Fraud Detection Sensitivity in Simulated HbA1c Trial Data by Sample Size

A simulation study by Philippe P. Hujoel, last updated in April 2026, examines the sensitivity of a 3-sigma lnCVR statistical test for detecting simulated fraud in clinical trial data. The 5.5 KB Excel file contains results where the worst 50–90% of HbA1c scores in an intervention arm were replaced by a best-responder value. The analysis varies the sample size per trial arm to assess the test's detection power.

Use Cases

• Benchmarking fraud detection methods based on simulated HbA1c score manipulation.
• Analyzing the relationship between sample size and statistical test sensitivity for data integrity checks.
• Designing clinical trial monitoring protocols based on simulated worst-case data replacement scenarios.

Strengths

• Dataset is openly licensed under CC-BY-4.0.
• File size is small (5.5 KB), ensuring quick download and inspection.
• The simulation parameters (50–90% score replacement, sample size variation) are explicitly defined in the description.

Limitations

• Row count and column-level documentation are absent; field semantics must be inferred after download.
• The data is simulated, not from real-world trials, which may limit generalizability to actual fraud cases.

Provenance

Source

figshare, authored by Philippe P. Hujoel.

Collection Method

Likely generated via statistical simulation, as described.

Freshness

Last updated 2026-04-15 17:45:10; freshness should be verified.