henhaobio
Sign in

bioRxivCancer BiologyPreprint

Suspected distortion of citations in high-impact cancer journals

Baptiste ScancariD, Jennifer Anne ByrneiD, David CauseuriD, Adrian Gerard BarnettiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Research and scholarship are shaped by article citations, which underpin the communication of ideas, assignment of credit, journal impact factors, and author career progression. Given their key influence on author and journal metrics, citations can be intentionally manipulated to inflate the reputation of journals and researchers. Paper mills, unethical organisations that produce and sell manuscripts and publishing services, may also be manipulating citations, but the extent of this manipulation is unknown. Here, we show that molecular cancer articles sharing features with retracted papers from paper mills display citation patterns that suggest systematic inflation. These articles were published in journals in the top decile of journal rankings. Suspected paper mill articles received 50 to 100% more citations than other papers 1 to 3 years after publication, while paradoxically attracting fewer readers and online accesses. Suspected paper mill articles also cited - and were cited by - other suspected paper mill articles, which frequently appeared in journals previously reported as paper mill targets. Citations from suspected paper mill articles measurably inflated journal citation metrics. These findings suggest that paper mills inflate the citation metrics of supported publications and affected journals at scale. This widespread manipulation of citation metrics may amplify unreliable findings, slowing scientific progress and spurring unreasonable citation benchmarks for research articles, journals and authors. Our findings point to measurable signatures of citation manipulation that could support article- and journal-level screening.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (16 Sept 2026) · CC BY 4.0

Molecular oncology papers that textually resemble retracted paper mill papers receive 50 to 100% more citations yet fewer readers, cite one another, and measurably inflate some journals' citation rates.

Paper mills have been linked to citation cartels and to high-impact molecular cancer journals, which raises concern that they inflate citations and impact factors. This work measures the effect by applying the authors' earlier text-based detector to 33,159 papers in 20 top molecular oncology journals, plus 2.4 million citing and 1.5 million referenced papers.

Key results

  1. ShownWithin journal and year, flagged index papers get substantially more citations than non-flagged papers but fewer Mendeley readers and online accesses. This split persists after adjusting for Chinese authorship; among non-coding RNA (ncRNA) papers, they still have higher citations and generally lower attention.

    Negative binomial rate ratios: citation counts 50 to 100% higher at both thresholds (0.6 and 0.9), largest early (over 100% at 0.9) and later about 50%. Mendeley readers and online accesses were about 10 to 20% lower. Dimensions.ai gave near-identical ratios; in n = 3,524 ncRNA papers, citations were still higher and attention generally lower. n = 33,159 (4,085 flagged vs 29,074 non-flagged at 0.6). · Flagged papers receive more citations but lower…; Statistical models and sensitivity analyses

  2. ShownFlagged papers are much more likely than non-flagged papers to cite, and be cited by, other flagged papers. This clustering holds after adjusting for Chinese authorship, restricting to the 20 study journals, and adding a research-topic covariate.

    Mixed-effects logistic regression on 2,389,626 citing and 1,538,819 referenced papers: odds ratio (OR) 6.5 (95% CI 6.3 to 6.8) for flagged citing papers and 10.1 (9.7 to 10.6) for flagged referenced papers. With China adjustment: 5.7 and 7.9. Within the 20 journals: 9.5 and 14.3. With the topic covariate: 4.5 to 5.7 and 7.1 to 9.3. Gini, flagged vs non-flagged: 0.36 vs 0.19 (citations), 0.35 vs 0.28 (references). · Flagged papers cite and are cited by other…; Statistical models and sensitivity analyses

  3. ShownFlagged papers can materially raise journal citation metrics. In two journals they account for more than half of all citations, and removing them lowers the 2-year citation rate (2YCR, the authors' impact-factor analogue) of the most affected journals.

    Share of total citations at threshold 0.6: 57% for Molecular Cancer and for Journal of Experimental & Clinical Cancer Research (JECCR), 32% for Cancer Letters, 17% for Oncogene. 2YCR contribution (2014 to 2023): JECCR 10 to 17% (peak 21% in 2018); Molecular Cancer 6 to 13% (peak 16% in 2016). Confidence intervals were sometimes wide; contributions declined from 2021 to 2023. · Flagged papers inflate citation metrics

  4. ShownFlagged papers are common in top molecular oncology journals but concentrated in a subset of titles, overrepresented among authors at Chinese institutions, and more often flagged by independent integrity sources.

    Flagged: 4,085 of 33,159 index papers (12.3%) at threshold 0.6 and 933 (2.8%) at 0.9. Present in 19 of 20 journals; at least 20% of papers flagged in half of the top ten. Chinese institutions: 70% of flagged vs 21% of all papers. PubPeer comments 18.8% vs 5.4% (integrity concerns 17.0% vs 4.5%), against a non-flagged sample of 4,088 papers. Problematic Paper Screener: 30% vs 8%. · Flagged papers cluster within leading molecular…; Screening for suspected paper mill papers

  5. SuggestedThe citation advantage builds early, mostly from other flagged papers, and flows through a reciprocal set of journals, some with documented integrity problems. The authors interpret this as organised citation exchange.

    At threshold 0.9, over 75% of first-three-year citations came from flagged papers (about 65% at 0.6), falling toward 50% after 10 to 12 years; for non-flagged papers the figure stayed at 10 to 20%. Reviews supplied nearly 40% of flagged papers' citations in 2024 (non-flagged: 32 to 37%). Chi-squared residuals placed Biomedicine & Pharmacotherapy and Bioengineered among cite-and-cited journals (p < 0.001). · Flagged papers cite and are cited by other…

Methods, data and reproducibility

Applied the authors' earlier BERT classifier (trained on 2,202 retracted paper mill papers' titles and abstracts) to 33,159 papers from 20 top-decile molecular oncology journals (2012 to 2023) and to their citing and referenced papers. Compared citations, readership and citation flows with regression models, and recomputed 2-year citation rates without flagged papers.

  • Index papers: original molecular cancer research articles (2012 to 2023) from 20 top-decile SCImago oncology journals, taken from PubMed and selected with a BERT classifier · n = 33,159
  • Citing papers of the index papers (OpenAlex) · n = 2,389,626
  • Referenced papers of the index papers (OpenAlex) · n = 1,538,819
  • Retracted paper mill papers from the Retraction Watch Database used to train the detector (earlier work) · n = 2,202
  • PubPeer comments for flagged papers and a proportional non-flagged comparison sample · n = 4,085 flagged; 4,088 non-flagged
  • Mendeley readership via Elsevier PlumX API · n = 33,149
  • Online accesses collected from publisher websites (unavailable for some journals) · n = 25,196
  • Dimensions.ai citation counts used in a sensitivity analysis
Methods & tools
BERT paper mill detection model (authors' earlier work; accuracy 0.91, sensitivity 0.87, specificity 0.96), BERT molecular cancer research classifier (accuracy 0.96), OpenAlex API, Elsevier PlumX API, Dimensions.ai, PubPeer, Problematic Paper Screener, Retraction Watch Database, SCImago Journal and Country Rank, PubMed, R ggplot2, R gt
Reproducibility gaps
The fine-tuned paper mill detection models and the code from the earlier paper are not released, to prevent evasion of fraud detection. Mendeley readership must be obtained from Elsevier's PlumX. Online-access data were collected from publisher websites and are missing for some journals.
Limitations (6)
  • The detector is a statistical text classifier that will misclassify some papers and gives no direct evidence of misconduct. Citation motives, and therefore paper mill intent, cannot be determined. · stated by the authors
  • Training labels come from Retraction Watch's staff-assigned paper mill category and from retracted papers only. This adds label noise, circularity toward journals with high retraction rates, uneven coverage over time, and blind spots for mill papers that are never retracted. · stated by the authors
  • Topic homophily is not ruled out: sub-classifying molecular oncology papers is difficult, and the topic covariate built from clustered embeddings is imperfect. · stated by the authors
  • The analysis covers only top-decile molecular oncology journals; patterns may differ in lower-impact venues or other fields. · stated by the authors
  • Some patterns may reflect China's research environment (home-bias citation) rather than paper mills, and lower Mendeley uptake in China may undercount readers. Chinese authorship was adjusted for, however. · stated by the authors
  • Every flag (index, citing and referenced papers) comes from one classifier whose models and code are withheld, so the core labels cannot be independently regenerated. · noted in this summary

Read this if you work on research integrity, bibliometrics or citation-based evaluation and want quantitative evidence of suspected paper mill citation manipulation in high-impact cancer journals, plus candidate screening signals.

Citations
4

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISCandal-Pedr '20Arabi '23Scancar '26Pavlovic '20Wren '20Zheng '23
65% semantic matchScientific Communication & Education
C. Candal-Pedreira, A. Ruano-Ravina, E. Fernandez+3 more·bioRxiv·Aug 11, 2020