Quantitative characterisation of extended- release tablets with quetiapine using NIR- chemometric methods

This study aims to develop and validate NIR-chemometric methods for quantifying the API (quetiapine) and two excipients in extended-release tablets without sample preparation. The calibration samples were prepared following an experimental design with three variables (quetiapine, HPMC and microcrystalline cellulose) and five levels (concentration 80-90-100-110-120% of API). The validation set included three concentration levels (90-100-110%). The best calibration algorithms have used the same pre-treatment method (SNV), and different factors: 7 PLS factors (R2 -0,966 and RMSEP-6,84) for quetiapine, 8 PLS factors (R2-0,927 and RMSEP 6,84) for HPMC and 3 PLS factors (R2-0,983 and RMSEP-7,26) for microcrystalline cellulose. The methods were fully validated according to the ICH guidance using these calibration models. Regarding the trueness of the methods, the recovery was between 98.51 and 99.43 for quetiapine, between 98.61 and 100.85 for HPMC, and between 100.61 and 101.78 for microcrystalline cellulose. According to data obtained, the accuracy profile was ± 5 for quetiapine and HPMC, and ± 6 for microcrystalline cellulose. Linearity profile was also in establish intervals at accuracy and the R2 value was 0.983 for quetiapine, 0.948 for HPMC and 0.997 for microcrystalline cellulose. In conclusion, the developed NIR-chemometric methods have suitable reproducibility, accuracy, linearity and can be used for quantitative characterisation of extended-release tablets with quetiapine, with any sample preparation


INTRODUCTION
In August 2002, FDA initiated PAT Guidance for Industry, named "a framework for innovative pharmaceutical development, manufacturing and quality assurance" [1]. By implementing this new concept, the analytical procedures will be more effective by reducing time and increasing the final products' quality [2,3].
In recent years, the application of Near Infrared (NIR) spectroscopy as a Process Analytical Technology (PAT) tool and monitoring technique within the pharmaceutical industry has been overgrown. NIR spectroscopy can be used for raw material testing, product quality control and process monitoring. The growing pharmaceutical interest in NIR is probably a direct result of its advantages of being simple, fast, and non-destructive and enabling the analysis of complex matrices without the need to manipulate samples [4]. NIR spectra are rich in chemical and physical information when is used in conjunction with appropriate chemometric. The primary step to develop a NIR assay method is the calibration procedure for model development. Once calibration is successful set and favourable predictions are expected, they must be validated in order to be accepted for routine use in the pharmaceutical industry [5,6].
In hydrophile matrix extended-release formulation, the ratio of the matrix-forming polymer plays a significant role in matrix swelling and erosion and, consequently, in the API kinetic release. Inadequate mixing or segregation problems may conduct to manufacture tablets with different content in HPMC/ tablets, which may have a consequence on the reproducibility of kinetic release profile [4,7,8]. So, to have a method accessible to evaluate the content uniformity of HPMC in the hydrophilic matrix extended-release formulation is beneficial in current quality control [9]. The NIR spectroscopy in combination with chemometric may provide this tool, on intact tablets, and without any sample preparation [4,10].
This study aimed to develop and validate a NIRchemometric method for directly and simultaneously quantification of one API (quetiapine) and two excipients (hydroxypropyl methylcellulose and microcrystalline cellulose) in extended-release tablets without any sample preparation.

Extended-release tablets manufacturing
Extended-release tablets (hydrophilic type matrix) with quetiapine were prepared through wet granulation following the following protocol. Powders were weighted and sieved. Quetiapine fumarate, lactose monohydrate and microcrystalline cellulose were granulated in an Aeromatic Fielder AG fluidised bed processor (Aeromatic, Switzerland) using a water solution of polyvinylpyrrolidone 10% (m/m) as a binder. After granulation, a wet calibration was performed, and granules were dried for 24 hours in an oven. The next step was dry calibration was performed to break up crowds. The powders blend for tabletting was obtained by mixing the granules with the rest of the excipients in an Erweka LK5 laboratory kneading equipment (Erweka, Germany). At the end, tablets were obtained using Korsch EK-0 tablet press (Korsch, Germany) equipped with a set of 12 mm punch and die. The tablet press was adjusted to obtain tablets with an average weight of 640 mg. Those tablets contain 230.26 mg quetiapine fumarate (corresponding to 200 mg quetiapine) were considered the 100% target formula (corresponding to level 3 in experimental design).

Calibration and validation protocol
For validation purposes three fully independent batches (levels 2-3-4 and four replicated for each level) were manufactured daily, on three different days, resulting 36 batches for the validation set (Table 1).

Experimental design and calibration and validation samples composition
The qualitative and quantitative composition of calibration and validation samples is presented in Table 2.
For the calibration set, in order to develop NIRchemometric methods for extended-release tablets with quetiapine characteristics prediction, an orthogonal experimental design, build in Modde 12.0 software (Umetrics, Sweeden), with three variables (quetiapine quantity -X1, HPMC quantity -X2, microcrystalline cellulose quantity -X3) and five levels (80% -90% -100% -110% -120%) was used. The advantage of using design of experiments is the study of a maximum numbers of factors by means of a minimum number of experiments. The amount of quetiapine and HPMC was varied between 80-120% of the targeted quantity in tablets composition and the quantity of microcrystalline cellulose (X3) was varied according to maintain a constant tablet weight of 640mg. Practically, the X1 levels (quetiapine quantity) was variated between 28.78% (m/m) and 43.07% (m/m), X2 levels (HPMC quantity) were varied between 28% (m/m) and 41.92% and X3 (microcrystalline cellulose quantity) were varied between 9.17% (m/m) and 38.8% (m/m). Colloidal silicon dioxide and magnesium stearate were kept constant in all formulations, at a percentage of 0.5% (m/m) and 1% (m/m) respectively. According to this experimental design, the calibration set contained 27 different formulations (Table 3).

NIR spectra recording
The NIR spectra were recorded on intact tablets using an FT-NIR spectrometer Antaris II (Thermoelectron, USA) in transmittance configuration. This configuration allows directly recording NIR spectra; no prior sample preparation is needed. In this configuration, NIR passes through the tablet and is measured by an InGaAS (indium galliumarsenide) detector positioned on the tablet. In order to increase the reproducibility of measurement, a sample holder was used. Spectra acquisition was performed by OMNIC (Thermoelectron, USA) for 15 different tablets from all calibration and validation batches. For each sample, the recorded spectra were an average of 32 scans integrated over the range from 10000 to 4000 cm-1, with a resolution of 16 cm-1.

NIR data processing
The development of the multivariate models was based on partial least squares regression -PLS. Both no processing and pre-processing methods were used single or combined. The following pre-processing methods, constant offset elimination (COE), straight-line subtraction (SLS), standard normal variate (SNV), minimum, maximum normalisation (mMN), multiplicative scattering correction (MSC), first derivative (FD) and second derivative (SD), FD+SLS, FD+SNV and FD+MSC, were tested for development of calibration models.
A cross-validation approach based on the leave-oneout procedure was applied to determine the optimal number of PLS factors. The prediction capability of multivariate models was evaluated considering: a small number of PLS factors, closed to 1 value for determination coefficient (R2) and as small as possible values for root mean standard error of prediction (RMSEP) and Bias [11,12].

NIR method validation
Models with the best prediction capability were subjected to full validation. Calculated validation parameters are recommended by the International Conference of Harmonization (ICH). The validation approach was based on the technique proposed by Hubert et al. [13][14][15] and was applied to validation batches ( Table 2). The calculations were performed using Microsoft Excel 2007 (Microsoft Corporation, USA) and included trueness (relative bias and recovery), precision (repeatability and intermediate precision) and accuracy (absolute and relative tolerance limits).

Spectral investigation and models selection
The NIR transmittance spectra of the tablets corresponding to the calibration set are presented in Figure 1.

Model development and selection
Based on both un-processed and pre-processed spectra, a number of 11 multivariate models were build for each analyte (quetiapine, HPMC and microcrystalline cellulose).  Table 4. For each analyte, the model with the best predictive capacity was selected based on the lowest number of PLS factors, for that, the values of RMSEP were not significantly higher than the model with one more PLS factor. To see the variation of RMSEP with number PLS factors, the graphical plotting the value of RMSEP number vs. PLS factors is very useful. The results obtained on the models developed for quetiapine, HPMC and microcrystalline cellulose are presented in Figure 2.

API (quatiapine) content quantification
Several models were developed for quetiapine quantification in extended-release tablets, and such as, c. SLS, d. SNV, g. FD, j. FD + SNV, k. FD + MSC, seems to indicate good prediction (Table 4). Based on the decrease of RMSEP, the value of correlation coefficient R² = 0.9663, a number of 7 factors for the PLS model with pre-treatment of samples using the SNV method was selected as the model for quetiapine content quantification. At this model also has a shallow bias (0.0222), it is also observed that RMSEP values decrease sharply with the increasing number of PLS factors (Figure 2.). So, model d. SNV was selected for validation. Four independent batches at three   Figure 3A and Figure  4A. According to obtain the results, the best recovery (99.43%), the lowest bias coefficient (-0.566) and the best repeatability (0.91%) were found for the tablets with the highest quetiapine concentration (252mg / tablet). Instead, for the tablets with the smallest amount of quetiapine (204 mg /tablet), a bias coefficient of 1,486 and a recovery of 98.51% were obtained, also value very close to 100%. The method precision, that was evaluated as intra-day precision (repeatability) and inter-day precision (intermediate repeatability), has excellent value in all cases. A maximum RSD value of 1.5% was obtained at inter-day precision at the lowest concentration level of quetiapine. The linearity profile is obtained by plotting the predicted concentration in the validation samples as a function of found (introduced) concentration in the samples when there were prepared [4,5,16]. In figure 3A, the plotline of the predicted quetiapine concentrations versus introduced concentrations shows an excellent linearity profile (an R2 close to 1). In terms of accuracy, relative tolerance limits were set at ± 5%, to make the method suitable for active substance quantification in pharmaceutical formulation. According to the graphical presentation of the accuracy profile, to have good accuracy, the β-expectation tolerance limits should not exceed the acceptance limits [4,15,16]. For quetiapine assay, the results obtained for accuracy profile were in the range of ± 5% at all three concentrations levels ( Table 5 and Figure 4A), so the method is accurate for quetiapine quantification in extended-release tablets.

HPMC content quantification
HPMC was used in quetiapine extended-release formulation as matrix-forming polymer, and its quantification may be useful in routine quality control.   Figure 3B shows an excellent linearity profile for the HPMC quantification, plotting the predicted concentration vs. found (introduced) concentration of HPMC in the validation samples. The accuracy profile, graphically presented in figure 5B, show that the accuracy is very good, as the β-expectation tolerance limits do not exceed the acceptance limits, for the profile range of ± 5% at all concentrations levels.

FIGURE 3. Linearity profile of NIR -chemometric methods for quetiapine (A), HPMC (B) and microcrystalline cellulose (C) quantification in extended-release tablets. The continuous line is the identity line y = x, dotted curves represent the acceptance limits set at ±5% and dashed lines correspond to the accuracy profile
Based on the obtain validation parameters (recovery, precision and accuracy) for HPMC quantification using model d (SNV pre-treatment and 8 PLS factors), it can be concluded that the NIRchemometric method can be used for matrix-forming polymer assay in extended-release tablets with quetiapine.

Microcrystalline cellulose content quantification
For microcrystalline cellulose quantification in extended-release tablets with quetiapine, eleven models were developed ( Table 4). Several of them (c. SLS, d. SNV, e. mMN, i. MSC, , J. FD + SNV, k. FD + MSC), seems to have a good prediction capacity. Based on RMSEP decrease Figure 2.)., SNV method as pre-treatment with a number of 3 PLS factors, (model d) was selected to be tested in the validation step for microcrystalline cellulose quantification. This model also has an excellent correlation coefficient (R² = 0.9830) and a low bias (0.0069). The validation results of the method for microcrystalline cellulose quantification in extended-release tablets using the d. SNV model are presented Figure  3C, 4C and Table 5. The best recovery (100.61%), the lowest bias coefficient (0.61) was found for the tablets with the highest microcrystalline cellulose concentration (163.1 mg/tablet). The best precision (repeatability 1.28 and intermediate repeatability 1.46) was found was also front for the highest concentration of microcrystalline cellulose. A maximum RSD value of 1.94% was obtained for inter-day precision at the lowest concentration level of microcrystalline cellulose. Figure  3C shows an excellent linearity profile for the microcrystalline cellulose. quantification, plotting the predicted concentration vs. found concentration of microcrystalline cellulose.in the validation samples. The accuracy profile, graphically presented in figure 5C, show that the accuracy is acceptable, as the β-expectation tolerance limits do not exceed the acceptance limits, for profile range of ± 6% at all concentrations levels. Based on the obtain validation parameters (recovery, precision and accuracy) for microcrystalline cellulose quantification using model d (SNV pre-treatment and 3 PLS factors), it can be concluded that the NIRchemometric method can be used for microcrystalline cellulose assay in extended-release tablets with quetiapine.

CONCLUSIONS
Different calibration models were developed for quantitative characterisation of quetiapine extended-release tablets using NIR-chemometric technique. A validation procedure was followed using the best calibration models for each analyte of interest (API, HPMC used as matrix-forming polymer and microcrystalline cellulose). The method was fully validated according to ICH guidance. The results demonstrate the developed NIR-chemometric method is suitable for directly and simultaneously quantification of quetiapine, HPMC and microcrystalline cellulose in extended-release tablets. The method is reproducible, has good linearity and excellent accuracy profile, and may be used for routine analysis of the three analytes in tablets.
In conclusion, the NIR spectrometry associated with chemometry can provide suitable tools for the chemical characterisation of extended-release tablets with quetiapine without any sample preparation.