The tabletting machine as an analytical instrument: consequences of uncertainties in punch force and punch separation data on some parameters describing the course of the tabletting process

The influence of the uncertainties involved in the measurement of punch forces and punch separation in an eccentric tableting machine on the validity of the analytical results was evaluated using six direct compression excipients. The analytical parameters considered were the maximum upper punch pressure, the minimum punch separation, the maximum relative density, the contact time, the area quotient according to Emschermann and Müller, the Weibull and Heckel parameters, as well as the total, expansion and apparent net work. The measuring uncertainties were divided into between-run deviations (BD) and within-run deviations (WD), which are constant and variable, respectively, during a tableting event. Both types of uncertainties were expressed as simple error limits. The effects of the measuring BD's were calculated by adding them to the force and displacement data and then computing the analytical parameters in the conventional way. The estimation of the effects of the measuring WD's needs special methods for each parameter. Their validation showed that they were in most cases able to include the true effects of some exemplary selected errors but tend to overestimate them. From the sum of the confidence interval (CI) of a mean parameter value from repeated tableting experiments, the confidence interval with respect to curve fitting, the BD and the WD of the analytical results, the total deviation (TD) of the results was obtained, which provides a worst case measure of the uncertainties. The TD makes a differentiation between materials with similar tableting behaviour impossible in many cases, thus providing too low a selectivity. The best case uncertainties account for the difference in the response of the data to be compared to the measuring errors. The uncertainty decreases considerably under best case conditions. However, the best case intervals predicted from the worst case limits are not generally valid. Thus, besides the TD, only the CI's remain for the assessment of the analytical results. However, in the presence of systematic errors the statistical analysis cannot assure the correctness of the conclusions drawn with the degree of certainty supposed, even if the systematic measuring errors are the same for the data to be compared.