It is a project of daunting scope, by anyone’s standards: a comprehensive review of data on nine medications collected during 24 clinical trials involving more than 4,000 children and adolescents. Officials at the U.S. Food and Drug Administration (FDA) believe the task is critical to confirming whether apparent increases in rates of suicidal and self-harming thoughts and behaviors observed in those clinical trials are truly tied to the antidepressant medications that were being studied.
It is a deceptively simple question—either the drugs are associated with increased rates or not. Yet, despite more than 12 years of controversy and study, the question remains unanswered by the agency. For the first time, however, the FDA revealed during a public advisory committee meeting last month on the subject significant amounts of data that the agency has been analyzing and made its case for why FDA officials have yet to answer the critical question definitively (Psychiatric News, March 5).
According to agency officials, at the heart of the controversy are data that are incomplete, largely contradictory, and seemingly riddled with errors. Critics of the FDA, as well as of the pharmaceutical companies that conducted the clinical trials in question, say the data are "straightforward," despite what critics allege are drug-company efforts to conceal negative data and massage and exploit positive data.
Officials at the FDA had thought—and pharmaceutical executives probably had hoped—that the issue of antidepressants and suicide was largely behind them after having studied a potential link in adults who took the medications and having held a controversial public advisory committee meeting in 1991. At that time, the agency determined there was not sufficient evidence to support a solid link. However, anecdotal evidence has continued to surface over the years, often in widely publicized cases of suicide or violent behavior while taking antidepressant medications.
Early last year, however, the issue again erupted, when regulators in both the United Kingdom and the United States were reviewing data that both countries require drug makers to submit on the efficacy and safety of antidepressants in pediatric populations. The initial concerns of the regulators in both countries surrounded data submitted by GlaxoSmithKline (GSK) for its Paxil brand of paroxetine.
Regulators in the United Kingdom, upon analyzing the clinical trials data submitted by GSK, noted what they termed a "signal" of an increased incidence of suicidal thoughts and behaviors in patients taking the active medication in the trials, compared with those patients taking placebo. They asked the company for more data and explanations of what the company thought the data meant. They also shared their concerns with officials at the FDA.
The FDA’s own concurrent analysis of the GSK data raised many more questions than it provided answers, leading the agency to request additional data from GSK and eventually from the makers of eight other similar antidepressant medications.
The agency had reviewed pediatric data on safety and efficacy for those eight other medications over about a three-year period and had noted, according to Thomas Laughren, M.D., FDA’s team leader for the psychiatric drug products group, that adverse events—including those suggestive of possible suicidality—had been coded by the trial sponsors of each company somewhat differently. However, suicidality did not emerge as a matter of concern until the agency looked at the data on paroxetine.
It appeared that GSK had lumped different types of adverse events that occurred during the trials, including those suggestive of possible suicidality, into a category labeled "emotional lability." It wasn’t clear to the agency exactly what types of events were combined into this category, so the agency asked GSK for clarification of the specific adverse events the company had included in the emotional lability category.
The FDA asked drug makers to report "risk ratios" for each clinical trial, reporting any "possibly suicide-related event" or suicide attempt. A risk ratio compares the percentage of patients taking the medication who experience the event with the percentage of those on placebo who experience the same event. The paroxetine studies had widely variable risk ratios, whereas the fluoxetine studies were all consistently close to 1.0, indicating no difference between the two groups.
In one of the three studies, GSK’s study known as "329," six of the 93 patients taking paroxetine had a possibly suicide-related event, compared with only one of the 88 patients taking placebo. That resulted in a risk ratio of 5.9, meaning patients taking paroxetine were nearly six times more likely to have a suicide-related event. In study 329, five patients on the drug attempted suicide, while none did on placebo. Although a risk ratio cannot be calculated due to division by zero, the suicide attempts are elevated relative to placebo.
For study 329, the conclusion is obvious: based on the data from 329 alone, paroxetine is associated with an elevated risk of suicidal thoughts and behaviors. But not so fast. When FDA and United Kingdom regulators looked at the other two studies, the picture became muddy all over again. In both study 377 and 701, the risk of a suicide-related event or attempt was nearly equal for both the active-drug group and placebo—a risk ratio very close to 1.0.
Regulators in the United Kingdom called the data in study 329 a "signal" of increased risk and issued strongly worded warnings not to prescribe paroxetine to anyone under 18—that in light of little or no evidence that paroxetine had any efficacy in child and adolescent depression, the risk outweighed any potential benefit.
Meanwhile, on this side of the big pond, FDA regulators took a different course. They issued a less-severe warning, noting that other safe and effective treatments are available for treatment of childhood depression.
Given the discrepancies in the data on paroxetine, the FDA proceeded in July 2003 to request new summary data from the makers of each of the eight other antidepressants it had already reviewed, breaking down adverse events like GSK had done into the categories of "possibly suicide related" and "suicide attempts." GSK had defined the two categories as "any events including thoughts or behaviors the sponsor considered to represent possible suicidality" and "a subset of possibly suicide-related events that included behaviors the sponsor considered to represent self-harm," respectively.
In summaries of the data from three clinical trials submitted on the use of fluoxetine (Prozac)—the only medication in the U.S. approved for pediatric depression—the suicide data are not only clear, they are consistent (see table below). In all three studies, there was no increased risk of suicide-related events or attempted suicide. Indeed, some statisticians would say that there is a "signal" that fluoxetine may be protective against these events, given that overall more suicide-related events and more attempts occurred on placebo than on fluoxetine. The difference between the two, however, was not statistically significant.
Data that the agency received on sertraline (Zoloft), citalopram (Celexa), nefazodone (Serzone), and mirtazapine (Remeron) were released at the February 2 advisory committee meeting. The data showed both more possibly suicide-related events and more suicide attempts in at least one trial for patients taking each of the active medications versus their respective placebo groups. Similar to the paroxetine data, the numbers differed between clinical trials of the same drug but were suggestive, FDA officials believed, of a "signal."
(Summary data of clinical trial adverse events for bupropion [Wellbutrin] and fluvoxamine [Luvox] were not released at the advisory committee meeting. Psychiatric News has filed a request with the agency for the data under the Freedom of Information Act.)
The FDA noted a more clear and disturbing trend in the data for venlafaxine (Effexor): one of the two studies showed patients taking venlafaxine were 5.3 times more likely to experience a suicide-related event compared with patients taking placebo. The second study showed patients taking venlafaxine to be about eight times more likely to experience a suicide-related event. As a result, Effexor’s maker Wyeth voluntarily changed the drug’s label to include a warning that the drug should not be prescribed to children under the age of 18 with major depression and issued a "Dear Doctor" letter explaining the data.
The FDA had three concerns with the numbers submitted by each antidepressant manufacturer.
"In looking at the summary data that sponsors gave us," FDA’s Laughren said at the advisory meeting, "it appeared that somewhat different approaches were used to capture and present these cases to us." Second, there was concern about how the adverse events being reported were classified into meaningful categories. Finally, he said, the FDA was concerned about the "inconsistency in the signal across individual studies within the programs."
The FDA had requested that each company not only provide the number of events and the terminology of how they were recorded, but also any narrative description of the adverse events, if available. The FDA received narratives for much of the data, but not all of them.
Analysis of the narratives revealed that a wide range of events was being listed under the broad category of "possibly suicide-related events" and even as suicide attempts. For example, according to Laughren, one event listed as a "possibly suicide-related event" described a child who became angry and banged his head into the wall two times. Without additional information, it wasn’t clear whether the child actually intended that action to be life threatening.
In contrast, Laughren said, an event that was classified in one study as an accidental injury involved a child who stabbed himself in the neck with a pencil at school after becoming frustrated with a test. Laughren noted that while that particular case may well have been an accident, it raised the issue that the agency might want to review all events of "accidental injury" to make sure no suicidal gestures were overlooked.
"One case classified as a suicide attempt," Laughren described, "was a girl who slapped herself in the face, and that was it." There were also six cases, he said, that were classified simply as "minor self-mutilation" with no identification of the self-harming behavior.
As the regulators reviewed the data, it became more and more clear that the data on hand were not sufficient to answer conclusively the question they were trying to answer. They not only needed additional data; they needed better data.
In the last three months of 2003, the FDA requested "patient-level data" for all 24 studies involving the nine medications under review. Essentially, the agency asked each pharmaceutical manufacturer to hand over individual patient records, collated into tables listing each patient in a given clinical trial, with numerous data points listed for each patient. This would give the agency a far more detailed look at what was happening than the summary tables normally supplied by drug manufacturers, in which only the number of events are listed and compared with the number of total patients exposed to the medication.
These data, the FDA believed, would allow the agency to reclassify all adverse events for risk of suicidal thoughts and behaviors using the same criteria and report the information in the same manner. With a clean, systematic approach, the agency would be able to answer the question at hand confidently. However, agency officials quickly realized that outside help would be needed for such an enormous undertaking, and so the agency turned to a group of suicidality experts at Columbia University and New York State Psychiatric Institute. The patient-level data will be given to the Columbia group for reclassification. The FDA plans to complete this analysis by mid to late summer.
In the meantime, the FDA has also analyzed the data it has received on each of the nine medications through its Adverse Event Reporting System.
Obviously, certain drugs have been on the market longer than others and thus are likely to have higher numbers of adverse-event reports associated with them compared with drugs on the market for only a brief time. To compare the nine medications in a systematic way, the FDA looked for any suicide-related adverse-event reports filed on each of the drugs during the first three years it was on the market (see table on page 1).
Interestingly, fluoxetine was associated with the most reports (34), nearly seven times the number filed for paroxetine (five). In contrast, escitalopram (Lexapro) and mirtazapine had only one report filed, and no reports were filed for nefazodone. Overall, most of the 78 reports involved females over the age of 12; the highest percentage of reports were classified as suicide attempts, which is consistent with epidemiological research on suicide, according to the FDA. There were seven completed suicides (four male, three female), of which six were by patients taking fluoxetine and one by a patient taking paroxetine.
"Interpreting these results, we would say that suicidality is reported with all drugs," said Andrew Mosholder, M.D., an FDA epidemiologist in the Office of Drug Safety, while testifying before the advisory committee. "The drugs with the largest numbers of reports coincided, roughly speaking, with the greatest amount of pediatric use."
Mosholder cautioned that the adverse-event reporting data are notoriously unreliable and represent underreporting of just about any event. While the data may be interesting, he said, they are of limited value in any quantitative comparison between drugs.
"The largest problem here," Mosholder noted, "is that the outcome of interest that we are tracking, which is suicidality, is also an outcome of the indication for which the drug is prescribed."
Data presented at the advisory committee meeting on SSRIs and suicidality, as well as a transcript of the meeting, are posted online at www.fda.gov/ohrms/dockets/ac/cder04.html#PsychopharmacologicDrugs. ▪