Early prediction of tablet defects during pan coating

The appearance of a tablet is a critical quality attribute that can be directly evaluated by patients. During tablet manufacturing, mechanical stress may lead to visual defects, such as edge chipping, the prediction of which remains a major challenge, particularly during technology transfer or scale-up processes. This study aimed to evaluate whether the number and severity of defects could be predicted under two mechanical stress environments: pan coating and friabilator testing. A breakage model was calibrated using drop tests with two placebo formulations: mannitol:microcrystalline cellulose (7:3) and microcrystalline cellulose:anhydrous dibasic calcium phosphate (7:3), both at target tensile strengths of 1.5-2.5 MPa. Discrete element method (DEM) simulations were used to estimate the collision velocities and frequencies in the coating pan, enabling defect prediction using the breakage model. One verification run was performed using the mannitol:MCC 7:3 formulation at 2.0 MPa, which was also used to predict the tablet appearance in the friabilator. The model accurately predicted the severity of defects (Classes I-III, where I is the least and III the most severe defect) in the coating trial but underestimated the number of Class I defects. The numerical predictions made by the model are exponentially affected by inaccuracies in the calibration of model parameters, particularly for small damage. Visual differences between tablets with equal mass loss in the friabilator and pan coater also suggest that wear from low-force collisions contributed to damage during the coating process. In contrast, the number of appearance defects was overpredicted in the friabilator, highlighting the model's limitations in systems with frequent collisions. Overall, with careful calibration and under conditions with few defect-causing events, the model can provide useful guidance for defect prediction.