mariuscomper.ukRO

The patients who never existed

Retracted medical studies that are still being cited

Since 1980, journals have retracted 2,731 papers that Retraction Watch classes as clinical studies, almost all of them done on patients: some for invented patients and data, others for manufactured papers, missing ethics approval or honest error. Retraction does not make them disappear. 11,142 times, a later paper cited one of them at least a full year after it was withdrawn, and 1,528 of those citations came from systematic reviews, meta-analyses or guidelines, the papers that treatment advice is meant to rest on.

Retractions from the Retraction Watch database (via Crossref), citations from Europe PMC, up to October 2026.

  • one study, from publication to retraction
  • retracted for fabricated or falsified data
  • a later paper that cited it
  • a review or guideline that cited it

Each diagonal trail of dots is one study, still cited year after year. Citations from the first twelve months after a retraction are left out, so papers already in press when it happened are not counted. Reviews and guidelines here means systematic reviews, meta-analyses and guidelines.

How long a study stands before it is retracted

Half of these studies were retracted within 21 months of publication. The rest could stay in the medical record for a long time: 411 stood in the literature for ten years or more before the journal withdrew them, the arcs that span the top of the picture.

The oldest stood for four decades. A 1983 study of a stimulant drug in children with intellectual disabilities, co-written by the psychologist Stephen Breuning, was retracted in 2023. Breuning had been found to have fabricated research and was convicted in the late 1980s; the journal acted more than three decades later.

Retractions also come in waves. In 2023, one publisher alone withdrew 644 studies on patients from journals that paper mills had infiltrated, the dense arcs that land in 2023 on the right of the picture.

Why studies on patients are retracted

Retraction Watch tags each notice with one or more reasons. To count each study once, it is placed under the most serious reason it carries, in the order below. “Fabricated” is used only when Retraction Watch tags the notice for invented or altered data, results or images.

  1. Fabricated or falsified data291Data, results or images made up or altered.
  2. Manufactured paper or peer-review problems795Paper mills, machine-generated text, fake, compromised or questioned peer review.
  3. Misconduct found262An official or journal finding of misconduct, without a specific fabrication label.
  4. No ethics approval or consent126The study ran without ethics approval or without patients' informed consent.
  5. Unreliable or missing data506Results the authors or journal could no longer stand behind, or data that could not be produced.
  6. Error, plagiarism or duplication722Honest mistakes, copied text, the same study published twice, and other reasons.
  7. Retracted and replaced29Withdrawn and republished in corrected form. Left out of the citations after retraction.

Only 291 of the 2,731 carry an explicit fabrication label; many more were pulled because nobody could check the data. 29 were retracted and replaced with a corrected version, like a PREDIMED analysis of the Mediterranean diet, withdrawn after randomisation errors were found at some sites (the trial's main paper was republished in 2018 and its main result held). Those are left out of every count of citations after retraction.

The patients who never existed

Between 1991 and 2011, the Japanese anaesthetist Yoshitaka Fujii published well over a hundred trials, most of them on drugs against nausea after surgery. In March 2012 the British anaesthetist John Carlisle published a test of 168 of them: in 141 studies on people, the reported weights, ages and heights were far more alike than real randomisation allows: the chance of genuine trials producing them was less than 1 in 1033. Then a committee of the Japanese Society of Anesthesiologists examined 212 of Fujii's papers and found fabricated data in 172, 126 of them randomised trials. The numbers of patients, it reported, were “totally different from the institutional records”. Most of his co-authors had not known; some of their signatures had been forged.

The 127 studies of Fujii's that are retracted in this dataset describe about 11,700 patients in their abstracts. Each dot below is one of them; 108 of the studies carry an explicit fabrication tag.

One dot per patient described in the abstract, one block per retracted study, oldest first. In amber: the 108 studies that Retraction Watch tags for fabricated or falsified data. Tap a block to open the study.

Fujii's work was not marginal. His trials were part of the evidence on how to prevent nausea after anaesthesia, and the papers that cite them did not stop when they were retracted: 223 citations came at least a year later.

The German anaesthetist Joachim Boldt is the other large case among studies on patients: 176 retracted here. Nearly half (82 of 176) carry the tag for missing ethics approval; a hospital inquiry in 2012 found that none of the 91 studies it reviewed had an opinion from the regional ethics committee, and false data in at least ten. Boldt studied hydroxyethyl starch, given by drip to restore blood volume. When a 2013 meta-analysis in JAMA removed his seven trials, the data on the remaining 10,290 patients showed that starch was linked to more deaths (risk ratio 1.09) and more kidney failure. His retractions were still arriving in 2026, fifteen years after the scandal broke.

The bone researcher Yoshihiro Sato, whose trials on hip fractures and osteoporosis were investigated by Hirosaki University in 2017, accounts for 58 more. A 2019 study followed 68 reviews, guidelines and trials that cited twelve of his trials: by 2018, only one had reassessed its findings.

After retraction: who keeps citing

Europe PMC links 46,418 citations to these studies. 11,142 of them appeared at least twelve months after the retraction, spread over 1,519 different retracted studies.

A citation is not always a mistake. A review can name a retracted trial in order to exclude it, and a paper about research fraud cites it on purpose. So wherever a citing review or guideline has full text in PubMed Central, every in-text citation of the retracted study was located and its paragraph checked automatically; then a sample was read one by one.

1,528 citations in systematic reviews, meta-analyses and guidelines, by what the citing text does
What the citing text doesCitations
No warning near the citation380
Mentions the retraction150
Lists it as excluded19
In the reference list only22
Reference not found in full text1
Full text not readable automatically285
No full text in PubMed Central671

A random sample of 40 citations marked “no warning” was then read one by one. In 36, the retracted study was used as ordinary evidence, with nothing in the paragraph or the reference to say it had been withdrawn; two left it out of an analysis for unrelated reasons and two could not be judged (one of them because of a reference-matching error, since fixed). None mentioned the retraction.

These are lower bounds. Europe PMC only knows reference lists it has read, and many guidelines are not in PubMed Central: for them, the citation is counted but the sentence could not be checked.

The sentences

Each pair shows the retraction and, years later, the sentence that cited the study. The citation is highlighted; nothing else has been changed.

Retracted · 2009

Combination treatment of angiotensin-II receptor blocker and angiotensin-converting-enzyme inhibitor in non-diabetic renal disease (COOPERATE): a randomised controlled trial

Lancet · 2003

Retracted by The Lancet after an investigation by the authors’ institution: no patient consent, no institutional approval, results that could not be reproduced.

Cited · 2022 · No warning near the citation

RSSDI Guidelines for the management of hypertension in patients with diabetes mellitus

Int J Diabetes Dev Ctries · 2022

A combination of ACEi and ARBs reduces urinary albumin excretion compared to monotherapy; however, they are associated with a further risk to the kidney and hence are not recommended [152].

COOPERATE reported that combining two blood-pressure drugs, an ACE inhibitor and an ARB, protected the kidneys. This guideline advises against the combination, but the only reference for the whole sentence, including the claim that the combination lowers protein in the urine, is the retracted trial.

PMC9750845

Retracted · 2016

Effect of Vitamin and Trace-element Supplementation on Immune Responses and Infection in Elderly Subjects

Lancet · 1992

Retracted by The Lancet in 2016, 23 years after publication, after an investigation: unreliable data and results, ethical violations by the author.

Cited · 2020 · No warning near the citation

Nutritional recommendations for CoVID-19 quarantine

Eur J Clin Nutr · 2020

Anti-oxidants increase the number of T-cell subsets, enhance lymphocyte response to mitogen, increased interleukin-2 production, potentiated natural killer cell activity, and increased response to influenza virus vaccine compared with placebo [9].

The 1992 trial reported that vitamin and trace-element supplements improved immune responses and reduced infections in older people. In 2020 it was cited as evidence in nutritional advice for the COVID-19 lockdown.

PMC7155155

Retracted · 2009

The Analgesic Efficacy of Celecoxib, Pregabalin, and Their Combination for Spinal Fusion Surgery

Anesthesia & Analgesia · 2006

Retracted in 2009 after an official investigation found fabricated data. Its author, Scott Reuben, was sentenced to six months in prison.

Cited · 2012 · No warning near the citation

Guidelines for the management of postoperative pain after total knee arthroplasty

Knee Surg Relat Res · 2012

Preemptive use of pain control drugs (opioids21,22), non-steroidal anti-inflammatory drugs [NSAIDs], cyclo-oxygenase [COX]-2 inhibitors23,24), and pregablin25-28)) or pain management procedures (regional anesthesia29,30) and peripheral nerve block31)) can be effective in reducing central & peripheral sensitization and postoperative pain level, which results in decreased consumption of the analgesic agents after surgery8-10,19,32).

Cited, three and a half years after the retraction, in a guideline on pain relief after knee replacement.

PMC3526756

Retracted · 2013

CPAP for the Metabolic Syndrome in Patients with Obstructive Sleep Apnea

NEJM: The New England Journal of Medicine · 2011

Retracted by the New England Journal of Medicine because errors were found in the data and the original records could not be located to check the corrections.

Cited · 2016 · No warning near the citation

Society of Anesthesia and Sleep Medicine Guidelines on Preoperative Screening and Assessment of Adult Patients With Obstructive Sleep Apnea

Anesth Analg · 2016

As OSA is associated with numerous poor health outcomes including risk of death, cardiovascular events, and average number of days hospitalized,188–190 and treatment of OSA appears to improve these outcomes,191,192 it is clinically important to diagnose OSA in those patients screened at high risk, and for these patients to have ongoing, long-term management.

PMC4956681

Retracted · 2013

Valsartan in a Japanese Population with Hypertension and Other Cardiovascular Disease (Jikei Heart Study): A Randomised, Open-label, Blinded Endpoint Morbidity-mortality Study

Lancet · 2007

Retracted by The Lancet in 2013 after an investigation found serious problems with the data.

Cited · 2022 · Mentions the retraction

2022 ACC/AHA guideline for the diagnosis and management of aortic disease: A report of the American Heart Association/American College of Cardiology Joint Committee on Clinical Practice Guidelines

Circulation (text from the J Thorac Cardiovasc Surg co-publication) · 2022

In the Jikei Heart Study,12 which supported the use of ARBs in the 2010 ACC/AHA thoracic aortic disease guidelines, Japanese patients on an antihypertensive drug regimen that included valsartan had a lower rate of adverse cardiovascular events, including mortality and, in particular, a reduction was showed in the incidence of aortic dissection. However, this study was subsequently retracted13 and, consequently, the LOE for use of ARBs has been downgraded to C from B.

How it should look. The guideline says what the trial had supported, that it was retracted, and that the level of evidence was lowered as a result. (Published in Circulation in November 2022; this is the 2023 co-publication.)

PMC10784847

Retracted · 2013

Infliximab in patients with primary Sjögren's syndrome: a pilot study

Arthritis and Rheumatism · 2001

Retracted in 2013, together with its one-year follow-up, for errors in methods and results found by an investigation.

Cited · 2024 · Mentions the retraction

British Society for Rheumatology guideline on management of adult and juvenile onset Sjögren disease

Rheumatology (Oxford) · 2024

Infliximab was initially reported as being beneficial on the basis of two open label studies [246, 247] but both of these apparently positive studies were subsequently retracted because of evidence that methodological errors had led to the wrong conclusions [248].

PMC12013823

Britain

66 retracted studies on patients list an institution in the United Kingdom. The one cited most after its retraction is British: Andrew Wakefield's 1998 Lancet paper on the MMR vaccine, retracted in 2010 (Retraction Watch lists falsified data and missing ethics approval among the reasons). It has been cited 663 times at least a year after its retraction, but most of those papers have titles about vaccines, autism, misinformation or fraud, and many cite it as a cautionary example. A count of citations is not a count of mistakes, which is why the citing text was checked.

The most recent British case is ADVOCATE, the trial of avacopan for ANCA vasculitis, retracted by the New England Journal of Medicine in June 2026: the primary-outcome assessments of nine patients had been re-adjudicated after the data were unblinded, without the knowledge of the academic leads, who asked for the retraction. The European Medicines Agency's medicines committee recommended revoking the drug's licence. ADVOCATE had been cited 753 times before its retraction; any citation from 29 June 2027 on will count here as one made after it.

Done well, it takes one sentence. When the British Society for Rheumatology published its guideline on Sjögren disease in 2024, it cited two early trials of infliximab and said, in the same sentence, that both had been retracted.

How to check a study before you trust it

Search the title in PubMed: a retracted paper carries a red “Retracted article” banner at the top. The Retraction Watch database covers everything else, including the reason.

Zotero flags retracted papers in a library automatically; EndNote can if its retraction alerts are switched on. A PDF saved years ago will not change by itself.

If a study you were shown has been retracted, the treatment it was about may still be sound. Ask your doctor what else the advice rests on.

How this was counted

The list of retractions is the Crossref–Retraction Watch dataset (downloaded 31 October 2026, 73,519 records, retractions up to October 2026). A “study on patients” is a record tagged Clinical Study and marked as a retraction: 2,731 records, 2,591 with a PubMed number. The label is used as given; PubMed indexes 47 of them as studies on animals only. Expressions of concern (352) and reinstated papers (9) are left out. 1,082 are randomised trials according to PubMed or their title.

Topics come from PubMed's subject headings (MeSH) and, for studies without them, from words in the title; one study can belong to several topics.

Citations come from Europe PMC (the CITES index). A citation counts as after retraction when the citing paper first appeared at least twelve months after the retraction notice (when Europe PMC gives only a year, two calendar years later). Systematic reviews, meta-analyses and guidelines are recognised by publication type or title; papers whose titles are about retraction or fraud are set aside. Where one had full text in Europe PMC or PubMed Central, every in-text citation of the retracted study was found in the XML and its paragraph checked for words such as retracted, withdrawn, fabricated or excluded.

The number of patients is the first count of people stated in each abstract, read by a pattern; a check of 30 random abstracts found two where a group size was read instead of the total, and five that could not be confirmed. It is shown per study and summed only for Fujii's studies.

This page names researchers only where an official investigation or the retraction notices themselves give the reason, and uses their wording. Reviewed for errors 31 October 2026; to be updated when the dataset is.

Sources

  • Crossref and Retraction Watch, Retraction Watch database (open data) gitlab.com
  • Europe PMC REST API: citations and open full text europepmc.org
  • NCBI E-utilities: PubMed records and PMC full text ncbi.nlm.nih.gov
  • Retraction Watch: report finds Fujii faked data in at least 172 papers (2012) retractionwatch.com
  • Carlisle JB. The analysis of 168 randomised controlled trials to test data integrity. Anaesthesia 2012 doi.org
  • Retraction Watch: Boldt inquiry finds false data in at least 10 studies (2012) retractionwatch.com
  • Zarychanski R et al. Hydroxyethyl starch and mortality: systematic review and meta-analysis. JAMA 2013 pubmed.ncbi.nlm.nih.gov
  • Retraction Watch: Hirosaki University investigation of Yoshihiro Sato (2017) retractionwatch.com
  • Avenell A et al. An investigation into the impact and implications of published papers from retracted research. BMJ Open 2019 pubmed.ncbi.nlm.nih.gov
  • Xu C et al. Retracted trials in evidence syntheses (VITALITY). BMJ 2025 doi.org
  • European Medicines Agency: recommendation to revoke Tavneos (avacopan), June 2026 ema.europa.eu
  • Isselbacher EM et al. 2022 ACC/AHA Guideline for the Diagnosis and Management of Aortic Disease pmc.ncbi.nlm.nih.gov

How many medical studies on patients have been retracted?

The Crossref–Retraction Watch dataset lists 2,731 retracted studies on patients, up to October 2026; 1,082 of them are randomised trials.

Do doctors still cite retracted studies?

Yes. Later papers cited these studies 11,142 times at least a year after their retraction, 1,528 times in systematic reviews, meta-analyses or guidelines. Some cite them to exclude them; many cite them as evidence.

Does a retracted study mean a medicine does not work?

No. A treatment usually rests on many studies. A retraction removes one of them; ask a doctor or pharmacist what the rest show before changing anything.

How can I tell if a study is retracted?

Search its title in PubMed, where retracted papers carry a red banner, or in the Retraction Watch database.