Special Article

Psychotherapy of Personality Disorders

Glen O. Gabbard, M.D.

Although personality disorders are often regarded as

“untreatable” by third-party payers, there is actually a
growing empirical literature suggesting that Axis II
P sychoanalysis and psychoanalytic therapy have
long been used in the treatment of patients with
personality disorders (PDs). The seminal work of Wil-
conditions may be eminently treatable by psychotherapy. helm Reich1 on analysis of the patient’s “character ar-
This literature is critically reviewed, the implications mor” laid the groundwork for this approach. More
for length of treatment are discussed, and cost- recently, other forms of psychotherapy have been
widely used to treat PDs. Research has been slow to
effectiveness issues are examined.
catch up with clinical practice, but in the last 20 years
(The Journal of Psychotherapy Practice and
behavior therapy, cognitive therapy, supportive ther-
Research 2000; 9:1–6)
apy, and short-term and long-term psychoanalytic/psy-
chodynamic therapy have all been studied, and the
results of these studies have given psychotherapists ev-
erywhere reason to feel cautiously optimistic about
treating long-standing and complex character pathol-
ogy.
Some managed care companies are still reluctant to
cover psychotherapy for PDs. At the same time, mental
health professionals have become increasingly aware
that these conditions are extraordinarily costly in terms
of utilization of medical services, violence, suicide, and
psychiatric hospitalization.2 It has also been widely rec-
ognized that PDs are not as intransigent to change as
we once thought. In addition to the psychotherapy stud-
ies, other research has demonstrated that some Axis II
disorders are variable over time and are influenced by
significant life events such as achievements and new re-
lationships.3–5 In this review, I summarize the major
psychotherapy research on personality disorders that
has appeared during the past two decades and comment
on its implications for generalized treatment planning.

Received August 25, 1999; accepted September 30, 1999. From The
Menninger Clinic, P.O. Box 829, Topeka KS 66601–0829. Send cor-
respondence to Dr. Gabbard at the above address.
Copyright q 2000 American Psychiatric Association

J Psychother Pract Res, 9:1, Winter 2000 1

 Personality Disorders

The studies identified were located through a MEDLINE 1. Prior to therapy, the patients were absent from
search and a review of relevant bibliographies. work an average of 4.7 months per year; following
the therapy, the average had declined to 1.37
months per year.
CONTROLLED STUDIES
2. The number of self-harm episodes after the therapy
was one-fourth the level of the pre-treatment rates.
Much of the research on the psychotherapy of PDs has
3. The number of visits to medical professionals
not employed randomization and control groups, partly
dropped to one-seventh of the pre-treatment rates
because of the difficulties in obtaining funding for such
after the psychotherapy.
studies and partly because the extended nature of many
4. The average time spent as an inpatient decreased
of the therapies leads to formidable design problems
by half.
(e.g., difficulty in finding suitable control subjects; sig-
5. The number of hospital admissions decreased by
nificant dropout rates; and the effect of intervening life
59% after the therapy.
events during the clinical trial). Nevertheless, several
controlled studies have been reported in the literature.
The durability of these changes was confirmed with
Linehan et al.6 randomly assigned 44 patients with
a 5-year follow-up assessment.9 Most of the outcome
borderline personality disorder (BPD) to dialectical be-
measures continued to show declines as compared to
havior therapy (DBT) or to “treatment as usual,” which
the pre-treatment rates. The only exception was that the
consisted of “hit-or-miss” treatment in the community. time away from work began to increase over the 5-year
The patients receiving DBT had once-weekly group follow-up period, but the investigators could not deter-
and once-weekly individual therapy. The individual mine how much of that employment difficulty related
therapy focused on correction of cognitions; the group to the recession that occurred in Sydney during that
meetings taught the patients behavioral coping skills. At time period.
the end of 1 year, the group that was treated with DBT The same investigators10 subsequently published a
had a median of 1.5 acts of self-mutilation in a 12-month comparison of their 30 BPD patients with a wait-list con-
period compared with 9 in the control group. Also, the trol group. The first 30 patients on the waiting list who
episodes of self-mutilation were less severe than those had been waiting 12 months or longer made up the
in the control group. There was a dramatic reduction in comparison group. These patients had their usual treat-
hospital days as a result of DBT, with the treatment ments during the waiting period, which included sup-
group needing only 8.46 days of hospitalization in the portive therapy, crisis intervention, and cognitive
entire year compared with 38.86 days in the control therapy. The investigators then compared the results of
group. In a subsequent report,7 the investigators also the treated patients with those of the wait-list control
determined that DBT subjects had lower anger scores subjects. Of the 30 treated patients, 30% no longer met
and showed improvements in social adjustment. criteria for BPD after 12 months of psychotherapy. The
Controlled studies of psychodynamic therapy for 30 patients on the waiting list for 1 year or more showed
borderline personality disorder have also begun to ap- no change in diagnosis. The treatment group also
pear. An Australian study initially used a “pre/post” de- showed a significant reduction in symptom checklist
sign to follow 30 patients with DSM-III-R borderline scores; the wait-list group showed no significant change
personality disorder prospectively.8 The patients were on those measures. Definitive conclusions cannot be
first identified and followed for 12 months prior to re- drawn from this study because randomization was not
ceiving treatment. The same patients then received employed, the length of time before follow-up data were
twice-weekly psychodynamic therapy influenced by the collected varied for the wait-list group, and different in-
ideas of Winnicott and Kohut for another 12 months. struments to measure symptomatic improvement were
Although the therapy was not manualized, the training used in the two groups. Nevertheless, the results are
therapists were intensively supervised. After termina- suggestive of substantial gains from the dynamic ther-
tion of the therapy, the same patients were followed for apy that was offered.
an additional 12 months. Substantial and enduring im- Promising results were also found in the Halliwick
provements were observed. Among the statistically sig- day unit study by Bateman and Fonagy.11 They com-
nificant changes were the following: pared 38 borderline patients in a psychoanalytically ori-

2 J Psychother Pract Res, 9:1, Winter 2000

Gabbard

ented partial hospital program with those in a control than the waiting-list patients on target complaints,
condition. The partial hospital condition consisted of symptom measures, and social adjustment. At follow-
once-weekly individual psychoanalytic psychotherapy, up (an average of 1.5 years), the improvements were
three-times-weekly group psychoanalytic therapy, maintained. The authors concluded that most patients
once-weekly expressive therapy informed by psycho- with Cluster C PDs as well as some patients with Cluster
drama techniques, weekly community meeting, meet- B disorders, primarily histrionic patients, respond to ei-
ing with a case coordinator, and medication review by ther modality. However, the exclusion criteria in this
a resident psychiatrist. The control treatment consisted study were broad, and therefore many patients with
of regular psychiatric review an average of two times poor prognoses were not included.
per month with a senior psychiatrist, inpatient admis- Separate studies of avoidant personality disor-
sion as appropriate, outpatient and community follow- der13–15 have employed brief behavioral treatments, in-
up, no psychotherapy, and medication similar to that of cluding systematic desensitization, graduated exposure,
the treatment group. and social skills training. In all three studies, signifi-
They found that the treatment group had a clear cantly more improvement was seen in the treatment
reduction in the proportion of the sample with suicide groups than in the waiting-list control groups. Stravyn-
attempts in the previous 6 months, from 95% on ad- ski et al.16 assigned 22 patients to 14 sessions of social
mission to 5.3% at 18 months. The average length of skills training alone or social skills with the addition of
hospitalization in the control group in the last 6 months cognitive techniques that challenged maladaptive be-
of the study increased dramatically; in the treatment liefs. Equal and significant gains were found for both
group, it remained stable at around 4 days per 6 groups. One of the reasons that avoidant personality
months. Both self-reported state and trait anxiety de- disorder, an Axis II condition with little psychodynamic
creased substantially in the treatment group but re-
tradition, has received so much study is because of its
mained unchanged in the control group. Beck
extensive overlap with social phobia. Indeed, many
Depression Inventory scores also significantly de-
contend that there is little validity in distinguishing be-
creased in the treatment group. There was a statistically
tween generalized social phobia and avoidant person-
significant decrease in severity of symptoms as mea-
ality disorder.17 At least two studies17,18 suggest that
sured by the Symptom Checklist–90 at 18 months.
patients who are comorbid for social phobia and avoid-
The investigators concluded that the improvement
ant personality disorder do as well with behavioral
of psychiatric symptoms and suicidal acts occurred after
treatments as social phobic patients without personality
the first 6-month measurement, but a reduction in fre-
disorder. One other study19 found that patients with so-
quency of hospital admission and length of inpatient
cial phobia comorbid for avoidant personality disorder
stay was only clear only in the last 6 months, indicating
a need for longer-term treatment. They also decided improved significantly with treatment but continued to
that partial hospitalization with psychoanalytic therapy report more severe impairment on all outcome mea-
seems to be a promising and cheaper alternative to spe- sures than social phobic patients without avoidant per-
cialist inpatient and general psychiatric treatment. sonality disorder.
Winston et al.12 randomly assigned 81 patients with Some promising results for patients with antisocial
PDs to one of three groups: brief adaptive psychother- personality disorder emerged from a study of opiate ad-
apy, short-term dynamic psychotherapy, or a waiting dicts,20 in which 110 male patients with opiate addiction
list for therapy. The therapies lasted 40 weeks and were were randomly assigned to either paraprofessional drug
compared with outcomes of people who were on the counseling alone or counseling plus professional psy-
waiting list for 15 weeks. Forty-four percent of the pa- chotherapy (either supportive-expressive or cognitive-
tients were diagnosed as having Cluster C PDs, with behavioral). Those who had antisocial personality
another 23% diagnosed as PD Not Otherwise Specified disorder made significant improvement in symptoms
with Cluster C features. Twenty-two percent were di- and employment, with reductions in drug use and ille-
agnosed with Cluster B PDs (antisocial, borderline, his- gal activity—but only if they also had a diagnosis of
trionic, narcissistic), and 4% came from Cluster A depression on Axis I. Antisocial personality disorder
(paranoid, schizoid, schizotypal). The patients in the patients without depression showed little gain from psy-
two therapy conditions improved significantly more chotherapy.

J Psychother Pract Res, 9:1, Winter 2000 3

 Personality Disorders

LENGTH OF PSYCHOTHERAPY AND OUTCOME who get in and out of therapy on the basis of whether
crises are present or absent in their lives. In the disser-
As many of the studies I have reviewed suggest, most tation research of Dr. Lisbeth Hoke,23 58 borderline pa-
personality disorders do not lend themselves to a quick tients were followed for up to 7 years. The BPD subjects
fix; “tincture of time” may be a critical ingredient. PDs in this study could be divided into two different groups
involve a mixture of temperamental characteristics that based on their natural course. The first group (approx-
are biologically based, characteristic patterns of internal imately half) had intermittent or inconsistent psycho-
object relations, particular cognitive styles, and specific therapeutic treatments; the second group had consistent
defense mechanisms. These are often ingrained in the psychotherapy over at least 2 years. Those who re-
individual in such a way that the patient has little aware- mained in a stable psychotherapy process showed
ness of the difficulties these personality traits create for greater improvement in mood functioning, a decreased
others. Hence a good deal of time is required to help need for more intensive psychiatric interventions (such
the patient begin to reflect on how he or she comes as hospitalization, emergency room visits, and day treat-
across. ment), decreased impulsiveness, and improved Global
A small body of research supports the notion that Assessment Scale scores.
PDs may require more extended psychotherapy than A Norwegian study24 supported the findings that
Axis I conditions. Howard et al.21 studied the dose– PDs may take longer to change than Axis I conditions.
effect relationship in psychotherapy and found that bor- In this study, 48 patients were treated with dynamic
derline patients take longer than other groups of psychotherapy that ranged from 9 to 53 sessions. Thirty
psychiatric patients to show improvement in psycho- patients received an Axis I diagnosis and 15 patients
therapy. Fifty percent of anxious and depressed patients had a PD; of the latter, 8 were dependent or avoidant
improved in 8 to 13 sessions, whereas, according to and 7 were histrionic, narcissistic, or borderline. All pa-
clinical chart ratings, borderline patients required 26 to tients were treated by psychoanalytically oriented psy-
52 sessions to achieve similar levels of improvement. chiatrists. For those patients with PDs, the number of
Some patients with BPD did not show significant im- treatment sessions was significantly related to acquisi-
provement until the second year of once-weekly treat- tion of insight 2 years after therapy and to overall dy-
ment. In a subsequent study,22 the same research group namic change 4 years after therapy. There was no such
examined the rapidity of change for different symptom correlation for patients without PDs. Significant long-
constellations. They studied 854 psychotherapy outpa- term dynamic changes in patients with PDs were ob-
tients who were in treatment with 141 psychotherapists. served after treatments that lasted 30 sessions or more.
The orientation of the therapists was predominantly Patients with Cluster B PDs seemed to do as well as
psychodynamic. Symptom checklists were adminis- those with PDs from Cluster C.
tered to all patients, and the symptoms were grouped In a small, intensive study of 5 borderline patients
into three classes: 1) chronic distress, 2) acute distress, successfully treated by experienced analytically ori-
and 3) characterological. The typical outpatient needed ented therapists, Waldinger and Gunderson25 noted that
about 1 year of psychotherapy to have a 75% chance of the therapists’ perseverance over time was an important
symptomatic recovery. The patients with acute symp- factor. A common thread was that all the therapists had
toms showed an improvement rate of 68% to 98% after an unusually strong commitment to persist at the diffi-
52 weekly sessions. For those with chronic distress cult work of therapy until a satisfactory outcome was
symptoms, the improvement rate was 60% to 86% over achieved. A major finding of the study was that none of
the same interval. When the investigators examined the the 5 patients were manifestly diagnosable as having
patients who had characterological symptoms, how- BPD after 4 years of treatment.
ever, only about 59% improved in that time interval;
they therefore concluded that a longer duration of in- DISCUSSION
dividual therapy appears to be necessary for symptoms
embedded in character. It has long been known that the presence of comorbid-
Evidence is accumulating that PD patients who can ity on Axis II complicates the treatment of Axis I con-
remain in a consistent, stable psychotherapy process ditions. Only recently, however, has treatment research
over an extended period of time fare better than those been systematically applied to PDs. The studies re-

4 J Psychother Pract Res, 9:1, Winter 2000

Gabbard

viewed in this article reflect the fact that research on the offices of other medical specialists, and in psychiatric
psychotherapy of PDs is in a very early stage of devel- hospitals. They are a highly treatment-seeking popula-
opment. Many studies are confounded by a lack of ran- tion and will one way or another find access to the
domization and controls, associated inpatient health care delivery system.
treatment, the presence of Axis I conditions, the possi- Both the studies by Linehan et al.6 and Stevenson
bility that maturational processes or life events may be and Meares8 demonstrate that providing extended psy-
responsible for part of the changes measured, and the chotherapy of 1 year or longer for patients with BPD is
use of medications along with the psychotherapy. More- highly cost-effective. Because the patients in the former
over, some PDs, notably those in Cluster A, have hardly study used only 8.46 hospital days per year compared
been studied at all, largely because patients with Cluster with 38.86 for control subjects, and also had much less
A PDs infrequently seek out psychotherapy or psycho- self-mutilation, the research group calculated that reg-
analysis. In a study of 100 patients applying for psycho- ular psychotherapy saved $10,000 per patient per
analysis at the Columbia Psychoanalytic Center, only 4 year.28
were diagnosed by the Personality Disorder Examina- Similar conclusions can be reached from the Ste-
tion26 with paranoid PD, 1 with schizoid PD, and none venson and Meares 1992 study,8 in which the use of
with schizotypal PD.27 hospitalization was cut by half after psychotherapy. Sav-
Despite the limitations of many of the studies re- ings can also be calculated in the reduced number of
viewed, we are now at a point where several tentative visits to medical professionals, the decreased number of
conclusions can be reached. Some of the major symp- self-harm episodes, and the impressive reduction in
toms of avoidant PD can be effectively addressed by work disability.29 When Stevenson and Meares30
use of social skills training and cognitive-behavioral looked at the cost of hospital admissions for the 30 pa-
techniques. Depressed antisocial patients with opiate tients for the year before their psychotherapy, they cal-
addiction may be far more treatable by psychotherapy culated that hospital treatment alone cost $684,346 in
than was previously thought. BPD patients who re- Australian dollars, with a range of $0 to $143,756 per
ceived once-weekly individual and once-weekly group patient. The cost of hospital admissions for the year af-
treatment following the dialectical behavior therapy ter treatment was $41,424, with a range of $0 to $12,333
model appeared to have reduced severity and fre- per patient. The average decrease in cost per patient
quency of parasuicidal behavior and a reduced need for was $21,431 over 12 months. The average cost of ther-
hospitalization. Internal feeling states of hopelessness apy per patient was $13,000, representing a savings per
and despair may show little improvement, but interper- patient of $8,431. Although cost savings were not cal-
sonal relatedness may change for the better. culated for the Halliwick day unit study,11 the figure
Psychodynamic psychotherapy also appears to be suggested similar savings were made in that setting.
highly effective for borderline personality disorder, es-
pecially when combined with an overall partial hospital SUMMARY
program. A psychodynamic approach seems to alter
some of the internal feeling states of depression that are Despite the frequent statements from insurance and
untouched by DBT. managed care companies that personality disorders are
In an era of managed care, it is incumbent on cli- not treatable, there is substantial evidence that they re-
nicians who treat PDs to make a strong cost-effective- spond to psychotherapy. Extended therapy appears to
ness argument for the treatments they prescribe. With be necessary for the full effect of the treatment. Medi-
patients who suffer from severe PDs, such as BPD, there cations may be helpful adjuncts to the psychotherapy
is a substantial medical morbidity, work disability, and with some personality disorders. Moreover, although
need for hospitalization, all of which can be quite costly. intensive and extended psychotherapy may be expen-
If regular outpatient therapy is not provided for these sive, in the long run it is highly cost-effective because it
patients, they will appear in emergency rooms, in the reduces inpatient stays and other costs.

J Psychother Pract Res, 9:1, Winter 2000 5

 Personality Disorders

REFERENCES
1. Reich W: Character Analysis, 3rd edition. New York, Farrar, 17. Hofmann SG, Newman MG, Becker E, et al: Social phobia with
Strauss, and Giroux, 1949 and without avoidant personality disorder: preliminary behav-
2. Gabbard GO, Lazar SG, Hornberger J, et al: The economic ior therapy outcome findings. J Anxiety Disord 1995; 9:427–
impact of psychotherapy: a review. Am J Psychiatry 1997; 438
154:147–155 18. Mersch PPA, Jansen MA, Arntz A: Social phobia and person-
3. Mattanah JJF, Becker DF, Levy KN, et al: Diagnostic stability ality disorders: severity of complaint and treatment effective-
in adolescents followed up 2 years after hospitalization. Am J ness. J Personal Disord 1995; 9:143–159
Psychiatry 1995; 152:889–894 19. Feske U, Perry KJ, Chambless DL, et al: Avoidant personality
4. Perry JC: Longitudinal studies of personality disorders. J Per- disorder as a predictor for treatment outcome among general-
sonal Disord 1993; 7(suppl):63–85 ized social phobics. J Personal Disord 1996; 10:174–184
5. Ronningstam E, Gunderson J, Lyons M: Changes in patholog- 20. Woody GE, McLellan T, Luborsky L, et al: Sociopathy and
ical narcissism. Am J Psychiatry 1995; 152:253–257 psychotherapy outcome. Arch Gen Psychiatry 1985; 42:1081–
6. Linehan MM, Armstrong HE, Suarez A, et al: Cognitive-be- 1086
havioral treatment of chronically parasuicidal borderline pa- 21. Howard KJ, Kopta SM, Krause MS, et al: The dose-effect re-
lationship in psychotherapy. Am Psychol (special issue: Psycho-
tients. Arch Gen Psychiatry 1991; 48:1060–1064
therapy Research) 1986; 41:159–164
7. Linehan MM, Tutek DA, Heard HL, et al: Interpersonal out-
22. Kopta SM, Howard KI, Lowry JL, et al: Patterns of symptom-
come of cognitive behavioral treatment for chronically suicidal atic recovery in psychotherapy. J Consult Clin Psychol 1994;
borderline patients. Am J Psychiatry 1994; 151:1771–1776 62:1009–1016
8. Stevenson J, Meares R: An outcome study of psychotherapy for 23. Hoke LA: Longitudinal Patterns of Behavior in Borderline Per-
patients with borderline personality disorder. Am J Psychiatry sonality Disorder. Doctoral dissertation, Boston University,
1992; 149:358–362 Boston, MA, 1989
9. Stevenson J, Meares R: Borderline patients at 5-year follow-up. 24. Høglend P: Personality disorders and long-term outcome after
Presented at the Annual Congress of the Royal Australia-New brief dynamic psychotherapy. J Personal Disord 1993; 7:168–
Zealand College of Psychiatrists, Cairns, Australia, May 1995 181
10. Meares R, Stevenson J, Comerford A: Psychotherapy with bor- 25. Waldinger RJ, Gunderson JG: Effective Psychotherapy With
derline patients, I: a comparison between treated and untreated Borderline Patients: Case Studies. Washington, DC, American
cohorts. Aust NZ J Psychiatry 1999; 33:467–472 Psychiatric Press, 1989
11. Bateman A, Fonagy P: The effectiveness of partial hospitaliza- 26. Loranger AW, Susman VL, Oldham JM, et al: Personality Dis-
tion in the treatment of borderline personality disorder: a ran- order Examination (PDE): A Structured Interview for DSM-III-
domized controlled trial. Am J Psychiatry 1999; 156:1563–1569 R Personality Disorders. White Plains, NY, New York Hospital–
12. Winston A, Laikin M, Pollack J, et al: Short-term psychotherapy Cornell Medical Center, Westchester Division
of personality disorders. Am J Psychiatry 1994; 151:190–194 27. Oldham JM, Skodol AE: Do patients with paranoid personality
disorder seek psychoanalysis? In Paranoia: New Psychoanalytic
13. Alden L: Short-term structured treatment for avoidant person-
Perspectives, edited by Oldham JM, Bone S. Madison, CT, In-
ality disorder. J Consult Clin Psychol 1989; 57:756–764
ternational Universities Press, 1994, pp 151–166
14. Cappe RF, Alden LE: A comparison of treatment strategies for
28. Heard H: Behavior therapies for borderline patients. Presented
clients functionally impaired by extreme shyness and social at the 147th annual meeting, American Psychiatric Association,
avoidance. J Consult Clin Psychol 1986; 54:796–801 Philadelphia, PA, May 1994
15. Marzillier JS, Lambert C, Kellett J: A controlled evaluation of 29. Gabbard GO: Borderline personality disorder and rational
systematic desensitization and social skills training for inade- managed care policy. Psychoanalytic Inquiry 1997;
quate psychiatric patients. Behav Res Ther 1976; 14:225–238 17(suppl):17–27
16. Stravynski A, Marks I, Yule W: Social skills problems in neu- 30. Stevenson J, Meares R: Psychotherapy with borderline patients,
rotic outpatients: social skills training with and without cogni- II: a preliminary cost-benefit study. Aust NZ J Psychiatry 1999;
tive modification. Arch Gen Psychiatry 1982; 39:1378–1385 33:473–477

6 J Psychother Pract Res, 9:1, Winter 2000