PAT SUPPLEMENT


 identification of influential and critical process parameters


 identification of correlation pattern among the process parameters


 generation of process signatures  relationship between process parameters and quality attributes by multivariate regression analysis


After the successful transfer of a product from pilot plant to commercial scale, the available and achieved process understanding needs to be embedded and translated into an appropriate process control strategy. The benefits of using MVDA in this context are:  efficient on-line tool for multivariate statistical control (MSPC)


 analysis of process variability  enabling on-line early fault detection  enabler for time resolved design space verification (real time quality assurance) – Real Time Release (RTR)


 predicting quality attributes based on process data


 excellent tool for root cause, trending analysis and visualisation


This paper will give two examples from pharma - ceutical and biopharmaceutical processes.


Case Study I: MVDA used for multivariate statistical process control (MSPC) in pharmaceutical unit operation content In the first example from pharmaceutical production, MVDA is used for an on-line assessment of the process parameters in order to detect any tendency of deviating from


FIGURE 1MSPC model for a granulation process including drying, water addition, and granulation and kneading


the normal operating ranges. In Figure 1a, an MVDA model of a granulation, comprising the different phases of drying, water addition, granulation and kneading is displayed. The control limits marked in red are set as three times the standard deviation from the average, which is shown as a pale green curve in Figure 1b. In such a model, many process parameters, for instance temperatures, agitation speeds, torque and power consumption, are included in the model as shown in Figure 1a. These univariate process variables are condensed into latent variables by means of linear combination, whereas the linear coefficient reflects the impact of the specific univariate variable on the impact on the process variability. These models reflect the inherent variability


of process and are based on empirical production data. New batches can be assessed by these models in order to evaluate whether


they fall within the expected ranges and therefore yield the desired quality. These models can be used for infor-


mation only in order to enhance process understanding or for providing release relevant information. For the latter application, the methods have to be embedded into the quality system of the firm. This entails a complete qualification of the used automation and IT infrastructure as well as the validation of the methodology itself. Special effort has to be taken on defining the procedures to handle deviations and the involved function from production, engineering to quality control and assurance. Furthermore, the maintenance plan of the


MVDA monitoring system has to be defined. As an example, batches of a review period of one year are shown in Figure 2, proving that the model still covers the normal variability of the process.


FIGURE 2Assessment of recently produce batched by the model 6


European Pharmaceutical Review Volume 16 | Issue 6 | 2011


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17