Fabricated Clinical Data and Misattributed Citations in 'Exploring the Functional Role of Programmed Death-Ligand 1 (PD-L1) in the Castration-Resistant Prostate Cancer Using Transcriptomic Sequencing Analysis' (Cancer Medicine, 2025; DOI: 10.1002/cam4.71225)

Academic fraud report · Geng Detector

Summary

Verdict: Strong evidence of data fabrication and citation hallucination. The report concludes that the paper by Lin Zhong, Pengxin Zhang, Jialin Ji, Jun Mao, and Lianhong Li contains clinically impossible percentages in Table 1: the claimed CRPC subgroup (N=10) shows a metastasis rate of 38.7%, which corresponds to 3.87 individuals, and matches exactly the fraction 12/31, suggesting the data were copied from a separate N=31 cohort. The ADPC subgroup (N=68) contains fractions (88.9%, 28.6%, 14.3%, 80%, 20%) that only arise from denominators of 7, 9, or 10, making the N=68 claim mathematically untenable. Additionally, two cited references in the Introduction are clearly mismatched: citation [14] attributes work to 'Christian et al.' but the listed paper is by Perry et al. (Nature Immunology, 2022), and citation [15] attributes findings to 'Massari et al.' while the actual first author is Modena. A typographical error ('altinative' for 'alternative') and a suspiciously long-lived ethics approval number (suffix 2002-452) further reinforce the impression of low-quality or AI-assisted drafting. Confidence is high for the table inconsistencies and citation mismatches; medium for the ethics-approval concern pending institutional verification. Pixel-level image analysis was not possible because raw images were not provided.

Verdict

🔴 Confirmed misconduct — Table 1 percentages are mathematically incompatible with the stated sample sizes, and the Introduction contains demonstrably misattributed citations. These findings, combined with low-quality textual errors, strongly suggest fabricated clinical data and AI-assisted drafting without proper verification.

Key findings

Fabricated clinical data in Table 1: Percentages cannot be derived from the stated N values.
Misattributed citations in the Introduction: Cited authors do not match the listed references.
Suspicious ethics-approval timeline: Approval number suffix "2002-452" implies a ~20-year coverage of patient enrollment (2017–2024).
Low-quality textual drafting: Keyword contains "altinative splicing" instead of "alternative splicing."
Image-based checks not performed: Raw Western blot, immunofluorescence, and RNA-seq images were not available for duplication or splicing analysis.

Evidence highlights

Finding 1 — Table 1 arithmetic inconsistencies (🔴 Critical)

Location: Table 1, Page 4
Claimed CRPC subgroup: N = 10 patients
Reported metastasis rate: 38.7% → 10 × 0.387 = 3.87 patients (clinically impossible).
Equivalent fraction: 12/31 = 38.7%, indicating the figures were copied from an N = 31 cohort.
Reported hypertension history: 48.4% → 15/31 = 48.39%, again consistent with N = 31, not N = 10.
Claimed ADPC subgroup: N = 68 patients
Metastasis rate 88.9% corresponds to 8/9 or 24/27.
Smoking history 28.6% corresponds to 2/7.
Hypertension history 14.3% corresponds to 1/7.
Acinar adenocarcinoma 80% vs ductal adenocarcinoma 20% corresponds to 8/10 vs 2/10.
Interpretation: Perfect percentages that only arise from denominators of 7, 9, 10, or 31 cannot coexist with claimed sample sizes of 68 and 10. This is strong evidence of fabricated or cobbled-together clinical data.

Finding 2 — Citation hallucination in the Introduction (🔴 Critical)

Location: Introduction, Page 2, paragraph 4
Citation [14]: The text attributes findings to "Christian and colleagues" on PD-L1 expression in CRPC. The actual reference [14] is Perry et al., Nature Immunology, 2022, titled "PD-L1-PD-1 Interactions Limit Effector Regulatory T Cell Populations…" — neither by Christian nor about CRPC clinical PD-L1 expression.
Citation [15]: The text attributes a finding (≈19% PD-L1 expression in CRPC) to "Francesco Massari et al." The actual reference [15] lists Modena as first author, not Massari.
Interpretation: These mismatches are classic symptoms of AI-generated reference hallucination or careless copy-paste without verification.

Finding 3 — Ethics-approval timeline anomaly (🟠 Moderate)

Location: Materials and Methods, Section 2.5 "Patients and Tissue Sample," Page 3
Claimed enrollment period: 2017–2024 (≈7–8 years).
Approval number suffix: "2002-452" (suggesting a 2002 batch approval).
Concern: A single, ~20-year-old ethics approval covering an enrollment span of 7–8 years without documented renewal is highly unusual and warrants institutional verification.

Finding 4 — Textual / spelling errors (🟡 Low)

Location: Keywords, Page 1
"altinative splicing" is used instead of "alternative splicing."
Such a visible error in the keyword field reflects insufficient proofreading, consistent with mass-produced or AI-assisted writing.

Limitation

Pixel-level image manipulation checks (Western blot, immunofluorescence, RNA-seq) could not be performed because raw image files were not supplied.

Notes

DOI preserved as published: 10.1002/cam4.71225
All numeric values (percentages, sample sizes, citation indices, year 2025, ethics number 2002-452) are reproduced exactly as they appear in the source report.
Recommended follow-up: request raw SPSS/CSV data behind Table 1 from the authors; post the citation-mismatch evidence on PubPeer; file a formal complaint with the Cancer Medicine editorial office and with the institutional research integrity committees of Dalian Medical University and its First Affiliated Hospital.
This assessment is generated by an AI-assisted tool for academic discussion; final determination of misconduct rests with the responsible institutions and editorial investigation.