Evaluation of Different Estimation Methods for Accuracy and Precision in Biological Assay Validation

Area of Science:

• Biostatistics
• Pharmacometrics
• Analytical Chemistry

Background:

• Biological assays (bioassays) are vital for estimating substance potency and monitoring biopharmaceutical products.
• Key performance metrics include relative accuracy (bias) and precision, with existing guidelines (USP <1033>) offering unclear definitions and methods.
• The utility of data transformations, like log-transformation, for intermediate precision estimation remains ambiguous, complicating method selection for analysts.

Purpose of the Study:

• To provide rigorous definitions for bias and precision in bioassay analysis.
• To introduce and evaluate three alternative methods for calculating relative standard deviation (RSD).
• To compare the performance of proposed RSD methods against the USP <1033> standard, particularly concerning bias and root-mean-square error (RMSE).

Main Methods:

• Development of precise definitions for bioassay bias and precision.
• Implementation and comparison of three novel RSD calculation methods.
• Statistical analysis to evaluate bias and RMSE across different methods and data distributions.

Main Results:

• All evaluated methods performed similarly when relative standard deviation (RSD) was ≤10%.
• The USP <1033> method exhibited larger bias and RMSE compared to the proposed methods for high-variation data.
• Estimates based on the original data scale performed well for moderately skewed data; log-transformation is suitable for noticeably skewed data.

Conclusions:

• The USP <1033> method for RSD calculation is not recommended for routine bioassay analysis due to potential bias and RMSE issues.
• The original scale method is preferred for estimating precision in bioassays with moderate skewness.
• Log-transformation offers a viable alternative for precision estimation with noticeably skewed bioassay data.