Research on antidepressants is 12 times as likely to be published in medical journals if the results show the drugs work, making it hard for doctors to pick the right medication for their patients, according to new findings in today's New England Journal of Medicine.
With 94 percent of the publicly available studies by pharmaceutical companies showing the drugs are effective, the information most accessible to doctors makes them appear to work much better than they do, the authors found.
When researchers included studies that had never been published, only half of the antidepressant trials could be considered successful, the new study shows.
"This is an open secret. Hardly anyone knows anything about this," said Dr. Erick Turner, the study's lead author and director of the mood disorders program at the Portland Veterans Affairs Medical Center in Oregon. "If selective publication is going on, we kind of want to know."
Turner, a psychiatrist who formerly reviewed psychiatric drugs at the Food and Drug Administration, said he took on the project because he saw many negative studies with new drug applications and was told they were commonplace.
On the other hand, he recalled, "It seemed every study couldn't miss if you looked at the published literature."
Turner and his colleagues looked at 74 industry-sponsored studies registered with the FDA involving 12 major antidepressants. They compared the number of published studies to the total conducted.
All 12 drugs had been deemed safe and effective in the two "adequate and well-controlled" studies required for FDA approval.
Even after Turner factored in negative studies, each antidepressant was still considered effective, though less effective than originally stated.
The FDA knows what studies have been conducted. When an antidepressant, for example, is being readied for market, every study conducted is supposed to end up in its files. More often than not, each approved antidepressant failed to perform better than a placebo in at least one trial, if not more.
"On average, you have two [positive] studies out of four or five [conducted]," said Dr. Thomas Laughren, director of the FDA's division of psychiatry drug products. "We're not too alarmed about this. We've come to accept that.
"Depression is a difficult disease to study," he added. "It's not a condition like hypertension or elevated cholesterol that you can measure objectively. ... It makes it more of a challenge to do these studies. We're not surprised that we see that half of these studies fail. We know that."
Trials show negative results for a variety of reasons, and they don't necessarily mean an antidepressant isn't working, Laughren said.
Early studies might use doses that are too small. In other cases, the drug won't work during the short duration of a particular trial - typically four to six weeks - but might work over a longer span.
Transparency in all medical research has been an issue for years. Some critics have complained that journals are biased toward the publication of positive results, although studies to quantify that suspicion have been few.
Others have complained that scientists try to massage their data after the fact, changing their hypotheses to reflect what their research has found - an effort to make the findings more impressive.
"This is a problem with drug trials in all branches of medicine," said Dr. Steven S. Sharfstein, president and CEO of Baltimore's Sheppard Pratt Health System.
The editors of 12 major journals have agreed in recent years to require all studies accepted for publication to be registered before they begin, in order to make sure the findings they publish match what the scientists say they intended to study.
Last month, the FDA began requiring registration of all government-funded clinical trials. As a result, researchers say, it will be more obvious when a trial goes unpublished - and scientists can ask appropriate questions about what might be hidden by missing results.
"You never hear from the drug companies about negative trials," said Sharfstein, a psychiatrist. "It's important to have the complete set and be able to weigh the positive and negative trials ... and get a much more nuanced sense of what works for what kinds of patients."
For example, according to Turner's study, there were nine published studies about the drug Paxil, seven of them positive. But there were five more negative studies about Paxil - none of them published.
"Doctors aren't getting the full picture," Sharfstein said. "They're just getting the good news. They're not getting the bad news. And that's not right."
Ken Johnson, senior vice president of the Pharmaceutical Research and Manufacturers of America, said in a statement that since 2004, drug makers have been proactive in getting "crucial information about medicines" to doctors and patients through an Internet database of study results for marketed medicines. The database currently contains summaries for 497 prescription drugs, he said.
"There should be no doubt that America's pharmaceutical research companies are committed to making available clinical trial information to make sure that patients and their physicians have access to relevant data from ongoing clinical testing," Johnson said.
Meanwhile, the FDA's Laughren said some of the negative studies noted in Turner's research could be what the FDA considers failed studies - too flawed to prove anything.
For example, a new drug could be tested against an older drug such as Prozac and against a placebo. If neither Prozac - a drug known to work - nor the new drug beat a placebo in the trial, the trial is considered a failure because Prozac should have worked. If Prozac beats the placebo but the new drug doesn't, the trial is considered negative.
"We tend to discount those studies and not use them" in determining a drug's effectiveness, Laughren said. "Those studies seem to be given equal weight" in Turner's research.
Turner said he had a difficult time getting copies of many of the studies kept by the FDA. Some had to be obtained through Freedom of Information Act requests. Newer studies are more readily available online, though not always, Turner said.
Several doctors said efforts should be made to get negative research results to doctors and patients.
Dr. Jeffrey M. Drazen, editor-in-chief of the New England Journal of Medicine, said about 35 percent of the results published in his journal are negative. There is an obligation, for example, to follow up a positive preliminary study with the final study results, particularly if the final analysis shows an opposite outcome.
Not knowing about negative results, doctors said, can keep them from best treating their patients.
"It's a misinformation of the public," said Dr. Teodore Postolache, an associate professor of psychiatry at the University of Maryland School of Medicine. "The effects of agents we have now in our hands is not as large as we believe."