- Email: [email protected]
Bipolar pharmacotherapy and suicidal behavior
Journal of Affective Disorders 103 (2007) 13 – 21 www.elsevier.com/locate/jad
Research report
Bipolar pharmacotherapy and suicidal behavior☆ Part 2. The impact of antidepressants Boghos I. Yerevanian a,b,⁎, Ralph J. Koek a,b , Jim Mintz a,c , Hagop S. Akiskal d a
Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles, United States b Sepulveda Ambulatory Care Center, VA Greater Los Angeles Health Care System, United States c Biostatistics Core, Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles, United States d International Mood Center, University of California at San Diego, La Jolla, CA and Veterans Medical Center, La Jolla, CA, United States Received 19 March 2007; received in revised form 26 April 2007; accepted 22 May 2007 Available online 6 July 2007
Abstract Antidepressant-induced mania and cycle acceleration is a potential risk in bipolar patients. Another serious risk of antidepressants, that of increasing suicidal behavior, has been identified in some affectively ill populations. However, there is a dearth of knowledge about the effects of antidepressants on suicidal behavior specifically in bipolar patients. Methods: Retrospective chart review of 405 veterans with bipolar disorder followed for a mean of three years, with month by month systematic assessment of current pharmacotherapy and suicide completion, attempt or hospitalization for suicidality. Chisquared comparison of (log) rates of suicidal events during mood stabilizer monotherapy, antidepressant monotherapy, and combination of mood stabilizer and antidepressant. Results: Suicidal behavior event rates (per 100 patient years) were greatest during treatment with antidepressant monotherapy (25.92), least during mood stabilizer monotherapy (3.48), and intermediate during mood stabilizer + antidepressant combination treatment (9.75). These differences were statistically significant. Limitations: In a clinical setting, antidepressants may have been prescribed because patients were deemed at greater risk of suicidality. Conclusions: During treatment with antidepressants (even when coupled with mood stabilizers), patients with bipolar disorder have significantly higher rates of non-lethal suicidal behavior compared to those on mood stabilizers without antidepressants, and thus require careful monitoring. Published by Elsevier B.V. Keywords: Suicide; Suicide attempt; Antidepressants; Bipolar disorder; Veterans; Longitudinal study; Public health
☆ Role of funding source: Abbott laboratories provided funding for a research assistant. The funding source had no role in the design of the study; nor in the collection, analysis and interpretation of data, writing of the report; and in the decision to submit the paper for publication. Conflict of interest: Dr. Yerevanian has served as a consultant to Abbott Laboratories and has been on the speaker's bureau of Astra Zeneca. Dr. Koek has been on the speaker's bureau of Janssen Pharmaceutica. ⁎ Corresponding author. Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles 16111 Plummer Street (116A-11), North Hills, CA 91343, United States. Tel.: +1 818 891 7711x7123; fax: +1 818 8760546. E-mail address: [email protected] (B.I. Yerevanian).
0165-0327/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.jad.2007.05.017
14
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
1. Introduction The controversy over the issue of antidepressants contributing to suicidal behavior continues in the literature. Numerous epidemiologic and clinical reports (Rihmer and Akiskal, 2006; Baldessarini et al., 2006; Moller, 2006a) actually point to reduction of suicidal behaviors with antidepressant treatment over the long term. However, meta-analyses of short-term clinical trials of antidepressants have found evidence of increased suicidal ideation (Khan et al., 2003) and attempts (Fergusson et al., 2005) during the course of treatment of depressive disorders. Such findings have led the FDA to mandate warnings for all antidepressants (FDA, 9-1404). Patients enrolled in clinical trials of antidepressant for major depression typically are assumed to be non-bipolar and recognized bipolars are explicitly excluded. The question of the effects of antidepressants on suicidality specifically in bipolar patients is controversial (Mc Elroy et al., 2006) and deserves further study. This is of particular importance in light of the emerging evidence that bipolarity – particularly bipolar II – plays a major role in suicide (Rihmer, 2002, 2007). Given that the course of bipolar disorder is dominated by depression (Judd et al., 2002, 2003) and that antidepressants prescribed to depressed patients could destabilize an unknown percentage of them towards suicidality (Akiskal and Mallya, 1987; Koukopoulos et al., 1992; Pompili et al., 2005), a systematic study of the effects of antidepressants in bipolar patients assumes great public health importance. Parallel developments in the field of bipolar research have emphasized the importance of recognizing subtle bipolar conditions. The clinical characteristics of these “pseudo-unipolar” patients have been described (Akiskal et al., 2005; Dilsaver and Akiskal, 2005) and caution has been issued about the negative consequences of treating both overt and sub-clinical bipolar patients with antidepressants alone (Ghaemi et al., 2003). Antidepressant-induced mania/hypomania, induction of rapid cycling as well as mood instability have all been described and the high risk of mixed states for suicidality has been emphasized (Akiskal and Benazzi, 2005). Relatively little attention has been paid to the impact of antidepressants on suicidal behavior in recognized bipolar patients who are often prescribed antidepressants (Post et al., 2003; Simon et al., 2004). In a recent study, Shi et al. (2004) found that unrecognized bipolar patients (who are more likely to be treated with antidepressant monotherapy compared to recognized bipolars) were significantly more likely to attempt suicide (0.9%) than recognized bipolar patients (0.3%) or non-bipolar patients (0.2%). In a cross sectional
study of 1000 bipolar patients in the STEP-BD program, Goldberg et al. (2005) found that suicidal ideation was more prevalent in patients taking antidepressants compared to those who were not (25% versus 14%). In a prospective report from the STEP-BD program, Bauer et al. (2006) reported no association between new onset suicidality (n = 24/425) and increased antidepressant exposure or initiation of antidepressants. Aizenberg et al. (2006), in a case controlled chart review of elderly hospitalized bipolar patients found no increased rate of antidepressant use among 16 admitted after a suicide attempt compared with 16 matched controls without suicide attempt prior to admission. There are thus contradictory findings concerning the impact of antidepressants on suicidal behavior in bipolar patients. In this retrospective study of a large group of bipolar veterans, we sought to answer the questions: Do antidepressants affect suicidal behavior in bipolar patients and if so how? To that end we compared the rates of suicidal behaviors in bipolar patients maintained on 1) monotherapy with mood stabilizers, 2) combination of mood stabilizer and antidepressant, or 3) antidepressant alone. The study was approved by the Institutional Review Board of the VA Greater Los Angeles Healthcare System. 2. Patients and methods 2.1. Clinical setting and patients The study is a retrospective chart review analysis of bipolar patients seen in a large Veterans Administration healthcare system in Southern California (VA Greater Los Angeles Healthcare System or VAGLAHS). The Computerized Patient Record System (CPRS) in use at the facility since 1994 includes progress notes of ALL treating personnel, discharge summaries of any hospitalizations, laboratory tests, records of medications dispensed including refills with dates, patient problem lists, demographic and other data. The period under study was January 1, 1994 to December 31, 2002. To identify patients with bipolar disorder within the system, we obtained pharmacy records of ALL lithium, divalproex, carbamazepine, gabapentin, topiramate, and lamotrigine prescriptions ever dispensed during that period of time. From this a list of 2650 patients was identified, and 405 patients met the following inclusion criteria: 1). A chart diagnosis of bipolar disorder, Type I, Type II, schizoaffective disorder bipolar type or bipolar NOS. For convenience, patients with cyclothymia, antidepressant-induced mania, or mania specified as secondary to other general
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
15
DSM-IV diagnostic criteria were generally in use in this academically-affiliated VA institution during the time period under study. For some patients, the diagnosis changed over time, as new symptoms emerged, or as treating psychiatrists changed. When this was observed, a hierarchical system was adopted with the more “severe” of any two or more bipolar diagnoses assigned. Thus, schizoaffective disorder, bipolar type took precedence over other diagnoses, followed by bipolar I, then bipolar II, and finally bipolar NOS. For example, a bipolar II patient who developed acute mania with hospitalization was considered as bipolar I and entered in the data system as such. Similarly, a patient who was treated for unipolar depression during the early part of the study was included as a bipolar subject if they later developed a (hypo)manic episode. Periods of treatment that such a patient received prior to bipolar disorder becoming evident or diagnosed were included.
Included were mood stabilizers (lithium carbonate or other lithium preparations, divalproex sodium/valproic acid, carbamazepine), antidepressants (SSRI's, tricyclics, MAO Inhibitors, nefazodone, venlafaxine, bupropion, mirtazapine, trazodone [see below]), and antipsychotics (both first and second generation). A month on a given medication or combination was recorded if chart review – progress notes, combined with medication administration history, and when applicable, serum levels – showed that they had been consumed by the patient for more than 15 days of any month. Periods of treatment with lamotrigine, topiramate and oxcarbazepine were excluded from this analysis. Periods of treatment with gabapentin, as well as benzodiazepines, were ignored, as was treatment with non-psychotropic drugs. Trazodone was considered as an antidepressant only at dose of 300 mg/day or more. Otherwise, it was ignored. The independent variables for this report were: months of exposure to 1) mood stabilizer monotherapy, 2) mood stabilizer plus antidepressant, and 3) antidepressant monotherapy. Suicide events occurring during treatment were the dependent variables. The latter were divided into completed suicide (S), attempted suicide (A), and psychiatric hospitalization for suicidal ideation or intent (H). Events were categorized hierarchically, with only the most serious event recorded. Thus, psychiatric hospitalization because of a suicide attempt was counted only as an attempt, since that is the more serious category. A completed suicide during a hospitalization for an attempt was counted as a suicide.
2.3. Stratification of patients by treatment
3. Statistical analysis
Our independent variable was medication treatment received during each month of follow-up, including periods of inpatient care. Our dependent variable was the occurrence of suicidal behavior during a given month of observation. Charts were manually reviewed month by month using a suicidal behavior life chart developed at the Sepulveda VA Mood Clinic. This permits longitudinal graphic display of medications received and the occurrence of suicidal behavior “events” during each month of treatment. A careful review of each note written by mental health professionals in the clinic was undertaken and data entered on the suicidal life chart. Every chart was reviewed by one of the authors (either BYor RK). Where information about medication status or suicidal behavior was conflicting or unclear, a consensus diagnosis was attempted, and if consensus could not be achieved, the period of observation was not included. This was done in order to achieve the highest possible degree of accuracy of the data obtained.
Time was recorded as total time observed rather than time-to-event. For most patients, this included time before and time subsequent to any suicidal events. Rates of events were calculated by medication status over the total time observed. These rates were then compared using the generalized linear regression model, with (log) number of events as the dependent measures, medication status as the independent variable, and (log) total time observed as an offset variable specifying Poisson error as described in Agresti (1990). This method models the rates (events/[time at risk]) as an exponential function of medication status. Details of these regression models are presented with the results.
medical conditions were also grouped with the NOS category; thus, the NOS category in this study subsumed a somewhat broader range of conditions than its DSM-IV counterpart. 2). A minimum of 6 months clinical care for bipolar disorder by an attending or post-graduate psychiatric physician (under attending supervision). 3). Chart documentation of inquiry about suicidal behavior by the psychiatrist. 2.2. Bipolar subtyping
4. Results Table 1 shows the demographic and diagnostic composition of patients who contributed months of exposure to antidepressants (AD ALONE) or mood stabilizer +
16
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
Table 1 Composition of patients contributing to mood stabilizer and antidepressant treatment categories a Category
Patients contributing to treatment with ALL PTS
N 405 M 366 (90.4) F 39 Age 50.1 B-I e 169 (42%) B-II f 68 (17%) SZA g 85 (21%) BIP_NOS h 83 (20.5%)
MS MONO b MS + AD c
AD ALONE d
192 (47.4%) 202 (49.9%) 181 (94.3) 174 (86.1) 11 28 51.4 49.6 103 (53.6) 84 (41.6) 29 (15.1) 52 (25.7) 21 (10.9) 12 (5.9) 39 (20.4) 64 (31.7)
112 (27.7%) 98 (87.5) 14 48.15 34 (30.4) 37 (33) 8 (7.1) 33 (29.5)
a During longitudinal follow-up, because treatment was naturalistic, the same patients could have contributed months to any one or more treatment exposure periods. b Mood stabilizer (lithium, divalproex or carbamazepine) monotherapy. c Mood stabilizer combined with antidepressant. d Antidepressant monotherapy. e Bipolar I disorder. f Bipolar II disorder. g Schizoaffective disorder. h Bipolar disorder, not otherwise specified.
antidepressant (MS + AD) with data from the total sample (ALL PTS), and those who received mood stabilizer monotherapy (MS MONO), presented for comparison. As shown, about half the population of our total sample was treated at some point during the study with MS + AD, and slightly over a quarter with AD
ALONE. Not surprisingly, patients who received antidepressants were somewhat less likely to be diagnosed with bipolar I or schizoaffective disorder, and more likely to receive bipolar II or NOS diagnoses. As shown in Table 2, rates of suicidal behavior were 2– 3 folds greater during MS + AD, compared with MS MONO for all three individual mood stabilizers, but the findings did not reach statistical significance probably due to sample size. However, when all exposure time to MS MONO was compared to total time of exposure to MS + AD, a significantly greater rate of all suicide events during MS + AD was seen. The only completed suicide in the study occurred in a patient on lithium plus antidepressants. In terms of serious events (completions + attempts), it is notable that none occurred during 3785 months of mood stabilizer monotherapy while 6 attempts/completions occurred during mood stabilizer plus antidepressants within 2954 months of observation. Rates of suicide attempts could not be compared using Chi-squared because of a zero value in the MS MONO category. However, the differences between MS MONO and MS + AD were statistically significant when a small value was substituted for the zero (Chi2 = 6.95, df = 1; p = .008, when 0.5 added to 0 event rate for MS MONO; Chi2 = 9.89, df = 1; p = .002, when 0.0001 added to 0 event rate for MS MONO). Table 3 compares suicidal behavior event rates during MS MONO with periods of treatment with AD ALONE. For these two types of treatment, event rates were
Table 2 Comparison of suicidal behavior during mood stabilizer versus mood stabilizer combined with antidepressant Med
Months (years) a d
On Li On Li + AD e On DVPX f On DVPX + AD On Carb g On Carb + AD On MS MONO h On MS + AD
1930 (160.83) 832 (69.33) 1540 (128.33) 1940 (153.33) 315 (26.25) 282 (23.5) 3785 (315.41) 2954 (246.16)
Sb
1
1
SA c
0 1 0 4 0 0 0 5
Hosps c
4 3 6 13 1 2 11 18
Rate per 100 years
Chi2
S+A
All
(All events)
0 2.88 0 2.61 0 0 0 2.44 i
2.49 7.21 4.67 11.08 3.80 8.51 3.48 9.75
2.5; df = 1; p = .11 3.25; df = 1; p = .07 0.46; df = 1; p = .5 8.71; df = 1; p = .003
a Patient months/years of longitudinal follow-up care by a psychiatrist during which it could be determined a) which medications the patient was taking, and b) whether or not a suicide event occurred. See Patients and methods for details. b Completed suicide. c Suicide attempts. d Lithium carbonate, citrate or any other preparation. e Antidepressant. See text for included agents. f Divalproex sodium or valproic acid. g Carbamazepine. h Combined rates for all mood stabilizer monotherapy months (Li + DVPX + Carb). i Different from MS monotherapy total, Chi2 = 6.95, df = 1; p = .008, when 0.5 added to 0 event rate for MS MONO; Chi2 = 9.89, df = 1; p = .002, when 0.0001 added to 0 event rate for MS MONO. See text for explanation.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
17
Table 3 Suicidal behavior during mood stabilizer monotherapy compared with antidepressant monotherapy Med
c
On Li On AD d On DVPX e On AD d On CBZ f On AD d On MS MONO g On AD a b c d e f g
Chi2
Exposure
Suicide events
Rate per 100 years
Months (years) a
SA b
Hosps b
SA
Both
All events
1930 (160.83) 1250 (104.17) 1540 (128.33) 1250 (104.17) 315 (26.25) 1250 (104.17) 3785 (315.41) 1250 (104.17)
0 7 0 7 0 7 0 7
4 20 6 20 1 20 11 20
0 6.72 0 6.72 0 6.72 0 6.72
2.49 25.92 4.67 25.92 3.80 25.92 3.48 25.92
19.14; df = 1; p b .0001 14.4; df = 1; p = .0001 3.55; df = 1; p = . 06 29.87; df = 1; p b .0001
Patient months/years of longitudinal follow-up care by a psychiatrist. See Patients and methods for details. Suicide attempts. Lithium carbonate, citrate or any other preparation. Antidepressant. See text for included agents. Divalproex sodium (or rarely, valproic acid). Carbamazepine. Combined rates for all mood stabilizer monotherapy months (Li + DVPX + Carb).
significantly higher for AD ALONE compared with lithium or divalproex monotherapy, and near-significantly higher for carbamazepine monotherapy. Total event rate during AD monotherapy (AD ALONE) was 7.5 times higher than during the combined time on MS monotherapy. This was highly statistically significant (Chi2 = 29.87, df = 1, p b 0.0001). Again notable is the fact that there were 7 attempts during 1250 months of AD ALONE, and none during 3785 months of MS MONO. Because of the complexity of the model, with individual patients contributing to one or several treatment categories, we identified a group of patients who contributed treatment time in all three treatment modalities under study. As can be seen in Table 4, event rates in this subgroup during each treatment period paralleled those in the overall sample, with the same patients faring best during MS MONO treatment, worst during AD ALONE, and intermediate during MS + AD.
5. Discussion 5.1. Main findings The results of this study suggest that bipolar patients on antidepressant monotherapy or needing antidepressants in the course of mood stabilization have a significantly higher likelihood of suicidal events in the course of follow-up. This observation was true with all three mood stabilizers; there was a 2–3 fold increase in the rate of all suicidal events for each mood stabilizer when used in combination with antidepressants, compared to when it was used as monotherapy. Taken as a group, when comparing all mood stabilizers with all mood stabilizers plus an antidepressant, there was a 2.8 fold statistically significant increase in the rate of suicidality (Table 2). The results suggest a significant instability of the clinical course, at least with respect to
Table 4 Suicidal behavior event rates during antidepressant and/or mood stabilizer in bipolar patients who received all three treatments (N = 24) Exposure period d
On MS MONO On MS + AD e On AD ALONE f
Months (years) a
Ab
Hc
All
Rate (all events/100 pt. years)
179 (14.92) 253 (21.08) 220 (18.33)
0 1 3
2 4 4
2 5 7
13.41 23.72 38.18
a Time these 24 patients received the specified type of treatment; not necessarily continuous, nor necessarily the same agent. (See text for definition). b Attempted suicide. c Hospitalization for suicidal ideation or intent. d Treatment with mood stabilizer (lithium, divalproex sodium, or carbamazepine) alone. e Treatment with mood stabilizer in combination with antidepressant. f Treatment with antidepressant alone.
18
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
suicidal behavior, when antidepressants are used in combination with mood stabilizers in bipolar patients. When comparing antidepressants alone to mood stabilizers alone in this group of bipolar patients, there was a 6–10 fold increase in the rates of suicidal behavior between mood stabilizer monotherapy and antidepressant monotherapy. (Table 3). These differences were highly statistically significant and we believe are also highly clinically significant. Thus, bipolar patients who for various reasons are treated with AD ALONE are at very high risk for suicidal behavior. In this retrospective study of a carefully reviewed group of bipolar patients in a veteran population, we found a high incidence of eventually bipolar patients who were maintained on monotherapy with antidepressants. Thus, 112/405 patients (27.7% of our total bipolar population) contributed to antidepressant monotherapy months. 202/405 patients (49.9% of the population sample) contributed to months of observation on mood stabilizer + antidepressants). These are similar to the incidence of antidepressant exposure in bipolars both from the Stanley foundation bipolar Network and the NIMH STEP-BD Program. There is thus a significant exposure of bipolar patients to antidepressants. 5.2. Methodological considerations The interpretation of our overall finding of antidepressants contributing to suicidality in bipolar patients must take into consideration a major limitation unavoidable in a retrospective study whereby treatments (in particular antidepressant prescription) could not have been randomly assigned. There are indeed many reasons why bipolar patients may be treated by AD monotherapy at least for a period of time: 1) bipolar patients often initially present with depression and the diagnosis
is changed later to bipolar with the onset of mania/ hypomania; 2) patients may refuse mood stabilizers or be unable to tolerate them; 3) disagreements between different treating clinicians about the accurate diagnosis; 4) disagreement among both experts and clinicians whether to prescribe antidepressants to bipolar patients. The latter is not a trivial disagreement (see Akiskal, 2006), since authorities in the field often disagree on the use of antidepressants in bipolar depression (Post et al., 2003; Moller et al., 2006b; Koukopoulos, et al., 2007). During combination of mood stabilizer and antidepressant treatment, patients had rates of suicidal behavior 3 times greater than during mood stabilizer monotherapy (Table 5 and Fig. 1). Thus, our data suggest that the addition of antidepressants in the depressed phase of bipolar disorder is associated with increased suicidal behavior even with the concomitant use of mood stabilizer. In other words, a mood stabilizer may not necessarily protect against antidepressant-associated destabilization. Because we wanted to examine a class effect of antidepressants, we did not, by design, collect information about specific antidepressants or specific AD + MS combinations, for purposes of this study. Although we found a significant statistical and clinical association between increased suicidality and use of antidepressants in bipolar patients, the results need to be interpreted carefully. Instead of a causal relationship, it is possible that confounding by indication may explain some of the differences. This confounding, in turn, could be of two types. First, patients who need (or are judged to need) antidepressants may be inherently more unstable and prone to suicidal behavior. Our subset of 24 patients (Table 4) who received all three types of treatment, however, argues against this, suggesting instead that it is differences in the treatments, not the patients, that account for differences in event rates. It is also
Table 5 Suicidal behavior during antidepressant and/or mood stabilizer treatment (N = 405) Sa
MS MONO e MS + AD f AD MONO g a b c d e f g
0 1 0
Ab
0 5 7
Hc
11 18 20
All
11 24 27
Months (years) of exposure d
Events/100 pt. years All events
Severe events (S + A)
3785 (315.41) 2954 (246.16) 1250 (104.17)
3.48 9.75 25.92
0 2.44 6.72
Completed suicide. Attempted suicide. Hospitalization for suicidal ideation or intent. Patient months/years of longitudinal follow-up care by a psychiatrist. See Patients and methods for details. Treatment with mood stabilizer (lithium, divalproex sodium, or carbamazepine) alone. Treatment with mood stabilizer in combination with antidepressant. Treatment with antidepressant alone. Data in this table are represented graphically in Fig. 1.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
19
Fig. 1. Suicidal behavior in bipolar veterans during mood stabilizer and/or antidepressant treatment (N = 405).
possible that confounding by phase of illness accounts for differences in event rates, with patients receiving antidepressants because they become more depressed or suicidal. The current dataset does not allow for controlling of this variable and further research with a prospective design may clarify the issue. However, even if those bipolar patients who developed suicidality on antidepressants (singly or in combination) received the antidepressant because of more severe depression or unstable illness, the fact remains that the addition of antidepressants did not diminish the rate of suicide events. These rates remained quite high in this group. The data in Fig. 1 (increased suicide events progressively across the three types of treatment exposure periods, from least with MS MONO, higher during MS + AD, through highest during AD ALONE) suggest that destabilization of bipolar patients during antidepressant treatment is only partially mitigated by concomitant mood stabilizer treatment. The data can be interpreted as meaning that even in the presence of mood stabilizers, antidepressant treated bipolar patients remain at higher risk for suicidal behavior. This study is limited by the fact that not all potential suicide risk factors were controlled for. Suicidal behavior is complex and many psychosocial factors and comorbidities are often involved (Rihmer, 2007). Meaningful control of these variables was not possible because of the retrospective design of the study and the fact that patients contributed months of observation in multiple treatment categories. We did not control for alcoholism in
this group, but we note that recent prospective studies of bipolar patients (Valtonen et al., 2006; Marangell et al., 2006) have shown that potential risk factors including alcoholism and co-morbidities per se were not significantly predictive of suicidal outcomes, yet previous suicide attempt and days depressed were. The latter acquire great significance in light of long-term prospective studies showing the domination of the course of bipolar disorder with greater percentage of days spent in depression than previously thought (Judd et al., 2002, 2003). We also did not control for the contributions of numerous other risk factors for suicide, including previous attempts, psychosocial factors, medical/substance use co-morbidity, and concomitant use of benzodiazepines and medications prescribed for general medical conditions. Because there was only one case of completed suicide, the results of this study are only applicable to non-lethal suicidal behavior. However, non-lethal suicidal behaviors in bipolar patients are clinically more serious than in the general population (Tondo and Baldessarini, 2000), and serious attempts are more like than unlike completed suicides (Maser et al., 2002). 5.3. Implications for clinical practice Our finding of a significant association of suicidal behavior and antidepressants in a bipolar cohort consistent with the hypotheses that the emergence of suicidality during treatment of major depression could be taken as a
20
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
marker of future bipolar course if not bona fide bipolar spectrum (Akiskal and Mallya, 1987; Akiskal and Benazzi, 2003). Furthermore, it has been suggested that (Dilsaver and Akiskal, 2005; Berk and Dodd, 2005) the excess of suicide attempts in younger patients given antidepressants in controlled trials may be attributable to unrecognized bipolar disorder in this population—given the younger age of onset of bipolar disorder, coupled with the frequent onset of bipolar illness with a depressive episode. Indeed, it has long been suggested that prepubertal depression is the precursor of bipolar transformation, especially in the presence of bipolar family history (Akiskal, 1995). In light of our results and the foregoing considerations, it is of great public health concern that the majority of bipolar patients continue to be prescribed antidepressant monotherapy (Baldessarini et al., 2007). Prospective randomized studies with larger populations are needed to address the potential cause–effect relationship between antidepressants and suicidality in bipolar patients but unlikely to be undertaken because of ethical constraints. In the meantime the results of the study suggest that antidepressants should be used cautiously in bipolar as well as potentially bipolar depressives. Depressive mixed states, unrecognized in the official DSM-IVTR classification, appear to be a particularly important condition to be vigilant about in clinical practice in adults (Akiskal and Mallya, 1987; Koukopoulos et al., 1992), in the community (Balazs et al., 2006), and of course in juvenile subjects (Dilsaver and Akiskal, 2005). Such depressive mixed states (activated depression) might also be particularly relevant in the context of PTSD and alcohol and substance abuse, two common conditions among veteran populations seen in mental health clinics. Our data of increased suicidality during antidepressant treatment of bipolar patients should be interpreted within a multifactorial model of suicidality in veterans as well as the general population. Acknowledgement The study was partially funded by Abbott Laboratories. References Agresti, A., 1990. Categorical Data Analysis. John Wiley and Sons, New York, NY, pp. 189–196. Aizenberg, D., Olmer, A., Barak, Y., 2006. Suicide attempts amongst elderly bipolar patients. J. Affect. Disord. 91, 91–94. Akiskal, H.S., Mallya, G., 1987. Criteria for the “soft” bipolar spectrum: treatment implications. Psychopharmacol. Bull. 23, 68–73. Akiskal, H.S., 1995. Developmental pathways to bipolarity: are juvenileonset depressions pre-bipolar? J. Am. Acad. Child Adolesc. Psych. 34, 754–763.
Akiskal, H.S., 2006. What is the optimum strategy for the clinical management of bipolar depression? J. Bipolar Disord. V 18. Akiskal, H.S., Benazzi, F., 2003. Family history validation of the bipolar nature of depressive mixed states. J. Affect. Disord. 73, 113–122. Akiskal, H.S., Benazzi, F., 2005. Psychopathologic correlates of suicidal ideation in major depressive outpatients: is it all due to unrecognized (bipolar) depressive mixed states? Psychopathology 38, 273–280. Akiskal, H.S., Benazzi, F., Perugi, G., Rihmer, Z., 2005. Agitated “unipolar” depression re-conceptualized as a depressive mixed state: implications for the antidepressant-suicide controversy. J. Affect. Disord. 85, 245–258. Baldessarini, R.J., Pompili, M., Tondo, L., 2006. Suicidal risk in antidepressant drug trials. Arch. Gen. Psychiatry 63, 246–248. Baldessarini, R.J., Leahy, L., Arcona, S., Gause, D., Zhang, W., Hennen, J., 2007. Patterns of psychotropic drug prescription for U.S. patients with diagnoses of bipolar disorders. Psychiatr. Serv. 58, 85–91. Balazs, J., Benazzi, F., Rihmer, Z., Rihmer, A., Akiskal, K.K., Akiskal, H.S., 2006. The close link between suicide attempts and mixed (bipolar) depression: implications for suicide prevention. J. Affect. Disord. 91, 133–138. Bauer, M.S., Wisniewski, S.R., Marangell, L.B., Chessick, C.A., Allen, M.H., Dennehy, E.B., Miklowitz, D.J., Thase, M.E., Sachs, G.S., 2006. Are antidepressants associated with new-onset suicidality in bipolar disorder? A prospective study of participants in the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD). J. Clin. Psychiatry 67, 48–55. Berk, M., Dodd, S., 2005. Are treatment emergent suicidality and decreased response to antidepressants in younger patients due to bipolar disorder being misdiagnosed as unipolar depression? Med. Hypotheses 65, 39–43. Dilsaver, S.C., Akiskal, H.S., 2005. High rate of unrecognized bipolar mixed states among destitute Hispanic adolescents referred for “major depressive disorder”. J. Affect. Disord. 84, 179–186. Fergusson, D., Doucette, S., Glass, K.C., Shapiro, S., Healy, D., Hebert, P., Hutton, B., 2005. Association between suicide attempts and selective serotonin reuptake inhibitors: systematic review of randomised controlled trials. BMJ 330 (7488), 396. Food and Drug Administration (9-14-04), www.FDA.gov/bbs/topics/ news/2004/new01116.html. Ghaemi, S.N., Hsu, D.J., Soldani, F., Goodwin, F.K., 2003. Antidepressants in bipolar disorder: the case for caution. Bipolar Disord. 5, 421–433. Goldberg, J.F., Allen, M.H., Miklowitz, D.A., Bowden, C.L., Endick, C.J., Chessick, C.A., Wisniewski, S.R., Miyahara, S., Sagduyu, K., Thase, M.E., Calabrese, J.R., Sachs, G.S., 2005. Suicidal ideation and pharmacotherapy among STEP-BD patients. Psychiatr. Serv. 56, 1534–1540. Judd, L.L., Akiskal, H.S., Schettler, P.J., Endicott, J., Maser, J., Solomon, D.A., Leon, A.C., Rice, J.A., Keller, M.B., 2002. The long-term natural history of the weekly symptomatic status of bipolar-I disorder. Arch. Gen. Psychiatry 59, 530–537. Judd, L.L., Akiskal, H.S., Schettler, P., Endicott, J., Maser, J., Solomon, D., Leon, A., Coryell, W., Keller, M., 2003. A prospective investigation of the natural history of the long-term weekly symptomatic status of bipolar II disorder. Arch. Gen. Psychiatry 60, 261–269. Khan, A., Khan, S., Kolts, R., Brown, W.A., 2003. Suicide rates in clinical trials of SSRIs, other antidepressants, and placebo: analysis of FDA reports. Am. J. Psychiatry 160, 790–792. Koukopoulos, A., Faedda, G., Proietti, R., D'Amico, S., de Pisa, E., Simonetto, C., 1992. Mixed depressive syndromes. Encéphale 18 (Spec No 1), 19–21.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21 Koukopoulos, A., Sani, G., Koukopoulos, A.E., Manfredi, G., Pacchiarotti, I., Girardi, P., 2007. Melancholia agitata and mixed depression. Acta. Psychiatr. Scand. Suppl. 433, 50–57. Marangell, L.B., Bauer, M.S., Dennehy, E.B., Wisniewsky, S.R., Allen, M.H., Miklowitz, D.J., Oquendo, M.A., Frank, E., Perlis, R.H., Martinez, J.M., Fagiolini, A., Otto, M.W., Chessick, C.A., Zboyan, H.A., Miyahara, S., Sachs, G., Thase, M.E., 2006. Prospective predictors of suicide and suicide attempts in 1556 patients with bipolar disorders followed for up to 2 years. Bipolar Disord. 8, 566–575. Maser, J.D., Akiskal, H.S., Zeller, P., Scheftner, W., Mueller, T., Endicott, J., Solomon, D., Clayton, P., 2002. Can temperament identify affectively ill patients who engage in lethal or near-lethal suicidal behavior? A 14-year prospective study. Suicide. LifeThreat. Behav. 32, 10–32. McElroy, S.L., Kotwal, R., Kaneria, R., Keck Jr, P.E., 2006. Antidepressants and suicidal behavior in bipolar disorder. Bipolar Disord. 8, 596–617. Moller, H.-J., 2006a. Evidence for beneficial effects of antidepressants on suicidality in depressive patients. A systematic review. Eur. Arch. Psychiatry Clin. Neurosci. 256, 229–343. Moller, H.-J., Grunze, H., Broich, K., 2006b. Do recent efficacy data on the drug treatment of acute bipolar depression support the position that drugs other than antidepressants are the treatment of choice? A conceptual review. Eur. Arch. Psychiatry Clin. Neurosci. 256, 1–16. Pompili, M., Tondo, L., Baldessarini, B., 2005. Suicidal risk emerging during antidepressant treatment: recognition and intervention. Clin. Neuropsychiatry 2, 66–72. Post, R.M., Leverich, G.S., Nolen, W.A., Kupka, R.W., Altshuler, L.L., Frye, M.A., Suppes, T., McElroy, S., Keck, P., Grunze, H., Walden, J.,
21
2003. A re-evaluation of the role of antidepressants in the treatment of bipolar depression: data from the Stanley Foundation Bipolar Network. Bipolar Disord. 5, 396–406. Rihmer, Z., 2002. Suicide risk in mood disorders. Curr. Opin. Psychiatry Jan. 20 (1), 17–22. Rihmer, Z., Akiskal, H.S., 2006. Do antidepressants t(h)reat(en) depressives? Toward a clinically judicious formulation of the antidepressant-suicidality FDA advisory in light of declining national suicide statistics from many countries. J. Affect. Disord. 94, 3–13. Rihmer, Z., 2007. Suicidal risk in mood disorders. Curr. Opin. Psychiatry 20, 17–22. Shi, L., Thiebaud, P., McCombs, J.S., 2004. The impact of unrecognized bipolar disorders for patients treated for depression with antidepressants in the fee-for-service California Medicaid (Medi-Cal) programme. J. Affect. Disord. 82, 373–383. Simon, N., Otto, M., Weiss, R., Bauer, M.S., Miyahara, S., Wisniewski, S.R., Thase, M.E., Kogan, J., Frank, E., Nierenberg, A.A., Calabrese, J.R., Sachs, G.S., Pollack, M.H., 2004. Pharmacotherapy for bipolar disorder and co-morbid conditions: baseline data from STEP-BD. J. Clin. Psychopharmacol. 24, 512–520. Tondo, L., Baldessarini, R.J., 2000. Reduced suicide risk during lithium maintenance treatment. J. Clin. Psychiatry, 61 (suppl 9), 97–104. Valtonen, H.M., Suominen, K., Mantere, O., Leppamaki, S., Arvilommi, P., Isometsa, E.T., 2006. Prospective study of risk factors for attempted suicide among patients with bipolar disorder. Bipolar Disord. 8, 576–585.
Research report
Bipolar pharmacotherapy and suicidal behavior☆ Part 2. The impact of antidepressants Boghos I. Yerevanian a,b,⁎, Ralph J. Koek a,b , Jim Mintz a,c , Hagop S. Akiskal d a
Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles, United States b Sepulveda Ambulatory Care Center, VA Greater Los Angeles Health Care System, United States c Biostatistics Core, Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles, United States d International Mood Center, University of California at San Diego, La Jolla, CA and Veterans Medical Center, La Jolla, CA, United States Received 19 March 2007; received in revised form 26 April 2007; accepted 22 May 2007 Available online 6 July 2007
Abstract Antidepressant-induced mania and cycle acceleration is a potential risk in bipolar patients. Another serious risk of antidepressants, that of increasing suicidal behavior, has been identified in some affectively ill populations. However, there is a dearth of knowledge about the effects of antidepressants on suicidal behavior specifically in bipolar patients. Methods: Retrospective chart review of 405 veterans with bipolar disorder followed for a mean of three years, with month by month systematic assessment of current pharmacotherapy and suicide completion, attempt or hospitalization for suicidality. Chisquared comparison of (log) rates of suicidal events during mood stabilizer monotherapy, antidepressant monotherapy, and combination of mood stabilizer and antidepressant. Results: Suicidal behavior event rates (per 100 patient years) were greatest during treatment with antidepressant monotherapy (25.92), least during mood stabilizer monotherapy (3.48), and intermediate during mood stabilizer + antidepressant combination treatment (9.75). These differences were statistically significant. Limitations: In a clinical setting, antidepressants may have been prescribed because patients were deemed at greater risk of suicidality. Conclusions: During treatment with antidepressants (even when coupled with mood stabilizers), patients with bipolar disorder have significantly higher rates of non-lethal suicidal behavior compared to those on mood stabilizers without antidepressants, and thus require careful monitoring. Published by Elsevier B.V. Keywords: Suicide; Suicide attempt; Antidepressants; Bipolar disorder; Veterans; Longitudinal study; Public health
☆ Role of funding source: Abbott laboratories provided funding for a research assistant. The funding source had no role in the design of the study; nor in the collection, analysis and interpretation of data, writing of the report; and in the decision to submit the paper for publication. Conflict of interest: Dr. Yerevanian has served as a consultant to Abbott Laboratories and has been on the speaker's bureau of Astra Zeneca. Dr. Koek has been on the speaker's bureau of Janssen Pharmaceutica. ⁎ Corresponding author. Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine, University of California, Los Angeles 16111 Plummer Street (116A-11), North Hills, CA 91343, United States. Tel.: +1 818 891 7711x7123; fax: +1 818 8760546. E-mail address: [email protected] (B.I. Yerevanian).
0165-0327/$ - see front matter. Published by Elsevier B.V. doi:10.1016/j.jad.2007.05.017
14
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
1. Introduction The controversy over the issue of antidepressants contributing to suicidal behavior continues in the literature. Numerous epidemiologic and clinical reports (Rihmer and Akiskal, 2006; Baldessarini et al., 2006; Moller, 2006a) actually point to reduction of suicidal behaviors with antidepressant treatment over the long term. However, meta-analyses of short-term clinical trials of antidepressants have found evidence of increased suicidal ideation (Khan et al., 2003) and attempts (Fergusson et al., 2005) during the course of treatment of depressive disorders. Such findings have led the FDA to mandate warnings for all antidepressants (FDA, 9-1404). Patients enrolled in clinical trials of antidepressant for major depression typically are assumed to be non-bipolar and recognized bipolars are explicitly excluded. The question of the effects of antidepressants on suicidality specifically in bipolar patients is controversial (Mc Elroy et al., 2006) and deserves further study. This is of particular importance in light of the emerging evidence that bipolarity – particularly bipolar II – plays a major role in suicide (Rihmer, 2002, 2007). Given that the course of bipolar disorder is dominated by depression (Judd et al., 2002, 2003) and that antidepressants prescribed to depressed patients could destabilize an unknown percentage of them towards suicidality (Akiskal and Mallya, 1987; Koukopoulos et al., 1992; Pompili et al., 2005), a systematic study of the effects of antidepressants in bipolar patients assumes great public health importance. Parallel developments in the field of bipolar research have emphasized the importance of recognizing subtle bipolar conditions. The clinical characteristics of these “pseudo-unipolar” patients have been described (Akiskal et al., 2005; Dilsaver and Akiskal, 2005) and caution has been issued about the negative consequences of treating both overt and sub-clinical bipolar patients with antidepressants alone (Ghaemi et al., 2003). Antidepressant-induced mania/hypomania, induction of rapid cycling as well as mood instability have all been described and the high risk of mixed states for suicidality has been emphasized (Akiskal and Benazzi, 2005). Relatively little attention has been paid to the impact of antidepressants on suicidal behavior in recognized bipolar patients who are often prescribed antidepressants (Post et al., 2003; Simon et al., 2004). In a recent study, Shi et al. (2004) found that unrecognized bipolar patients (who are more likely to be treated with antidepressant monotherapy compared to recognized bipolars) were significantly more likely to attempt suicide (0.9%) than recognized bipolar patients (0.3%) or non-bipolar patients (0.2%). In a cross sectional
study of 1000 bipolar patients in the STEP-BD program, Goldberg et al. (2005) found that suicidal ideation was more prevalent in patients taking antidepressants compared to those who were not (25% versus 14%). In a prospective report from the STEP-BD program, Bauer et al. (2006) reported no association between new onset suicidality (n = 24/425) and increased antidepressant exposure or initiation of antidepressants. Aizenberg et al. (2006), in a case controlled chart review of elderly hospitalized bipolar patients found no increased rate of antidepressant use among 16 admitted after a suicide attempt compared with 16 matched controls without suicide attempt prior to admission. There are thus contradictory findings concerning the impact of antidepressants on suicidal behavior in bipolar patients. In this retrospective study of a large group of bipolar veterans, we sought to answer the questions: Do antidepressants affect suicidal behavior in bipolar patients and if so how? To that end we compared the rates of suicidal behaviors in bipolar patients maintained on 1) monotherapy with mood stabilizers, 2) combination of mood stabilizer and antidepressant, or 3) antidepressant alone. The study was approved by the Institutional Review Board of the VA Greater Los Angeles Healthcare System. 2. Patients and methods 2.1. Clinical setting and patients The study is a retrospective chart review analysis of bipolar patients seen in a large Veterans Administration healthcare system in Southern California (VA Greater Los Angeles Healthcare System or VAGLAHS). The Computerized Patient Record System (CPRS) in use at the facility since 1994 includes progress notes of ALL treating personnel, discharge summaries of any hospitalizations, laboratory tests, records of medications dispensed including refills with dates, patient problem lists, demographic and other data. The period under study was January 1, 1994 to December 31, 2002. To identify patients with bipolar disorder within the system, we obtained pharmacy records of ALL lithium, divalproex, carbamazepine, gabapentin, topiramate, and lamotrigine prescriptions ever dispensed during that period of time. From this a list of 2650 patients was identified, and 405 patients met the following inclusion criteria: 1). A chart diagnosis of bipolar disorder, Type I, Type II, schizoaffective disorder bipolar type or bipolar NOS. For convenience, patients with cyclothymia, antidepressant-induced mania, or mania specified as secondary to other general
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
15
DSM-IV diagnostic criteria were generally in use in this academically-affiliated VA institution during the time period under study. For some patients, the diagnosis changed over time, as new symptoms emerged, or as treating psychiatrists changed. When this was observed, a hierarchical system was adopted with the more “severe” of any two or more bipolar diagnoses assigned. Thus, schizoaffective disorder, bipolar type took precedence over other diagnoses, followed by bipolar I, then bipolar II, and finally bipolar NOS. For example, a bipolar II patient who developed acute mania with hospitalization was considered as bipolar I and entered in the data system as such. Similarly, a patient who was treated for unipolar depression during the early part of the study was included as a bipolar subject if they later developed a (hypo)manic episode. Periods of treatment that such a patient received prior to bipolar disorder becoming evident or diagnosed were included.
Included were mood stabilizers (lithium carbonate or other lithium preparations, divalproex sodium/valproic acid, carbamazepine), antidepressants (SSRI's, tricyclics, MAO Inhibitors, nefazodone, venlafaxine, bupropion, mirtazapine, trazodone [see below]), and antipsychotics (both first and second generation). A month on a given medication or combination was recorded if chart review – progress notes, combined with medication administration history, and when applicable, serum levels – showed that they had been consumed by the patient for more than 15 days of any month. Periods of treatment with lamotrigine, topiramate and oxcarbazepine were excluded from this analysis. Periods of treatment with gabapentin, as well as benzodiazepines, were ignored, as was treatment with non-psychotropic drugs. Trazodone was considered as an antidepressant only at dose of 300 mg/day or more. Otherwise, it was ignored. The independent variables for this report were: months of exposure to 1) mood stabilizer monotherapy, 2) mood stabilizer plus antidepressant, and 3) antidepressant monotherapy. Suicide events occurring during treatment were the dependent variables. The latter were divided into completed suicide (S), attempted suicide (A), and psychiatric hospitalization for suicidal ideation or intent (H). Events were categorized hierarchically, with only the most serious event recorded. Thus, psychiatric hospitalization because of a suicide attempt was counted only as an attempt, since that is the more serious category. A completed suicide during a hospitalization for an attempt was counted as a suicide.
2.3. Stratification of patients by treatment
3. Statistical analysis
Our independent variable was medication treatment received during each month of follow-up, including periods of inpatient care. Our dependent variable was the occurrence of suicidal behavior during a given month of observation. Charts were manually reviewed month by month using a suicidal behavior life chart developed at the Sepulveda VA Mood Clinic. This permits longitudinal graphic display of medications received and the occurrence of suicidal behavior “events” during each month of treatment. A careful review of each note written by mental health professionals in the clinic was undertaken and data entered on the suicidal life chart. Every chart was reviewed by one of the authors (either BYor RK). Where information about medication status or suicidal behavior was conflicting or unclear, a consensus diagnosis was attempted, and if consensus could not be achieved, the period of observation was not included. This was done in order to achieve the highest possible degree of accuracy of the data obtained.
Time was recorded as total time observed rather than time-to-event. For most patients, this included time before and time subsequent to any suicidal events. Rates of events were calculated by medication status over the total time observed. These rates were then compared using the generalized linear regression model, with (log) number of events as the dependent measures, medication status as the independent variable, and (log) total time observed as an offset variable specifying Poisson error as described in Agresti (1990). This method models the rates (events/[time at risk]) as an exponential function of medication status. Details of these regression models are presented with the results.
medical conditions were also grouped with the NOS category; thus, the NOS category in this study subsumed a somewhat broader range of conditions than its DSM-IV counterpart. 2). A minimum of 6 months clinical care for bipolar disorder by an attending or post-graduate psychiatric physician (under attending supervision). 3). Chart documentation of inquiry about suicidal behavior by the psychiatrist. 2.2. Bipolar subtyping
4. Results Table 1 shows the demographic and diagnostic composition of patients who contributed months of exposure to antidepressants (AD ALONE) or mood stabilizer +
16
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
Table 1 Composition of patients contributing to mood stabilizer and antidepressant treatment categories a Category
Patients contributing to treatment with ALL PTS
N 405 M 366 (90.4) F 39 Age 50.1 B-I e 169 (42%) B-II f 68 (17%) SZA g 85 (21%) BIP_NOS h 83 (20.5%)
MS MONO b MS + AD c
AD ALONE d
192 (47.4%) 202 (49.9%) 181 (94.3) 174 (86.1) 11 28 51.4 49.6 103 (53.6) 84 (41.6) 29 (15.1) 52 (25.7) 21 (10.9) 12 (5.9) 39 (20.4) 64 (31.7)
112 (27.7%) 98 (87.5) 14 48.15 34 (30.4) 37 (33) 8 (7.1) 33 (29.5)
a During longitudinal follow-up, because treatment was naturalistic, the same patients could have contributed months to any one or more treatment exposure periods. b Mood stabilizer (lithium, divalproex or carbamazepine) monotherapy. c Mood stabilizer combined with antidepressant. d Antidepressant monotherapy. e Bipolar I disorder. f Bipolar II disorder. g Schizoaffective disorder. h Bipolar disorder, not otherwise specified.
antidepressant (MS + AD) with data from the total sample (ALL PTS), and those who received mood stabilizer monotherapy (MS MONO), presented for comparison. As shown, about half the population of our total sample was treated at some point during the study with MS + AD, and slightly over a quarter with AD
ALONE. Not surprisingly, patients who received antidepressants were somewhat less likely to be diagnosed with bipolar I or schizoaffective disorder, and more likely to receive bipolar II or NOS diagnoses. As shown in Table 2, rates of suicidal behavior were 2– 3 folds greater during MS + AD, compared with MS MONO for all three individual mood stabilizers, but the findings did not reach statistical significance probably due to sample size. However, when all exposure time to MS MONO was compared to total time of exposure to MS + AD, a significantly greater rate of all suicide events during MS + AD was seen. The only completed suicide in the study occurred in a patient on lithium plus antidepressants. In terms of serious events (completions + attempts), it is notable that none occurred during 3785 months of mood stabilizer monotherapy while 6 attempts/completions occurred during mood stabilizer plus antidepressants within 2954 months of observation. Rates of suicide attempts could not be compared using Chi-squared because of a zero value in the MS MONO category. However, the differences between MS MONO and MS + AD were statistically significant when a small value was substituted for the zero (Chi2 = 6.95, df = 1; p = .008, when 0.5 added to 0 event rate for MS MONO; Chi2 = 9.89, df = 1; p = .002, when 0.0001 added to 0 event rate for MS MONO). Table 3 compares suicidal behavior event rates during MS MONO with periods of treatment with AD ALONE. For these two types of treatment, event rates were
Table 2 Comparison of suicidal behavior during mood stabilizer versus mood stabilizer combined with antidepressant Med
Months (years) a d
On Li On Li + AD e On DVPX f On DVPX + AD On Carb g On Carb + AD On MS MONO h On MS + AD
1930 (160.83) 832 (69.33) 1540 (128.33) 1940 (153.33) 315 (26.25) 282 (23.5) 3785 (315.41) 2954 (246.16)
Sb
1
1
SA c
0 1 0 4 0 0 0 5
Hosps c
4 3 6 13 1 2 11 18
Rate per 100 years
Chi2
S+A
All
(All events)
0 2.88 0 2.61 0 0 0 2.44 i
2.49 7.21 4.67 11.08 3.80 8.51 3.48 9.75
2.5; df = 1; p = .11 3.25; df = 1; p = .07 0.46; df = 1; p = .5 8.71; df = 1; p = .003
a Patient months/years of longitudinal follow-up care by a psychiatrist during which it could be determined a) which medications the patient was taking, and b) whether or not a suicide event occurred. See Patients and methods for details. b Completed suicide. c Suicide attempts. d Lithium carbonate, citrate or any other preparation. e Antidepressant. See text for included agents. f Divalproex sodium or valproic acid. g Carbamazepine. h Combined rates for all mood stabilizer monotherapy months (Li + DVPX + Carb). i Different from MS monotherapy total, Chi2 = 6.95, df = 1; p = .008, when 0.5 added to 0 event rate for MS MONO; Chi2 = 9.89, df = 1; p = .002, when 0.0001 added to 0 event rate for MS MONO. See text for explanation.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
17
Table 3 Suicidal behavior during mood stabilizer monotherapy compared with antidepressant monotherapy Med
c
On Li On AD d On DVPX e On AD d On CBZ f On AD d On MS MONO g On AD a b c d e f g
Chi2
Exposure
Suicide events
Rate per 100 years
Months (years) a
SA b
Hosps b
SA
Both
All events
1930 (160.83) 1250 (104.17) 1540 (128.33) 1250 (104.17) 315 (26.25) 1250 (104.17) 3785 (315.41) 1250 (104.17)
0 7 0 7 0 7 0 7
4 20 6 20 1 20 11 20
0 6.72 0 6.72 0 6.72 0 6.72
2.49 25.92 4.67 25.92 3.80 25.92 3.48 25.92
19.14; df = 1; p b .0001 14.4; df = 1; p = .0001 3.55; df = 1; p = . 06 29.87; df = 1; p b .0001
Patient months/years of longitudinal follow-up care by a psychiatrist. See Patients and methods for details. Suicide attempts. Lithium carbonate, citrate or any other preparation. Antidepressant. See text for included agents. Divalproex sodium (or rarely, valproic acid). Carbamazepine. Combined rates for all mood stabilizer monotherapy months (Li + DVPX + Carb).
significantly higher for AD ALONE compared with lithium or divalproex monotherapy, and near-significantly higher for carbamazepine monotherapy. Total event rate during AD monotherapy (AD ALONE) was 7.5 times higher than during the combined time on MS monotherapy. This was highly statistically significant (Chi2 = 29.87, df = 1, p b 0.0001). Again notable is the fact that there were 7 attempts during 1250 months of AD ALONE, and none during 3785 months of MS MONO. Because of the complexity of the model, with individual patients contributing to one or several treatment categories, we identified a group of patients who contributed treatment time in all three treatment modalities under study. As can be seen in Table 4, event rates in this subgroup during each treatment period paralleled those in the overall sample, with the same patients faring best during MS MONO treatment, worst during AD ALONE, and intermediate during MS + AD.
5. Discussion 5.1. Main findings The results of this study suggest that bipolar patients on antidepressant monotherapy or needing antidepressants in the course of mood stabilization have a significantly higher likelihood of suicidal events in the course of follow-up. This observation was true with all three mood stabilizers; there was a 2–3 fold increase in the rate of all suicidal events for each mood stabilizer when used in combination with antidepressants, compared to when it was used as monotherapy. Taken as a group, when comparing all mood stabilizers with all mood stabilizers plus an antidepressant, there was a 2.8 fold statistically significant increase in the rate of suicidality (Table 2). The results suggest a significant instability of the clinical course, at least with respect to
Table 4 Suicidal behavior event rates during antidepressant and/or mood stabilizer in bipolar patients who received all three treatments (N = 24) Exposure period d
On MS MONO On MS + AD e On AD ALONE f
Months (years) a
Ab
Hc
All
Rate (all events/100 pt. years)
179 (14.92) 253 (21.08) 220 (18.33)
0 1 3
2 4 4
2 5 7
13.41 23.72 38.18
a Time these 24 patients received the specified type of treatment; not necessarily continuous, nor necessarily the same agent. (See text for definition). b Attempted suicide. c Hospitalization for suicidal ideation or intent. d Treatment with mood stabilizer (lithium, divalproex sodium, or carbamazepine) alone. e Treatment with mood stabilizer in combination with antidepressant. f Treatment with antidepressant alone.
18
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
suicidal behavior, when antidepressants are used in combination with mood stabilizers in bipolar patients. When comparing antidepressants alone to mood stabilizers alone in this group of bipolar patients, there was a 6–10 fold increase in the rates of suicidal behavior between mood stabilizer monotherapy and antidepressant monotherapy. (Table 3). These differences were highly statistically significant and we believe are also highly clinically significant. Thus, bipolar patients who for various reasons are treated with AD ALONE are at very high risk for suicidal behavior. In this retrospective study of a carefully reviewed group of bipolar patients in a veteran population, we found a high incidence of eventually bipolar patients who were maintained on monotherapy with antidepressants. Thus, 112/405 patients (27.7% of our total bipolar population) contributed to antidepressant monotherapy months. 202/405 patients (49.9% of the population sample) contributed to months of observation on mood stabilizer + antidepressants). These are similar to the incidence of antidepressant exposure in bipolars both from the Stanley foundation bipolar Network and the NIMH STEP-BD Program. There is thus a significant exposure of bipolar patients to antidepressants. 5.2. Methodological considerations The interpretation of our overall finding of antidepressants contributing to suicidality in bipolar patients must take into consideration a major limitation unavoidable in a retrospective study whereby treatments (in particular antidepressant prescription) could not have been randomly assigned. There are indeed many reasons why bipolar patients may be treated by AD monotherapy at least for a period of time: 1) bipolar patients often initially present with depression and the diagnosis
is changed later to bipolar with the onset of mania/ hypomania; 2) patients may refuse mood stabilizers or be unable to tolerate them; 3) disagreements between different treating clinicians about the accurate diagnosis; 4) disagreement among both experts and clinicians whether to prescribe antidepressants to bipolar patients. The latter is not a trivial disagreement (see Akiskal, 2006), since authorities in the field often disagree on the use of antidepressants in bipolar depression (Post et al., 2003; Moller et al., 2006b; Koukopoulos, et al., 2007). During combination of mood stabilizer and antidepressant treatment, patients had rates of suicidal behavior 3 times greater than during mood stabilizer monotherapy (Table 5 and Fig. 1). Thus, our data suggest that the addition of antidepressants in the depressed phase of bipolar disorder is associated with increased suicidal behavior even with the concomitant use of mood stabilizer. In other words, a mood stabilizer may not necessarily protect against antidepressant-associated destabilization. Because we wanted to examine a class effect of antidepressants, we did not, by design, collect information about specific antidepressants or specific AD + MS combinations, for purposes of this study. Although we found a significant statistical and clinical association between increased suicidality and use of antidepressants in bipolar patients, the results need to be interpreted carefully. Instead of a causal relationship, it is possible that confounding by indication may explain some of the differences. This confounding, in turn, could be of two types. First, patients who need (or are judged to need) antidepressants may be inherently more unstable and prone to suicidal behavior. Our subset of 24 patients (Table 4) who received all three types of treatment, however, argues against this, suggesting instead that it is differences in the treatments, not the patients, that account for differences in event rates. It is also
Table 5 Suicidal behavior during antidepressant and/or mood stabilizer treatment (N = 405) Sa
MS MONO e MS + AD f AD MONO g a b c d e f g
0 1 0
Ab
0 5 7
Hc
11 18 20
All
11 24 27
Months (years) of exposure d
Events/100 pt. years All events
Severe events (S + A)
3785 (315.41) 2954 (246.16) 1250 (104.17)
3.48 9.75 25.92
0 2.44 6.72
Completed suicide. Attempted suicide. Hospitalization for suicidal ideation or intent. Patient months/years of longitudinal follow-up care by a psychiatrist. See Patients and methods for details. Treatment with mood stabilizer (lithium, divalproex sodium, or carbamazepine) alone. Treatment with mood stabilizer in combination with antidepressant. Treatment with antidepressant alone. Data in this table are represented graphically in Fig. 1.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
19
Fig. 1. Suicidal behavior in bipolar veterans during mood stabilizer and/or antidepressant treatment (N = 405).
possible that confounding by phase of illness accounts for differences in event rates, with patients receiving antidepressants because they become more depressed or suicidal. The current dataset does not allow for controlling of this variable and further research with a prospective design may clarify the issue. However, even if those bipolar patients who developed suicidality on antidepressants (singly or in combination) received the antidepressant because of more severe depression or unstable illness, the fact remains that the addition of antidepressants did not diminish the rate of suicide events. These rates remained quite high in this group. The data in Fig. 1 (increased suicide events progressively across the three types of treatment exposure periods, from least with MS MONO, higher during MS + AD, through highest during AD ALONE) suggest that destabilization of bipolar patients during antidepressant treatment is only partially mitigated by concomitant mood stabilizer treatment. The data can be interpreted as meaning that even in the presence of mood stabilizers, antidepressant treated bipolar patients remain at higher risk for suicidal behavior. This study is limited by the fact that not all potential suicide risk factors were controlled for. Suicidal behavior is complex and many psychosocial factors and comorbidities are often involved (Rihmer, 2007). Meaningful control of these variables was not possible because of the retrospective design of the study and the fact that patients contributed months of observation in multiple treatment categories. We did not control for alcoholism in
this group, but we note that recent prospective studies of bipolar patients (Valtonen et al., 2006; Marangell et al., 2006) have shown that potential risk factors including alcoholism and co-morbidities per se were not significantly predictive of suicidal outcomes, yet previous suicide attempt and days depressed were. The latter acquire great significance in light of long-term prospective studies showing the domination of the course of bipolar disorder with greater percentage of days spent in depression than previously thought (Judd et al., 2002, 2003). We also did not control for the contributions of numerous other risk factors for suicide, including previous attempts, psychosocial factors, medical/substance use co-morbidity, and concomitant use of benzodiazepines and medications prescribed for general medical conditions. Because there was only one case of completed suicide, the results of this study are only applicable to non-lethal suicidal behavior. However, non-lethal suicidal behaviors in bipolar patients are clinically more serious than in the general population (Tondo and Baldessarini, 2000), and serious attempts are more like than unlike completed suicides (Maser et al., 2002). 5.3. Implications for clinical practice Our finding of a significant association of suicidal behavior and antidepressants in a bipolar cohort consistent with the hypotheses that the emergence of suicidality during treatment of major depression could be taken as a
20
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21
marker of future bipolar course if not bona fide bipolar spectrum (Akiskal and Mallya, 1987; Akiskal and Benazzi, 2003). Furthermore, it has been suggested that (Dilsaver and Akiskal, 2005; Berk and Dodd, 2005) the excess of suicide attempts in younger patients given antidepressants in controlled trials may be attributable to unrecognized bipolar disorder in this population—given the younger age of onset of bipolar disorder, coupled with the frequent onset of bipolar illness with a depressive episode. Indeed, it has long been suggested that prepubertal depression is the precursor of bipolar transformation, especially in the presence of bipolar family history (Akiskal, 1995). In light of our results and the foregoing considerations, it is of great public health concern that the majority of bipolar patients continue to be prescribed antidepressant monotherapy (Baldessarini et al., 2007). Prospective randomized studies with larger populations are needed to address the potential cause–effect relationship between antidepressants and suicidality in bipolar patients but unlikely to be undertaken because of ethical constraints. In the meantime the results of the study suggest that antidepressants should be used cautiously in bipolar as well as potentially bipolar depressives. Depressive mixed states, unrecognized in the official DSM-IVTR classification, appear to be a particularly important condition to be vigilant about in clinical practice in adults (Akiskal and Mallya, 1987; Koukopoulos et al., 1992), in the community (Balazs et al., 2006), and of course in juvenile subjects (Dilsaver and Akiskal, 2005). Such depressive mixed states (activated depression) might also be particularly relevant in the context of PTSD and alcohol and substance abuse, two common conditions among veteran populations seen in mental health clinics. Our data of increased suicidality during antidepressant treatment of bipolar patients should be interpreted within a multifactorial model of suicidality in veterans as well as the general population. Acknowledgement The study was partially funded by Abbott Laboratories. References Agresti, A., 1990. Categorical Data Analysis. John Wiley and Sons, New York, NY, pp. 189–196. Aizenberg, D., Olmer, A., Barak, Y., 2006. Suicide attempts amongst elderly bipolar patients. J. Affect. Disord. 91, 91–94. Akiskal, H.S., Mallya, G., 1987. Criteria for the “soft” bipolar spectrum: treatment implications. Psychopharmacol. Bull. 23, 68–73. Akiskal, H.S., 1995. Developmental pathways to bipolarity: are juvenileonset depressions pre-bipolar? J. Am. Acad. Child Adolesc. Psych. 34, 754–763.
Akiskal, H.S., 2006. What is the optimum strategy for the clinical management of bipolar depression? J. Bipolar Disord. V 18. Akiskal, H.S., Benazzi, F., 2003. Family history validation of the bipolar nature of depressive mixed states. J. Affect. Disord. 73, 113–122. Akiskal, H.S., Benazzi, F., 2005. Psychopathologic correlates of suicidal ideation in major depressive outpatients: is it all due to unrecognized (bipolar) depressive mixed states? Psychopathology 38, 273–280. Akiskal, H.S., Benazzi, F., Perugi, G., Rihmer, Z., 2005. Agitated “unipolar” depression re-conceptualized as a depressive mixed state: implications for the antidepressant-suicide controversy. J. Affect. Disord. 85, 245–258. Baldessarini, R.J., Pompili, M., Tondo, L., 2006. Suicidal risk in antidepressant drug trials. Arch. Gen. Psychiatry 63, 246–248. Baldessarini, R.J., Leahy, L., Arcona, S., Gause, D., Zhang, W., Hennen, J., 2007. Patterns of psychotropic drug prescription for U.S. patients with diagnoses of bipolar disorders. Psychiatr. Serv. 58, 85–91. Balazs, J., Benazzi, F., Rihmer, Z., Rihmer, A., Akiskal, K.K., Akiskal, H.S., 2006. The close link between suicide attempts and mixed (bipolar) depression: implications for suicide prevention. J. Affect. Disord. 91, 133–138. Bauer, M.S., Wisniewski, S.R., Marangell, L.B., Chessick, C.A., Allen, M.H., Dennehy, E.B., Miklowitz, D.J., Thase, M.E., Sachs, G.S., 2006. Are antidepressants associated with new-onset suicidality in bipolar disorder? A prospective study of participants in the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD). J. Clin. Psychiatry 67, 48–55. Berk, M., Dodd, S., 2005. Are treatment emergent suicidality and decreased response to antidepressants in younger patients due to bipolar disorder being misdiagnosed as unipolar depression? Med. Hypotheses 65, 39–43. Dilsaver, S.C., Akiskal, H.S., 2005. High rate of unrecognized bipolar mixed states among destitute Hispanic adolescents referred for “major depressive disorder”. J. Affect. Disord. 84, 179–186. Fergusson, D., Doucette, S., Glass, K.C., Shapiro, S., Healy, D., Hebert, P., Hutton, B., 2005. Association between suicide attempts and selective serotonin reuptake inhibitors: systematic review of randomised controlled trials. BMJ 330 (7488), 396. Food and Drug Administration (9-14-04), www.FDA.gov/bbs/topics/ news/2004/new01116.html. Ghaemi, S.N., Hsu, D.J., Soldani, F., Goodwin, F.K., 2003. Antidepressants in bipolar disorder: the case for caution. Bipolar Disord. 5, 421–433. Goldberg, J.F., Allen, M.H., Miklowitz, D.A., Bowden, C.L., Endick, C.J., Chessick, C.A., Wisniewski, S.R., Miyahara, S., Sagduyu, K., Thase, M.E., Calabrese, J.R., Sachs, G.S., 2005. Suicidal ideation and pharmacotherapy among STEP-BD patients. Psychiatr. Serv. 56, 1534–1540. Judd, L.L., Akiskal, H.S., Schettler, P.J., Endicott, J., Maser, J., Solomon, D.A., Leon, A.C., Rice, J.A., Keller, M.B., 2002. The long-term natural history of the weekly symptomatic status of bipolar-I disorder. Arch. Gen. Psychiatry 59, 530–537. Judd, L.L., Akiskal, H.S., Schettler, P., Endicott, J., Maser, J., Solomon, D., Leon, A., Coryell, W., Keller, M., 2003. A prospective investigation of the natural history of the long-term weekly symptomatic status of bipolar II disorder. Arch. Gen. Psychiatry 60, 261–269. Khan, A., Khan, S., Kolts, R., Brown, W.A., 2003. Suicide rates in clinical trials of SSRIs, other antidepressants, and placebo: analysis of FDA reports. Am. J. Psychiatry 160, 790–792. Koukopoulos, A., Faedda, G., Proietti, R., D'Amico, S., de Pisa, E., Simonetto, C., 1992. Mixed depressive syndromes. Encéphale 18 (Spec No 1), 19–21.
B.I. Yerevanian et al. / Journal of Affective Disorders 103 (2007) 13–21 Koukopoulos, A., Sani, G., Koukopoulos, A.E., Manfredi, G., Pacchiarotti, I., Girardi, P., 2007. Melancholia agitata and mixed depression. Acta. Psychiatr. Scand. Suppl. 433, 50–57. Marangell, L.B., Bauer, M.S., Dennehy, E.B., Wisniewsky, S.R., Allen, M.H., Miklowitz, D.J., Oquendo, M.A., Frank, E., Perlis, R.H., Martinez, J.M., Fagiolini, A., Otto, M.W., Chessick, C.A., Zboyan, H.A., Miyahara, S., Sachs, G., Thase, M.E., 2006. Prospective predictors of suicide and suicide attempts in 1556 patients with bipolar disorders followed for up to 2 years. Bipolar Disord. 8, 566–575. Maser, J.D., Akiskal, H.S., Zeller, P., Scheftner, W., Mueller, T., Endicott, J., Solomon, D., Clayton, P., 2002. Can temperament identify affectively ill patients who engage in lethal or near-lethal suicidal behavior? A 14-year prospective study. Suicide. LifeThreat. Behav. 32, 10–32. McElroy, S.L., Kotwal, R., Kaneria, R., Keck Jr, P.E., 2006. Antidepressants and suicidal behavior in bipolar disorder. Bipolar Disord. 8, 596–617. Moller, H.-J., 2006a. Evidence for beneficial effects of antidepressants on suicidality in depressive patients. A systematic review. Eur. Arch. Psychiatry Clin. Neurosci. 256, 229–343. Moller, H.-J., Grunze, H., Broich, K., 2006b. Do recent efficacy data on the drug treatment of acute bipolar depression support the position that drugs other than antidepressants are the treatment of choice? A conceptual review. Eur. Arch. Psychiatry Clin. Neurosci. 256, 1–16. Pompili, M., Tondo, L., Baldessarini, B., 2005. Suicidal risk emerging during antidepressant treatment: recognition and intervention. Clin. Neuropsychiatry 2, 66–72. Post, R.M., Leverich, G.S., Nolen, W.A., Kupka, R.W., Altshuler, L.L., Frye, M.A., Suppes, T., McElroy, S., Keck, P., Grunze, H., Walden, J.,
21
2003. A re-evaluation of the role of antidepressants in the treatment of bipolar depression: data from the Stanley Foundation Bipolar Network. Bipolar Disord. 5, 396–406. Rihmer, Z., 2002. Suicide risk in mood disorders. Curr. Opin. Psychiatry Jan. 20 (1), 17–22. Rihmer, Z., Akiskal, H.S., 2006. Do antidepressants t(h)reat(en) depressives? Toward a clinically judicious formulation of the antidepressant-suicidality FDA advisory in light of declining national suicide statistics from many countries. J. Affect. Disord. 94, 3–13. Rihmer, Z., 2007. Suicidal risk in mood disorders. Curr. Opin. Psychiatry 20, 17–22. Shi, L., Thiebaud, P., McCombs, J.S., 2004. The impact of unrecognized bipolar disorders for patients treated for depression with antidepressants in the fee-for-service California Medicaid (Medi-Cal) programme. J. Affect. Disord. 82, 373–383. Simon, N., Otto, M., Weiss, R., Bauer, M.S., Miyahara, S., Wisniewski, S.R., Thase, M.E., Kogan, J., Frank, E., Nierenberg, A.A., Calabrese, J.R., Sachs, G.S., Pollack, M.H., 2004. Pharmacotherapy for bipolar disorder and co-morbid conditions: baseline data from STEP-BD. J. Clin. Psychopharmacol. 24, 512–520. Tondo, L., Baldessarini, R.J., 2000. Reduced suicide risk during lithium maintenance treatment. J. Clin. Psychiatry, 61 (suppl 9), 97–104. Valtonen, H.M., Suominen, K., Mantere, O., Leppamaki, S., Arvilommi, P., Isometsa, E.T., 2006. Prospective study of risk factors for attempted suicide among patients with bipolar disorder. Bipolar Disord. 8, 576–585.