High-Lethality status in patients with Borderline Personality Disorder
Journal of Personality Disorders, 19(4), 386–399, 2005
Ó 2005 The Guilford Press
Paul H. Soloff, MD, Anthony Fabio, PhD, Thomas M. Kelly, PhD, Kevin M. Malone, MD, and J. John Mann, MD
From the University of Pittsburgh, Western Psychiatric Institute and Clinic, Pittsburgh, PA (P.H.S., A.F., and T.M.K.). From the Department of Adult Psychiatry, Mater Hospital, Univer- sity College, Dublin, Ireland (K.M.M.). From the New York State Psychiatric Institute, Colum- bia University, New York (J.J.M.).
This paper was presented at the 157th Annual Meeting of the American Psychiatric Associa- tion, May 3, 2004 in New York City.
This research was supported by the National Institute of Mental Health grant MH048463 (P.H.S.).
Address correspondence and reprint requests to Dr. Paul H. Soloff, Western Psychiatric Insti- tute and Clinic, 3811 O’Hara Street, Pittsburgh, PA 15213; E-mail: soloffph@upmc.edu
Introduction
Recurrent suicidal behaviors in patients with Borderline Personality Disorder (BPD) are often considered communicative gestures; however, 10% complete suicide. This study seeks to identify risk factors for sui- cide within a BPD sample by comparing patients with High- and Low- Lethality attempts. BPD attempters (n = 113) were assessed on dem- ographic, diagnostic, and personality variables: clinical symptoms, suicidal behaviors; childhood, family, and treatment histories; social adjustment; and recent life events. Forty-four High-Lethality attempt- ers, defined by a score of 4 or more on Beck’s Medical Lethality Scale, were compared to 69 Low-Lethality attempters. Discriminating vari- ables were entered in a multivariate logistic regression model to define predictors of High-Lethality status. High-Lethality attempters were older, with children, less education, and lower socioeconomic class (SES) than Low-Lethality attempters. They were more likely to have Ma- jor Depressive Disorder (MDD), co-morbid Antisocial Personality Disor- der (ASPD), and family histories of substance abuse. They reported greater intent to die, more lifetime attempts, hospitalizations, and time in the hospital. High-Lethality status was best predicted by low SES, co-morbid ASPD, extensive treatment histories, and greater intent to die. These characteristics resemble profiles of patients who complete suicide, are not specific for BPD, and do not include impulsivity, aggres- sion, or severity of BPD criteria.
Suicide attempts by patients with Borderline Personality Disorder (BPD) are widely characterized as “communicative gestures,” or “ambivalent” at- tempts, yet 10% die by suicide, making BPD one of the most lethal of psy- chiatric disorders (Gunderson & Ridolfi, 2001; Paris & Zweig-Frank, 2001).
BPD is the only psychiatric diagnosis defined, in part, by recurrent suicidal behavior (DSM-IV). The prevalence rate for suicide attempts among con- secutively studied inpatients and outpatients with BPD is over 70%, with an average of 3 lifetime attempts per patient (Soloff, Lynch, Kelly, Malone, & Mann, 2000). Suicidal ideation and attempt behavior precede comple- tion by years, becoming a chronic and habitual response to adverse life events (Mehlum, Friis, Vaglum, & Karterus, 1994). Surprisingly, little is known of risk factors for suicide completion in patients with BPD, or how an early history of repeated “gestures” progresses to death.
In a stress-diathesis model for suicidal behavior, comorbidity with BPD represents a chronic vulnerability to suicide attempts and completion (Mann, Waternaux, Haas, & Malone, 1999). Comorbidity with BPD increases the likelihood of attempt behavior across diagnoses among inpatients, in- creases the number and lethality of suicide attempts among patients with Major Depressive Disorder (MDD) or substance use disorders (SUD), and is a risk factor for suicide completion in epidemiologic studies (Mann et. al., 1999; Fyer, Frances, Sullivan, Hurt, & Clarkin, 1998; Malone, Haas, Sweeney, & Mann, 1995; Corbitt, Malone, Haas, & Mann, 1996; Rich & Runeson, 1992). It is unclear to what extent the diathesis to suicide at- tempts and completion accompanying BPD is associated with its charac- teristic personality traits, or, alternatively, with the clinical, social, and vocational consequences of this chronic disorder.
Impulsivity, aggression, and affective instability are prominent personal- ity traits related to temperament in BPD. These personality traits have been associated with high-frequency, low-lethality suicide attempts inde- pendent of diagnostic context (Hirschfeld & Davidson 1988). Among inpa- tients with BPD, impulsivity, aggression, and antisocial personality traits are associated with both occurrence and frequency of suicide attempts; however, the relationship of these traits to medical lethality and suicide completion in BPD remains ill-defined (Soloff et. al., 2000; Brodsky, Ma- lone, Ellis, Dulit, & Mann, 1997; Soloff et. al., 1994a). These personality traits are also associated with evidence of diminished central serotonergic regulation in BPD, and may represent a biologic diathesis for suicidal be- havior (Oquendo & Mann, 2000). However, the biological component of BPD is just one part of a complex contribution to suicide risk.
The risk of suicidal behavior in BPD may also be influenced by high rates of comorbidity with acute stressors such as MDD, or SUD. In a stress- diathesis model of suicide, MDD or SUD may be conceptualized as acute disorders (superimposed on the personality disorder) which increase sui- cide intent. However, the effects of these acute stressors on suicide at- tempts in patients with BPD are inconsistent across studies. MDD and SUD increase the number and seriousness of attempts among BPD adults and adolescents in some studies (Fyer et. al., 1998; Friedman, Aronoff, Clarkin, Corn, & Hurst, 1983), but not all (Soloff et. al., 1994a; Shearer, Peters, Quaytman, & Wadman, 1988).
Risk factors among BPD patients who complete suicide are derived from retrospective, epidemiologic surveys of suicide in general populations (Runeson & Beskow, 1991; Isomesta et al., 1996), or follow-up studies of BPD patients treated at a single facility (Paris & Zweig-Frank, 2001; Kull- gren, 1988; Kjelsberg, Eikeseth, & Dahl, 1991; Stone, 1989). These studies highlight the social and vocational consequences of BPD as risk factors for suicide completion. BPD patients who complete suicide are often char- acterized by lower socioeconomic status, and problems with school, em- ployment, and finances (Runeson & Beskow, 1991). Mirroring the attempt literature, studies of suicide completion in BPD (or other Cluster B disor- ders) are inconsistent in regard to the effects of comorbid MDD or SUD, though less severe depressive syndromes (e.g., Depressive Disorder, NOS) appear highly prevalent among BPD completers (Isomesta et. al., 1996). Adverse childhood experiences, antisocial behavior, and comorbid Anti- social Personality Disorder (ASPD) are all more prevalent among BPD com- pleters than comparison groups (Rich & Runeson, 1992; Runeson & Bes- kow, 1991; Isomesta et. al., 1996).
Retrospective studies lack reliable, structured methods for making diag- noses, and systematic assessments of risk factors using standardized in- struments. They generally do not assess core borderline traits such as im- pulsivity; clinical variables such as hopelessness, depressed mood, and hostility; or characteristics of suicidal behavior such as lethal intent and objective planning, which require prospective study designs.
In this study, we used a prospective design to assess demographic, diag- nostic, clinical, and psychosocial variables of High- and Low-Lethality at- tempters with BPD in an effort to define predictors of high-lethality behav- ior within the borderline disorder. Using a stress-diathesis model, we predicted that high-lethality attempts in BPD would be more likely when acute stressors, such as MDD or SUD, hopelessness, and anxiety, which intensify suicidal intent, are combined with high degrees of temperamental impulsivity or aggression. Although it is well understood that attempters are not the same as completers, High-Lethality attempters share many of the characteristics of completers, and may be studied prospectively as sur- rogates for patients who complete suicide (e.g., as “failed suicides”) (Beck, Steer, Kovacs, & Garrison,1975).
METHOD
This study was approved by the University of Pittsburgh Institutional Re- view Board. Subjects were recruited from the inpatient and outpatient ser- vices of the Western Psychiatric Institute and Clinic, and by advertise- ment, from the surrounding community. After describing the study, written informed consent was obtained. Experienced clinical raters diag- nosed Axis I disorders using the Structured Clinical Interview for DSM III- R (SCID; Spitzer, Williams, Gibbon, & First, 1988) and Axis II disorders using the International Personality Disorders Examination (IPDE; Loran- ger, Sussman, Oldham, & Russakoff, 1987). (The DSM III-R was used for purposes of continuity with longitudinal studies.) The Diagnostic Interview for Borderline Patients (DIB; Gunderson, Kolb, & Austin, 1981), provided confirmation of BPD diagnosis. Final diagnoses were determined in a con- sensus conference of raters, using all available data. Patients with psy- chotic disorders, organic mood disorders, or bipolar disorders were ex- cluded. Demographic data, suicide history, childhood and abuse histories, and family and treatment histories were obtained by semi-structured in- terviews, described previously (Soloff et al., 2000; Mann et al., 1999; So- loff, Lynch, & Kelly, 2002).
A suicide attempt was defined as any self-injurious behavior with intent to die. Suicidal ideation and intent were assessed on the Scale for Suicidal Ideation (SSI) and Suicide Intent Scale (SIS), respectively (Beck, Schuyler, & Herman, 1974). A Subjective Intent factor and an Objective Planning factor were derived from the SIS using the method of Mieczkowicz et al. (Mieczkowski et. al., 1993). The Medical Lethality Scale, an ordinal scale quantifying degree of lethality by medical consequences, was scored for each attempt. Descriptive anchors define degrees of medical severity ac- cording to suicide method (Beck, Beck, & Kovacs, 1975). Using the median score on the Medical Lethality Scale, subjects were defined as High-Lethal- ity attempters by a score of 4 or more for any lifetime attempt. (e.g., a score of 4 following a sedative drug overdose is defined as “comatose,” or “requiring hospitalization.”) Low-Lethality attempters had scores of 3 or less. Attempts were also classified according to violence of method (Soloff et al., 2000). Subjects were assessed for current state symptoms, person- ality traits, current social adjustment, reasons for living, and recent life events using standardized measures, described previously (Tables 1–3; Soloff et al., 2000; Mann et al., 1999; Kelly, Soloff, Lynch, Haas, & Mann, 2000).
STATISTICAL METHOD
After testing for homogeneity and normality, univariate differences be- tween High- and Low-Lethality attempters were tested using the Student’s t-test (two-tailed), the Chi-Square, or Mann-Whitney U-tests as appro- priate. Continuous variables were correlated using Pearson or Spearman correlations. Missing data resulted in variable sample sizes for some anal- yses. Factor analyses were used where possible within content categories to reduce the number of intercorrelated variables. Multiple logistic regres- sion was used to identify independent predictors of Lethality status follow- ing the modeling procedures of Hosmer and Lemeshow (Hosmer & Leme- show, 2000). Because it is defined as an ordinal scale, the Medical Lethality Scale was not used as a continuous variable in these analyses. In a first exploratory step, variables discriminating between groups in the univariate analyses, with a significance threshold of p £ .1, were grouped by content category (Tables 1–3), and regressed on Lethality status one category at a time. This exploratory step was conducted without statistical adjustment for multiple comparisons to decrease the possibility of exclud- ing important variables due to Type II error (Hosmer & Lemeshow, 2000). A final multivariate regression model was built using significant predictors from these individual content category regressions, retaining in the final model only those variables significant at p < .05. This model was then chal- lenged by adding back each significant variable from the content category regressions one variable at a time. Odds ratios and 95% confidence inter- vals were calculated for the logistic regression models. Sensitivity and specificity of the final model in identifying High-Lethality attempters were calculated using the standard cutoff of 0.5. All analyses were performed using SPSS for Windows 10.1.0 (2000, Chicago, IL).
TABLE 1. Characteristics of High- vs. Low-Lethality Attempters: Demographic, Diagnostic and Clinical State Variables
| High Leth. | Low Leth. | t, df, p | |||
| Risk Factor | N = 44 | N = 69 | x2, df, p | ||
| 1. Demographic Age: mean (SD) |
31.3(8.8) |
27.5(7.6) |
2.37, 111df, p. = .02 |
||
| Sex (male/female) | 16(m) 28(f) | 16(m) 53(f) | 2.30, 1df, p ns | ||
| Race (Caucasian) | 35(Cau.) | 52(Cau) | 0.27, 1df, p ns | ||
| (Other) | 9(Other) | 17(Other) | |||
| SES (I-III = High) | 19(43.2%) | 43(63.3%) | 3.97, 1df, p < .05 | ||
| (IV + V = Low) | 25(56.8%) | 26(37.7%) | |||
| Ed. (£HS/Trade) | 25(56.8%) | 26(37.7%) | 3.97, 1df, p < .05 | ||
| Married (ever) | 12(45.5%) | 25(36.2%) | 0.95, 1df, p ns | ||
| Children (%yes) | 22(51.2%) | 19(28.4%) | 5.83, 1df, p = .016 | ||
| Religion (%yes) | 37(84.1%) | 47(69.1%) | 3.19, 1df, p ns | ||
| 2. Diagnostic | |||||
| Major Depression | 29(65.9%) | 32(46.4%) | 4.13, 1df, p = .04 | ||
| Substance Abuse | 31(70.5%) | 40(58.0%) | 1.79, 1df, p ns | ||
| Any Axis II | 30(68.2%) | 32(46.4%) | 5.16, 1df, p = .02 | ||
| Any Cluster B | 22(50.0%) | 19(27.5%) | 5.86, 1df, p = .015 | ||
| ASPD | 17(38.6%) | 10(14.5%) | 8.61, 1df, p = .003 | ||
| 3. Clinical State | |||||
| Beck Depression Inv. | 28.8(13.1) | 25.8(13.1) | 1.12, 97df, p ns | ||
| Hamilton Depr.-24 item | 21.9(8.7) | 21.9(7.9) | 0.014, 98df, p ns | ||
| Beck Hopelessness | 11.7 (6.1) | 10.9 (6.4) | MW: z 0.58, p ns | ||
| GAS | 46.6(10.4) | 49.0(13.3) | 0.97, 106df, p ns | ||
MW = Mann Whitney U-test, z-score
Table 2. Suicide History and Personality Variables
| High Leth. | Low Leth. | t, df, p | |||
| Risk Factor | N = 44 | N = 69 | c2, df, p | ||
| 4. Suicide Hx | |||||
| Self-injury | 21 (37.5%) | 35 (62.5%) | 1.11, 1df, p ns | ||
| Age 1st attempt (yrs.) | 17.9(12.4) | 20.3(7.3) | 1.27,111df, p ns | ||
| # lifetime attempts | 4.4 (3.7) | 2.8 (2.7) | 2.77,111df, p < .01 | ||
| SIS intent (max)* | 9.1(3.2) | 7.7(3.9) | 2.15,95df, p < .05 | ||
| SIS plan (max) | 7.0(3.4) | 6.0(3.1) | 1.57,100df, p ns | ||
| SIS total (max) | 16.5(5.9) | 12.1(8.8) | 3.11,106df, p = .002 | ||
| Violent methods | 29(65.9%) | 33(47.8%) | 3.55,1df, p = .06 | ||
| 5. Personality | |||||
| Barratt Impulsiveness | 77.1(9.9) | 73.9(20.4) | 0.82,85df, p ns | ||
| Buss-Durkee Hos. Inv. | 46.(10.1) | 45.7(11.9) | 0.12,93df, p ns | ||
| LHA** | 24.9(7.3) | 26.0(6.0) | 0.86,101df, p ns | ||
| MMPI-Pd | 21.5(5.6) | 21.0(7.4) | 0.38,93df, p ns | ||
| DIB Section Total | 27.4(4.6) | 27.8(5.2) | 0.35,82df, p ns | ||
| BPD criteria | 5.1 (1.5) | 5.4 (1.0) | MW: z 1.05, p ns | ||
| #SPD criteria | 1.4 (1.4) | 1.1 (1.2) | MW: z 1.23, p ns | ||
| IPDE total | 55.7(24.5) | 48.3(20.8) | 1.62,99df, p = 0.1 |
* SIS: Suicide Intent Scale, lifetime maximum score. ** LHA: Brown-Goodwin Lifetime History of Aggression. MW = Mann Whitney U-test, z-score.
Table 3. Childhood and Family History, Social Adjustment and Treatment Variables
| High Leth. | Low Leth. | t, df, p | |||
| Risk Factor | N = 44 | N = 69 | c2, df, p | ||
| 6. Child/Fam. Hx. | |||||
| Sexual Abuse | 15(50%) | 19(43.2%) | 0.33,1df, p ns | ||
| Sep. par. <15y.o.* | 22(91.7%) | 33(89.2%) | 0.10,1df, p ns | ||
| Fam. Hx: Subst. abuse | 32(76.2%) | 36(55.4%) | 4.77,1df, p <.03 | ||
| Fam. Hx. Suicide | 13 (29.5) | 21 (30.4) | 0.01,1df, p ns | ||
| 7. Social Adjustment | |||||
| Reasons For Living | 137.0(48.4) | 146.3(47.9) | 0.86,80df, p ns | ||
| Recent Life Changes | 141.0 (8.5) | 139.8 (7.5) | 0.61,77df, p ns | ||
| Social Adj. Scale** | 1.45 (0.43) | 1.50 (0.41) | 0.408,88df, p ns | ||
| 8. Treatment | |||||
| Age 1st hosp. | 22.8 (8.2) | 22.4 (7.1) | MW: z .018, p ns | ||
| Previous hosps. | 41(93.2%) | 44(65.7%) | 11.2,1df, p = .001 | ||
| #hosp. adms. | 5.7 (5.5) | 4.2 (8.3) | MW: z 2.83, p = .005 | ||
| Time in hosp. (mos.) | 9.3 (16.2) | 2.5 (4.5) | MW: z 2.78, p = .005 | ||
| OPD Psychotherapy | 39(97.5%) | 50(86.2%) | 3.62,1df, p < .1 |
*n = 76 subjects with any childhood separation from either parent.
**Social Adjustment Scale, Self-rated. MW = Mann Whitney U-test, z-score.
RESULTS
There were 113 BPD subjects, 81 females (71.7%) and 32 males (28.3%). Eighty-seven subjects were Caucasian (77%); 26 were minority subjects (23%), predominately African American. The mean (SD) age was 29.0 (8.3) years, with a range of 18 to 50 years. Socioeconomic class (SES), defined by Hollingshead’s (1957, 1958) two-factor classification, placed 54.9% in Class III, 34.4% in Class IV, and 9.7% in Class V. Half of our subjects (54.9%) had at least a high school education; 45.1% had a trade school education or less than 12 years of high school. A majority of subjects had never married (60.2%) and had no children (61.1%). A defined religion was endorsed by 74.3%, the largest groups identifying themselves as Catholic (32%) or Protestant (31%). Fifty-three subjects (46.9%) were recruited as inpatients, 25 (22.1%) as outpatients. Thirty-five subjects (31 %) were not currently in treatment.
Forty-four High-Lethality attempters were compared to 69 Low-Lethality attempters. High-Lethality attempters were significantly older, had less education, and were more likely to have children (see Table 1) There was no interaction between gender (female), responsibility for children, and le- thality status. For group comparisons of SES, the middle Class III (n = 62), was compared to the combined lower Classes IV and V (n = 51). Signifi- cantly more High-Lethality attempters were found in the lower classes and more Low-Lethality attempters in the middle class. There were no signifi- cant differences between groups by gender, race, religion, or patient status at time of recruitment (see Table 1).
Comorbidity with one or more Axis I depressive disorder (MDD, Dysthy- mic Disorder, Depression NOS) were present in 79 subjects (70%). Sixty- eight subjects (50%) had comorbid MDD. High-Lethality attempters had more comorbid MDD compared to Low-Lethality attempters (see Table 1). This finding was specific for MDD, as comorbidity with “any depressive disorder,” or “any Axis I disorder” did not discriminate groups. Anxiety disorders were infrequent (e.g., Panic Disorder [5.9%], Posttraumatic Stress Disorder [PTSD] [4.4%], Social Phobia [2.9%] and Generalized Anxi- ety Disorder [2.9%]) and not related to lethality status.
A current diagnosis of SUD, including alcohol, was found in 71 subjects (62.8%). Substance abusers were compared to 42 subjects (37.2%) with no current SUD. A current diagnosis of SUD did not discriminate High- from Low-Lethality groups; however, a family history of SUD, found in 60.2% of all cases, was significantly more prevalent among High-Lethality attempters.
Axis II comorbidity (in addition to BPD) was noted in 62 subjects (54.9%), a second Cluster B diagnosis (Histrionic, Narcissistic or Antisocial PD) in 41 subjects (36.3%), and ASPD in 27 (23.9%). Comorbidity with an additional Axis II Disorder, a second Cluster B disorder, or ASPD was significantly more prevalent among High-Lethality attempters. Differences between groups for Axis II and Cluster B disorders were rendered nonsig- nificant by controlling for ASPD.
Severity of BPD, assessed by IPDE, or DIB Section Score Total, was not significantly related to number of attempts or lethality status. Schizotypal PD symptoms (one or more IPDE-SPD criteria) were found in 61.0% of BPD attempters, but were not related to lethality status. Similarly, a general measure of PD severity, (the IPDE total dimensional score), and comorbid Cluster B severity (IPDE dimensional scores for three Cluster B disorders), did not significantly discriminate groups.
Severity of depressive mood, hopelessness, and global function, as- sessed as state variables at the time of study intake, were not related to lethality status (see Table 1). However, each of these state variables had a modest relationship to the lifetime number of suicide attempts. (HamD-24: r = +0.22, n = 100, p. < .05; BDI: r = +0.30, n = 92, p < .01; Hopelessness: r = +0.37, n = 105, p < .001; GAS: r = -0.26, n = 108, p < .01). State anxiety symptoms (BPRS) were not significantly related to lethality status.
The mean (SD) age at first suicide attempt was 19.6 (9.8) years, with no significant difference between High- and Low-Lethality attempters (see Table 2), and no significant relationship to lifetime number of attempts. High-Lethality attempters had a greater lifetime number of attempts than Low-Lethality attempters. Suicidal ideation (SSI) did not discriminate be- tween groups. Intent to die (SIS) was significantly greater among High- Lethality attempters for both the most recent attempt and lifetime most serious attempt. Within the two-factor structure of the SIS, Subjective In- tent, but not Objective Planning, discriminated High- from Low-Lethality attempters. High-Lethality attempters tended to choose violent methods more often than Low-Lethality attempters (p = .06). A history of (nonsui- cidal) self-injurious behavior (SIB) did not discriminate High- from Low- Lethality attempters.
Impulsivity (BIS), aggression (LHA), state hostility (BDHI), and antisocial trait (MMPI-Pd) failed to discriminate High- from Low-Lethality attempters. Among these measures, only antisocial trait (MMPI-Pd) was significantly related to the lifetime number of suicide attempts (r = 0.36, n = 72, p.002). Our subjects reported high rates of adverse childhood experiences, in- cluding physical abuse (42.9%), sexual abuse (40.7 %), prolonged separa- tion (>6 mos) from either parent during childhood (56.7%), and separation or divorce of parents (44.9%). None of these adverse early life experiences was related to lethality status. Similarly, a family history of suicide, re- ported by 30% of subjects, did not discriminate High- from Low-Lethality attempters.
A majority of subjects (70.3%) were living in households with other peo- ple, including parents, other relatives, and friends, while 29.7% lived alone. At the time of assessment, only 15.9% were living with a spouse or conjugal partner. Measures of social adjustment, including Reasons for Living (RFL), Recent Life Changes and the Social Adjustment Scale, failed to discriminate High- from Low-Lethality attempters. Fewer Reasons for Living were associated with greater lifetime number of suicide attempts (r =-0.30, n = 82, p < .01; see Table 3).
While less than half of our subjects were recruited in the hospital, 75.2% had a history of prior hospitalization, and 78.8% had prior outpatient psy- chotherapy. Age at first hospitalization was not related to lethality status; however, High-Lethality attempters were more likely to have extensive treatment histories (i.e., prior hospitalization, more frequent hospitaliza- tions, and greater cumulative time spent in the hospital). As expected, number of hospital admissions was related to lifetime number of suicide attempts (+0.34, n = 93, p = .001).
MULTIVARIATE REGRESSION ANALYSES
Our multivariate regression method required complete data on all vari- ables; therefore, the sample size was reduced to 108 (of 113 attempters). Risk factors most associated with High-Lethality status include: Low SES (O.R. 2.01, 95% CI: 1.01–4.01), co-morbidity with ASPD (O.R. 3.66, 95%
CI: 1.27–10.52), the Treatment Factor (O.R. 2.22, 95% CI: 1.25–3.96), and a high score on the Suicide Intent Scale (i.e., lifetime maximum, O.R.1.12, 95% CI: 1.04–1.21). The Suicide Intent Scale displaced Major Depression in the final model. These variables explain 36% of variance, (Nagelkerke R-Square = 0.36) and identify High-Lethality attempters with a sensitivity of 80.0% and specificity of 65.1%.
DISCUSSION
Characteristics of High-Lethality attempters in our study resemble many of the risk factors of BPD (or Cluster B) patients who complete suicide (Paris & Zweig-Frank, 2001; Runeson & Beskow, 1991; Heikkinen et al., 1997; Kjelsberg et al., 1991; Stone, 1989; Kullgren, 1988). These risk fac- tors are not specific to BPD, or related to severity of the borderline diagno- sis or associated symptoms. Consistent with the stress-diathesis model, we found that suicide intent, which is acutely increased by stressors such as comorbid MDD, predicted High- Lethality status. However, there was no significant relationship between lethality and the borderline patients’ temperamental impulsivity or aggression. Instead, High-Lethality attempt- ers with BPD are characterized by many of the same demographic, diag- nostic, clinical, and psychosocial variables that characterize completers in other high-risk groups.
RISK FACTORS FOR HIGH LETHALITY IN BPD
Low SES. Low SES, unemployment, and financial problems, are associ- ated with suicide completion in some (Runeson & Beskow, 1991; Heikki- nen et al., 1997), but not all (Paris & Zweig-Frank, 2001), studies of com- pleted suicide in BPD or Cluster B patients. Lower levels of personal income and fewer years of education are also associated with suicide com- pletion in nonclinical population studies (Angst & Clayton, 1998), and with attempt behavior across diagnostic categories among psychiatric inpa- tients (Mann et al., 1999), suggesting that low SES is a nonspecific stressor for suicide. The St. Louis Health Study, a community survey of personality disorders, found that subjects with BPD have lower educa- tional achievement and less employment than subjects with other PDs, including ASPD (Cloninger, Bayon, & Przybeck, 1997). Early onset and chronic course of symptoms in BPD result in social and vocational disabil- ities which may contribute to suicide risk in the long term.
Older Age. Borderline patients with high lethality attempts and suicide completion are older than those in BPD comparison groups (Rich & Rune- son, 1992; Soloff et. al., 1994a; Shearer et al., 1988), but younger than sui- cides in the general population. A Finnish study of suicide, which carefully examined PD diagnoses, found that completers with a PD were younger (at 38.3 years) than victims with no PD (at 48.4 years; Isomesta et al., 1996). The mean age at suicide completion was 37.3 years among BPD patients treated in a Montreal general hospital, and followed up to 27 years (Paris & Zweig-Frank, 2001). Younger BPD patients accounted for numer- ous Low-Lethality attempts, as impulsive communicative gestures, while older BPD patients completed suicide.
Prior Suicide Attempts. A history of prior attempts is a robust predictor of future attempts or completion in adults with MDD (Leon, Friedman, Sweeney, Brown, & Mann, 1990; Fawcett et al., 1990), BPD, or Cluster B personality disorders, (Soloff et al., 1994a; Shearer et al., 1988; Isomesta et al., 1996), or adult psychiatric inpatients independent of diagnoses (Mann et al., 1999). Among inpatients with MDD, medical lethality of a recurrent attempt is higher than lethality of an index attempt, suggesting progres- sion toward completion (Malone et al., 1995). In BPD patients, high life- time number of attempts is strongly correlated with lethality, increasing the likelihood of a High-Lethality event. Our current data suggest a step- wise progression in lethality (Soloff et al., unpublished data).
Comorbid Depression. MDD, and other affective disorders, are well- established risk factors for suicide, and frequently comorbid with BPD (Malone et al., 1995; Hirschfeld & Davidson, 1988). Borderline patients with comorbid MDD have risk factors for suicide from both disorders, in- cluding severity of depressed mood and hopelessness related to MDD alone, and impulsive-aggression related to BPD alone. They have more life- time attempts, and more objective planning for suicide than either disor- der alone (Soloff et al., 2000).
SUD. A family history of substance abuse was associated with High- Lethality status, consistent with previous reports among BPD attempters, (Shearer et al., 1988) and completers (Rich & Runeson, 1992; Runeson & Beskow, 1991). The relevance of a family history of substance abuse for BPD probands may lie in shared genetic influences (e.g., impulsivity, anti- sociality), or psychosocial consequences, or both. A comorbid diagnosis of SUD was not related to lethality in our BPD sample. The co-occurence of affective disorder and substance abuse may be a particularly lethal combi- nation, associated with a higher rate of serious attempts, (Fyer et al., 1988) and lethality in patients with BPD (Stone, 1989) or Cluster B disorders (Isomesta et al., 1996). Worsening of Major Depression and Substance Use Disorders in the prior month is a significant predictor of suicide attempt among patients with PDs, including BPD (Yen et al., 2003). Substance use may increase suicide intent associated with depression.
ASPD. Antisocial behavior is associated with impulsivity, aggression, and suicide independent of diagnostic context (Plutchik & Van Praag, 1997). A prospective, longitudinal study of men tested at the time of mili- tary conscription found that antisocial personality traits were associated with increased rates of suicide later in life (Angst & Clayton, 1998). In the borderline patient, ASPD or antisocial traits increase the risk of attempts and completion (Rich & Runeson, 1992; Runeson & Beskow, 1991; Kull- gren, 1988; Kjelsberg et al., 1991). ASPD is the strongest predictor of High- Lethality status in our study, increasing threefold the risk of high lethality attempts. The strong association between ASPD and suicidal behavior may be due to the synergy of multiple risk factors associated with this disorder (i.e., impulsivity and aggression, substance use, and social and vocational consequences of a lifetime of antisocial behavior).
Clinical State. Intent to die (SIS, lifetime max.) was a stronger predictor of High-Lethality status than MDD in our multivariate analysis. It has long been reported that suicide intent is greater in completers than attempters, and in High- versus Low-Lethality attempters across diagnoses (Beck et al., 1974; Goldney, 1981; Michel, 1987; Hamdi, Amin, & Mattas, 1991). In the stress-diathesis model of suicide, MDD increases suicidal intent. Subjective intent (SIS), not objective planning, discriminated High- from Low-Lethality attempters in our BPD sample. Since objective planning, not subjective intent, is predictive of actual medical damage in high-risk disor- ders such as MDD, clinicians often discount the borderline patients’ ex- pressions of suicidal intent as a “cry for help.” Suicidal behavior in BPD often occurs on impulse, in the context of anger, and without objective planning. The combination of impulsivity and subjective intent to die in- creases the risk of High-Lethality attempts.
Treatment History. A long history of hospitalizations and treatment ef- forts reflects the chronicity of the suicide history and may be the predictor of High Lethality most characteristic of borderline patients. As Gunderson and Ridolfi have noted, recurrent suicidality is the borderline patient’s “be- havioral specialty” (Gunderson & Ridolfi, 2001). Multiple and long hospi- talizations also suggest relative refractoriness to treatment. Paris and Zweig-Frank (2001) suggest that suicide completion in BPD is the end re- sult of a chronic process that involves persistent or recurrent comorbid affective disorders, years of unsuccessful treatment, and progressive loss of family and social supports. After years of treatment effort, BPD patients may distance themselves from professional help before completing suicide. Patients with BPD or a Cluster B disorder who complete suicide have less treatment contacts in the months prior to suicide than BPD controls (or suicides without a PD) (Kjelsberg et al., 1991, Isometsa et al., 1996).
LIMITATIONS
By design, we did not include cross-diagnostic comparison groups. Risk factors that discriminate BPD attempters from attempters with other dis- orders may not predict High Lethality within a BPD sample. Impulsivity is an important example. In cross-diagnostic studies, impulsivity predicts the lifetime number of suicide attempts, and discriminates attempters from nonattempters independent of diagnosis (Soloff et al., 2000; Mann et al., 1999). Within a BPD sample, impulsivity is highly prevalent, and does not discriminate High- from Low-Lethality attempters. We found no rela- tionship between impulsivity and medical lethality in BPD across two sep- arate inpatient studies (Soloff et al., 2000; Soloff et al., 1994a).
Characteristics that discriminate BPD attempters from BPD nonat- tempters increase the likelihood of suicidal behavior but may not predict medical lethality. Suicide attempts in BPD are associated with stressors such as depressed mood and hopelessness, poor social adjustment, and few reasons for living (Soloff et al., 2000, Kelly et al., 2000, Rietdijk, van den Bosch, Verheul, Koeter, & van den Brink, 2001). A childhood history of sexual abuse increases ten-fold the risk of suicide attempts in BPD (So- loff et al., 2002). Similarly, a history of self-injurious behavior in BPD may contribute a diathesis to suicide attempts (Stone, 1989; Soloff et al., 1994b; Stanley, Gameroff, Michalson, & Mann, 2001). While clinically im- portant, none of these risk factors were associated with High Lethality in our borderline patients.
Risk factors for suicidal behavior are not specific by diagnosis, though some are more closely associated with one disorder compared to others. For example, hopelessness is associated with suicide completion among patients with MDD (Fawcett et al., 1990; Beck et al., 1985; Beck, Brown, Berchick, Stewart, & Steer, 1990), with intent to die among inpatients with BPD, MDD, or MDD+BPD (Soloff et al., 2000). Although BPD attempters endorse more hopelessness than nonattempters, hopelessness was not re- lated to lethality of attempts in our study. Similarly, panic attacks and severe anxiety, are associated with suicide completion in MDD, (Fawcett et al., 1990), but are not related to lethality in BPD subjects. A lethal out- come depends on the synergy of multiple risk factors (e.g., the depressed borderline patient with hopelessness or severe anxiety is at higher risk of completion).
The risk of high lethality attempts in BPD is related to older age. At a mean age of 29 years, our borderline subjects may not have reached the age of highest risk for suicide completion. Many are currently symptomatic with recurrent suicidal behavior and affective comorbidity, and involved in active treatment efforts. A majority (70.3%) still reside with other people. Risk factors for High-Lethality attempts may change as the sample ages. However, based on prior studies, we expect the social and vocational dis- abilities of this chronic disorder to remain relevant to suicidal outcome over time (Heikkinen et al., 1997; Paris & Zweig-Frank, 2001).
REFERENCES
Angst, J., & Clayton, P. J. (1998). Personal- psychiatric outpatients. American Jour-
ity, smoking and suicide: Aprospective nal of Psychiatry, 147, 190–195.
study. Affective Disorders, 51(1), 55–62. Beck, A. T., Schuyler, D., & Herman, I.
Beck, A. T., Beck, R., & Kovacs, M. (1975). (1974). Development of suicidal intent
Classification of suicidal behavior: I: scales. In A. T. Beck, H. L. P. Resnick,
Quantifying intent and medical lethal- & D. Lettiem (Eds.), The prediction of
ity. American Journal of Psychiatry, suicide (pp. 45–56). Bowie, MD: Char-
132, 285–287. les Press.
Beck, A. T., Brown, G., Berchick, R. J., Stew- Beck, A. T., Steer, R. A., Kovacs, M., & Garri-
art, B. L., & Steer, R. A. (1990). Rela- son, B. (1985). Hopelessness and
tionship between hopelessness and eventual suicide: A 10 year prospective
ultimate suicide: A replication with study of patients hospitalized with sui-
cidal ideation. American Journal of Hirschfeld, R. M. A., & Davidson, L. (1988).
Psychiatry, 142, 559–563. Risk factors for suicide. In A. J. Fran-
Brodsky, B. S., Malone, K. M., Ellis, S. P., ces & R. E. Hales (Eds.), American Psy-
Dulit, R. A., & Mann, J. J. (1997). chiatric Press review of psychiatry (Vol.
Characteristics of Borderline Personal- 7, pp. 307–333). Washington, DC:
ity Disorder associated with suicidal American Psychiatric Press.
behavior. American Journal of Psychia- Hollingshead, A. B. (1957). Two-factor index
try, 154, 1715–1719. of social position. New Haven, CT: Au-
Cloninger, C. R., Bayon, C., & Przybeck, T. R. thor.
(1997). Epidemiology and Axis I co- Hollingshead, A. B., & Redlich, F. C. (1958).
morbidity of antisocial personality. In Social class and mental illness. New
- M. Stoff, J. Breiling, & J. D. Maser York: John Wiley & Sons.
(Eds.), Handbook of antisocial behavior Hosmer, D. W, & Lemeshow, S. (2000). Ap-
(pp. 12–21). New York: Wiley and Sons. plied logistic regression (2nd ed.). New
Corbitt, E. M., Malone, K. M., Haas, G. L., & York: John Wiley and Sons.
Mann, J. J. (1996). Suicidal behavior Isomesta, E. T., Henriksson, M. M., Heikkinen,
in patients with Major Depression and M. E., Aro, H. M., Marttunen, M. J., Ku-
Comorbid Personality Disorders. Jour- oppasalmi, K. I., & Lonnqvist, J. K.
nal of Affective Disorders, 39, 61–72. (1996). Suicide among subjects with
Fawcett, J., Scheftner, W. A., Fogg, L., Clark, personality disorders. American Jour-
- C., Young, M. A., Hedecker, D., & nal of Psychiatry, 153, 667–673.
Gibbons, R. (1990). Time-related pre- Kelly, T. M., Soloff, P. H., Lynch, K. G., Haas,
dictors of suicide in major affective G. L., & Mann, J. J. (2000). Recent life
disorders. American Journal of Psychi- events, social adjustment, and suicide
atry, 147, 1189–1194. attempts in patients with Major De-
Friedman, R. C., Aronoff, M. S., Clarkin, J. F., pression and Borderline Personality
Corn, R., & Hurst, S. W. (1983). His- Disorder. Journal of Personality Disor-
tory of suicidal behavior in depressed ders, 14(4), 316–326.
borderline inpatients. American Jour- Kjelsberg, E., Eikeseth, P. H., & Dahl, A. A.
nal of Psychiatry, 140, 1023–1026. (1991). Suicide in borderline patients:
Fyer, M. R., Frances, A. J., Sullivan, T., Hurt, Predictive factors. Acta Psychiatrica
S.W., & Clarkin, J. (1988). Suicide at- Scandinavica, 84, 283–287.
tempts in patients with Borderline Per- Kullgren, G. (1988). Factors associated with
sonality Disorder. American Journal of completed suicide in Borderline Per-
Psychiatry, 145, 737–739. sonality Disorder. Journal of Nervous
Goldney, R. D. (1981). Attempted suicide in and Mental Disorders, 176, 40–44.
young women. Correlates of lethality. Leon, A. C., Friedman, R. A., Sweeney, J. A.,
British Journal of Psychiatry, 139, Brown, R. P., & Mann, J. J. (1990).
382–390. Statistical issues in the identification
Gunderson, J. G., Kolb, J. E., & Austin, V. of risk factors for suicidal behavior:
(1981). The Diagnostic Interview for The application of survival analysis.
Borderlines. American Journal of Psy- Psychiatry Research, 31, 99–108.
chiatry, 138, 896–903. Loranger, A. W., Susman, V. L., Oldham,
Gunderson, J. G., & Ridolfi, M. E. (2001). J. M., & Russakoff, L. M. (1987). The
Borderline Personality Disorder: Sui- Personality Disorder Examination: A
cidality and self-mutilation. Annals of preliminary report. Journal of Person-
the New York Academy of Sciences, ality Disorders, 1, 1–3.
932, 61–77. Malone, K. M., Haas, G. L., Sweeney, J. A., &
Hamdi, E., Amin, Y., & Mattas, T. (1991). Mann, J. J. (1995). Major depression
Clinical correlates of intent in at- and the risk of attempted suicide. Jour-
tempted suicide. Acta Psychiatrica nal of Affective Disorders 4, 173–185.
Scandinavica, 83, 406–411. Mann, J. J., Waternaux, C., Haas, G. L., &
Heikkinen, M. E., Isometsa, E. T., Henriks- Malone, K. M. (1999). Toward a clinical
son, M. M., Marttunen, M. J., Aro, model of suicidal behavior in psychiat-
- M., & Lonnqvist, J. K. (1997). Psy- ric patients. American Journal of Psy-
chosocial factors and completed sui- chiatry, 156, 181–189.
cide in personality disorders. Acta Psy- Mehlum, L., Friis, S., Vaglum, P., & Kar-
chiatrica Scandinavica, 95, 49–57. terus, S. (1994). The longitudinal pat-
tern of suicidal behavior in Borderline Soloff, P. H. (2005). Unpublished data.
Personality Disorder: A prospective fol- Soloff, P. H., Lis, J. A., Kelly, T. M., Corne-
low-up study. Acta Psychiatrica Scan- lius, J., & Ulrich, R. (1994a). Risk fac-
dinavica 90, 124–130. tors for suicidal behavior in Borderline
Michel, K. (1987). Suicide risk factors: A Personality Disorder. American Jour-
comparison of suicide attempters with nal of Psychiatry, 151, 1316–1323.
suicide completers. British Journal of Soloff, P. H., Lis, J. A., Kelly, T., Cornelius,
Psychiatry, 150, 78–82. J., & Ulrich, R. (1994b). Self-mutila-
Mieczkowski, T. A., Sweeney, J. A., Haas, G. tion and suicidal behavior in Border-
L., Junker, B. W., Brown, R. P., & line Personality Disorder. Journal of
Mann, J. J. (1993). Factor composition Personality Disorders, 8(4), 257–267.
of the Suicide Intent Scale. Suicide and Soloff, P. H., Lynch, K. G., & Kelly, T. M.
Life Threatening Behavior, 23, 37–45. (2002). Childhood abuse as a risk fac-
Norusis, M. J. (2000). SPSS for Windows tor for suicidal behavior in Borderline
10.1.0. Chicago, Illinois. Personality Disorder. Journal of Per-
Oquendo, M., & Mann, J. J. (2000). The biol- sonality Disorders, 16(3), 201–214.
ogy of impulsivity and suicidality. Psy- Soloff, P. H., Lynch, K. G., Kelly, T. M., Ma-
chiatric Clinics of North America, 23(1), lone, K. M., & Mann, J. J. (2000).
11–25. Characteristics of suicide attempts of
Paris, J., & Zweig-Frank, H. (2001). A 27- patients with major depressive episode
year follow-up of patients with Border- and Borderline Personality Disorder: A
line Personality Disorder. Comprensive comparative study. American Journal
Psychiatry, 42, 782–487. of Psychiatry, 157, 601–608.
Plutchik, R., & Van Praag, H. M. (1997). Sui- Spitzer, R. L., Williams, J. B. W., Gibbon, M.,
cide, impulsivity and antisocial behav- & First, M. B. (1988). Instruction man-
ior. In D. M. Stoff, J. Breiling, & J. D. ual for the Structured Clinical Interview
Maser (Eds.), Handbook of antisocial for DSMIII-R. Biometrics Research De-
behavior (pp. 101–108). New York: partment, New York State Psychiatric
John Wiley & Sons. Inst., 722 West 168th Street, New York,
Rich, C. L., & Runeson, B. S. (1992). Similar- NY. Rev. 4/1/88
ities in diagnostic comorbidity between Stanley, B., Gameroff, M. J., Michalson, V.,
suicide among young people in Swe- & Mann, J. J.(2001). Are suicide at-
den and the United States. Acta Psy- tempters who self-mutilate a unique
chiatrica Scandinavica, 86, 335–339. population? American Journal of Psy-
Rietdijk, E. A., van den Bosch, L. M. C., Ver- chiatry, 158, 427–432.
heul, R., Koeter, W. J., & van den Brink, Stone, M. H. (1989). The course of Borderline
- (2001). Predicting self-damaging Personality Disorder. In A. Tasman,
and suicidal behaviors in female bor- R. E. Hales, & A. Frances (Eds.), Amer-
derline patients: Reasons for living, ican Psychiatric Press review of psychi-
coping and Depressive Personality atry (Vol. 8, pp. 103–125). Washing-
Disorder. Journal of Personality Disor- ton, DC: American Psychiatric Press.
ders 15(6), 512–520. Yen, S., Shea, M. T., Pagano, M., Sanislow,
Runeson, B., & Beskow, J. (1991). Border- C. A., Grilo, C. M., McGlashan, T. H.,
line Personality Disorder in young Skodol, A. E., Bender, D. S., Zanarini,
Swedish suicides. Journal of Mental M. C., Gunderson, J. G., & Morey, L. C.
and Nervous Diseases, 179, 153–156. (2003). Axis I and Axis II disorders as
Shearer, S. L., Peters, C. P., Quaytman, M. S., predictors of prospective suicide at-
& Wadman, B. E. (1988). Intent and le- tempts: Findings from the Collabora-
thality of suicide attempts among fe- tive Longitudinal Personality Disorders
male borderline inpatients. American Study. Journal of Abnormal Psychol-
Journal of Psychiatry, 145, 1424–1427. ogy, 112(3), 375–381.