When Medical Science Isn't
A scandal over hidden data on youth suicide lights a dark corner of our drug approval system

by David Dobbs

from Scientific American Mind, Winter 2004

______________________________________________


The use of antidepressants in children, always a warm subject, this year grew hot enough to burn, with revelations that both the drug industry and the FDA hid evidence that most SSRI (selective serotonin uptake inhibitor) antidepressants double suicide risk in depressed juveniles while helping children no more than placebos do. The story started as a trickle in spring and grew all summer, reaching a torrent at high-profile Congressional and FDA hearings in September , where parents told of good children moving from moderate depression to suicide within days of starting SSRIs. Some of these children died during the year during which the FDA delayed action.  Congressional ubcommittee chair Joe Barton (R-Tex.) expressed the general reaction when he called the drug companies' withholding of data appalling and said the FDA's connivance suggested its initials stood for "foot-dragging and alibis." Even Dr. John Hayes, product team leader for Eli Lilly (whose Prozac was the one SSRI found both effective and safe) acknowledged the crisis atmosphere, noting with marked understatement that , "These hearings are evidence É there is a great deal of mistrust."

Given the stellar rise of SSRI antidepressants such as Prozac, Zoloft, and Paxil, perhaps a fall was due. Almost from the moment of Prozac's introduction in 1988, SSRIs were hailed for being as effective as other types of antidepressants while avoiding their pitfalls; SSRIs cause serious side-effects far less often than tricyclic antidepressants and create fewer and milder reactions with other drugs than the monomase oxidine inhibitors. As published studies  and doctors' experience confirmed this early assessment, SSRI use skyrocketed in both adults and children. SSRI prescriptions for minors in the U.S. grew at a double-digit pace through most of the 1990s and early 2000s, until some 2.4% of all minors in the U.S. Ð about 2 million kids Ñ were prescribed SSRIs by 2003. This pediatric rise occured even though no  SSRI won FDA approval for use in children until Prozac did in 2003. Once a drug wins approval for general use (based on testing in adults), doctors may  use it for children too (or even for other conditions) unless a specific ban forbids; such "off-label" use is common, legal, and, given due care, generally safe. Winning specific approval for use in children requires a drug company to run additional trials on pediatric patients.

It was the data from such trials that came into question this summer.  In both Britain and the U.S., government epidemiologists, running meta-analyses on both the published, positive numbers from pediatric trials and data from less flattering, previously withheld trials, found that the total evidence showed that SSRI use in juveniles gave only a placebo-level benefit (helping around 30%-35% of patients) while doubling the incidence of suicidal thoughts and tendencies. This reversed the positive benefit-risk balance the companies had shown in their selected, positive studies.

Britain responded by banning pediatric use of all SSRIs except Prozac. But when the FDA's epidemiologist, Dr. Andrew Mosholder, recommended similarly strong action, the agency deemed his findings inconclusive, ordered another study, forbade him from publishing, and blocked him from testifying at FDA hearings on the issue in February 2004.

Of course, truth will out, at least sometimes, and the tale of Mosholder's suppressed findings leaked to the press soon after the FDA hearing in February. A month later, the Journal of the Canadian Medical Association published a leaked a 1998 GlaxoSmithKline memo urging its staff to suppress findings showing that its SSRI, Paxil, worked no better than placebos. Headlines and talk of cover-ups started to fly. By June, New York State Attorney General Eliot Spitzer had sued GSK for consumer fraud, and GSK and other companies soon faced both individual and class-action lawsuits from families of children who had attempted or committed suicide. Finally, September brought that unmistakeable certification of scandal, the Congressional hearing, where under hot lights both the drug industry and the FDA had to face ugly music, much bipartisan thrashing, and wrenching testimony from parents of suicide victims. 

The entire episide, the British medical journal The Lancet put it, was "a story of confusion, manipulation, and institutional failure." Fortunately, the publicity seems likely to spur a much-needed revision of warnings and protocols for pediatric antidepressant use. A day after its own September meeting, the FDA's advisory committee voted to recommend a "black box" warning for  SSRIs (the strongest measure short of a ban). This would require every SSRI dispensed to display a prominent, black-bordered warning on its label and come with a pamphlet describing the suicide risk and urging close monitoring. The warning would also appear in all ads. Most observers felt this a good solution, for it will inspire more discretion and better monitoring while allowing use of SSRIs when needed. (Few dispute that the drugs help some patients, sometimes profoundly; the drugs appear to create suicidal tendencies in some patients while stopping or preventing them in many more.) As this story went to press, it was unclear whether the FDA would accept these recommendations. While most observers thought the highly publicized recommendations would leave the agency no choice, others note that the FDA's chief counsel, Daniel Troy, is a former drug-industry lawyer who has often intervened on drug companies' behalf since joining the FDA in 2001.

If instituted, a black-box warning will surely do much to alleviate the child-antidepressant problem, causing both doctors and patients to consider the benefits and risks of SSRIs more soberly and monitor their use more closely. Already, the publicity over the data controversy has caused a drop in pediatric antidepressant prescription of almost 20% form late 2003 to September 2004. While doctors worry this will prevent some needy patients from taking the drugs, it should improve monitoring across the board and slow what many felt was an over-readiness to prescribe these drugs.

Unfortunately, the deeper problem Ñ a drug-approval system that allows industry to highlight flattering results and hide negative ones Ñ  will be harder to fix. Drugmakers have cherry-picked their drug-trial data for FDA consideration for decades, defending the practice in the name of protecting proprietary information; only about 50% of all drug trials over the last half-century were reported or published, according to a 2003 Journal of American Medical Association (JAMA) study of clinical trials . As a result, the FDA routinely approves drugs based on partial and often highly unrepresentative data Ñ forcing doctors to rely on the same skewed information.

Given the vast and growing role that prescription medications play in our medical system, critics say that even more an occasional death lies at stake; at issue is whether the "science" underlying much of our health care deserves the name. As University of California at San Francisco Medical School professor and Journal of the American Medical AssociationJAMA deputy editor Dr. Drummond Rennie put it, "If a company does ten trials on a drug and two show it helps but eight show it works no better than Rice Krispies, I'm not exactly getting a scientific view if they publish only the two positive studies. And this affects me as a patient. I've got a good doctor, and I watch his prescribing hand closely. We like to think we're sophisticated. But how can we practice sophisticated medicine if the drug companies are hiding their results? That's not science. That's marketing."

The solution, say Rennie and other observers, lies in establishing a system that makes all drug trials, not just successful ones, part of the public record. The recent pledges by drug companies to publish their studies in an industry database won't answer, as the companies want to make trial registration voluntary, allowing the same manipulations now at issue. More constructive was the September announcement by eleven major medical journals, including JAMA, the New England Journal of Medicine, and The Lancet, }that, beginning next July, they will require drug makers to register drug trials at their outset if they want the option to eventually publish them Ð a move designed to prevent the companies from hiding studies that don't pan out. However, this system could also be manipulated unless all of the hundreds of existing medical journals observe that policy.

Most doctors and patient advocates feel the only sure fix will be to require registration of all drug trials at their inception in a centralized, publicly accessible database that includesa single unique identifier for each drug, the intended therapeutic use in each trial, and each trials' protocols, outcomes, and results. Most advocate a nonprofit or FDA government register, perhaps building on the existing, voluntary register at clinicaltrials.gov, which already lists several thousand drug trials. If mandatory, such a register would enable the FDA to easily consider all trial results (negative, neutral, and positive) when weighing a drug's approval. It would also allow doctors and patients to review trial data by drug, and if sufficiently detailed it might allow independent researchers to do meta-analyses of data from multiple trials, providing the kind of vital perspective the British and U.S. government reviews of SSRIs did.

For such a trial register to really work, advocates say, Congress must not only make trial registrations mandatory, it must give the FDA or the Department of Health and Human Services strong enforcement powers (such as extremely painful fines) to ensure that the drug companies actually register every trial. (The one mandatory trial register already existing, established in 1997 for drugs and devices aimed at life-threatening conditions, gets only 50% compliance from the industry.) "Any law establishing a new database has to give the government a big stick," says Kay Dickersin, the director of Brown University's Center for Clinical Trials and Evidence-Based Healthcare.

Will it happen? Representatives Edward Markey (R--MA) and Henry Waxman (D-CA) proposed a mandatory trial-registration bill in September, and the bipartisan outrage at the Congressional hearings suggested its chances were good. But the pharmaceutical industry lobby, one of Washington's most powerful, has resisted this idea for years and will probably oppose the measure vigorously, hoping to satisfy Congress that a voluntary register will suffice. This argument may find sympathy with a Congress known to be skeptical of regulation. (The industry says providing all trial data jeopardizes proprietary information and competitiveness; its critics say the industry resists mandatory disclosure so it can hide information that is unflattering but vital for public health.)   Much depends on the outcome. A well-enforced mandatory database seems the only thing capable of repairing the present system's data-quality and confidence problems; anything less seems likely to leave both science and confidence wanting.

To trial-registration advocates like JAMA's Rennie and Brown University's Dickersin, there's a painful irony in all this. "We've been pushing trial registration for over two decades," says Dickersin. "But the drug companies have always fended it off by claiming it infringes on their proprietary interests. It's terrible that we had to get to something that involved children and death to make people see the seriousness of this issue. But perhaps this will finally get the job done."