The Oncologist has tagged three review papers that share a first author with an expression of concern. The three papers, which have together been cited more than 1,000 times, focus on HER2, a gene that can contribute to breast cancer.
Though the papers contain errors, the conclusions — about how the HER2 gene serves as a predictive factor for breast cancer, and a target for therapies — remain unchanged, according to the EOC. The editor of the journal notes that the conclusions of the papers have been confirmed by other publications. Two of the papers are more than 10 years old, and today many patients continue to be treated with medications that target HER2, such as Herceptin.
Here’s the expression of concern:
The Editors have received information that three articles [1–3] on the subject of HER2 amplified breast cancer, previously published by this Journal, contained errors in the citation of methods and interpretation of primary data cited by the authors. Although the general conclusion of the papers remains unchanged, the Editors of The Oncologist bring these inaccuracies to the attention of readers and suggest that this information should be taken into account in making reference to these three articles [1–3] and in judging their content. This statement has been reviewed by the senior authors, who endorse this Expression of Concern.
Jeffrey S. Ross, who works at Albany Medical Center, is the first author on all three papers that the EOC applies to:
- “The HER-2/neu Oncogene in Breast Cancer: Prognostic Factor, Predictive Factor, and Target for Therapy,” published in 1998 and cited 449 times, according to Thomson Reuters Web of Science. Jonathan A. Fletcher, who works at Brigham And Women’s Hospital, is the last author.
- “The HER-2/neu Gene and Protein in Breast Cancer 2003: Biomarker and Target of Therapy,” published in 2003 and cited 361 times. Kenneth Bloom, who is the Head of Oncology & Immunotherapy at Human Longevity, Inc., is the last author. Fletcher is also a co-author on this paper.
- “The HER-2 Receptor and Breast Cancer: Ten Years of Targeted Anti–HER-2 Therapy and Personalized Medicine,” published in 2009 and cited 378 times (making it a “highly cited” paper). Gabriel N. Hortobagyi, affiliated with the MD Anderson Cancer Center, is the last author.
We asked Bruce A. Chabner, the editor in chief of The Oncologist, for more information on the specifics of the errors. He told us:
The reader who brought the concerns to the Editors’ attention has raised issues with the meta-analysis of published data linking the amplification of HER2 detected by FISH and/or overexpression of the HER2 protein detected by IHC and prognosis in breast cancer for patients not treated with anti-HER2 targeted therapies, which only became available in 1998. The reader identified a number of errors in classification of the results of the studies such as whether the data was evaluated by univariate or multivariate statistical analysis. In addition, in the review, prognosis for breast cancer was not clearly defined in terms of separating time to progression from overall survival.
The journal editor and the authors recognize the deficiencies in the analysis of prior literature in the three review articles called into question, and have published our concern. However, these deficiencies do not invalidate the papers’ primary conclusion, which has been confirmed by a number of concurrent and subsequent publications. It is generally accepted that amplification and/or overexpression of HER2 in newly diagnosed breast cancer in the pre-targeted therapy era is significantly associated with an adverse clinical outcome independent of both clinical (patient’s age for example) or pathologic (tumor grade, size, estrogen receptor status for example) parameters classically used to evaluate newly diagnosed patients. Therefore, we feel that our notice of concern is sufficient to alert our readers.
The reader who alerted the journal to the issues with the paper is Robert Fortner, a writer who is researching breast cancer for a book he’s working on. He has described his concerns with the papers on his blog.
Fletcher, who is a co-author on author on the 1998 paper, told us that:
To my understanding, the concern was with the way some of the data from various series were summarized in these reviews. My role was to provide information and examples on molecular cytogenetic detection of Her2 (which was then a less routine assay), and I wasn’t involved in summarizing the previously published data, which was the main focus of the reviews.
Hortobagyi, a senior editor at The Oncologist, told us more about the concerns regarding the 2009 paper, on which he is a co-author:
The errors mentioned referred to statements regarding “independent” prognostic value for HER2 overexpression and/or amplification. The information in question included some papers in which multivariable analyses were not performed and therefore, the word “independent’ was inappropriately used, since in the absence of such analyses, one cannot state with [certainty] that the variable indeed contributes independently to the prediction of prognosis. In some other papers referenced in the publication, HER2 expression/amplification was correlated to other prognostic biomarkers and not directly to overall survival. Therefore, the claim of HER2 being prognostic would not be directly supported by those papers. In addition, the conflict of interest of the first author was not accurate.
Here’s how the conflict of interest disclosure for Ross reads on the 2009 paper:
Jeffrey S. Ross: None
We asked Hortobagyi if the errors have affected papers that cite the 2009 review. He said:
Yes, although the weight of the evidence that exists today clearly supports the claim that HER2 overexpression/amplification is a adverse prognostic factor. Therefore, whether you include this review or not, currently accepted knowledge about HER2 would remain unchanged.
We have reached out to Ross (who has a building named after him at Albany Medical College), and Bloom. We’ll update this post with anything else we learn.
Like Retraction Watch? Consider making a tax-deductible contribution to support our growth. You can also follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up on our homepage for an email every time there’s a new post, or subscribe to our new daily digest. Click here to review our Comments Policy. For a sneak peek at what we’re working on, click here.
In article concerning Ross what relevance to EOC is there to the fact that there is a building named for him?
“However, these deficiencies do not invalidate the papers’ primary conclusion, which has been confirmed by a number of concurrent and subsequent publications.”
Although I’m not in the field, and this case may very well be different since it’s a review of some sort, this is all too typical of a terrible attitude that I’ve been seeing in a lot of replies by authors and editors on this blog. Saying ‘the study has been confirmed by subsequent publications’ only makes me think it and the authors are more suspect from now on; this should not be said by authors/editors. Never. At all.
There is a well-known recipe for quickly publishing substandard work with lots of shortcuts in a glamour mag so you can take all the glory, grants, and citations, while someone who is more careful and has a stronger study, gets second place. It has led to the downfall of some who took so many shortcuts that what they published actually turned out to be wrong when someone subsequently filled in the holes. And even if they are not wrong, it just looks bad when the shortcuts get exposed. You don’t want to (erroneously or not) associate yourself with this kind of practice by putting out statements such as the one I quoted above. The primary conclusion should stand on the basis of the current paper. If not, then the subsequent publications and the associated authors get all the glory.
Disclamer: I’m not saying that this is necessarily true of this study and I’m not qualified to judge the work. Just making a general statement.
By the very early 1990s a series of primary research papers had clearly established c-erbB2 (or HER-2 as it became known) as an independent prognostic marker that identified a subset of patients who were going to do badly. There were confounding groups patients, who had poor markers (hormone receptor negative, c-erbB2 positive), but who did well, an issue which is now understood.
So the ‘problem’ with these review papers is that they put things into the same basket that should have been kept in different ones, in terms of their pulling together published data.
A contributing factor may have been be space – even in 2016 many journals still stick to the print concept and restrict space. Another contributing factor will have been the complete absence of access to original data. This means the authors will have had to glean such data form publications, and that is extremely difficult.
So I’m on the fence on this one, I can see why some may consider the EOC to be a weak response, but this position is not entirely clear to me – it is difficult to write a review with closed data, but the reviews are necessary.
On Mark Moasser’s telling (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3021475/):
“Initial conflicting reports regarding the prognostic relevance of HER2 were resolved with improved methodologies and the overwhelming data now confirms this initial landmark genetic-biologic finding (nicely reviewed in (Ross et al. 2003)).”
So it is my sense that it is not quite right to say that HER2’s prognostic value was established by the early 1990s. Even in the Ross et al. literature reviews, which do not specify selection criteria, there are numerous studies that do not find HER2 to be prognostic.
In addition, the Ross et al. reviews simply mis-report a number of findings. For example, a paper from Battifora et al. (1991) is categorized by Ross et al. as “yes” under multivariate analysis of prognostic factors. However, Battifora and colleagues clearly found HER2 was not independently prognostic:
“Stepwise Cox Regression: This analysis identified independent prognostic factors of DFS and OS when all variables were considered together. Independent predictors of DFS included stage of disease, histology, and nuclear grade. Nuclear grade and stage were the only significant predictors of OS.”
I found 10 errors of this sort out of 107 papers reviewed in Ross et al. (2009); the papers found “no” but Ross et al. report “yes.”
Also, a separate group of 7 of the 107 papers conducted no multivariate analysis of whether HER2 was prognostic, but Ross et al. (2009) report that those studies did and that each found HER2 prognostic. Six of the seven didn’t conduct any multivariate analysis; one of the seven did but all patients were HER2 positive. (The seven papers are: Dykins et al., Molina et al., Press et al., Charpin et al., Agrup et al., Horita et al., and Tanner et al.)
Co-author Gabriel Hortobagyi, quoted above in this RW post, observed that in some papers included in Ross et al., “HER2 expression/amplification was correlated to other prognostic biomarkers and not directly to overall survival.”
11 of the 107 papers included in Ross et al. (2009) correlated HER2 with other biomarkers and not clinical outcomes. Among the 11 is the paper with by far the largest number of patients, Lal et al. (2005), n=3,655. A total of 7,511 (19%) of the 39,730 patients in Ross et al. should not have been included.
When I corrected the errors in Ross et al. (1998), it inverted the paper’s conclusion so that 60% of studies found HER2 not prognostic. (http://robertfortner.posthaven.com/prognostic-findings-for-her2-in-breast-cancer-not-reproducible)
I invite others to confirm or disconfirm these errors. Overall, I documented issues with 30 of the 107 papers reviewed in Ross et al. (2009), an error rate of 28%.
Mr. Fortner, the article you cite by Mark Moasser explicitly states, “Numerous experimental models now solidly support the hypothesis that HER2 overexpression promotes tumorigenesis and the finding of HER2 amplification and overexpression in many human breast cancers and its link with the biology of this disease now clearly implicate this oncogene in the pathogenesis of this type of human cancer…The critical driving role of HER2 in human cancers, and the large number of patients affected by this type of cancer has made HER2 an ideal target for the development of rationally designed anti-cancer drugs.”
So even if there hadn’t been a consensus in the 1990s, there certainly is one now.
This raises a question for me: why do you care? With HER2 overexpression as a prognostic factor clearly established, what difference does it make whether these few papers in particular have (apparently serious) inaccuracies? You can have a legitimate argument about whether The Oncologist should have corrected them or identified the specific errors in the EOC (I’m inclined to believe that it’s unnecessary at this point, but I can appreciate other points of view on this). But why turn this into a crusade?
I think it is true there is a consensus (indeed confident unanimity) that HER2 is prognostic. What is that based on if not these reviews?
Measured by quality of evidence, I believe experimental models come in below clinical studies of the type reviewed in the Ross et al. papers. Indeed, experimental models found that Herceptin blocks HER2 signaling but, again, if Mark Moasser is to be believed, Herceptin does not block HER2 signaling. (I believe lapatinib does, and researchers other than Mark have arrived at similar findings.)
I asked Mark (some time ago) what citation might substitute for Ross et al. in his paper but received no reply. After reading this post, I also emailed Bruce Chabner and Gabriel Hortobagyi inquiring about specifics for what Chabner called “a number of concurrent and subsequent publications” showing that HER2 is prognostic. To this point, I haven’t heard back.
It seems possible that the consensus about HER2 as prognostic rests at least in part on these papers whose validity is being examined. If the status of these papers changes, consensus about HER2’s prognostic value perhaps should track changes in the scientific evidence.