BACKGROUND: Granule characteristics are some of the important intermediate qualities that determine tablet properties. However, the relationships between granule and tablet characteristics are poorly understood. The aim of this study was to elucidate relationships among formulation factors, granule characteristics, and tablet properties using a non-linear response surface method (RSM) incorporating a thin-plate spline interpolation (RSM-S) and a Bayesian network (BN). METHOD: Tablets containing lactose (Lac), cornstarch (CS), and microcrystalline cellulose (MCC) were prepared by wet granulation. Ten formulations were prepared by an extreme vertices design. The angle of repose (Y1), compressibility (Y2), cohesion force (Y3), internal friction angle (Y4), and mean particle size (Y5) were measured as granule characteristics. Tensile strength (TS) and disintegration time (DT) were measured as tablet properties. RESULTS: RSM-S results showed that TS increased with increasing amounts of MCC and Lac. DT decreased with increasing amounts of MCC and CS. The optimal BN models were predicted using four evaluation indices -Y3 was shown to be the most important factor for TS, whereas Y2, Y3, and Y4 were relatively important for predicting DT. Moreover, tablets with excellent tablet properties (i.e. high TS and low DT) were produced by relatively high Y1, low Y2, high Y3, high Y4, and middle Y5 values, and resulted from the middle of MCC, middle-to-low CS, low Lac, and middle-to-low magnesium stearate (Mg-St) amounts. CONCLUSION: The RSM-S and BN techniques are useful for revealing complex relationships among formulation factors, granule characteristics, and tablet properties.
BACKGROUND: Granule characteristics are some of the important intermediate qualities that determine tablet properties. However, the relationships between granule and tablet characteristics are poorly understood. The aim of this study was to elucidate relationships among formulation factors, granule characteristics, and tablet properties using a non-linear response surface method (RSM) incorporating a thin-plate spline interpolation (RSM-S) and a Bayesian network (BN). METHOD: Tablets containing lactose (Lac), cornstarch (CS), and microcrystalline cellulose (MCC) were prepared by wet granulation. Ten formulations were prepared by an extreme vertices design. The angle of repose (Y1), compressibility (Y2), cohesion force (Y3), internal friction angle (Y4), and mean particle size (Y5) were measured as granule characteristics. Tensile strength (TS) and disintegration time (DT) were measured as tablet properties. RESULTS: RSM-S results showed that TS increased with increasing amounts of MCC and Lac. DT decreased with increasing amounts of MCC and CS. The optimal BN models were predicted using four evaluation indices -Y3 was shown to be the most important factor for TS, whereas Y2, Y3, and Y4 were relatively important for predicting DT. Moreover, tablets with excellent tablet properties (i.e. high TS and low DT) were produced by relatively high Y1, low Y2, high Y3, high Y4, and middle Y5 values, and resulted from the middle of MCC, middle-to-low CS, low Lac, and middle-to-low magnesium stearate (Mg-St) amounts. CONCLUSION: The RSM-S and BN techniques are useful for revealing complex relationships among formulation factors, granule characteristics, and tablet properties.