Pfizer has retracted a 2009 Journal of Clinical Oncology study purporting to show a benefit of their experimental drug for lung cancer figitumumab after discovering that its clinical lead on the project had done analyses improperly.
Here’s the text of the notice:
Pfizer, Inc., has recently learned that the overall response rate and progression-free survival data published in “Phase II Study of the Anti–Insulin-Like Growth Factor Type 1 Receptor Antibody CP-751,871 in Combination With Paclitaxel and Carboplatin in Previously Untreated, Locally Advanced, or Metastatic Non-Small-Cell Lung Cancer” (J Clin Oncol 27:2516-2522, 2009) are incorrect. These results, which have no impact on the safety profile of figitumumab, were discovered by Pfizer researchers during routine study close-out activities.
Specifically, in some cases, responses were not confirmed by performing a 4-week follow-up computed tomography scan per RECIST, or the follow-up scan did not verify continued tumor shrinkage. In addition, the response rate may have been overestimated as a result of including both confirmed and unconfirmed responses in the analysis. These errors in the objective response rates and progression-free survival data appear to be the result of inaccurate clinical data analyses that were performed outside of Pfizer’s standard operating procedures and are not the result of problems arising from study conduct. The revised data, included in the table below, were obtained through strict application of RECIST criteria and assignment of patients to a dosing level on the basis of original treatment received.
Because these results are fundamental to the conclusions of the paper, Pfizer and the lead author, Daniel D. Karp, MD, respectfully request that this article be retracted. We have shared the proposed retraction notice with all of the authors, and all agreed with this retraction notice. Pfizer has no reason to believe that the external authors contributed to the publication of these incorrect data and deeply regrets the impact of this action on the work of other investigators and to the readers of Journal of Clinical Oncology.
In a table included in the notice, there’s a clear difference in the results before and after the corrected analysis. In the original paper, for example, the control group, who received standard lung cancer treatments paclitaxel and carboplatin, had a median progression-free survival of about 3.5 months, while those who had the highest dose of CP-751,871 had a median of 5.0 months without progression. In the new analysis, there was no statistically significant difference between those two groups, with the control at 4.3 months and the high-dose group at 4.5 months.
Pfizer tells Retraction Watch that the study
…demonstrated that figitumumab was likely to be effective when used in combination with two chemotherapy drugs, paclitaxel and carboplatin, in patients with NSCLC; but not to the same degree as the previously published results indicated. The study did not demonstrate that the combination of the drugs was effective in the studied subset of patients with non-adenocarcinoma.
CP-751,871 is also known as figitumumab, which, as the company explains,
is a Pfizer-discovered monoclonal antibody that was developed for the treatment of various forms of cancer, including lung, prostate, breast and colorectal cancers and Ewing’s sarcoma.
The company stopped developing the drug in January 2011 after two Phase 3 studies in non-small-cell lung cancer (NSCLC) failed. The first of those trials ended in December 2009 after “independent monitors had found more serious adverse events, including deaths, with patients receiving figitumumab,” Reuters reported at the time. (The company stopped enrolling new patients in October of that year.) Pfizer is “no longer seeking regulatory approval for this compound.”
The 2009 JCO study has been cited 152 times, according to Google Scholar.
Pfizer gave Retraction Watch a bit more detail on what happened, including the fact that their clinical lead on the compound has left the company:
While conducting study close-out activities, Pfizer became aware that final data from a Phase 2 study of figitumumab, known as the 1002 study, and certain biomarker analyses performed in connection with the figitumumab program differed from the results that had been reported previously to regulators and published in medical journals.
We conducted a thorough and comprehensive review to determine how this may have occurred, and enlisted the services of independent experts to assist with this review. Our review determined that these errors were the result of inaccurately performed clinical data analyses. These incorrect analyses were performed outside of Pfizer’s standard operating procedures by the clinical lead, who left the Company in September 2010. The problems were not based on study conduct or the work of the external investigators.
We expect our employees to adhere to the highest levels of personal and professional integrity. We also expect colleagues to take ownership and accountability, and to understand that failure to do so can lead to serious consequences including termination.
The clinical lead on the project appears to have been Antonio Gualberto, who is listed as a faculty member at Brown University’s medical school but does not appear to be employed there anymore, as his email bounces.
Pfizer continued:
Patient health and safety is our first concern. We want you to know that these new findings do not alter the safety profile of figitumumab.
Since Pfizer learned of these issues, we have been carefully analyzing our internal controls and ensuring that issues such as these did not exist in any other development programs. Our internal review has confirmed that this was an isolated incident and the controls in place at that time were rigorous.
We are committed to ensuring that accurate clinical study results are disclosed to medical professionals, patients and the public. As per our standard practice, we are making the final study results available in a timely and responsible manner.
We’ve tried to reach lead author Daniel Karp, of MD Anderson, and will update with anything we learn.
How could Pfizer claim that “figitumumab was likely to be effective” when the revised results showed a progression free survival of 4.5 vs 4.3 months? (That’s six days.) Even the initially reported results were pathetic: a month and a half (admittedly, a third more life than the control group.)
As to the side effects: ANY regimen that includes carboplatin is likely to have intolerable side effects for, probably, half the patients treated.
Once again, I’m diving into an area I’m not qualified to critique (being an ex-family practicioner rather than a treating oncologist) but I think this stinks. Surely the established research community has better ideas for treating a cancer that has such poor survival rates than just combining another monoclonal antibody with carboplatin and paclitaxel? Many victims of NSCLC would rather just go home to die than suffer through a regimen like that to gain an average of 1.5 months (or is it six days?) of longer life.
Some of such patients (the more cynical ones) would say that the research oncologists involved are just playing around with the patients’ lives to justify their salaries. At least, after this retraction, there’s one less researcher working for Pfizer.
I am guessing the treatment was ‘very effective’ and ‘cured cancer’ during invitro studies and in the pre-clinical models (probably due to data manipulation), but when the treatment was actually used in people there was no effect whatsoever (6 days = no effect). It is far more difficult (though not impossible) to data manage deaths during trials – but some have tried and published (see science fraud blog)
Cancer research seems to be prevaded with frauds at the moment. The more I look, the more scary it gets.
The massive fraud at the moment gives due reason to include a module in data fraud detection (rather than an ethics module) in all undergraduate courses.
Clearly, the ethics teaching has failed, and miserably so.
I am guessing it was NOT data manipulation in in-vitro studies and pre-clinical models. Going into clinical trials costs many hundred of millions, and a company will only do so if it has solid data that the treatment may well work. Companies can try and spin the results of clinical trials, but they have absolutely no reason to do so with the preclinical data!
I do not think that the authors had to guess.
In Clin Cancer Res. 2010 Sep 15;16(18):4654-65.
http://clincancerres.aacrjournals.org/content/16/18/4654.long
the authors have figured out that only a subset of patietns with non-small cell lung cancer (NSCLC) to benefit from non-small cell lung cancer (NSCLC) to anti-insulin-like growth factor receptor (IGF-IR) therapy (the inclusion of figitumumab).
The first author is up front when he mentioned
Disclosure of Potential Conflicts of Interest
A. Gualberto: ownership interest, Pfizer, Inc.
In Br J Cancer. 2011 Jan 4;104(1):68-74
http://www.nature.com/bjc/journal/v104/n1/full/6605972a.html
the authors quite frankly state “Phase III trials of the anti-insulin-like growth factor type 1 receptor (IGF-IR) antibody figitumumab (F) in unselected non-small-cell lung cancer (NSCLC) patients were recently discontinued owing to futility.”
Did it not occur to the authors to inform the Journal of Clinical Oncology on knowing this that some notice should be placed on J Clin Oncol. 2009 May 20;27(15):2516-22.
http://jco.ascopubs.org/content/27/15/2516.long
so that people would know the latest developments?
The retraction of J Clin Oncol. 2009 May 20;27(15):2516-22. did not appear until the 19th of October 2012.
http://jco.ascopubs.org/content/suppl/2012/10/24/JCO.2008.19.9331.DC1/Karp_retraction_FINAL.pdf
There are 40 publications in pubmed which refer to “figitumumab”.
http://www.ncbi.nlm.nih.gov/pubmed/?term=figitumumab%20
Including a a very recent publication: J Natl Cancer Inst. 2012 Jul 3;104(13):975-81.
http://www.ncbi.nlm.nih.gov/pubmed/22761272
“Although early results from clinical trials that targeted the IGF1R showed some evidence of response, larger randomized phase III trials have not shown clear clinical benefit of targeting this pathway in combination with conventional strategies. These disappointing results have resulted in the discontinuation of several anti-IGF1R programs. However, the conduct of these trials has brought to the forefront several important factors that need to be considered in the conduct of future clinical trials.”
Some of the other publications seem to confuse a biomarker with the disease.
http://www.ncbi.nlm.nih.gov/pubmed/22553344
Clin Cancer Res. 2012 Jun 15;18(12):3407-13.
“Figitumumab is biologically active in prostate cancer. PSA declines in treatment-naive patients were observed, potentially mediated by IGF-IR effects on androgen receptor expression.”
to Marco
‘I am guessing it was NOT data manipulation in in-vitro studies and pre-clinical models. Going into clinical trials costs many hundred of millions, and a company will only do so if it has solid data that the treatment may well work. Companies can try and spin the results of clinical trials, but they have absolutely no reason to do so with the preclinical data!’
Yes, you’re right, there is no data manipulation using cell culture and preclinical models (usually mice).
The fact the ‘company’ spent alot of money on the ‘treatment’ and then had to do all sorts of data, ahem, management, shows they have been victims of fraud, and if they have any sense, will sue.
Now that I’ve got your attention: I’m going to ask a stupid question.
What practical issues prevent researchers from offering a new agent by itself rather than at the same time as the conventional treatment?
ie why not, rather than a trial of carboplatin plus paclitaxel plus figitumumab, a trial of paclitaxel plus figitumumab, or a trial of figitumumab alone? [“figitumumab” needs to be edited for conciseness; I suggest “fumab”.]
If you run a trial of fumab alone, and by chance or whatever your set of patients is healthier than the patients that were once on the paclitaxel trial, you will be lead to believe that fumab is better than paclitaxel even if fumab is no better or perhaps even slightly worse. Trials were run in this fashion 50 years ago, but researchers and regulators learned that such comparisons were easily prone to such biases, intentional or not. Hence the current trial guidelines to always have a control arm, and also to analyze the data according to “intention to treat” (analyze all patients randomized to fumab as if they did indeed take all their fumab, even those who had a reaction and had to stop taking fumab).
Because Taxols and cisplatins are proven treatments and shouldnt interact with targeting a receptor via a monoclonal.
It is not ethical to run a trial which denies patients proven effective treatments.
I understand the part about having a “control” arm, which has to include the previously accepted “best” treatment; and the “intention to treat” analysis, which includes patients who weren’t actually treated(given the intended regimen) but were randomized to that group. I was trying to advocate for trials of the new drug alone, without the previously “proven effective treatments.”
It’s just because I’m so skeptical of the effectiveness of cisplatin; not so much paclitaxel, which is probably somewhat effective… on the other hand, on searching, I quickly found studies of fumab alone, such as the one with stage I prostate Ca, which was not really very impressive. It might have been more impressive if it had lasted a year, or five years; but the author’s intent wasn’t to obtain mortality data. The author intended to evaluate “surrogate markers” of effectiveness in order to do a study in a reasonable amount of time.
I’m sure cisplatin has studies showing effectiveness, at least I hope I’m sure. But what if those studies were flawed or inaccurate? Anyone care to discuss the robustness of clinical trial data about cisplatin?
I guess I’m just squeamish about giving cisplatin to people who are already dying from cancer. It’s very unscientific, but I feel bad for patients who suffer intolerable side effects from treatment when they’re staring death in the face. Maybe I’ll write a story about a couple of patients I saw.
Um, a 25-50% decline in a PSA that wasn’t very high to start with (no more than 35) just doesn’t impress, does it? (Study of figitumumab alone in treatment-naive Stage I prostate cancer patients.)
Sorry for asking this question: is Google Scholar authentic source to look for citations? has this been validated? Pubmed, ISI and Scopus seem to be somewhat selective in chosing journal citations as new journals are not included in their database.
I suppose you can always check out the publications in which the citations appear.
If the issue is simply about citations, then you have to say that if the journals exist that they are citations.
This site is a blog, but leading newspapers such as the New York Times, Canada’s National Post, the Guardian in the U.K., do cite it.
All journals were new at one time. Moving too slowly with the times might miss important things out.
Until very recently journals like Nature didn’t like the internet or e-mails (as if they were subversive in themselves), not realising that they are swifter and softer on the environment than sail-mail.
None of the citation indices have a “fraud” or “scientific misconduct” app with them.
Many believe that is the case, but it is not true.
Google scholar also finds citations in (some) books and (some) PhD theses. I even found one of my papers cited in a project proposal. That automatically means you can get some quite different numbers.
Some people are so desperate that they scavenge the internet for any mention of their work. But not all citations are born equal. If a peer-reviewed paper in a journal with an assigned impact factor cites your work, than this is a valuable citation. If some cooking magazine mentions your work than the value of this citation is close to zero. The most nonsensical sociology papers (like the recently mentioned on this blog paper on the woman’s time of the month influencing her political views) get all the buzz in the mass media (that is, a gazillion of citations) but the value of such citations matches the value of the paper itself – Scopus & co. will not list such “citations”.
I don’t know which citation index is the best. I was simply trying to point out that it might depend on what you are looking for. The standard ones do put up retraction notices, although that can take some time.
Marco November 5, 2012 at 1:21 am
“Google scholar also finds citations in (some) books and (some) PhD theses.” Was that meant to be a criticism or a staement?
Books are publications. Reviews are common enough in journals. What’s the difference between an eBook and an eJournal? A bit 20th century to make such distinctions is one opinion. Even in the 1920s people like
Walter Benjamin were saying that the form of the book would soon change.
Finding theses using scholar can be very useful, for example a published record of reagents you can ask for.
I don’t think Marco was making any derogatory inference about books (or MSc/PhD theses)! However books (and theses) serve different purposes from published papers, by and large. The published paper is the publication of record in which (hopefully) the authors have given it their best shot to get the data correct, the stats done properly, and are comfortable with the interpretations etc.. The journal article is peer-reviewed as a secondary quality check.
Books (usually) and theses are not peer-reviewed, and may contain preliminary data/analyses and less well justifiable conclusions. They are much more often reviews that primary research. Often conference proceedings are published in book form, but these don’t constitute the publication of record. It would be very strange indeed to cite a book as a source of primary research data- anything of scientific value is published in the peer-reviewed literature. Books may be cited as general background or because you might be referring to some interpretation that someone has written about in a book and so on…
As for some citation indexes not containing some new journals….
(i) Some citation indexes take a while to compile up to date lists of papers. ISI Web of Science can be quite slow (maybe several months post publication).
(ii) Some citation indexes make very rapid additions to the database including abstracts of papers in press or in on-line “pre-publication” journal web-pages. PubMed does this for example.
(iii) Citation Indexes need to make some judgement about the quality of a journal before adding it to their database. There are loads of rubbish journals…ISI Web of science doesn’t list some of these, or else it may be slow to incoporate these into the database.
etc. etc…
reply to chris November 6, 2012 at 6:59 am
Just because something has a hardback to it does not make it so different.
I hope that books also show the correct data. Could you point me in the direction of scientific books where this is not the case?
“Books (usually) and theses are not peer-reviewed, and may contain preliminary data/analyses and less well justifiable conclusions. They are much more often reviews that primary research.”
I don’t know the tradition in your country, but if you present a review as your Ph.D. thesis in many countries you will be disappointed. The data in a Ph.D. thesis has to stand up.
I believe that the average Ph.D. thesis gets much more attention than the average paper.
Springer Press has series upon series of journal/books. It does retract if there is overlap.
From this very blog we know that “The journal article is peer-reviewed as a secondary quality check.” is very variable.
Reviews stuff many journals.
It was meant as a statement. I have a higher h-index using Google Scholar than e.g. ISI, so guess which one I use when I am asked about my h-index :-).
But you could read it as a criticism, too, in the sense that Google Scholar is a bit more arbitrary than the other organizations that provide metrics. “Some” is an important qualifier in this context, as “some” books show up in the “cited by” list, but others do not. Same goes for PhD theses. Also, my PhD thesis contains reprints (sort of) of the papers I wrote, and thus there would be double-counting of citations if my thesis would have been included in the statistics for the papers I cited.
I take the point about double-counting. Something best to avoid.That can be unfair.
If the thesis were the source of the information then it would be right to cite it. Many are electronic now to make them more available, also more available for checking (e.g. zu Guttenberg).
In the U.K. mostly the thesis is not published as papers, in Sweden it often is, or even encouraged to be published as papers.
In the U.K. you do not have to publish papers to get your Ph.D., but have to do an original piece of work by yourself which is presented in your thesis. You have to write at the front that you did the work/experiments. In the Netherlands you publish papers as you go along so that the final thesis can look quite small, but there are the papers to go with it. The details might be absolutely correct for each university or country, but this is what I have gathered from talking to people from these countries.