Comparing the predictions by NIR spectroscopy based multivariate models for distillation fractions of crude oils by F-test.

The whole range of distillation fractions in industrially relevant crude oil samples is predicted by using two multivariate models based on near-infrared (NIR) spectra. The first versions of the models as well as the respective model updates are considered, with the updates largely aimed at expanding the models. The prediction results are compared across all the fractions and F-test is used to critically compare the performance of the models and the effectiveness of the limited updates. The results suggest that both multivariate methods perform very comparably, and the updates do not lead to statistically significant changes, which differs from what one could conclude from the nominal prediction errors. The near-equivalency of the prediction accuracy of the updated models is additionally illustrated by perusing predictions of a number of batches from one sour and one sweet crude arriving at the refinery during a four month period.

Raman chemical mapping of low-content active pharmaceutical ingredient formulations. III. Statistically optimized sampling and detection of polymorphic forms in tablets on stability.

Several tablets of a formulation containing 1% w/w of the desired active pharmaceutical ingredient (API) form are spiked with minimal amounts of two different anhydrous polymorphs and an amorphous form. The amount of contaminant form was 2.5 to 10% of the total API concentration (0.025 to 0.1% w/w in the tablet), with five spiked tablets prepared. The presence of these contaminant particles are then identified using Raman microscopy/mapping. The entire surface of each of these tablets is Raman-probed through a grid based on our previous proposal (Sasic, S.; Whitlock, M. Appl. Spectrosc.2008, 62, 916) about the minimal number of spectra to acquire that would guarantee identification of the targeted component (taking into account the limit of detection). All three forms have been clearly identified in the Raman mapping spectra of prepared "calibration" tablets; particularly of note is the 2.5% spike (0.025% w/w in the tablet) of the relatively weakly scattering amorphous form. The same method is then applied to packaged tablets on stability and demonstrates that none of the previously analyzed contaminant forms is detected, hence building confidence that the desired API form does not change during stability testing.

Freeze-dry microscopy: impact of nucleation temperature and excipient concentration on collapse temperature data.

The objective of this study was to investigate the impact of nucleation temperature (T(n)) and excipient concentration on the collapse temperature data obtained from freeze-dry microscopy (FDM) experiments. T(n), the temperature of the onset of collapse (T(oc)), and the full collapse temperature (T(fc)) were determined for aqueous solutions of polyvinylpyrrolidone (PVP) 40 kDa and 2-(hydroxypropyl)-beta-cyclodextrin. Concentrations were varied from 1% to 20% (w/w) for PVP and from 1% to 30% (w/w) for the 2-(hydroxypropyl)-beta-cyclodextrin. Mutual correlation coefficients were calculated for the observed T(n), T(oc), and concentrations of the solutions. In addition, outliers were detected and eliminated by applying the leaving-one-out routine and calculating correlation coefficients without it. T(n) was found to be non-correlated with concentrations and only weakly correlated with T(oc). The correlation between these two temperatures was particularly poor for the solutions of the highest and lowest concentrations. In contrast, T(oc) correlated much better with the corresponding concentrations, resulting in a quadratic fit for PVP and a linear fit for 2-(hydroxypropyl)-beta-cyclodextrin.

Fast Raman chemical imaging of tablets with non-flat surfaces.

A curved area with embossment on an analgesic tablet with three APIs is imaged with a Raman instrument equipped with hardware options that allow for continuous focus adjustment and very fast acquisition of micro-Raman spectra, about sixty spectra per second in this study. Univariate, self-modelling curve resolution, and regression derived images via the pure component spectra are obtained for all three APIs. The quality of the images and reasons for thresholding are discussed, as well as the relation between the histograms of the images and respective thresholds. The univariate image of the API of lowest concentration (caffeine) is found to be most affected by the Raman signal of the more abundant components, which is much less noticeable in the multivariate images. Particle size analysis is conducted on the thresholded caffeine images only, as the other two components are too abundant for such type of analysis. Depending on the method used for imaging, the area coverage for caffeine varies from 1% in univariate to 4% in regression images.

Two-dimensional correlation analysis of nuclear magnetic resonance metabonomics data.

Two-dimensional (2D) correlation analysis has been used in this study to identify changes in complex nuclear magnetic resonance (NMR) metabonomics spectra of rat urine samples obtained during a study in which vasculitis (vascular injury), an important safety element in preclinical trials, was induced. Two types of correlation analysis were performed, along the variables and along the samples, and both 2D covariance and correlation coefficient maps were calculated. The binned and ''raw'' NMR spectra were analyzed (0.04 and 0.001 ppm resolution, respectively). Good correlation was found among the major peaks of the binned spectra, and two groups of samples were identified using sample-sample 2D correlation maps. Much more complex correlation features were obtained from the ''raw'' spectra, in which the specific, butterfly-like patterns were obtained in the covariance map but with only a few significant correlation coefficients in the corresponding 2D correlation maps. In terms of classification, the same group of the last nine spectra that indicated the end of the process and clustered in the 2D sample-sample covariance map of the binned data was also found in the 2D sample-sample covariance map of the raw NMR spectra but, again, not in the 2D correlation coefficient map. A discussion is given on the details of the application of the correlation analysis with regard to spectral data resolution, alignment, the effect of actual intensities of the NMR signal, and reference to various results from 2D correlation analysis of vibrational spectra.

Raman mapping of low-content active-ingredient pharmaceutical formulations. part II: statistically optimized sampling for detection of less than 1% of an active pharmaceutical ingredient.

Several tablets are prepared with two forms of an active pharmaceutical ingredient (API) of which one (less than 1% w/w) is considered undesirable. The presence of this component is tested for by Raman microscopy in a series of mapping experiments. These experiments are conducted with a statistically based sampling routine in which the number of spectra to collect across the whole surface of a tablet is set so as to theoretically ensure spectral detection of the low-concentration form. Such experiments are then repeated a number of times to achieve approximately 95% confidence that the strictly limited number of sampling points suffice to detect the low-concentration form and that Raman microscopy is technically a reliable method for analytical analysis of this type.

Multivariate modeling of diffuse reflectance infrared fourier transform (DRIFT) spectra of mixtures with low-content polymorphic impurities with analysis of outliers.

Diffuse Reflectance Fourier Transform Infrared Spectroscopy (DRIFTS)-based multivariate models were developed to quantify the content of two polymorphic impurities in mixtures with the desired active pharmaceutical ingredient (API) form, with the impurities not exceeding 2% wt/wt. In addition, close attention was paid to the outlier detection criteria: Q residuals; Hotelling T2; and score bi-plot. While reasonably accurate results were obtained for the relatively simple calibration models for both forms of the impurity, the predictions for "blank" samples (separately verified to be impurity-free) were apparently biased. Thus, the model training sets were augmented with spectra from calibration mixtures incorporating some of the API from batches used in the prediction. The performance of the updated models as assessed by cross-validation was somewhat degraded as a result, while predictions against independent batches of API showed a decrease in bias indicating robustness had improved. Nevertheless, the Q residuals criterion disqualified a large number of prediction samples as outliers in contrast to the other two criteria that reported no issues at all. The results here demonstrated the effectiveness of DRIFTS for quantifying low concentration polymorphic impurities, while simultaneously highlighting the variability issues that can be encountered in practice and which need to be understood and managed appropriately to ensure the success of any automated or Good Manufacturing Practice (GMP) compliant application of multivariate modeling.

DRIFTS-based multivariate calibration and prediction of low-concentration polymorphic impurities in multiple lots of an active pharmaceutical ingredient, and outlier criteria.

Mixtures of two polymorphic impurities with one lot of the desired form of an Active Pharmaceutical Ingredient (API), mostly binary mixtures, with up to 2% wt/wt of an impurity, were used for multivariate modeling via Diffuse Reflectance Infrared Fourier Transform (DRIFT) spectra. The two obtained cross-validated models, significantly differing in accuracy, were used to predict the concentrations of these impurities in independent API lots. The predictions were found to be biased and with outliers, as revealed by the Q residuals criterion but not the other two outlier criteria. Updating the models with the spectra from the mixtures using multiple API lots produced very different calibration results: the model of the impurity with the strong IR response became noticeably worse, while the model of the impurity with less responsive IR signal changed only marginally. The updated models performed much better in the prediction as the bias for both polymorphs was reduced and the outlier-related issues mostly disappeared.

An in-depth analysis of Raman and near-infrared chemical images of common pharmaceutical tablets.

This study reports on the application of Raman and near-infrared (NIR) imaging techniques for determining the spatial distribution of all (five) components in a common type of pharmaceutical tablet manufactured in two different ways. Multivariate chemical images were produced as principal component (PC) scores, while univariate images were produced by using the most unique spectra selected by the orthogonal projection approach (OPA), a searching algorithm. Multivariate Raman images were obtained for all five components in both tablets, while only two or three components could be imaged with the NIR instrument. Very interesting PC results are reported that in effect cast doubt on the effectiveness of the established criteria for determining signal-related PCs in the Raman data. PCA has been found to be indispensable for imaging the minor components using the Raman data. Significant similarity between the multivariate and univariate chemical images has been noted despite there being considerable spectral overlap within the Raman and, especially, within the NIR mapping data sets. Gray-scale images are carefully thresholded, which allowed for quantitative comparison of the obtained binarized images. A thorough discussion is given on the problems and approximations needed for producing composite images.

Defining a strategy for chemical imaging of industrial pharmaceutical samples on Raman line-mapping and global illumination instruments.

The performance of line-mapping and global illumination Raman systems for two pharmaceutical tablets and a powder blend are assessed in this study. The chemical images were obtained from the placebo, real tablets, and powder blend by using x20, x50, and x100 objectives, as well as via the (pseudo) confocal set-up. The chemical images were produced via univariate wavenumbers and as re-folded principal component (PC) scores (known as score images). In most cases it was easy to image two or three major components of the tablets directly, while the minor components were only imaged via PC scores. The active pharmaceutical ingredients (APIs) were located relatively easily even if present in quite low concentrations (less than 1%) owing to the high Raman scattering coefficients of these materials. The strength of the Raman signal of the API makes it almost ubiquitous in the chemical images of real tablets. Thorough discussion is given on the strategies used to produce chemical images, the prospects of making composite images of all components present in the tablets, and the effects of packing density with relation to the diffusion of the excitation laser light inside the sample. The strengths and weaknesses of the Raman imaging techniques used are emphasized and suggestions are given regarding which instrument is preferable with respect to the goal of the experiment and material under study. For example, mapping technology is preferred for analyzing minor components, while the global illumination approach is recommended for imaging of spatially isolated strong Raman scatterers.

Raman spectroscopy for noninvasive glucose measurements.

We report the first successful study of the use of Raman spectroscopy for quantitative, noninvasive ("transcutaneous") measurement of blood analytes, using glucose as an example. As an initial evaluation of the ability of Raman spectroscopy to measure glucose transcutaneously, we studied 17 healthy human subjects whose blood glucose levels were elevated over a period of 2-3 h using a standard glucose tolerance test protocol. During the test, 461 Raman spectra were collected transcutaneously along with glucose reference values provided by standard capillary blood analysis. A partial least squares calibration was created from the data from each subject and validated using leave-one-out cross validation. The mean absolute errors for each subject were 7.8%+/-1.8% (mean+/-std) with R2 values of 0.83+/-0.10. We provide spectral evidence that the glucose spectrum is an important part of the calibrations by analysis of the calibration regression vectors.

Analyzing Raman maps of pharmaceutical products by sample-sample two-dimensional correlation.

Sample-sample (SS) two-dimensional (2D) correlation spectroscopy is applied in this study as a spectral selection tool to produce chemical images of real-world pharmaceutical samples consisting of two, three, and four components. The most unique spectra in a Raman mapping spectral matrix are found after analysis of the covariance matrix. (This is obtained by multiplying the original mapping data matrix by itself.) These spectra are identified by analyzing the slices of the covariance matrix at the positions where covariance values are at maxima. Chemical images are subsequently produced in a univariate fashion by visually selecting the wavenumbers in the extracted spectra that are least overlapped. The performance of SS 2D correlation is compared with principal component analysis in terms of highlighting the most prominent spectral differences across the whole data set (which typically comprises several thousand spectra) and determining the total number of species present. In addition, the selection of the unique spectra by SS 2D correlation is compared with the selection obtained by the orthogonal projection approach (OPA). Both comparisons are found to be satisfactory and demonstrate that a quite simple SS 2D correlation routine can be used for producing reliable images of unknown samples. The main benefit of using SS 2D correlation is that it is based on a few data processing commands that can be executed separately and produce results that are closely related to the chemical features of the system.

Detailed analysis of the online near-infrared spectra of pharmaceutical blend in a rotary tablet press feed frame.

The present study is an in-depth analysis of the online near-infrared (NIR) spectra acquired with a spectrometer mounted on the rotary tablet press feed frame. A 3.5% active pharmaceutical ingredient (API) formulation is analyzed. An attempt is made to determine the API univariately via the 2nd derivative spectra due to favorable appearance of the API and placebo bands in the formulation. However, the signal at the univariate API peak is ambiguous and principal component analysis is hence applied to understand better the structure of the data. To eliminate effect of the bias along the wavelength axis which is found to dominate the data, the analysis is restricted only to the spectral region that covers the API band of interest. This leads to significantly better results in terms of the univariate profile of the API now heavily overlapping with the first principal component and the elimination of the bias. Having thus proven that the univariate analysis is a viable option, an attempt is made to calibrate the API response by using some previous runs that involved alternation of the placebo and the formulation. This analysis produces mixed results due to baseline differences in the two sets of spectra. The final univariate profile of the API is therefore compared with the assay of the tablets and the two are found to agree very well. It is, therefore, concluded that the NIR probe in the feed frame can rapidly detect moderate changes in API concentration in the blend and be a good predictor of tablet potency at the same time point.

Moving window two-dimensional correlation spectroscopy and determination of signal-to-noise threshold in correlation spectra.

In this paper we report two new developments in two-dimensional (2D) correlation spectroscopy; one is the combination of the moving window concept with 2D spectroscopy to facilitate the analysis of complex data sets, and the other is the definition of the noise level in synchronous/asynchronous maps. A graphical criterion for the latter is also proposed. The combination of the moving window concept with correlation spectra allows one to split a large data matrix into smaller and simpler subsets and to analyze them instead of computing overall correlation. A three-component system that mimics a consecutive chemical reaction is used as a model for the illustration of the two ideas. Both types of correlation matrices, variable-variable and sample-sample, are analyzed, and a very good agreement between the two is met. The proposed innovations enable one to comprehend the complexity of the data to be analyzed by 2D spectroscopy and thus to avoid the risks of over-interpretation, liable to occur whenever improper caution about the number of co-existing species in the system is taken.

Raman chemical mapping of magnesium stearate delivered by a punch-face lubrication system on the surface of placebo and active tablets.

Raman chemical mapping was used to determine the distribution of magnesium stearate, a lubricant, on the surface of tablets. The lubrication was carried out via a punch-face lubrication system with different spraying rates applied on placebo and active-containing tablets. Principal component analysis was used for decomposing the matrix of Raman mapping spectra. Some of the loadings associated with minuscule variation in the data significantly overlap with the Raman spectrum of magnesium stearate in placebo tablets and allow for imaging the domains of magnesium stearate via corresponding scores. Despite the negligible variation accounted for by respective principal components, the score images seem reliable as demonstrated through thresholding the one-dimensional representation and the spectra of the hot pixels that show a weak but perceivable magnesium stearate band at 1295 cm(-1). The same approach was applied on the active formulation, but no magnesium stearate was identified, presumably due to overwhelming concentration and spectral contribution of the active pharmaceutical ingredient.

Monitoring the dissolution of Active Pharmaceutical Ingredient and TPGS in real time via IR spectroscopy during the manufacturing of liquid dosage formulation.

Infrared spectroscopy is used to monitor the dissolution of the Active Pharmaceutical Ingredient (API) and an excipient (vitamin E - TPGS) during manufacturing of a liquid pharmaceutical formulation. The goal of the analysis is to explore options for real-time, on screen, and quantitative monitoring of these two components by using an iC10 instrument. As is common, the first step in the approach is to create respective calibration models for the two components and then apply those models on the spectra obtained from scale-up batches. Interestingly, while the API dissolves at the room temperature, TPGS dissolves at an acceptable rate at 50 °C so both temperatures have to be considered. It is shown that univariate models of sufficient accuracy can be developed with a straightforward applicability to the scale-up batches spectra and providing reasonably accurate estimates of the API and TPGS concentrations. Some limitations of the software on the employed instrument may diminish the prospect for the quantitative analysis of the components of interest in this formulation.

Internal multiple-scattering hole-enhanced Raman spectroscopy: improved backscattering Fourier transform Raman sampling in pharmaceutical tablets utilizing cylindrical-conical holes.

The benefits of Raman signal enhancement and improved measurement precision are demonstrated using 180° backscattering Fourier transform Raman (FT-Raman) spectroscopy from drilled cylindrical-conical holes within pharmaceutical tablet cores. Multiple scattering of the incident laser light within the holes results in an increased Raman signal due to the larger Raman sampling volume. This is important for overcoming typical sub-sampling issues encountered when employing FT-Raman backscattering of heterogeneous pharmaceutical tablets. Hole depth and diameter were found to be important experimental parameters and were optimized to yield the greatest signal enhancement. The FT-Raman spectra collected using backscattering from cylindrical-conical holes is compared to typical 180° backscattering from flat surfaces using tablet cores of Excedrin® and Vivarin®. Raman chemical images are used to establish a representative sampling area. We observe a three- to five-fold increase in the Raman intensity and a two-fold improvement in the measurement precision when sampling from cylindrical-conical holes rather than classic backscattering from flat tablet cores. Self-absorption effects on analyte band ratios are negligible in the fingerprint region but are more significant at the higher near-infrared (NIR) absorbances found in the C-H/O-H/-N-H stretching region. The sampling technique will facilitate developing quantitative FT-Raman methods for application to pharmaceutical tablets using the fingerprint spectral region.

Multivariate statistical analysis of Raman images of a pharmaceutical tablet.

This paper describes the application of principal component analysis (PCA) and independent component analysis (ICA) to identify the reference spectra of a pharmaceutical tablet's constituent compounds from Raman spectroscopic data. The analysis shows, first with a simulated data set and then with data collected from a pharmaceutical tablet, that both PCA and ICA are able to identify most of the features present in the reference spectra of the constituent compounds. However, the results suggest that the ICA method may be more appropriate when attempting to identify unknown reference spectra from a sample. The resulting PCA and ICA models are subsequently used to estimate the relative concentrations of the constituent compounds and to produce spatial distribution images of the analyzed tablet. These images provide a visual representation of the spatial distribution of the constituent compounds throughout the tablet. Images associated with the ICA scores are found to be more informative and not as affected by measurement noise as the PCA based score images. The paper concludes with a discussion of the future work that needs to be undertaken for ICA to gain wider acceptance in the applied spectroscopy community.

Monitoring of API particle size during solid dosage form manufacturing process by chemical imaging and particle sizing.

Agglomeration of API during a solid dosage form manufacturing process is followed from the bulk API, through the initial blend with the excipients to the ribbons by a combination of chemical imaging and particle sizing experiments. Particle size of the ingoing API was characterized using a Sympatec HELOS laser diffractometer. Chemical images of the API were obtained from the blends, granules, and ribbons using near-infrared (NIR) and Raman mapping instruments. All the chemical images are obtained in the univariate fashion through the API-characteristic wavenumbers. Light microscopy and laser diffraction were used to assess presence of large agglomerates in the bulk API. NIR chemical images of the sparsely distributed blend particles confirmed that the large agglomerates were not dispersed during the blending. Also, it was found that normal microscopy may be efficient at detecting those API agglomerates due to their distinct appearance (whiteness and size). The agglomerates were not detected in the NIR chemical images of the granules and the ribbons. This was more reliably confirmed by Raman chemical images in which small API domains were clearly identified which was not attainable by NIR mapping.

Characterizing the structure of pharmaceutical granules obtained by wet granulation with varying amounts of water via Raman chemical imaging.

A pharmaceutical formulation containing metformin hydrochloride (MET), hydroxypropyl cellulose (HPC), and microcrystalline cellulose (MCC) was wet granulated with varying amounts of water and the structure of the obtained granules was characterized by Raman chemical mapping. Univariate Raman mapping was found to be satisfactory for producing the images of the two components of interest (HPC and MCC). In addition to the images, the average Raman spectra from the maps as well as the micro-Raman spectra from the hot pixels were analyzed. HPC is found to strongly respond to the addition of water, with its domain dissipating and Raman bands becoming weaker as the water addition increases. MCC is also responsive to water, reacting similarly to HPC but to a much smaller extent and only for the largest amounts of water. Granules made with increasing water content also have improved tabletting properties and flow.