Category: Manufacturing and Bioprocessing
Purpose: In this research, simulation model predicting tablet strength was established. Tablet strength was evaluated by impact tester (BIT-ST, Hitachi Technologies and Services, Ltd.). Correlation between tablet strength and tablet density distribution was evaluated by both actual data and simulation data. Actual tablet density distribution was evaluated by NIR chemical imaging (Compovision, Sumitomo Electric Industries, Ltd.). Discrete element method (DEM) was used to evaluate tablet density distribution in computer simulation.
Methods: Impact tester was used to evaluate tablet strength. 6 kinds of tablet shapes and 3 kinds of tablet densities were used to evaluate the correlation between tablet strength and tablet density distribution. Drop test was repeatedly performed 5 times in each tablets with the fall height of the probe raised by 5 mm at a time. The minimum value of the fall height at which all the tablets became defective tablets was recorded. Actual drop height was calculated by the following equation to cancel the effect of the prove weight.
Actual drop height (cm)=Minimum fall height (cm) × (Tablet weight (mg) + Probe weight (mg))/Tablet weight (mg)
Tablet density distribution was evaluated by NIR chemical imaging. A calibration model predicting tablet density was constructed by employing flat-faced tablets to measure tablet bulk density accurately. 56 measurements were conducted for constructing calibration model including various tablet density, tablet thickness, and tablet angle to cancel the artifact related to various tablet shapes. Tablet angle were changed by jig to lean the flat-faced tablets.
Calibration model was constructed by Unscrambler X (Ver. 10.3, S.T. Japan Inc.). 56 kinds of NIR spectra were used for constructing calibration and validation model. 1510-1900 nm regions of NIR spectra were used for calibration model. Partial least squares (PLS) were conducted to construct the calibration model. Tablet density distribution of 6 kinds of tablet shapes were evaluated by calibrated model.
To simulate the tablet density distribution and predict the tablet strength, DEM simulation software (R-flow) was used. Compression in simulation was conducted by moving lower punch and total contact force was used to evaluate tablet density. Total contact force was calculated as below.
Total contact force in each particles (N) = Σkxi
Where k is spring coefficient, xi is distance between each particles, and i is number of particles whose distance is less than particle diameter.
To evaluate the correlation between tablet strength and density distribution by NIR chemical imaging and DEM simulation, statistical analysis software (JMP Ver. 6.0.0, SAS Institute Inc.) was used. As input parameters, average tablet density (or average contact force) in each area and tablet bulk density were used. As an output parameter, actual drop height was used. Step wise analysis was used to eliminate the parameters not correlating with tablet strength.
Results: Tablet density prediction model was constructed from the spectra of flat-faced tablets. R2 of calibration model was 0.965 and R2 of validation was 0.940. These results mean prediction model that can cancel the artifact related to various tablet shapes.
Results of tablet density distribution were shown in Figure 1 (NIR chemical imaging) and Figure 2 (DEM simulation). Tablets strength results showed lower bulk density had higher cracking tendency (lower actual drop height). Tablets having round cap or higher cap-depth (like No. 1 and No. 4) had higher cracking tendency. Flat tablets (No. 5) also had the cracking tendency because edge side of the tablets tended to make the crack.
Results of the strength prediction model by DEM simulation and NIR chemical imaging were shown in Figure 3. These analysis results showed both NIR chemical imaging and DEM simulation could predict tablet strength.
Conclusion: In this research, tablet strength and tablet density distribution were evaluated by NIR chemical imaging and DEM simulation. It was found that NIR chemical imaging was useful to evaluate tablet density distribution. DEM simulation was also used to evaluate the tablet density distribution. Tablet strength had a correlation with tablet density distribution by NIR chemical imaging and DEM. R2 value of prediction model by DEM simulation was larger than that of NIR chemical imaging because simulation had less artifact like variation of filling process, particle size distribution, and contents of excipients. Simulation tool established in this research could be useful to predict tableting process because of their simplicity and less artifact.