The editors of the New England Journal of Medicine yesterday accused the authors of an article on the painkiller Vioxx of underreporting the number of heart attacks that occurred in a study published five years ago.
Although the article did report a higher rate of heart attacks in Vioxx users compared with people taking another painkiller, the unreported cardiac events "call into question the integrity of the data . . . in this article." The editors asked the authors to submit a correction.
Vioxx was removed from the market last year by its maker, Merck & Co., after evidence emerged that it increased the risk of heart attack in some users. The question of when Merck discovered Vioxx had that effect is a major point of dispute between the company and some Vioxx critics.
As of late September, heart attack victims or relatives had filed 6,500 lawsuits against Merck, alleging that the company should have stopped selling Vioxx earlier, or at least warned patients of the drug's hazards. Two trials are complete; Merck won one and the plaintiff the other. A jury in Houston is deliberating in a third case.
Merck is facing potentially huge losses, and last month announced plans to cut 7,000 jobs and close five plants worldwide.
In a statement released last night, the company said it did not suppress any data about heart attacks in the study. It said the three heart attacks the journal editors say should have been mentioned occurred "after the pre-specified cut-off date and therefore were not included in the primary analysis."
The statement further said the three heart attacks were reported to the Food and Drug Administration in 2000, and presented publicly to an FDA advisory panel in February 2001.
The article does not name a cutoff date. However, it does say that the analysis included only events that occurred during treatment or within 14 days after treatment stopped.
Asked if the journal knew about a cutoff date, Executive Editor Gregory D. Curfman said: "We never knew about it, and it was never in any version of any version of the manuscript that we had." He added that even if Merck had reported only heart attacks that occurred before a certain date, "we would have expected an update."
A company document that emerged last month in one of lawsuits showed that two of the article's 12 authors knew about the three additional heart attacks in July 2000. That was months after they submitted the first draft of the article to the journal, but two months before it was accepted in September. It was published on Nov. 23, 2000.
"Even though it's a small number of heart attacks, it really does have a substantial impact on the conclusions from the entire data set," Curfman said.
The lead author of the article, Claire Bombardier of the University of Toronto, reiterated the points made in the Merck statement, and added that "the authors of the paper will be preparing a response" to the editors' charges.
The study that Bombardier led was trying to determine whether patients with rheumatoid arthritis who took Vioxx developed fewer ulcers and stomach hemorrhages than people taking naproxen, a nonprescription painkiller.
The article said that a higher percentage (0.4 percent) of those taking Vioxx had heart attacks compared with those taking naproxen (0.1 percent). However, it did not say how many of those events were in each group.
It turns out the numbers (rather than just percentages) were in an early version of the paper, but the authors had removed them. This was discovered in November 2004 when the journal's managing editor, Stephen Morrissey, found a computer disk in the article's file that had not been examined.
The disk contained an electronic version of the study that Bombardier and her colleagues had submitted in paper form on May 18, 2000. However, when the cursor was pointed over the text, changes made by the authors, and the dates they were made, appeared.
On May 16 -- two days before the paper was submitted -- one of the authors had removed the information describing the number of heart attacks. There were 17 in the Vioxx group and four in the naproxen group. The journal editors subsequently learned there were three more in the Vioxx group, making the final tally 20 vs. four, an even more striking disparity.