Detection of data manipulation in bioequivalence trials

Area of Science:

• Pharmacokinetics and Pharmaceutical Sciences
• Regulatory Science
• Biostatistics

Background:

• Pharmaceutical companies may manipulate bioequivalence trial data for non-approvable formulations.
• Methods include interim analysis, re-analysis with new subject aliases, and altering test/reference products or dilutions.
• This manipulation artificially forces failing products towards bioequivalence and is undetectable by standard pharmacopoeial methods.

Purpose of the Study:

• To demonstrate methods for detecting signals of fraudulent study conduct in bioequivalence trials.
• To introduce and describe the "Buster" and "SaToWIB" routines used by regulators.

Main Methods:

• The "Buster" routines visualize trends using partial statistics, residual plots, and cumulative confidence intervals.
• Runs tests on the sign of residuals are proposed as a potential manipulation test.
• The "SaToWIB" routines rank profile pairs based on numerical similarity using an objective function.

Main Results:

• The "Buster" routines can detect trends indicative of data manipulation.
• The rank determined by the "SaToWIB" score indicates fraud, with actual fraud cases showing higher ranks.
• European Union regulators have publicly questioned trial validity based on pharmacokinetic profile similarity since 2020.

Conclusions:

• The "Buster" and "SaToWIB" routines offer effective methods for detecting fraudulent bioequivalence trial data.
• There is a need for formal statistical tests to address manipulation, especially considering multiplicity issues.