The NIR instrument used in the study was XDS MasterLab (FOSS NIRSystems, Laurel, MD), which was capable of automatically measuring
multiple tablets after they were positioned in a special tray (see Figure 1). In the insert, a tablet tray is to the left,
and to the right is the standards tray containing the wavelength standard (traceable to NIST SRM-2035) for photometric standards.
The universal tablet tray used for this study had 20 positions for four different tablet sizes and five positions for the
0.25-in. diameter tablets under test. The tray was loaded twice to scan all 10 tablets. The 10 tablets were scanned in less
than 5 min, taking a reference spectrum before scanning each set of five tablets. Spectra were collected in the transmission
mode from 800 nm to 1650 nm with 0.5-nm data intervals, and 32 scans were coadded to produce a single spectrum. HPLC analysis
was run on each individual calibration and validation tablet after spectra were collected with the NIR instrument. The HPLC
reference values and the NIR spectra were used to develop the regression model.
Figure 2 shows the raw NIR spectra from the calibration set and a spectrum of pure CPM in green. The pure CPM spectrum was
scanned in reflectance and multiplied by a scaling factor to superimpose it over the transmission calibration spectra. By
taking the second derivative of the spectra, as shown in Figure 3, the baseline was normalized and the spectral features were
enhanced so that the fanning out of the analytical region for CPM was observed at 1138 nm. Figure 4 shows the expanded analytical
band demonstrating the linear response from 0.1 mg to 2.0 mg CPM. Smoothing was done on the derivative with a segment of 10
and a gap of zero. A thickness correction was applied as a math pretreatment to correct for tablet thickness and density variance
over the region of 1250–1350 nm. The raw spectral variance of the calibration set is large over this region, and the pure
CPM has an absorption minimum as seen in Figure 2. Thickness correction is a normalization function offered in Vision software
(Foss NIRSystems, Inc.) as a spectral mathematical pretreatment used to correct for path length variance. The integral correction
factor was calculated over the range of 1250–1350 nm with a unity-scaling factor.
Integral correction factor =
The correction factor equals the 0.5-nm increment at which the data were collected times the sum of the y-axis values (S'i) and the y-axis value plus 1 (S'i+1), divided by 2. Then, the original spectrum was divided by this correction factor throughout the analytical region of 1120–1380