A group of cancer researchers from Tunisia has been seeding the oncology literature with plagiarized articles that steal liberally — both text and data — from the work of others.
The group has one retraction, in the journal Obesity — whose splash page has the jaunty, if disconcerting, invite: “Welcome to Obesity!” — and at least two withdrawn papers. However, we have been alerted to at least one other case of apparent plagiarism involving an article in the Annals of Saudi Medicine that ought to receive careful scrutiny.
The M.O. of the group — there’s some ambiguity about who is involved, which we’ll get to in a moment — appears to be quite simple: Find a study that looks easy to “replicate,” change a few of the particulars and submit as if it were a piece of local, original work.
For example, in May 2010, the researchers published an article in Obesity titled,”Effect of obesity at the pathologic response to neoadjuvant chemotherapy among premenopausal Tunisian women with breast cancer.” Here’s the abstract from that paper:
In this article, we evaluated BMI and response to neoadjuvant chemotherapy (NC) in premenopausal Tunisian women with operable breast cancer. From May 2006 to July 2009, 800 patients were diagnosed and received NC from CHU Farhat Hached (Sousse, Tunisia). Patients were categorized as obese (BMI >/=30 kg/m(2)), overweight (25 </= BMI < 30 kg/m(2)), or normal/underweight (BMI <25 kg/m(2)). Logistic regression was used to examine associations between BMI and pathologic complete response (pCR). Breast cancer-specific, progression-free, and overall survival times were examined using the Kaplan-Meier method and Cox proportional hazards regression analysis. Median age was 42 years; 27% of patients were obese, 25% were overweight, and 48% were normal or underweight. In the univariate model, there was a significant difference in pCR to NC for obese compared with normal/underweight patients. In multivariate analysis, there was no significant difference in pCR for obese compared to normal weight patients. Overweight and the combination of overweight and obese patients were significantly less likely to have a pCR (odds ratio (OR) = 0.59; 95% confidence interval (CI), 0.37-0.95; and OR = 0.67; 95% CI, 0.45-0.99, respectively). Higher BMI was associated with worse pCR to NC. So, its association with worse overall survival suggests that greater attention should be focused on this risk factor to optimize the care of breast cancer patients.
Compare that with this abstract from a 2008 article in the Journal of Clinical Oncology by a team from MD Anderson Cancer Center:
PURPOSE:
To understand the mechanism through which obesity in breast cancer patients is associated with poorer outcome, we evaluated body mass index (BMI) and response to neoadjuvant chemotherapy (NC) in women with operable breast cancer.
PATIENTS AND METHODS:
From May 1990 to July 2004, 1,169 patients were diagnosed with invasive breast cancer at M. D. Anderson Cancer Center and received NC before surgery. Patients were categorized as obese (BMI >or= 30 kg/m(2)), overweight (BMI of 25 to < 30 kg/m(2)), or normal/underweight (BMI < 25 kg/m(2)). Logistic regression was used to examine associations between BMI and pathologic complete response (pCR). Breast cancer-specific, progression-free, and overall survival times were examined using the Kaplan-Meier method and Cox proportional hazards regression analysis. All statistical tests were two-sided.
RESULTS:
Median age was 50 years; 30% of patients were obese, 32% were overweight, and 38% were normal or underweight. In multivariate analysis, there was no significant difference in pCR for obese compared with normal weight patients (odds ratio [OR] = 0.78; 95% CI, 0.49 to 1.26). Overweight and the combination of overweight and obese patients were significantly less likely to have a pCR (OR = 0.59; 95% CI, 0.37 to 0.95; and OR = 0.67; 95% CI, 0.45 to 0.99, respectively). Obese patients were more likely to have hormone-negative tumors (P < .01), stage III tumors (P < .01), and worse overall survival (P = .006) at a median follow-up time of 4.1 years.
CONCLUSION:
Higher BMI was associated with worse pCR to NC. In addition, its association with worse overall survival suggests that greater attention should be focused on this risk factor to optimize the care of breast cancer patients.
Here’s what Obesity had to say about the similarity, which prompted first a withdrawal then a retraction in May 2011:
The editors of the journal Obesity have come to realize that significant portions of the text and data of this article were plagiarized from a previously published article. Because of this plagiarism, Drs. Richard N. Bergman (Editor-in-Chief), Marilyn Ader (Deputy Editor), and Steven D. Mittelman (Associate Editor) hereby retract the manuscript from the journal Obesity.
The following authors have stated in correspondence to the editors that they had no knowledge of submission of this manuscript: Landolsi Amel, Maaloul Jihen, and Ben Ahmed Slim.
There are only two other authors on that paper, Mahmoudi Kacem and Msolly Awatef. Both appear with Slim on a 2011 article in the Annals of Saudi Medicine, “Association between body mass index and risk of breast cancer in Tunisian women,” which also has strong echoes in an earlier article from an unrelated group.
Here’s the abstract from the Annals of Saudi Medicine paper:
BACKGROUND AND OBJECTIVES:
The number of breast cancer in women has increased dramatically in Tunisia. The cause is perceived to stem from adaptation to a westernized life style which increases body mass index (BMI). This study aimed to investigate the association between BMI and breast cancer among Tunisian women.
DESIGN AND SETTING:
Hospital-based case control study of breast cancer patients seen between November 2006 and April 2009 at the University College Hospital Farhat Hached in Sousse, Tunisia.
PATIENTS AND METHODS:
Standardized questionnaires concerning BMI and other anthropometric data were completed on 400 breast cancer cases and 400 controls. The controls were frequency-matched to the cases by age. Results: BMI at diagnosis was positively correlated with the risk of breast cancer among postmenopausal women (P<.001 for trend). When compared with women with a low BMI (<19), women with a BMI of 23-27 and 27-31 had a 1.7-fold (95% CI, 1.1-2.9) and 2.1-fold (95% CI, 1.1-3.9) increased risk of breast cancer, respectively, after adjustment for non-anthropometric risk factors. BMI at diagnosis was not related to the risk of breast cancer among premenopausal women. The odds ratios for premenopausal women with a BMI of 23-27 and 27-31 were 1.5 (95% CI, 0.8-2.8) and 1.3 (95% CI, 0.4-4.5), respectively. Furthermore, present BMI was not associated with breast cancer risk in either pre- and postmenopausal women.
CONCLUSIONS:
Weight control in obese women may be an effective measure of breast cancer prevention in postmenopausal women
And its predecessor, “Association between body mass index and risk of formation of breast cancer in Chinese women,” which appeared in 2005 in the Asian Journal of Surgery:
OBJECTIVE:
To analyse the association between body mass index (BMI) and breast cancer risk among Chinese women in Hong Kong.
METHODS:
We conducted a population-based case control study of breast cancer in June 2002. Standardized questionnaires concerning BMI and other anthropometric data were completed by patients at the Queen Mary Hospital (QMH). The cases were 198 women aged 24-85 years who had documented breast cancer in 1995-2000 by triple assessment criteria, and the controls were 353 women who were followed up at QMH for benign breast disease after breast cancer had been excluded by triple assessment. The controls were frequency-matched to the cases by age.
RESULTS:
BMI at diagnosis was positively correlated with the risk of breast cancer among postmenopausal women (p < 0.001 for trend). Also, when compared with women with a low BMI (< 19), women with a BMI of 23-27 and 27-31 had a 1.73-fold (95% confidence interval, CI, 1.04-2.86) and 2.06-fold (95% CI, 1.08-3.93) increased risk of breast cancer, respectively, after adjustment for non-anthropometric risk factors. BMI at diagnosis, however, was not related to the risk of breast cancer among premenopausal women. The odds ratios for premenopausal women with a BMI of 23-27 and 27-31 were 1.5 (95% CI, 0.82-2.71) and 1.32 (95% CI, 0.39-4.43), respectively. Furthermore, present BMI and BMI 5 years before diagnosis were poorly associated with breast cancer risk among both pre- and postmenopausal women.
CONCLUSION:
Weight control in obese women may be an effective measure for breast cancer prevention in postmenopausal women.
Slim was senior author on a 2011 article in Cancer Epidemiology — on which Kacem, Awatef and Amel also appeared — that was withdrawn, although there seems to be another version of the paper that isn’t marked that way. The reason for the move isn’t stated (such is the standard for withdrawals, alas), but we put parts of the abstract, which is still available on ScienceDirect through Google and came up with a 2007 paper in Cancer Detection and Prevention. We’re guessing that’s not merely a coincidence.
A MedLine search turns up four articles with Slim, Awatef and Kacem as co-authors. Same for just Awatef and Kacem.
We emailed Slim and the managing editor of the Saudi journal for comment and will update this post if we learn more.
We’ll note that this case underscores how withdrawal notices can be cheap tricks for unethical researchers. If Obesity — one of whose editors tipped us off to this case — hadn’t pursued the retraction instead of allowing the group simply to withdraw its paper, readers would not know to be wary of their other studies.
Kudos for managing to write an entire article about obesity without making a pun about the lead author being named Slim!
On another Tunisian science related topic, there’s an interesting post over on “In The Pipeline” this morning, about life extenstion by C60 buckminsterfullerene (http://bit.ly/J90wRU) from a French/Tunisian group. I’ll reproduce my comment (posting under a different alias) from that blog here…
“The oldest lived animals were 66 months (5.5 years), and the paper was submitted in Jan 2012. Thus, the studies were begun in July 2006. Many of the papers they cite regarding the rationale for the dosing regimen were published AFTER the study had already begun. The paper providing the rationale for stopping dosing at 7 months (due to accumulation of c60 in the liver) was not published until 2010, more than 3 years after they’d already stopped dosing! You can’t cite something as rationale, when it wasn’t even published when you were designing the study”. Something is odd about this paper for sure.
Provided the Tunisian authors cited the original paper I really don’t see the problem here. Tunisians first language will be Arabic, 2nd language French and then third language English.
I can’t say whether the original paper was of such clinical significance that it was worth replicating in a different population, but the Tunisian doctors seemed to think so. Since we think it beneficial if studies are repeated in different populations, whether or not the language is repeated seems a fairly minor thing to get uptight about.
One can’t help wondering if there was an element of anti-Arab racism on the part of Berman, Ader and Mittelman, who might have proved to be far more understanding if the authors had hailed from the Charite Hospital Berlin.
Well, you maybe already guess it: they did not cite the paper by Litton et al. Also, do you have any concern that three authors claim they had no knowledge that this paper was submitted?
Look more closely. In both pairs of papers the results are numerically identical. This is either a coincidence of cosmic proportions or the research on “Tunisian women” was never performed at all, meaning that the papers were fraudulent from soup to nuts.
Generally I assume that if you work in an oncology department, have lots of cancer patients, then its a breeze to turn out this type of paper as it is purely descriptive. I mean why not measure the BMI of your patients and determine if that is associated with responsiveness to therapy?
However looking at some of the raw data
http://www.nature.com/oby/journal/vaop/ncurrent/fig_tab/oby2010101t1.html#figure-title
We learn such gems as 143.5 patients have positive lymph nodes and 192.5 patients don’t. 36.9 patients are in tumour stage four etc etc.
The mind-bogglingly idiocy of such fabrications is quite extraordinary. He seems to know enough how to reverse engineer stats from a given population without a clue what they mean.
in theory reviewers ought to catch this, but its not a level of cheating most would look out for.
This reminds me of the time George Washington Carver declared himself a scientist and started cranking out patents for products white people had been making and using for decades.
Retraction
http://www.sciencedirect.com/science/article/pii/S1550728912001062
Surgery for Obesity and Related Diseases
Available online 24 April 2012
Retraction: Shang E, Hasenberg T. Aerobic endurance training improves weight loss, body composition, and co-morbidities in patients after laparoscopic Roux-en-Y gastric bypass. Surg Obes Relat Dis 2010;6:260–6
A minor detail: Tunisians follow the French practice of listing their last names first. So Slim, Kacem, Awatef and Amel are, in this case, the first names of the authors.
Is the country of Tunisia relevant enough to this story mention it in the title and make into a plagiarism pun?
I sense the presence of compelling forces that propel these authors and the publishers to publish no matter what !
littlegreyrabbit – I initially shared your concern that this was a case of Arab authors being bashed for replicating work locally while using apparently standard English phrasing borrowed from other abstracts. However, look at the stats toward the end of the results. In the Journal of Clinical Oncology study, there were 1169 patients, and the stats on reduced likelihood of pCR in overweight or overweight plus obese women were “OR = 0.59; 95% CI, 0.37 to 0.95; and OR = 0.67; 95% CI, 0.45 to 0.99, respectively”. (Do the math, this means obese women were more likely to obtain a pCR than overweight women, and the difference compared to normal-weight women would not have been significant. Hmmm.) Now check out the paper in Obesity. Supposedly done in 800 women, 8 years younger on average, with a different distribution of weights. Yet the stated odds ratios and confidence intervals were identical. What are the odds of THAT happening?