Geng Integrity Report: Analysis of Bound Caffeine Proportions in Different Teas (DOI: 10.16211/j.issn.1004-390X(n).2015.02.010)

Academic fraud report · Geng Detector

Summary

Verdict: Strong evidence (🔴 confirmed) of data fabrication and methodological negligence. The authors, led by Huang Yuewei and corresponding author Sheng Jun, report HPLC measurements of bound caffeine across 25 tea samples in this 2015 Journal of Yunnan Agricultural University article. Key issues: (1) In Table 2, 22 of 25 standard errors are identical at ±0.01, with three at ±0.03—an implausibly uniform error pattern for independent triplicate extractions across diverse tea matrices. (2) Table 3 shows the same unnatural SE clustering (mostly 0.01–0.02) and an impossible 0.63 ± 0.00 for sun-dried green maocha, indicating fabricated decimals. (3) Methods state 'weigh 20 mg/mL tea powder,' a unit mismatch revealing copy-paste or unfamiliarity with the protocol. (4) SPSS 17.0 is cited but no inferential statistics appear. Limitations: these patterns strongly suggest fabrication but cannot prove intent without raw chromatograms; alternative explanations (extreme rounding, single-batch instrument behavior) cannot be entirely excluded.

Verdict

🔴 Confirmed — strong evidence of data fabrication and serious methodological carelessness.

Key findings

Uniform SE pattern in Table 2 (22/25 samples at exactly ±0.01) is statistically implausible for independent HPLC triplicates across heterogeneous tea matrices.
Impossible SE of ±0.00 for sample #10 (晒青毛茶, 0.63 ± 0.00) in Table 3.
Methods section contains a dimensional error: "称取 20 mg/mL 的茶粉" (mg/mL is a concentration unit, not a mass).
SPSS 17.0 declared in 1.2.4 but no p-values, F-values, t-values, or ANOVA reported anywhere in the paper.
Evidence of salami-slicing: the work builds on refs [8] (low-pH precipitation method) and [9] (bound caffeine across Pu-erh fermentation stages), reusing the established pipeline across six tea categories.

Evidence highlights

Table 2: 22 samples report SE = ±0.01; only 云抗10号, 大渡岗, and 宝洪茶 report ±0.03.
Table 3: SE values clustered at 0.01 or 0.02 across a concentration range spanning 0.60% to 22.30%; sample 10 (晒青毛茶) shows 0.63 ± 0.00.
Section 1.2.3: dimensional error "20 mg/mL" confuses mass with concentration.
Section 1.2.4: SPSS 17.0 declared; only Mean ± SE shown with no inferential tests.
DOI: 10.16211/j.issn.1004-390X(n).2015.02.010.

Notes

The 22/25 identical-SE pattern is the strongest single indicator; however, alternative explanations such as aggressive rounding to two significant figures or a single shared instrument calibration artifact cannot be fully excluded without raw data.
The ±0.00 SE is mechanically inconsistent with a non-trivial average and is difficult to explain by rounding.
The dimensional error and unused SPSS citation suggest substantial portions of the methods were copied without verification.
Recommended follow-up: request original HPLC chromatograms, raw integration areas, and laboratory notebooks for all 25 samples; flag Table 2 and Table 3 on PubPeer; request editorial data audit by the Journal of Yunnan Agricultural University.