(W1130-10-57) Assessing Topical Dosage Form Manufacturing Methods and Their Impact on Globule Microstructure with Non-Invasive Micro-CT Analysis

Purpose: In emulsion-based creams systems, the quantification of dispersed globules becomes a critical quality attribute when globules are correlated to the location of the active pharmaceutical ingredient. Drug stability and performance are affected by globule microstructure, making it important to study the dispersed phase during stages of product development and scale-up. In this project, two samples were produced with different manufacturing methods. Non-invasive X-ray microscopy (XRM) combined with artificial intelligence (AI)-based image analysis was used to assess the globule structure in 3D with minimal sample preparation.  Analysis of light microscopy data was also performed to compare with XRM and understand the impact sample preparation can have on the delicate cream microstructure.

Methods: Two oil-in-water emulsion cream samples were prepared for this study. Sample A was produced using the direct conventional method that combined all ingredients at high temperature. Sample B had similar composition but was produced with a cosmetic industrial technique that created lamellar crystalline structure in the matrix and a ring along the globule edges. The lamellar network was to allow for emollient to slip into the spaces provided by the crystals in the last step. For light microscopy, the samples were prepared in thin slides and imaged at 20x magnification with the Nikon Eclipse Ci-L instrument. Deep learning algorithms segmented the globules, matrix, and the shadow in sample A. The globules were characterized via size and roundness distributions. For X-ray microscopy, a Zeiss Versa 520 was used to obtain the 3D image volumes. The samples were loaded inside 2 mL plastic tubes and mounted inside the instrument. Images were obtained through filtered back projection reconstruction of 1601 radiographic projections over an angular sweep of 180 degrees. The final images had a pixel resolution of 1 µm/voxel and 2 µm/voxel for Sample A and B respectively. Image volumes were analyzed with digiM I2S software for image processing, AI-based segmentation, and quantification. Machine learning algorithms segmented phases of the globule, matrix, large air bubbles, air-in-matrix, and air-in-globule. Globule characterization included size distributions, sphericity analysis, spatial distributions, and autocorrelation function.

Results: Three phases were segmented in the conventional light microscopy images: globules, matrix, and shadow (Figure 1f). The shadow phase represented darker pixels caused by the diffraction of light at the edges of the globules. Size distribution revealed an average globule size of 9 µm and roundness analysis showed highly circular globules with an average roundness of 0.93 (Figures 2a,b). XRM similarly captured the globule and matrix phase, but further phases such as air-in-matrix and air-in-globule were elucidated for both samples. Sample B also classified air-in-globule into outer air-in-globule and inner air-in-globule. The separation of the air-in-globule phases indicated the presence of the lamellar ring produced by the manufacturing method. In addition, the globule volume fraction matched the formulation weight percentage. The globule size distributions, sphericity analysis, spatial distributions and inter-globule spacing were quantified from the XRM images. Sample B had globules with an average size of 40 µm, while Sample A had an average size of 60 µm (Figure 2a). Both samples exhibited a decrease in globule sphericity as size increased, and Sample A showed less spherical globules compared to Sample B overall (Figure 2b-d). Both samples showed a homogenous distribution of globule sizes spread throughout the image data volume indicating good intra-sample globule uniformity irrespective of manufacturing method (Figure 3a,b). From autocorrelation function, an average spacing of 6 µm between globules for sample B and 8 µm for sample A was observed (Figure 3c). The technique used to produce sample B created smaller and rounder globules that were spaced closer together than in Sample A. Compared to light microscopy, globules quantified by XRM were larger and less round. Light microscopy can only capture a single cross section of each globule due to its 2D nature, thus potentially obscuring the globules’ true structure. In addition, the globules could become altered during thin slide preparation whereby the cream was manually smeared onto the slide then pressed with a cover slip. In contrast, XRM was both non-invasive and 3D which offered additional characterization parameters as well as little to no disturbance from sample preparation needed.

Conclusion: In this work, X-ray microscopy was shown to be able to directly and quantitatively assess the impact of manufacturing on globule structure in cream-based formulations. Sample B had smaller, rounder, and more closely packed together globules than sample A. The globule differences were attributed to the different production techniques. In addition, analyzed light microscopy data from sample A had significantly smaller and rounder globules than those from XRM. However, XRM offered a better comprehensive characterization of globules due to its 3D non-invasive nature which introduced less bias from imaging and sample preparation. In the future, more cream samples will be characterized with different formulation parameters to study the influence of ingredients on the globule microstructure and corresponding drug performance.

Figure 1: a) A representative slice from X-ray microscopy scan at 1 µm resolution on the sample A cream sample. b) The same representative slice of sample A overlaid with the segmentation. c) A representative slice from X-ray microscopy scan at 2 µm resolution on the sample B cream sample. d) The same representative slice in sample B overlaid with the segmentation. e) light microscopy image of sample A. f) Segmented version of the light microscopy image of Sample A. All segmented phases are color coded according to legend to the right of the preview.

Figure 2: a) Globule size distribution including light microscopy and XRM data. b) Globule roundness distribution including light microscopy and XRM data. c) Globule sphericity vs size for sample A XRM dataset. d) Globule sphericity vs size for sample B XRM dataset.

Figure 3: a) Axial spatial distribution vs size for sample A XRM dataset. b) Axial spatial distribution for sample B XRM dataset. c) Autocorrelation function for the XRM datasets of samples A and B.