Alison Avenell first came across The Yamaguchi Osteoporosis Study (YOPS) when she was working on a 2014 Cochrane Review on bone fractures.
She cited the study but felt something was off about it. “I suppose, together with my collaborators over the years, we developed sort of antennae for rather suspicious looking studies,” Avenell, of the University of Aberdeen in Scotland, told Retraction Watch. “And when you see a relatively large trial with just two authors, you think to yourself, that’s not possible.”
Avenell and her colleagues, whose work we’ve written about before, were critical to the retraction of fraudulent research by the late Yoshihiro Sato and his collaborator Jun Iwamoto, who rank third and fourth, respectively, on our retraction leaderboard.
This time the study in question was called “Comparative efficacy of hormone replacement therapy, etidronate, calcitonin, alfacalcidol, and vitamin K in postmenopausal women with osteoporosis: The Yamaguchi Osteoporosis Prevention Study.” The paper, which did not include Sato as an author, had a large scope, claiming to include nearly 400 postmenopausal women and comparing multiple treatment methods. It was published in the American Journal of Medicine (AJM), an Elsevier journal, in 2004.
The study was co-authored by Yoichiro Ishida and Shinya Kawai while they were at Yamaguchi University in Japan. Since then it has been cited 83 times according to Clarivate Analytics Web of Science. It has also been cited in five clinical practice guidelines, Avenell says.
Several of Sato and Iwamoto’s retracted papers cited the study as well.
Avenell and her collaborators began digging into the study in earnest the summer of 2019. “And so when we started to look at this particular trial approach, there are many what we call red flags, about the conduct of the trial, improbable data, authorship that varied, impossible workload in terms of recruitment, lots of numbers that doesn’t add up at all, and data that were completely impossible,” she says.
Avenell and her team initially wrote to AJM about their concerns just under a year ago.
That’s when things got a bit weird.
First, Joseph Alpert, the editor of AJM, invited the team to write about their concerns in the form of a published letter to the editor, which Avenell found an “unusual approach.”
But they went along with it. The critique was published in the June, 2020 issue of the AJM along with the editor’s response. That response, from chief editor Joseph Alpert, reads, “The Editor-in-Chief and the editorial staff of The American Journal of Medicine (AJM) have carefully reviewed the information sent to us by Bolland, Grey, and Avenell from New Zealand. We have attempted without success to contact Dr. Ishida in Japan. These articles were accepted for publication by the previous editor and editorial board of the AJM at a time when all submissions were in paper form.”
Alpert explained that because of how long ago the paper was published, the journal couldn’t access the notes from the original peer review process. He continued: “The allegations of Bolland, Grey, and Avenell seem valid, and the AJM and Elsevier are currently in a final review phase prior to retracting these articles. In my personal opinion, given the documentation of Bolland, Grey, and Avenell, these articles should be considered as likely fraudulent and should not be cited in the scientific literature.”
So far, so good, right? Well, maybe.
A fact-check, first: Avenell is in fact still in Scotland, not New Zealand, as the editor implied. But more important, Avenell is concerned that the article shows up on indexing sites without context — typically without a link to her critique or the editor’s expression of concern.
“The letter from the editor at American Journal of Medicine… indicates that hopefully a retraction will happen eventually,” says Avenell, “although it’s somewhat puzzling to us that, given the information that we’ve provided it hasn’t yet happened.”
When Retraction Watch contacted Alpert to find out more, he responded, “Not my area of expertise.” That was puzzling, given that he is the editor, but he responded the same way to follow-up questions.
Retraction Watch’s emails to Ishida and Kawai produced “delivery incomplete” warnings. The researchers don’t appear to be in Yamaguchi University’s faculty database. Yamaguchi University did not reply to an email request for comment, and both authors apparently ceased publishing in 2009.
Avenell was concerned not just by the slowness of the journal’s response, but by the manner of it. In her case, she and collaborators had little to risk as whistleblowers, but she wonders how the journal would have handled a case where the investigating authors felt at risk and wished to avoid identifying themselves publicly.
“If the authors had said, I’m sorry, we’re not prepared to write a letter because… we don’t want to display our identity, particularly if it turned out that they were writing about people who knew them or even worked at the same institution,” she says.
“So, it is a rather unusual approach to adopt,” says Avenell. “And not really… following the best practice as recommended by” the Committee on Publication Ethics (COPE).
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
Hi, just wanted to note that it appears the Yamaguchi article has now been retracted https://www.amjmed.com/article/S0002-9343(04)00429-2/fulltext
A Retraction Note is finally on-line: https://www.sciencedirect.com/science/article/pii/S0002934325001524
Admirably detailed.
—————————-
This article has been retracted at the request of the Editor in Chief because of the stated concerns listed below. This article was accepted for publication by a previous editor and editorial board, at a time when submissions and documentation were in paper form, prior to the transition of The American Journal of Medicine to a digital submission and review process. These records are no longer extant and consequently we are unable to review the comments of the reviewers and editors involved at that time. We have attempted to contact the authors regarding these concerns and received no response. We are therefore retracting this article since the evidence presented below strongly argues for scientific misconduct.
The integrity of these publications is severely compromised by wide-ranging and serious concerns about governance, ethics, authorship, implausible study conduct, implausible workload, discrepant participant numbers and treatment groups, impossible data, implausible data, implausible outcome data, and discrepant methodology.
Governance
No funding source is reported for a 2-year randomized controlled trial of up to 396 participants. The work was reportedly undertaken at Tsushimi Hospital, with ethics oversight from the Tsushimi Hospital ethics committee. An internet search does not identify other research overseen by this committee.
Implausible workload
The article is co-authored by 2 investigators, only 1 of whom was affiliated to Tsushimi Hospital. To meet the timeline for the study as reported (3 years), screening and recruitment would have had to be completed within a year. This means that about 3 participants were screened each working day for a year and 1.5 recruited every working day. This workload seems beyond the capacity of 1 researcher, who presumably was also engaged in clinical work at the same time.
Discrepant participant numbers and treatment groups
Randomization to 6 groups of 66 participants is described. In (Ishida Y, Mine T, Taguchi T. Effect of progestins with different glucocorticoid activity on bone metabolism. Clin Endocrinol (Oxf). 2008;68:423-8, designated below as Y2), the 2 randomized groups that overlap with Y1 (control and HRT/CE + MPA) included only 34 and 35 participants, respectively, and one randomized group (CE + NET, n=35) is reported that was not reported in Ishida and Kawai’s paper.
Implausible and discrepant participant recruitment
Eligible participants had forearm BMD ≥30% below the young adult mean (T score ≤-3.7), or ≥1 vertebral fracture and BMD ≥20% below the young adult mean (T score ≤-2.5). Reported mean baseline BMD was 65% of young adult mean, corresponding to a T score of -4.3. Thus for 739 women screened, 54% (n=396) had osteoporosis (T score ≤-2.5) and about 25% (n∼188) must have had BMD T score ≤-4.3. It is implausible that such a high proportion of a screened population would meet these very, very low BMD criteria.
Impossible data
The age range for participants was 50-75 y but mean ± standard deviation (SD) ages of HRT, etidronate, calcitonin, and alfacalcidol groups were 70±15, 70±15, 69±17, and 71±12, respectively. These summary data cannot occur within the stated age range. As an approximation the SD should be (Max 75 – Min 50)/4 = 6.3.
Implausible and discrepant baseline data
Baseline data for age at menopause, height, and weight are implausible, e.g. mean of 48-50 y (SD 7-11 y) for age at menopause indicates a significant proportion must have had menopause 60 y (usually 2-5%). Similarly, the baseline data suggest that a substantial proportion of participants must have weighed < 40 kg and been < 140 cm tall.
The SDs for these variables are much greater than those for the participants from the same trial, reported in Y2. These data are highly unlikely to have been generated by recruiting from the same source population.
Implausible baseline data
a. Implausibly similar
The mean levels of the bone turnover markers N-telopeptide of type I collagen and deoxypyridinoline in each of the 6 randomized treatment groups are within 1 unit and 0.1 unit of each other, respectively. This is implausible, given the variability that is widely recognized for each measurement (and reflected in the large SDs for each).
Each of the 6 randomized treatment groups has exactly the same mean BMD value. Age and prevalence of vertebral fracture were the two balancing factors used in minimisation of randomization. BMD was not a minimising factor, yet it shows much less variability between randomized groups than age. These data are unlikely to have occurred by random assignment.
b. Implausibly different
Randomization was said to have been by a minimization method, balanced for age and prevalence of vertebral fractures. In Y2, the mean ages of the study groups differed by < 1 y. In Y1, however, the mean ages differed by up to 3 y. These data are implausible.
Implausible fracture outcomes
Incident vertebral fractures occurred in 26% and 21% of the participants in the control group, thus 11-13% per annum. These are extraordinarily high rates, far in excess of those reported in clinical trials in osteoporotic populations, in Japan and internationally.
60 participants experienced an incident vertebral fracture during follow-up, compared to 5 participants who experienced a non-vertebral fracture. This dramatic preponderance of vertebral fractures over non-vertebral fractures does not occur in clinical trials of osteoporosis, nor in epidemiological studies, and is highly implausible.1
To resolve these concerns, the authors should have provided documentary evidence that the trial was conducted as reported (ethics committee, participant records and consent forms), and raw data for independent review and analysis.
Numerous attempts to contact the authors, including queries to their affiliated institutions, have been unsuccessful.
Joseph S. Alpert, MD
Professor of Medicine, University of Arizona College of Medicine, Tucson, Arizona; Editor in Chief, The American Journal of Medicine
References
1 A. Avenell, A. Grey, G.D. Gamble, M.J. Bolland
Concerns About the Integrity of the Yamaguchi Osteoporosis Prevention Study (YOPS) Report
Am J Med, 117 (2004), pp. 549-555