Journal of Personality Disorders, Volume 32, 1-19, 2018

© 2018 The Guilford Press

SOLOFF AND CHIAPPETTA
SUICIDAL BEHAVIOR IN BPD

10-YEAR OUTCOME OF SUICIDAL BEHAVIOR IN

BORDERLINE PERSONALITY DISORDER
Paul H. Soloff, MD, and Laurel Chiappetta, MS

Prospective predictors of suicide attempts were assessed in 118 subjects

with borderline personality disorder (BPD) after 10 or more years of
follow-up. Mean (SD) time to follow-up was 14.4 (4.7) years. Subjects were
predominately female (78.8%), Caucasian (81.4%), and of lower socioeco-
nomic status. Initial recruitment was evenly balanced between inpatient,
outpatient, and non-patient (community) sources. In the 10-year interval,
55 subjects (46.6%) attempted suicide. Compared to baseline, suicidal
ideation, number of attempts, and non-suicidal self-injury diminished mark-
edly. Core symptoms of BPD, substance abuse, and alcohol use disorders
decreased significantly; however, major depressive disorder (MDD) re-
mained constant at 50%. Forty-four percent of subjects had poor psychoso-
cial, vocational, and economic outcomes. Psychosocial outcome was inde-
pendent of suicide history and any treatment. Increased risk was associated
with interval hospitalization prior to any attempt (illness severity), as well
as poor social, vocational, and psychosocial functioning at baseline.

Personality disorders (PD), especially Cluster B disorders, increase the risk of

suicide when comorbid with high risk disorders such as depression, substance
use, and alcohol dependence (Allebeck, Allgulander, & Fisher, 1988). They
are frequently associated with suicide in concert with socio-demographic risk
factors such as living alone, low educational level, and adverse life events,
even after adjustment for affective and substance use disorders (Schneider
et al., 2006, 2008). Borderline personality disorder (BPD) is the only Diag-
nostic and Statistical Manual of Mental Disorders, fifth editoon (DSM-V;
American Psychiatric Association [APA], 2013), category defined, in part,
by recurrent suicidal behavior. Among PD subjects followed prospectively
over a 10-year period, BPD is uniquely associated with the risk of “ever at-
tempting suicide” compared to all others (Ansell et al., 2015). Among BPD
patients, suicide attempts are reported in 46%–92% at first intake, with a
completion rate of 3% to 10% in longitudinal studies (Black, Blum, Pfohl,
& Hale, 2004). Among BPD inpatients and outpatients recruited for pro-

From the Department of Psychiatry, University of Pittsburgh School of Medicine (P. H. S.); and the Statis-
tics Department (L. C.), University of Pittsburgh, Pittsburgh, Pennsylvania.
Supported by a grant from the National Institute of Mental Health to Dr. Soloff (RO1 MH 048463).
This article is based on a paper presented at the annual meeting of North American Society for the Study
of Personality Disorders, New York City, April 8, 2017.
Address correspondence to Paul H. Soloff, MD, 19200 Space Center Blvd., Apt. 1932, Houston, TX
77058. E-mail: soloffph@upmc.edu

1
2 SOLOFF AND CHIAPPETTA

spective longitudinal studies, 20%–25% attempt suicide within the first 2

years of follow-up (Soloff & Fabio, 2008; Yen et al., 2003). Most of our
knowledge of suicidal behavior in BPD is derived from retrospective studies
of suicide completion, or cross-diagnostic studies of attempt behavior (Sol-
off, 2005, for review). In contrast, a prospective, longitudinal design has the
advantage of using standardized multidimensional baseline assessments, and
systematic follow-ups focused on risk factors specifically relevant to suicidal
behavior. We are conducting a prospective longitudinal study of suicidal be-
havior in BPD, seeking predictors of suicide attempts among demographic,
diagnostic, clinical, and socioeconomic risk factors associated with suicidal
behavior in prior retrospective and cross-diagnostic studies. In previous re-
ports, we identified prospective predictors of attempt behavior at 1-, 2-, 4-,
6-, and 8-year intervals (Soloff & Chiappetta, 2012a, 2017; Soloff & Fabio,
2008). An important finding of this longitudinal research is that risk factors
associated with suicidal behavior in BPD change over time.
Time-varying risk factors present specific challenges to the clinician at
each interval. An early finding of this study was that an acute Axis I disor-
der, major depressive disorder (MDD), was predictive of interval attempts
only in the first 12 months of follow-up, despite prevalence in half of sub-
jects through the 8-year assessment. Instead, global measures of psychosocial
functioning, social and vocational achievement, and illness severity assumed
prominence as consistent predictors through the 8-year interval. Frequent
changes in employment and minority status were also significant predictors
of interval suicide attempts in this time frame (Soloff & Chiappetta, 2017).
Minority status represented multiple measures of poor socioeconomic func-
tion including: lower socioeconomic status, less employment, lower house-
hold income, and greater reliance on government support compared to Cau-
casian subjects. Non-specific measures of illness severity (such as inpatient
recruitment and/or hospitalization in the interval prior to any attempt) were
also consistent predictors. Contributions to illness severity may include in-
teractive effects of diagnostic comorbidities and specific personality dimen-
sions of BPD, such as mood instability and impulsive aggression, which were
individually predictive of attempt behavior at the 8-year follow-up (Soloff &
Chiappetta, 2017). However, overall measures of BPD syndrome severity or
lifetime number of BPD criteria did not contribute to prediction of attempt
behavior (Soloff & Chiappetta, 2017).
The long-term prognosis for BPD is for symptomatic and diagnostic re-
mission in 85%–93% of subjects over a 10-year period (Gunderson et al.,
2011; Zanarini, Frankenburg, Reich, & Fitzmaurice, 2010a). The landmark
McLean Study of Adult Development (MSAD) found that prevalence of in-
terval suicide attempts among BPD subjects recruited as inpatients dimin-
ished from 20.1% at the 4-year follow-up to 8.2% at the 16-year assess-
ment (Wedig et al., 2012). Despite symptomatic and diagnostic remission for
the vast majority of subjects, only 40% sustained stable functional recovery
(Zanarini, Frankenburg, Reich, & Fitzmaurice, 2012). At 8-year follow-up
in the current study, we found no direct relationship between poor psycho-
social outcome (defined by a Global Assessment Scale score (GAS) < 61) and
suicidal behavior in the interval—that is, suicidal behavior in the interval did
SUICIDAL BEHAVIOR IN BPD 3

not determine a poor psychosocial outcome. However, socioeconomic fac-

tors such as employment, income, and minority race were significant predic-
tors of suicidal behavior in the interval.
In previous work, we proposed subtyping BPD by suicidal behavior.
Using trajectory analysis, high risk subjects were identified by a pattern of
increasing medical lethality of recurrent attempts over time (Soloff & Chiap-
petta, 2012b). A high lethality BPD subtype was characterized by inpatient
recruitment (a measure of illness severity), poor psychosocial functioning at
baseline, poor work history, and poor relationships in the immediate family.
Measures of BPD psychopathology or syndrome severity were not signifi-
cant characteristics of the high lethality subtype. A low lethality subtype was
characterized by an attitude of negativism, histrionic and/or narcissistic PD
traits, and substance use disorders, a clinical presentation commonly associ-
ated with BPD. The current study assesses changes in prospective predictors
of attempt behavior at 10-year follow-up, addresses the relationship of psy-
chosocial and socioeconomic risk factors to suicide attempts, and seeks to
identify BPD subjects at highest risk.

METHOD
PARTICIPANTS

This study was approved by the University of Pittsburgh’s institutional review

board. Male and female subjects between the ages of 18 and 45 years were
recruited from inpatient, outpatient, and non-patient (community) sources.
Diagnostic interviews were conducted by master’s–prepared research clini-
cians using standardized semi-structured interviews. These included the
Structured Clinical Interview for DSM-IV-TR (SCID-I/P) for Axis I disorders
and the International Personality Disorders Examination (IPDE) for Axis II
(First, Spitzer, Gibbon, & Williams, 2005; Loranger, 1999). BPD subjects
were first required to meet diagnostic criteria for BPD (probable or definite)
on the IPDE, with a lifetime time frame. A separate interviewer (PHS) then
conducted the Diagnostic Interview for Borderline Patients (DIB), requiring
a score of 7 or more (definite) for inclusion (Gunderson, Kolb, & Austin,
1981). After 2001, the Revised Diagnostic Interview for Borderlines (DIB-r)
was adopted as the primary diagnostic interview for BPD, requiring a score
of 8 or more (definite) for inclusion (Zanarini, Gunderson, Frankenburg, &
Chauncey, 1989). For the sake of continuity, both forms were scored concur-
rently. Both the DIB and the DIB-r have a two-year time frame. All diagnoses
were confirmed in a consensus conference of raters, using a best estimate
process and all available data.
Exclusion criteria included: a lifetime (past or current) Axis I diagno-
sis of schizophrenia, delusional (paranoid) disorder, schizoaffective disorder,
and any bipolar disorder or psychotic depression; clinical evidence of central
nervous system (CNS) pathology of any etiology (including seizure disor-
der, acquired brain injury, or developmental deficits); physical disorders or
treatments with known psychiatric consequence; or borderline intellectual
4 SOLOFF AND CHIAPPETTA

functioning. All subjects provided written informed consent after receiving a

complete description of the study.

MEASURES

“Psychobiology of Suicidal Behavior in BPD” originated as a component of

the Mental Health Clinical Research Center for the Study of Suicidal Behav-
ior (MHCRC; J. J. Mann, MD, PI). The core assessment battery of the MH-
CRC was adapted for the study of BPD and has been presented elsewhere
(Soloff, Fabio, Kelly, Malone, & Mann, 2005). Measures from this multidi-
mensional assessment include:

• The MHCRC demographic interview

• Diagnostic interviews (SCID-I/P, IPDE, and the DIB/DIB-r)
• Clinical state assessments: Beck Depression Inventory (BDI-II; Beck,
Ward, Mendelson, Mock, & Erbaugh, 1961), Hamilton Rating Scale for
Depression–24 item format (HamD-24; Guy, 1976), Beck Hopelessness
Scale (Beck & Steer, 1988), and Global Assessment Scale (GAS; Endicott,
Spitzer, Fleiss, & Cohen, 1976)
• Suicide scales: MHCRC Suicide History and Lethality Rating Scale
(LRS; Oquendo, Halberstam, & Mann, 2003), Suicide Intent Scale (SIS)
and Scale for Suicidal Ideation (SSI; Beck, Beck, & Kovacs, 1975; Beck,
Schuyker, & Herman, 1974), and Reasons for Living Inventory (RFL;
Linehan, Goodstein, Nielson, & Chiles, 1983)
• Personality traits: Barratt Impulsiveness Scale–version-11(BIS; Bar-
ratt, 1965; Barratt & Stanford, 1995), Buss-Durkee Hostility Inventory
(BDHI; Buss & Durkee, 1957), Life History of Aggression, adult score
(LHA; Brown & Goodwin, 1986), and MMPI–Psychopathic Deviate sub-
scale (MMPI–Pd; Hathaway & Meehl, 1951)
• MHCRC family history and abuse history (Soloff, Lynch, & Kelly, 2002)
• Social Adjustment Scale–self-report (SAS-sr; Weissman & Bothwell, 1976)
• MHCRC treatment history

High lethality suicide attempts were defined by a lethality rating scale score >
4. The MHCRC follow-up interview defined interval changes in demograph-
ic, diagnostic, clinical, psychosocial, and treatment histories. The follow-
up Axis I diagnostic interview utilized Diagnostic and Statistical Manual of
Mental Disorders, fourth edition (DSM-IV-TR; APA, 2000) criteria. Follow-
up interviews were biannual, during which all clinical state measures, sui-
cide, personality trait, and social adjustment assessments were repeated.
In assessing economic status, the federal definition of poverty was used:
for a single person household, poverty level income was $10,000; for a three-
person household, $20,000 (www.hhs.gov). For subjects at remote locations
or unable to travel, follow-up interviews were conducted by telephone and
self-ratings by mail.
SUICIDAL BEHAVIOR IN BPD 5

STATISTICAL ANALYSES

Baseline variables were compared to 10-year outcomes using t tests and

McNemar’s test (McNemar, 1947). If data were not available on the 10-
year anniversary date, the nearest values beyond 10 years were used. Interval
attempters were compared to non-attempters at 10-year follow-up using t
tests and chi square tests as appropriate. Significance for differences between
groups was determined using Bonferroni correction for multiple compari-
sons, set at p < .001 (group differences of noteworthy trends are reported
for p < .01). Suicide attempts were defined by subjective intent to die, with
behavior in furtherance of suicide, and medical consequences. All attempts
in this analysis had LRS scores of 2 or greater, which require some physi-
cal consequences. Individual Cox proportional hazards models were used
to examine predictive associations between predefined risk factors assessed
at both baseline and follow-up, and suicide attempts in the interval (Table
1). Treatment variables (e.g., hospitalization, outpatient [OPD] treatment)
were counted only “prior to any interval attempt,” as medically significant
suicide attempts in our setting resulted in either inpatient admission or OPD
referral. “Poor” psychosocial outcome was defined by a GAS score less than
61; “good” psychosocial outcome by a GAS score of 61 or greater. The
GAS score at 10-year follow-up determined poor versus good psychosocial
outcomes. “Time-to-attempt” was computed using the difference between
the date of the baseline interview and the first suicide attempt in the 10-year
interval (Allgulander & Fisher, 1986). Estimates of risk (odds ratio) and as-
sociated 95% confidence intervals (CI) were obtained for each analysis. All
individually significant variables were entered into a final regression model.

RESULTS
SAMPLE CHARACTERISTICS (TABLE 1)

A total of 118 BPD subjects who had been followed for 10 or more years
were included in the analysis, drawn from a total study population of 358.
Mean (SD) time to follow-up was 14.4 (4.7) years, with a range of 10–25 yrs.
At follow-up, subjects were predominately female (78.8%), single (77.1%),
and without children (67.8%). One-third were living alone. Mean (SD) age
at follow-up was 44.2 (9.0) years. Participation by race mirrored the de-
mographics of our catchment area: 81.4% Caucasian and 18.6% minority,
largely African American. Subjects were drawn predominately from lower
socioeconomic classes; 65.8% were in Hollingshead Classes IV and V. They
had a mean (SD) education of 14.1 (2.2) years. A majority (68.4%) identified
with a specific religious denomination.
A prior history of psychiatric hospitalization was reported by 69.5%,
with first admission at mean (SD) age of 22 (6.9) years. A larger number
(85.7%) had received outpatient treatment, beginning at a mean (SD) age of
18.4 (6.9) years. Recruitment was well balanced between inpatient (32.2%),
outpatient (33.9%), and non-patient community sources (33.9%).
6 SOLOFF AND CHIAPPETTA
TABLE 1. Risk Factors From Baseline to 10-Year Follow-Up
Baseline Follow-Up (FU) Statistic, p valuea
A. Demographic
Age (at long-term FU) 29.54 ± 7.91 44.2 ± 9.0 t(1) = 34.13, < .001
Married (% no) 78.0 77.1 McN(1) = 0.80, 1.00
Children (% none) 71.2 67.8 McN(1) = 2.25, .134
B. Socioeconomic Changes
SES (% Hollingshead Cl. IV,V) 65.8 57.6 McN(1) = 1.42, .233
Lives alone (% yes) 25.4 33.9 McN(1) = 2.13, .143
Source of income (% on gov’t. support) 26.7 32.5 McN(1) = 0.83, .361
Household income (% < 20K) 59.8 35.5 McN(1) = 13.1, < .001
Patient personal income (% < 10K) 68.6 41.5 McN(1) = 13.6, < .001
Employed (% yes) 47.5 56.8 McN(1) = 2.13, .145
Social Adjustment Scale (SAS-SR) total 2.53 ± 0.55 (106) 2.34 ± 0.63 (62) t(54) = 2.48, .016
Work (102/61) 1.52 ± 1.05 2.14 ± 0.84 t(52) = 2.28, .027
Social (105/61) 2.83 ± 0.74 2.69 ± 0.88 t(52) = 1.91, .062
External Family (106/59) 2.35 ± 0.68 2.26 ± 0.81 t(51) = 0.69, .492
Marital (39/25) 2.51 ± 0.48 2.24 ± 0.66 t(13) = 0.63, .540
Parental (16/22) 1.88 ± 0.86 1.72 ± 0.63 t(4) = 0.53, .625
Family unit (67/44) 2.63 ± 0.95 2.48 ± 1.08 t(29) = 1.38, .179
C. Treatment History
Psychiatric hospitalizationb 69.5 21.2 z = 7.45, p < .001
Outpatient treatmentb 85.7 16.1 z = 10.69, p < .001
D. Diagnostic Variables
Major depressive disorder (MDD) 50.8 50.0 McN(1) = 0.00, 1.00
Substance use disorder (SUD) 32.2 11.4 McN(1) = 15.6, < .001
Alcohol abuse/dependence (ALC) 31.4 8.8 McN(1) = 18.6, < .001
Post-traumatic stress disorder (PTSD) 12.7 9.6 McN(1) = 0.45, .503
E. Clinical State Variables
Beck Depression Inventory (101/68) 26.16 ± 12.58 18.21 ± 12.69 t(50) = 3.03, .004
Hamilton Depression (HamD) (115/114) 20.94 ± 8.22 18.54 ± 11.93 t(110) = 1.49, .139
Beck Hopelessness Scale (115/53) 11.75 ± 6.01 8.72 ± 5.88 t(51) = 2.32, .024
Global Assessment Scale (GAS) 51.97 ± 12.23 61.42 ± 12.19 t(114) = 6.81, < .001
Poor psychosocial function (%GAS<61) 72.6 44.0 McN(1) = 19.5, < .001
F. Suicidal Behavior
Suicide attempt (%Yes) 83.1 46.6 McN(1) = 36.0, < .001
Suicide Intent Scale (SIS)
Total score, max lethality 15.16 ± 5.92 16.00 ± 6.16 t(16) = 0.74, .468
Total score, most recent attempt 14.34 ± 6.28 15.90 ± 5.27 t(16) = 0.77, .451
Lethal intent factor, max lethality 8.4 ± 3.5 8.9 ± 3.6 t(16) = 0.79, .439
Lethal intent factor, most recent 7.7 ± 3.6 8.7 ± 3.0 t(16) = 0.39, .700
Planning factor, max lethality 6.4 ± 3.2 7.0 ± 3.3 t(16) = 1.01, .327
Planning factor, most recent 6.3 ± 3.6 7.0 ± 2.9 t(16) = 0.27, .788
Scale for Suicidal Ideation (SSI)
Prior 2 weeks 12.20 ± 11.53 2.84 ± 5.58 t(95) = 7.23, < .001
Current 6.54 ± 8.76 1.92 ± 4.35 t(87) = 4.93, < .001
Number of attempts (baseline vs. interval) 3.16 ± 4.21 1.47 ± 3.02 t(117) = 4.29, < .001
SUICIDAL BEHAVIOR IN BPD 7

TABLE 1. (continued)
Lethality Rating Scale (LRS) (max score) 3.09 ± 1.74 3.08 ± 1.60 t(97) = 0.09, p = .93
High lethality attempter (LRS > 4) (% yes) 27.1 13.6 McN(1) = 7.03, .007
Number of high lethal attempts (LRS > 4) 0.47 ± 0.92 0.28 ± 0.83 t(117) = 1.95, .053
G. Personality Traits
Aggression, adult only (Brown-Goodwin LHA) 24.05 ± 6.46 14.85 ± 3.85 t(74) = 11.29, < .001
Barratt Impulsiveness Scale (BIS) 75.16 ± 4.51 74.78 ± 4.72 t(97) = 0.81, .418
Buss-Durkee Hostility Inventory (BDHI) 45.52 ± 11.41 33.77 ± 16.05 t(44) = 4.38, < .001
Assault 4.65 ± 2.77 3.71 ± 2.90 t(44) = 2.02, .05
Indirect hostility 5.64 ± 1.82 4.15 ± 2.11 t(44) = 3.74, .001
Irritability 7.85 ± 2.28 5.58 ± 3.30 t(44) = 3.81, < .001
Negativism 2.99 ± 1.46 2.13 ± 1.51 t(44) = 3.80, < .001
Resentment 5.32 ± 2.01 3.62 ± 2.55 t(44) = 3.25, .002
Suspicion 5.86 ± 2.39 3.94 ± 3.18 t(44) = 2.85, .007
Verbal hostility 7.79 ± 3.13 6.67 ± 3.33 t(44) = 1.23, .226
Guilt 5.41 ± 2.05 4.35 ± 2.19 t(44) = 2.34, .024
MMPI Psychopathic Deviate Subscale 59.23 ± 6.72 62.49 ± 7.26 t(77) = 4.28, < .001
Diagnostic Interview for Borderline Patients
(DIB) Section Score Total 27.65 ± 4.22 15.63 ± 6.90 t(51) = 11.35, < .001
Social adaptation 4.81 ± 1.61 3.87 ± 1.65 t(51) = 3.65, .001
Impulse action patterns 7.21 ± 2.03 3.04 ± 2.77 t(51) = 9.50, < .001
Affects 6.38 ± 1.30 3.98 ± 2.30 t(51) = 6.94, < .001
Psychosis 2.90 ± 2.28 1.50 ± 1.64 t(51) = 3.75, < .001
Interpersonal relations 6.35 ± 2.27 3.25 ± 2.65 t(51) = 7.50, < .001
Revised Diagnostic Interview for Borderlines
(DIB-R) Scaled Section Score 8.38 ± 1.26 3.97 ± 3.17 t(31) = 8.58, < .001
Affect 9.06 ± 1.32 6.53 ± 2.77 t(31) = 4.83, < .001
Cognition 2.97 ± 1.84 1.84 ± 2.16 t(31) = 2.34, .026
Impulse action patterns 7.22 ± 1.79 3.00 ± 2.64 t(31) = 7.76, < .001
Interpersonal relationships 9.03 ± 2.65 4.28 ± 3.54 t(31) = 7.37, < .001
a
McN = McNemar test; statistic is a chi-square test, when a value is not reported, binomial distribution (z) was used.
b
At baseline, psychiatric hospitalization and outpatient treatment are reported for “% yes ever prior to baseline.” For
FU, these are reported as “% yes prior to first attempt in interval, since baseline.”

The most prevalent Axis I comorbidities at intake included major de-

pressive disorder (MDD; 50.8%), substance use disorder (SUD; 32.2%), al-
cohol abuse or dependence (ALC; 31.4%), and post-traumatic stress disor-
der (PTSD; 12.7%). An additional Axis II disorder was reported by 73.6%
of subjects, a second Cluster B disorder in 45.3%, with 18.9% meeting full
criteria for antisocial personality disorder. The NEO Five Factor Inventory
indicated high scores (given in the following as mean [standard deviation])
in Neuroticism (32.3 [9.1]) and Openness (32.3 [7.4]), using the combined
gender profile for reference (Form S, adult). Scores for Extraversion (28.9
[10.4]), Agreeableness (33.1 [7.0]) and Conscientiousness (32.0 [9.8]) were
all in the average range.
8 SOLOFF AND CHIAPPETTA

By 10-year follow-up, the prevalence of comorbid SUD and ALC had

diminished significantly, while MDD remained essentially unchanged, and
PTSD minimally decreased. Scores on the observer-rated HamD-24 changed
little over time, remaining in the highly symptomatic range at follow-up.
Subjects endorsed less depressed mood on the self-rated Beck Depression
Inventory (BDI-II) (p = .004), and the Beck Hopelessness scales (p = .024),
though falling short of significance after Bonferroni correction. Aggression
(LHA) and hostility (BDHI total score) both diminished significantly over
time, though not trait impulsiveness (BIS). A small but significant increase
in antisocial traits (MMPI–Psychopathic Deviate scale) was noted over time
(Table 1). Severity of BPD criteria diminished significantly across all five sec-
tion and total scores of the DIB, and three (of four) scaled-section and total
scores of the DIB-r (only the DIB-r Cognition section score fell short of sig-
nificance at p = .026).
Many subjects came from highly dysfunctional families. A family history
of alcohol abuse was highly prevalent (49.2%), as was a family history of
substance abuse (33.9%). Childhood sexual abuse was reported by 47.6%
of subjects, physical abuse by 41.3%. Adult sexual abuse was reported by
18.0% and adult physical abuse by 21%. A family history of suicide was
reported by only two subjects (1.7%), though nine (7.6%) reported family
members with suicide attempts.
Treatment utilization decreased markedly over the 10-year interval. Psy-
chiatric hospitalization decreased from 69.5% for the time prior to baseline
to 21.2% during the 10-year interval (i.e., for attempters, this was for the
time prior to a first interval attempt; for non-attempters, the entire inter-
val). Similarly, utilization of outpatient treatment declined significantly from
baseline (85.7%) to 10-year follow-up (16.1%).

SUICIDAL OUTCOMES (TABLE 1)

At the time of intake, a past history of suicide attempt was reported by

83.1% of subjects, with a mean (SD) of 3.2 (4.2) attempts per attempter,
and a first attempt at mean (SD) age of 20.3 (8.0) years. A past history of
non-suicidal self-injury (NSSI) was reported by 48.7%. In the 10-year in-
terval, 55 subjects (46.6%) reported suicide attempts, a significant decrease
from the attempt rate at baseline. Similarly, NSSI diminished to 21.7% at
10-year follow-up. Suicidal ideation (SSI) decreased significantly over the
interval compared to baseline; however, attempt characteristics—that is, the
degree of subjective intent, objective planning, and medical lethality associ-
ated with interval attempts—was similar to that established at baseline for
past attempts. Fewer subjects reported high lethality attempts in the interval
(13.6%) compared to baseline (27.1%).
In the course of the longitudinal study, there were 16 deaths, four with
clear suicidal intent. Deaths by suicide involved hanging (1) and polydrug
overdose with prescribed psychotropic medications (3). The mean age of
suicides was 34.8 years, with last research follow-ups at 3 months, 1 year, 5
years, and 9 years, respectively. Eight deaths were associated with acute drug
toxicity (e.g., opiates, cocaine, alcohol) or chronic effects of alcohol abuse
SUICIDAL BEHAVIOR IN BPD 9

TABLE 2. Comparisons Between Interval Attempters and Non-Attempters at 10-Year Follow-Up

Non-Attempter Attempter
(n = 63) (n = 55) Statistic, p
A. Demographic Variables
Age (at most recent follow-up [FU]) 42.3 ± 9.1 46.4 ± 8.3 t = 2.4, df = 116, p = .012
Education (at baseline, no. years FT) 14.3 ± 2.4 13.9 ± 2.1 t = 0.87, df = 115, p = .39
Sex (% female) 74.6 83.6 χ2 = 1.4, df = 1, 0.23
Source of recruitment χ2 = 5.2, df = 2, 0.08
Inpatient 23.8 41.8
Outpatient 34.9 32.7
Non-patient 41.3 25.5
Race (% Caucasian) 85.7 76.4 χ2 = 1.7, df = 1, p = .19
Religion (% none) 31.7 31.5 χ2 = 0.001, df = 1, p = .98
Married (% no) 81.0 72.7 χ2 = 1.13, df = 1, p = .29
Children (% none) 68.3 67.3 χ2 = 0.01, df = 1, p = .91
B. Socioeconomic
Hollingshead SES (% low) 50.8 65.5 χ2 = 2.6, df = 1, p = .11
Lives alone (% yes) 25.4 43.6 χ2 = 4.4, df = 1, p = .04
Major source of income (% government) 27.4 38.5 χ2 = 1.6, df = 1, p = .21
Household income (% < 20K) 24.1 49.0 χ2 = 7.2, df = 1, p = .007
Patient personal income (% poverty – < 10K) 33.3 50.9 χ2 = 3.7, df = 1, p = .053
Change in employment (% yes) 73.0 72.7 FET = 0.87, p = 1.00
Employed (% no, from occupation) 33.3 54.4 χ2 = 5.4, df = 1, p = .02
*Hospitalization (% yes before att) 38.1 45.5 χ2 = 0.66, df = 1, p = .042
*Outpatient treatment (% yes before att) 87.3 34.5 χ2 = 34.95, df = 1, p < .001
C. Treatment History
*Hospitalization (% yes before att) 38.1 45.5 χ2 = 0.66, df = 1, p = .042
*Outpatient treatment (% yes before att) 87.3 34.5 χ2 = 34.95, df = 1, p < .001
D. Diagnostic Variables
MDD 46.7 53.7 χ2 = 0.56, df = 1, p = .45
Substance 10.0 13.0 χ2 = 0.25, df = 1, p = .62
Alcohol 8.3 9.3 FET, p = 1.00
PTSD 8.3 11.1 χ2 = 0.25, df = 1, p = .62
BDI 18.0 ± 12.5 18.4 ± 13.1 t = 0.13, df = 66, p = .90
Aggression (LHA) 14.8 ± 3.9 14.9 ± 3.8 t = 0.16, df = 77, p = .87
Impulsivity (BIS) 74.9 ± 4.6 74.7 ± 4.9 t = 0.19, df = 107, p = .85
Hostility (Buss-Durkee) 34.5 ± 17.0 33.1 ± 15.4 t = 0.31, df = 50, p = .76
HamD 18.5 ± 10.6 18.6 ± 13.5 t = 0.05, df = 95.6, p = .96
Hopelessness 9.0 ± 6.2 8.4 ± 5.6 t = 0.4, df = 51, p = .69
SAS-SR 2.3 ± 0.6 2.4 ± 0.7 t = 0.55, df = 60, p = .58
Poor psychosocial outcome, 10-yr (% GAS
< 61) 37.7 50.9 χ2 = 2.0, df = 1, p = .15
Poor psychosocial outcome, baseline (% GAS
< 61) 61.9 85.2 χ2 = 7.9, df = 1, p = .005
E. Personality Traits
IPDE no. borderline criteria met 5.75 ± 1.11 5.30 ± 1.38 t = 1.97, df = 113, p = .051
MMPI psychopathic deviant subscale 61.98 ± 6.87 63.14 ± 7.76 t = 0.79, df = 97, p = .44
DIB-R total 3.06 ± 2.79 5.04 ± 3.30 t = 2.45, df = 55, p = .02
Affect 5.35 ± 3.08 6.96 ± 3.26 t = 1.91, df = 55, p = .06
Cognition 1.45 ± 2.05 2.77 ± 2.22 t = 2.33, df = 55, p = .02
Impulse action patterns 2.13 ± 2.25 3.46 ± 2.93 t = 1.90, df = 46.3, p = .06
Interpersonal Relationships 4.10 ± 3.63 5.62 ± 3.97 t = 1.51, df = 55, p = .14
NEO Neuroticism 31.78 ± 9.06 32.89 ± 9.35 t = 0.46, df = 57, p = .65
NEO Extraversion 29.59 ± 10.56 28.04 ± 10.38 t = 0.57, df = 57, p = .57
NEO Openness to Experience 34.38 ± 6.27 31.93 ± 8.50 t = 1.27, df = 57, p = .21
NEO Agreeableness 33.97 ± 5.63 32.04 ± 8.32 t = 1.06, df = 57, p = .29
NEO Conscientiousness 33.81 ± 8.79 32.00 ± 10.99 t = 0.70, df = 57, p = .48
*Hospitalization and Outpatient were re-coded so that for attempters, it reflects the interval from B→ 1st attempt in
FU; for non-attempters, B→ most recent visit. MDD: major depressive disorder, PTSD: posttraumatic stress disorder,
BDI: Beck Depression Inventory, LHA: Life History of Aggression, BIS: Barratt Impulsiveness Scale, HRSD: Hamilton
Rating Scale for Depression, SAS-SR = Social Adjustment Scale, GAS: Global Assessment Scale, IPDE: International
Personality Disorders Examination, DIB-R: Revised Diagnostic Interview for Borderlines.
10 SOLOFF AND CHIAPPETTA

(alcoholic pancreatitis, end-stage liver disease) at a mean age of 42.3 years.

Finally, 4 subjects died of natural causes, at a mean age of 55.8 years.

PSYCHOSOCIAL AND SOCIOECONOMIC OUTCOMES (TABLE 1)

Overall psychosocial functioning (GAS) improved modestly but significantly

from baseline to follow-up, with a 10-year mean (SD) of 61.2 (12.3); how-
ever, 44% of subjects still met our criterion for poor psychosocial outcome
(i.e., GAS < 61). While subjects tended to report themselves as improved in
overall social and vocational functioning (SAS-sr), objective vocational and
economic data indicated a poor outcome for many. At 10-year follow-up,
43.2% were unemployed, which was not statistically different from baseline.
Personal incomes, though significantly improved, remained below the federal
poverty level for 41.5% of subjects, household incomes for 35.5%, while
32.5% were dependent on government assistance.

ATTEMPTERS VERSUS NON-ATTEMPTERS (TABLE 2)

Interval attempters were compared to non-attempters at 10-year follow-up.

There were surprisingly few significant differences between groups after Bon-
ferroni correction; however, several important trends were noted. Attempters
tended to be older than non-attempters and more likely to have had hospital
admissions prior to any attempt. Compared to non-attempters, significantly
fewer attempters sought outpatient treatment in the interval. The two groups
did not differ in severity of BPD criteria, number of BPD criteria met (IPDE),
or Five Factor personality dimensions. Similarly, there was no difference be-
tween groups in comorbidity with Axis I disorders, or measures of mood,
aggression, impulsivity, or hostility. At baseline, more attempters tended to
have poor psychosocial function (GAS < 61) compared to non-attempters;
however, by 10-year follow-up, psychosocial outcomes and overall social
and vocational adjustment (SAS-sr) did not differ between groups. Similarly,
mean GAS scores at follow-up were modestly but not significantly lower
among attempters. Among socioeconomic indicators, attempters tended to
live alone or in households with incomes below the federal poverty limit. (A
similar trend is apparent for personal incomes.) Half of interval attempters
were unemployed compared to one-third of non-attempters.

PREDICTORS OF INTERVAL SUICIDE ATTEMPTS (TABLE 3)

Risks of an interval attempt associated with baseline and follow-up variables

were examined individually using Cox regression models. Increased risk was
significantly associated with non-specific measures of illness severity (e.g.,
hospitalization and/or OPD treatment in the interval prior to any attempt),
with poor social and vocational adjustment (SAS-sr), and with poor psy-
chosocial function (GAS) at baseline. Among clinical variables, there were
notable trends for baseline depressed mood (BDI-II, HamD, Hopelessness)
to predict interval attempts (p < .01), but not for MDD at baseline or follow-
up. Similarly, there was a trend for comorbidity with SUD and a history of
SUICIDAL BEHAVIOR IN BPD 11

TABLE 3a. Predictors of Increased Risk: Baseline and Follow-Up (FU)

Risk Factors (Individual Cox Regression Models)
Risk Factor Risk Ratio 95% CI p value
A. BASELINE
Any substance use disorder 2.007 1.171, 3.440 .01
Beck Depression Inventory 1.037 1.013, 1.063 .003
Hamilton Depression (HamD) 1.051 1.017, 1.085 .003
Beck Hopelessness Scale 1.070 1.018, 1.123 .007
Global Assessment Scale (GAS) 0.943 0.919, 0.968 < .001
Poor psychosocial function (%GAS < 61) 2.729 1.285, 5.794 .009
Social Adjustment Scale (SAS-SR) 2.851 1.681, 4.835 < .001
SAS-SR Work 1.379 1.085, 1.752 .009
SAS-SR Social 1.891 1.262, 2.834 .002
History of child sexual abuse 2.149 1.191, 3.877 .01
Scale for Suicidal Ideation
Prior 2 weeks 1.031 1.006, 1.056 .01
Current 1.042 1.012, 1.073 .009
Reasons for Living Total Score (Protective) 0.988 0.981, 0.995 .001
RFL mean item score 0.577 0.413, 0.806 .001
RFL survival coping beliefs 0.569 0.426, 0.759 < .001
RFL fear of suicide 0.680 0.509, 0.908 .009
B. FOLLOW-UP
High lethality attempter status 5.354 2.883, 9.941 < .001
Number of high lethality attempts 1.839 1.471, 2.299 < .001
Any outpatient treatment over FU 2.232 1.209, 4.123 .010
Psych hospitalization in interval before attempt 4.442 2.543, 7.759 < .001
Any outpatient treatment in interval before attempt 3.127 1.767, 5.533 < .001

childhood sexual abuse to increase attempt risk (p = .01). Suicidal ideation

at baseline (SSI) was related to risk (p < .01), but not at follow-up. It is
noteworthy that no measures of personality pathology contributed signifi-
cantly to attempt risk, including severity of baseline BPD criteria (DIB/DIB-r,
IPDE), impulsiveness (BIS), aggression (LHA), and the dimensional traits of
the Five-Factor Inventory (NEO-FFI-3). The Reasons for Living scale, ob-
tained at baseline, diminished attempt risk, especially the Survival Coping
Beliefs subscale (p = < .001) and the Fear of Suicide subscale (p = .009). At
follow-up, status as a high lethality attempter and number of high lethality
attempts in the interval significantly contributed to attempt risk (this was
not true at baseline). There were no predictive associations for Axis I comor-
bidities (including MDD, SUD), mood, hostility, or aggression measures at
follow-up.
A final Cox regression model examined risk of interval attempts using
significant variables from the individual models. Increased risk was associ-
ated with: (a) psychiatric hospitalization in the interval, but prior to any at-
tempt, (b) poor social and vocational adjustment at baseline (SAS-sr), and (c)
poor psychosocial functioning at baseline (GAS) (Table 3).
12 SOLOFF AND CHIAPPETTA

TABLE 3b. Predictors of Increased Risk: Final Cox Regression Model

Risk Factor RR 95% CI p value
Psychiatric hospitalization in interval 3.42 1.86, 6.30 < .001
(Higher) SAS overall score at baseline 2.18 1.3, 3.66 .003
(Poor) Global Assessment Score at baseline 0.96 0.92, 0.99 .007
Note. Bonferroni correction for individual models p < 0.001; trends shown for p < .01.

DISCUSSION

After 10 years, the strongest predictors of interval suicide attempts are non-
specific measures of illness severity and baseline measures of social, voca-
tional, and psychosocial functioning. They have appeared as predictors of at-
tempt behavior in every previous analysis, though in differing measures (e.g.,
GAS, SAS-sr; Soloff & Chiappetta, 2012a; Soloff & Fabio, 2008). These risk
factors also characterize a high lethality BPD subtype (Soloff & Chiappetta,
2012a, 2017).
Illness severity is a non-specific risk factor, represented by inpatient re-
cruitment, inpatient treatment, and outpatient treatment events in the inter-
val occurring prior to any attempt behavior. Individual regression models
suggest contributions to illness severity from comorbid substance use disor-
ders, depressed mood, suicidal ideation, childhood sexual abuse, and poor
social, vocational, and psychosocial functioning. Each of these risk factors
has previously been associated with suicidal behavior in BPD in retrospective
and cross-diagnostic studies (Soloff, 2005, for a review).
Prospective predictors of suicidal behavior change over time (Soloff &
Chiappetta, 2012b, 2017). Prior studies of suicidal behavior in BPD identi-
fied suicide risk factors related to core dimensions of borderline psychopa-
thology, especially impulsivity, impulsive-aggression, affective instability, and
negative affectivity (Soloff, 2005, for review). In cross-diagnostic studies, im-
pulsive aggression increases likelihood of suicidal behavior in both depressed
and non-depressed inpatients with BPD (Soloff, Lynch, Kelly, Malone, &
Mann, 2000), increases the frequency of suicide attempts (Brodsky, Malone,
Ellis, Dulit, & Mann, 1997), and is associated with completed suicide in
subjects with BPD (McGirr et al., 2007, 2009).
Affective instability and negative affectivity are frequently associated
with attempt behavior in BPD and other PD subjects (Koenigsberg et al.,
2002). The Collaborative Longitudinal Personality Disorder Study (CLPDS)
reported that affective instability was the strongest predictor of interval at-
tempt behavior among PD subjects followed for 2 years; 78% of attempt-
ers were diagnosed with BPD (Yen et al., 2004). Negative affectivity (with
impulsivity) was the most robust predictor of interval attempt behavior at
7-year follow-up (Yen et al., 2009). In a stress-diathesis model of suicide,
impulsivity, impulsive aggression, affective instability, and negative affectiv-
ity are traits of temperament that may reflect an underlying neurobiologic
diathesis to suicidal behavior at times of stress (Mann, 2003; Siever, 2008).
Core BPD traits, such as negative affectivity and impulsive aggression, were
associated with increased risk of attempt in our 8-year analysis, but not after
SUICIDAL BEHAVIOR IN BPD 13

10 years of follow-up (Soloff & Chiappetta, 2017). These core BPD traits are
time-varying risk factors that diminish in severity over time (Zanarini et al.,
2007). It is unclear if remission is due to developmental maturation (“time
alone”) or treatment in the interval (Zanarini et al., 2007).
Remission of BPD psychopathology over time, previously described in
retrospective studies, has now been well documented in two remarkable pro-
spective longitudinal studies. The CLPDS followed 668 treatment-seeking
patients, predominately but not exclusively outpatients, diagnosed with 1
of 4 PDs (only 26.2% BPD), compared to an MDD/no PD control group
(Gunderson et al., 2000). Remission of the categorical BPD diagnosis was
noted in 85% of subjects at 10-year follow-up (Gunderson et al., 2011).
The McLean Study of Adult Development (MSAD) followed 290 former
inpatients with BPD and also reported diagnostic remission in 93% at 10-
year follow-up (Zanarini, Frankenburg, Reich, & Fitzmaurice, 2010a). It is
important to note that remission based on categorical DSM and DIB-r di-
agnoses is a statistical construct and does not imply sustained improvement
or psychosocial recovery (defined as remission with good social and voca-
tional functioning). Acute symptoms in BPD wax and wane as situational
stressors act upon chronic personality vulnerabilities. At 10-year follow-up
in the MSAD study, 30% of patients who had achieved remission for 2 years
subsequently experienced a recurrence, and 34% of patients who had re-
covered lost their psychosocial recovery (Zanarini, Frankenburg, Reich, &
Fitzmaurice, 2010a). We suggest that sustained remission depends, in part,
on diminished temperamental vulnerability to negative life events, as well as
enhanced coping skills.
Both CLPDS and MSAD studies addressed suicidality, though it was not
a primary aim of either study. Choice of suicide risk factors was opportunis-
tic and did not include characteristics of attempt behavior or measures spe-
cific for suicide risk (e.g., SSI, SIS, RFL). Among BPD subjects in the CLPDS,
20.5% attempted suicide in the first 2 years of follow-up, similar to our own
2-year experience of 24.8% interval attempters (Soloff & Fabio, 2008; Yen
et al., 2003). Among all subjects in the CLPDS sample, a baseline diagnosis
of BPD and substance use disorder predicted attempts in the 2-year interval.
Worsening of MDD, drug abuse, or alcohol use predicted a suicide attempt
in the ensuing month. The presence of MDD also predicted attempt behavior
in the first year of our study, though not in following years (Soloff & Fabio,
2008). Using all study participants, the 2-year CLPDS report also found that
affective instability, female gender, and a childhood history of sexual abuse
predicted attempts (Yen et al., 2003). At 3-year follow-up, assessing only PD
subjects, negative life events (involving love/marriage and crime/legal events)
were associated with interval attempts (Yen et al., 2005). At 7 years, nega-
tive affectivity and lack of premeditation (impulsivity) predicted interval at-
tempts in a sample that included all study participants (Yen et al., 2009).
The MSAD study assessed predictive associations of baseline charac-
teristics of BPD subjects with interval suicide attempts at 16-year follow-up
(Wedig et al., 2012). Though severity and significance of risk factors vary
over time, time-varying risk factors were not reassessed at follow-up in this
analysis. A final multivariate model found significant associations between
14 SOLOFF AND CHIAPPETTA

interval attempt behavior and baseline MDD, SUD, PTSD, self-harm, adult
sexual assault, caretaker suicide, affective instability, and severe dissociation.
Comparisons between these studies and our own are made difficult by
the important sample and method differences noted above. Nonetheless,
there are some noteworthy overlapping results. Our study noted the impor-
tance of comorbid MDD as a predictor of attempt behavior in the first year
of follow-up, impulsivity and negative affectivity at the 8-year follow-up,
and SUD and childhood sexual abuse at 10 years.
Time-varying risk factors were assessed at each follow-up in our study.
One consequence of the progressive remission of BPD symptoms over time
was the paucity of significant differences between interval attempters and
non-attempters at 10-year follow-up. Differences between groups were pri-
marily related to severity of baseline risk factors. Contrary to expectation,
attempters did not differ from non-attempters in severity of BPD criteria,
personality traits, or Axis I comorbidity. Instead, they tended to have more
non-specific illness severity and poor psychosocial and socioeconomic func-
tion at baseline, reflecting predictors of interval attempts identified in the
final regression model. We found no significant relationship between suicidal
behavior and psychosocial outcome at 10-year follow-up, that is, a favorable
outcome could still result despite a history of suicidal behavior.
Poor psychosocial functioning and low levels of social adjustment are
related to attempt behavior, high lethality attempts, and suicide completion
in BPD (Kelly, Soloff, Lynch, Haas, & Mann, 2000; Soloff, 2005). Many ret-
rospective studies have detailed the social and vocational impairment of sui-
cides with BPD (and other PDs; Heikkinen et al., 1997; Paris & Zweig-Frank,
2001; Runeson & Beskow, 1991; Runeson, Beskow, & Waern, 1996). In a
retrospective survey of suicide among BPD patients 27 years after hospital
discharge, Paris and Zweig-Frank (2001) reported that years of illness, loss
of supportive relationships, and social isolation were prominent risk factors
for suicide. Recent life events, interpersonal loss and conflict, job problems,
and unemployment were significantly associated with death by suicide in an
epidemiological study of PD subjects (61% Cluster B), compared to non–PD
suicides (Heikkinen et al., 1997). Unemployment, poor social integration,
and poor psychosocial functioning are well-known risk factors for suicide
across diagnoses and in non-clinical populations (Angst & Clayton, 1998;
Angst et al., 2014; Duberstein, Conwell, Eberly, Evinger, & Caine, 2004;
Kposowa, 2001). At 10-year follow-up, the mean GAS score for subjects in
our study was only 61.2, very close to the definition of poor psychosocial
outcome (GAS < 61). Similarly, the CLPDS reported a mean GAF of 57 at
10-year follow-up (Gunderson et al., 2011). The MSAD reported psychoso-
cial recovery in only 50%, defined by remission of diagnosis, good social and
vocational functioning, and a GAF > 61 (Zanarini et al., 2010b).
At 10-year follow-up in the MSAD study, poor vocational attainment,
and not social adjustment, contributed most to poor psychosocial outcomes
(Zanarini et al., 2010b). Even among subjects who started the study with
good psychosocial functioning, a large majority (87%) lost good functioning
by 10-year follow-up. The primary cause for loss of previously good psycho-
social functioning (in 77.6% of cases) was loss of vocational achievement.
SUICIDAL BEHAVIOR IN BPD 15

In the CLPDS, 64% of BPD subjects were unemployed at the 10-year as-
sessment, more than other PD comparison groups (Gunderson et al., 2011).
Our subjects reported no significant changes in social relationships; however,
more attempters (43.6%) reported living alone, compared to non-attempters
(25.4%). Vocational impairment was manifested across multiple measures,
including high rates of unemployment, household incomes at poverty lev-
el, and continued dependence on government support. At 10-year follow-
up, 32.5% of our subjects were receiving some government assistance (up
from 26.7% at baseline), somewhat less than the 44% who were receiving
SSDI in the 10-year MSAD analysis (Zanarini, Jacoby, Frankenburg, Reich,
& Fitzmaurice, 2009). Educational and vocational deficits associated with
years of acute illness early in life may contribute to these poor socioeconomic
outcomes and, thereby, to long-term risk of suicidal behavior in BPD.
Symptomatic and diagnostic improvement in BPD is necessary but not
sufficient to produce functional recovery. The important contribution of Axis
I comorbidity, especially MDD, to poor psychosocial outcome was reported
in our 8-year follow-up, where MDD was the only risk factor predictive of
poor psychosocial outcome (Soloff & Chiappetta, 2012b). The prevalence of
MDD remained unchanged in half of subjects over 10 years, despite wide-
spread participation in treatment. As this was not designed as a treatment
study, it was not possible to assess treatment effectiveness. Nonetheless, psy-
chosocial outcomes, assessed as good (GAS > 61) versus poor (GAS < 61) or,
as a change in GAS from baseline to follow-up, were statistically indepen-
dent of treatment received in the interval.
Affective symptoms of BPD, including depressed mood, are among the
most refractory to change over time (Zanarini, Frankenburg, Hennen, &
Silk, 2003; Zanarini et al., 2007). The refractoriness of MDD among bor-
derline patients begs the question of whether characterologic features of the
disorder (such as negative affectivity, low self-esteem, and rejection sensi-
tivity) confound the diagnosis of MDD on structured interviews. An older
literature described “characterologic depression” in patients with BPD, as
reflecting long-standing negative attitudes about the self. These attitudes are
relatively refractory to pharmacotherapy, and they indicate a need for psy-
chotherapeutic intervention (Soloff, Cornelius, & George, 1991).
The persistence of MDD may have an adverse effect on attaining and
maintaining employment, and on workplace productivity. Workplace pro-
ductivity is a significant predictor of long-term remission in MDD (Jha et al.,
2016). Work-focused interventions (e.g., job coaching) have been effective as
adjunctive treatments in improving work outcomes for employed dysthymic
subjects (Adler et al., 2015; Hees, de Vries, Koeter, & Schene, 2013) and may
have relevance as an adjunctive treatment for depressed subjects with BPD.
Although vocational rehabilitation programs have been proposed in the past
for patients with BPD, such programs have not been widely implemented
(Links, 1993).
Suicidal behavior diminished markedly over the 10-year interval across
multiple indicators. Frequency of attempts dropped from 83.1% of subjects
at baseline to 46.6% at follow-up and self-injury from 48.7% at baseline to
21.7% at follow-up. Zanarini and colleagues (2008) also reported decreased
16 SOLOFF AND CHIAPPETTA

suicidal behavior for the inpatient MSAD sample at 10-year follow-up. A

baseline history of suicide attempt diminished from 79% at baseline to less
than 13% at 10-year follow-up, and self-mutilation from 90% at baseline to
less than 18% over the 10-year period (Zanarini et al., 2008). Zanarini and
colleagues (2008) suggested that the decline in suicidal behavior may be due
to effects of treatment or maturation (time alone). It is likely that diminished
intensity in personality trait vulnerabilities, especially in impulsive aggres-
sion and negative affectivity, mediate this favorable outcome.
Our longitudinal study confirms core findings of the CLPDS and MSAD
studies in regard to the long-term outcome of BPD. A majority of subjects
will experience diagnostic remission and symptomatic improvement over 10
years’ time; however, many will not attain functional recovery in terms of
psychosocial, vocational, and economic achievement. At 10-year follow-up,
44% of subjects in our study had poor psychosocial outcomes, with un-
employment, poverty-level incomes, and reliance on government assistance.
Long-term psychosocial outcome for these subjects resembles that for other
serious and persistent mental illnesses. Empirical evidence from all three
studies strongly supports an additional focus on educational and vocational
rehabilitation for BPD in order to mitigate poor psychosocial outcomes and
risk of suicidal behavior in the long term.

LIMITATIONS

The absence of clinical comparison groups limits generalization of our find-

ings. Similarly, our analyses are limited to volunteer subjects who continued
participation for 10 or more years. To address this issue, we compared the
118 10-year subjects with the remaining subjects in the database on all base-
line variables. After Bonferroni correction (at p < .001), subjects in the 10-
year sample were older, had lower GAS scores, and were more often suicide
attempters. This suggests greater baseline impairment among subjects who
remained in the study for 10 or more years; however, the long-term outcome
for these cooperative subjects in terms of illness severity and social and vo-
cational achievement may differ from those who dropped out of the study.

REFERENCES

Adler, D. A., Lerner, D., Visco, Z. L., Greenhill, the Cox regression model: Methods for lon-
A., Chang, H., Cymerman, E., … Rogers, gitudinal psychiatric research. Acta Psychi-
W. H. (2015). Improving work outcomes of atrica Scandinavica, 74(6), 529–535.
dysthymia (persistent depressive disorder) American Psychiatric Association. (2000). Diag-
in an employed population. General Hospi- nostic and statistical manual of mental dis-
tal Psychiatry, 37(4), 352–359. orders (4th ed., text rev.). Washington, DC:
Allebeck, P., Allgulander, C., & Fisher, L. D. Author.
(1988). Predictors of completed suicide American Psychiatric Association. (2013). Diag-
in a cohort of 50,465 young men: Role of nostic and statistical manual of mental dis-
personality and deviant behaviour. BMJ, orders (5th ed.). Washington, DC: Author.
297(6642), 176–178. Angst, J., & Clayton, P. J. (1998). Personality,
Allgulander, C., & Fisher, L. D. (1986). Survival smoking and suicide: A prospective study.
analysis (or time to an event analysis) and
SUICIDAL BEHAVIOR IN BPD 17

Journal of Affective Disorders, 51(1), 55– tifact? A case-control study. Psychological

62. Medicine, 34, 1331–1337.
Angst, J., Hengartner, M. P., Rogers, J., Schnyder, Endicott, J., Spitzer, R. L., Fleiss, J. L., & Cohen,
U., Steinhausen, H. C., Ajdacic-Gross, V., J. (1976). The global assessment scale. A
& Rossler, W. (2014). Suicidality in the pro- procedure for measuring overall severity of
spective Zurich study: Prevalence, risk fac- psychiatric disturbance. Archives of Gen-
tors and gender. European Archives of Psy- eral Psychiatry, 33(6), 766–771.
chiatry and Clinical Neuroscience, 264(7), First, M. B., Spitzer, R. L., Gibbon, M., & Wil-
557–565. doi:10.1007/s00406-014-0500-1 liams, J. B. W. (2005). Structured Clinical
Ansell, E. B., Wright, A. G. C., Markowitz, J. C., Interview for DSM-IV-TR Axis I Disor-
Sanislow, C. A., Hopwood, C. J., Zanarini, ders–Patient Edition (SCID-I/P, 4/2005
M. C., … Grilo, C. M. (2015). Personality revision). New York, NY: Biometrics Re-
disorder risk factors for suicide attempts search Department, New York State Psychi-
over 10 years of follow-up. Personality Dis- atric Institute.
orders: Theory, Research, and Treatment, Gunderson, J. G., Kolb, J. E., & Austin, V. (1981).
6(2) 161–167. doi:10 1037/per0000089 The diagnostic interview for borderline
Barratt, E. S. (1965). Factor analysis of some psy- patients. American Journal of Psychiatry,
chometric measures of impulsiveness and 138(7), 896–903.
anxiety. Psychological Reports, 16, 547– Gunderson, J. G., Shea, M. T., Skodol, A. E., Mc-
554. Glashan, T. H., Morey, L. C., Stout, R. L.,
Barratt, E. S., & Stanford, M. S. (1995). Impul- . . . Keller, M. B. (2000). The Collaborative
siveness. In C. G. Costello (Ed.), Personality Longitudinal Personality Disorders Study:
characteristics of the personality disordered Development, aims, design, and sample
(pp. 91–118). New York, NY: Wiley. characteristics. Journal of Personality Dis-
Beck, A. T., Beck, R., & Kovacs, M. (1975). Clas- orders, 14(4), 300–315.
sification of suicidal behaviors: I. Quantify- Gunderson, J. G., Stout, R. L., McGlashan, T. H.,
ing intent and medical lethality. American Shea, M. T., Morey, L. C., Grilo, C. M., .
Journal of Psychiatry, 132(3), 285–287. . . Skodol, A. E. (2011). Ten-year course
Beck, A. T., Schuyker, D., & Herman, I. (1974). of borderline personality disorder: Psy-
Development of suicidal inent scales. In chopathology and function from the Col-
A. T. Beck, H. L. P. Resnick, & D. Let- laborative Longitudinal Personality Disor-
tiem (Eds.), The prediction of suicide (pp. ders study. Archives of General Psychiatry,
45–56). Bowie, MD: Charles Press. 68(8), 827–837. doi:10.1001/archgenpsy-
Beck, A. T., & Steer, R. A. (1988). Beck Hopeless- chiatry.2011.37
ness Scale: Manual. San Antonio, TX: Psy- Guy, W. (1976). ECDEU Assessment Manual of
chological Corporation. Psychopharmacology–Revised. Rockville,
Beck, A. T., Ward, C. H., Mendelson, M., Mock, MD: National Institute of Mental Health,
J., & Erbaugh, J. (1961). An inventory for Psychopharmacology Research Branch.
measuring depression. Archives of General Hathaway, S., & Meehl, P. (1951). An atlas for
Psychiatry, 4, 561–571. the clinical use of the MMPI. Minneapolis,
Black, D. W., Blum, N., Pfohl, B., & Hale, N. MN: University of Minnesota Press.
(2004). Suicidal behavior in borderline Hees, H. L., de Vries, G., Koeter, M. W. J., &
personality disorder: Prevalence, risk fac- Schene, A. H. (2013). Adjuvant occupa-
tors, prediction, and prevention. Journal tional therapy improves long-term depres-
of Personality Disorders, 18(3), 226–239. sion recovery and return-to-work in good
doi:10.1521/pedi.18.3.226.35445 health in sick-listed employees with major
Brodsky, B. S., Malone, K. M., Ellis, S. P., Dulit, R. depression: Results of a randomized con-
A., & Mann, J. J. (1997). Chartacteristics trolled trial. Occupational and Environ-
of borderline personality disorder associat- mental Medicine, 70, 252–260.
ed with suicidal behavior. American Journal Heikkinen, M., Isometsa, E. T., Henriksson, M.
of Psychiatry, 154, 1715–1719. M., Marttunen, M. J., Aro, H. M., & Lon-
Brown, G. L., & Goodwin, F. K. (1986). Cerebro- nqvist, J. K. (1997). Psychosocial factors
spinal fluid correlates of suicide attempts and completed suicide in personality disor-
and aggression. Annals of the New York ders. Acta Psychiatrica Scandinavica, 95(1),
Academy of Sciences, 487, 175–188. 49–57.
Buss, A. H., & Durkee, A. (1957). An inventory Jha, M. K., Minhajuddin, A., Greer, T. L., Carmo-
for assessing different kinds of hostility. dy, T., Rush, J. A., & Trivedi, M. H. (2016).
Journal of Consulting and Clinical Psychol- Early improvement in work productivity
ogy, 21(4), 343–349. predicts future clinical course in depressed
Duberstein, P. R., Conwell, K. R., Eberly, S., Ev- outpatients: Findings from the CO-MED
inger, J. S., & Caine, E. D. (2004). Poor trial. American Journal of Psychiatry,
social integration and suicide: Fact or ar- 173(12), 1196–1204.
18 SOLOFF AND CHIAPPETTA

Kelly, T. M., Soloff, P. H., Lynch, K. G., Haas, G. Runeson, B., & Beskow, J. (1991). Borderline per-
L., & Mann, J. J. (2000). Recent life events, sonality disorder in young Swedish suicides.
social adjustment, and suicide attempts in Journal of Nervous and Mental Disease,
patients with major depression and border- 179(3), 153–156.
line personality disorder. Journal of Person- Runeson, B. S., Beskow, J., & Waern, M. (1996).
ality Disorders, 14(4), 316–326. The suicidal process in suicides among
Koenigsberg, H. W., Harvey, P. D., Mitropoulou, young people. Acta Psychiatrica Scandi-
V., Schmeidler, J., New, A. S., Goodman, navica, 93(1), 35–42.
M., . . . Siever, L. J. (2002). Characterizing Schneider, B., Schnabel, A., Wetterling, T., Bar-
affective instability in borderline personali- tusch, B., Weber, B., & Georgi, K. (2008).
ty disorder. American Journal of Psychiatry, How do personality disorders modify sui-
159(5), 784–788. cide risk? Journal of Personality Disorders,
Kposowa, A., J. (2001). Unemployment and sui- 22(3), 233–245.
cide: A cohort analysis of social factors Schneider, B., Wetterling, T., Sargk, D., Schnei-
predicting suicide in the US National Lon- der, F., Schnabel, A., Maurer, K., & Fritze,
gitudinal Mortality Study. Psychological J. (2006). Axis I disorders and personality
Medicine, 31, 127–138. disorders as risk factors for suicide. Euro-
Linehan, M. M., Goodstein, J. L., Nielson, S. L., pean Archives of Psychiatry and Clinical
& Chiles, J. A. (1983). Reasons for stay- Neuroscience, 256(1), 17–27. doi:10.1007/
ing alive when you are thinking of killing s00406-005-0593-7
yourself: The Reasons for Living Inventory. Siever, L. J. (2008). Neurobiology of aggression
Journal of Consulting and Clinical Psychol- and violence. American Journal of Psychia-
ogy 51, 276–286. try, 165(4), 429–442.
Links, P. S. (1993). Psychiatric rehabilitation Soloff, P. H. (2005). Risk factors for suicidal be-
model for borderline personality disorder. havior in borderline personality disorder: A
Canadian Journal of Psychiatry, 38(Suppl. review and update. In M. C. Zanarini (Ed.),
1), S35–S38. Borderline personality disorder (pp. 333–
Loranger, A. W. (1999). International Personality 365). Boca Raton, FL: Taylor & Francis.
Disorder Examination: DSM-IV and ICD- Soloff, P. H., & Chiappetta, L. (2012a). Subtyp-
10 Interviews. Lutz, FL: Psychological As- ing borderline personality disorder by
sessment Resources. suicidal behavior. Journal of Personality
Mann, J. J. (2003). Neurobiology of suicidal be- Disorders, 26(3), 468–480. doi:10.1521/
haviour. Nature Reviews Neuroscience, pedi.2012.26.3.468
4(10), 819–828. Soloff, P. H., & Chiappetta, L. (2012b). Pro-
McGirr, A., Alda, M., Seguin, M., Cabot, S., Les- spective predictors of suicidal behavior in
age, A., & Turecki, G. (2009). Familial ag- borderline personality disorder at 6-year
gregation of suicide explained by Cluster B follow-up. American Journal of Psychia-
traits: A three-group family study of suicide try, 169(5), 484–490. doi:10.1176/appi.
controlling for major depressive disor- ajp.2011.11091378
der. American Journal of Psychiatry, 166, Soloff, P. H., & Chiappetta, L. (2017). Suicidal
1124–1134. behavior and psychosocial outcome in bor-
McGirr, A., Paris, J., Lesage, A., Renaud, J., & derlione personality disorder at 8-year fol-
Turecki, G. (2007). Risk factors for suicide low-up. Journal of Personality Disorders,
completion in borderline personality disor- 31, 774–789. Advance online publication.
der: A case-control study of Cluster B co- doi:10.1521/pedi_2017_31_280
morbidity and impulsive aggression. Jour- Soloff, P. H., Cornelius, J., & George, A. (1991).
nal of Clinical Psychiatry, 68(5), 721–729. The depressed borderline: One disorder or
McNemar, Q. (1947). Note on the sampling error two? Psychopharmacology Bulletin, 27(1),
of the difference between correlated pro- 23–30.
portions or percentages. Psychometrika, Soloff, P. H., & Fabio, A. (2008). Prospective pre-
12(2), 153–157. dictors of suicide attempts in borderline
Oquendo, M. A., Halberstam, B., & Mann, J. J. personality disorder at one, two, and two-
(2003). Risk factors for suicidal behavior. to-five year follow-up. Journal of Personali-
In M. B. First (Ed.), Standardized evalu- ty Disorders, 22(2), 123–134. doi:10.1521/
ation in clinical practice (pp. 103–129). pedi.2008.22.2.123
Washington, DC: American Psychiatric Soloff, P. H., Fabio, A., Kelly, T. M., Malone, K.
Publishing. M., & Mann, J. J. (2005). High-lethality
Paris, J., & Zweig-Frank, H. (2001). A 27-year status in patients with borderline per-
follow-up of patients with borderline per- sonality disorder. Journal of Personality
sonality disorder. Comprehensive Psychia- Disorders, 19(4), 386–399. doi:10.1521/
try, 42(6), 482–487. pedi.2005.19.4.386
SUICIDAL BEHAVIOR IN BPD 19

Soloff, P. H., Lynch, K. G., & Kelly, T. M. (2002). spective follow-up of the phenomenology of
Childhood abuse as a risk factor for suicidal borderline personality disorder. American
behavior in borderline personality disorder. Journal of Psychiatry 160(2), 274–283
Jourmal of Personality Disorders, 16(3), Zanarini, M. C., Frankenburg, F. R., Reich, D. B.,
201–214. & Fitzmaurice, G. (2010a). Time to attain-
Soloff, P. H., Lynch, K. G., Kelly, T. M., Malone, K. ment of recovery from borderline personal-
M., & Mann, J. J. (2000). Characteristics of ity disorder and stability of recovery: A 10-
suicide attempts of patients with major de- year prospective follow-up study. American
pressive episode and borderline personality Journal of Psychiatry, 167(6), 663–667.
disorder: A comparative study. American Zanarini, M. C., Frankenburg, F. R., Reich, D. B.,
Journal of Psychiatry, 157(4), 601–608. & Fitzmaurice, G. M. (2010b). The 10-year
Wedig M. M., Silverman M. H., Frankenburg F. course of psychosocial functioning among
R., Reich B., Fitzmaurice G., & Zanarini patients with borderline personality disor-
M. C. (2012). Predictors of suicide attempts der and axis II comparison subjects. Acta
in patients with borderline personality dis- Psychiatrica Scandinavica, 122, 103–109.
order over 16 years of prospective follow- doi:10.1111/j.1600-0447.2010.01543.x
up. Psychological Medicine, 42(11), 2395– Zanarini, M. C., Frankenburg, F. R., Reich, D. B.,
2404. doi:10.1017/S0033291712000517 & Fitzmaurice, G. M. (2012). Attainment
Weissman, M. M., & Bothwell, S. (1976). Assess- and stability of sustained symptomatic re-
ment of social adjustment by patient self-re- mission and recovery among patients with
port. Archives of General Psychiatry, 33(9), borderline personality disorder and Axis II
1111–1115. comparison subjects: A 16-year prospective
Yen, S., Pagano, M. E., Shea, M. T., Grilo, C. follow-up study. American Journal of Psy-
M., Gunderson, J. G., Skodol, A. E., . . . chiatry, 169, 476–483.
Zanarini, M. C. (2005). Recent life events Zanarini, M. C., Frankenburg, F. R., Reich, D. B.,
preceding suicide attempts in a personality Fitzmaurice, G., Weinberg, I., & Gunder-
disorder sample: Findings from the collab- son, J. G. (2008). The 10-year course of
orative longitudinal personality disorders physically self-destructive acts reported by
study. Journal of Consulting and Clinical borderline patients and axis II comparison
Psychology, 73(1), 99–105. subjects. Acta Psychiatrica Scandinavica,
Yen, S., Shea, M. T., Pagano, M. E., Sanislow, C. 117(3), 177–184. doi:10.1111/j.1600-
A., Grilo, C. M., McGlashan, T. H., . . . 0447.2008.01155.x
Morey, L. C. (2003). Axis I and axis II dis- Zanarini, M. C., Frankenburg, F. R., Reich, D. B.,
orders as predictors of prospective suicide Silk, K. R., Hudson, J. I., & McSweeney, L.
attempts: Findings from the collaborative B. (2007). The subsyndromal phenomenol-
longitudinal personality disorders study. ogy of borderline personality disorder: A
Journal of Abnormal Psychology, 112(3), 10-year follow-up study. American Journal
375–381. of Psychiatry, 164(6), 929–935.
Yen, S., Shea, M. T., Sanislow, C. A., Grilo, C. M., Zanarini, M. C., Gunderson, J. G., Frankenburg,
Skodol, A. E., Gunderson, J. G., . . . Morey, F. R., & Chauncey, D. L. (1989). The Re-
L. C. (2004). Borderline personality disor- vised Diagnostic Interview for Borderlines:
der criteria associated with prospectively Discriminating BPD from other Axis II dis-
observed suicidal behavior. American Jour- orders. Journal of Personality Disorders,
nal of Psychiatry, 161(7), 1296–1298. 3(1), 10–18. doi:10.1521/pedi.1989.3.1.10
Yen, S., Shea, M. T., Sanislow, C. A., Skodol, Zanarini, M. C., Jacoby, R. J., Frankenburg, F.
A. E., Grilo, C. M., Edelen, M. O., . . . R., Reich, D. B., & Fitzmaurice, G. (2009).
Gunderson, J. G. (2009). Personality traits The 10-year course of Social Security dis-
as prospective predictors of suicide at- ability income reported by patients with
tempts. Acta Psychiatrica Scandinavica, borderline personality disorder and Axis II
120(3), 222–229. doi:10.1111/j.1600- comparison subjects. Journal of Personality
0447.2009.01366.x Disorders, 23(4), 346–356. doi:10.1521/
Zanarini, M. C., Frankenburg, F. R., Hennen, J., & pedi.2009.23.4.346
Silk, K. R. (2003). The longitudinal course
of borderline psychopathology: 6-year pro-