Anil Menon

Identifying fluid-bed parameters affecting product variability

Recent statistical/graphical methodology identifies the importance of bringing the mean to target while maintaining a low variance. Fluid-bed literature abounds in optimization strategies while overlooking product variability. In the present study, active factors (location effects) related to fluid-bed granulation of lactose with povidone (PVP) were identified using normal probability plotting and basing the slope on the calculated critical pseudo standard error value. An unreplicated fractional factorial design was employed. Residual-Fit spread plots indicated adequate empirical models through least squares. Dispersion analysis using computed residuals and resulting variances indicated spray rate binder state, and certain interactions effecting particle size variability. Additional runs confirmed the feasibility of the approach in maximizing crushing strength and minimizing particle size variability. Various factor effects could be attributed to the solubility of lactose during granulation. Variability due to spray rate was ascribed to the high inlet temperature and the surface area of the fluid-bed charge exposed to the binder solution. Rapid binder film formation promoted intragranular bonds and enhanced compactibility. Using simple empirical modeling and data visualization approaches, factors influencing product characteristics and variability were identified. The applicability of this approach in robust process development is discussed.

Sequential Experimentation in Product Development. I

Investigations proceed as an iterative and recursive learning process, involving a sequence of experiments. Evaluating the results from a prior set of experiments allows the investigator to design and predict the next experimental cycle better. The investigator looks at what has gone before with a view to improve and extend the inquiry. This communication reiterates the value of sequential experimentation in product development. A fluid-bed granulation study is used to illustrate the sequential approach. Acceptable solutions were achieved using technical knowledge, a sequence of factorial designs, and graphical data analysis.

Case studies in blend sampling using a front-loading thief.

In-process testing is employed in pharmaceutical systems for monitoring and detecting process failures. Blend sampling is a useful in-process test. The present article outlines case studies in which sampling thieves with front-loading and side-vented mechanisms were compared. Blend uniformity analysis (BUA) using a front-loading sampling thief helped confirm or discover blend homogeneity issues and guide the product development process to rational decision points by providing quick responses to actual or impending failure with a low probability of falsely declaring a failure that is not present.

Effective graphical displays.

A graph is a visual display of data to achieve an understanding of the underlying patterns and interrelationships of the data. Such visual representations of data often form the basis for inferences and decisions. However, a poorly chosen graphical form can lead to erroneous inferences. In this paper, previously published graphs depicting biopharmaceutical, drug metabolism, formulation, and pharmacokinetic data are recast using alternate display methods in an attempt to better clarify the data structure. The display methods used include dot plots and trellis plots.

Communicating Tabular Data Using ORACLE

“You see but you do not observe.” Sherlock Holmes to Watson Arthur Conan Doyle A Scandal in Bohemia