Category: Manufacturing and Bioprocessing
Purpose: Evaluate blend uniformity using an in-line near infrared (NIR) spectrometer and a qualitative moving block F-test method
Methods: Two separate direct compression blends each containing a single active pharmaceutical ingredient (API) were studied. The APIs in both blends were present in similar concentrations. A PharmaTech MB100 blender equipped with 60 L and 100 L intermediate bulk containers (IBCs) and SentroPAT BU II spectrometer (Sentronic) were used in this study.
NIR spectra were taken in absorbance mode with a 1350-1800 nm spectral range, 1 nm spectral resolution, 25 mm spot size, 4 scans per spectrum, and 1 spectrum per bin revolution. For inline measurements, the spectrometer was mounted to a sapphire sight glass located on the IBC cone. For offline measurements, the spectrometer was used separately.
Second derivative Savitzky-Golay preprocessing (17 smoothing points, 2nd order polynomial) was performed to remove physical attributes from all spectra. NIR spectra for each raw material in each blend were obtained offline to identify unique wavelengths for the two APIs and to confirm method specificity. A spectral range of 1585-1785 nm was chosen for the moving block F-test analysis due to the presence of unique API peaks in this range.
Block sizes of 46 spectra for API 1 blend and 46-69 spectra for API 2 blend were used, respectively. The variable block size for API 2 blend was required due to differences in the IBC volumes and corresponding blend masses. Principal component analysis was used to sum variance within each block, thus allowing variance between blocks to be compared using the F-test. The critical F-value of 1.66 (46 spectra block size) and 1.49 (69 spectra block size) was calculated from the F distribution table selecting 95% statistical confidence. After the critical F-value was reached, a lag time of one block was chosen to derisk false positive results. Spectra were analyzed using Unscrambler X v. 10.1 (Camo Analytics).
Results: Analysis of four batches for each API showed distinct mixing patterns for both APIs with the blending end points (critical F-value plus lag time) reached at 557, 596, 621, and 672 revolutions for API 1 and 435, 492, 518, 589 revolutions for API 2, respectively. The larger range for API 2 was attributed to the different blend masses and IBCs used in the manufacture of the batches as well as the larger lag time for batches blended in the 100 L IBCs. Only one blend mass and IBC were used for blends containing API 1.
Conclusion: The moving block F-test showed distinct mixing patterns during the blending process for both APIs. Blending endpoints of 557-672 revolutions and 435-589 revolutions were determined for APIs 1 and 2, respectively. The data suggest that NIR detection combined with application of the moving F-test method can be applied to reliably assess the blending endpoint of direct blends.
Adam Worrall– Wilmington, Ohio