The investigators of the Australian study also did a preliminary cost-benefit analysis (138) in which they compared the direct cost of treatment for the 12 months preceding psychodynamic therapy with the direct cost of treatment for the 12 months following this therapy. In Australian dollars, the cost of the treatment for all patients decreased from $684,346 to $41,424. Including psychotherapy in the cost of treatment, there was a total savings per patient of $8,431 per year. This cost-effectiveness was accounted for almost entirely by a decrease in the number of hospital days. Without a control group, however, one cannot definitively conclude that the cost savings were the result of the psychotherapy.

d) Length and frequency of treatment

Most clinical reports of psychodynamic psychotherapy involving patients with borderline personality disorder refer to the treatment duration as "extended" or "long term." However, there are only limited data about how much therapy is adequate or optimal. In the aforementioned randomized controlled trial of psychoanalytically focused partial hospitalization treatment (9), the effect of psychotherapy on reducing hospitalization was not significant until after the patients had been in therapy for more than 12 months. There are no studies demonstrating that brief therapy or psychotherapy less than twice a week is helpful for patients with borderline personality disorder. Howard and colleagues (142), to study the psychotherapeutic dose-effect relationship, conducted a meta-analysis comprising 2,431 subjects from 15 patient groups spanning 30 years. One study they examined in detail involved a group of 151 patients evaluated by self-report and by chart review; 28 of these patients had a borderline personality disorder diagnosis. Whereas 50% of patients with anxiety or depression improved in eight to 13 sessions, the same degree of improvement occurred after 13�26 sessions for "borderline psychotic" patients according to self-ratings (the same degree of improvement occurred after 26�52 sessions according to chart ratings by researchers [143]). Seventy-five percent of patients with borderline personality disorder had improved by 1 year (52 sessions) and 87%�95% by 2 years (104 sessions). While this study confirms the conventional wisdom that more therapy is needed for patients with borderline personality disorder than for patients with an axis I disorder, it is unclear whether raters were blind to diagnosis. It appears that a standardized diagnostic assessment and standard threshold for improvement were not used, there are no data on treatment dropouts, and little information is provided about the type of therapy or the therapists except that they were predominantly psychodynamically oriented. What can be concluded is that in a naturalistic setting, outpatients who are clinically diagnosed as "borderline psychotic" will likely need more extended therapy than will depressed or anxious patients.

e) Adverse effects

While no adverse effects were reported in the aforementioned studies, psychodynamic psychotherapy has the potential to disorganize some patients if the focus is too exploratory or if there is too much emphasis on transference without an adequately strong alliance. Intensive dynamic psychotherapy may also activate strong dependency wishes in the patient as transference wishes and feelings develop in the context of the treatment. It is the exploration of such dependency that is often essential to help the patient to achieve independence. This dependence may elicit countertransference problems in the therapist, which can lead to inappropriate or ineffective treatment. The most serious examples of this include unnecessary increases in the frequency or duration of treatment or transgression of professional boundaries.

f) Implementation issues

i) Difficulties with adherence

Most studies report a high dropout rate from dynamic psychotherapy among patients with borderline personality disorder. However, this is true for almost all approaches to the treatment of these patients, and it has not been demonstrated to be any higher for dynamic therapy. It does, however, emphasize the paramount importance of adequate attention to the therapeutic alliance as well as to transference and countertransference issues.

ii) Need for therapist flexibility

Early in the treatment, and periodically in the later stages, a therapist who is also functioning as primary clinician may need to take a major role in management issues, including limit-setting, attending to suicidality, addressing pharmacotherapy, and helping to arrange hospitalization. A stance in which the therapist only explores the patient�s internal experience and does not become involved in management of life issues may lead to adverse outcomes for some patients.

iii) Importance of judicious transference interpretation

Excessive transference interpretation or confrontation early in treatment may increase the risk that the patient will drop out of therapy. One process study of psychoanalytic therapy with patients with borderline personality disorder (11) found that for some patients, transference interpretation is a "high-risk, high-gain" phenomenon in that it may improve the therapeutic alliance but also may cause substantial deterioration in that alliance. Therapists must use transference interpretation judiciously on the basis of their sense of the state of the alliance and the patient�s capacity to hear and reflect on observations about the therapeutic relationship. A series of empathic and supportive comments often paves the way for an effective transference interpretation. Other patients may be able to use transference interpretation effectively without this much preparatory work.

iv) Role of therapist training and competency

Psychodynamic therapy for patients with borderline personality disorder is uncommonly demanding. Consultation from an experienced colleague is highly recommended for all therapists during the course of the therapy. In some situations, personal psychotherapy can help the clinician develop skills to manage the intense transference/countertransference interactions that are characteristic of these treatments.

2. Cognitive behavior therapy

a) Definition and goals

Although cognitive behavior therapy has been widely used and described in the clinical literature, it has more often been used to treat axis I conditions (e.g., anxiety or depressive disorders) than personality disorders. Cognitive behavior therapy assumes that maladaptive and distorted beliefs and cognitive processes underlie symptoms and dysfunctional affect or behavior and that these beliefs are behaviorally reinforced. It generally involves attention to a set of dysfunctional automatic thoughts or deeply ingrained belief systems (often referred to as schemas), along with learning and practicing new, nonmaladaptive behaviors. Utilization of cognitive behavior methods in the treatment of the personality disorders has been described (19), but because persistent dysfunctional belief systems in patients with personality disorders are usually "structuralized" (i.e., built into the patient�s usual cognitive organization), substantial time and effort are required to produce lasting change. Modifications of standard approaches (e.g., schema-focused cognitive therapy, complex cognitive therapy, or dialectical behavior therapy) are often recommended in treating certain features typical of the personality disorders. However, other than dialectical behavior therapy (17, 144�147), these modifications have not been studied.

b) Efficacy

Most published reports of cognitive behavior treatment for patients with borderline personality disorder are uncontrolled clinical or single case studies. Recently, however, several controlled studies have been done, particularly of a form of cognitive behavior therapy called dialectical behavior therapy. Dialectical behavior therapy consists of approximately 1 year of manual-guided therapy (involving 1 hour of weekly individual therapy for 1 year and 2.5 hours of group skills training per week for either 6 or 12 months) along with a requirement for all therapists in a study or program to meet weekly as a group. Linehan and colleagues (8) reported a randomized controlled trial of dialectical behavior therapy involving patients with borderline personality disorder whose symptoms included "parasuicidal" behavior (defined as any intentional acute self-injurious behavior with or without suicide intent). Control subjects in this study received "treatment as usual" (defined as "alternative therapy referrals, usually by the original referral source, from which they could choose"). Of the 44 study completers, 22 received dialectical behavior therapy, and 22 received treatment as usual; patients were assessed at 4, 8, and 12 months. At pretreatment, 13 of the control subjects had been receiving individual psychotherapy, and nine had not. Patients who received dialectical behavior therapy had less parasuicidal behavior, reduced medical risk due to parasuicidal acts, fewer hospital admissions, fewer psychiatric hospital days, and a greater capacity to stay with the same therapist than did the control subjects. Both groups improved with respect to depression, suicidal ideation, hopelessness, or reasons for living; there were no group differences on these variables. Because there were substantial dropout rates overall (30%) and the number of study completers in each group was small, it is unclear how generalizable these results are. Nonetheless, this study is a promising first report of a manualized regimen of cognitive behavior treatment for a specific type of patient with borderline personality disorder.

A second cohort of patients was subsequently studied; the same study design was used (148). In this report, there were 26 intent-to-treat patients (13 received dialectical behavior therapy, 13 received treatment as usual). One patient who received dialectical behavior therapy committed suicide late in the study, and three patients receiving dialectical behavior therapy and one patient receiving treatment as usual dropped out. Nine of the 13 control patients were already receiving individual psychotherapy at the beginning of the study or entered such treatment during the study. Patients who received dialectical behavior therapy had greater reduction in trait anger and greater improvement in Global Assessment Scale scores.

One year after termination of their previously described study (8), the Linehan group reevaluated their patient group (5). After 1 year, the greater reduction in parasuicide rates and in severity of suicide attempts seen in the dialectical behavior therapy group relative to the control subjects did not persist, although there were significantly fewer psychiatric hospital days for the dialectical behavior therapy group during the follow-up year. These findings suggest that although dialectical behavior therapy produces a greater reduction in parasuicidal behavior than treatment as usual, the durability of this advantage is unclear.

In a subsequent report, Linehan and colleagues (149) compared dialectical behavior therapy with treatment as usual in patients with borderline personality disorder with drug dependence. Only 18 of the 28 intent-to-treat patients completed the study (seven who received dialectical behavior therapy, 11 given treatment as usual). Patients receiving dialectical behavior therapy had more drug- and alcohol-abstinent days after 4, 8, and 16 months. All patients had reduced parasuicidal behavior as well as state and trait anger; there was no difference between the groups. This study, too, involved small numbers of patients and had substantial dropout rates, but it represents an important attempt to evaluate the impact of dialectical behavior therapy with severely ill patients with borderline personality disorder and comorbid substance abuse.

In all of these studies, it is difficult to ascertain whether the improvement reported for patients receiving dialectical behavior therapy derived from specific ingredients of dialectical behavior therapy or whether nonspecific factors such as either the greater time spent with the patients or therapist bias contributed to the results. In a small study in which skills training alone was compared with a no-skills training control condition, no difference was found between the groups (unpublished 1993 study of MM Linehan and HL Heard). The researchers concluded that the specific features of individual dialectical behavior therapy are necessary for patients to show greater improvement than control groups. Linehan and Heard (150) reported that more time with therapists does not account for improved outcome. Nonetheless, other special features of dialectical behavior therapy, such as the requirement for all therapists to meet weekly as a group, could contribute to the results.

Springer et al. (151) used an inpatient group therapy version of dialectical behavior therapy for patients with personality disorders, 13 of whom had borderline personality disorder. The patients with borderline personality disorder exhibited improvement in depression, hopelessness, and suicidal ideation, but the improvement was not greater than it was for a control group. In this study, compared with control subjects, patients receiving the dialectical behavior therapy treatment showed a paradoxical increase in parasuicidal acting out during the brief hospitalization (average length of stay was 12.6 days).

Barley and colleagues (152) compared dialectical behavior therapy received by patients with borderline personality disorder on a specialized personality disorder inpatient unit with treatment as usual on a similar-sized inpatient unit. They found that the use of dialectical behavior therapy was associated with reduced parasuicidal behavior. It is unclear whether improvement was due to dialectical behavior therapy per se or to other elements of the specialized unit.

Perris (153) reported preliminary findings from a small uncontrolled, naturalistic follow-up study of 13 patients with borderline personality disorder who received cognitive behavior therapy similar to dialectical behavior therapy. Twelve patients were evaluated at a 2-year follow-up point, and all patients maintained the normalization of functioning that had been evident at the end of the study treatment.

Other controlled studies reported in the literature of cognitive behavior approaches are difficult to interpret because of small patient group sizes or because the studies focused on mixed types of personality disorders without specifying borderline cohorts (154�156).

In summary, there are a number of studies in the literature suggesting that cognitive behavior therapy approaches may be effective for patients with borderline personality disorder. Most of these studies involved dialectical behavior therapy and were carried out by Linehan and her group. Replication studies by other groups in other centers are needed to confirm the validity and generalizability of these findings.

c) Cost-effectiveness

Published data are not available on the cost-effectiveness of cognitive behavior approaches for treatment of borderline personality disorder, although Linehan and colleagues (8) reported that patients receiving dialectical behavior therapy had fewer psychiatric inpatient days and psychiatric hospital admissions than did control subjects.

d) Length and frequency of treatment

Short-term cognitive therapy involving 16�20 sessions has been described as a generic treatment approach; however, the patient characteristics thought to be necessary for a successful treatment outcome are not typical of patients with personality disorders (147). Instead, longer forms of treatment, such as "schema-focused cognitive therapy" (147), "complex cognitive therapy" (144), or dialectical behavior therapy (17), are usually recommended.

The standard length of dialectical behavior therapy is approximately 1 year for the most commonly administered phase of the treatment. It involves 1 hour of individual therapy per week, more than 2 hours of group skills training per week (for either 6 or 12 months), and 1 hour of group process for the therapists per week. Other versions of dialectical behavior therapy, such as that administered in a brief inpatient setting (151), may be useful but are not necessarily more effective than other forms of inpatient treatment.

e) Adverse effects

Although there are no reports of adverse effects of cognitive behavior therapy, including dialectical behavior therapy, as administered on an outpatient basis, one inpatient study (151) reported a paradoxical increase in parasuicidal acting out in the dialectical behavior therapy group compared with the control group�a finding thought perhaps to be due to the contagion effect within a closed, intensive milieu.

f) Implementation issues

Many components of cognitive behavior therapy are similar to elements of psychodynamic psychotherapy, although they may have different labels. For example, as Linehan (17) pointed out, focusing on "therapy-interfering behavior" is similar to the psychodynamic emphasis on transference behaviors. Similarly, the notion of validation resembles that of empathy. Beck and Freeman (19) noted that cognitive therapists and psychoanalysts have the common goal of identifying and modifying "core" personality disorder problems. However, psychodynamic therapists view these core problems as having important unconscious roots that are not available to the patient, whereas cognitive therapists view them as largely in the realm of awareness. It is not clear how successfully psychiatrists who have not been trained in cognitive behavior therapy can implement manual-based cognitive behavior approaches.

Although dialectical behavior therapy has been well described in the literature for many years, it is not clear how difficult it is to teach to new therapists in settings other than that where it was developed. Variable results in other settings could be due to a number of factors, such as less enthusiasm for the method among therapists, differences in therapist training in dialectical behavior therapy, and different patient populations. Although the Linehan group has developed training programs for therapists, certain characteristics recommended in dialectical behavior therapy (e.g., "a matter-of-fact, somewhat irreverent, and at times outrageous attitude about current and previous parasuicidal and other dysfunctional behaviors" [17]) may be more effective when carried out by therapists who are comfortable with this particular style.

3. Group therapy

a) Goals

The goals of group therapy are consistent with those of individual psychotherapy and include stabilization of the patient, management of impulsiveness and other symptoms, and examination and management of transference and countertransference reactions. Groups provide special opportunities for provision of additional social support, interpersonal learning, and diffusion of the intensity of transference issues through interaction with other group members and the therapists. In addition, the presence of other patients provides opportunities for patient-based limit setting and for altruistic interactions in which patients can consolidate their gains in the process of helping others.

b) Efficacy

Some uncontrolled studies suggest that group treatment (157), including process-focused groups in a therapeutic community setting (158), may be helpful for patients with borderline personality disorder. However, these studies had no true control condition, and the efficacy of the group treatment is unclear, given the complexity of the treatment received. Another small chart review study of an "incest group" for patients with borderline personality disorder (159) suggested shorter subsequent inpatient stays and fewer outpatient visits for treated patients than for control subjects. A randomized trial (160) involving patients with borderline personality disorder showed equivalent results with group versus individual dynamically oriented psychotherapy, but the small sample size and high dropout rate make the results inconclusive. Wilberg et al. (161) did a naturalistic follow-up study of two cohorts of patients with borderline personality disorder. This quasi-experimental, nonrandomized study showed that patients with borderline personality disorder discharged from a day program with continuing outpatient group therapy (N=12) did better than those who did not have group therapy (N=31). They had better global health and lower global severity index symptoms, lower Health-Sickness Rating Scale scores, lower SCL-90 scores, lower rehospitalization rates, fewer suicide attempts, and less substance abuse. There were, however, important differences between the two comparison groups that could account for outcome differences.

Perhaps the most interesting aspect of group therapy is the use of groups to consolidate and maintain improvement from the inpatient stay. Linehan and colleagues (8) combined individual and group therapy, making the specific effect of the group component unclear. They reported that, contrary to expectations, the addition of group skills training to individual dialectical behavior therapy did not improve clinical outcome. For those patients with borderline personality disorder who have experienced shame or have become isolated as a result of trauma, including those with comorbid PTSD, group therapy with others who have experienced trauma can be helpful. Such groups provide a milieu in which their current emotional reactions and self-defeating behaviors can be seen and understood. Groups may also provide a context in which patients may initiate healthy risk-taking in relationships. Group treatment has also been included in studies of psychodynamic psychotherapy; although the overall treatment program was effective, the effectiveness of the group therapy component is unknown (9, 162). Clinical wisdom indicates for many patients combined group and individual psychotherapy is more effective than either treatment alone.

c) Cost-effectiveness

Group psychotherapy is substantially less expensive than individual therapy because of the favorable therapist-patient ratio. Marziali and Monroe-Blum (163) calculated that group psychotherapy for borderline personality disorder costs about one-sixth as much as individual psychotherapy, assuming that the fee for individual therapy is only slightly higher than that for group therapy. However, this potential saving is tempered by the fact that most treatment regimens for borderline personality disorder combine group interventions with individual therapy.

d) Length and frequency of treatment

Groups generally meet once a week, although in inpatient settings sessions may occur daily. In some studies, groups are time-limited�for example, 12 weekly sessions�whereas in other studies they continue for a year or more.

e) Adverse effects

Acute distress from exposure to emotionally arousing group issues has been reported. Other potential risks of treating patients with borderline personality disorder in group settings include shared resistance to therapeutic work, hostile or other destructive interactions among patients, intensification of transference problems, and symptom "contagion."

f) Implementation issues

Groups take considerable effort to set up and require a group of patients with similar problems and willingness to participate in group treatment. Patients in group therapy must agree to confidentiality regarding the information shared by other patients and to clear guidelines regarding contact with other members outside the group setting. It is critical that there be no "secrets" and that all interactions among group members be discussed in the group, especially information regarding threats of harm to self or others.

4. Couples therapy

a) Goals

The usual goal of couples therapy is to stabilize and strengthen the relationship between the partners or to clarify the nonviability of the relationship. An alternative or additional goal for some is to educate and clarify for the spouse or partner of the patient with borderline personality disorder the process that is taking place within the relationship. Partners of patients with borderline personality disorder may struggle to accommodate the patient�s alternating patterns of idealization and depreciation as well as other interpersonal behaviors. As a result, spouses may become dysphoric and self-doubting; they may also become overly attentive and exhibit reaction formation. The goal of treatment is to explore and change these maladaptive reactions and problematic interactions between partners.

b) Efficacy

The literature on the effectiveness of couples therapy for patients with borderline personality disorder is limited to clinical experience and case reports. In some cases, the psychopathology and potential mutual interdependence of each partner may serve a homeostatic function (164�166). Improvement can occur in the relationship when there is recognition of the psychological deficits of both parties. The therapeutic task is to provide an environment in which each spouse can develop self-awareness within the context of the relationship.

c) Adverse effects

One report (41) described an escalation of symptoms when traditional marital therapy was used with a couple who both were diagnosed with borderline personality disorder. Clinical experience would indicate the need for careful psychiatric evaluation of the spouse. When severe character pathology is present in both, the clinician will need to use a multidimensional approach, providing a holding environment for both partners while working toward individuation and intrapsychic growth. Because the spouse�s own interpersonal needs or behavioral patterns may, however pathological, serve a homeostatic function within the marriage, couples therapy has the potential to further destabilize the relationship.

d) Implementation issues

At times, it might be helpful for the primary clinician to meet with the spouse or partner and evaluate his or her strengths and weaknesses. It is important to recognize the contingencies of the extent of the partner�s loyalty and his or her understanding of what can be expected from the patient with borderline personality disorder before recommending couples therapy. Couples therapy with patients with borderline personality disorder requires considerable understanding of borderline personality disorder and the attendant problems and compensations that such individuals bring to relationships.

5. Family therapy

a) Goals

Relationships in the families of patients with borderline personality disorder are often turbulent and chaotic. The goal of family therapy is to increase family members� understanding of borderline personality disorder, improve relationships between the patient and family members, and enhance the overall functioning of the family.

b) Efficacy

The published literature on family therapy with patients with borderline personality disorder consists of case reports (167�170) and one published study (12) that found a psychoeducational approach could improve communication, diminish alienation and burden, and diminish conflicts over separation and independence. The clinical literature suggests that family therapy may be useful for some patients� in particular, those who are still dependent on or significantly involved with their families. Some clinicians report the efficacy of dynamically based therapy, whereas others support the efficacy of a psychoeducational approach in which the focus is on educating the family about the diagnosis, improving communication, diminishing hostility and guilt, and diminishing the burden of the illness.

c) Adverse effects

Some clinicians report that traditional dynamically based family therapy has the potential to end prematurely and have a poor outcome, since patients may alienate their family members or leave the treatment themselves because they feel misunderstood (171) when family involvement is indicated. A psychoeducational approach appears to be less likely to have such adverse effects; however, even psychoeducational approaches can upset family members who wish to avoid knowledge about the illness or involvement in the family member�s treatment.

d) Implementation issues

Traditional dynamically based family therapy requires considerable training and sufficient experience with patients with borderline personality disorder to appreciate their problems and conflicts and to be judicious in the selection of appropriate families.

C. Review of Pharmacotherapy and Other Somatic Treatments

1. SSRI antidepressants

a) Goals

In borderline personality disorder, SSRIs are used to treat symptoms of affective dysregulation and impulsive-behavioral dyscontrol, particularly depressed mood, anger, and impulsive aggression, including self-mutilation.

b) Efficacy

Early case reports and small open-label trials with fluoxetine, sertraline, and venlafaxine (a mixed norepinephrine/serotonin reuptake blocker) indicated significant efficacy for symptoms of affective dysregulation, impulsive-behavioral dyscontrol, and cognitive-perceptual difficulties in patients with borderline personality disorder (44�49, 67). Aggression, irritability, depressed mood, and self-mutilation responded to fluoxetine (up to 80 mg/day), venlafaxine (up to 400 mg/day), or sertraline (up to 200 mg/day) in trials of 8�12 weeks (45). An unexpected finding in some of these early reports was that improvement in impulsive behavior appeared rapidly, often within the first week of treatment, and disappeared as quickly with discontinuation or nonadherence. Improvement in impulsive aggression appeared to be independent of effects on depression and anxiety and occurred whether or not the patient had comorbid major depressive disorder (67). Nonresponse to one SSRI did not predict poor response to all SSRIs. For example, some patients who did not respond to fluoxetine, 80 mg/day, responded to a subsequent trial of sertraline. Similarly, patients who did not respond to sertraline, paroxetine, or fluoxetine subsequently responded to venlafaxine. In one study, higher doses and a longer trial (24 weeks) with sertraline converted half of sertraline nonresponders to responders (45).

Three double-blind, placebo-controlled studies have been conducted. Salzman and colleagues (44) conducted a 12-week trial of fluoxetine (20�60 mg/day) in 27 relatively high-functioning subjects (mean Global Assessment Scale score of 74) with borderline personality disorder or borderline traits. Other axis I or axis II comorbid diagnoses were absent, as were recent suicidal behavior, self-mutilation, substance abuse, and current severe aggressive behavior (i.e., behaviors typical of patients with borderline personality disorder seeking treatment). This strategy diminishes generalizability to more seriously ill patients but has the advantage of allowing for a test of efficacy in the absence of comorbidity. For the 22 subjects who completed the study (13 given fluoxetine and nine who received placebo), significant reduction in symptoms of anger and depression and improvement in global functioning were reported for subjects given fluoxetine compared with those given placebo. Improvement in anger was independent of improvement in depressed mood. Improvement was modest, with no subject improving more than 20% on any measure. In addition, a large placebo response was noted.

Markovitz (45) studied 17 patients (nine given fluoxetine, 80 mg/day, and eight given placebo) for 14 weeks. This patient group was noteworthy for the high rate of comorbid axis I mood disorders (10 with major depression, six with bipolar disorder), anxiety disorders, and somatic complaints (e.g., headaches, premenstrual syndrome, irritable bowel syndrome). While this group is more typical of an impaired borderline personality disorder patient population, comorbidity with affective and anxiety disorders confounds interpretation of results. Patients receiving fluoxetine improved significantly more than those given placebo in depression, anxiety, and global symptoms. Measures of impulsive aggression were not included in this study. Some patients with premenstrual syndrome and headaches noted improvement in these somatic presentations with fluoxetine, whereas none improved with placebo.

A double-blind, placebo-controlled study by Coccaro and Kavoussi (67) focused attention on impulsive aggression as a dimensional construct (i.e., a symptom domain found across personality disorders but especially characteristic of borderline personality disorder). Forty subjects with prominent impulsive aggression in the context of a personality disorder, one-third of whom had borderline personality disorder, participated. There was a high rate of comorbidity with dysthymic disorder or depressive disorder not otherwise specified; subjects with major depression and bipolar disorder were excluded. Anxiety disorders, as well as alcohol and drug abuse, were common. In this 12-week, double-blind, placebo-controlled trial, fluoxetine (20�60 mg/day) was more effective than placebo for treatment of verbal aggression and aggression against objects. Improvement was significant by week 10, with improvement in irritability appearing by week 6. Global improvement, favoring fluoxetine, was significant by week 4. As in the open-label trials and the aforementioned Salzman et al. study (44), these investigators found that the effects on aggression and irritability did not appear as a result of improvement in mood or anxiety symptoms.

In summary, these three randomized, double-blind, placebo-controlled studies show efficacy for fluoxetine for affective symptoms�specifically, depressed mood (44, 45), anger (44), and anxiety (45, 67)�although effects on anger and depressed mood appear quantitatively modest. Efficacy has also been demonstrated for impulsive-behavioral symptoms�specifically, verbal and indirect aggression (67)�and global symptom severity (44, 45, 67). Effects on impulsive aggression (67) and anger (44) were independent of effects on affective symptoms, including depressed mood (44, 67) and anxiety (67). Although the three published double-blind, placebo-controlled trials used fluoxetine, open-label studies and clinical experience suggest potential usefulness for other SSRIs.

c) Side effects

The side effect profile of the SSRIs is favorable compared with that of older tricyclic, heterocyclic, or MAOI antidepressants, including low risk in overdose. Side effects reported in these studies are consistent with routine clinical usage.

d) Implementation issues

The SSRI antidepressants may be used in their customary antidepressant dose ranges and durations (e.g., fluoxetine, 20�80 mg/day; sertraline, 100�200 mg/day). One investigator used very high doses of sertraline (200�600 mg/ day) for nonresponders, with some improved efficacy (45). At these high doses, peripheral tremor was noted. There are no published studies of continuation and maintenance strategies with SSRIs, although anecdotal reports suggest continuation of improvement in impulsive aggression and self-mutilation for up to several years while the medication is taken and rapid return of symptoms upon discontinuation (49, 172, 173). The duration of treatment is therefore a clinical judgment that depends on the patient�s clinical status and medication tolerance at any point in time.

2. Tricyclic and heterocyclic antidepressants

a) Goals

In borderline personality disorder, antidepressants are used for affective dysregulation, manifested most commonly by depressed mood, irritability, and mood lability. Evaluation of antidepressant trials in the treatment of borderline personality disorder must take into account the presence of comorbid axis I mood disorders, which are common in patients with borderline personality disorder. Studies in which there is a preponderance of comorbid axis I depression would be expected to demonstrate a favorable response to antidepressant treatments but may not reflect the pharmacological responsiveness of borderline personality disorder.

b) Efficacy

Double-blind, placebo-controlled trials of tricyclic antidepressants in borderline personality disorder have used amitriptyline, imipramine, and desipramine in both inpatient and outpatient settings. Mianserin, a tetracyclic antidepressant not available in the United States, has been used in an outpatient setting. Most of these studies were parallel comparisons with another medication and placebo. A 5-week inpatient study of patients with borderline personality disorder that compared amitriptyline (mean dose=149 mg/day) with haloperidol and placebo found that amitriptyline decreased depressive symptoms and indirect hostility and enhanced attitudes about self-control compared with placebo (51). It is interesting to note that amitriptyline was not effective for the "core" depressive features of the Hamilton Depression Rating Scale but rather was effective for the seven "associated" symptoms of diurnal variation, depersonalization, paranoid symptoms, obsessive-compulsive symptoms, helplessness, hopelessness, and worthlessness. Patients who had major depression were not more likely to respond. Schizotypal symptoms and paranoia predicted a poor response to amitriptyline.

A small crossover study comparing desipramine (mean dose=162.5 mg/day) with lithium carbonate (mean dose=985.7 mg/day) and placebo in outpatients with borderline personality disorder and minimal axis I mood comorbidity found no significant differences between desipramine and placebo in improvement of affective symptoms, anger, or suicidal symptoms or in therapist or patient perceptions of improvement after 3 and 6 weeks (61).

A small open-label study that assessed the use of amoxapine (an antidepressant with neuroleptic properties) in patients with borderline personality disorder with or without schizotypal personality disorder found that it was not effective for patients with only borderline personality disorder (174). However, it was effective for patients with borderline personality disorder and comorbid schizotypal personality disorder, who had more severe symptoms. This latter group had improvement in cognitive-perceptual, depressive, and global symptoms (174).

In outpatients with a primary diagnosis of atypical depression (which required a current diagnosis of major, minor, or intermittent depression plus associated atypical features) and borderline personality disorder as a secondary diagnosis, imipramine (200 mg/day) produced global improvement in 35% of patients with comorbid borderline personality disorder. In contrast, phenelzine had a 92% response rate in the same sample (57). The presence of borderline personality disorder symptoms predicted a negative global response to imipramine but a positive global response to phenelzine.

One longer-term study was conducted in patients hospitalized for a suicide attempt who were diagnosed with borderline personality disorder or histrionic personality disorder but not axis I depression (175). In this 6-month, double-blind, placebo-controlled study of a low dose of mianserin (30 mg/day), no antidepressant or prophylactic efficacy was found for mianserin compared with placebo for mood symptoms or recurrence of suicidal acts. (The same investigators did demonstrate efficacy against recurrent suicidal acts in this high-risk population with a depot neuroleptic, flupenthixol [80].)

These data suggest that the utility of tricyclic antidepressants in patients with borderline personality disorder is highly questionable. When a clear diagnosis of comorbid major depression can be made, SSRIs are the treatment of choice. When atypical depression is present, the MAOIs have demonstrated superior efficacy to tricyclic antidepressants; however, they must be used with great caution given the high risk of toxicity. (Although the SSRIs have not been extensively studied in atypical depression, at least one double-blind study has indicated comparable efficacy for fluoxetine and phenelzine for the treatment of atypical depression [176].) The efficacy of SSRIs in borderline personality disorder and their favorable safety profile argue for their empirical use in patients with borderline personality disorder with atypical depression.

At best, the response to tricyclic antidepressants (e.g., imipramine) in patients with borderline personality disorder appears modest. The possibility of behavioral toxicity and the known lethality of tricyclic antidepressants in overdose support the preferential use of an SSRI or related antidepressant for patients with borderline personality disorder.

c) Side effects

Common side effects of tricyclic antidepressants include sedation, constipation, dry mouth, and weight gain. The toxicity of tricyclic antidepressants in overdose, including death, indicates that they should be used with caution in patients at risk for suicide. Patients with cardiac conduction abnormalities may experience a fatal arrhythmia with tricyclic antidepressant treatment. For some inpatients with borderline personality disorder, treatment with amitriptyline has paradoxically been associated with behavioral toxicity, consisting of increased suicide threats, paranoid ideation, demanding and assaultive behaviors, and an apparent disinhibition of impulsive behavior (50, 177).

d) Implementation issues

Other antidepressants are generally preferred over the tricyclic antidepressants for patients with borderline personality disorder. If tricyclic antidepressants are used, the patient should be carefully monitored for signs of toxicity and paradoxical worsening. Doses used in published studies were in the range of 150� 250 mg/day of amitriptyline, imipramine, or desipramine. Blood levels may be a useful guide to whether the dose is adequate or toxicity is present.

3. MAOI antidepressants

a) Goals

MAOIs are used to treat affective symptoms, hostility, and impulsivity related to mood symptoms in patients with borderline personality disorder.

b) Efficacy

MAOIs have been studied in patients with borderline personality disorder in three placebo-controlled acute treatment trials (55�57). In an outpatient study of phenelzine versus imipramine that selected patients with atypical depression (with borderline personality disorder as a secondary comorbid condition), global improvement occurred in 92% of patients given 60 mg/day of phenelzine compared with 35% of patients given 200 mg/day of imipramine (57). In a study of tranylcypromine, trifluoperazine, alprazolam, and carbamazepine in which borderline personality disorder was a primary diagnosis but comorbid with hysteroid dysphoria (55), tranylcypromine (40 mg/day) improved a broad spectrum of mood symptoms, including depression, anger, rejection sensitivity, and capacity for pleasure. Cowdry and Gardner (55) noted that, "the MAOI proved to be the most effective psychopharmacological agent overall, with clear effects on mood and less prominent effects on behavioral control." Tranylcypromine also significantly decreased impulsivity and suicidality, with a near significant effect on behavioral dyscontrol. When borderline personality disorder is the primary diagnosis, with no selection for atypical depression or hysteroid dysphoria, results are clearly less favorable. Soloff and colleagues (56) studied borderline personality disorder inpatients with comorbid major depression (53%), hysteroid dysphoria (44%), and atypical depression (46%); the patient group was not selected for presence of a depressive disorder. Phenelzine was effective for self-rated anger and hostility but had no specific efficacy, compared with placebo or haloperidol, for atypical depression or hysteroid dysphoria. These three acute trials were 5�6 weeks in duration. A 16-week continuation study of the responding patients in a follow-up study (68) showed some continuing modest improvement over placebo beyond the acute 5-week trial for depression and irritability. Phenelzine appeared to be activating, which was considered favorable in the clinical setting.

On balance, these studies suggest that MAOIs are often helpful for atypical depressive symptoms, anger, hostility, and impulsivity in patients with borderline personality disorder. These effects appear to be independent of a current mood disorder diagnosis (56), although one study found a nonsignificantly higher rate of MAOI response for patients with a past history of major depression or bipolar II disorder (55).

c) Side effects

Phenelzine can cause weight gain (56) and can be difficult to tolerate. Other side effects include orthostatic hypotension (55). Fatal hypertensive crises are the most serious potential side effect of MAOIs, although no study reported any hypertensive crises due to violation of the tyramine dietary restriction. The initial clinical picture of MAOI poisoning is one of agitation, delirium, hallucinations, hyperreflexia, tachycardia, tachypnea, dilated pupils, diaphoresis, and, often, convulsions. Hyperpyrexia is one of the most serious problems (178).

d) Implementation issues

Doses of phenelzine and tranylcypromine used in published studies ranged from 60�90 mg/day and 10�60 mg/day, respectively. Experienced clinicians may vary doses according to their usual practice in treating depressive or anxiety disorders. Adherence to a tyramine-free diet is critically important and requires careful patient instruction, ideally supplemented by a printed guide to tyramine-rich foods and medication interactions, especially over-the-counter decongestants found in common cold and allergy remedies. Given the impulsivity of patients with borderline personality disorder, it is helpful to review in detail the potential for serious medical consequences of nonadherence to dietary restrictions, the symptoms of hypertensive crisis, and an emergency treatment plan in case of a hypertensive crisis. Patients must be instructed to discontinue an SSRI long enough in advance of instituting MAOI therapy to avoid precipitating a serotonin syndrome.

4. Lithium carbonate and anticonvulsant mood stabilizers

a) Goals

Lithium carbonate and the anticonvulsant mood stabilizers carbamazepine and divalproex sodium are used to treat symptoms of behavioral dyscontrol in borderline personality disorder, with possible efficacy for symptoms of affective dysregulation.

b) Efficacy

The efficacy of lithium carbonate for bipolar disorder led to treatment trials in patients with personality disorders characterized by mood dysregulation and impulsive aggression. Rifkin and colleagues (179, 180) demonstrated improvement in mood swings in 21 patients with emotionally unstable character disorder, a DSM-I diagnosis characterized by brief but nonreactive mood swings, both depressive and hypomanic, in the context of a chronically maladaptive personality resembling "hysterical character." In this placebo-controlled crossover study (each medication was taken for 6 weeks), there was decreased variation in mood (i.e., fewer "mood swings") and global improvement in 14 of 21 patients during lithium treatment. Subsequent case reports demonstrated that lithium had mood-stabilizing and antiaggressive effects in patients with borderline personality disorder (181, 182).

One double-blind, placebo-controlled crossover study compared lithium with desipramine in 17 patients with borderline personality disorder (61). All patients took lithium for 6 weeks (mean dose=985.7 mg/day) and received concurrent psychotherapy. Among 10 patients completing both lithium and placebo treatments, therapists� blind ratings indicated greater improvement during the lithium trial, although patients� self-ratings did not reflect significant differences between lithium and placebo. The authors noted that therapists were favorably impressed by decreases in impulsivity during the lithium trial, an improvement not fully appreciated by the patients themselves. There has never been a double-blind, placebo-controlled trial of the antiaggressive effects of lithium carbonate in patients with borderline personality disorder selected for histories of impulsive aggression.

The anticonvulsant mood stabilizer carbamazepine has been studied in two double-blind, placebo-controlled studies that used very different patient groups, resulting in inconsistent findings. Gardner and Cowdry (55, 62), in a crossover trial, studied female outpatients with borderline personality disorder and comorbid hysteroid dysphoria along with extensive histories of behavioral dyscontrol. Patients underwent a 6-week trial of carbamazepine (mean dose=820 mg/day) and continued receiving psychotherapy. Patients had decreased frequency and severity of behavioral dyscontrol during the carbamazepine trial. Among all patients, there were significantly fewer suicide attempts or other major dyscontrol episodes along with improvement in anxiety, anger, and euphoria (by a physician�s assessment only) with carbamazepine treatment compared with placebo.

De la Fuente and Lotstra (63) failed to replicate these findings, although this may be due to their small study group size (N=20). These investigators conducted a double-blind, placebo-controlled trial of carbamazepine in inpatients with a primary diagnosis of borderline personality disorder. Patients with any comorbid axis I disorder, a history of epilepsy, or EEG abnormalities were excluded. Unlike in the Cowdry and Gardner study (55), patients were not selected for histories of behavioral dyscontrol. There were no significant differences between carbamazepine and placebo on measures of affective or cognitive-perceptual symptoms, impulsive-behavioral "acting out," or global symptoms.

Divalproex sodium has been used in open-label trials targeting the agitation and aggression of patients with borderline personality disorder in a state hospital setting (70) and mood instability and impulsivity in an outpatient clinic (66). Wilcox (70) reported a 68% decrease in time spent in seclusion as well as improvement in anxiety, tension, and global symptoms among 30 patients with borderline personality disorder receiving divalproex sodium (with dose titrated to a level of 100 mg/ml) for 6 weeks in a state hospital. Patients did not have "psychiatric comorbid conditions" (by clinical assessment), although five had an EEG abnormality (but no seizure disorders); concurrent psychotropic medications were allowed. An abnormal EEG predicted improvement with divalproex sodium. The author noted that both the antiaggressive and antianxiety effects of divalproex sodium appeared instrumental in decreasing agitation and time spent in seclusion.

An open-label study by Stein and colleagues (66) enrolled 11 cooperative outpatients with borderline personality disorder, all of whom had been in psychotherapy for a minimum of 8 weeks and were free of other medications before starting divalproex sodium treatment, which was titrated to levels of 50�100 mg/ml. Among the eight patients who completed the study, four responded in terms of global improvement and observed irritability; physician ratings of mood, anxiety, anger, impulsivity, and rejection sensitivity; and patient ratings of global improvement. There were no significant changes in measures specific for depression and anxiety, but baseline depression and anxiety scores were low in this population.

Kavoussi and Coccaro (69) also reported significant improvement in impulsive aggression and irritability after 4 weeks of treatment with divalproex sodium in 10 patients with impulsive aggression in the context of a cluster B personality disorder, five of whom (four completers) had borderline personality disorder. Among the eight patients who completed the 8-week trial, six had a 50% or greater reduction in aggression and irritability. All patients had not responded to a previous trial with fluoxetine (up to 60 mg/day for 8 weeks).

Only one small, randomized controlled trial of divalproex has been reported that involved patients with borderline personality disorder (65). Among 12 patients randomly assigned to divalproex, only six completed a 10-week trial, five of whom responded in terms of global measures. There was improvement in depression, albeit not statistically significant, and aggression was unchanged. None of the four patients randomly assigned to placebo completed the study.

In summary, preliminary evidence suggests that lithium carbonate and the mood stabilizers carbamazepine and divalproex may be useful in treating behavioral dyscontrol and affective dysregulation in some patients with borderline personality disorder, although further studies are needed. The only report on the newer anticonvulsants (i.e., gabapentin, lamotrigine, topiramate) in borderline personality disorder is a case series in which three of eight patients had a good response to lamotrigine (183). Because of the paucity of evidence concerning these agents, careful consideration of the risks and benefits is recommended when using such medications pending the publication of findings from systematic studies.

c) Side effects

Although lithium commonly causes side effects, most are minor or can be reduced or eliminated by lowering the dose or changing the dosage schedule. More common side effects include polyuria, polydipsia, weight gain, cognitive problems (e.g., dulling, poor concentration), tremor, sedation or lethargy, and gastrointestinal distress (e.g., nausea). Lithium may also have renal effects and may cause hypothyroidism. Lithium is potentially fatal in overdose and should be used with caution in patients at risk of suicide.

Carbamazepine�s most common side effects include neurological symptoms (e.g., diplopia), blurred vision, fatigue, nausea, and ataxia. Other side effects include skin rash, mild leukopenia or thrombocytopenia, and hyponatremia. Rare, idiosyncratic, but potentially fatal side effects include agranulocytosis, aplastic anemia, hepatic failure, exfoliative dermatitis, and pancreatitis. Carbamazepine may be fatal in overdose. In studies of patients with borderline personality disorder, carbamazepine has been reported to cause melancholic depression (64).

Common dose-related side effects of valproate include gastrointestinal distress (e.g., nausea), benign hepatic transaminase elevations, tremor, sedation, and weight gain. With long-term use, women may be at risk of developing polycystic ovaries or hyperandrogenism. Mild, asymptomatic leukopenia and thrombocytopenia occur less frequently. Rare, idiosyncratic, but potentially fatal adverse events include hepatic failure, pancreatitis, and agranulocytosis.

d) Implementation issues

Full guidelines for the use of these medications can be found in the APA Practice Guideline for the Treatment of Patients With Bipolar Disorder (85). Lithium carbonate and the anticonvulsant mood stabilizers are used in their full therapeutic doses, with plasma levels guiding dosing. Routine precautions observed for the use of these medications in other disorders also apply to their use in borderline personality disorder, e.g., plasma level monitoring of thyroid and kidney function with prolonged lithium use, periodic measure of WBC count with carbamazepine therapy, and hematological and liver function tests for divalproex sodium.

5. Anxiolytic agents

a) Goals

Anxiolytic medications are used to treat the many manifestations of anxiety in patients with borderline personality disorder, both as an acute and as a chronic symptom.

b) Efficacy

Despite widespread use, there is a paucity of studies investigating the use of anxiolytic medications in borderline personality disorder. Cowdry and Gardner (55) included alprazolam in their double-blind, placebo-controlled, crossover study of outpatients with borderline personality disorder, comorbid hysteroid dysphoria, and extensive histories of behavioral dyscontrol. Use of alprazolam (mean dose=4.7 mg/day) was associated with greater suicidality and episodes of serious behavioral dyscontrol (drug overdoses, self-mutilation, and throwing a chair at a child). This occurred in seven (58%) of 12 patients taking alprazolam compared with one (8%) of 13 patients receiving placebo. However, in a small number of patients (N=3), alprazolam was noted to be helpful for anxiety in carefully selected patients with borderline personality disorder (52). Case reports suggest that clonazepam is helpful as an adjunctive agent in the treatment of impulsivity, violent outbursts, and anxiety in a variety of disorders, including borderline personality disorder (54).

Although clinicians have presented preliminary experiences with nonbenzodiazepine anxiolytics in patients with borderline personality disorder (e.g., buspirone) (184), there are currently no published studies of these anxiolytics in borderline personality disorder.

c) Side effects

Behavioral disinhibition, resulting in impulsive and assaultive behaviors, has been reported with alprazolam in patients with borderline personality disorder. Benzodiazepines, in general, should be used with care because of the potential for abuse and the development of pharmacological tolerance with prolonged use. These are particular risks in patients with a history of substance use.

d) Implementation issues

In the absence of clear evidence-based recommendations, dose and duration of treatment must be guided by clinical need and judgment, keeping in mind the potential for abuse and pharmacological tolerance.

6. Opiate antagonists

a) Goals

It has been suggested that the relative subjective numbing and physical analgesia that patients with borderline personality disorder often feel during episodes of self-mutilation, as well as the reported sense of relative well-being afterward, might be due to release of endogenous opiates (185�187). Opiate antagonists have been employed in an attempt to block mutilation-induced analgesia and euphoria and thereby reduce self-injurious behavior in patients with borderline personality disorder.

b) Efficacy

Clinical case reports (188) and several small case series have assessed the efficacy of opiate antagonists for self-injurious behavior, and two suggested some improvement in this behavior (189, 190). One small, double-blind study involving female patients with borderline personality disorder with a history of self-injurious behavior who underwent a stress challenge showed no effect of opiate receptor blockade with naloxone on cold pressor pain perception or mood ratings (191). While the stress level may not have been high enough to mimic clinical situations, the study does not support the theory that opiate antagonism plays a role in reducing self-injurious behavior.

Despite the few promising clinical case reports, these reports are very preliminary, and there is no clear evidence from well-controlled trials indicating that opiate antagonists are effective in reducing self-injurious behavior among patients with borderline personality disorder.

c) Side effects

Nausea and diarrhea are occasionally reported (190).

d) Implementation issues

In published reports, the typical dose of naltrexone was 50 mg/day. No time limit for treatment emerges from the literature, but the effect is presumably reversed when the medication stops.

7. Neuroleptics

a) Goals

The primary goal of treatment with neuroleptics in borderline personality disorder is to reduce acute symptom severity in all symptom domains, particularly schizotypal symptoms, psychosis, anger, and hostility.

b) Efficacy

Early clinical experience with neuroleptics targeted the "micropsychotic" or schizotypal symptoms of borderline personality disorder. However, affective symptoms (mood, anxiety, anger) and somatic complaints also improved with low doses of haloperidol, perphenazine, and thiothixene. An open-label trial of thioridazine (mean dose=92 mg/day) led to marked improvement in impulsive-behavioral symptoms, global symptom severity, and overall borderline psychopathology (78). Similar findings were reported for adolescents with borderline personality disorder treated with flupentixol (mean dose=3 mg/day) (77), with improvement in impulsivity, depression, and global functioning.

Systematic, parallel studies that compared neuroleptics without a placebo control condition also reported a broad spectrum of efficacy. Leone (73) found that loxapine succinate (mean dose=14.5 mg/day) or chlorpromazine (mean dose=110 mg/day) improved depressed mood, anxiety, anger/hostility, and suspiciousness. Serban and Siegel (74) reported that thiothixene (mean dose=9.4 mg/day, SD=7.6) or haloperidol (mean dose=3.0 mg/day, SD=0.8) produced improvement in anxiety, depression, derealization, paranoia (ideas of reference), general symptoms, and a global measure of borderline psychopathology.

Subsequent double-blind, placebo-controlled trials also suggested a broad spectrum of efficacy for low-dose neuroleptics in the treatment of borderline personality disorder. Acute symptom severity improved in cognitive-perceptual, affective, and impulsive-behavioral symptom domains, although efficacy for schizotypal symptoms, psychoticism, anger, and hostility were most consistently noted.

Many of the double-blind, placebo-controlled studies of neuroleptics in borderline personality disorder are noteworthy for biases in sample selection that strongly affected outcomes. In a study of patients with borderline or schizotypal personality disorder and at least one psychotic symptom (which biased the sample toward cognitive-perceptual symptoms), thiothixene (mean dose=8.7 mg/day for up to 12 weeks) was more effective than placebo for psychotic cluster symptoms� specifically illusions and ideas of reference�and self-rated obsessive-compulsive and phobic anxiety symptoms but not depression or global functioning (75). The more severely symptomatic patients were at baseline (e.g., in terms of illusions, ideas of reference, or obsessive-compulsive and phobic anxiety symptoms), the better they responded to thiothixene (75).

Cowdry and Gardner (55) conducted a complex, placebo-controlled, four-drug crossover study in borderline personality disorder outpatients with trifluoperazine (mean dose=7.8 mg/day). Patients were required to meet criteria for hysteroid dysphoria and have a history of extensive behavioral dyscontrol, introducing a bias toward affective and impulsive-behavioral symptoms. All patients were receiving psychotherapy. Those patients who were able to keep taking trifluoperazine for 3 weeks or longer (seven of 12 patients) had improved mood, with significant improvement over placebo on physician ratings of depression, anxiety, rejection sensitivity, and suicidality.

Soloff and colleagues (50, 51) studied acutely ill inpatients, comparing haloperidol with amitriptyline and placebo in a 5-week trial. Patients who received haloperidol (mean dose=4.8 mg/day) improved significantly more than those receiving placebo across all symptom domains (50), including global measures, self- and observer-rated depression, anger and hostility, schizotypal symptoms, psychoticism, and impulsive behaviors (51). Haloperidol was as effective as amitriptyline for depressive symptoms.

However, a second study by the same group (56) that used the same design but compared haloperidol with phenelzine and placebo failed to replicate the broad-spectrum efficacy of haloperidol (mean dose=3.9 mg/day). Efficacy for haloperidol was limited to hostile belligerence and impulsive-aggressive behaviors, and placebo effects were powerful. Patients in this study had milder symptoms, especially in the cognitive-perceptual and impulsive-behavioral symptom domains, than patients in the first study.

Cornelius and colleagues (68) followed a subset of the aforementioned group who had responded to haloperidol, phenelzine, or placebo for 16 weeks following acute treatment. Patients� intolerance of the medication, a high dropout rate, and nonadherence were decisive factors in this study. The attrition rates at 22 weeks were 87.5% for haloperidol, 65.7% for phenelzine, and 58.1% for placebo. Further significant improvement with haloperidol treatment (compared with placebo) occurred only for irritability (with improvement for hostility that was not statistically significant). Depressive symptoms significantly worsened with haloperidol treatment over time, which was attributed, in part, to the side effect of akinesia. Clinical improvement was modest and of limited clinical importance.

Montgomery and Montgomery (80) controlled for nonadherence by using depot flupenthixol decanoate, 20 mg once a month, in a continuation study of recurrently parasuicidal patients with borderline personality disorder and histrionic personality disorder. Over a 6-month period, patients receiving flupentixol had a significant decrease in suicidal behaviors compared with the placebo group. Significant differences emerged by the fourth month and were sustained through 6 months of treatment. This important study awaits replication.

The introduction of the newer atypical neuroleptics increases clinicians� options for treating borderline personality disorder. To date, findings from only two small open-label trials have been published, both with clozapine. Frankenburg and Zanarini (81) reported that clozapine (mean dose=253.3 mg/day, SD=163.7) improved positive and negative psychotic symptoms and global functioning (but not depression or other symptoms) in 15 patients with borderline personality disorder and comorbid axis I psychotic disorder not otherwise specified who had not responded to (or were intolerant of) other neuroleptics. Improvement was modest but statistically significant. Patients were recruited from a larger study of patients with treatment-resistant psychotic disorders, raising the question of whether their psychotic symptoms were truly part of their borderline personality disorder.

These concerns were addressed by Benedetti and colleagues (71), who excluded all patients with axis I psychotic disorders from their cohort of patients with refractory borderline personality disorder. Target symptoms included "psychotic-like" symptoms that are more typical of borderline personality disorder. Patients had not responded to at least 4 months of prior treatment with medication and psychotherapy. In a 4-month, open-label trial of 12 patients treated with clozapine (mean dose= 43.8 mg/day, SD=18.8) and concurrent psychotherapy, a low dose of clozapine improved symptoms in all domains�cognitive-perceptual, affective, and impulsive-behavioral.

Despite a lack of data, clinicians are increasingly using olanzapine, risperidone, and quetiapine for patients with borderline personality disorder. These medications have less risk than clozapine and may be better tolerated than the typical neuroleptics. Schulz and colleagues (83) presented preliminary data from a double-blind, placebo-controlled, 8-week trial of risperidone in 27 patients with borderline personality disorder who received an average dose of 2.5 mg/day (to a maximum of 4 mg/ day). On global measures of functioning, there was no significant difference between risperidone and placebo, although the authors noted that risperidone-treated patients were "diverging from the placebo group" in paranoia, psychoticism, interpersonal sensitivity, and phobic anxiety (83). The same group conducted an 8-week, open-label study of olanzapine in patients with borderline personality disorder and comorbid dysthymia (82). Patients received an average dose of 7.5 mg/day (range=2.5�10 mg/day). Among the 11 completers, significant improvement was reported across all domains, with particular improvement noted in depression, interpersonal sensitivity, psychoticism, anxiety, and anger/hostility. These medications require further investigation in double-blind studies.

In summary, neuroleptics are the best-studied psychotropic medications for borderline personality disorder. The literature supports the use of low-dose neuroleptics for the acute management of global symptom severity, with specific efficacy for schizotypal symptoms and psychoticism, anger, and hostility. Relief of global symptom severity in the acute setting may be due, in part, to nonspecific "tranquilizer" effects of neuroleptics, whereas symptom-specific actions against psychoticism, anger, and hostility may relate more directly to dopaminergic blockade. Acute treatment effects of neuroleptic drugs in borderline personality disorder tend to be modest but clinically and statistically significant.

Two studies that addressed continuation and maintenance treatment of a patient with borderline personality disorder with neuroleptics had contradictory results. The Montgomery and Montgomery study (80) reported efficacy for recurrent parasuicidal behaviors, whereas the Cornelius et al. study (68) suggested very modest utility for only irritability and hostility. More controlled trials are needed to investigate low-dose neuroleptics in continuation and maintenance treatment.

c) Side effects

Dropout rates in neuroleptic trials in borderline outpatients range from 13.7% for a 6-week trial (73) to 48.3% for a 12-week trial (75) to 87.5% for a 22-week continuation study (68). In acute studies, patient nonadherence is often due to typical medication side effects, e.g., extrapyramidal symptoms, akathisia, sedation, and hypotension. Patients with borderline personality disorder who have experienced relief of acute symptoms with low-dose neuroleptics may not tolerate the side effects of the drug with longer-term treatment. The risk of tardive dyskinesia must be considered in any decision to continue neuroleptic medication over the long term. Thioridazine has been associated with cardiac rhythm disturbances related to widening of the Q-T interval and should be avoided. In the case of clozapine, the risk of agranulocytosis is especially problematic. While the newer atypical neuroleptics promise a more favorable side effect profile, evidence of efficacy in borderline personality disorder is still awaited. Neuroleptics should be given in the context of a supportive doctor-patient relationship in which side effects and nonadherence are addressed frequently.

d) Implementation issues

All studies have used a low dose and demonstrated beneficial effects within several weeks. With the exception of one study that used a depot neuroleptic (flupentixol, which is not available in the United States), all medications were given orally and daily. Acute treatment studies are a good model for acute clinical care and typically range from 5 to 12 weeks in duration. There is insufficient evidence to make a strong recommendation concerning continuation and maintenance therapies. At present, this is best left to the clinician�s judgment after carefully weighing the risks and benefits for the individual patient. CBC monitoring must be done if clozapine is used.

8. ECT

a) Goals

The goal of ECT in patients with borderline personality disorder is to decrease depressive symptoms in individuals with a comorbid axis I mood disorder, which is present in as many as one-half of hospitalized patients with borderline personality disorder.

b) Efficacy

Most of the clinical and empirical literature that describes experience with ECT in patients with major depression comorbid with personality disorders does not report results specifically for borderline personality disorder. Although studies that used a naturalistic design have had inconsistent findings, patients with major depression and a comorbid personality disorder were generally less responsive to somatic treatments than patients with major depression alone.

In one naturalistic follow-up study (based on chart review), there was no significant difference in recovery rates for 10 patients with major depressive disorder and a personality disorder (40% recovery) compared with 41 patients with major depressive disorder alone (65.9% recovery) (192). In another study, involving 1,471 depressed inpatients, depressed patients with a personality disorder were 50% less likely to be recovered at hospital discharge than depressed patients without a personality disorder (193).

Several uncontrolled studies found that outcome was dependent on the time of assessment. In one small study (194), there were no significant differences in immediate response to ECT between depressed subjects with or without a personality disorder; however, at a 6-month follow-up evaluation, the patients with a personality disorder had more rehospitalizations and more severe depression symptoms. Conversely, in another uncontrolled study of inpatients with major depression (195), compared with depressed patients without a personality disorder, those with a personality disorder had a poorer outcome in terms of depression and social functioning immediately following treatment. However, after 6 and 12 weeks of follow-up, there were no differences between the two groups in terms of depression and social functioning. The number of rehospitalizations did not differ between groups at the 6-month and 12-month follow-up evaluations.

In another small study (N=16) (196�198) that used the self-rated Millon Clinical Multiaxial Inventory�II and assessed borderline personality disorder, there was significant improvement in avoidant, histrionic, aggressive/sadistic, and schizotypal personality traits with ECT. Improvements were noted in passive-aggressive and borderline personality traits that did not reach statistical significance. The presence of pretreatment borderline traits predicted poorer outcome with ECT (198).

Although the results of these studies appear somewhat divergent, most found that patients with major depression and a personality disorder have a less favorable outcome with ECT than depressed patients without a personality disorder.

c) Adverse effects

Because ECT is not recommended for borderline personality disorder per se, adverse effects are not described here and can be found in the APA Practice Guideline for the Treatment of Patients With Major Depressive Disorder (84).

d) Implementation issues

The affective dysregulation, low self-esteem, pessimism, chronic suicidality, and self-mutilation of patients with borderline personality disorder are often misconstrued as axis I depression. Clinical experience suggests that, not infrequently, these characterological manifestations of borderline personality disorder are treated with ECT, often resulting in a poor outcome. Although there is a paucity of ECT studies involving patients with borderline personality disorder, a recommendation for ECT in these patients with comorbid major depression should be guided by the presence and severity of verifiable neurovegetative symptoms, e.g., sleep disturbance, appetite disturbance, weight change, low energy, and anhedonia. These symptoms should ideally be confirmed by outside observers, as they provide an objective way to assess treatment response. Perhaps the greatest challenge for the clinician is not when to institute ECT in the depressed patient with borderline personality disorder but when to stop. As the neurovegetative symptoms of major depression resolve, many patients continue to have borderline features that clinical experience suggests are unresponsive to ECT. Knowledge of the patient�s personality functioning before the onset of major depression is critical to knowing when the "baseline" has been achieved. Many patients with borderline personality disorder who are considered nonresponsive to ECT because of persistence of depressive features are, in fact, already in remission from their axis I depression but continue to experience chronic characterological depressive features.

Notable progress has been made in our understanding of borderline personality disorder and its treatment. However, there are many remaining questions regarding treatments with demonstrated efficacy, including how to optimally use them to achieve the best health outcomes for patients with borderline personality disorder. In addition, many therapeutic modalities have received little empirical investigation for borderline personality disorder and require further study. The efficacy of various treatments also needs to be studied in populations such as adolescents, the elderly, forensic populations, and patients in long-term institutional settings. The following is a sample of the types of research questions that require further study.

PART C:
FUTURE RESEARCH NEEDS

VII. PSYCHOTHERAPY

Many aspects of psychotherapy in the treatment of borderline personality disorder require further investigation. For example, further controlled treatment studies of psychodynamic psychotherapy, dialectical behavior therapy, and other forms of cognitive behavior therapy are needed, particularly in outpatient settings. In addition, psychotherapeutic interventions that have received less investigation, such as group therapy, couples therapy, and family interventions, require study. The following are some specific questions that need to be addressed by future research:

 

• What is the relative efficacy of different psychotherapeutic approaches? Which types of patients respond to which types of psychotherapy?
• What components of dialectical behavior therapy and psychodynamic psychotherapy are responsible for their efficacy? What common elements of these treatments are responsible for their efficacy?
• What are the indications for use of psychodynamic psychotherapy and dialectical behavior therapy? How does the presence of certain clinical features (e.g., prominent self-destructive behavior or dissociative features) affect response to these treatments?
• To what extent is a good outcome due to the unique components of these treatments versus the amount of treatment received?
• How effective are psychodynamic psychotherapy and dialectical behavior therapy when used in the community rather than in specialized treatment settings, and how can these treatments be optimally implemented in community settings?
• What is the optimal duration of psychotherapy for patients with borderline personality disorder?
• Is there a model of brief psychotherapy (12�30 sessions) that is effective for borderline personality disorder?
• What are the optimal frequencies of psychotherapeutic contact for different psychotherapies during different stages of treatment?
• What is the relative efficacy of psychotherapy versus pharmacotherapy for patients with borderline personality disorder? Do certain patients respond better to one treatment modality than to the other?
• What is the relative efficacy of a combination of psychotherapy and pharmacotherapy versus either treatment modality alone?

VIII. Pharmacotherapy and Other Somatic Treatments

Many aspects of pharmacotherapy in the treatment of borderline personality disorder also require investigation. Further controlled treatment studies of medications�in particular, those that have received relatively little investigation (for example, atypical neuroleptics)�are needed. Studies of continuation and maintenance treatment as well as treatment discontinuation are especially needed, as are systematic studies of treatment sequences and algorithms. The following are some specific questions that need to be addressed by future research:

• What is the relative efficacy of different pharmacological approaches for the behavioral dimensions of borderline personality disorder?