Evaluating manufacturing process profile comparability with multivariate equivalence testing: Case study of...

Area of Science:

• Biotechnology
• Chemical Engineering
• Statistical Analysis

Background:

• Comparing multivariate Gaussian distributions with unequal sample sizes presents statistical challenges.
• The Generalized Mahalanobis Distance (GMD) offers a method for measuring dissimilarity between such distributions.

Purpose of the Study:

• To investigate and validate the adjusted Generalized Mahalanobis Distance (GMD) approach for testing statistical equivalence.
• To apply the adjusted GMD method to assess the equivalence of biopharmaceutical harvest profiles across different scales.

Main Methods:

• The study involved adjusting the Generalized Mahalanobis Distance (GMD) approach to control Type I error rates.
• The adjusted GMD was applied to analyze harvest profiles from a 12L small-scale model and a 2000L manufacturing scale.
• Equivalence testing was performed on critical parameters: Productivity, Lactate Production, and Total Cell Density.

Main Results:

• The adjusted GMD approach demonstrated the ability to achieve nominal Type I error rates.
• Sufficient statistical power was achieved for testing equivalence between profile populations.
• Harvest profiles from the 12L and 2000L scales were found to be practically equivalent based on the defined equivalence margin.

Conclusions:

• The adjusted Generalized Mahalanobis Distance (GMD) is a validated statistical tool for equivalence testing in bioprocessing.
• The GMD approach confirms practical equivalence of harvest profiles between small-scale and manufacturing-scale bioprocesses.
• This methodology supports robust scale-up and technology transfer in biopharmaceutical manufacturing.