameters are about 1.0 cm, up to about 1.0 cm of thickness [16]. And their NIRDRSs were also measured in the above way.
2.5. Prediction of TW 2.5.1. Pretreatment of spectral information The spectral information was pretreated by the chemometric techniques selected from no preprocessing, multiplicative signal correction (MSC) or standard normal variate (SNV) for eliminating the interferences from the particle size and the compactness, first derivative (FD) or second derivative (SD) for deducting the background and separating the overlapping signals, Savitzky–Golay smoothing (SGS) or Norris derivative smoothing (NDS) for removing the noise, and mean centering (MC) and variance scaling (VS) for improving the performances of PLS model for predicting TW (PLSTW). 2.5.2. Selection of spectral sub-range The spectral sub-range, used to establish the PLSTW, was selected primarily on the basis of the characteristic absorptions of C–H and N–H groups of SMZ to reduce data redundancy for enhancing the PLSTW performances and increasing the computational speed.
2.5.3. Establishment and validation of PLSTW The NIRDRSs for 257 samples of SMZCO from 13 batches and four manufacturers were divided into the calibration and prediction sets in the ratio of 2:1. The reference values of 86 prediction samples from 13 batches and four manufacturers were approximately uniformly distributed over the range of the reference values of 171 calibration samples from 13 batches and four manufacturers. After the outliers were diagnosed by TQ Analyst 8.0 and ignored, the PLSTW was built using the TW reference values, the NIRDRSs for calibration samples and the optimal number of factors to reach the lowest root mean square error of cross-validation (RMSECV). And the PLSTW was validated with the NIRDRSs and the TW reference values for prediction samples. The performances of PLSTW were assessed with the root mean square errors of calibration (RMSEC), RMSECV and prediction (RMSEP), and with the correlation coefficients of calibration (Rc), cross-validation (Rcv) and prediction (Rp).
80
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
Fig. 2. Raw NIRDRSs were obtained with the resolution 8 cm−1 and the number of scans 64. (a) shows 257 NIRDRSs for SMZCO and (b) gives a typical NIRDRS for SMZCO (black) and two NIRDRSs for the APIs of SMZ (red) and TMP (blue). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
NIRDRS for SMZCO (black) and two NIRDRSs for the APIs of SMZ (red) and TMP (blue) in Fig. 2 (b).
2.6. Prediction of TMP content The NIRDRSs used in the prediction of TW were simultaneously used to predict the TMP content. And the PLS model for predicting TMP content (PLSTMP) was built and validated using the reference values of TMP contents, similarly to the PLSTW.
3.3. Prediction of TW 3.3.1. Pretreatment of spectral information The main pretreatment methods for spectral information and the corresponding performances of PLSTW are shown in Table 1. The combination of MSC, FD, SGS and MC was selected to pretreat the spectral information and then the high overall performance of PLSTW was reached.
3. Results and discussion 3.1. Reference values The TW reference values of 257 samples of SMZCO from 13 batches and four manufacturers were 0.4999–0.5897 g, equivalently to 90.6%– 106.9% of the label claim (400.0 mg of SMZ in each tablet). The reference values of TMP contents in 257 samples were 0.1301– 0.1658 g/g, equivalently to 94.6%–111.8% of the label claim (80.0 mg of TMP in each tablet).
3.3.2. Selection of spectral sub-range The red spectrum in Fig. 2 (b) represents that the characteristic absorptions of C–H and N–H groups of SMZ are mainly in the range of 7500–5900 cm−1, including the combinations of C–H, 7300 cm−1, 5985 cm−1 and 5920 cm−1, and the overtone of N–H, near 6800 cm−1 [16]. And Table 1 gives that the PLSTW performances are higher when the spectral sub-range 7500–5900 cm−1 is used than the wider spectral sub-range 9900–4100 cm−1. Therefore, the former was selected to build the PLSTW.
3.2. Measurement of NIR spectra 257 NIRDRSs for SMZCO were obtained with the resolution 8 cm−1 and the number of scans 64 as illustrated in Fig. 2 (a), and a typical
Table 1 Main pretreatments, spectral sub-ranges and corresponding performances of the PLSTW and the PLSTMP. Item
Model
Pretreatment
Sub-range (cm−1)
Factor Number
RMSEC
Rc
RMSECV
Rcv
RMSEP
Rp
TW (g)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
MSC + FD + SGS + MC No preprocessing MSC SNV FD SD SGS NDS MC VS MSC + FD + SGS + MC SGS + MC No preprocessing MSC SNV FD SD SGS NDS MC VS SGS + MC
7500–5900
10 7 10 11 13 11 7 15 13 15 7 8 12 9 11 15 11 12 14 8 11 7
0.00398 0.01010 0.00447 0.00439 0.00627 0.00833 0.01010 0.00816 0.00435 0.00645 0.00472 0.00118 0.00275 0.00137 0.00130 0.00086 0.00109 0.00276 0.00213 0.00118 0.00293 0.00123
0.9543 0.7551 0.9420 0.9443 0.9014 0.8410 0.7546 0.8412 0.9453 0.8903 0.9351 0.9205 0.6218 0.8909 0.9024 0.9591 0.9386 0.6203 0.7653 0.9205 0.5791 0.9139
0.00514 0.01160 0.00509 0.00511 0.01160 0.01770 0.01160 0.01140 0.00505 0.00870 0.00546 0.00129 0.00337 0.00156 0.00157 0.00287 0.00373 0.00339 0.00315 0.00129 0.00356 0.00132
0.9230 0.6966 0.9245 0.9238 0.7390 0.5611 0.6962 0.7258 0.9260 0.8038 0.9127 0.9050 0.4741 0.8573 0.8565 0.6249 0.4438 0.4712 0.5632 0.9050 0.4384 0.8995
0.00460 0.01020 0.00503 0.00479 0.00887 0.01500 0.01020 0.01050 0.00466 0.00905 0.00511 0.00125 0.00371 0.00174 0.00182 0.00265 0.00292 0.00370 0.00359 0.00125 0.00340 0.00131
0.9312 0.7258 0.9175 0.9272 0.7805 0.5703 0.7241 0.7920 0.9281 0.8110 0.9121 0.9025 0.5262 0.8262 0.8160 0.7340 0.6722 0.5294 0.6508 0.9017 0.5459 0.8953
TMP content (g/g)
SGS: 7-point SGS; NDS: 5-point NDS (FD). The optimal pretreatment methods are highlighted in bold.
9900–4100 8920–4450
9900–4100
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
81
Fig. 3. Scatter plots for the TW (a and b) and TMP content (c and d) of SMZCO: (a and c) show the relationships between RMSECV and factor number of PLSTW and PLSTMP, respectively; the optimal factor numbers are highlighted by red . (b and d) show the relationships between reference value and prediction value for the calibration (red ) and prediction (blue ) sets of PLSTW and PLSTMP, separately. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.3.3. Establishment and validation of PLSTW The TW reference values 0.5294–0.5875 g of prediction samples were approximately uniformly distributed over 0.4999–0.5897 g of calibration samples. Fig. 3 (a) shows that the PLSTW using 10 factors reaches the lowest RMSECV 0.00514 g. Fig. 3 (b) gives a good linear relationship (y = 0.9105 x + 0.0492, Rc 0.9543) between the reference value and the prediction value of TW. The main PLSTW models and their performances are shown in Table 1. It is obvious that the optimal PLSTW is the Model 1. Comparing with the optimal PLSTW (Model 1), the second-best PLSTW (Model 9) uses more factors, has more RMSEC and RMSEP, and has less Rc and Rp. 3.4. Prediction of TMP content 3.4.1. Pretreatment of spectral information The main pretreatment methods for spectral information and the corresponding performances of PLSTMP are shown in Table 1. The combination of SGS and MC was selected to pretreat the spectral information on the basis of the high overall performance of PLSTMP. 3.4.2. Selection of spectral sub-range The blue spectrum in Fig. 2 (b) shows that the characteristic absorptions of C-H and N-H groups of TMP are mainly in the range of 8920–4450 cm−1, involving the first overtones of C–H, 5880 cm− 1, 5797 cm−1, 5770 cm−1 and 5666 cm−1, the second overtone of C–H, 8230 cm−1, the combinations of C-H, 5985 cm− 1, 5920 cm− 1 and 4655–4532 cm−1, and the first overtone of N-H, 6897–6803 cm−1 [16]. And Table 1 indicates that the PLSTMP performances are higher when the spectral sub-range 8920–4450 cm−1 is used than 9900–4100 cm−1, which includes high levels of noise, especially near 4100 cm−1, as seen in Fig. 2 (a) and (b). Therefore, the former was selected to build the PLSTMP. 3.4.3. Establishment and validation of PLSTMP The reference values of TMP contents, 0.1312–0.1513 g/g, in prediction samples were approximately uniformly distributed over 0.1301– 0.1658 g/g of calibration samples. Fig. 3 (c) presents that the PLSTMP
using 8 factors has the lowest RMSECV 0.00129 g/g. Fig. 3 (d) shows a good linear relationship (y = 0.8482 x+ 0.0216, Rc 0.9205) between the reference value and the prediction value of TMP content. The main PLSTMP models and their performances are shown in Table 1. It is clear that the optimal PLSTMP is the Model 12. Comparing with the optimal PLSTMP (Model 12), the second-best PLSTMP (Model 20) has less Rp. 4. Conclusion The established high-throughput and rapid NIR spectroscopy can be used to simultaneously predict TW and TMP content of SMZCO for controlling UODU. This novel strategy expands the NIR application in controlling UODU, especially in the high-throughput prediction of TWs and contents of the compound pharmaceutical tablets, and it is an important complement to the common NIR on-line analytical method for pharmaceutical tablets. Acknowledgments We are sincerely grateful for the funding support of the Municipal Science and Technology Committee of Chongqing (cstc2012ggyyjs10039). References [1] The United States Pharmacopeial Convention, The United States Pharmacopeia, thirty-eighth ed., 2015 675 (Rockville, MD). [2] European Directorate for the Quality of Medicines & Healthcare, The European Pharmacopoeia, eighth ed., 2013 357 (Strasbourg). [3] The Society of Japanese Pharmacopeia, The Japanese Pharmacopoeia, sixteenth ed., 2011 127 (Tokyo). [4] Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China 2010, volume II, China Medical Science Press, Beijing, 2010 (pp. Appendix I A and X E). [5] Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China 2010, volume II, China Medical Science Press, Beijing, 2010 594. [6] M.L. Landrum, N.G. Conger, M.A. Forgione, Trimethoprim-sulfamethoxazole in the treatment of stenotrophomonas maltophilia osteomyelitis, Clin. Infect. Dis. 40 (2005) 1551–1552. [7] The United States Pharmacopeial Convention, The United States Pharmacopeia, thirty-eighth ed., 2015 5399 (Rockville, MD). [8] K. Järvinen, W. Hoehe, M. Järvinen, S. Poutiainen, M. Juuti, S. Borchert, In-line monitoring of the drug content of powder mixtures and tablets by near-infrared
82
[9]
[10]
[11]
[12]
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82 spectroscopy during the continuous direct compression tableting process, Eur. J. Pharm. Sci. 48 (2013) 680–688. W. Zhang, Z.Y. Qu, Y.P. Wang, C.L. Yao, X.Y. Bai, S. Bian, B. Zhao, Near-infrared reflectance spectroscopy (NIRS) for rapid determination of ginsenoside Rg1 and Re in Chinese patent medicine naosaitong pill, Spectrochim. Acta A 139 (2015) 184–188. Q. Kang, Q.G. Ru, Y. Liu, L.Y. Xu, J. Liu, Y.F. Wang, Y.W. Zhang, H. Li, Q. Zhang, Q. Wu, On-line monitoring the extract process of Fu-fang shuanghua oral solution using near infrared spectroscopy and different PLS algorithms, Spectrochim. Acta A 152 (2016) 431–437. J. Li, Y. Jiang, Q. Fan, Y. Chen, R.Q. Wu, Simultaneous determination of the impurity and radial tensile strength of reduced glutathione tablets by a high selective NIRPLS method, Spectrochim. Acta A 125 (2014) 278–284. C.V. Möltgen, T. Herdling, G. Reich, A novel multivariate approach using sciencebased calibration for direct coating thickness determination in real-time NIR process monitoring, Eur. J. Biopharm. 85 (2013) 1056–1063.
[13] J. Wu, W. Luo, X.K. Wang, Q. Cheng, C.G. Sun, H. Li, A new application of WT-ANN method to control the preparation process of metformin hydrochloride tablets by near infrared spectroscopy compared to PLS, J. Pharm. Biomed. Anal. 80 (2013) 186–191. [14] I. Tomuta, L. Rus, R. Iovanov, L.L. Rus, High-throughput NIR-chemometric methods for determination of drug content and pharmaceutical properties of indapamide tablets, J. Pharm. Biomed. Anal. 84 (2013) 285–292. [15] A.C.O. Neves, G.M. Soares, S.C. de Morais, F.S.L. da Costa, D.L. Porto, K.M.G. de Lima, Dissolution testing of isoniazid, rifampicin, pyrazinamide and ethambutol tablets using near-infrared spectroscopy (NIRS) and multivariate calibration, J. Pharm. Biomed. Anal. 57 (2012) 115–119. [16] J. Workman, L. Weyer, Practical guide to interpretive near-infrared spectroscopy, first ed. CRC Press, USA, 2008.
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
High-throughput prediction of tablet weight and trimethoprim content of compound sulfamethoxazole tablets for controlling the uniformity of dosage units by NIR Yanhong Dong, Juan Li, Xiaoxiao Zhong, Liya Cao, Yang Luo, Qi Fan ⁎ School of Pharmacy, Chongqing Medical University, Chongqing 400016, China
a r t i c l e
i n f o
Article history: Received 15 November 2015 Received in revised form 15 January 2016 Accepted 20 January 2016 Available online 22 January 2016 Keywords: Uniformity of dosage units Compound sulfamethoxazole tablets Tablet weight Trimethoprim content Near infrared spectroscopy Partial least squares regression
a b s t r a c t This paper establishes a novel method to simultaneously predict the tablet weight (TW) and trimethoprim (TMP) content of compound sulfamethoxazole tablets (SMZCO) by near infrared (NIR) spectroscopy with partial least squares (PLS) regression for controlling the uniformity of dosage units (UODU). The NIR spectra for 257 samples were measured using the optimized parameter values and pretreated using the optimized chemometric techniques. After the outliers were ignored, two PLS models for predicting TW and TMP content were respectively established by using the selected spectral sub-ranges and the reference values. The TW model reaches the correlation coefficient of calibration (Rc) 0.9543 and the TMP content model has the Rc 0.9205. The experimental results indicate that this strategy expands the NIR application in controlling UODU, especially in the highthroughput and rapid analysis of TWs and contents of the compound pharmaceutical tablets, and may be an important complement to the common NIR on-line analytical method for pharmaceutical tablets. © 2016 Elsevier B.V. All rights reserved.
1. Introduction The tablet is the most common dosage form as the finished pharmaceutical products (FPPs). In the quality control (QC) of orally administered pharmaceutical tablets, the uniformity of dosage units (UODU) is a vitally important test item stated in United States Pharmacopeia (USP), European Pharmacopoeia (EP), Japanese Pharmacopoeia (JP) and Chinese Pharmacopoeia (CP) [1–4] because it is related to the safety and effectivity of dosage units. The compound sulfamethoxazole tablets (SMZCO), containing 400.0 mg of sulfamethoxazole (SMZ) and 80.0 mg of trimethoprim (TMP) in each tablet, are prescribed for the treatment of infections, such as urinary tract infection, pneumocystis carinii pneumonia and Stenotrophomonas maltophilia osteomyelitis [5,6]. The UODU of SMZCO is demonstrated by weight variation (WV) for SMZ (its ratio is more than 25% by weight) and by content uniformity (CU) for TMP (its ratio is less than 25% by weight) in USP and CP [1,4]. The WV and CU are calculated separately on the basis of tablet weight (TW) and TMP content. The classical measurement method for TW is weighing method [1–4], and one for TMP content is high performance liquid chromatography (HPLC) [5,7]. However, weighing method is low-throughput and HPLC is sample-destructive, environment polluting and time consuming.
⁎ Corresponding author. E-mail address: [email protected] (Q. Fan).
http://dx.doi.org/10.1016/j.saa.2016.01.030 1386-1425/© 2016 Elsevier B.V. All rights reserved.
Near infrared (NIR) spectroscopy with chemometric techniques can extract simultaneously multiple chemical and physical information of an analyte, which contains X–H (X = C, O, N). Because it is highthroughput, nondestructive, pollution-free, rapid, easy-to-use and available for on-line process control, NIR spectroscopy has been extensively used for the quantitative analysis of multiple chemical and physical characteristics of pharmaceuticals, for example, content [8–10], impurity [11], coating thickness [12], hardness [13], strength [11], disintegration time [14], and dissolution [15]. In this work, the feasibility of high-throughput and rapid prediction of TW and TMP content of SMZCO for controlling UODU is evaluated by NIR spectroscopy with partial least squares (PLS) regression principally based on the reasons below. Firstly, the relationship between TW and SMZ content can be described by Eq. (1) [1]. TW ¼ CSMZ TW=A
ð1Þ
where TW is the individual weight of the tablet tested, CSMZ is the individual estimated SMZ content of the tablet tested, TW is the mean of TWs, and A is the SMZ assay. Secondly, the molecular structures of both SMZ and TMP include many C–H and N–H groups, as seen in Fig. 1. In a word, both TW and TMP content could be predicted by NIR spectroscopy. This work may expand the NIR application in controlling UODU, especially in the high-throughput prediction of TWs and contents of the compound pharmaceutical tablets. And the established
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
79
Fig. 1. Chemical structural formulas of SMZ (a) and TMP (b).
high-throughput and rapid strategy may be an important complement to the common NIR on-line analytical method for pharmaceutical tablets. 2. Materials and methods 2.1. Instruments The NIR spectra were measured by FT-NIR analyzer (Antaris II, Thermo Fisher Scientific, USA), which was furnished with integrating sphere and sample holder and was controlled by the software package RESULT 3.0. The chemometric calculations were executed using the software TQ Analyst 8.0. The reference values of TW and TMP content were determined respectively by analytical balance (AB204-E, Mettler Toledo, Switzerland) and HPLC system (Agilent 1100, Agilent, USA). 2.2. Reagents and samples The reference material of TMP (batch number: 100031–200304) was purchased from the Chinese National Institutes for Food and Drug Control (Beijing, China). The active pharmaceutical ingredients (APIs) of SMZ (batch number: SMXB01312; assay: ≥99.0%) and TMP (batch number: A-10111204022; assay: ≥99.0%) were obtained respectively from Suzhou Sanray Pharmaceutical Co., Ltd. (Suzhou, China) and Shouguang Fukang Pharmaceutical Co., Ltd. (Shouguang, China). Acetonitrile was HPLC grade, other chemicals were analytical grade, and experimental water was purified water. In this experiment, 257 samples of SMZCO (400.0 mg of SMZ and 80.0 mg of TMP in each tablet) from 13 batches were produced by four Chinese manufacturers. The appearances of all samples are consistent: white cylindrical flat-faced tablets coded with “SMZ” and “CO” on the scored side. 2.3. Reference methods 2.3.1. Determination of TW The TW reference value for each tablet was determined by using an analytical balance [1–4]. 2.3.2. Determination of TMP content The reference value of TMP content in each tablet was determined by HPLC and calculated using the external standard method. The separation was performed on an Agilent ZORBAX SB-C18 column (150 mm × 4.6 mm, 5 μm), whose temperature was maintained at 30 °C, by using acetonitrile–water–triethylamine (200:799:1, v/v/v; pH 5.9) as the mobile phase at a flow rate of 1.0 mL/min. The TMP content was detected at 240 nm. And the injection volume was 10 μL [5]. 2.4. Measurement of NIR spectra Each tablet was put on the circle detection window of integrating sphere module of FT-NIR analyzer and the scored side of tablet was
upward. The sample holder was adjusted to ensure the position of each tablet concentric with the detection window. A Fourier transform near infrared diffuse reflectance spectrum (NIRDRS) for each tablet was acquired in the range of 10,000–4000 cm−1 with same resolution and same number of scans, which were selected from the resolutions 4, 8 and 16 cm−1 and the numbers of scans 32 and 64 on the basis of small spectral variation and short measuring time. The ambient temperature was about 25 °C. The background spectrum was determined under the same conditions to eliminate the interferences from H2O and CO2 in air. The API of SMZ and the API of TMP were respectively poured into two same empty circular glass vials, which diameters are about 1.0 cm, up to about 1.0 cm of thickness [16]. And their NIRDRSs were also measured in the above way.
2.5. Prediction of TW 2.5.1. Pretreatment of spectral information The spectral information was pretreated by the chemometric techniques selected from no preprocessing, multiplicative signal correction (MSC) or standard normal variate (SNV) for eliminating the interferences from the particle size and the compactness, first derivative (FD) or second derivative (SD) for deducting the background and separating the overlapping signals, Savitzky–Golay smoothing (SGS) or Norris derivative smoothing (NDS) for removing the noise, and mean centering (MC) and variance scaling (VS) for improving the performances of PLS model for predicting TW (PLSTW). 2.5.2. Selection of spectral sub-range The spectral sub-range, used to establish the PLSTW, was selected primarily on the basis of the characteristic absorptions of C–H and N–H groups of SMZ to reduce data redundancy for enhancing the PLSTW performances and increasing the computational speed.
2.5.3. Establishment and validation of PLSTW The NIRDRSs for 257 samples of SMZCO from 13 batches and four manufacturers were divided into the calibration and prediction sets in the ratio of 2:1. The reference values of 86 prediction samples from 13 batches and four manufacturers were approximately uniformly distributed over the range of the reference values of 171 calibration samples from 13 batches and four manufacturers. After the outliers were diagnosed by TQ Analyst 8.0 and ignored, the PLSTW was built using the TW reference values, the NIRDRSs for calibration samples and the optimal number of factors to reach the lowest root mean square error of cross-validation (RMSECV). And the PLSTW was validated with the NIRDRSs and the TW reference values for prediction samples. The performances of PLSTW were assessed with the root mean square errors of calibration (RMSEC), RMSECV and prediction (RMSEP), and with the correlation coefficients of calibration (Rc), cross-validation (Rcv) and prediction (Rp).
80
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
Fig. 2. Raw NIRDRSs were obtained with the resolution 8 cm−1 and the number of scans 64. (a) shows 257 NIRDRSs for SMZCO and (b) gives a typical NIRDRS for SMZCO (black) and two NIRDRSs for the APIs of SMZ (red) and TMP (blue). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
NIRDRS for SMZCO (black) and two NIRDRSs for the APIs of SMZ (red) and TMP (blue) in Fig. 2 (b).
2.6. Prediction of TMP content The NIRDRSs used in the prediction of TW were simultaneously used to predict the TMP content. And the PLS model for predicting TMP content (PLSTMP) was built and validated using the reference values of TMP contents, similarly to the PLSTW.
3.3. Prediction of TW 3.3.1. Pretreatment of spectral information The main pretreatment methods for spectral information and the corresponding performances of PLSTW are shown in Table 1. The combination of MSC, FD, SGS and MC was selected to pretreat the spectral information and then the high overall performance of PLSTW was reached.
3. Results and discussion 3.1. Reference values The TW reference values of 257 samples of SMZCO from 13 batches and four manufacturers were 0.4999–0.5897 g, equivalently to 90.6%– 106.9% of the label claim (400.0 mg of SMZ in each tablet). The reference values of TMP contents in 257 samples were 0.1301– 0.1658 g/g, equivalently to 94.6%–111.8% of the label claim (80.0 mg of TMP in each tablet).
3.3.2. Selection of spectral sub-range The red spectrum in Fig. 2 (b) represents that the characteristic absorptions of C–H and N–H groups of SMZ are mainly in the range of 7500–5900 cm−1, including the combinations of C–H, 7300 cm−1, 5985 cm−1 and 5920 cm−1, and the overtone of N–H, near 6800 cm−1 [16]. And Table 1 gives that the PLSTW performances are higher when the spectral sub-range 7500–5900 cm−1 is used than the wider spectral sub-range 9900–4100 cm−1. Therefore, the former was selected to build the PLSTW.
3.2. Measurement of NIR spectra 257 NIRDRSs for SMZCO were obtained with the resolution 8 cm−1 and the number of scans 64 as illustrated in Fig. 2 (a), and a typical
Table 1 Main pretreatments, spectral sub-ranges and corresponding performances of the PLSTW and the PLSTMP. Item
Model
Pretreatment
Sub-range (cm−1)
Factor Number
RMSEC
Rc
RMSECV
Rcv
RMSEP
Rp
TW (g)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
MSC + FD + SGS + MC No preprocessing MSC SNV FD SD SGS NDS MC VS MSC + FD + SGS + MC SGS + MC No preprocessing MSC SNV FD SD SGS NDS MC VS SGS + MC
7500–5900
10 7 10 11 13 11 7 15 13 15 7 8 12 9 11 15 11 12 14 8 11 7
0.00398 0.01010 0.00447 0.00439 0.00627 0.00833 0.01010 0.00816 0.00435 0.00645 0.00472 0.00118 0.00275 0.00137 0.00130 0.00086 0.00109 0.00276 0.00213 0.00118 0.00293 0.00123
0.9543 0.7551 0.9420 0.9443 0.9014 0.8410 0.7546 0.8412 0.9453 0.8903 0.9351 0.9205 0.6218 0.8909 0.9024 0.9591 0.9386 0.6203 0.7653 0.9205 0.5791 0.9139
0.00514 0.01160 0.00509 0.00511 0.01160 0.01770 0.01160 0.01140 0.00505 0.00870 0.00546 0.00129 0.00337 0.00156 0.00157 0.00287 0.00373 0.00339 0.00315 0.00129 0.00356 0.00132
0.9230 0.6966 0.9245 0.9238 0.7390 0.5611 0.6962 0.7258 0.9260 0.8038 0.9127 0.9050 0.4741 0.8573 0.8565 0.6249 0.4438 0.4712 0.5632 0.9050 0.4384 0.8995
0.00460 0.01020 0.00503 0.00479 0.00887 0.01500 0.01020 0.01050 0.00466 0.00905 0.00511 0.00125 0.00371 0.00174 0.00182 0.00265 0.00292 0.00370 0.00359 0.00125 0.00340 0.00131
0.9312 0.7258 0.9175 0.9272 0.7805 0.5703 0.7241 0.7920 0.9281 0.8110 0.9121 0.9025 0.5262 0.8262 0.8160 0.7340 0.6722 0.5294 0.6508 0.9017 0.5459 0.8953
TMP content (g/g)
SGS: 7-point SGS; NDS: 5-point NDS (FD). The optimal pretreatment methods are highlighted in bold.
9900–4100 8920–4450
9900–4100
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82
81
Fig. 3. Scatter plots for the TW (a and b) and TMP content (c and d) of SMZCO: (a and c) show the relationships between RMSECV and factor number of PLSTW and PLSTMP, respectively; the optimal factor numbers are highlighted by red . (b and d) show the relationships between reference value and prediction value for the calibration (red ) and prediction (blue ) sets of PLSTW and PLSTMP, separately. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.3.3. Establishment and validation of PLSTW The TW reference values 0.5294–0.5875 g of prediction samples were approximately uniformly distributed over 0.4999–0.5897 g of calibration samples. Fig. 3 (a) shows that the PLSTW using 10 factors reaches the lowest RMSECV 0.00514 g. Fig. 3 (b) gives a good linear relationship (y = 0.9105 x + 0.0492, Rc 0.9543) between the reference value and the prediction value of TW. The main PLSTW models and their performances are shown in Table 1. It is obvious that the optimal PLSTW is the Model 1. Comparing with the optimal PLSTW (Model 1), the second-best PLSTW (Model 9) uses more factors, has more RMSEC and RMSEP, and has less Rc and Rp. 3.4. Prediction of TMP content 3.4.1. Pretreatment of spectral information The main pretreatment methods for spectral information and the corresponding performances of PLSTMP are shown in Table 1. The combination of SGS and MC was selected to pretreat the spectral information on the basis of the high overall performance of PLSTMP. 3.4.2. Selection of spectral sub-range The blue spectrum in Fig. 2 (b) shows that the characteristic absorptions of C-H and N-H groups of TMP are mainly in the range of 8920–4450 cm−1, involving the first overtones of C–H, 5880 cm− 1, 5797 cm−1, 5770 cm−1 and 5666 cm−1, the second overtone of C–H, 8230 cm−1, the combinations of C-H, 5985 cm− 1, 5920 cm− 1 and 4655–4532 cm−1, and the first overtone of N-H, 6897–6803 cm−1 [16]. And Table 1 indicates that the PLSTMP performances are higher when the spectral sub-range 8920–4450 cm−1 is used than 9900–4100 cm−1, which includes high levels of noise, especially near 4100 cm−1, as seen in Fig. 2 (a) and (b). Therefore, the former was selected to build the PLSTMP. 3.4.3. Establishment and validation of PLSTMP The reference values of TMP contents, 0.1312–0.1513 g/g, in prediction samples were approximately uniformly distributed over 0.1301– 0.1658 g/g of calibration samples. Fig. 3 (c) presents that the PLSTMP
using 8 factors has the lowest RMSECV 0.00129 g/g. Fig. 3 (d) shows a good linear relationship (y = 0.8482 x+ 0.0216, Rc 0.9205) between the reference value and the prediction value of TMP content. The main PLSTMP models and their performances are shown in Table 1. It is clear that the optimal PLSTMP is the Model 12. Comparing with the optimal PLSTMP (Model 12), the second-best PLSTMP (Model 20) has less Rp. 4. Conclusion The established high-throughput and rapid NIR spectroscopy can be used to simultaneously predict TW and TMP content of SMZCO for controlling UODU. This novel strategy expands the NIR application in controlling UODU, especially in the high-throughput prediction of TWs and contents of the compound pharmaceutical tablets, and it is an important complement to the common NIR on-line analytical method for pharmaceutical tablets. Acknowledgments We are sincerely grateful for the funding support of the Municipal Science and Technology Committee of Chongqing (cstc2012ggyyjs10039). References [1] The United States Pharmacopeial Convention, The United States Pharmacopeia, thirty-eighth ed., 2015 675 (Rockville, MD). [2] European Directorate for the Quality of Medicines & Healthcare, The European Pharmacopoeia, eighth ed., 2013 357 (Strasbourg). [3] The Society of Japanese Pharmacopeia, The Japanese Pharmacopoeia, sixteenth ed., 2011 127 (Tokyo). [4] Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China 2010, volume II, China Medical Science Press, Beijing, 2010 (pp. Appendix I A and X E). [5] Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China 2010, volume II, China Medical Science Press, Beijing, 2010 594. [6] M.L. Landrum, N.G. Conger, M.A. Forgione, Trimethoprim-sulfamethoxazole in the treatment of stenotrophomonas maltophilia osteomyelitis, Clin. Infect. Dis. 40 (2005) 1551–1552. [7] The United States Pharmacopeial Convention, The United States Pharmacopeia, thirty-eighth ed., 2015 5399 (Rockville, MD). [8] K. Järvinen, W. Hoehe, M. Järvinen, S. Poutiainen, M. Juuti, S. Borchert, In-line monitoring of the drug content of powder mixtures and tablets by near-infrared
82
[9]
[10]
[11]
[12]
Y. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 159 (2016) 78–82 spectroscopy during the continuous direct compression tableting process, Eur. J. Pharm. Sci. 48 (2013) 680–688. W. Zhang, Z.Y. Qu, Y.P. Wang, C.L. Yao, X.Y. Bai, S. Bian, B. Zhao, Near-infrared reflectance spectroscopy (NIRS) for rapid determination of ginsenoside Rg1 and Re in Chinese patent medicine naosaitong pill, Spectrochim. Acta A 139 (2015) 184–188. Q. Kang, Q.G. Ru, Y. Liu, L.Y. Xu, J. Liu, Y.F. Wang, Y.W. Zhang, H. Li, Q. Zhang, Q. Wu, On-line monitoring the extract process of Fu-fang shuanghua oral solution using near infrared spectroscopy and different PLS algorithms, Spectrochim. Acta A 152 (2016) 431–437. J. Li, Y. Jiang, Q. Fan, Y. Chen, R.Q. Wu, Simultaneous determination of the impurity and radial tensile strength of reduced glutathione tablets by a high selective NIRPLS method, Spectrochim. Acta A 125 (2014) 278–284. C.V. Möltgen, T. Herdling, G. Reich, A novel multivariate approach using sciencebased calibration for direct coating thickness determination in real-time NIR process monitoring, Eur. J. Biopharm. 85 (2013) 1056–1063.
[13] J. Wu, W. Luo, X.K. Wang, Q. Cheng, C.G. Sun, H. Li, A new application of WT-ANN method to control the preparation process of metformin hydrochloride tablets by near infrared spectroscopy compared to PLS, J. Pharm. Biomed. Anal. 80 (2013) 186–191. [14] I. Tomuta, L. Rus, R. Iovanov, L.L. Rus, High-throughput NIR-chemometric methods for determination of drug content and pharmaceutical properties of indapamide tablets, J. Pharm. Biomed. Anal. 84 (2013) 285–292. [15] A.C.O. Neves, G.M. Soares, S.C. de Morais, F.S.L. da Costa, D.L. Porto, K.M.G. de Lima, Dissolution testing of isoniazid, rifampicin, pyrazinamide and ethambutol tablets using near-infrared spectroscopy (NIRS) and multivariate calibration, J. Pharm. Biomed. Anal. 57 (2012) 115–119. [16] J. Workman, L. Weyer, Practical guide to interpretive near-infrared spectroscopy, first ed. CRC Press, USA, 2008.