A fraudulent cancer breakthrough, a test for the future president of MIT, and a new age of doubt in science

Boston Globe

2 days ago

'No, not today,' Juliet said, according to a transcript of the meeting that would emerge years later in a lawsuit.
The mortality tables were hard to contemplate — life expectancy for stage IV lung cancer is often less than a year. Since Juliet's recent diagnosis, at the age of 69, the Jacobses had scoured the country for cutting-edge treatments. They had narrowed their search to a handful of options: a specialized hospital out West, a renowned cancer surgeon in Boston, and the National Cancer Institute, which they hoped would sponsor a relevant clinical trial. But they had come to Duke University Medical Center to hear about one more.
Get Starting Point
A guide through the most important stories of the morning, delivered Monday through Friday.
Enter Email
Sign Up
The doctor sitting before them was young, just 37. Potti's soft features made him look boyish, harmless even. He called Juliet
dear
and spoke in aphorisms.
Advertisement
'Hope beats despair every day of the week and twice on Sundays,' he told them.
Potti came off as approachable and self-effacing. But in the world of biomedical research, he was also a rising star. He claimed to have developed algorithms — computational tools that can process large amounts of data and generate predictions — to analyze the genetic material inside cancerous tumors and then select the best chemotherapy cocktail to kill them. Some of his colleagues believed he might eventually win a Nobel Prize. Some called his algorithms a Holy Grail.
Advertisement
In his exam room, Potti explained to the Jacobses how it worked.
'We take a piece of tissue,' he said, 'we process it, pull out the DNA and the RNA because that's what tells you how your tumor is behaving. ...Based on your genetics, because there's multiple options, we'll be able to guide you.
'So I don't call myself a lung cancer doctor,' Potti said. 'I tell people I'm here to treat your lung cancer.'
As Potti spoke, the Jacobses nodded along. He was offering them something special, a personalized treatment they couldn't find anywhere else. All they had to do was sign Juliet up for a clinical trial at Duke that was testing Potti's algorithms on patients for the first time.
They'd already booked a flight to see the star surgeon in Boston. But Potti was winning them over. 'Maybe we can just postpone the Boston [trip],' Juliet said.
'We can do that,' Walter replied.
Potti assured them they were in good hands. 'I will help you,' he said. 'Trust me.'
W
hen the Jacboses left their meeting with Potti, they could hardly believe their luck. They trusted Potti — and Duke. They had lived nearby in Chapel Hill since Walter retired after a long career in management consulting. Juliet, a former schoolteacher, now served on the board of a brain tumor center at Duke Medical School.
Shortly after the meeting, the Jacobses' daughter, Melissa Niccum, was cooking dinner when a commercial for Duke came on TV. Potti appeared on screen, promising that 'genomics will revolutionize cancer therapy.'
That's the man who's going to save my mom,
Niccum thought.
Advertisement
What the Jacobs family didn't know — couldn't have known — was that they were now involved in what would become one of the worst medical research scandals of this century. Prominent scientists would see their careers derailed. Duke, an emerging biomedical powerhouse, would be disgraced. Patients would die of their cancers not knowing their final months of treatment had been compromised by scientific fraud.
The scandal would also prove a crucial test for a leader quickly rising through the ranks of academia. The dean overseeing clinical research at Duke Medical School at the time was Sally Kornbluth. Today, she is the president of the Massachusetts Institute of Technology. She is helming one of the top research institutions in the world during a period of unprecedented upheaval in science, at a time when President Trump and his allies have precipitated — and celebrated — the ouster of at least five leaders of elite universities, four of them women.
Walter Jacobs, 86, at his home in Chapel Hill, North Carolina.
Kate Medley for the Boston Globe
The Trump administration has pointed to a rash of scientific misconduct scandals to justify cuts to research funding and proposals to reimagine the way science is conducted. Top officials, including director of the National Institutes of Health Jay Bhattacharya and Secretary of Health and Human Services Robert F. Kennedy Jr., have questioned the ability of peer-reviewed journals to enforce high standards by researchers and adhere to scientific objectivity. Others, including Vice President JD Vance, have cast universities as cynical institutions that advance their own interests at the expense of the public good.
Trump's critics say the administration is trying to seize control of federally-funded science and squelch dissent. The campaign has often been shambolic. In their attempts to defund research on transgender issues, Trump officials have ensnared unrelated studies that include the term 'trans-' — as in 'transgenic genetic material.' In the name of combatting antisemitism, they have put Jewish scientists out of work.
Advertisement
Still, in the past two decades, research misconduct scandals have become alarmingly common, and the scientific establishment has not created safeguards that reliably root out fraud. Theranos, the multibillion-dollar blood testing company, collapsed in 2018 after its technology was revealed to be bogus. The president of Stanford University, Marc Tessier-Lavigne, resigned in 2023 after a research scandal in his lab. Last year, a top scientist at the National Institutes of Health was removed after allegations of fabrication in his research papers. This year, a Harvard behavioral scientist, Francesca Gino, lost her tenure over allegations of data falsification, which she denies. MIT
For this story, the
Globe
reviewed thousands of pages of court records and investigative documents — the vast majority of them never previously reported — that detail how Duke leaders, including Kornbluth, responded to the mounting warnings about Potti's research and allowed his algorithms to be tested on humans for years before the experiments were shut down. Kornbluth was not alone in guiding Duke's response. But as vice dean for research, she played a key role.
She was, by her own account, Duke's 'fact-finding person' reporting directly to Duke University Health System's CEO at the height of the scandal. Many of the warnings about Potti's research went to Kornbluth personally and she shared in the decisions about handling them.
Many of the records the
Globe
reviewed, including Kornbluth's sworn deposition testimony filed in legal proceedings, were stored in a Durham, North Carolina, courthouse for a decade — attracting little, if any, notice. Those records stemmed from lawsuits brought by the relatives of patients who had enrolled in the Duke clinical trials. Duke settled for an undisclosed sum in 2015, although its lawyers argued there was no evidence that patients were harmed by the chemotherapy regimens the algorithms chose.
Advertisement
The same year, a federal investigation concluded that Potti had 'engaged in research misconduct by including false research data' in his published work. To settle the case, Potti, while not admitting wrongdoing, accepted sanctions including strict supervision of any federally-funded research for five years. In reality, he was banished from the academic research world for good. (Potti declined multiple requests to speak with the
Globe.
)
Kornbluth, a cell biologist with a PhD in molecular oncology, is known to close colleagues as a straight shooter who is unimpeachably honest. In the Trump era, she has been a deft operator, promoting MIT as a beacon of scientific inquiry while avoiding the worst of the administration's onslaught against universities. She was the only university president to hold onto her job after the infamous December 2023 congressional hearing on campus antisemitism that contributed to the resignations of Harvard's
MIT President Sally Kornbluth reads her opening statement during a hearing of the House Committee on Education and the Workforce in Washington, D.C., in 2023.
Mark Schiefelbein/Associated Press
But well before that, in the aftermath of the Potti scandal, Kornbluth was a target of fierce criticism, including from scientists inside Duke who felt that she downplayed the warning signs about Potti and prioritized Duke's interests over those of its patients — charges Kornbluth has always denied. Kornbluth and other former Duke leaders have contended they were duped by a con man who knew how to exploit their trust. Their response to the challenges to Potti's research 'was a slow and deliberative process — perhaps too slow and deliberative in the eyes of some — complicated by the active and ongoing deception of the perpetrator,' Kornbluth said in a statement to the
Globe.
(She declined a request for a sit-down interview.)
Advertisement
'Scientific research is a complex, integrated enterprise that requires the engagement and trust of many different collaborators and specialists,' Kornbluth continued. 'Without that trust, the system cannot function.'
T
he first signs of trouble reached Kornbluth in the fall of 2007.
She was in a meeting of Duke deans when one of her colleagues mentioned Joseph Nevins, an eminent cancer researcher and Potti's mentor. 'Hey,' the colleague asked, 'did anybody see that there's this correspondence in
Nature Medicine
going on about some work from Nevins's lab?'
Potti and Nevins had published a key paper about their algorithm in the journal. But now two scientists in Texas, Keith Baggerly and Kevin Coombes, along with a colleague, had sent a letter to
Nature Medicine'
s editor saying that the algorithm 'did not work.' They had tried to reproduce the results, but the numbers just didn't add up.
A challenge to Potti's research would have been very unwelcome to Duke.
At the time, Potti was the school's golden child. An immigrant from India, he completed his medical residency in the United States and crafted an impressive résumé , which listed him as a Rhodes Scholar.
Potti had come to Duke in 2003 and joined Nevins's laboratory. It was a heady time in the cancer research world. Advances in human genome sequencing seemed to open a new horizon of personalized, gene-based medical treatments. The ability to better target chemotherapy — and save patients from some of the ravages of the toxic drugs — was a tantalizing possibility, with vast financial and humanitarian benefits.
Soon Nevins and Potti were publishing blockbuster research articles about their algorithms in the world's top scientific journals:
The New England Journal of Medicine, Nature Medicine, The Lancet Oncology,
and others. The work seemed astounding. The algorithms purported to identify patterns in the behavior of a tumor's genes, making it possible to identify whether one chemotherapy drug or another would be more effective. They seemed capable of taking a lot of the guesswork out of oncology. In a lab setting, the algorithms — they called them genomic predictors — could select the better of two chemotherapy options with stunning accuracy, sometimes exceeding 85 percent, the lab claimed.
This was a coup for Duke. The school was sometimes called the Harvard of the South, but hadn't been able to compete with elite institutions of the Northeast. By the mid-2000s, no scientist had ever won a Nobel Prize for work primarily conducted at Duke. Many believed the Potti and Nevins research finally gave the university a shot. It was also a business opportunity — a big one — that could enrich Potti, Nevins, and Duke itself. Their genomic predictors, an internal analysis at the time estimated, represented 'a potential market of over $2 billion.'
Anil Potti during his time at Duke.
Duke Photography
To reap those rewards, Duke had to prove the predictors worked in the real world — that they could actually help cancer patients. Nevins, Potti, and other Duke scientists launched three clinical trials in 2007 and 2008. In all, more than 100 patients with lung and breast cancer would enroll in the trials and most would have their chemotherapy treatments guided by the Nevins-Potti predictors.
To many patients and their relatives, the opportunity seemed like a godsend. Some tapped their savings to travel from out of state, seeking what they believed to be their last, best chance. The first patient to enroll, Joyce Shoffner, had a dangerous form of breast cancer that was likely to recur. When she learned about Potti's algorithms, she later said, they seemed almost 'too good to be true.'
E
ven as many at Duke saw immense promise in the trials, others were worried — especially scientists who had the most intimate knowledge of the data.
One of them was Holly Dressman, a Duke researcher and coauthor on some of the Nevins-Potti papers. When she tried to reproduce the findings later, she found the predictor's accuracy seemed to be little better than a coin toss.
'I get awful predictions,' she emailed Nevins in 2008, shortly after the third clinical trial had launched.
Bruce Sullenger, another Duke medical researcher, was also worried. In December 2007, he emailed Robert Califf, Duke's vice chancellor for clinical and translational research, and other university officials, attaching the
Nature Medicine
letter by Baggerly and Coombes. Sullenger said he wanted to 'make sure that we are all aware that significant questions still appear to exist ...' He also notified Victor Dzau, then the CEO of Duke University Health System.
Kornbluth would later say that she understood the challenges to the Nevins-Potti work as merely a dispute over methodology, a common occurrence in high-level research. Dzau said the same. '[W]hat we saw was a methodologic controversy ...,' he testified at his deposition in the lawsuit brought by patients and their relatives. Kornbluth and Dzau believed, they would later say, that they were merely witnessing the sometimes messy, sometimes contentious, but ultimately salutary process of scientific inquiry.
But then something far more unusual happened: A whistle-blower came forward from inside the lab.
Text from the memo written by Brad Perez.
Brad Perez, a third-year medical student at Duke, had joined Potti's lab as part of a fellowship in 2007. It was a career-making opportunity. Under Potti's mentorship, he was able to take the lead role on a research paper about the predictors, a huge honor for a student.
But as he drafted the article, Perez grew suspicious. Potti insisted that he drastically pare back the methods section. Perez wondered why Potti wanted to be vague about how they performed their analysis. After Christmas break, he started digging through the data that had been used to develop the predictors. He didn't like what he saw. In some cases, samples — that is, whole columns of numbers — seemed to have been removed without explanation. In others, data appeared to be what he later called 'misrepresented.' It looked to him like someone was cherry-picking data to get a desired result.
By the spring of 2008, Perez had seen enough to confront Nevins and Potti. He hoped they would tell him that he had it all wrong. But they shut him down. They said he had identified 'mistakes' that could be easily corrected and were 'in no way intentional,' according to a letter to him signed by both Nevins and Potti­.
Perez felt overwhelmed by 'the stress of knowing these problems exist' while the clinical trials with patients were underway, he would later say. So he took an extraordinary step. He wrote a detailed, scathing memo and shared it with Duke leaders. Even though it would set back his career, he left the lab and demanded not to be listed as a coauthor on any papers.
'In raising these concerns, I have nothing to gain and much to lose,' he wrote in the 2008 memo, which was first
Other Duke deans took the lead on handling the Perez complaints, but Kornbluth was aware of them, according to emails she exchanged with colleagues at the time. In her deposition, she said she probably did not read Perez's memo until about a year later.
Duke University campus.
Cornell Watson
Even as criticism of Potti's research mounted, his star continued to rise at Duke. The university featured him in TV commercials, such as the one Melissa Niccum saw shortly after her mother, Juliet Jacobs, enrolled in the clinical trials.
At some point in the winter of 2008 to 2009, months after Perez sent his memo, Kornbluth gave a tour of Duke Medical School to a prospective donor, according to her deposition testimony. She took the donor, whose wife had been treated for cancer, to meet Potti in his lab — a kind of 'show and tell,' as she described it.
Potti showed them the small slides used in the genetic testing process. He commiserated with the donor over the plight of cancer patients. Potti could relate, Kornbluth recalled him saying, because he had once been treated for lymphoma himself.
L
ess than a year later, on September 24, 2009, Kornbluth convened a meeting with Nevins and Potti.
The challenges to their work had been escalating, and the two scientists had been forced to issue several corrections to their journal articles. Baggerly and Coombes had also just published their latest broadside, writing in an article in
The Annals of Applied Statistics
that 'errors' in the predictors might be endangering patients enrolled in the trials by putting them on less effective medications.
Then the feds had weighed in. 'Serious concerns ... have come to our attention,' Jeffrey Abrams, a top official at the National Cancer Institute, wrote in a September 22 email. '[I]t is possible that patients are being treated on the basis of flawed genomic predictors ...' Duke officials forwarded the email to Kornbluth the same day.
Two days later, Kornbluth sat down with Nevins and Potti. Another scientist might have appeared chastened, but Nevins went on the attack, according to Kornbluth's recollection of the meeting. Baggerly and Coombes were not being 'collegial,' he said. The whole ordeal had begun with a 'gotcha' when they found some minor errors. Nevins gave her the impression they were on a 'crusade.' (Nevins didn't respond to multiple requests for comment.)
Kornbluth was inclined to believe Nevins, she later said in her deposition. She had known him for more than 20 years, since her time as a graduate student at Rockefeller University in New York. They hadn't been especially close, but in the years since, she had watched his career flourish.
@font-face {
font-family: BentonSansCond-Regular;
src: url("https://apps.bostonglobe.com/common/font/bentonsanscond-regular.woff2") format('woff2'), url("https://apps.bostonglobe.com/common/font/bentonsanscond-regular.woff") format('woff');
}
@font-face {
font-family: BentonSansCond-Bold;
src: url("https://apps.bostonglobe.com/common/font/bentonsanscond-bold.woff2") format('woff2'), url("https://apps.bostonglobe.com/common/font/bentonsanscond-bold.woff") format('woff');
}
.dip3_cap_cred {
font-family: "BentonSansCond-Regular", "Times New Roman", Times, Georgia, serif;
font-size: 14px;
letter-spacing: .5px;
text-align: left;
margin: 3px 15px 0px 0px;
font-weight: 200;
color: #000;
}
.dip3_photo {
max-width: 100%;
height: auto;
padding-top: 15px;
}
.dip3_cap_cred {
width: 100%;
padding: 5px 5px 5px 5px;
}
.pic3__container {
width: 100%;
position: relative;
margin: 0 auto;
}
.dip3__content {
width: 100%;
display: grid;
grid-template-columns: 1fr;
}
@media(min-width: 600px) {
.dip3__content {
grid-template-columns: 1fr 1fr;
grid-column-gap: 40px;
}
}
.cdip3container {
display: block;
width:100%;
height: auto;
margin:0 auto;
padding:1rem 1rem;
-moz-box-sizing: border-box;
overflow: hidden;
}
Retired statistician and scientist Keith Baggerly at home in Houston in June.
ANNIE MULLIGAN FOR THE BOSTON GLOBE
Kevin Coombes in his office at the Georgia Cancer Center at Augusta University in June.
SAM WOLFE FOR THE BOSTON GLOBE
In the unofficial points system that establishes the hierarchy of academic researchers, Nevins was an undisputed winner. He sometimes published a dozen peer-reviewed articles a year. His lab brought in millions of dollars of federal research funding.
Nevins was 'revered' as an 'outstanding scientist,' Kornbluth said. 'There was absolutely no reason for me to question [his] interpretation.'
Baggerly and Coombes were well-regarded statisticians, but they didn't rank highly in the scientific pecking order. They were sometimes treated as hired hands more than pathbreaking scientists in their own right. Their publications tended to focus on the technical aspects of statistical analysis and study design — important work but not the kind that wins major awards or attracts big federal grants.
'Baggerly and Coombes don't do their own real original research,' Kornbluth contended.
By the time of the meeting, Kornbluth had moved into a new role. She was now vice dean of research with operational oversight of clinical trials, an area where she was not an expert. 'While she was very experienced in basic science research' — the kind of work she did as a cell biologist — 'she was very inexperienced in clinical and translational research,' recalls Scott Gibson, then the chief operating officer of Duke Medical School.
Around the same time, Kornbluth's boss, Nancy Andrews, the dean of Duke Medical School, recused herself from decision-making about the controversy. Her husband, a cancer researcher named Bernard Mathey-Prevot, was working with Nevins on another project, so there was a potential conflict of interest. Kornbluth's new role combined with Andrews' recusal left Kornbluth with an unusual level of authority over the controversy.
After her September 24 meeting with Nevins and Potti, Kornbluth was not especially alarmed. 'At this point,' she wrote to colleagues, 'I'm not concerned about the underpinnings/validity of the currently running clinical trials.'
But the number of people who were alarmed was growing. On October 2,
In response to the concerns, Kornbluth and her colleagues hired two outside experts to check Potti's and Nevins's research. It was 'important that we find a middle ground, someone to kind of adjudicate between [the] two sides of the story,' Kornbluth said in her deposition. Duke leaders also barred any new patients from joining the clinical trials — at least until the outside experts completed their review.
But even as Kornbluth and other Duke leaders set up the review, warnings kept mounting.
Images from court documents, Adobe Stock; Illustration by Maura Intemann/Globe Staff
On October 7, Holly Dressman, the Duke researcher, emailed Nevins again. She was still worried about the data and wondering 'does this really work or not,' she wrote. Now she raised a provocative idea: Should they retract the paper they had published in
Nature Medicine
'and start fresh again ...?'
Nevins wouldn't hear it. 'I'm just not sure if you realize what you're suggesting in all of this,' he responded. If they started retracting papers, they'd have to shut down the trials. '[N]ot put them on hold,' he wrote, 'but stop them completely.' Dressman apologized and dropped it.
The next day, Kornbluth received an email from another concerned Duke scientist: Mathey-Prevot.
This was the man whose work with Nevins had forced Kornbluth's boss to recuse herself.
But far from defending his colleague, Mathey-Prevot was withering. A former member of Nevins's lab, Erich Huang, had told him the research 'has fatal flaws.' At least one of their papers should be retracted, Mathey-Prevot believed. He also questioned Nevins's judgment. He was acting, he wrote, like 'a zealot with a mission.'
M
eanwhile, in Texas, Baggerly kept digging. He and Coombes worked at a leading specialized hospital, The University of Texas MD Anderson Cancer Center. They had first scrutinized the Potti-Nevins research papers simply because they looked so promising — researchers at their cancer center wanted to develop something similar.
But Baggerly and Coombes had grown concerned about shoddy statistics and obvious errors almost immediately. A few times, Nevins thanked them for finding what he called 'clerical errors.' But as the problems accumulated and Baggerly's and Coombes' questions became more pointed, Nevins and Potti stopped responding.
The Texas researchers couldn't come up with any explanation except that Nevins and Potti were hiding something. By the fall of 2009, Baggerly says, he suspected the data underlying the predictors had been falsified. Deliberately. But he couldn't prove it.
Then, in the first week of November 2009, he was trawling the Nevins lab's website and found something. Someone had added a new page that contained links to data and spreadsheets. Baggerly clicked through and couldn't believe what he was seeing. Here was some of the key data he'd been seeking for two years.
The spreadsheets contained the data Potti had used to 'validate' some of his predictors. To prove the algorithms worked, Potti had used medical records from past patients with known responses to different chemotherapy drugs. Then he had his algorithms predict which chemotherapy drugs would have been best for each patient and compared the result with what had happened in real life.
Baggerly already had the data from those original patients — it was publicly available. What he'd been missing was Potti's validation data. Now that he could compare the two data sets, something was immediately obvious: They didn't match. Many of the numbers had been scrambled and mislabeled in ways that were inexplicable.
This was, Baggerly believed, the closest thing to a smoking gun he'd found. He spent a couple of days writing a detailed report and, on the evening of November 9, 2009, typed out an email to Kornbluth.
'I hadn't contacted you before ... because I assumed you knew where to find me,' he wrote. 'However, new information has recently come to my attention that I think you should be aware of.'
He explained his findings as clearly as he could. The problems he had discovered 'would seem so severe as to be 'fatal'' to the integrity of the research, he wrote.
The sensitivity labels are wrong.
The sample labels are wrong.
The gene labels are wrong.
All are 'wrong' in ways that could lead to assignment of patients to the wrong treatment.
He attached his report and hit send.
And that would be it, he thought. Kornbluth would read his report and understand its significance. She would share it with the external reviewers. Duke would announce that the science was flawed beyond repair. They would shut down the trials, and the patients would seek care that wasn't compromised by the predictors. Any other outcome was inconceivable.
A
bout 10 weeks later, Baggerly was in an airport when his cellphone rang. It was Paul Goldberg, the publisher of
The Cancer Letter,
the newsletter that had been covering the Duke controversy since the previous fall. 'You're not going to believe this,' Goldberg said.
Goldberg had just received notice from Kornbluth and Dr. Michael Cuffe, another dean, that the external review had come back clean. The outside experts had 'validated this pioneering science,' Kornbluth and Cuffe wrote in an open letter that they asked Goldberg to publish. Now they planned to reopen enrollment in the trials to new patients.
Baggerly, usually mild-mannered, cursed out loud. 'You've got to be kidding me!' he said.
He couldn't understand how Duke's experts could have given the all clear — especially if they had seen his November 9 email. So he and Coombes asked Duke to share a copy of the reviewers' final report. Duke declined.
'At this point,' Baggerly recalls, 'I started feeling somewhat adversarial toward Duke as an institution.'
In the recriminations that would come later, the handling of Baggerly's November 9 report would become a key point of contention and criticism.
Kornbluth had responded to the email by telling Baggerly she would share his report with the external reviewers. Later, in her deposition, she would say that she had intended to send it to them.
Keith Baggerly reviews articles and papers related to the Duke research scandal.
Annie Mulligan for The Boston Globe
But that never happened. Nevins sharply objected to the idea of sharing Baggerly's findings. 'This is insidious and completely unethical from my point of view,' he wrote on an email chain that included Kornbluth, Duke hospital CEO
Victor Dzau, and others. Seeing the report would only 'bias' the outside experts against him, he wrote. The situation was 'perverse' and 'absurd.'
Ultimately, Dzau made the decision to withhold the report. He didn't want to 'bias the review process,' he later said in his deposition, 'and I meant it.' He said that he didn't believe the Baggerly report contained new information, and that sharing it would have the effect of 'flaming the process.'
'I might not have made the same decision,' Kornbluth said in her deposition, 'but it didn't strike me as outrageous.'
'It was really Victor who was in charge,' says Andrews, the then-dean of Duke Medical School who is now a member of the MIT board. 'They made mistakes. Some of the mistakes were because I think Victor was too ready to believe Joe Nevins in particular and too excited about the possibilities' of the research.
The outside review itself would come under withering criticism as details of it became known within Duke and the wider scientific community. Duke had commissioned two outside statisticians to conduct it, and Kornbluth and Cuffe later said the intention was to give them 'unfettered access' to Nevins's and Potti's data and lab. But critics would come to see the review as a stage-managed affair. Rather than having unfettered access, the reviewers received their information from Nevins and Potti — Potti himself compiled the data. It was as if the suspects in an investigation were responsible for selecting the evidence for the detectives.
Potti and Nevins also gave the outside reviewers a written rebuttal of the Baggerly and Coombes critiques, and delivered an in-person presentation. The reviewers never heard from Baggerly and Coombes directly.
Baggerly and Coombes knew nothing of the internal deliberations or how the review had been conducted. All they could see was that they had sent what they believed to be damning evidence to key decision-makers and nothing had happened. They felt defeated.
I have no idea what we can do beyond this to actually get the right outcome for patients,
Coombes recalls thinking at the time.
In the meantime, Baggerly says, Potti's validation data had disappeared from the lab's website.
T
wo weeks after the trials reopened, in February of 2010, Juliet and Walter Jacobs walked into Potti's exam room for their first meeting. They hadn't been told anything about the controversy.
Like others, they left their meeting with Potti charmed. They were sure that he could give Juliet the best chance to extend her life.
The first step was a biopsy to extract a piece of a tumor for genetic analysis. Getting a sample meant that Juliet would have a large needle plunged into her neck and through one of her lungs.
It would be horrible. At least one Duke doctor had refused to perform some biopsies for the clinical trials because he thought they could cause a collapsed lung. But to the Jacobses, the risks seemed worth it — a necessary price to pay for access to Potti's predictor.
Walter and Juliet Jacobs in 2007.
From the Jacobs Family
In the weeks after the biopsy, however, pain from the puncture in Juliet's neck radiated downward to her shoulder, her arm, and her back. She lost some mobility. It was hard to sleep.
By early March, Juliet was even weaker than she'd been when she met Potti just a few weeks earlier. But she was still confident she had made the right choice. She went back to the hospital with Walter for her first infusion of the chemotherapy chosen by the predictor.
The couple allowed themselves to imagine that Juliet might survive for years thanks to Potti's predictor. Maybe she would even get her breath back, and return to playing tennis. There was another benefit, too, a less tangible one. They believed they were helping future patients by participating in the clinical trials, that Juliet's pain and disease could serve a greater good.
But that vision was a mirage. Within months, Juliet would be dead and the promise of Potti's breakthrough would collapse.
B
y the spring of 2010, Potti's endeavor had survived challenges from internal critics, outside scientists, and an intervention from the federal government. Many of the skeptics felt there was little more they could do. But then, around April, an official at the National Cancer Institute was reviewing old paperwork and spotted something strange. The NCI, it turned out, was providing funding for one of the clinical trials at Duke through a grant to Potti's lab — a fact Potti had not properly disclosed. (In a letter to the NCI, Potti called that an 'oversight.')
This wasn't a mere clerical matter. The fact that its grant money was being spent on the trials meant the NCI wasn't just a concerned third party, but had actual authority. It could now demand information from the lab, not just request it.
For Lisa McShane, an NCI statistician, the newfound authority changed everything. She had been scrutinizing Nevins's and Potti's work for years. In the fall of 2009, she had spoken with Kornbluth directly about her concerns, telling her in a September conversation about 'inconsistencies' in the data and following up in a January email to say she and her colleagues 'continue to observe puzzling findings.'
She felt her concerns had not broken through.
But now, there was something more she could do. She had come to believe that Potti's algorithms were based on data manipulation — fraud, in other words. And she wanted to prove it.
Throughout that spring, she demanded that Potti show his work. Bit by bit, she extracted nearly every detail about his data and methodology. 'Please provide more specific information about the variables associated with the 30 Gyorffy cell lines,' she wrote to him at one point.
Potti, meanwhile, presented himself as just a 'junior investigator,' who was 'not entirely sure of the process involved here,' he wrote. '[P]lease pardon any ignorance or naivety ...'
As McShane worked, Kornbluth lost patience. She knew that McShane knew Baggerly and Coombes, and she didn't trust her judgment. 'We are under the strong impression that McShane is biased,' she wrote to Potti and Nevins. She planned to ask the NCI to have a different statistician assigned to the case, she told them.
At the time, Kornbluth thought McShane was acting like 'a dog with a bone,' she recalled in her deposition.
But other Duke leaders were far more skeptical of Nevins and Potti. For months, a communications executive, Douglas Stokke, had been warning colleagues they shouldn't trust the scientists. In a March email to Kornbluth, he referred drily to 'their pattern of less than transparent communication.'
Califf, the Duke vice chancellor, put it more directly. 'You guys had better wake up,' he recalled writing in an email to Kornbluth and other Duke Medicine leaders. In May, Califf pulled Kornbluth aside for a face-to-face meeting. His feedback was unsparing. Kornbluth, he believed, had been 'blinded.' The external review had been inadequate, he said. A much deeper investigation had been called for.
Sally Kornbluth answers a question during a press conference held to introduce her as MIT's 18th president.
Jessica Rinaldi/Globe Staff/File
Kornbluth was facing yet another person taking shots at her judgment. 'I'll be fine,' she wrote in an email to colleagues after the meeting. 'I just have to cool down.'
Months into McShane's investigation, Kornbluth felt like the challenges would never end. She recalled telling McShane, 'Look, you don't know me, but enough is enough already. We're — you know, we're trying to do things the right way. If you have a problem with the way things are functioning, let's just sit down and work it out.'
Later that month, Kornbluth, Nevins, Potti and others traveled to the NCI's campus in Rockville, Maryland. They sat down in a conference room with McShane and other officials. McShane booted up a slide show.
What followed was an interrogation.
McShane grilled Potti about discrepancies in his data. She pulled up a slide that tracked the way his data sets had changed over time. Potti squirmed. He gave feeble explanations. Maybe he had sent the wrong version of the data, he offered at one point. When Potti gave evasive answers, McShane pulled out documents and quoted his own words back to him.
Kornbluth watched, astonished. It was like something 'out of a TV show,' she said in the deposition. 'She basically nailed him to the wall.'
In her deposition, Kornbluth said that she and other Duke leaders did not consider shutting down the trials at that time. But McShane and her colleagues could now force their hands. They gave Duke one last chance to prove the predictors worked. The task fell to William Barry, a statistician who had collaborated with Nevins and Potti.
That work was interrupted by a final challenge to Potti's credibility. It had nothing to do with the trials, or statistical methods, or corrupted data sets — it was about Potti himself.
The table of contents from a 2010 issue of The Cancer Letter.
From court documents
In mid-July,
The Cancer Letter
published
The news rocketed through the cancer research world. The Duke student newspaper ran articles about it. Scientists from Harvard, Johns Hopkins, and other top research universities sent a joint letter to the head of the NCI saying that the clinical trials should be shut down.
Inside Duke, the revelation about Potti's résumé seemed finally to spur decisive action, three years after challenges to Potti's work first emerged. Within days, the medical school placed Potti on leave. The deans launched an internal scientific misconduct investigation. They shut down enrollment in the clinical trials.
In her statement to the
Globe,
Kornbluth said, 'The decision to suspend clinical trials was based on several factors.' The résumé discrepancies, however, 'were important because they were indisputably fabricated, and thus impossible to paint as a difference of opinion. . . . [T]hey made it clear that Dr. Potti was not trustworthy, and that was a relevant and significant piece of data.'
Shortly after Potti was suspended, he and Kornbluth passed each other on campus, she recalled in her deposition. They made eye contact, and then Potti looked away.
O
n the morning of November 1, 2010, Walter Jacobs walked out of his home in Chapel Hill to pick up the newspaper. He had been a widower for half a year — Juliet had died in May. Potti's predictor hadn't saved her. It didn't even seem to have bought her a little extra time.
Now, after nearly 50 years of marriage, Walter was alone, bumping around the house they had shared, smoking cigars because he no longer had to worry about Juliet's lungs, lying awake at night contending with what it meant to be depressed.
He carried his newspaper inside and spotted a headline on the front page:
Jacobs called his daughter, Niccum, and asked if she knew what this all meant. She had no idea. Duke had never told them anything was wrong with the trials.
At Duke, officials had been questioning everything about Potti's stories. At one point, Potti had said his wife had cancer. 'Actually [she] was perfectly healthy and had thrown him out of the house,' Kornbluth alleged in her deposition. Some even began wondering if Potti had really had lymphoma, according to Kornbluth's testimony, as he had told the donor who visited his lab. (Potti did not respond to a question about this.)
Walter Jacobs' wife, Juliet, died of lung cancer a few months after enrolling in a Duke study that was later found to be based on fraudulent research.
Kate Medley for the Boston Globe
Barry's assignment shifted. Instead of trying to prove the predictors were legitimate, his mandate was to scrutinize Potti's work. He went back to the original, raw data and compared it to the data Potti had given to the external reviewers, to McShane, and to statisticians on the clinical trials. Nevins helped with the search. What Barry found, according to an account of the scandal later written by Duke leaders, was devastating: 'mismatches' and 'non-random mislabeling.' Those were euphemistic terms for something Baggerly, Coombes, McShane, Perez, and others had long suspected: fraud.
Potti has always denied tampering with data. But several Duke leaders — including Nevins, Kornbluth, and Dzau — later said under oath that they believed the data had been manipulated to make it appear the predictors worked. Eventually, an investigation by the US Department of Health and Human Services concluded that Potti 'engaged in research misconduct by including false research data' in research papers, a grant application, and the 'research record.'
Potti's algorithm was fake: Rather than a Holy Grail, it was no better than a random number generator.
After Barry confirmed the data problems, Nevins's and Potti's blockbuster papers about the predictors were retracted.
The Institute of Medicine, a nongovernmental organization viewed as an authority on best scientific practices, had begun a review of the field, including what had happened at Duke. In a statement, Dzau said he requested that the institute 'conduct a transparent and independent study ... to provide lessons learned for the broader research community.' The institute was later renamed the National Academy of Medicine, which Dzau now leads as president.
As the institute's review began, Nevins appeared to still be attempting damage control. He implored Kornbluth not to share the original data with the institute, according to emails cited during Kornbluth's deposition. That is, he wanted her to withhold the single most damning piece of evidence in the whole affair.
Now, Kornbluth stood up to him. '[I]t is NOT going to be possible,' she wrote in an email. 'We would be pilloried if we allowed the IOM to do extensive analyses using flawed data.'
Stokke, the communications executive who had long distrusted Nevins and Potti, wrote, 'At the risk of sounding politically incorrect and harsh, I think an abundance of patience and deference has been shown to Joe and Anil ...'
As the Nevins-Potti research unraveled, the full scope of the challenges to their work became clear to many Duke scientists for the first time.
For some, the most shocking revelation was the existence of the November 2009 report Baggerly had sent to Kornbluth, which Dzau had decided not to send to the external reviewers.
'When Bill [Barry] disclosed to me that he was NOT EVEN AWARE of the November, 2009 document until October of 2010, I fully realized the extent of concealment,' Liz Delong, chair of Duke's department of biostatistics and bioinformatics, wrote in an early 2011 email to Barry and Califf. 'It is clear that there was significant effort to sweep the challenges under the rug and to try to disclose as little as possible.'
Bruce Sullenger, the Duke researcher who had tried to make leaders pay attention to Baggerly and Coombes three years earlier, was furious. 'Rob, I hope that the IOM process is open and transparent to all and not like the previous coverup/white wash institutional review of the Potti case,' he wrote to Califf. 'Many of our patients, faculty including me now, and peers have lost confidence that certain Duke Medicine leaders will not put its sacred cows ... above the welfare of our patients and scientific knowledge.'
Others came to a more philosophical view. The whole ordeal was a kind of tragedy with only one villain at its center: Anil Potti.
For Huang, the former Nevins colleague who warned the algorithms had 'fatal flaws,' Potti was the consummate con man. 'He had a talent for saying the right things to the right people at the right times,' he said in a recent interview. Nevins would later say he had been 'deceived' by 'someone that I viewed as a friend.'
Califf, the vice chancellor who was critical of Kornbluth and other Duke leaders, said in a deposition that he didn't believe there was 'some purposeful effort to not get to the truth ... I think they just botched it.'
Kornbluth herself has conceded that she and other Duke leaders 'may have been naive.' But she has always denied that there was any intent to cover up wrongdoing. 'We may have done things that weren't, you know, the smartest in the world,' she said in her deposition. 'But, that's, you know, 20/20 hindsight.'
After the Potti scandal, as well as two more research controversies that did not involve Kornbluth, the National Institutes of Health placed grants to Duke researchers under a strict oversight program. Meanwhile, Duke implemented new policies governing how researchers stored data. Kornbluth announced plans to place a biostatistician — that is, a person like Baggerly or Coombes — in every clinical research group.
It is unclear if any of these measures could have averted the Potti scandal. 'The reality is if somebody wants to lie and deceive, it's very hard for policies to stop them,' says Andrews, the former Duke Medical School dean.
Jonathan Kimmelman, a McGill University expert on clinical trials, said that not much has changed since the scandal. Ethics panels that greenlight trials are still 'basically taking investigators at their word when they say they've got good science behind their study,' he says.
'There are researchers who cut corners, who are so captivated by the allure of professional advancement that they end up doing research that just shouldn't be done,' he adds.
But he also sees an alternate reading of the Duke story. 'There are people who are so deeply committed to the ideals of science that they spend their spare time looking for work that doesn't fully adhere to what they consider to be [those] ideals,' he says. 'That's partly why science is self-correcting.'
'I learned a lot from this case,' Kornbluth said in her written comments to the
Globe,
'and the changes we implemented at the time reflect that. Of course, one thing to keep in mind is that the real-time information you use to make decisions in these situations can be imperfect, messy and even contradictory.'
In 2011, at least two surviving patients from the clinical trials and the relatives of many more who had died sued Duke University Health System, Nevins, Potti, Kornbluth, and others. Many of them said they felt betrayed. Duke 'had every reason and every warning to know that this fraudulent program was going on,' Walter Jacobs said in his deposition. They 'not only did not act then to hold anybody accountable, but continued to try and cover up that program to the detriment of past and future patients.'
In 2015, Duke settled for an undisclosed sum.
Kornbluth changed roles at Duke in 2011. She stopped overseeing clinical trials at the medical school and returned to her previous role as a dean in charge of basic science, that is, research that does not involve human subjects. In 2014, Duke promoted her to provost, the second highest administrative role at the university.
Sally Kornbluth at her inauguration as MIT's President in 2023.
Suzanne Kreiter/Globe Staff/File
In 2022, MIT's longtime president Rafael Reif announced he would step down, and the school's board began the search for his successor. Early in the interview process, Kornbluth 'proactively raised the Potti case,' Diane Greene, the then-chair of MIT's board, told the
Globe.
'As part of its due diligence, the search committee conducted extensive reference checks with Duke leadership and the broader community, all of which were uniformly positive,' she said.
When Kornbluth was selected, Lily Tsai, then chair of MIT's faculty and a member of the presidential search committee, described her as 'a bold leader with exceptional judgment' and 'experience handling crises with wisdom and calm.'
The Potti scandal effectively ended Nevins's scientific career. He retired from Duke in 2013 and never returned to the heights of academic research.
Potti left Duke in late 2010 and joined an oncology practice in South Carolina as a doctor. After a
North Dakota seems to suit him. In the bio, he writes, 'sure, the weather may be cold, but the people sure are warm.'
Mike Damiano can be reached at