Suicide and Suicide Attempts in the Systematic Treatment Enhancement Program for Bipolar Disorder

Abstract

Background

The current report describes individuals with bipolar disorder who attempted or completed suicide while participating in the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) study.

Methods

Baseline and course features of individuals with suicide events are described.

Results

Among the 4360 people with bipolar disorder enrolled, 182 individuals made 270 prospectively observed suicidal acts, including 8 completed suicides. This represents a suicide rate of .014 per 100 person years in STEP-BD, which included frequent clinical visits, evidence based care, and standardized assessment at each patient contact. Approximately 1/3 of those who attempted suicide had more than one attempt during study participation. Those who completed suicide tended to do so early in study participation, and half of them did so on their first attempt.

Limitations

While this study is limited to description of individuals and precipitants of completed suicides and attempts in STEP-BD, further analyses are planned to explore risk factors and potential interventions for prevention of suicidal acts in persons with bipolar disorder.

Conclusions

Persons with bipolar disorder are at high risk for suicide. Overall rates of suicide events in STEP-BD were lower than expected, suggesting that the combination of frequent clinical visits (i.e., access to care), standardized assessment, and evidence-based treatment were helpful in this population.

Introduction

Persons with bipolar disorder (“BD”) are at increased risk for suicide. Angst and colleagues (2002) reported a 7% suicide case rate for 106 hospitalized bipolar patients followed for 34-38 years. A meta-analysis found the risk of suicide in BD to be at least 15 times higher than that of the general population (Harris and Barraclough; 1997). It is likely that the proportion of all suicides attributed to BD has been underestimated due to inclusion of patients with bipolar depression in the major depression estimates (Rihmer & Kiss, 2002), and that BD may account for one quarter of all completed suicides.

STEP-BD is a large, NIMH-funded trial that enrolled a total of 4360 individuals with BD from 22 US sites (Sachs et al., 2003; Sachs, 2004). In a cohort of 1500 participants from this study, observed for over 2 years, only previous history of suicide attempt (OR=4.52, p<.0001) and percent of days spent depressed in the past year (OR=1.16, p=.036) were significantly associated with prospectively observed suicide attempts or completions (Marangell et al., 2006).

The current report describes the characteristics of those participants in STEP-BD with documented suicide attempts or completed suicides during study participation, and describes temporal relationships between treatment and course variables and suicidality.

Methods

Study Overview

STEP-BD combines a prospective study utilizing a common disease-management model and a series of randomized, controlled trials. Detailed information on the methodology and inclusion criteria can be found elsewhere (Sachs et al., 2003; Sachs, 2004). After approval by the respective Human Subjects Panel at each site, patients provided both oral and written consent, in accordance with the Declaration of Helsinki.

Participants

STEP-BD enrolled a total of 4360 participants who were at least 15 years of age and diagnosed with bipolar I, bipolar I, cyclothymia, bipolar NOS, or schizoaffective disorder, bipolar subtype, according to DSM-IV criteria (APA, 1994).

Assessments and Procedures

The Mini International Neuropsychiatric Interview (MINI, Version 5.0; Sheehan et al., 1998) was administered at study entry. Clinical course characteristics were obtained from the Affective Disorders Evaluation (ADE; Sachs, 2004). Use of services is recorded on the Care Utilization (“CU”) Form, a semi-structured interview administered quarterly over the first year, and biannually after that point. Serious Adverse Event (SAE) forms were completed at all visits. Suicide attempts were defined as intentional, self-injurious behaviors with some potential for serious harm and/or lethality. An independent safety officer and SAE committee reviewed all potential suicide events to ensure accurate classification.

Prescription information is entered on the Clinical Monitoring Form (“CMF”, Sachs, Guille & McMurrich, 2002). The CMF also included symptom ratings, assignment of clinical status, and a rating of suicidal ideation.

Intervention

Since STEP-BD was designed as an effectiveness study, subjects received pharmacological interventions as clinically indicated. Study clinicians completed training in the principles of evidence based treatment, and pharmacotherapy guidelines based on published treatment guidelines were supplied in an annually-updated Clinicians Handbook (STEP-BD Clinicians’ Handbook, unpublished).

Statistical Analyses

The current paper is designed to describe suicide events in individuals with BD participating in a longitudinal treatment study, including the 8 completed suicides that were observed in STEP-BD. Given discrepancy in cell size, we present means and standard deviations or percentages of individuals experiencing each attribute.

Results

Over the course of STEP-BD, a total of 182 participants experienced 270 suicide events. This includes 8 completed suicides, and 262 suicide attempts. The 8 completed suicides per 4000 individuals in the STEP-BD observed for an average of 15.74 months represent a suicide rate of .014 per 100 person years.

Within the eight people who completed suicide, one had a previous suicide attempt documented during STEP-BD participation. Among the remainder, 119 individuals had one documented suicide attempt, and 55 made more than one suicide attempt (range from 2-5 events). Table 1 describes the demographic and clinical course characteristics of individuals across three groups: those with no prospectively observed suicide event, those with a prospectively observed attempt(s), and those with a prospectively observed completed suicide. Members in all three groups reported a similar age of onset of bipolar symptoms in their teens. While the sample sizes in our completed group are too small to draw any statistical conclusions, our descriptive data are consistent with the published literature in that females were more likely to attempt suicide while males were more likely to complete. Those who completed suicide were more likely to live alone and have no history of a committed relationship than those in the “attempt” and “no event” groups. While the ratio of bipolar diagnoses was similar across those with no event or attempts, and mirrored enrollment in STEP-BD, those who completed suicide were equally likely to be diagnosed with bipolar I and II (n=4 each).

Table 1.

Baseline Descriptive and Course Features of Individuals with no Prospective Suicide Event, and those with Prospective Attempted and Completed Suicides.

	No Prospective Suicide Event (n=4178)	Prospective Suicide Attempt (n=174)	Completed Suicide (n=8)
Age (x, SD)	39.77 (12.94)	36.11 (11.47)	35.63 (13.65)
Female (n, %)	2359 (56.46%)	118 (67.82%)	2 (25.0%)
White or Caucasian (n, %)	3724 (89.13%)	158 (90.80%)	7 (87.50%)
Marital Status (n, %)
Married or Living as Married	1361 (35.35%)	61 (37.66%)	0 (0%)
Separated, Divorced, or Widowed	1014 (26.34%)	35 (21.60%)	2 (25.0%)
Never Married or Lived as Married	1417 (36.81%)	64 (39.51%)	6 (75.0%)
Diagnosis (n, %)
Bipolar I	2669 (63.88%)	128 (73.56%)	4 (50.0%)
Bipolar II	1131 (27.07%)	35 (20.11%)	4 (50.0%)
Other	378 (9.05%)	11 (6.32%)	0 (0.0%)
Current or Lifetime Alcohol Use Disorder (n, %)	1626 (43.27%)	88 (53.99%)	5 (62.5%)
Current or Lifetime Substance Use Disorder (n, %)	1119 (29.78%)	63 (38.65%)	6 (75.0%)
Current or Lifetime Anxiety Disorder (n, %)	1894 (50.43%)	114 (69.94%)	3 (37.5%)
Current or Lifetime History of Panic Disorder (n, %)	809 (21.56%)	55 (33.74%)	3 (37.5%)
History of Psychotic Disorder Diagnosis (n, %)	1506 (36.05%)	76 (43.68%)	4 (50.0%)
Age of Onset of Bipolar Disorder (x, SD)	17.21 (8.74)	14.76 (7.05)	16.63 (8.72)
Percentage with age of onset <18 years	2470 (61.08%)	130 (76.47%)	5 (62.50%)
History of Suicide Attempt (n, %)	1444 (34.56%)	121 (69.54%)	4 (50.0%)
Family History of Suicide (n, %)	268 (9.34%)	16 (11.51%)	0 (0%)
Mean Time in STEP before First Event (in months)	N/A	10.28 (9.92)	6.57 (3.70)

Those who completed suicide had a lower frequency of prior attempts than those who attempted. Half of those who completed suicide during STEP-BD did so on their first lifetime attempt. On average, the suicide events occurred relatively early in treatment, and all patients had been seen by a clinician within a month before the event. There was more current or lifetime history of substance use disorders in those who completed suicide.

Table 2 presents characteristics from the clinical visit that occurred prior to the suicide event. On average, these visits occurred within a month of the event. The numbers suggest that the majority of patients that attempted and completed suicide were not in full syndromal mood episodes at the visits immediately preceding the suicide event, but instead, were more likely to be experiencing subsyndromal symptoms, or were in some stage of recovery from an episode. Patients in the attempted vs. completed groups were taking similar numbers of psychotropic medications, and there was no suggestion that a change in prescribed medications at the most recent clinical visit precipitated suicide events.

Table 2.

Characteristics of the Visit Prior to the first^* Event, by Event Group

	Prospective Suicide Attempt (n=174)	Prospective Completed Suicide (n=8)
Time Between Assessment (CMF) and Event (in months)	0.76 (1.42)	0.70 (.57)
GAF in Past Week	58.67 (11.40)	62.0 (8.37)
GAF in Past Month	57.74 (12.41)	61.14 (8.88)
Mean estimate of % of Days with Abnormal Anxiety in last 10 days	44.66 (40.73)	54.29 (46.14)
Clinical Status at last CMF
Depressed	52 (32.30%)	2 (28.57%)
Hypomanic/Manic/Mixed	11 (6.83%)	0
Subsyndromal Symptoms	35 (21.74%)	2 (28.57%)
Recovering/Recovered	63 (39.13)	3 (42.86%)
Additional psych treatment noted on CMF (includes ER, hospital, outpatient)	33 (20.50%)	1 (14.29%)
Presence of Delusions	4 (2.48%)	0 (0%)
Presence of Hallucinations	12 (7.45%)	0 (0%)
Mean Number of Medications Prescribed at last CMF (mean, SD)	4.17 (2.04)	4.14 (1.57)
Lithium discontinuation within 30 days prior to event	4 (2.30%)	0
SSRI started within 30 days prior to event**	8 (4.60%)	0
Non-SSRI Antidepressant started within 30 days prior to the event***	8 (4.60%)	0

^*for those individuals with multiple attempts, only the characteristics of the visit prior to the first attempt are described here.

^**SSRI Antidepressants = citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline,

^***Non-SSRI antidepressants = bupropion, mirtazapine, nefazadone, venlafaxine, duloxetine

There was no strong association of event to calendar month. The greatest number of events occurred in January-March (n=86, 32%) with the rest fairly evenly distributed across the remaining three quarters (n=55, 66, and 63). Among completed suicides, confirmed method of death was drug overdose (n=2), hanging (n=2), and gunshot (n=2); definitive cause of death was not available for two participants. For those suicide attempts with detail on the exact events, overdose was the dominant method of attempt.

Of those who completed suicide, six were male, ages ranged from 18-57 years, all were single, and five were unemployed. Most were in the early days of treatment in STEP-BD. One individual completed an enrollment visit and never returned, and 4 others had 4 or fewer clinical visits. Only two reported any suicidal ideation at the visit immediately prior to the event.

Discussion

This report presents descriptive detail on individuals who made a prospectively observed suicide attempt or completed suicide while participating in STEP-BD. It is noteworthy that there were only 8 completed suicides among the 4000 individuals with BD who enrolled in the STEP-BD and were followed an average of 16 months (15.74 months). This represents a suicide rate of .014 per 100 person years under care. This is compared to a weighted average rate of .48 per 100 patient years in 21,783 patients reported in six studies of bipolar patients followed for varying lengths but often averaging 10 years (Osby et al., 2001; Brown et al., 2000; Carlson et al., 1974; Angst & Preisig, 1995; Hoyer et al, 2000; Tondo et al., 1998).

Generally, shorter observation periods have been associated with higher rather than lower suicide rates. However, STEP-BD differed from a naturalistic follow-up study in that it emphasized use of evidence-based psychosocial and pharmacological interventions (Dennehy et al., 2007; Miklowitz & Otto, 2006), and provided stable access to quality care. This study suggests a number of challenges in the prevention of suicide in clinical populations. Clinicians are well aware of the finding that history of suicide incurs additional risk (Hawton et al., 2005), and a previous report from STEP-BD demonstrated that prior history of attempt was predictive of suicide events (attempts and completions) in patients with bipolar disorder (Marangell et al., 2006). When the eight completed suicides were isolated, however, 4 (50%) did not have a previous lifetime attempt, consistent with findings from other authors (Roy, 1982; Conwell et al., 1998; Isometsa & Lonnqvist, 1998). While a previous history of attempt was present in only half the completed suicides in STEP-BD, when present, it significantly increases risk (Marangell et al., 2006). Given the lack of sensitivity of previous attempts in predicting suicide in this small number of cases, however, it is important to pursue identification and awareness of other potential predictors.

Of those who completed suicide, most were not severely ill at their last contact, and 75% reported no suicidal ideation at the visit prior to the suicide. They tended to be in the early stages of study participation, with five having four or fewer visits, highlighting the importance of discussing suicidal thoughts, risk factors, and establishing contracts and safety plans early in treatment (Osby et al., 2001). Having comorbid anxiety disorder, or anxiety symptoms, was typical for those who experienced any type of suicide event (MacKinnon et al, 2005, Lee & Dunner, 2008; Kauer-Sant’Anna et al, 2007; Rihmer, 2007; Quarantini et al., 2010; Simon et al., 2004).

Given concerns about the association between antidepressant prescription and suicide, we assessed recent changes in antidepressant prescriptions. Similar to previous findings from STEP-BD (Bauer et al., 2005), very few individuals had a new prescription for an SSRI medication (n=8, 5%) or another category of antidepressant (n=8, 5%) in the 30 days prior to the suicide event. None of those in the completed suicide group had received a new prescription for antidepressant medication.

We looked at the association of suicide event to calendar months. Given the wide geographic variability of STEP sites, and small number of cases, we used a crude division of event by month to assess for trends. While there was no signal for any association between season and event, further work looking at seasonal variation may be warranted (Rocchi et al., 2007; Voracek et al., 2007).

This report is limited by our reliance on prospectively observed and verified suicide events only, without regard for patient report of prior history. Importantly, 35% of participants with no prospectively observed suicide event while participating in STEP-BD reported a prior history of attempt, and may be somehow different from those with no history or prospectively observed suicide event. We will consider this in future analyses. In this report, we chose to focus only on prospectively observed suicide events, as prior history of suicide was collected via patient self-report only, without corroboration of the potential lethality of the act. In contrast, prospectively observed events were entered into SAE forms and monitored by a study clinician. All those that were potential suicide events were reviewed by a committee, and in many cases, queried extensively, before being coded as an “attempt”. Completed suicides required independent confirmation of the method of death, either through the presence of a clear suicide note, or examination by a coroner or other professional. We are confident in the veracity and significance of the events described in the current report.

This study suggests a number of additional challenges in the prevention of suicide in clinical populations. While clinicians rely on the predictive value of past suicide events predicting future risk, only 50% of completed suicides occurred in individuals with prior attempts. Clearly, newly identified or acquired patients also require special consideration. Intensive contact in the early stages of the relationship may help establish the patient-provider relationship and a collaborative approach to risk assessment and prevention. Recent data also point to the importance of including family or significant others in crisis planning (Chessick et al., 2009).

We plan additional multivariate analyses to further explore risk factors and predictors of suicide attempts in this population.