Doctor doctors data – surgeon Roger Poisson admits falsifying data in landmark breast-cancer study – special issue: 1994 – the Year in Science
Yet another wayward researcher, playing fast and loose with his data, has sullied the name of science. In March the Chicago Tribune reported that Roger Poisson, a Montreal surgeon, had submitted false information to a landmark study on breast cancer. The study had apparently proved that for many women with cancer in its early stages, mastectomy – removing the entire breast – was not necessary; removing only the cancerous lump appeared to be just as effective in preventing the disease from recurring.
The findings, published in 1985 and 1989 in the widely respected New England Journal of Medicine, changed medical practice almost immediately. Mastectomy, once the standard treatment, was replaced by lumpectomy in tens of thousands of cases.
Hence the shock waves when Poisson’s actions were revealed – he had enrolled 354 of the 2,163 patients in the breast-cancer study and had falsified data on 6 of them. Furthermore, that deception was part of a larger pattern of fraud: between 1977 and 1990, Poisson apparently faked data in 13 other clinical trials on cancer, all of them part of the National Surgical Adjuvant Breast and Bowel Project (NSABP), which was funded by the U.S. govermnent. In all, investigators found 115 instances of falsified data involving 99 of Poisson’s patients.
News of Poisson’s altered data cast doubt on the entire mastectomy study, and women who had refiised mastectomy began to fear their lives might be at stake. Fortunately, a reanalysis of the data without Poisson’s patients has since confirmed the original study’s findings, but the public perception of large-scale clinical trials may never be the same.
As it is, the fraud generated a huge uproar, complete with congressional hearings. Poisson was barred from future studies paid for by the United States and fired from his position as head of cancer research at Saint-Luc Hospital in Montreal. Worse, he took odiers down with him: Bernard Fisher of the University of Pittsburgh, who has had a long and distinguished career in cancer research, was removed from the post he’d held for 3 5 years as head of the NSABP, in response to accusations that he had failed to investigate and report the fraud in a timely fashion. Fisher is angrily disavowing Poisson and fighting to be reinstated; his colleagues on the project are supporting him.
Hoping to avert filrther scandals in medical research, the National Institutes of Health recently set up a Clinical Trials Monitoring Branch, which will resort to both scheduled and surprise audits at study sites to keep tabs on researchers. Indignant researchers say those visits may prove costly and time-consuming, not to mention degrading: the intellectual equivalent of being frisked. They’ve already labeled the auditors “data police.”
Meanwhile, Poisson, the man behind the brouhaha, argues that his falsifications were merely “white lies,” bits of rule bending he knew wouldn’t interfere with the studies’ goals. He changed his patients’ records, he claims, only to make them appear eligible for studies they otherwise would have been barred from entering. In the mastectomy study, for example, biopsy dates were altered – in four of the six cases by a month or less – to make patients’ diagnoses appear more recent. Most of the changes in the other studies were similar, involving dates of diagnoses or the results of lab tests used to categorize a patient’s cancer.
Poisson says he wanted to help patients get into studies that he felt would provide them with better treatment. “I believed I understood the reasons behind die rules,” he told investigators, “and I felt that the rules per se were made to be understood and not necessarily blindly followed.”
Dwight Kaufman, a deputy director at the National Cancer Institute, called Poisson’s arguments “farcical.” Others point out that the prestige and grant money derived from enrolfing patients in these trials belie Poisson’s protests of having fibbed with the noblest of intentions. But in truth, Poisson didn’t harm any patients (though one woman with a heart condition was enrolled in a trial that involved potentially cardiotoxic drugs) or lie about treatment results. And because the records were altered before the patients were randomly assigned to one treatment group or another, Poisson’s fraud should, in theory, have been split equally between them and thus shouldn’t have skewed any of the studies toward a particular form of therapy.
In hindsight, Poisson says he should have tried to have the study criteria changed officially. “I would not consider for a minute altering anything again,” he says.
Of course, it’s doubtful he’ll ever get that chance.
COPYRIGHT 1995 Discover
COPYRIGHT 2004 Gale Group