Data Fabrication and Statistical Impossibilities in a Rat Study of Yucang Granules on Tripterygium Glycoside–Induced Premature Ovarian Insufficiency

Academic fraud report · Geng Detector

Summary

This forensic review of a 2020 paper in Global Chinese Medicine (doi:10.3969/j.issn.1674-1749.2020.02.004) reports multiple severe data-integrity failures. Findings 1 and 3 demonstrate statistical impossibilities: the authors claim one-way ANOVA but report Z-values (Z=10.774, 10.402, 13.272), which is methodologically self-contradictory because ANOVA yields F-statistics. More damning, with n=10 per group a non-parametric Z-value cannot mathematically exceed ~3.78, so the reported values are physically impossible. Recalculation of the ER comparison (0.2621±0.0767 vs 0.1998±0.0683) yields a t≈1.94 and P≈0.06–0.07, contradicting the claimed P<0.05. Finding 2 documents cloned standard deviations (both 0.059) across two independent treatment arms in Table 3, an outcome incompatible with real biological variance. Image analysis (Findings 4) was not possible due to lack of raw files. Confidence is high for the statistical and SD-cloning findings; image-level concerns remain pending. Overall verdict: confirmed substantive data fabrication.

Verdict

🔴 Confirmed (实锤). The paper exhibits systematic numeric fabrication and a fundamental misunderstanding of inferential statistics, supporting a determination of research misconduct. Image-level duplication could not be assessed.

Key findings

Method-vs-statistic contradiction: Methods state one-way ANOVA (which produces F-values), yet Table 1 and Table 2 footnotes report Z-values (10.774, 10.402, 13.272).
Mathematically impossible Z-values: With n=10 per group, the theoretical maximum of a two-sample non-parametric Z is approximately 3.78; reported values are unattainable from real data.
Fabricated significance: Recomputing the ER comparison (0.2621±0.0767 vs 0.1998±0.0683, n=10) gives t≈1.94, P≈0.06–0.07, which contradicts the paper's claim of P<0.05.
Cloned standard deviations: In Table 3, the Yucang granule group reports PKA 0.221±0.059 and the clomiphene group reports 0.325±0.059; identical SDs to three decimals across independent interventions are a hallmark of copy-paste fabrication.
Images unassessed: Figures 1–3 (IHC and Western blot) could not be checked because raw image files were not provided.

Evidence highlights

DOI: 10.3969/j.issn.1674-1749.2020.02.004
Table 1 footnotes cite Z=10.774 and Z=10.402; Table 2 footnotes cite Z=13.272.
Table 2 ER values: treatment 0.2621±0.0767 (n=10) vs control 0.1998±0.0683 (n=10); recomputed t≈1.94, P≈0.06–0.07.
Table 3 PKA values: 0.221±0.059 vs 0.325±0.059 — identical SDs (0.059) across independent arms.
n=10 per group establishes the ~3.78 Z-value ceiling for two-sample non-parametric tests.

Notes

Confidence is high for the statistical and SD-cloning findings because they rely solely on the numbers the authors themselves printed.
Western blot and IHC figures should be requested from the authors (un-cropped original blots and raw image files) for follow-up image forensics.
An independent statistical re-analysis of all tables by the journal is recommended.