At the forefront of this week's medical news is a surprising clash between America's premier medical journal, the New England Journal of Medicine, and the nation's most important business paper, the Wall Street Journal. These publications are dueling over a problem at the highest junction of medicine and business: the Vioxx crisis. At issue is whether the New England Journal's editors were sufficiently diligent five years ago when they published the paper that described the benefits of Vioxx, before the drug's fall from grace. It's a serious charge with some merit, but the WSJ's argument is also flawed.
The back story is about pain relief. Over time, two main classes of anti-inflammatory medicines have evolved: corticosteroids and nonsteroidal anti-inflammatory drugs, or NSAIDs. The steroids are very powerful and effective. But they work by disabling some of the body's crucial defensive mechanisms, so they often have dangerous side effects, especially if taken for any length of time.
The other group of drugs—the NSAIDs—are also quite good at relieving the effects of inflammation. And in general, they're less risky. This group includes familiar products like aspirin and ibuprofen (Advil and Motrin). These older pain-relievers decrease inflammation but often cause a serious side effect: gastrointestinal irritation that can sometimes cause fatal bleeding in the stomach or intestines. Vioxx and Celebrex (and, more recently, Bextra) are also NSAIDS. They were developed as powerful anti-inflammatories that wouldn't cause the gastrointestinal complications. These drugs are called COX-2 inhibitors, because they interfere with the action of an important enzyme, cyclooxygenase-2, which begins the cascade in the body that results in inflammation and pain.
Before the COX-2 drugs were released, the expectations for them were enormous—both among patients who anticipated symptom relief and in the financial markets, which sensed huge opportunity. Merck, the maker of Vioxx, was a highly admired member of the pharmaceutical industry—a company known for the quality of its research and products, as well as for its sometimes altruistic policies. (Disclosure: My daughter-in-law is a microbiologist who works for Merck.) Merck's corporate culture was more like the culture of medicine than the culture of industry. Perhaps for those reasons, at least in part, Merck was not as well-regarded on Wall Street as some other drug firms. So, a potential blockbuster drug was tremendously important to the company. Vioxx was released with great fanfare and early success, accounting on its own for 20 percent of the company's profits in 2003, the year before it was withdrawn from the market.
Then the worm began boring through the apple. To test the thesis that Vioxx might help prevent bowel cancer, researchers sponsored by Merck launched a study of 2,500 patients during the course of which it became apparent that whatever the drug's effect on cancer risk, it clearly increased the risk of heart attacks and of stroke. As a result of these findings, Merck took Vioxx off the market, leading to an estimated loss of $28 billion in market value for its stock. (It's worth noting that this increased cardiovascular risk appears to be a class effect that applies to all COX-2 drugs, not just Vioxx.)
It's a different, earlier study that's at the heart of the controversy between the New England Journal of Medicine and the Wall Street Journal. This study, published in the New England Journal in November 2000,was conducted to examine how well Vioxx prevented gastrointestinal side effects—but also showed that patients given this drug were at greater risk for heart attack. The published paper counted 17 heart attacks that occurred during the study. But it failed to take note of an additional three heart attacks that occurred during the study but were not reported until shortly after the previously agreed-on date for closing the data-collection period. These data were not incorporated because they didn't strictly meet the criteria for inclusion. If they had been, the analysis would have shown a slightly higher heart attack rate in patients given Vioxx.
TheNew England Journal apparently learned about the relevance of the three excluded heart attacks to the study long after publication, when they were unearthed as a byproduct of Vioxx-related lawsuits against Merck. *NEJM printed an "expression of concern" about the original study last December. But the WSJ suggests that the journal wasn't critical enough when the paper was first submitted for publication and distanced itself far too late. The WSJ also hints (but is careful not to say outright) that because NEJM permitted these data to be omitted, it earned almost $700,000 from the sale of reprints, mainly to Merck.
There's more. The WSJ also re-examined the study that led Merck to withdraw Vioxx from the market—the one that began by looking for a cancer-preventing effect but produced the more important finding that Vioxx increases the risk of heart attacks and stroke. That study included a graph that compared the cumulative cardiovascular effects as time passed in patients treated with Vioxx and patients treated with a placebo. For the first 18 months of the study, the two curves of the graph pretty much overlap. Then they begin to diverge in a way that suggests that Vioxx is riskier to cardiovascular health than the placebo. * That increased risk is precisely why Merck withdrew Vioxx from the market.
Now the WSJ, again using newly released data elicited by the lawsuits, argues that if patients were studied in a new way—if they were followed for signs of heart disease for a much longer period than they were in the original research protocol—then the two curves would diverge at four months after starting Vioxx, not 18 months. In other words, such a study would show Vioxx users to be at greater risk of heart attack after taking the drug for only four months. If true, that's a great boon to potential plaintiffs and their lawyers because it expands the class of patients with apparently supported claims. *
But there are two problems with reanalyzing the results in this way. First, there is great benefit to analyzing data according to standards defined before there is even a hint of a result. That way, there can be no question that the standards were picked—consciously or unconsciously—to manipulate a study's outcome. Second, trends shown by differences in graphed curves are only obvious in retrospect. Early on, when the two lines begin to diverge, the apparent divergence might be due to real differences, but it might equally well be due to a random wobble in the data. Only in the fullness of time can the true answer emerge. So, this hindsight reanalysis of the data against Merck is improper and unfair.
Merck has defended itself by saying that standard statistical practice obliged it to exclude the three additional cases of heart attack and not to change the way heart disease was recorded after the study began. That's quite correct from the standpoint of good statistics. It precluded skewing the data to fit a desired result. But it's not a wholly satisfying response. The real question is whether medical ethics demanded that Merck not only do the correct statistical analysis, as I think it did, but also point out that additional data, not appropriate for inclusion in the study, nonetheless raised questions that required further examination. And here is where I suspect Merck's responsible past practices gave way to its hope of keeping a blockbuster drug online. Put to the test, the company failed to follow its tradition of ethical leadership and instead became just another pharmaceutical company (not a compliment).
What of the New England Journal, a generally excellent medical journal, which I read regularly and very much rely on? Dr. Jeffrey Drazen, its editor in chief, now argues that Merck deceived patients and doctors about the safety of Vioxx—by concealing data from the editors that, had they been made known, would have more clearly and earlier revealed the drug's risks. So, to which standard should the journal have held the authors? The standard of science, which requires punctilious adherence to the preset boundaries of a research project, or the standard of medicine, which requires that not just the truth, but the whole truth, be revealed?
Here's my answer: I was once a scientist and am now a doctor, and I want it both ways. I want my statistics to be honest and solid. But when a researcher has a hunch that there is more to a question than the formal presentation of the data can reveal, or if new information comes to light outside the formal analysis, I want him or her to share that information (properly labeled, of course). I think that drug companies have a moral obligation to reveal everything—good and bad—about their products. And medical journals have a moral obligation to rigorously examine research to make sure the companies meet that bar.
Corrections, May 30, 2006: The original sentence stated that the New England Journal learned about the three excluded heart attacks long after publication. In fact, the journal learned of the heart attacks in 2001, several months after publication, but says it did not understand their relevance to the study until 2005, when litigation led to the release of a memo showing that at least two of the study's authors had known about them before publication. Return to that corrected sentence. Also, the article originally stated that the study compared Merck to the drug naproxen. In fact, the study compared Merck to a placebo. Return to that corrected sentence.
Clarification, June 15, 2006: The Vioxx study at issue here included a data and safety monitoring board that, throughout the study, tracked the two curves of the study separately from the scientists who undertook it. The board noted small differences between the Vioxx takers and the control group beginning at four months. Only with the benefit of 18 months of data did the extent of the risk emerge and the board recommend shutting down the study. (Return to the sentence above.)