Meeting Summary
Assessing Suicidality During Antidepressant Treatment
November 9, 2005 – November 10, 2005
Bethesda, Maryland
Sponsored by:
Department of Health and Human Services (DHHS)
National Institutes of Health (NIH)
National Institute of Mental Health (NIMH) and NIH Office of Rare Diseases (ORD)
The purpose of this 1 1/2-day workshop was to consider systematic approaches to assessing and studying suicidality and other related adverse events potentially occurring during treatment with antidepressant medications, in order to enhance both basic research on understanding the pharmacologic mechanisms of these agents, as well as research that addresses practical questions of relevance to practitioners, patients, and families (see agenda). The meeting was sponsored by the National Institutes of Health (NIH) Office of Rare Diseases and the National Institute of Mental Health. Participants included suicidologists, other researchers with expertise in a broad array of disciplines, Food and Drug Administration (FDA) participants, and other stakeholders (see participant list). The following is a summary of the discussion and recommended points for consideration.
Issues in the Assessment/Measurement of SSRI Side Effects in Trials
To set the context for the meeting, the workshop began with a set of presentations that summarized the series of events that led to the FDA’s review of antidepressant medications and subsequent warnings. Based on variability in classification of spontaneously-reported adverse events (AEs) across industry-sponsored trials, the FDA consulted outside experts to develop and apply guidelines for operationalizing potential suicide-related AEs. Results based on re-coded data from 24 pediatric trials in 9 drug development programs (5 SSRIs, 4 atypicals) including the NIMH-sponsored Treatment for Adolescents with Depression Study, indicated that antidepressant treatment was associated with an increased risk for suicide-related AEs but not suicide deaths compared with placebo; there was variability in risk across trials and across agents http://www.fda.gov/cder/drug/antidepressants/.
Federally sponsored clinical trials typically include systematically collected AE data. The ongoing Treatment of Adolescent Suicide Attempters study employs the Safety Monitoring Uniform Report Form (general inquiry, body systems review, drug-specific inquiries) plus a review of 13 drug-specific putative triggers for suicidality (e.g., irritability, emotional lability). Similarly, typical adult trials include the use of scales that include systems review and specific inquiries.
Discussion highlighted the following points for consideration when interpreting data from extant trials and designing future research:
- Individual investigator effects should be considered (e.g., in procedures for ascertaining AEs, in provisions for data integrity / quality control, in adequacy of statistical support).
- FDA review utilized essentially a meta-analytic approach; application of more sensitive analytical approaches to clinical trial databases using individual patient-level data may shed further light on the possible association between antidepressants and suicidality.
- Safety of a treatment must be evaluated in the context of its efficacy. This highlights the importance of relying on well-controlled trials that allow for a proper analysis of risks and benefits.
- Analyses to date do not address whether suicide AEs or suicidality was observed among youths who responded to treatment or not (i.e., events might in part reflect sub-optimal response).
- Patients with suicidality were typically excluded from trials; representative samples (with AE assessments prior to initiating treatment and attention to prior suicidality history in the randomization scheme) are needed.
- Treatment in these pediatric trials involved monotherapy; adjuncts are often used to address side effects (SEs; e.g., benzodiazepines to treat agitation), thus, the observed rates might not reflect those in actual practice.
Strategies and Methods that Might Illuminate the Putative Relationship between SSRIs and Suicidality
Additional presentations illustrated a broad array of research approaches and methods that either have been applied or could potentially be applied to improve our understanding of pressing questions.
- Public/Private Databases (e.g., VA, Medicaid, HMO, Indian Health surveillance data, Practice Research Networks) are potential resources for clinical epidemiologic studies or for surveillance; typically available data include demographics, patient data, pharmacy data, procedures/services, and cost data. Advantages include: larger, more representative samples, longer observational periods than randomized controlled trials (RCTs), searchable data elements; and data that are amenable to techniques for addressing selection bias. Disadvantages include the fact that records are only easily searchable by coded fields, key information may be in clinical note text (e.g. adherence), constructs are not optimally operationalized, and disenrollment can lead to missing data.
- Psychological Autopsy techniques and efforts to examine evidence of antidepressants in the system at the time of death might be explored, but drug exposure cannot be determined using these methods. Potential alternatives include something akin to an elaborated medical examiner report, or novel strategies such as those that have followed a cohort of patients admitted to hospice care for cancer.
- Judgment/decision-making theories/models and cognitive psychology methods could be used, for example, to examine the impact of black-box information on consumer/provider decision making regarding whether to initiate/terminate medication. Similarly, cues (e.g., demographics, symptom profile, illness history) used by clinicians in judging potential for suicidality or in making judgments about whether to initiate/discontinue medication could be examined using vignettes or confederate patients.
- Self-report measures, alone and in combination with other methods: Potentially relevant distinctions among self-report assessments include differentiating more trait- vs. state-like (e.g., aggression, impulsivity) measures and more objective (clinician-rated) vs. subjective (self-reported) measures. Preliminary findings suggest that more subjective measures may more closely relate to suicide attempts than objective ratings; and imaging data suggest “objective” and “subjective” map onto different brain regions, suggesting specific parts of the brain may be correlated with specific features of illness.
- Electronic Diary Assessment and Ecological Momentary Assessment (EMA) methods (e.g., via PDA or cell phone) could be used to assess respondents’ experience in the real-time, in real-world settings and obtain a representative sample of experience, for example, including: medication compliance and effects of changes in doses; AE/SE occurrence; suicidal ideation/behavior at random intervals or in response to key events; occurrence of putative proxies (e.g., impulsivity, agitation, sleep problems); use of coping strategies.
- Pharmacogenetics (PG) and pharmacokinetics (PK): Both PG and PK , alone or in combination, can contribute to understanding the mechanisms underlying AEs and facilitate identification of patients at risk. PG involves the study of the impact of genetic variation on the clearance and action of drugs; PK involves study of the absorption, distribution, metabolism, and excretion of a drug (i.e., what the body does to a drug). Individual/subgroup variability in how drugs are processed leads to variable concentration in drug plasma, which in turn can lead to periodic loss of effectiveness, emergence of drug resistance, unnecessary increases in doses, and toxicity/AEs. Efficient methods for sampling patient populations to assess adherence can also be used to better estimate assumed PK and PG effects.
Additional Points for Consideration and Summary
The presentations and discussion illustrated how limitations in current assessment methods, including definitional problems (e.g., differentiating suicidal vs. self-harm behavior, operationalizing “intent”) and lack of reliable methods, and the transitory, private nature of the phenomenon itself, complicate assessment of suicidality. Discussion also highlighted points for consideration when developing, selecting, or administering instruments to assess suicidality, including:
- Rater training and rater influences (e.g., rater’s own risk-aversiveness, caseload demands)
- Timing of administration (e.g., prior to initiating medication, frequency during treatment)
- Selection of informant (parent vs. youth vs. both)
- Report method (self-report vs. clinician rated vs. independent rater, as well as order of obtaining reports)
- Intended use/purpose of assessment (e.g., to determine eligibility, as AE, as outcome)
- Reporting context (e.g., privacy, presence of parents/others) and anonymity of format (e.g., self-report via computer vs. face-to-face interview)
- Impact of negative affect/arousal on reporting
- Site/organizational -level factors that impact likelihood of assessing/reporting AEs
- Broader individual-level influences (culture, language [spoken and “language of thought”], literacy, religiosity, social support)
- Developmental influences (e.g., on comprehension, recall, social desirability)
- Validity of measures/methods among subgroups with different base rates of risk/events
- Implications of repeated assessment (i.e., when is it neutral, therapeutic, iatrogenic?)
- Transportability of instruments (e.g., issues of ease of administration, clinical utility, burden)
A number of relevant, pressing, potentially researchable questions were noted, including:
- Do suicidal events fall along a continuum/hierarchy (e.g., is a preparatory event “worse than” ideation)?
- Are dimensions of events (e.g., lethality, intent, frequency or duration of ideation) or discernable patterns (waxing/waning, common trajectories) predictive of future suicidal behavior?
- Are certain subgroups of patients (e.g., previous attempters, those with comorbid problems) more at risk for treatment-related AEs?
- How do putative proxies(e.g., impulsivity, aggressiveness, insomnia), especially those that emerge early, relate to suicidal behavior?
- Are there discernable differences in associated suicidality among different treatments (e.g., SSRIs, other drug classes, psychotherapy)?
- What underlies the anecdotally described drug-induced “activation”S; and how does this relate to other constructs (e.g., akathisia)?
- Can approaches (time-line follow-back) and technologies (EMA) counter recall biases?
- Can laboratory-based assessments (e.g., behavioral measures of impulsivity) be usefully applied?
Summary
The workshop presentations and discussion suggest that there might be important qualitative differences in the type of information obtained from different sources (e.g., self-report/subjective vs. clinician-rated/more objective), thus, highlighting the need for broad assessment and data integration. The presentations also demonstrated the potential relevance of various research platforms (e.g., RCTs, large public/private databases) and approaches (applications of basic behavioral/cognitive research methods, pharmacogentic/pharmacokinetic approaches). Given the low base rate of suicidality during antidepressant treatment, and given the challenges inherent in studying it, innovative methods and study designs need to be considered to address if, when, and how antidepressant treatment might be associated with suicidality.
