Reply to the American College of Neuropharmacology’s Report on SSRI and Suicidal Behavior in Children
Jonathan Leo, Ph.D.
Introduction
The American College of Neuropsychopharmacology (ACNP) recently released an Executive Summary from its Task Force on SSRIs and Suicidal Behavior. The task force concluded that "taking Selective Serotonin Reuptake Inhibitors (SSRIs) or other new generation antidepressant drugs does not increase the risk of suicidal thinking or suicide attempts" (p. 9). However, the conclusions of the task force are based on a biased review of the data, and their report essentially provides a framework for how the FDA should not proceed.
Problems with the ACNP’s First Task Force
This is not the first time the ACNP has officially addressed the issue of SSRIs causing increased thoughts of suicide in some patients. In 1993 the first such task force addressed the same issue in adults, and released a report concluding that there was no scientific evidence of a link between suicide and SSRIs (ACNP, 1993). At that time the ACNP based its conclusions, in large part, on a study published in the British Medical Journal, written by Dr. Charles Beasley (Beasley, Dornseif, Bosomworth et al., 1991). The Beasley study was a meta-anlysis of 3,067 patients who received either fluoxetine (Prozac), a reference antidepressant, or placebo; and the study found a greater degree of improvement in item 3 of the Hamilton Depression Rating Scale for those patients receiving fluoxetine compared to those receiving placebo. However, there are significant problems with the Beasley study which neither the FDA nor the ACNP acknowledged at the time, but which Dr. David Healy has since pointed out. For instance:
- None of the studies included in the Beasley analysis were designed to test whether fluoxetine could be associated with the emergence of suicidality.
- Some of the fluoxetine studies used in this analysis had in fact been rejected by the FDA.
- Only 3,067 patients of the approximately 26,000 patients entered into clinical trials of fluoxetine were included in the meta-analysis.
- No mention was made that benzodiazepines had been co-prescribed in the clinical trial program in order to minimize the agitation that Lilly had recognized fluoxetine could cause.
- No reference was made to the 5% of patients who dropped out for anxiety and agitation – a drop out rate which was statistically significantly greater than placebo (Healy, 2003a, 2003b).
In retrospect it is surprising that the ACNP did not see the problems with the Beasley study. However, in a deposition regarding the trial of Tobin v. SmithKline Beecham, Dr. John Mann a member of the first task force reported that the task force may not have reviewed all the data but instead reviewed only "summary tables." Dr. Mann is also a co-chair of the second conference, leading to obvious questions about how extensive this second task force has been when it comes to analyzing the data.
Problems with the Second Task Force
Just like the first task force, the current 2003 Executive Summary of the ACNP has also based its findings on a selective reading of the evidence. Perhaps the most serious error they make is contained in the following paragraph on page nine:
The most important of these studies pooled and analyzed the evidence from a very large FDA database of clinical trials. The database contained information from more than 20,000 adults studied in randomized controlled trials. Analyses of the database found no relationship between SSRIs and suicidal attempts or actual suicides in adults (Khan et al, 2003, 2002, 2002, 2000).
The ACNP’s summary of the Khan studies is hard to understand because it is so obviously biased towards one side of the story. Dr. Khan’s main scientific interest is in examining the placebo effect in studies of antidepressants to determine whether patients randomized to a placebo group are being exposed to substantial morbidity and morality. Thus Khan and his colleagues analyzed the data in terms of Patient Exposure Years (PEY); and as the ACNP report states, it is true that when the data is analyzed this way there is no significant difference in the rate of suicidal ideation between the patients receiving an SSRI or those receiving placebo. However for curious, diligent, and concerned readers of the Khan et al. 2000 study, there is an interesting statistic: Twenty-seven patients receiving the investigational antidepressant committed suicide, while only 2 receiving placebo committed suicide (Khan, Warner, and Brown, 2000).
Thus, in 2003 David Healy used the same data set as Arif Khan did, but instead analyzed the data looking at suicides and suicide attempts per number of patients, not PEYs. The reasoning for this is that the issue of suicidal ideation is problematic during the first several weeks of antidepressant therapy – virtually no one disagrees with the idea that if there is a problem it occurs at the start of pharmacotherapy. Thus while using PEYs is the correct way to examine questions about the placebo effect, as Khan did; it is not the way to examine a problem, like suicidal ideation, that occurs at the start of therapy. If the most dangerous time for people driving their car to work is when they exit from their own driveway, and you wanted to do a risk analysis involved with driving to work, then "miles driven" would not be the benchmark; instead you would use "the number of departures" from the driveway. Just as suicides should not be reported per patient years; but, rather, should be reported per number of patients starting treatment.
And, as Dr. Healy has shown, when you analyze the data this way there is a problem, because for virtually every single antidepressant analyzed there are more suicides and suicide attempts on the investigational SSRI than on placebo or active comparator – and this is in adults.
Also of note regarding the ACNP’s discussion of the Khan study is that they fail to mention Khan’s findings regarding efficacy. Khan et al. 2000 reported that symptom reduction was 40.7% with investigational drugs, but that with patients receiving placebo it was 30.9% – not a very large difference. The ACNP’s conclusion regarding the Khan studies is based on a superficial reading of the data, yet, a more in-depth reading of Khan’s work leads to a much different conclusion.
The ACNP’s refusal to even cite David Healy, much less deal with his data, must certainly raise some eyebrows. It is hard to imagine why they did not make at least one reference to Healy, and one can only imagine the Task Forces’ deliberations about "The Healy Dilemma." If the FDA also ignores David Healy it is certainly a decision that scientists will be writing about for years.
Conclusions
This reply to the ACNP is not meant to be an exhaustive discussion of the entire issue by any means. It is instead meant to point out the problems with selectively analyzing the data, as the ACNP task force did ten years ago, and has done again in 2004. In 1993 they missed the problems with the Beasley study; and in 2004 they have presented a one-sided explanation of the Khan study. In retrospect, the problems with the Beasley study seem obvious, yet somehow the ACNP missed the problems. In light of the fact that the Beasley study was funded by Eli Lilly there are some problematic implications for conflict of interest issues. But these conflict of interest issues also plague the ACNP’s task force which is made up of, not only investigators with substantial ties to the pharmaceutical industry, but investigators whose studies have led to the current debate.
Whenever conflict of interest issues are raised, the principals involved typically claim that the money they have received has not swayed their judgment; yet, when pharmaceutical companies selectively release data to paid consultants who are writing a favorable review of their product, while simultaneously refusing to release the same data to scientists who are critical of the product, the problem seems obvious. According to the ACNP’s own summary they have been granted access to data which the general public and other scientists are not privy to.
In summary the ACNP report is no different from numerous publications that these authors have previously written. What is somewhat unique is the tone of their document. They have declared that, based on their analysis of confidential data there is no link between suicide and SSRIs; and while they cannot share all that data with the scientific community (at least at this point), because of their credentials, we should trust their analysis – conflicts of interest aside. Yet, this is not the way that "science" functions, and hopefully it is not the way the FDA will function when it comes to analyzing this issue.
For a meaningful discussion to take place the process must be open. As a start, all the data about the studies of anti-depressants and children should be made available to the public. Second, voices other than those responsible for the widespread use of SSRIs in children, and other than those with strong ties to the pharmaceutical industry, should be heard.
References
ACNP. (1993). Suicidal behavior and psychotropic medication: Accepted as a consensus statement by the ACNP council, March 2, 1992. Neuropscyhopharmacology, 8, 177-183.
Beasley, C., Dornseif, B., Bosomworth, J., Sayler, M., Rampey, A., and Heiligenstein, J. (1991). Fluoxetine and suicide: A meta-analysis of controlled trials of treatment for depression. British Medical Journal, 303, 685-692.
Healy, D. (2003a). Let them eat prozac. Toronto: James Lorimer & Company.
Healy, D. (2003b). Lines of evidence on the risks of suicide with selective serotonin reuptake inhibitors. Psychotherapy and Psychosomatics, 72, 71-79.
Khan, A., Warner, H. A., and Brown, W. A. (2000). Symptom Reduction and suicide risk in patients treated with placebo in antidepressant clinical trials. Archives of General Psychiatry, 57, 311-317.
Contact Information:
Jonathan Leo, Ph.D.
Associate Professor of Anatomy
Western University of Health Sciences
309 East Second Street
Pomona, CA 91767