Principles of Analysis of Solid Drug Formulations Using the Tablet Processing Workstation: A Case Study of Two Drug Candidates under Clinical Development

Sample preparation of solid dosage forms is a laborious but mandatory requirement prior to liquid chromatographic assay. The recent introduction of robotic sample prep systems such as the tablet processing workstations (TPW) has significantly increased sample throughput and overall productivity but the lack of generic approaches to automated method development still undermines the effective use of this technology.

This work leveraged existing and new sample preparation strategies to develop methods for generic use in the preparation of solid drug formulations on the TPW for end analysis by High-Performance Liquid Chromatography using CPD1201 capsules and CPD1202 tablets as case studies.

The five factors normally expected to influence sample preparation on the TPW (extraction volume, homogenisation speed, number of pulses, pulse duration, and vortex speed) were screened using a design of experiments (DoE) approach employing a resolution V factorial design to determine which of the factors were having a significant effect on the assay of content. The factional design used was a half replicate of the full five-factor factorial consisting of 16 sample runs. Three true centre points for the design were additionally used to determine repeatability and to detect any curvature in the experimental space, giving a total of 19 runs for each sample formulation. Each of the 19 sample runs was completely randomised and carried out in duplicate to obtain assay means and relative standard deviations (RSD). Statistical significance of variability, interaction, and curvature of the design space were all tested at an alpha level of 0.05 by Analysis of Variance (ANOVA) using the Design-Expert® software.

The results showed that only homogenisation speed had a statistically significant effect on the response for the CPD1201 capsule, accounting for nearly 40% of the total variability in the results. The remaining four factors were all inert. Lower homogenisation speeds resulted in lower assay results but no significant curvature was detected in the response surface. For the CPD1202 tablet, the response was found to occur in a three-dimensional factor space comprising extraction volume, homogenisation speed, and number of pulses. Both pulse duration and vortex speed were found to be inert. Extraction volume was found to account for 47% of the total variability in the results, with lower extraction volume resulting in lower assay values. A statistically significant interaction was found to occur between homogenisation speed and number of pulses, which accounted for a further 36% of the total variability. Significant curvature in the response surface was also detected. Empirical evaluation of various solid drug samples and TPW usage data led to the successful elucidation of three decision trees whose applicability was also illustrated.

The studies successfully determined the critical factors for sample prep on the TPW for the two drug formulations used. By optimising these factors and using them alongside the three proposed decision trees as tools to guide the sample prep process, the basis for generic automated method development was demonstrated. However, a further optimisation design is recommended to obtain a more detailed picture of the design space, including details of which of the significant parameters have non-linear relationships with the response and the location of the optimum. By augmentation, the additional runs required for each sample formulation could be kept to a minimum.