Integrity Review Report: Construction of an Online Database for Functional Dyspepsia (J South Med Univ, 2015)

Academic fraud report · Geng Detector

Summary

This review evaluates a 2015 database-construction paper (DOI: 10.3969/j.issn.1673-4254.2015.06.26) reporting a clinical database for functional dyspepsia (FD), authored by Huang Zaiwei, Chen Liping, and Xiao Bing of Nanfang Hospital. The work contains no typical manipulable experimental figures (no Western blots, microscopy, or electrophoresis), limiting forensic imaging opportunities to two UI screenshots (a CRF form and a web interface) and three statistics tables. Automated image forensics flagged copy-move block matches at pure vertical offsets (0,48) for the CRF image and (0,40) for the UI image, plus heterogeneous noise variance in the UI image. These flags are consistent with benign explanations: form layouts and repeating UI controls naturally generate vertically aligned repetitive blocks. Statistical review of Table 1 (n=126) found terminal-digit bias toward 0/5, but with n=126 and a uniform denominator, retaining one decimal inevitably produces this pattern. Table 2 covers only 56 of 126 FD cases without explicit justification, and Table 3 reports rounded β values and P-values of 0.000, consistent with SPSS default truncation. Overall verdict: SUSPICIOUS but not fraud. Confidence: moderate; recommendations are limited to PubPeer-style queries.

Verdict

🟡 Suspicious / Inconclusive — insufficient evidence for fraud

Key findings

F1 (Image forensics, downgraded): Copy-move block matches detected at vertical offsets (0,48) and (0,40) in img-000.jpg (CRF form) and img-001.jpg (web UI), respectively; img-001.jpg also showed heterogeneous noise variance (CV=1.214). Benign explanations (regularly repeating form/UI elements) are directly supported by the article's own figure captions, so the signals are retained as leads rather than evidence of fabrication.
F2 (Table 1 terminal-digit bias): Percentages across 28 cells (e.g., 36.5%, 63.5%, 4.8%, 50.0%, 19.8%, 15.1%, 22.2%, 15.9%, 43.7%, 28.6%, 11.8%, 33.3%, 66.7%, 82.5%, 70.6%, 10.3%, 7.2%, 22.2%, 25.4%, 23.8%, 23.0%, 26.2%, 17.5%, 37.3%, 19.0%, 38.0%, 2.4%, 10.3%) show prevalence of 0/5 terminal digits. With a uniform denominator of n=126 FD cases and one-decimal reporting, this distribution is expected. Benign explanation explicitly supported by the article ("符合纳入标准的FD病例126例").
F3 (Table 2 sample-size gap): Table 2 reports drug-efficacy data for only 56 patients (33 + 23), while Table 1 covers all 126 FD cases. The article does not state how the 56 were selected from the 126. Table 3 also reports β to only three decimal places (e.g., 0.820 for Sleeping Time, 0.734 for Tea, -0.271 for Stress, Std.Error=0.217) and P-values with inconsistent precision (0.000, 0.001 vs 0.212, 0.065, 0.896, 0.705, 0.535). The 0.000 truncation is consistent with SPSS default behavior, as the article specifies "应用SPSS对治疗效果进行分析", but degrees of freedom and t-statistics are not reported.
F4 (Methods consistency): 1.2节 states data collection "from July 2013", while 2.2节 reports 312 enrolled cases by July 2014, of which 126 met FD criteria and 186 were "其它功能性胃肠病和正常人对照组". The 126/312 = 40.4% proportion therefore represents the FD subgroup share rather than a low inclusion rate. The benign explanation is directly supported by the article.
F5 (Preprint timing): The journal received the manuscript on 2015-02-03 and published it in 2015 Vol. 35(6):907–911, but the associated ChinaXiv copy is 201712.00866v1 (December 2017), creating an inverted preprint-postprint sequence. ChinaXiv permits back-archiving of published papers, so this is not a violation, only atypical.

Evidence highlights

DOI: 10.3969/j.issn.1673-4254.2015.06.26
Journal: Journal of Southern Medical University / 南方医科大学学报, 2015, 35(6):907-911
Manuscript received: 2015-02-03; archival copy: ChinaXiv 201712.00866v1
I1: img-000.jpg copy-move at offset (0,48), 16 matched pairs
I3: img-001.jpg copy-move at offset (0,40), 21 matched pairs
I2: img-001.jpg noise-variance CV=1.214
Sample sizes: 312 total enrolled → 126 FD + 186 other functional GI / healthy controls
Table 1 percentages all denominated by 126 FD cases
Table 2 drug-efficacy N=33 and N=23 (total 56); Table 3 β examples: Sleeping Time 0.820 (P=0.000), Tea 0.734 (P=0.001), Stress -0.271 (Std.Error=0.217, P=0.212); software: SPSS per methods text

Notes

Confidence in flagged imaging signals is low; UI-element repetition is the dominant benign hypothesis and is textually supported.
Confidence in Table 1 digit-pattern flag is low; small n and uniform denominator fully account for it.
Main residual concerns are transparency issues (Table 2 selection criteria for the 56 patients; Table 3 missing t and df), not fabrication.
Recommended follow-up is PubPeer-style inquiry requesting raw data, SPSS output, and the selection criteria for Table 2; no formal institutional complaint is warranted on current evidence.
Possible OCR/parsing artifact caused Table 2 numeric tokens (8, 25, 33, 1, 22, 23) to register as "未找到" during automated cross-check; manual inspection confirmed they are present in the table.