Real-Time Release Testing of Herbal Extract Powder by Near-Infrared Spectroscopy considering the Uncertainty around Specification Limits

Shi, Guolin; Xu, Bing; Wang, Xin; Xue, Zhong; Shi, Xinyuan; Qiao, Yanjiang

doi:https://doi.org/10.1155/2019/4139762

Journal of Spectroscopy

On this page

Abstract Introduction Experimental Results and Discussion Conclusions Data Availability Additional Points Conflicts of Interest Acknowledgments Supplementary Materials References Copyright Related Articles

Special Issue

State-of-the-Art Infrared Applications in Drugs, Dietary Supplements, and Nutraceuticals

View this Special Issue

Research Article | Open Access

Volume 2019 | Article ID 4139762 | https://doi.org/10.1155/2019/4139762

Real-Time Release Testing of Herbal Extract Powder by Near-Infrared Spectroscopy considering the Uncertainty around Specification Limits

Guolin Shi,¹Bing Xu,^1,2Xin Wang,³Zhong Xue,⁴Xinyuan Shi,^1,2and Yanjiang Qiao^1,2

Academic Editor: Alessandra Durazzo

Received05 Oct 2018

Revised15 Jan 2019

Accepted23 Jan 2019

Published03 Mar 2019

Abstract

The concept of real-time release testing (RTRT) has recently been adopted by the production of pharmaceuticals in order to provide high-level guarantee of product quality. Process analytical technology (PAT) is an attractive and efficient way for realizing RTRT. In this paper, near-infrared (NIR) determination of cryptotanshinone and tanshinoneII_A content in tanshinone extract powders was taken as the research object. The aim of NIR analysis is to reliably declare the extract product as compliant with its specification limits or not. First, the NIR quantification method was developed and the parameters of the multivariate calibration model were optimized. The reliable concentration ranges covering the specification limits of two APIs were successfully verified by the accuracy profile (AP) methodology. Then, with the designed validation data from AP, the unreliability graph as the decision tool was built. Innovatively, the β-content, γ-confidence tolerance intervals (β-CTIs) around the specification limits were estimated. During routine use, the boundary of β-CTIs could help decide whether the NIR prediction results are acceptable. The proposed method quantified the analysis risk near the specification limits and confirmed that the unreliable region was useful to release the product quality in a real-time way. Such release strategy could be extended for other PAT applications to improve the reliability of results.

1. Introduction

Radix Salvia Miltiorrhizae is the dried root of Salvia miltiorrhiza Bge [1]. It is widely used in several therapy systems for the treatment of angina pectoris, coronary heart disease and myocardial infarction, atherosclerosis, chronic renal failure, and liver fibrosis [2]. Tanshinone extract, the important components in Radix Salvia Miltiorrhizae, is listed in the Chinese Pharmacopoeia (ChP, 2015 edition). The tanshinone extract powders were generally manufactured using a series of batch operations, including extracting, filtering, concentrate, washing, drying, and milling.

Traditionally, the quality of tanshinone extract powder was assured by laboratory testing after the manufacturing was completed. And two active pharmaceutical ingredients (APIs), i.e., the cryptotanshinone and the tanshinoneII_A, were assayed by the HPLC method. However, the HPLC analysis is time-consuming and requires labor-intensive protocols including sample collection, sample pretreatment, sample analysis, and data processing procedures. Besides, this conventional approach was conducted on limited samples and had been at risk in providing qualified products to the public.

Since the promulgation of the process analytical technology (PAT) guidance in September 2004 [3], the American Food and Drug Administration (FDA) has encouraged the pharmaceutical manufacturers to adopt new technologies in pharmaceutical process, mainly for designing, analyzing, and controlling manufacturing through timely measurements of critical quality and performance attributes of raw and in-process materials and processes with goal of ensuring final product quality. Real-time release testing (RTRT) [4], which is advocated to substitute the end-point testing, is the ability to evaluate and ensure the quality of in-process and/or final product based on process data, which typically include a valid combination of measured material attributes and process controls. Advances in quality by design (QbD) have shown that the application of RTRT in any stage of the manufacturing process and any type of finished product may provide greater assurance of product quality than finished product testing alone [5].

The rapid spectroscopy techniques are important part of RTRT plans. Near-infrared (NIR) spectroscopy has proven to be effective for both qualitative [6–8] and quantitative analysis [9–12] in the pharmaceutical industry due to its high efficiency, nondestructive nature, and capacity to measure both physical and chemical properties with minimal or no sample preparation [13]. It is more and more considered as an attractive and promising analytical tool for PAT. Recently, the NIR spectroscopy has been introduced into the Chinese herbal medicine production processes, such as the extraction process of Epimedium brevicornum Maxim [14] and Qizhiweitong granules [15], the alcohol precipitation process of Reduning injection [16], the purification process of Aesculi semen extracts [17], the enrichment process of Danhong injection [18], the fluidizing drying process of Poria cocos formula granules [19], and the mixing process [20]. By summarizing the NIR analysis results of Chinese medicine materials from the reported literatures, it was found that the relative standard errors of prediction (RSEP) were in the range of 1.51% to 10.41% [13–19] and the prediction error of some herbal samples exceeded the general target acceptance criteria for bulk drug (2%) or dosage form (10%) [21]. Therefore, the operable region and unreliability region of these analytical methods should be judged to confirm their scope of application.

Conventionally, the performance of the NIR analytical method was evaluated by chemometric indicators [22–25], such as the root-mean-square error (RMSE) [26], the correlation coefficient (r), the ratio of performance to deviation (RPD) [27], which only gave the average level of information about errors and bias of NIR method and did not provide the uncertainty of each individual prediction over the range of measurement [28]. Therefore, more and more researchers adopted the accuracy profile (AP) approach to evaluate the risk or confidence of the NIR method [29–32]. The core of the AP methodology, in agreement with the ICH Q2A guidance, is to use the β-expectation tolerance interval (β-ETI) to integrate the trueness, the intermediate precision coefficient variation, as well as the β chance for future results [33]. Based on the validation results from the AP approach, Rozet et al. [34] further identified the unreliable region around the specification limit by intersection of the upper and lower β-ETIs with the specification limit. Such methods were successfully applied for HPLC-UV quantification of (R)-timolol impurity in (S)-timolol drug substance and for NIRS quantification of acetaminophen in the uniformity of dosage units (UDU) test [34, 35].

As reported by Saffaj et al., the β-content, γ-confidence tolerance interval (β-CTI) could provide a better estimate of measurement risk than β-expectation tolerance interval and gave the best guarantee concerning the decision of declaring a method as valid and reliable [36–38]. Our previous work also revealed that the overall uncertainty estimated by the β-CTI from the total error (bias and standard deviation) was similar to the overall uncertainty assessed from validation data according to the trueness, precision, and robustness experiments [39]. In this work, NIR was used as a rapid detective tool to assay the APIs content of tanshinone extract powder. The traditional figures of merit were used to optimize the multivariate calibration model. A full factorial design generating the validation data was used to calculate statistical intervals. The β-content, γ-confidence tolerance interval was for the first time used to develop the unreliable region around the specification limit of tanshinone extract powder, in order to increase the confidence when releasing the multicomponents natural product in a real-time way.

2. Experimental

2.1. Reagents and Materials

The tanshinone extract powders were purchased from Xi’an Changyue Phytochemistry Co., Ltd (Xi’an, China, lot: 140420), Xi’an Honson Biotechnology Co., Ltd. (Xi’an, China, Lot: 141029.) and Shanxi Undersun Bimedtech Co., Ltd (Shanxi, China, Lot: Udst130507). The cryptotanshinone reference standard (lot number: 110852–200806) and tanshinoneII_A reference standard (lot number: 110776–200619) were purchased from the National Institutes for Food and Drug Control (Beijing, China). The acetonitrile and phosphoric acid of HPLC grade were purchased from the Thermo Fisher Scientific Inc. (Massachusetts, USA), and pure water was purchased from Wahaha Co., Ltd. (Hangzhou, China).

2.2. Acquisition of Spectroscopic Data

The sample was held in a circular sample cuvette with a solid cap, and the NIR spectra were collected in the integrating sphere diffuse reflectance mode with the Antaris Nicolet FT-NIR system (Thermo Fisher Scientific Inc., USA) at ambient temperature. Each spectrum was the average of 64 scans with 8 cm⁻¹ resolution. The range of spectra was from 10000 to 4000 cm⁻¹. The background spectrum was taken daily in air.

2.3. Reference Method

The reference method used for cryptotanshinone and tanshinoneII_A determination was HPLC assay recommended by the Chinese Pharmacopoeia (2015 Edition) for the extract of Salvia miltiorrhiza Bge. Firstly, samples were dissolved by methanol properly after NIR scanning. Then, the solution was filtered through a Millipore membrane filter with an average pore diameter of 0.45 μm. Finally, 10 μL of filtrate was injected into the HPLC system for analysis.

An Agilent 1100 series HPLC apparatus, equipped with a quaternary solvent delivery system, an auto sampler, a DAD detector, and HP workstation for data processing were used. The concentration of cryptotanshinone and tanshinoneII_A were analyzed by reverse-phase chromatography on an Agilent XDB C18 column (4.6 × 250 mm, 5 μm) with gradient. The mobile phase A is acetonitrile, and the mobile phase B is phosphoric acid water (0.026%). The elution procedures are as follows: 0∼25 min, 60%∼90% A; 25∼30 min, maintaining 90% A; 30∼31 min, 90%∼60% A; 31∼40 min, 60%∼60% A. The column temperature was 25°C, the flow rate was 1.2 mL·min⁻¹, and the detection wavelength at 270 nm was set.

2.4. Calibration and Validation Protocols

The experimental protocols were created for both calibration and validation sets in order to obtain a robust model. A total of 103 samples were collected in the calibration set. Four grams of tanshinone extract powder sample was weighed and then directly measured by NIR under the conditions specified in Section 2.2.

The external validation set was built with the same method as the calibration set. The validation protocol used the “8 × 5 × 3” full factorial experimental design. Eight different concentration levels of cryptotanshinone, i.e., 0.20%, 0.31% 0.50%, 1.18%, 2.05%, 2.26%, 5.29%, and 9.68%, were investigated. Eight different concentration levels of tanshinoneII_A, i.e., 0.10%, 0.15%, 0.36%, 0.54%, 2.04%, 6.03%, 18.76%, and 27.64%, were investigated. Each concentration level was performed in 5 replicates on 3 different days, resulting in 120 samples in the validation set for both two components. Moreover, all validation samples were from different batches of tanshinone extract powders to test the robustness of the NIR model.

2.5. NIR Method Development

To perform the quantitative determination of cryptotanshinone and tanshinoneII_A content in tanshinone extract powders, the partial least squares (PLS) regression was applied for the sake of linking the NIR spectra with the reference values analyzed by the HPLC method [39]. In order to improve the performance of the PLS model, a variety of spectroscopic data pretreatment methods were investigated to extract the useful information. For example, the first-order derivatives (1std) [40], the second-order derivatives (2ndd) [41] could be used to remove the baseline drift and decrease the overlapping. The multiplicative scatter correction (MSC) [42, 43] and the standard normal variate transformation (SNV) [44] could reduce the light scattering effects. The wavelet denosing of spectra (WDS) [45] and the Savitzky–Golay (SG) smoothing [46] can effectively eliminate the noise.

During the NIR method development process, correlation coefficients r for both the calibration and validation sets, the root-mean-square error of calibration (RMSEC), the root-mean-square error of cross-validation (RMSECV), the root–mean-square error of prediction (RMSEP), and RPD were used to evaluate and select the best NIR calibration model. The optimal latent variables (LVs) used to build PLS model were selected according to comprehensive consideration of RMSEC, RMSECV, RMSEP, and cumulative prediction error sum of square (PRESS) values.

2.6. The Unreliability Graph

The unreliability graph as a decision making tool is a 2D-graphical representation of tolerance intervals aiming at helping the analyst to decide whether an analytical result is reliable or not. For details about the theory, the authors are recommended to refer to the published literatures [36, 47].

2.6.1. Estimation of the β-Content, γ-Confidence Tolerance Interval

The “I × J × K” full factorial validation protocol was designed to obtain the validation dataset, where the effect of three aspects, i.e., conditions (I, ), the number of repetitions (J, ), and the level of concentrations (K, k = 1, 2, … , a), were taken into account [48]. The β-content, γ-confidence tolerance interval can be expressed by the following formula:

In equation (1), is the average value of the results at each concentration level K; , , and , respectively means the intermediate precision, the interseries, and the intraseries variances; denotes the coverage factor and is related to and .

Mee’s approach is used for estimating the β-content, γ-confidence tolerance interval as follows [36, 49]:where the lower () and upper () limits denote a specified proportion of measured results that will fall within the interval [] at specified confidence level. is an approximation to k. denotes the quantile of a noncentral chi-square distribution under the freedom degree of 1. τ means the noncentral parameter. with degrees of freedom denotes the quantile of a noncentral chi-square distribution. And, denotes the mean square ratio MS_B/MS_E. MS_B and MS_E, respectively, denote the mean square of the interseries and the intraseries variances. Under degree freedom and is the percentile of an F distribution. The recommended values of are 0.85, 0.905, and 0.975, which are corresponding to 0.90, 0.95, and 0.99 for , respectively [47].

The β-content, γ-confidence tolerance interval could also be written in a relative form:where is the theoretical value.

2.6.2. Establishment of the Accuracy Profile

In order to globally validate accuracy and robustness of the NIR quantitative method, the accuracy profile is developed as follows [50]:(1)Set acceptance limits ± 20% for natural product in this paper(2)Calculate the β-content, γ-confidence relative tolerance intervals [L (%), U (%)] for each concentration level based on equation (8) at a desired confidence level γ(3)Construct a 2D-accuracy graph with the horizontal axis for the validation standards concentration and vertical axis for the relative tolerance interval limits [L (%), U (%)] and accuracy(4)If [L (%), U (%)] at given concentration levels are within acceptance limits (±20%), it demonstrates that the developed method is accuracy and robustness; otherwise, the method cannot be accepted

2.6.3. Establishment of the Unreliability Graph

The unreliability graph was used as a decision making tool to increase the confidence of real-time release testing at the specification limit. The procedures for developing the unreliability graph are as follows [35, 39]:(1)Set the specification limit (λ) according to the requirement.(2)Calculate the [L, U] for each concentration level based on equation (2) at the desired confidence level γ.(3)Develop a 2D graph with the horizontal axis for the validation standards concentrations and vertical axis for the observed concentrations.(4)Locate the tolerance limits L and U for each validation concentration on the 2D graph. The tolerance limit L at each concentration level was connected into a broken line in turn. The same procedures were also performed on tolerance limits U.(5)Make a straight line perpendicular to the horizontal axis at the specification limit (λ). The intersections of the lower and upper tolerance interval with the specification limit line are defined as L_API and U_API, respectively.(6)Make two straight lines parallel to horizontal axis through the intersections. The area between the two parallel straight lines is the unreliability region around the specification limits.(7)If the analytical results exceed the U_API, the target product can be immediately released; otherwise, it cannot be directly released.

2.7. Software

The SIMCA-P 11.5 (Umetrics, US) and Unscrambler 7.0 (CAMO, Norway) software were used to perform spectral pretreatments. The Matlab 7.0 (Mathwork, USA) with PLS Toolbox 2.1 (Eigenvector Research Inc., USA) was used to carry out the PLS regression. For calculation of the β-content, γ-confidence tolerance intervals, the Matlab codes were referred to [47].

3. Results and Discussion

3.1. NIR Method Development

In this study, raw spectra of 103 samples were obtained by NIR scanning of the extract powders, as shown in. Figure 1. It was difficult to observe the differences from the original spectra because the wave bands were seriously overlapped. Partial least square, one of the most commonly used chemometric methods, was applied to handle the overdetermined problem in the calibration process. And, the PLS1 algorithm was used predict the concentrations of each API in tanshinone extract. Before ascertaining the structure and finetuning the coefficients of PLS model, the Kennard–Stone (K-S) algorithm [51] was used to split the original data set into a calibration set (75 samples) and a test set (28 samples).

Then, various data preprocessing methods in Section 2.5 were used to extract useful information from the noisy spectral data. Tables 1 and 2 show the PLS modeling results in both calibration and prediction of cryptotanshinone and tanshinoneII_A content, respectively. The PLS model based on the second-order derivative NIR absorption spectrum has the optimal results, where the RMSECV and RMSEP values were smallest and the RPD values were highest. This revealed that the robustness of the quantitative models with 2ndd pretreatment was satisfactory and the models had excellent predictive ability. Figure 2 shows the full spectra of all samples through 2ndd preprocessing method. It can be seen that this method can obviously eliminate the baseline drift, remove the background interference, and distinguish the absorption peaks. The characteristic absorption waveband was from 6400 cm⁻¹ to 4000 cm⁻¹. Thereby, 2ndd was chosen as the preprocessing method.

The number of latent variables (LVs) was an important parameter and could directly affect the accuracy of the model. LVs listed in Tables 1 and 2 are optimized by the leave-one-out cross-validation. As can be seen from Figure 3, the RMSE and cumulative PRESS values gradually decreased and eventually did not change as the number of latent variables increased. Consequently, 10 LVs and 8 LVs were, respectively, used to establish the PLS models for cryptotanshinone and tanshinoneII_A, respectively. The RPD value in prediction of tanshinoneII_A was 15.6, which was larger than that in prediction of cryptotanshinone (RPD = 5.3). The reason may be that the standard deviation of tanshinoneII_A content in the test set was higher than that of cryptotanshinone.

(a)

(b)

3.2. NIR Method Validation

According to the ICH Points to consider (R2) document [52], the PLS quantitative model could be classified into high impact model, since the APIs content of tanshinone extract powder predicted by the multivariate calibration models were key indicators for quality control. Consequently, it is necessary to implement method validation to ensure the accuracy and robustness of the quantitative model. After 120 validation samples were prepared according to validation protocols illustrated in Section 2.4, the concentrations of cryptotanshinone and tanshinoneII_A were predicted by the developed NIR calibration model and are listed in appendix Tables S1 and S2, respectively.

The validation statistics for NIR quantitative method are shown in Tables 3 and 4 for cryptotanshinone and tanshinoneII_A, respectively. Taking the validation results of cryptotanshinone for example, the range of concentration studied in Table 3 can be divided into two parts. Part 1 and part 2 were the concentration range of [0.20–1.18]% and [2.05–9.68]%, respectively. In the first part, variances of trueness and precision were exceptionally severe, indicating that the precision and the accuracy of analytical method were anomalous. In part 2, it can be seen that the recovery was closed to 100% and the relative bias was small, indicating that the precision and accuracy of the quantitative model were acceptable within this range. With the same analysis procedures for tanshinoneII_A, it was easy to draw a conclusion that the contents of tanshinoneII_A within [0.10–2.04]% cannot be accurately detected since the precision and accuracy were anomalous. By contrast, the analytical method can accurately determine the tanshinoneII_A in the range [6.03–27.64]%. In conclusion, the NIR method can be used for the determination of cryptotanshinone and tanshinoneII_A both in the second parts of the concentration range.

According to Section 2.6.2, the accuracy profile (AP) was used to globally assess the NIR quantitative analysis method, as shown in Figure 4. The β = 66.7% and γ = 90% suggested by Saffaj’s were applied to estimate the β-content, γ-confidence tolerance interval [36], and the results were listed in Tables 5 and 6. It was clearly seen from Figure 4(a) that the relative tolerance intervals for the first 4 concentrations visibly came out of the two acceptance limits, revealing that the contents within [0.20–1.18]% cannot be accurately measured. Although at content of 2.26% the relative tolerance intervals slightly exceeded upper acceptance limit, the most contents measured by calibration model were acceptable and the recovery in this concentration was no more than 110%. It indicated that the PLS model could determine the contents of cryptotanshinone in this concentration level. However, the tolerance intervals of the last 4 concentration levels fell within the acceptance limits. Consequently, the method was considered to be valid in the concentration range within [2.05–9.68]%. By using the same analysis method, a conclusion can be clearly drawn that the NIR method was valid only for the last 3 concentration levels in detecting the tanshinoneII_A, as shown in Figure 4(b). These findings are consistent with the validation results above.

(a)

(b)

3.3. Real-Time Release Testing for Tanshinone Extract Powders

3.3.1. The Specification Limit

According to the ChP (2015 Edition) [1], the minimum mass content of cryptotanshinone and tanshinoneII_A in tanshinone extract powders are 2.1% and 9.8%, respectively. The specification limits were well located within the reliable concentration regions in Section 3.2, indicating that the developed NIR analysis method can be used to release the tanshinone extract powders in real time.

3.3.2. Unreliability Graph Development

By, respectively, connecting the tolerance limits L and U listed in Tables 5 and 6 at each concentration level, the unreliability graph could be drawn. Figures 5(a) and 6(a) show the unreliability graphs for cryptotanshinone and tanshinoneII_A at full concentration ranges, respectively. The β-content, γ-confidence tolerance interval at the specification limit was then estimated. Taking cryptotanshinone for instance, a line passing through two points, i.e., the upper tolerance limits at 2.39% and 2.98%, was dropped, and the linear function was regressed as (x and y mean validation and observed concentration, respectively). The specification limit of cryptotanshinone was substituted into this function, and U_API of cryptotanshinone were calculated to be 2.53%. Then, another line passing through the lower tolerance limits at 1.72% and 1.98% was regressed as . 2.1% was substituted into this function, and L_API of cryptotanshinone were calculated to be 1.78%. Similarly, as for tanshinoneII_A, the tolerance interval around 9.8% was estimated to be from 8.86% to 10.37%. Figures 5(b) and 6(b) show the unreliable regions for cryptotanshinone and tanshinoneII_A around the specification limits, respectively.

(a)

(b)

(a)

(b)

3.3.3. Real-Time Release Testing

The unreliable region around the specification limit can be seen as risk region or guard-banding, since there was a certain risk in the results of NIR quantitative analysis. During routine use, if the NIR analysis result is larger than the upper limit of the built unreliable region, it is assured that the content would satisfy the specification. And, for the analysis result within the unreliable region, it cannot accurately determine whether the content meets the specification or not. If the NIR analysis result is located under the lower limit of the unreliable region, it absolutely does not meet the target requirements.

For real-time release testing of tanshinone extract powders by NIR analytical method, the release standard was that the contents of cryptotanshinone and tanshinoneII_A must be no less than 2.53% and 10.37%, respectively. Only in this way, the tanshinone extract powder can be directly released to the next pharmaceutical manufacture units or to markets. Otherwise, the tanshinone extract powders cannot be released. The NIR spectroscopy combined with the unreliability graph significantly increases the confidence about the compliance of the product in a real-time way.

4. Conclusions

In this paper, a new release strategy based on the unreliability graph methodology which incorporated the β-content, γ-confidence tolerance intervals, has successfully been achieved. Firstly, the cryptotanshinone and tanshinoneII_A content in tanshinone extract powders were rapidly detected by NIR using the PLS quantitative model. And secondly, this quantitative model was validated by the accuracy profile. The NIR methods can accurately determine the cryptotanshinone in the range [2.05–9.68]% and the tanshinoneII_A in the range [6.03–27.64]%. Finally, the release strategy with NIR quantitative model based on the unreliability graph was applied to real-time release test of tanshinone extract powders. The proposed approach offered a formal statistical framework to show when the analytical methods will provide daily results that can be used efficiently to make adequate decisions. Besides, the release strategy proposed can be applied to any quantitative analytical method and provide greater assurance for the quality of the final products, to achieve the purpose of real-time release.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Additional Points

(1) A release strategy was proposed to determine whether the analytical results were reliable in real-time release testing (RTRT). (2) The β-content, γ-confidence tolerance intervals were applied to establish the unreliability graph as the decision tool. (3) The new release strategy can be used for quality control of the complex system of Chinese medicine product.

Conflicts of Interest

There are no conflicts of interests regarding the publication of this manuscript.

Acknowledgments

The authors are thankful to the research funding supports from the National Science and Technology Major Projects (No. 2018ZX09201011-006, China) and Scientific Research Project of Beijing University of Chinese Medicine (No. 2019-JYB-JS-015).

Supplementary Materials

Table S1: predicted concentrations of cryptotanshinone validation samples expressed as mass content (%). Table S2: predicted concentrations of tanshinoneIIA validation samples expressed as mass content (%). (Supplementary Materials)

References

Chinese Pharmacopoeia Commission, Chinese Pharmacopoeia 2015 Edition, Beijing Chemical industry press, Beijing Shi, China, 2015.
J. Luo, W. Song, G. Yang, H. Xu, and K. Chen, “Compound Danshen (Salvia miltiorrhiza) dripping pill for coronary heart disease: an overview of systematic reviews,” American journal of Chinese medicine, vol. 43, pp. 25–43, 2015.
View at: Publisher Site | Google Scholar
Food and Drug Administration, Guidance for Industry Guidance for Industry PAT–a Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance, US Department of Health and Human Services, Food and Drug Administration, Center for Biologics Evaluation and Research, Rockville Google Scholar, Washington, DC, USA, 2004.
European Commission, EU Guidelines to Good Manufacturing Practice Medicinal Products for Human and Veterinary Use, European Commission, Brussels, Belgium, 2009.
H. F. Xu, X. Zhang, and Y. L. Ma, Real time release testing extends to the overall manufacturing process, FDA, Silver Spring, MD, USA, 2004.
M. Blanco and I. Villarroya, “NIR spectroscopy: a rapid-response analytical tool,” TrAC Trends in Analytical Chemistry, vol. 21, no. 4, pp. 240–250, 2002.
View at: Publisher Site | Google Scholar
H. Yang, J. Irudayaraj, and M. Paradkar, “Discriminant analysis of edible oils and fats by FTIR, FT-NIR and FT-Raman spectroscopy,” Food Chemistry, vol. 93, no. 1, pp. 25–32, 2005.
View at: Publisher Site | Google Scholar
Y. Roggo, L. Duponchel, and J.-P. Huvenne, “Comparison of supervised pattern recognition methods with McNemar’s statistical test,” Analytica Chimica Acta, vol. 477, no. 2, pp. 187–200, 2003.
View at: Publisher Site | Google Scholar
M. Blanco and M. Alcalá, “Content uniformity and tablet hardness testing of intact pharmaceutical tablets by near infrared spectroscopy,” Analytica Chimica Acta, vol. 557, no. 1-2, pp. 353–359, 2006.
View at: Publisher Site | Google Scholar
P. R. Wahl, G. Fruhmann, S. Sacher, G. Straka, S. Sowinski, and J. G. Khinast, “PAT for tableting: inline monitoring of API and excipients via NIR spectroscopy,” European Journal of Pharmaceutics and Biopharmaceutics, vol. 87, no. 2, pp. 271–278, 2014.
View at: Publisher Site | Google Scholar
M. Blanco, M. Alcalá, J. M. González, and E. Torras, “A process analytical technology approach based on near infrared spectroscopy: tablet hardness, content uniformity, and dissolution test measurements of intact tablets,” Journal of Pharmaceutical Sciences, vol. 95, no. 10, pp. 2137–2144, 2006.
View at: Publisher Site | Google Scholar
M. Otsuka and I. Yamane, “Prediction of tablet properties based on near infrared spectra of raw mixed powders by chemometrics: scale-up factor of blending and tableting processes,” Journal of Pharmaceutical Sciences, vol. 98, no. 11, pp. 4296–4305, 2009.
View at: Publisher Site | Google Scholar
M. Jamrógiewicz, “Application of the near-infrared spectroscopy in the pharmaceutical technology,” Journal of Pharmaceutical and Biomedical Analysis, vol. 66, pp. 1–10, 2012.
View at: Publisher Site | Google Scholar
X. U. Ding-Zhou, L. I. Jing, C. H. Liu, L. Yang, L. I. Yu, and M. Zhong, “Application of near-infrared spectroscopy technology in online detection of Epimedium brevicornum Maxim extraction process,” China Modern Medicine, vol. 23, pp. 4–6, 2014.
View at: Publisher Site | Google Scholar
X. S. Meng, Y. Tan, Y. F. Zhang et al., Method and Application for Rapid Detection of the Water Extraction Process of Qizhi weitong Granules by Near Infrared Spectroscopy, 2017, CN106383096A.
Y. X. Wang, H. J. Mi, C. L. Zhang et al., “Near infrared spectroscopy on-line and real-time monitoring of alcohol precipitation process of reduning injection,” China Journal of Chinese Materia Medica, vol. 39, pp. 4608–4614, 2014.
View at: Publisher Site | Google Scholar
Y. L. Xue, A. X. Zuo, X. S. Liu et al., An Automatic Judgment and Control Method in the Process of Extracting Liquid Receiving Eluent Aesculi, 2013, CN103203123B.
Y. Jin, Y. J. Wu, and X. S. Liu, Online Detection Method for Double Effect Concentration Process of Danhong Injection, 2014, CN103913433A.
J. W. Zhang, Y. Y. Zhang, Y. Liu, M. Jia, and G. L. Zhao, NIR On-Line Detection Method for Paeoniflorin Content in Extract of Paeonia Lactiflora Pall, 2011, CN102058682A.
H. Y. Chen, L. I. Qiong-Ya, J. L. Chen, L. J. Luan, and Y. X. Dai, “End point judgment of blending process of zhengtian pills by near infrared spectroscopy,” Chinese Journal of Experimental Traditional Medical Formulae, vol. 22, pp. 13–16, 2016.
View at: Google Scholar
V. P. Shah, K. K. Midha, S. Dighe et al., “Analytical methods validation: bioavailability, bioequivalence, and pharmacokinetic studies,” Journal of Pharmaceutical Sciences, vol. 81, no. 3, pp. 309–312, 1992.
View at: Publisher Site | Google Scholar
M. Blanco and A. Peguero, “Analysis of pharmaceuticals by NIR spectroscopy without a reference method,” TrAC Trends in Analytical Chemistry, vol. 29, no. 10, pp. 1127–1136, 2010.
View at: Publisher Site | Google Scholar
C. V. Liew, A. D. Karande, and P. W. S. Heng, “In-line quantification of drug and excipients in cohesive powder blends by near infrared spectroscopy,” International Journal of Pharmaceutics, vol. 386, no. 1-2, pp. 138–148, 2010.
View at: Publisher Site | Google Scholar
H. Grohganz, D. Gildemyn, E. Skibsted, J. M. Flink, and J. Rantanen, “Towards a robust water content determination of freeze-dried samples by near-infrared spectroscopy,” Analytica Chimica Acta, vol. 676, no. 1-2, pp. 34–40, 2010.
View at: Publisher Site | Google Scholar
C. C. Corredor, D. Bu, and D. Both, “Comparison of near infrared and microwave resonance sensors for at-line moisture determination in powders and tablets,” Analytica Chimica Acta, vol. 696, no. 1-2, pp. 84–93, 2011.
View at: Publisher Site | Google Scholar
L. Norgaard, A. Saudland, J. Wagner, J. P. Nielsen, L. Munck, and S. B. Engelsen, “Interval partial least-squares regression (iPLS): a comparative chemometric study with an example from near-infrared spectroscopy,” Applied Spectroscopy, vol. 54, no. 3, pp. 413–419, 2000.
View at: Publisher Site | Google Scholar
R. Leardi and L. Nørgaard, “Sequential application of backward interval partial least squares and genetic algorithms for the selection of relevant spectral regions,” Journal of Chemometrics, vol. 18, no. 11, pp. 486–497, 2005.
View at: Publisher Site | Google Scholar
A. C. Olivieri, N. M. Faber, J. Ferré, R. Boqué, J. H. Kalivas, and H. Mark, “Uncertainty estimation and figures of merit for multivariate calibration (IUPAC Technical Report),” Pure and Applied Chemistry, vol. 78, no. 3, pp. 633–661, 2006.
View at: Publisher Site | Google Scholar
P. Hubert, J.-J. Nguyen-Huu, B. Boulanger et al., “Harmonization of strategies for the validation of quantitative analytical procedures,” Journal of Pharmaceutical and Biomedical Analysis, vol. 36, no. 3, pp. 579–586, 2004.
View at: Publisher Site | Google Scholar
Z. Wu, B. Xu, M. Du, C. Sui, X. Shi, and Y. Qiao, “Validation of a NIR quantification method for the determination of chlorogenic acid in Lonicera japonica solution in ethanol precipitation process,” Journal of Pharmaceutical and Biomedical Analysis, vol. 62, pp. 1–6, 2012.
View at: Publisher Site | Google Scholar
Y.-C. Feng and C.-Q. Hu, “Construction of universal quantitative models for determination of roxithromycin and erythromycin ethylsuccinate in tablets from different manufacturers using near infrared reflectance spectroscopy,” Journal of Pharmaceutical and Biomedical Analysis, vol. 41, no. 2, pp. 373–384, 2006.
View at: Publisher Site | Google Scholar
J. Mantanus, E. Rozet, K. Van Butsele et al., “Near infrared and Raman spectroscopy as Process Analytical Technology tools for the manufacturing of silicone-based drug reservoirs,” Analytica Chimica Acta, vol. 699, no. 1, pp. 96–106, 2011.
View at: Publisher Site | Google Scholar
P. Hubert, J.-J. Nguyen-Huu, B. Boulanger et al., “Harmonization of strategies for the validation of quantitative analytical procedures,” Journal of Pharmaceutical and Biomedical Analysis, vol. 45, no. 1, pp. 82–96, 2007.
View at: Publisher Site | Google Scholar
E. Rozet, E. Ziemons, R. D. Marini, B. Boulanger, and P. Hubert, “Quality by design compliant analytical method validation,” Analytical Chemistry, vol. 84, no. 1, pp. 106–112, 2011.
View at: Publisher Site | Google Scholar
E. Rozet, E. Ziemons, R. D. Marini, B. Boulanger, and P. Hubert, “Methodology for the validation of analytical methods involved in uniformity of dosage units tests,” Analytica Chimica Acta, vol. 760, pp. 46–52, 2013.
View at: Publisher Site | Google Scholar
T. Saffaj and B. Ihssane, “Uncertainty profiles for the validation of analytical methods,” Talanta, vol. 85, no. 3, pp. 1535–1542, 2011.
View at: Publisher Site | Google Scholar
T. Saffaj and B. Ihssane, “Response to comments on “Uncertainty profiles for the validation of analytical methods”,” Talanta, vol. 94, pp. 361-362, 2012.
View at: Publisher Site | Google Scholar
T. Saffaj and B. Ihssane, “Remarks on “Reply to the responses to the comments on “uncertainty profiles for the validation of analytical methods” by Saffaj and Ihssane”,” Talanta, vol. 106, pp. 155–157, 2013.
View at: Publisher Site | Google Scholar
Z. Xue, B. Xu, X. Shi et al., “Overall uncertainty measurement for near infrared analysis of cryptotanshinone in tanshinone extract,” Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, vol. 170, pp. 39–47, 2017.
View at: Publisher Site | Google Scholar
V. J. Barclay, R. F. Bonner, and I. P. Hamilton, “Application of wavelet transforms to experimental spectra: smoothing, denoising, and data set compression,” Analytical Chemistry, vol. 69, no. 1, pp. 78–90, 1997.
View at: Publisher Site | Google Scholar
M. C. Sarraguça and J. A. Lopes, “Quality control of pharmaceuticals with NIR: from lab to process line,” Vibrational Spectroscopy, vol. 49, no. 2, pp. 204–210, 2009.
View at: Publisher Site | Google Scholar
T. Isaksson and T. Næs, “The effect of multiplicative scatter correction (MSC) and linearity improvement in NIR spectroscopy,” Applied Spectroscopy, vol. 42, no. 7, pp. 1273–1284, 2016.
View at: Publisher Site | Google Scholar
I. Helland, T. Næs, and T. Isaksson, “Related versions of the multiplicative scatter correction method for preprocessing spectroscopic data,” Chemometrics and Intelligent Laboratory Systems, vol. 29, no. 2, pp. 233–241, 1995.
View at: Publisher Site | Google Scholar
R. J. Barnes, M. S. Dhanoa, and S. J. Lister, “Standard normal variate transformation and de-trending of near-infrared diffuse reflectance spectra,” Applied Spectroscopy, vol. 43, no. 5, pp. 772–777, 2016.
View at: Publisher Site | Google Scholar
D. Cozzolino, M. Kwiatkowski, M. Parker et al., “Prediction of phenolic compounds in red wine fermentations by visible and near infrared spectroscopy,” Analytica Chimica Acta, vol. 513, no. 1, pp. 73–80, 2004.
View at: Publisher Site | Google Scholar
Y. de Micalizzi, N. Pappano, and N. Debattista, “First and second order derivative spectrophotometric determination of benzyl alcohol and diclofenac in pharmaceutical forms,” Talanta, vol. 47, no. 3, pp. 525–530, 1998.
View at: Publisher Site | Google Scholar
T. Saffaj, B. Ihssane, F. Jhilal, H. Bouchafra, S. Laslami, and S. A. Sosse, “An overall uncertainty approach for the validation of analytical separation methods,” The Analyst, vol. 138, no. 16, pp. 4677–4691, 2013.
View at: Publisher Site | Google Scholar
B. Govaerts, W. Dewé, M. Maumy, and B. Boulanger, “Pre-study analytical method validation: comparison of four alternative approaches based on quality-level estimation and tolerance intervals,” Quality and Reliability Engineering International, vol. 24, no. 6, pp. 667–680, 2008.
View at: Publisher Site | Google Scholar
R. W. Mee, “β-Expectation and β-content tolerance limits for balanced one-way ANOVA random model,” Technometrics, vol. 26, no. 3, pp. 251–254, 1984.
View at: Publisher Site | Google Scholar
Z. Xue, B. Xu, C. Yang et al., “Method validation for the analysis of licorice acid in the blending process by near infrared diffuse reflectance spectroscopy,” Analytical Methods, vol. 7, no. 14, pp. 5830–5837, 2015.
View at: Publisher Site | Google Scholar
R. Galvao, M. Araujo, G. Jose, M. Pontes, E. Silva, and T. Saldanha, “A method for calibration and validation subset partitioning,” Talanta, vol. 67, no. 4, pp. 736–740, 2005.
View at: Publisher Site | Google Scholar
ICH 2011, ICH Points to Consider (R2), ICH-Endorsed Guide for ICH Q8/Q9/Q10 Implementation, 2011, http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_9_10_QAs/PtC/Quality_IWG_PtCR2_6dec2011.pdf.

Copyright

Copyright © 2019 Guolin Shi et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

1194

Downloads

888

Citations