NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Institute of Medicine (US) Forum on Neuroscience and Nervous System Disorders. CNS Clinical Trials: Suicidality and Data Collection, Workshop Summary. Washington (DC): National Academies Press (US); 2010.

Cover of CNS Clinical Trials

CNS Clinical Trials: Suicidality and Data Collection, Workshop Summary.

Show details

1Perspectives from the FDA, Academia, and Patients


Thomas Laughren, director of the Food and Drug Administration’s (FDA’s) Division of Psychiatry Products, explained the FDA’s current policies and recommendations surrounding the relationship between suicidal ideation and suicidal behaviors and use of antidepressants. Laughren’s presentation addressed the methodology and main findings of his division’s meta-analysis of proprietary clinical trial data dealing with antidepressants and suicidality. His presentation also covered what Laughren characterized as the FDA’s “evolving” policy on a requirement to assess suicidality prospectively in future clinical trials for all new psychiatric drugs. The details of that policy are due out in the form of a guidance document, which has not yet been released.

Acting on troubling case reports and some systematic data, the FDA issued a warning about suicidal risk and antidepressants in children. In 2004 the FDA required the addition of a “black box” warning on antidepressants regarding suicide risk to children and adolescents (see the Introduction for more information). Warnings of this kind do not preclude clinicians from prescribing antidepressants to these groups of patients, but they alert clinicians to monitor these patients more closely for signs of clinical worsening, suicidality, or unusual changes in behavior. The FDA recognized the importance of a strong evidence base from randomized clinical trials (RCTs) on which to make policy. The FDA worked with researchers at Columbia University to develop an algorithm for classifying suicidality events from case narratives in previous clinical trials. The classification tool that emerged, with the assistance of Kelly Posner of Columbia University, was C-CASA, the Columbia Classification Algorithm for Suicide Assessment (Box 1-1). Four of the algorithm’s major categories served, for the purpose of the meta-analysis, as primary outcome measures: completed suicide, suicide attempt, preparatory acts toward imminent suicidal behavior, and suicidal ideation. The FDA sought to collect and further analyze additional data beyond what drug sponsors previously submitted to them from clinical trials. The FDA’s solicitation covered all psychiatric drugs and all psychiatric indications. The FDA made a specific request for the case narratives, which would enhance the FDAs capacity to reclassify and pool the data into a meta-analysis, and used these C-CASA categories to classify trial-level and patient-level data that it required drug sponsors to submit.

Box Icon

BOX 1-1

C-CASA Domains. Suicidal ideation Passive (wish to be dead)

C-CASA domains were used to classify case narratives of more than 300 double blind, placebo-controlled RCTs involving 4,600 children and adolescents and more than 77,000 adults. It became clear, said Laughren, that the dataset was unlikely to be a direct source of evidence for the risk of completed suicide. The reason was the dearth of completed suicides. In the pediatric dataset (<18 years of age), there were no suicides. Among adults, there were a total of eight suicides. “[D]o these meta-analyses tell us anything about completed suicide? . . . The answer is no,” said Laughren emphatically. However, Gail Griffith, patient representative to the FDA Psychopharmacological Drugs Advisory Committee, felt that the media, practitioners, and the public focused heavily on the risks despite the documented benefits of psychiatric drug treatments, leading to increased suspicion and alarm.

There are two main findings of the meta-analysis, according to the data in Laughren’s presentation. First, the risk of suicidality is strongest among children and adolescents taking antidepressants versus placebo, OR = 2.2 (95% CI 1.4–3.6), followed by young adults (ages 18–24), OR = 1.55 (95% CI 0.91–2.7) and the risk appears to go down with age (e.g., ages 25–30, OR = 1.00; ages 31–64, OR = 0.77; ages 65+, OR = 0.39) (Figure 1-1). Second, the risk of suicidality is stronger for non-depressed psychiatric patients as compared to depressed ones. The last finding led the FDA to conclude that anyone treated with an antidepressant, regardless of diagnosis, is at risk for suicidality—not just depressed patients. The reasons behind the epidemiological associations are unknown. As more data came to light, the FDA revised its black box warnings, with the most recent changes occurring in 2007. It extended the age range to cover young adults and the diagnosis to cover any psychiatric disorder (Box 1-2).

FIGURE 1-1. Suicidal behavior and ideation psychiatric indications, odds ratio.


Suicidal behavior and ideation psychiatric indications, odds ratio. NOTE: CI = confidence interval, OR = odds ratio. SOURCES: Adapted from Hammad et al., 2006, and Stone et al., 2009.

Box Icon

BOX 1-2

Suicidality Black Box Warning Required by the Food and Drug Administration for Antidepressant Drugs. Antidepressants increased the risk compared to placebo of suicidal thinking and behavior (suicidality) in children, adolescents, and young adults in short-term (more...)

Laughren addressed concerns about ascertainment bias (non-random sampling of patients) being responsible for erroneous results. Critics have raised two major arguments for ascertainment bias. One is that patients who take antidepressants are more likely to be less symptomatic and more talkative due to effective treatment than patients receiving placebo, and thus more likely to disclose their suicidality. The other is that suicidality is more likely to be detected because patients taking antidepressants, in contrast to patients taking placebo, might have other adverse events from the medication, which in turn might provoke more reporting of suicidal thoughts and behaviors. Laughren pointed out that even if patients were more talkative, that would not explain the age-relatedness of the finding because expressivity would act uniformly across age groups. Similarly, if ascertainment bias was responsible, he asked how the presence of treatment-emergent adverse events could explain the age-relatedness of the finding. Furthermore, how would ascertainment bias explain the other finding that the relationship is stronger in non-depressed psychiatric patients, usually those with anxiety disorders, rather than in depressed patients? How could ascertainment bias explain yet another finding that suicidality is stronger for behavior than for ideation? While recognizing legitimate disagreement over the FDA’s findings, Laughren pointed out that a case-control study found that young adults (<19 years old) receiving antidepressant treatment were more likely to make a suicide attempt or completion than adults ages 19 or older (Olfson et al., 2006).

Finally, Laughren was unambiguous about the low predictive value of suicidal ideation as a surrogate endpoint for suicidality. The relationship between antidepressant use and suicidality was far stronger for suicidal behavior than for ideation, persuading him to conclude, “I think in subsequent meta-analyses we probably won’t be looking at that broad endpoint [ideation].” Regarding the use of excluding suicidal patients from clinical trials, Laughren responded, “We do need to expand into more complicated, sicker patients. I think we will learn a lot more if we do that.”


The development of C-CASA and C-SSRS, an instrument for use prospectively in RCTs intended to systematically ascertain and document the occurrence of events defined by C-CASA as suicidality events, was addressed by Kelly Posner of Columbia University, who worked closely with the FDA to analyze and classify its data. She opened her presentation with the observation that psychiatry as a field possesses neither the clarity on how to define suicidal occurrences nor the accompanying terminology, standardization, and training from which clarity would flow. Lack of clarity has tremendous implications because it limits confidence in epidemiological statistics, which can miss signals or amplify false signals. Furthermore, clinical descriptions and patient management and treatment are also jeopardized.

Her team at Columbia University developed the classification and rating system C-CASA to overcome these problems for the purpose of distinguishing suicidal from non-suicidal events. C-CASA provides a common language to classify suicidality data derived from retrospective examination of clinical trials submitted to the FDA (Figure 1-2). It includes a set of operationalized guidelines to infer suicidal intent. C-CASA was found to have strong interrater reliability (Posner et al., 2007). The FDA recommends the use of a similar instrument in antidepressant clinical trials and many other trials for central nervous system disorders. The application of C-CASA to the FDA datasets revealed that a striking one-third of suicidality classifications from drug sponsors were misclassified, Posner noted. But she also acknowledged that C-CASA has its own limitations, especially from ascertainment bias. This bias might have occurred as a result of receiving an antidepressant versus a placebo because sometimes patients who are on a medication are more likely to report side effects of any kind. That gives the medication recipients more opportunities to report any adverse event, including a suicidal occurrence, said Posner.

FIGURE 1-2. C-CASA classification scheme.


C-CASA classification scheme. NOTE: Light gray boxes are Food and Drug Administration (FDA) “primary analysis” (includes events deemed suicidal); light and dark gray boxes are FDA “sensitivity analysis” (includes any event (more...)

C-CASA is the retrospective counterpart of the more detailed classification instrument C-SSRS. The C-SSRS tool was first developed for a prospective national study of treatment for adolescent suicide attempts. C-SSRS was developed by reliance on evidence stemming from two decades of research, noted Posner. It contains a 1-to-5 rating scale for suicidal ideation of increasing severity (from a “wish list to die” to an “active thought of killing oneself with plan and intent”), in contrast to C-CASA, which only has one ideation item. Further details are shown in Figure 1-3. Having been used in hundreds of studies worldwide, C-SSRS has long-standing feasibility.

FIGURE 1-3. C-SSRS prospective C-CASA.


C-SSRS prospective C-CASA. SOURCE: Posner, 2009.

C-SSRS has been translated into more than 90 languages, typically taking only a few minutes to administer. Psychiatric professionals are trained on this tool via a web-based DVD and more sophisticated means. It is used by national and international agencies, including the World Health Organization, Japanese National Institute, Centers for Disease Control and Prevention, and other Public Health Service agencies, as well as in primary care and schools and on college campuses, Posner said.


As an informed parent and as patient representative to the FDA Psychopharmacological Drugs Advisory Committee, Gail Griffith offered her perspective. Feeling that her family was at the “nexus of the debate,” she described her high school–aged son’s initial response to antidepressants. Just as he began to feel well, he wrote four suicide notes and attempted suicide. Fortunately, his near-fatal attempt did not succeed, and he made a full recovery. The experience initially inspired Griffith to provide vigorous support in 2004 to an FDA-imposed black box warning. Yet, as she read more of the national news, she became troubled about misunderstanding by the media, which blurred the important distinctions among suicidal ideation, behaviors, and completed suicide. The media, in her opinion, unduly alarmed the public. Yet she still thought that the black box warning minimized the “somewhat cavalier approach we had seen in [physician] prescribing.” As the FDA attempted to clarify and make transparent its concern about a possible relationship between antidepressants and suicidality, she believed the anxious public was so suspicious of psychiatry and psychiatric drugs that it was far more inclined to exaggerate the risks and ignore the potential benefits. At that point, she reached the conclusion that the black box warning “constituted a real setback in my mind for treatment and for the field of mental health.”

In retrospect, Griffith said she regretted her support in 2004 in favor of the initial black box warning, despite the possible link between suicidality and antidepressants. Among her reasons was the difficulty of evaluating the risks versus benefits with respect to psychiatric drugs. She noted that mental illness is “treated as a suspect claim,” carrying less legitimacy than does physical illness. The public, in her view, emphasized the risks and ignored the benefits around the time of the black box warning. More fundamentally, she asserted, our society does not know how to begin to engage in public discourse about suicidal thoughts and actions, much less completed suicide.

In contrast, she pointed to the way the epilepsy community tackled the possibility of a relationship between antiepileptic drugs and suicidality. An FDA meta-analysis found that antiepileptic drugs also were associated with increased suicidality (FDA, 2007). The epilepsy advocacy community, which she said garners far more legitimacy than the psychiatric community, argued that the treatments have great benefits. The possibility of undertreatment or treatment avoidance carried greater risks than did suicidal thoughts and actions, in that group’s view. Being squarely against a black box warning, epilepsy advocates succeeded in convincing the FDA not to proceed with one. Griffith said she believes that “until psychiatric illnesses are treated like physical illnesses, then all the efforts to end distortions and misperceptions about psychiatric medication benefits are for naught.”

Copyright © 2010, National Academy of Sciences.
Bookshelf ID: NBK52959


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (818K)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...