Clinical Trials: Fraud and Ethics Charges Hit Stroke Drug Trial

Martin Enserink

Martin Enserink is a science writer based in Amsterdam.

The trial was designed to study the effect of dipyridamole, also known by its brand name Persantin Retard, in preventing a recurrent stroke in patients who have already suffered a previous stroke or a transient ischemic attack, a very mild stroke with symptoms lasting less than 24 hours. Currently, aspirin is the only generally accepted drug for this purpose, but it prevents recurrent strokes in roughly one in five patients only. Researchers had failed, so far, to find any significant improvement on aspirin. The CAPRIE study, a huge multicenter trial published in The Lancet in November, showed that another potential drug, clopidogrel, was only marginally more effective than aspirin.

Like aspirin and clopidogrel, dipyridamole is an antiplatelet drug--it prevents blood from clotting and thus cerebral arteries from getting blocked--and has been a mainstay of the German drug giant Boehringer Ingelheim for more than 15 years in the treatment of all cardiovascular afflictions, although its efficacy has been debated for many years. The results of the Boehringer-sponsored ESPS 2 trial indicate, however, that the drug has great potential in preventing recurrent strokes. Dipyridamole alone reduced the risk of a new stroke by 16% during the 2 years the patients were followed; aspirin, which was also tested, gave a risk reduction of 18%; but combined the two lessened the chances of a stroke by 37%, with relatively few side effects. The death rates in all three treatment groups were about the same, however, and did not differ significantly from that of the placebo group.

But the encouraging news in the ESPS 2 results was soured by the paper's guarded suggestions of data problems discovered during the study. The paper refers to "serious inconsistencies in patient case record forms and compliance assay determinations" at one center. Armand Lowenthal of Middelheim Hospital in Antwerp, Belgium, who chaired the trial's coordinating committee, says that a thorough examination by an independent lab revealed that some of the 438 patients enrolled in the trial at that center turned out to be "entirely made up."

Although Lowenthal declined to name the doctor at the center of the allegations, others have identified him as H. J. Gelmers of Twenteborg Hospital, a regional clinic in Almelo, the Netherlands. Gelmers was originally a member of the trial's protocol and publishing committee, but left the committee after the alleged fraud was discovered in 1993; his resignation is recorded in a 1995 paper by the ESPS 2 study group in the Journal of the Neurological Sciences. In this month's paper, Gelmers was not listed as one of the authors and Twenteborg Hospital was not listed among the 59 institutions whose data were used in the study.

Gelmers denied in a brief telephone interview with Science that he fabricated data in the trial and said of alleged problems with blood samples: "It's more probable that this happened elsewhere." Gelmers said he was not allowed to comment further and referred questions to H. J. Sissingh, the director of Twenteborg Hospital. Sissingh says that he was notified of the case some time ago by Boehringer Ingelheim. "We subsequently had some conversations and carried out an investigation, but we were unable to establish either guilt or innocence. So we left it at that," says Sissingh.

Chris Verhorst, medical director of Boehringer Ingelheim Netherlands, says the trial's organizers first became concerned about the large number of patients enrolled by a single physician and the unusual dedication with which they took their daily medication. Suspicions grew, says Lowenthal, when blood pressure data submitted by the physician turned out to be distributed along a perfect Gaussian curve--which is highly unlikely in a patient sample of this size.

According to the trial protocol, patient compliance was tested by assaying drug levels in plasma from 15% of the patients. Lowenthal says analysis of the suspect samples showed that all of them came from just two individuals, and all contained both drugs, which would have been very unlikely because only one in four patients got a combination of both drugs. The concentrations of the drugs also varied widely, reaching impossibly high levels in some of the samples, Lowenthal said.

Asked why the problems with the data went undetected for years, Verhorst says this was an old study, and "monitoring was less careful and less frequent than it would be today. ... But we did discover the irregularities thanks to the monitoring." Verhorst says Boehringer Ingelheim paid participating centers about $1500 for each patient who completed the trial, which would have amounted to about $640,000 for the suspect data. Verhorst says that payments were stopped after the alleged fraud was discovered, and as far as Boehringer Ingelheim is concerned, the case is closed--the company decided not to claim its money back.

The Dutch Association for Neurology may not let the matter rest, however. It is intending to set up an independent committee to investigate, says Marianne de Visser, chair of the association. "This is something that bears on the reputation of Dutch neurology," she says. "There may be something wrong, but we will investigate the matter very carefully before drawing conclusions."

While the neurology association investigates the implications of the alleged fraud for Dutch neurology, researchers in several countries have expressed ethical concerns about the ESPS 2 trial. During the trial, patients were divided into four groups, which were given different treatments: dipyridamole alone, aspirin alone, a combination of the two, and a placebo. A placebo is generally used only when there is no proven effective drug to test against, but critics argue that aspirin had already been proven effective in reducing the risk of recurrent stroke.

Clinical epidemiologist Michael Gent of Canada's McMaster University in Hamilton, Ontario, who was the principal investigator in the CAPRIE study, says that "When they started this study in 1988, there was clear evidence from a number of published studies that antiplatelet drugs, particularly aspirin, prevented stroke ... in these patients." The 1649 people in the placebo group were unnecessarily exposed to a higher risk, he says.

Erasmus's Koudstaal says that there may have been room to doubt aspirin's efficacy at the beginning of the trial. But later studies, such as the 1991 Swedish Aspirin Low-dose Trial (SALT) and the "aspirin papers," a series of meta-analyses published in the British Medical Journal in 1994, "convincingly proved" that aspirin worked, he says, and should have been reason to stop the placebo arm of the trial. Yet, the study continued until March 1995. "It was probably a mistake at the time," adds Charles Warlow, a neurologist at Western General Hospital in Edinburgh, U.K. "I personally wouldn't have put people in the trial. But there are many perfectly respectable people who did. People vary in what they regard as definite evidence."

Verhorst and Lowenthal dismiss these criticisms, arguing that the efficacy of aspirin, as well as the optimal dosage, was not known when the trial started. Says Verhorst: "The study was submitted to ethical committees in every center. If 59 ethical committees approve, then there is no ethical problem." And Lowenthal says that the trial was monitored by a central ethics committee, which discussed the study each year. It reported in 1995 that use of placebos was justified because of doubts about aspirin's efficacy and side effects. However, Verhorst confirms that the study was turned down by The Lancet largely due to ethical concerns.

Neurologists are now debating whether the results of the study should guide clinical practice. Some previous studies have found no benefit from the combination of dipyridamole and aspirin compared with aspirin alone in stroke prevention, although one other study, published 10 years ago by the same group, reported very positive results with a smaller number of patients. A meta-analysis of all relevant trials will be needed to give a definitive answer. "There is some discomfort about this," says Gent, "but the results are interesting."