Dr. Anil Potti and his co-authors on Friday formally withdrew a 2008 study published in the Journal of the American Medical Association (JAMA) that purported to offer a “genetic signature” that would show which patients with breast cancer would respond to a particular treatment.
The retraction was the eighth of what officials at Duke, in Durham, North Carolina, said this past summer would be a series of about a dozen pulled studies that mainly focused on lung cancer, along with another dozen corrections and partial retractions. The Duke team had published about 40 papers, many of which have been cited hundreds of times by other scientists, according to Thomson Scientific’s Web of Knowledge.
The authors write in a statement on the JAMA website — available at — that they are retracting the paper because it was based on a method they now believe is unreliable. That approach, which no one has been able to reproduce, was described in a now-withdrawn paper in the journal Nature Medicine in 2006.
The case is another example of researchers, some of whom had financial interests in the method, being overly enthusiastic about findings that held great promise for very sick patients, but that did not hold up to scrutiny.
Journals and universities began subjecting the work to intense scrutiny, however, after it emerged that Potti had claimed awards and scholarships that he never received, in biographical sketches for grant applications.
The whole episode has been a “huge setback,” said Dr. Otis Brawley, chief medical officer of the American Cancer Society.
“What happened at Duke is that you have a whole bunch of folks worried that this genomic information is just too complicated, making it too easy for someone to basically create fraudulent science,” Brawley said.
Keith Baggerly, who studies genomics at the M.D. Anderson Cancer Center in Houston, has been looking into the Duke team’s data since shortly after the Nature Medicine report came out. A colleague approached Baggerly and Kevin Coombes, another M.D. Anderson researcher, hoping the work could help improve the treatment of patients at the cancer center.
But Baggerly and Coombes found numerous problems in the research, including mislabeling of data. They asked the Duke researchers for their data, but they were slow to provide it.
Eventually, some journals began publishing the critiques, often alongside defenses of the work by Potti and his colleagues. Meanwhile, clinical trials using the genetic signature continued at Duke.
Baggerly’s questions went largely unheeded until July 2010, when The Cancer Letter, a trade publication, reported that Potti had falsely claimed to be a Rhodes Scholar on an American Cancer Society grant application. Within months, the trials were halted, and Duke returned $729,000 to the cancer organization.
Potti resigned from Duke in November 2010, and has since joined a private oncology practice with offices in North Carolina and South Carolina. Potti was unavailable to comment for this story.
He and his colleagues, along with Duke, face two lawsuits from nine patients who took part in the trials. Eleven other such cases have already been settled for at least $75,000 each, according to the North Carolina Medical Board, which has reprimanded Potti.
For plaintiffs’ attorneys, proving that patients were actually harmed by being in the trials will be difficult, said Brawley.
“I don’t myself think that patients were harmed by being steered to a particular chemotherapy, at one level,” he said. The choices in the trial were widely accepted therapies that their own doctors would have likely given them, Brawley said.
“I’m a little bit concerned that some of the patients may have gone back and had new biopsies, for new analysis in the laboratory,” Brawley said. “If that was purely for this trial, those were unnecessary procedures, with risks. I hope no one was harmed by that.”
One potential move being considered as part of a review of cancer genomics studies by the US Institute of Medicine (IOM) is to require genetic tools such as the one used by the Duke team to go through the same approval process that a medical device would go through at the US government’s Food and Drug Administration.
The Institute of Medicine is an independent, nonprofit organization that provides advice to US policymakers.
The institute’s report is expected out in the next few months, and there are similar changes being considered at the US government’s National Cancer Institute — which funded some of the Duke team’s work — and the FDA. Duke itself has created a team to establish new safeguards as genetic research moves into the clinic.
“To a certain extent, what I’m worried about is that this may show aspects of how it is becoming increasingly difficult to check the scientific literature and how that difficulty stems at least in part from lack of immediate access to data but also lack of code and documentation,” Baggerly said.
Given the highly technical nature of the work, it’s not surprising that the flaws in the papers were not caught before they were published, according to Baggerly.
“That’s actually okay,” he said. “It’s not okay that it took so long for the challenges to be accepted once the research was questioned ... (and) the fact that it made it into guiding clinical trials.”
Brawley said the issue also shows how problematic it is to have universities police their researchers.
“As science progresses, those universities are more and more interested in how many patents they have, and how many licenses they have for those patents,” he said, pointing out that Duke had an ownership stake in CancerGuide Diagnostics, formerly Oncogenomics, Inc.
CancerGuide, which cut ties with Potti in July 2010 after the Rhodes misrepresentation came to light, had licensed technology based on Potti’s work from Duke . The university has since divested from the company.
Despite the setback, cancer treatment will continue to make more and more use of genetic information, Brawley said. Oncologists already use tools such as the OncotypeDX genetic test to tailor treatments for breast cancer and colon cancer.
“We’ve got these new tools, and people are very excited about using them,” Baggerly said. “Some of these genetic level screw-ups are indeed what cause the disease, so there is the potential there. That said, some of the initial claims about how quickly we’re going to be able to turn these (findings) into clinically useful assays have been over-optimistic.”
“We’re going to get there, but a number of the early studies thought this would be an easy problem,” he said. “The biology turns out to be quite complex.” er.”
