Tablet Blend Morphology - A case study for solid dosage forms

A near-infrared chemical image of a tablet blend morpholpgy shows the location of the active ingredient as well as all of the excipients. The image provides an intuitive understanding of the relationship between the components. Far more than 'pretty pictures,' chemical images reveal the extent of ingredient blending, particle size distributions, agglomeration of component particles, the presence of polymorphs, and trace contaminants. The wealth of chemical and physical information in a chemical image provides a fingerprint of the tablet's production properties…and ultimately, its performance.

Important attributes like dissolution can be radically altered by tablet blend morphology - the integrated physical and chemical associations. As a result, millions of dollars are invested in optimizing tablet blend. The keys to unlocking these secrets are contained in the tablet itself.

We have developed an analytical technique that provides complete information on tablet morphology, highlighting component interdependencies. This information can be used to solve some of the most difficult problems in solid dosage development and manufacturing. A single instrument can provide detailed comparisons with already-marketed drugs, valuable information in pre-formulation development, formulations assessment, scale-up, failure mode analysis, and OOS root cause detection. It is fast, free from sample preparation, and non-destructive. This technique is Near-infrared Chemical Imaging.

Study Parameters

In this case study, near-infrared (NIR) Chemical Imaging is used to determine the concentrations of three active pharmaceutical ingredients (APIs) in commercial Excedrin® Migraine formula. No concentration standards are required for the experiment: all values are derived directly from the tablet chemical images. The NIR spectra of the three APIs are recorded from loose powder samples and used to determine the optimum 'marker' bands.

Image acquisition time is approximately 2 minutes. Method development takes about 2 hours, and subsequent application of the method requires fewer than 10 minutes. No duplicate samples are needed, and no sample preparation is required. The technique is reagent free.

The Spectral Dimensions Chimera 2.5, long wavelength Chemical Imaging System was used for this work.

Step 1: Determine pure compound spectral 'markers'

Pure component spectra for the three pure APIs are collected by arranging powder samples side-by-side on a single sample holder. All three samples are in the field-of-view of the Chimera 2.5, and the data for all is collected simultaneously. Data acquisition time is about 2 minutes.

Spectra from each compound are converted to their 2nd derivative to remove baseline variations. Acetaminophen, shown at right in red, has a distinct band at 2042 nm. Aspirin®, shown in green, has a distinct band at 1637 nm, and the blue caffeine spectrum shows a distinct band at 2420 nm. These are the bands that maximize contrast between components - the so-called 'marker' bands.

Step 2: Observe tablet images at 'marker' bands

The chemical images shown here are all taken from a single data set and represent different views in chemical space. They provide an immediate, and intuitive, understanding of the component distributions: the tablet blend is quite heterogeneous, the domains are as large as 500 microns, although there are significantly smaller particles as well, this particle size variation exists for each API. It is even possible to make relative assessments of the quantity of each component - acetaminophen and Aspirin® appear to be present in approximately equal amounts, caffeine is present to a much lesser extent. As we proceed, we can precisely determine the quantities of API.

Acetaminophen image	Aspirin image	Caffeine image

Step 3: Compare chemical image & pure spectra

		In addition to the single wavelength images shown previously, the data set includes a complete spectrum of the material at each spatial location. In this experiment, each 30x30 micron pixel has a unique NIR spectrum associated with it. Unlike bulk NIR techniques, which provide only a superposition of the spectra of all components, each pixel in a chemical image retains the spectral features of the pure component. At left, the 2nd derivative spectrum of the pure compound (shown as a dotted line) is compared against the spectrum at a single pixel identified with this API (shown as a solid line). Note that the correlation of chemical identifying bands is incontrovertible. This is true provided the component particle size, or localized high concentration of component, is large enough that a single pixel spectra represents relatively pure material. The relationships between all three APIs can be viewed in a single RGB image by assigning each component distribution to a separate color, as shown below. Acetaminophen and caffeine, red and blue channels, show distinctive agglomerations -the green channel image for Aspirin® shows less distinct patches. The wealth of information in a single chemical image dataset acts as a definitive fingerprint of a production tablet.

Step 4: Data reduction and results

It is possible to calculate statistically the API concentrations by leveraging the tremendous amount of data in a chemical image.

First, the information is summarized using a histogram. The x-axis of the histogram represents 'brightness' in the acetaminophen chemical image, and the y-axis represents the number of pixels at any given 'brightness' level. The histogram shows a distinct population of pixels corresponding to the acetaminophen at high 'brightness' values. The number of pixels that belong to this population can be determined by fitting a Gaussian curve. The ratio of the number of pixels in this population to the number of pixels in the entire tablet provides a concentration estimate.

The same process is used to determine the concentration for caffeine. Since the Aspirin distribution is less distinct, a multivariate chemometric technique is necessary - partial least squares (PLS) is used.

Although the NIR technique used in this study is a surface technique, with penetration depths on the order of 50 microns, the statistics provided by sampling the entire tablet surface enable a remarkably accurate 'pixel-count' quantification. No concentration standards are required.