Multivariate Data Analysis to Assess Process Evolution and Systematic Root Causes Investigation in Tablet Manufacturing at an Industrial Scale-A Case Study Focused on Improving Tablet Hardness.

Background/Objectives: Only a few studies performed at industrial scale in non-simulated conditions have investigated the effect of input variability from the product's lifecycle on product quality. The purpose of this work was to identify the root causes for the low and variable hardness of core tablets prepared using high-shear wet granulation through batch statistical modeling and to verify the short- and long-term effectiveness of the improvement actions. Methods: The novelty of this study is the use of multivariate methods for the complex assessment of a wide data set belonging to two proportional composition strengths, manufactured at an industrial scale, with different tablet shapes and sizes, with the aim of identifying inter-related active ingredient and process variables with the highest impact on hardness value and for defining optimal processing conditions leading to a robust product. Results: Four main variables affecting the output variable were identified: API particle size, nozzle type used for granulation, wet discharge, and drying intensity. These were included in an updated control strategy (3 out of 4 variables having to be within the desired ranges: API d0.5 < 45 microns; granulation nozzle that ensures liquid dispersion into droplets; gentle wet discharge and drying processes). In the case of the product studied, the newly defined process conditions could even accommodate d0.5 up to 70 microns and still ensure adequate core tablet hardness (at least 30% above the lower specification limit) for the successive film-coating step. Conclusions: Besides the beneficial impact of reducing the risk for out-of-specification hardness results, this study also offered the benefit of cost avoidance and yield improvement. The improvement was confirmed through the significant average hardness increase (15-20%) and between-batch variability decrease, leading to decent sigma quality levels (2.5) for the control phase batches.