Journal withdraws article after complaints from drug manufacturers

A paper suggesting that two new antidiabetes drugs could greatly increase the risk of pancreatitis and several cancers has been withdrawn from the website of the journal Gastroenterology after complaints from Novo Nordisk and Merck, the drugs’ manufacturers.

The companies wrote letters to Anil Rustgi, the journal’s editor in chief, after the paper appeared online. They expressed concern that the analysis, which used data derived from the US Food and Drug Administration’s adverse event reporting system, reached conclusions that could not be justified. Merck warned that the paper could have a negative effect on the care of patients, while Novo Nordisk claimed that it could spark an unnecessary health scare.

One of the authors of the paper, Peter Butler, of the David Geffen School of Medicine at the University of California at Los Angeles (UCLA), said that the drug companies wanted the paper deleted. Gastroenterology’s website initially recorded the paper as “retracted,” which might imply that the authors themselves had chosen to remove it. This was not the case, Dr Butler said: the decision had been taken by the editors. The journal has since changed the paper’s status from “retracted” to “withdrawn.”

If unchallenged, the paper could have severe consequences for the two drugs. The German Diabetes Society has already said that new patients should be treated with the drugs only in very special circumstances and that all patients currently being treated with them should be told about the findings.

The two drugs involved are Novo Nordisk’s exenatide (marketed as Byetta) and Merck’s sitagliptin (Januvia). They belong to a class called incretin mimetics or glucagon-like peptide 1 (GLP-1) based therapies and aim to give better control of type 2 diabetes.

In the disputed paper the team responsible (Michael Elashoff, Aleksey Matveyenko, Belinda Geir, Robert Elashoff, and Peter Butler, from the Larry L Hillblom Islet Research Center at UCLA) used the FDA database of adverse events from 2004 to 2009 to compare the number of cases of disease reported by users of the two drugs with the number reported by users of four other antidiabetes drugs.

They concluded that the risk of pancreatitis was increased by 6.8 times in users of sitagliptin (95% confidence interval 4.7 to 10.2) and by 11.8 times (8.5 to 16.6) in users of exenatide. For pancreatic cancer the risk ratios were 2.4 and 2.0, respectively, and for thyroid cancer 3.4 and 7.6. There were 970 pancreatitis events among patients taking exenatide and 131 among patients taking sitagliptin. The number of cases of pancreatic cancer was 72 among patients taking exenatide and 14 among patients taking sitagliptin. The paper did not report confidence intervals for the risk ratios for pancreatic cancer.

They conclude: “These data are consistent with case reports and animal studies indicating an increased risk for pancreatitis with GLP-1 based therapy. The findings also raise caution about the potential long-term actions of these drugs to promote pancreatic cancer.”

The paper went online on 21 February. On 6 March Mads Krogsgaard Thomsen, chief science officer for research and development at Novo Nordisk, wrote a five page letter to Professor Rustgi expressing his “considerable concern.” The FDA warned, he said, that its database was not intended for drawing inferences about rates of adverse events and should be considered alongside evidence from randomised controlled trials and matched cohort studies.

Two days later Nancy Thornberry and Barry Goldstein from Merck Research Laboratories wrote a letter to Professor Rustgi couched in similar terms. The adverse event reporting system database, they said, cannot be used to quantify risk or calculate the incidence of an adverse event, for several reasons: the absence of a reliable denominator; the uncontrolled, voluntary nature of the submitted reports; the incomplete nature of these reports, including lack of important information on confounding factors; and the greater tendency for adverse events of new drugs to be reported, as opposed to older drugs.

The team responsible for the paper acknowledges the limitations of the database and tried to get around them by comparing outcomes of interest, such as pancreatitis, against the control drugs and against control events (such as back pain, urinary tract infections, and cough) for which it was presumed that no link between drug and event existed. But this is illegitimate, say the Merck scientists, because serious events such as pancreatitis are more likely to be reported than trivial ones.

Dr Thomsen says that his company’s attempts to reproduce the analysis for exenatide in the period before 2007 (when the FDA warned about a potential rise in the incidence of pancreatitis among patients taking it) had failed to reach the same conclusions. In contrast to the findings of Elashoff and colleagues, the Novo Nordisk analysis showed no increase in pancreatitis—a discrepancy “we are unable to explain,” he said.

In response to an inquiry from the BMJ Professor Rustgi said that the paper would be published in the July print edition, together with an editorial. In Gastroenterology, as in many journals (but not the BMJ), putting a paper online doesn’t count as definitive publication: that occurs only when the print version appears and replaces the previous online version.

Dr Butler said he was pleased that the two drug companies had not achieved “complete deletion” of the paper.