The FDA alert on suicidality and antiepileptic drugs: Fire or false alarm?


Address correspondence to Dale C. Hesdorffer, Ph.D., GH Sergievsky Center, Columbia University, P & S Unit 16, 630 West 168th Street, New York, NY 10032, U.S.A. E-mail:


In January 2008, the U.S. Food and Drug Administration (FDA) issued an alert about an increased risk for suicidality in 199 clinical trials of 11 antiepileptic drugs (AEDs) for three different indications, including epilepsy. An advisory panel voted against a black-box warning on AED labels, and the FDA has accepted this recommendation. We discuss three potential problems with the alert. First, adverse event data were used rather than systematically collected data. Second, the 11 drugs grouped together as a single class of AEDs have different mechanisms of action and very different relative risks, many of which were not statistically significant and some of which were smaller than one. These facts suggest that they should not be grouped as a class. Third, the risk of adverse effects from uncontrolled seizures almost certainly outweighs the small risk of suicidality. We place our comments in the context of a review of the literature on suicidality and depression in epilepsy and the sparse literature on AEDs and suicidality. We recommend that all patients with epilepsy be routinely evaluated for depression, anxiety, and suicidality, and that future clinical trials include validated instruments to systematically assess these conditions to determine whether the possible signal observed by the FDA is real.

On January 31, 2008, the U.S. Food and Drug Administration (FDA) issued an alert that a meta-analysis had found a statistically significant 1.80-fold increased risk for suicidality associated with all antiepileptic drugs (AEDs). The meta-analysis had grouped data from trials of 11 AEDs: carbamazepine, felbamate, gabapentin, lamotrigine, levetiracetam, oxcarbazepine, pregabalin, tiagabine, topiramate, valproate, and zonisamide. The analysis included data from 199 randomized clinical trials, including 43,892 patients treated for epilepsy, psychiatric disorders, and other disorders, predominantly pain. In July, an FDA advisory board recommended against a black-box warning. However, the board’s report said, “all current antiepileptic drugs pose an increased risk of suicidality (defined as suicidal ideation and behavior), and prescriptions should be accompanied by a patient medication guide describing this risk” (Busco, 2008). The FDA has accepted the advisory board recommendation against the black box warning.

For many in the epilepsy community, the FDA alert was particularly surprising in that it stated that all 11 AEDs were associated with the risk. A 2006 Epilepsy Research article had previously mentioned plans to undertake such a grouped analysis (Katz, 2006). The reasons included one sponsor’s report of a slightly higher risk of suicidality in the AED-treated group than in the placebo group, and postmarketing reports of suicidality related to the brand formulation of gabapentin (Neurontin) (Rosack, 2005). These reports led the FDA to request data from sponsors in 2005 (U.S. Department of Health and Human Services et al., 2008). Specifically, the FDA asked sponsors to search their databases for text strings of adverse events that might be related to suicidality and submit these to the FDA for categorization by Columbia University experts and later statistical analysis.

The Columbia University methodology (Posner et al., 2007) used to retrospectively identify episodes of suicidality involved the expert evaluation of text strings from adverse event reports supplied by the sponsors. These text strings were classified according to a standardized suicidality rating system, the Classification Algorithm for Suicide Assessment (C-CASA), by evaluators blind to the treatment group. The C-CASA has eight levels of classification under three rubrics: nonsuicidal events (self-injurious behavior without suicidal intent and no deliberate self-harm); indeterminate or potentially suicidal events (self-injurious behavior with unknown suicidal intent); and suicidal events (ideation, preparatory acts toward imminent suicidal behavior, suicide attempt, and completed suicide). Only suicidal events were included in the FDA analysis. The C-CASA has previously been used to classify suicidality in association with the selective serotonin reuptake inhibitors (SSRIs) (Mann et al., 2006).

Suicidality in Epilepsy: A Complex Problem

Completed suicide is the rarest manifestation of suicidality, followed by suicide attempt, and suicidal ideation (Kessler et al., 1999, 2005). The rate of completed suicides is 12.0/100,000 (Li et al., 2001), with a marked predominance in males (McIntosh & Jewell, 1986). In contrast, the lifetime prevalence of suicide attempts is 0.6–4.9% overall (Kessler et al., 1999; Weissman et al., 1999; Oquendo et al., 2004; Kessler et al., 2005), with a female preponderance.

Epilepsy and suicidality

Suicidality should be of concern in the care of patients with epilepsy. Epilepsy is comorbid with suicidality (Henriksen et al., 1970; Barraclough, 1987; Nilsson et al., 1997; Rafnsson et al., 2001; Jones et al., 2003; Hesdorffer et al., 2006; Christensen et al., 2007) and with major depression (Dominian et al., 1963; Jacoby et al., 1996; O’Donoghue et al., 1999; Hesdorffer et al., 2000; Beghi et al., 2002; Boylan et al., 2004; Cramer et al., 2004; Hesdorffer et al., 2006). A history of suicidality (Hesdorffer et al., 2006) and a history of major depression (Forsgren & Nystrom, 1990; Hesdorffer et al.,2000, 2006) are also associated with an increased risk for developing epilepsy. After a diagnosis of epilepsy, the risk for completed suicide increases, with the overall standardized mortality ratio (SMR) ranging from 3.5 to 5.0 (Nilsson et al., 1997; Rafnsson et al., 2001). It is higher in the presence of known psychiatric diagnosis [SMR = 20; 95% confidence interval (CI) = 12.7–31.2] than in the absence of such diagnosis (SMR = 2; 95% CI = 1.7–2.3) (Christensen et al., 2007). The increased risk for completed suicide after epilepsy onset may represent recurrence of premorbid suicidality, but this has never been studied. Likewise, although many cross-sectional studies have shown that depression is more common in people with prevalent epilepsy than in controls, it is not known whether this represents recurrence of premorbid depression, emergence of a first episode of depression, or both. Further work is needed to determine if associations between epilepsy, suicidality and depression are truly bidirectional.

In addition, suicidal ideation and behavior have been identified as postictal psychiatric phenomena in patients with pharmacoresistant epilepsy. Therefore, in a study of 100 consecutive patients with refractory partial epilepsy, 13% of patients reported suicidal ideation within 72 h after more than 50% of their seizures (median duration: 24 h; range: 1–240 h) (Kanner et al., 2004).

Forced normalization, although rare, is another epilepsy-related process that could result in increased suicidality. It consists of the development of depressive or psychotic episodes in patients who become seizure-free after having suffered from a chronic pharmacoresistant epilepsy (Wolf, 1991).

AED and suicidality: Not much is known

Very few studies have been published specifically examining the risk of suicidal ideation and behavior associated with AEDs. In these studies, the incidence has been very low. In an observational study of 517 consecutive patients treated with levetiracetam, only four (0.7%) reported suicidal ideation, all of whom had experienced a depressive disorder in the past (Mula & Sander, 2007). In another observational study of 224 consecutive patients treated with levetiracetam, the drug was discontinued in four (1.8%) due to episodes of major depression associated with suicidal ideation. All four patients had a prior history of a depressive disorder (Kanner AM, unpublished data). Still, the actual incidence of suicidal ideation and behavior resulting from AED exposure itself remains unknown, as most published data group all psychiatric adverse events together (Brent et al., 1987; Chadwick, 1999; Trimble et al., 2000; French et al., 2001; Mula et al., 2003; Cramer et al., 2004), rather than reporting suicidality by itself, and those reporting suicidality have not adjusted for prior depression and prior suicidality.

More information is available about the occurrence of psychiatric disorders in patients taking AEDs (Brent et al., 1987; Chadwick, 1999; Trimble et al., 2000; French et al., 2001; Mula et al., 2003; Cramer et al., 2004), but these studies have not been able to rule out the contribution of other factors such as family history of psychiatric disorders or past psychiatric disorder. We mention these studies because psychiatric disorders, which can potentially facilitate the development of suicidal ideation and behavior, have been identified with phenobarbital and primidone among the older AEDs, and with vigabatrin, levetiracetam, topiramate, and zonisamide among the second-generation AEDs (Brent et al., 1987; Chadwick, 1999; French et al., 2001; Cramer et al., 2004). Importantly, a previous psychiatric history or a family psychiatric history appears to be a common denominator in several of these reports. For example, depressive episodes were identified in 40% of 15 children with epilepsy treated with phenobarbital and 4% of 24 children treated with carbamazepine (p = 0.02). Yet this differential prevalence of depression was accounted for by a family history of a major affective disorder among first-degree relatives (Brent et al., 1987). In a case series of 89 patients treated with topiramate, tiagabine, and vigabatrin as add-on regimens, Trimble et al. (2000) investigated psychiatric adverse events and found that nearly two-thirds had a psychiatric history with a strong association between the type of past psychiatric illness and the emerging illness. In a study of 517 consecutive patients treated with levetiracetam, cited above, a family history of psychiatric disorder was a significant predictor for the development of psychiatric symptoms (Mula & Sander, 2007). Among 491 consecutive patients treated with topiramate, 23% were found to exhibit psychiatric symptoms. The risk was greater in patients with a family psychiatric history [odds ratio (OR) 6.1; 95% CI: 2.0–18.9] and in those with a previous psychiatric history [OR: 4.5; 95% CI: 2.8–7.3] (Mula et al., 2003). Therefore, a prior personal or family history of psychiatric disorders is associated with an increased risk for manifesting psychiatric disorders, predominantly depression, while taking AEDs. This may reflect the natural course of an underlying recurrent psychiatric illness with no effect from AEDs, or it could suggest that AEDs lower the threshold for manifesting psychiatric symptoms in individuals who are vulnerable because of a history of psychiatric disorders.

FDA Alert: Is It Warranted?

The C-CASA suicidality classification was used in 199 trials for the treatment of epilepsy (31%), pain (41%), and psychiatric disorders (28%). The FDA included only 33% of these trials in the main analysis, using the odds ratio to examine the association between AEDs and suicidality, because trials without reports of suicidality were excluded from the analysis. The agency conducted the statistical analysis and reported that suicidality occurred in 4.3 per 1,000 patients treated with AEDs in the active arm compared to 2.2 per 1,000 patients in the comparison arm (U.S. Department of Health and Human Services et al., 2008). This represents a risk difference of 2.1 per 1,000 with a statistically significant 95% CI of 0.7–4.2 (significant because it does not include 0, which would indicate no difference). The term comparison arm is used here rather than placebo, because many of the trials, particularly in epilepsy, compared adjunctive therapy, not monotherapy, against placebo.

In analyses within each AED, there was a nonsignificant protective effect on suicidality (i.e., ORs <1) for carbamazepine [OR = 0.65; 95% CI = 0.08–4.42] and divalproex [OR = 0.72; 95% CI = 0.29–1.84]. All other AEDs had odds ratios greater than 1, indicating an increased risk of suicidality, although the increase reached statistical significance only with lamotrigine [OR = 2.08; 95% CI = 1.03–4.40] and topiramate [OR = 2.53; 95%CI = 1.21–5.85], both of which already list suicidality on their package inserts.

As expected from the general population data on suicidality (Kessler et al., 1999, 2005; Weissman et al., 1999), ideation was the most frequent manifestation of suicidality in adverse event reports in the AED trials and accounted for 67.6% of all suicidality. Preparatory acts accounted for 2.8%, attempt for 26.8%, and completed suicide for 2.8%. In analyses separately considering suicidal ideation and suicidal behavior, the magnitude of the increased risk for ideation [OR = 1.45] was less than that for behavior [OR = 2.92], although the difference was not statistically significant. No analyses took into account suicidality before entrance into the clinical trial.

Subanalyses examined whether the effect of AEDs on suicidality differed according to indication, gender, age, race, inpatient status, North American location, or presumed mechanism of action of the AEDS. None of these subanalyses showed statistically significant differences by subgroup. However, the association between AEDs and suicidality was stronger in the non–North American trials [OR = 4.53; 95% CI = 1.86–13.18] than in the North American trials [OR = 1.38; 95% CI = 0.90–2.13].

Additional subanalyses were undertaken after GlaxoSmithKline submitted three further lamotrigine trials fulfilling the FDA’s inclusion criteria (Dr. Katz memo and briefing document to PCNS and PD Advisory Committees, 2008). These trials, which were ongoing when the FDA originally requested data from the drug companies, added nine suicidality events; one in the treatment arm and eight in the placebo arm. The magnitude of the association between lamotrigine and suicidality decreased and was no longer statistically significant [OR = 1.32; 95% CI = 0.75–2.38]. Importantly, the overall association between the 11 AEDs and suicidality decreased from 1.80 [95% CI = 1.24–2.66] to 1.55 [95% CI = 1.09–2.21] further highlighting the vulnerability of the suicidality signal detected by the FDA.

Methodologic issues

Several methodologic difficulties with the FDA analyses must be taken into account. These difficulties are related to the source of outcome data on suicidality, the statistical methodologies used, heterogeneity of the trials with respect to whether or not adjunctive therapy was used, and the different mechanisms of the 11 AEDs evaluated.

The use of adverse event data, rather than systematically collected data on suicidality, is problematic for several reasons. Adverse event data arise from spontaneous reporting and not from systematic collection of the event of interest using a validated scale. Their completeness and validity suffer consequently, and the FDA would never accept such data as the primary endpoint in a clinical trial. A major limitation of such spontaneous reports is the tendency for patients on the active compound to report more adverse events in general than patients taking placebo (GlaxoSmithKline, 2006;Dr. Katz memo and briefing document to peripheral and central nervous systems (PCNS) and psychiatric drugs (PD) advisory board, 2008:51). This phenomenon, referred to as reporting bias, is a flaw in measuring outcome that results in differential accuracy of information between compared groups.

The inherent problems with adverse event data could be overcome by using systematically collected data on suicidality. These data are typically collected as part of trials conducted in patients with psychiatric disorders, but the FDA has not analyzed these data in the AED trials conducted for psychiatric indications. In at least one important recent case, systematically collected adverse event data has led to diametrically opposite results from spontaneous adverse event data. In the FDA analysis of suicidality in association with SSRIs, the adverse event data indicated a 1.95-fold statistically significant increased risk for suicidality (Hammad et al., 2006). In contrast, when systematically collected data from rating scales employed during the SSRI trials to assess changes in depressive symptoms were used to assess suicidality, there was no association between suicidality and SSRIs [relative risk (RR) = 0.92; 95% CI = 0.76–1.11] (Mann et al., 2006). Particularly telling were the consequences of the black-box warning. Prescription rates for youth fell after the warning (Gibbons et al., 2007; Nemeroff et al., 2007), but subsequently the incidence of completed suicide actually increased in that age group (Gibbons et al., 2007). If SSRIs were truly associated with an increased risk for suicidality as the FDA suggested, then a reduced prescription rate should have been associated with a decrease in completed suicide. When the opposite is seen, one must question the value of such analyses of passively reported adverse events. Systematic data collection eliminates the reporting bias common to adverse events and would provide a more rigorous test of whether or not AEDs increase risk for suicidality. Clearly future AED trials must include validated instruments to systematically assess suicidality (Baldessarini et al., 2006).

The FDA’s main statistical analysis involved separate examination of the increased risk for suicidality for each randomized clinical trial. These were then combined to yield one summary estimate of the increased risk for suicidality across all AEDs and indications, using the exact method (Cytel, 2005). Unfortunately, this method eliminates trials with no suicidality events in either arm, considering them uninformative. According to the FDA report, this represented 133 of the 199 trials (66.8%), leaving only 33 trials for the main analysis (U.S. Department of Health and Human Services et al., 2008). In an ancillary analysis, the FDA included the zero event trials, using methodology described by Greenland and Robins (1985). However, if all trials for a given drug have no events, then the confidence interval cannot be calculated using this method.

Although eight of 11 AEDs were associated with an increased risk for suicidality, only two of these associations reached statistical significance. No suicidality events were identified for felbamate in the active or placebo arm. Two AEDs had small protective effects for suicidality: carbamazepine and valproate. Patients on these drugs accounted for 7.8% of all patients on active drug and 5.7% of all patients on placebo. Despite these differences in the individual drugs, the FDA analysis justified grouping all the drugs together because formal statistical tests for heterogeneity across trials produced nonsignificant results. However, the very small number of patients on carbamazepine and valproate makes it unlikely that any heterogeneity would be found. The FDA report also acknowledges that statistical power was low for testing consistency across trials due to the small number of suicidality events.

Excluding the low-dose trials that were not part of the FDA’s detailed analysis of 199 trials, adjunctive therapy trials accounted for 92% of epilepsy trials, 14% of psychiatric trials, and 15% of other trials. This is in sharp contrast to the SSRI analysis where all but one trial compared SSRIs to a no-treatment placebo (Mann et al., 2006). As previously mentioned, AEDs were associated with a greater risk for suicidality in epilepsy [OR = 3.53; 95% CI = 1.28–12.10] than in psychiatric disorders [OR = 1.51; 95% CI = 0.95–2.45] and other disorders [OR = 1.87; 95% CI = 0.81–4.76]. It is unclear whether the greater magnitude of association in the epilepsy trials was due to drug interactions, given the high proportion of epilepsy trials designed with adjunctive therapy, or whether it was potentially due to the low suicidality risk associated with carbamazepine and valproate, both protective for suicidality, and the most common comparison drugs in these trials (French et al., 2005).

The FDA chose to consider the 11 AEDs as a class despite their different mechanisms of action, because they all share “the ability to decrease the frequency of seizures” (Dr. Katz memo and briefing document to PCNS and PD advisory board, 2008:4). Dr. Katz explains that the FDA typically considers all drugs approved for a specific indication to constitute a class and that problems would arise if only some of the AEDs were considered to be associated with suicidality and others not because this might change prescribing practices. This argument is not scientifically based and ignores the context in which AED drug development has occurred. Consider the example of antihypertensives: drugs with extremely different mechanisms of action (e.g., diuretics, calcium channel blockers, and beta blockers) that nonetheless share the common ability to decrease blood pressure. If a hypothetical analysis of adverse events revealed diuretics were associated with suicidality, then the FDA’s logic would indicate that all drugs that lower blood pressure should be considered as a class and labeling applied to all even if, for example, beta blockers were protective.

Nothing can be done to correct the problematic nature of adverse event outcome data, or the associated reporting bias, other than to compare results of analyses of these spontaneous reports with analyses using systematically collected data in the same patients. Unfortunately, existing systematically collected data on suicidality have not yet been analyzed by the FDA. It, therefore, falls to the epilepsy, psychiatry, and pain communities to collect these data in a systematic fashion in future AED clinical trials.

Implications of the FDA Alert in Clinical Practice

Alternative treatments

Although AEDs are not the only option for treatment of psychiatric disorders and pain, currently there are no alternative treatments for epilepsy. Therefore, assessment of the risk–benefit balance for epilepsy may be more important than for the other disorders for which AEDs are prescribed. The FDA advisory panel recommended that information about the potential risk for suicidality be sent to doctors and detailed within the package insert. In addition, the FDA has recommended that patients receive a medication guide detailing the potential for suicidality each time a prescription is filled. It is too early to tell what effect this may have on AED prescription rates or on adherence to AEDs in epilepsy. However, after the black-box warning was issued for SSRIs, prescription rates fell among children and adolescents younger than 19 years of age (Gibbons et al., 2007; Nemeroff et al., 2007). This suggests the possibility of two consequences with AEDs that have already been observed for SSRIs: doctors may issue fewer prescriptions (Bhatia et al., 2008), and patients or their caretakers may decide that the drugs are not worth the potential risk for suicidality (Bhatia et al., 2008).

Although the issue of suicidality with AEDs is controversial, the adverse effects of failing to control epilepsy are not. If AEDs are less frequently prescribed or taken, seizure frequency will increase with associated increases in accidents (Buck et al., 1997; Van den Broek et al., 2004) and mortality, including sudden unexplained death in epilepsy (SUDEP) (Sander & Bell, 2004). Seizure frequency is an important predictor of both accidents and SUDEP. The risk for these serious adverse outcomes of epilepsy is more severe than the 3.4/1,000 suicidality rate reported by the FDA in the AED-treated group, especially since 64.4% of the suicidality incidents were ideation rather than attempts. Among people with active epilepsy who had had at least one seizure during the previous year (Buck et al., 1997), 24% experienced at least one head injury, 16% were burned or scalded, 10% had a dental injury, and 6% had a fracture. Seizure severity and seizure frequency were strong predictors of experiencing such injuries. Another large study of people with epilepsy reported that 21% had an accident during the study, of which 24% were seizure-related (Van den Broek et al., 2004). SUDEP occurs in 0.35–2.7 per 1,000 people with epilepsy in population-based studies (Annegers & Coan, 1999), and is more common in people with uncontrolled seizures (Annegers & Coan, 1999; Nilsson et al., 1999; Sander & Bell, 2004), and among those with a generalized tonic–clonic seizure in the preceding 3 months (Langan et al., 2005). But even if the FDA is correct, the risk for suicidality, predominantly suicidal ideation, is only 3.4/1,000 in people with epilepsy treated with the 11 AEDs. Therefore, AED treatment protects people with epilepsy from severe adverse outcomes associated with the occurrence of seizures.

What should clinicians look for?

Psychiatric disorders in the general population and in epilepsy are more often than not unrecognized and untreated (Simon et al., 1999; Kanner et al., 2004; Olfson et al., 2005), even if they are severe. In a study of 100 consecutive patients with a depressive episode sufficiently severe to warrant pharmacologic treatment, 63% of patients with spontaneous depressive episodes and 53% of “iatrogenic” episodes had been symptomatic for more than one year before they were identified (Kanner et al., 2004). Yet good medical practice demands an early identification of comorbid depressive and anxiety disorders in patients with epilepsy, not only because of the significant negative impact they have on patients’ quality of life (Gilliam, 2002) but also because of their association with an increased risk of suicidal ideation and behavior (Jones et al., 2003; Tellez-Zenteno et al., 2007).

At a minimum, clinicians should obtain the following clinical data: (1) Current or past history of the principal depressive and anxiety disorders (e.g., major depressive disorder, generalized anxiety disorder, panic disorder); (2) current or past history of suicidal ideation or behavior, both during the interictal and postictal periods; and (3) family psychiatric history of mood disorders and suicidal behavior. When clinicians treating epilepsy identify current psychiatric symptoms or suicidality, they should refer the patient to a mental health professional for care.

The acquisition of such data need not be a taxing demand on clinicians who can initially inquire if patients are experiencing anhedonia, that is, the inability to find pleasure in most activities. This symptom is a very good predictor of current depression (Mohr et al., 2007). Anhedonia is not usually related to physical complaints secondary to drugs or underlying illness, and it is a barometer of the intensity of the depression in the medically ill. Similar single questions are not possible for anxiety and suicidality.

Clinicians could also use screening instruments to identify current symptoms of depression. These include a six-item self-rating screening instrument, the Neurological Disorders Depression Inventory for Epilepsy (NDDI-E), that takes less than 3 min to complete and is validated to identify major depressive episodes in patients with epilepsy (Gilliam et al., 2006). This instrument was specifically constructed to minimize the potential that symptoms reported are not adverse events related to AEDs (e.g., fatigue or experiencing cognitive problems), whereas other instruments do not make this distinction. A score of 15 and higher is suggestive of a major depressive disorder and serves as a red flag referral to a psychiatrist for further evaluation. The Beck Depression Inventory (BDI-II) (Beck et al., 1996) is another self-rating screening instrument developed to identify symptoms of depression in the general population that has been validated in patients with epilepsy (Jones et al., 2005). It has 21 items and it takes 5–10 min to complete. A score of 14–19 suggests a depressive episode of mild severity; 20–28 moderate, and 29–63 severe. Both the NDDI-E and the BDI-II have items that identify suicidal ideation.

Several validated self-rating instruments to identify symptoms of anxiety are available, although none has been validated in patients with epilepsy. The recently developed the Patient’s Health Questionnaire-Generalized Anxiety Disorder-7 (PHQ-GAD7) (Kroenke et al., 2007), is a seven-item self-rating instrument that takes 3 min to administer and was developed to identify symptoms of generalized anxiety disorders. None of the items can be confounded with adverse events of AEDs or cognitive symptoms associated with the underlying neurologic insult that caused the epilepsy.

The identification of any current or past symptoms of depression and anxiety or suicidality as well as family history of these conditions should never be considered a reason to delay the start of AED therapy, as the likelihood of seizure recurrence poses a greater safety risk for patients, as previously discussed. The identification of such a prior or current history may help the clinician in the choice of AED, such as avoiding drugs that are known to increase the risk of psychiatric adverse events (barbiturates, levetiracetam, topiramate, zonisamide), particularly in the presence of these factors.

The development of psychiatric symptoms after the start of an AED should not be automatically considered causally related to the AED. Indeed, several mechanisms may be involved in the development of AED-related psychiatric symptoms and each should always be carefully considered. These include: (1) the administration of an AED with negative psychotropic properties. In such cases, the AED should be discontinued; (2) during a switch to a new AED, the discontinuation of the previous AED with mood-stabilizing properties (e.g., carbamazepine, valproic acid, lamotrigine) or anxiolytic ones (e.g., benzodiazepines, pregabalin, gabapentin), in patients with a prior mood or anxiety disorder that had been in remission with the discontinued AED. In such cases, an AED with similar psychotropic properties should be selected, or the previous AED restarted; and (3) the introduction of AEDs with enzyme-inducing properties in patients with a psychiatric disorder treated with a psychotropic drug metabolized in the liver. In such patients, the psychiatric symptoms result from an increased metabolism of the concurrent psychotropic drug, leading to lower serum concentrations associated with loss of efficacy. In such cases, the dose of the psychotropic drug must be adjusted to compensate for its increased clearance.

It is not rare for some patients with refractory epilepsy to become seizure-free or experience a significant reduction of seizure frequency with a new AED that can also cause psychiatric adverse events. If no other AEDs are available, the clinician can discuss with the patient the option to treat the psychiatric symptoms with a psychotropic drug without having to discontinue the AED.

Obviously, the presence of suicidality requires an immediate in-depth evaluation and treatment of the underlying psychiatric disorder ideally by a psychiatrist, and if not available (e.g., in rural areas) by another mental health professional. Likewise, current depressive and anxiety symptoms call for a more detailed evaluation by a mental health professional and a change to therapy more likely to yield total remission of symptoms (e.g., pharmacotherapy or cognitive behavior therapy, or both). Patients with current psychiatric symptoms not associated with suicidal ideation should be followed closely after the start of an AED and advised to contact their physician if their symptoms worsen or if they start experiencing suicidal ideation. Finally, patients with a family psychiatric history or a past psychiatric history should be advised of the increased risk of psychiatric symptoms in association with some AEDs. They should be told to contact their physician as soon as they occur.

How should patients and family be counseled about the FDA alert?

This alert can be expected to cause great concern among patients and family members. Accordingly, clinicians must provide a careful explanation of the “meaning” of the alert, placing the AED-“related” suicidality risk in the context of the complexity of suicidality in epilepsy as discussed previously. Clinicians must emphasize the very low frequency with which these symptoms were attributed to the drugs in the FDA meta-analysis and indicate that many questions remain to be answered with respect to the validity of the FDA’s conclusions. Finally, patients and family members must understand the grave risk incurred by stopping AED without the advice and close supervision of their physician. This is particularly important, since physicians have reported that 22% of caregivers refused SSRI treatment for depressed adolescents after the black-box warning (Bhatia et al., 2008).

What is the expected impact of the FDA alert on AED regulatory trials?

It is very likely that the FDA in the U.S.A., and at some future time the European Agency for the Evaluation of Medicinal Products (EMEA), will require a prospective investigation of suicidality in every regulatory AED trial. It is also possible that patients will need to be screened for suicidality, depression, and anxiety before randomization and during the course of future trials. In our opinion, an appropriate evaluation should include: the identification of suicidality, using the suicide module of the Mini International Neuropsychiatric Interview (MINI) (Sheehan et al., 1998)), as well as a brief structured interview to identify current and past history of depression and anxiety disorders, such as the depression and anxiety modules of the Structured Clinical Interview for Axis I DSM-IV Diagnosis (SCID) (First et al., 1999); and self-rating screening instruments like the NDDI-E (Gilliam et al., 2006), the BDI-II (Beck et al., 1996), and the PHQ-GAD7 (Kroenke et al., 2007). In addition, the Columbia Suicide Severity Rating Scale (C-SSRS), a new instrument, has been developed specifically to identify current and past suicidal ideation and behavior and is being used in trials for antidepressants (Posner K., personal communication).

What should be done if patients screen positive for these conditions?

The exclusion of patients with any past history of suicidal ideation or behavior from drug trials is of great concern and, in our opinion, probably unnecessary. It is reasonable to exclude patients with a history of suicidal behavior within the last 2 years but not patients with any previous history. Furthermore, a history of suicidal ideation without any evidence of intention to act on the thoughts should not automatically call for the exclusion of the patient from a trial, even if the ideation has occurred in the last 2 years. This applies particularly to postictal suicidal ideation, as patients typically learn to live with such symptoms, knowing that they are self-limited in duration.

Concluding Remarks

The relationship between suicidality and epilepsy is a complex, multifactorial problem, and AEDs probably have little impact. The meta-analysis completed by the FDA has raised more questions than it has provided answers. These data need to undergo further careful statistical analyses in order for clinicians and researchers to better understand their meaning. Unfortunately, even the best statistical analysis cannot compensate for the fact that the data on suicidality in these trials was not collected prospectively in a systematic fashion. Serial systematic data collection of suicidality is the only way to eliminate the biases already discussed. What this means is that every patient must be screened for current or past history of suicidality, depression, and anxiety before randomization, using validated instruments. During the course of the trial, these conditions must be reassessed on a schedule using the same validated instruments. Information collected in this manner will help us to determine whether the FDA alert indicates a fire or is merely a false alarm.


The authors thank Jessica Ancker, MS, for her thoughtful editorial assistance. In addition, we confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Disclosures: Dr. Hesdorffer has no financial interest in relation to this manuscript or its results. Dr. Kanner reports the following disclosures. Research grants: GlaxoSmithKline, Novartis. Advisory Boards: UCB, GlaxoSmithKline, Ortho McNeill, Valeant Laboratories, Pfizer. Speaker’s Bureau: GlaxoSmithKline, UCB, Pfizer.