Diagnosing fraudulent baseline data in clinical trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Research Integrity

Background:

• The initial table in randomized clinical trials (RCTs) typically presents baseline characteristics across treatment arms.
• Deviations from expected random variation in these baseline comparisons can raise suspicions of data manipulation.
• A notable case involved accusations and confirmation of data falsification stemming from such baseline inconsistencies.

Purpose of the Study:

• To theoretically confirm findings from simulation analyses regarding the difficulty of proving data inconsistency in RCT baseline tables.
• To propose a practical diagnostic method for identifying potential data falsification in RCTs.
• To enhance the reliability and integrity of clinical trial reporting.

Main Methods:

• Theoretical confirmation of simulation analysis results.
• Analysis of the challenges in detecting baseline data inconsistencies without covariate correlation information.
• Development of a diagnostic approach to flag potential issues in baseline tables.

Main Results:

• Confirmed that proving inconsistency with chance in baseline covariates is statistically challenging without knowledge of inter-covariate correlations.
• Demonstrated the theoretical basis for why such inconsistencies are hard to detect definitively.
• Proposed a diagnostic criterion to identify potentially problematic baseline comparisons.

Conclusions:

• Detecting data falsification based solely on baseline table inconsistencies is difficult due to statistical complexities.
• The proposed diagnostic offers a pragmatic approach to flag suspicious findings for further scrutiny.
• Emphasizing rigorous checks on baseline data is crucial for maintaining the integrity of clinical trial evidence.