(M1030-08-55) Correlating Tablet Performance Differences to Tablet Microstructure Via MFV-SEM and XRM Imaging Coupled with AI Image Analytics

Purpose: Amorphous solid dispersions (ASDs) are a prominent formulation approach in oral solid dosage forms of small molecule active pharmaceutical ingredients (APIs) with poor solubility. A common approach to manufacturing ASDs is through spray drying which yields a powder with controlled particle size of homogeneously dispersed API molecules in a polymeric matrix. If the powder has a good flowability i.e., large particle size and high bulk density, the ASD powder can be directly compressed (DC) with excipients to form tablets. In case of small spray dried particles and low bulk density the powder has poor flowability and an additional granulation step, such as roller compaction (RC) followed by milling, is often required to aggregate the spray dried particles and improve processability. DC and RC processes can produce tablets with distinct differences in microstructures, which can lead to large variations in performance attributes such as disintegration time and dissolution rate. Decoding such variation is further complicated by additional parameters within each method and formulation, such as RC compaction forces, ASD powder volume fractions, and API loading in ASD. Conventional characterization techniques are challenged by lack of sensitivity and difficulties in deconvoluting the impact from multiple parameters. Employing innovative microstructure imaging and artificial intelligence (AI)-based quantification tools, this study sought to correlate observed variations in tablet disintegration with tablet microstructure, in particular tablet micro-porosity.

Methods: ASD composed of API and HPMC-AS were manufactured by spray drying. Ten lots of tablets were manufactured from the 3 ASD powders. Each lot featured a unique combination of tableting method, API to excipient ratio, processing conditions, and coating. Finally, each lot underwent disintegration study. Full tablet three-dimensional x-ray microscopy (3D XRM) was performed on a DC and RC tablet for gross tablet morphology characterization of both tableting concepts. High-resolution 3D XRM was performed on the center of five tablets. Micro-porosity was further studied with 200nm resolution mosaic field of view scanning electron microscopy (MFV-SEM) on all ten samples. Artificial intelligence (AI)-based analyses of MFV-SEM data was performed for phase separation of tablet solid and micro-porosity. Porosity and pore size distribution was then calculated for each sample. AI image analysis was then performed on the high-resolution 3D XRM data to reconstruct tablet solid and porosity. Porosity, pore size distribution, and permeability was calculated for each high-resolution XRM dataset. Finally, investigation into correlations between image-based porosity quantification data and formulation/processing/performance data was performed.

Results: Correlative imaging with 3D XRM and MFV-SEM revealed several microstructure characteristics correlating with observed differences in disintegration among the samples studied. Qualitative analysis of full tablet XRM revealed varied homogeneity among comparable DC and RC tablets. Qualitative analysis of corresponding MFV images further revealed the number of spray dried particles that survived compaction greatly varied among otherwise comparable DC and RC tablets. Quantitative porosity analysis of all MFV-SEM datasets revealed a strong correlation between pore volume fraction and disintegration time. When disintegration time was plotted against calculated porosity, disintegration time was seen to have a strong inverse correlation to pore volume fraction. Permeability analysis of XRM data also revealed a trend of increased permeability as sample porosity increased.

Conclusion: High-resolution correlative imaging with XRM and MFV-SEM, coupled with advanced AI image processing and computational tools, elucidated microstructure differences in samples prepared by different tableting methods/parameters. Microstructure differences were correlated with observed differences in disintegration. Porosity and permeability were computed from the images and correlated with disintegration and dissolution between DC and RC samples. The impact of different granulation forces in RC process was investigated.

Figure 1. Plot of sample disintegration time versus calculated MFV-SEM pore area percentage. Data point annotation is differentiated by color; blue corresponds to DC samples, red corresponds to low force RC, and yellow corresponds to target force RC samples.