CLARKIN ET AL.
PSYCHODYNAMIC TREATMENT FOR BPD
THE DEVELOPMENT OF A PSYCHODYNAMIC
TREATMENT FOR PATIENTS WITH BORDERLINE
PERSONALITY DISORDER: A PRELIMINARY
STUDY OF BEHAVIORAL CHANGE
John F. Clarkin, PhD, Pamela A. Foelsch, PhD, Kenneth N.
Levy, PhD, James W. Hull, PhD, Jill C. Delaney, MSW, and
Otto F. Kernberg, MD
This study examines the effectiveness of a modified psychodynamic
treatment called Transference Focused Psychotherapy (TFP) designed
specifically for patients with borderline personality disorder (BPD).
Twenty-three female patients diagnosed with DSM-IV BPD began
twice-weekly TFP. Patients were assessed at baseline and at the end of 12
months of treatment with diagnostic instruments, measures of
suicidality, self-injurious behavior, and measures of medical and psychi-
atric service utilization. Compared to the year prior to treatment, the
number of patients who made suicide attempts significantly decreased,
as did the medical risk and severity of medical condition following self-in-
jurious behavior. Compared to the year prior, study patients during the
treatment year had significantly fewer hospitalizations as well as number
and days of psychiatric hospitalization. The dropout rate was 19.1%.
This uncontrolled study is highly suggestive that this structured and
manualized psychodynamic treatment modified for borderline patients
shows promise for the ambulatory treatment of these patients and war-
rants further study.
Borderline personality disorder (BPD) is a highly prevalent and chronic psy-
chiatric problem and constitutes one of the most important sources of
long-term impairment in both treated and untreated populations
(Weissman, 1993; Oldham, et al., 2000). Approximately 11% of psychiatric
outpatients and 19% of inpatients meet the DSM-IV (American Psychiatric
Journal of Personality Disorders, 15(6), 487-495, 2001
© 2001 The Guilford Press
From the Personality Disorders Institute and Department of Psychiatry, Weill Medical College
of Cornell University.
This research was supported in part by grant MH-53705-02from the National Institute of Men-
tal Health, Washington, DC. (PI: Dr. Clarkin), and the DeWitt Wallace Reader’s Digest Fund.
The authors wish to thank Jack Barchas, MD, for institutional support. We acknowledge the
consultation of Marsha Linehan, PhD, to this study, the assistance of Heidi Heard, PhD, in
training with the PHI and THI, and the suggestions of Gerhard Dammann, MD. The authors
wish to thank our colleagues who served as therapists in the study. We would also like to thank
Ann Appelbaum, MD, Michael Stone, MD, and Frank Yeomans, MD, for providing therapy su-
pervision. Finally, we would like to thank members of the Personality Disorders Institute.
Address correspondence to John F. Clarkin, PhD, Personality Disorders Institute, Macy Villa,
The New York Presbyterian Hospital-Weill Medical College of Cornell University, Westchester
Division, 21 Bloomingdale Road, White Plains, NY 10605; E-mail: email@example.com.
Association, 1994) criteria for BPD (Skodol, et al., 2001). Suicidal
(McGlashan, 1986; Stone, 1993) and self-injurious behavior is particularly
prevalent among BPD patients, with rates ranging from 69% to 75%
(Cowdry, Pickar, & Davies, 1985; Clarkin, Widiger, Frances, Hurt, &
Psychotherapy is the most widely practiced technique for treating border-
line patients. Among other common treatment approaches to BPD is the ob-
ject relations approach based on Kernberg’s clinical theorizing (1984,
1996). Kernberg and his colleagues have published a treatment manual de-
scribing a modified psychodynamic treatment of patients with borderline
personality organization called Transference-Focused Psychotherapy (TFP;
Clarkin, Yeomans, & Kernberg, 1999).
TFP relies principally on the techniques of clarification, confrontation,
and interpretation within the evolving transference relationship between
the patient and the therapist. The primary focus of TFP is on the dominant
affect-laden themes that emerge in the relationship between borderline pa-
tients and their therapists in the here-and-now of the transference. During
the first year of treatment, TFP focuses on a hierarchy of issues: the contain-
ment of suicidal and self-destructive behaviors, the various ways of destroy-
ing the treatment, and the identification and recapitulation of dominant
object relational patterns, as they are experienced and expressed in the
here-and-now of the transference relationship. Although psychoanalytic
psychotherapy based on Kernberg’s theory is a widely practiced technique
for treating BPD, research concerning the effectiveness and efficacy is lim-
ited and is greatly needed.
THE PRESENT STUDY
We report findings from an NIMH-funded treatment development study ex-
amining pre-post changes observed in the 1-year outpatient treatment of
borderline patients with TFP. Our primary hypotheses were that subjects
would show a significant reduction in the number and severity of suicidal
and self-injurious behavior; a decrease in physical harm resulting from sui-
cidal and self-injurious behavior; a significant reduction in hospitaliza-
tions, emergency room visits, and number of days hospitalized; and
improved psychosocial functioning.
Subjects were recruited from all treatment settings (i.e., inpatient, day hos-
pital, and outpatient clinics) within the New York-Presbyterian Hospi-
tal—Weill Cornell Medical Center system. Written informed consent was
obtained after all study procedures had been explained. Potential subjects
were screened with both clinical and semi-structured interviews. Women
who met the following selection criteria were eligible for the study: (1) five or
more DSM-IV criteria for BPD as assessed on the SCID-II; (2) at least two in-
cidents of suicidal or self-injurious behavior in the last 5 years; (3) absence
488 CLARKIN ET AL.
of DSM-IV criteria for schizophrenia, bipolar disorder, organic pathology,
and/or mental retardation as assessed using the SCID-I; (4) between the
ages of 18 and 50; and (5) agreement to the study conditions, including ter-
mination from other individual psychotherapy. Upon admission to the
study, patients were given a number of additional assessment instruments
described in detail below. Subjects were reevaluated after 12 months of
Twenty-three patients met criteria for BPD, agreed to the study conditions,
and entered the treatment. Two patients dropped out around the 4-month
mark and two patients dropped out after 8 months of treatment. An addi-
tional two patients were administratively discharged from TFP early in the
process because of protocol violations (i.e., consistent failure to adhere to
the treatment contract despite verbal assurances to the contrary).
The treatment group (N = 17) had a mean age of 32.7 years (SD = 7.52;
range of 19 to 48). Thirteen (76.5%) subjects were Caucasian, and four
(23.5%) were Hispanic. Ten were single and never married, four were mar-
ried, and three divorced. At the beginning of the treatment period, eight
were unemployed, seven worked at the technical or clerical level, one was a
minor professional, and one was a student/homemaker. Most subjects met
criteria for more that one Axis I disorder and at least one Axis II personality
disorder. The most common clinical Axis I diagnoses in this sample were:
major depression (n = 8; 47.1%); dysthymia (n = 4; 23.5%) and eating disor-
der (n = 3; 17.6%). The most common Axis II conditions comorbid with BPD
were narcissistic (82%), paranoid (76%), obsessive compulsive (71%), and
avoidant personality disorder (65%).
The therapists ranged from six experienced, senior individuals with at least
10 years of experience, to six faculty/staff psychologists and postdoctoral
trainees in psychology, all of whom had 2 or more years of experience treat-
ing BPD patients with psychodynamic treatment and training in TFP. All
therapists selected for this phase of the study were judged by independent
supervisory ratings to be both competent and adherent to the TFP manual.
Throughout the study, all therapists regularly videotaped sessions and
were supervised on a weekly basis. Consensual adherence and competence
ratings were made during the weekly supervision sessions.
The Parasuicidal History Interview. (PHI; Linehan, Wagner, & Cox, 1989)
was used to assess the number of suicidal and parasuicidal behaviors as
well as the medical severity and physical outcomes of these behaviors for
the year prior to treatment and during the treatment year. The methodology
for assessing medical risk and physical condition was derived using the
scales described by Linehan (Linehan, Wagner, & Cox, 1989).
PSYCHODYNAMIC TREATMENT FOR BPD 489
The Treatment History Interview. (THI; Linehan, 1987) was used to as-
sess the types and amount of treatment received during the target period,
including emergency room visits, and number and length of psychiatric
The Global Assessment of Functioning Scale. (GAF; APA, 1994) provides
a single global rating of functioning and symptomatology. We used a modi-
fied version of the GAF included in DSM-III-R. In this version, scores range
from a low of 1 (e.g., needs constant supervision, serious suicide act with
clear intent and expectation of death) to a high of 90 (e.g., superior function-
ing in a wide range of activities, no symptoms).
Two sets of analyses were conducted. The first set compared pre- and
post-treatment scores in the intent-to-treat group, that is, all patients who
agreed to enter treatment. The intent-to-treat group included the 17 pa-
tients who completed TFP, the 4 therapy dropouts, and the 2 subjects who
were administratively discharged (N= 23). In order to be statistically conser-
vative, pretreatment scores were carried forward in those cases where
post-test scores were not available. The two subjects who were administra-
tively discharged were only available at pre-treatment. The second set com-
pared patients who completed TFP (N=17) pre- and post-treatment. For both
analyses, three repeated multivariate analyses of variance measures
(MANOVAs) were performed, one on the set of suicide variables (number of
suicide attempts, average medical risk of all attempts, average resulting
physical condition after the attempts), one on the set or parasuicide vari-
ables (number of parasuicide attempts, average medical risk of all attempts,
average resulting physical condition after the attempts), and finally, one for
the set of treatment utilization variables (number of hospitalizations and
length of hospitalizations). In each MANOVA, time was the repeated factor
and there was no between-subjects grouping variable. Individual treatment
effects were considered significant only when the overall multivariate model
was significant and the
level for the individual effect was less than 0.05. All
comparisons employed two-tailed tests. For each outcome variable, we also
investigated the effect size that treatment had on outcome using a formula
provided by Cohen (1992). In that formula, effect size equals the difference
in mean score pre- and post-treatment divided by the pre-therapy standard
Before analyzing pre-post changes for the intent-to-treat group and treat-
ment completers, demographic and study variables were compared for pa-
tients who completed treatment and those who declined to enter treatment,
dropped out, or were administratively discharged using chi-square analysis
and t-test for independent samples. There were no significant differences
490 CLARKIN ET AL.
between the groups in terms of age, education, employment, marital status,
ethnicity, or religion. Similarly, no significant differences were found be-
tween the groups in distribution of Axis I and II diagnoses except for the di-
agnosis of schizoid personality disorder based on the SCID-II-Q. Those in
the treatment group were more likely to meet criteria for schizoid personal-
ity disorder (treatment completers = 6 [35%]; treatment decliners, dropouts,
and discharged = 0;
= 6.52, p
.01). We also compared the two groups on
the amount of psychotropic medications prescribed and used. Results of
chi-square analyses indicated that at pre-treatment there were no be-
tween-group differences in the number of subjects using psychotropic med-
For both the intent-to-treat and the completer analyses, the overall
multivariate model was not significant for the set of suicide variables. Al-
though the multivariate results were not significant, we did observe a sugges-
tive decrease in the number of suicide attempts from the year prior to the
treatment year (see Table 1). In addition, a decrease is also reflected in the
number of patients who made suicide attempts. In the prior year, 9 of the 17
(53%) patients made a suicide attempt, and during the 1-year treatment, only
3 of 17 (18%) made such an attempt (McNemar’s test (1) = 4.64, p =
For the completer analyses, the overall multivariate model was significant
for the parasuicide variables (Pillai’s trace = 0.42, F(3,14) = 3.38, p
the intent-to-treat group, this model approached significance (Pillai’s trace
= 0.31, F(3,20) = 3.01, p
.06). Means, standard deviations, F tests, and ef-
fect sizes from subsequent univariate tests are shown in Table 1. There was
not a significant decrease in the number of self-injurious behaviors, but
there was a significant decrease in average medical risk and average physi-
cal condition following such incidents. The intent-to-treat results parallel
these results for completers.
With regard to service utilization, the overall MANOVA was significant for
the completer group (Pillai’s trace = 0.35, F(2,15) = 3.97, p
.05) as well as
the intent-to-treat group (Pillai’s trace = 0.26, F(2,21) = 3.61, p
Means, standard deviations, F tests, and effect sizes from subsequent
univariate tests for these models are shown in Table 1. There was a signifi-
cant reduction in the number of hospitalizations (72%) for the completer
group, with the reduction in days hospitalized approaching significance
.06). Parallel results were obtained for the intent-to-treat group.
The magnitude of effect was investigated by calculating the effect size (d’) for
each of the variables. Almost all of the effect sizes indicated favorable
change. The average effect sizes were 0.38 and 0.56 for the intent-to-treat
analyses and completer analyses, respectively.
PSYCHODYNAMIC TREATMENT FOR BPD 491
We examined the treatment outcome for patients diagnosed with BPD who
were treated in a 1-year modified psychodynamic outpatient psychother-
apy. The major finding in this study is that TFP appears to be a promising
psychotherapeutic technique that warrants additional research. In both the
intent-to-treat and treatment completion groups, borderline patients re-
ceiving TFP showed considerable improvement in a number of important ar-
The 1-year dropout rate was low (19.1%; 4 of 21 of patients dropped out of
treatment) and no patient committed suicide. This dropout rate compares
well with previous studies (Bateman & Fonagy, 1999; Linehan, Armstrong,
Suarez, Allmon, & Heard, 1991; Linehan, Schmidt, Dimeff, & Craft et al.,
1999; Stevenson & Meares, 1992), which reported a range between 16.7% to
21.0%. Additionally, none of the treatment completers deteriorated or were
adversely affected by the treatment. Therefore, it appears that TFP is toler-
ated quite well.
There was a significant reduction in the number of patients who had made
a suicide attempt during the treatment year compared with the year prior to
treatment (18% vs. 53%). However, the number of suicide attempts, the
medical risk of these acts, and the patients’ physical condition afterward
were not significantly improved. Although non-suicidal self-injurious be-
492 CLARKIN ET AL.
TABLE 1. Univariate Tests on the Intent-to-Treat Group and the Computer Group
Intent-To-Treat Analysis (N = 23)
M SD M SD F df p d prime
Number of Incidents
4.39 6.34 3.44 4.57 0.60 1,22 0.45 0.15
2.06 1.17 1.62 1.24 6.88 1,22 0.02 0.37
Physical Condition 2.10 1.24 1.54 1.17 8.46 1,22 0.01 0.46
Pillai’s trace = 0.31,
(3,20) = 3.01,
Hospitalizations 1.48 1.59 0.83 1.4 6.89 1,22 0.02 0.41
Days Hospitalized 55.33 84.32 29.7 70.83 4.21 1,22 0.06 0.31
Pillai’s trace = 0.26, F(2,21) = 3.61,
Completer Analysis (N = 17)
M SD M SD F df p d prime
Number of Incidents 5.18 7.25 4.24 5.08 0.6 1,16 0.45 0.13
1.72 1.13 1.14 0.99 7.61 1,16 0.02 0.51
1.89 1.31 1.12 0.99 9.64 1,16 0.01 0.58
Pillai’s trace = 0.42,
(3,14) = 3.38,
1.24 1.35 0.35 0.61 7.63 1,16 0.02 0.61
Days Hospitalized 39.21 67.03 4.53 9.61 4.45 1,16 0.06 0.52
Pillai’s trace = 0.35,
(2,15) = 3.97,
havior did not decrease in frequency, the medical risk was significantly re-
duced, and the physical condition of the patients was significantly
Compared to the year prior, study patients during the treatment year had
significantly fewer psychiatric hospitalizations (67% reduction) and days of
inpatient hospitalization (89% reduction). While 64.7 % of patients were
hospitalized the year prior, only 29.4 % were hospitalized during the treat-
Two other points are worth noting. First, TFP relies principally on trans-
ference interpretations, which are controversial (Piper, Azim, Joyce, &
McCallum, 1991; Gabbard, 1991). Future research should examine the role
of transference interpretation more directly as one of the specific mecha-
nisms of change in TFP. Second, Linehan et al. (1991), Stevenson and
Meares (1992), Bateman and Fonagy (1999), and this study all employed
close supervision, a team approach, and provided structured treatments.
Future research should examine these as common factors in the successful
treatment of BPD.
As a preliminary study, there are a number of design issues that limit the
interpretation and the generalizability of our results. First, the absence of a
comparison group limits the interpretation of positive change, as the
changes we observed in our patients may have occurred over time without
TFP or even without treatment. However, previous research with borderline
patients has found fair stability over 2- to 5-year periods in terms of the di-
agnosis and severity of emotional difficulties (Hoke, Lavori, & Perry, 1992;
Vaglum, Friis, Karterud, Mehlum, & Vaglum, 1993). Nevertheless, a ran-
domized controlled trial of TFP would constitute a more stringent test of the
efficacy of this treatment, and with the positive results presented here we
are proceeding to such a study.
The sample in the present study was a relatively homogeneous group of
severely disturbed, chronically self-destructive borderline women. It is un-
clear if our results would generalize to less severely disturbed borderline in-
dividuals or whether the treatment would be as effective for men.
Another important issue with regard to generalizability concerns the fact
that none of our treated patients was currently abusing substances at the
time of entry into the study. A number of studies have found a high preva-
lence of alcohol and substance use in individuals with BPD. Although none
of our patients met criteria for alcohol or substance dependence at the be-
ginning of treatment, more than half of our treatment completers (N = 9) had
significant drug use/abuse histories. Therefore, while our findings may not
generalize to patients with current alcohol or substance dependence, cer-
tainly our findings are generalizable to borderline patients with significant
drug and/or alcohol histories.
This study, which used the patients as their own controls, is highly sugges-
tive that extended TFP is well-tolerated and may result in considerable im-
provement in functioning in a broad range of areas. Based on the present
findings, future research with TFP is warranted. Future studies should in-
PSYCHODYNAMIC TREATMENT FOR BPD 493
clude a randomized controlled trial of TFP, as well as an exploration of the
treatment process and underlying mechanisms of action that result in
change for these patients (Shea, Benjamin, Clarkin, & Magnativa, 1999).
Additionally, given the severity and chronicity of BPD, follow-up data is im-
perative to establish the long-term significance of these findings.
494 CLARKIN ET AL.
American Psychiatric Association (1994). Di-
agnostic and statistical manual of men-
tal disorders, 4th edition. Washington,
Bateman, A., & Fonagy, P. (1999). Effective-
ness of partial hospitalization in the
treatment of borderline personality dis-
order: A randomized controlled trial.
American Journal of Psychiatry, 156,
Clarkin, J. F., Widiger, T., Frances, A., Hurt,
S. W., & Gilmore, M. (1983). Prototypic
typology and the borderline personality
disorder. Journal of Abnormal Psychol-
ogy, 92, 263-275.
Clarkin, J. F., Yeomans, F., & Kernberg, O. F.
(1999). Psychotherapy of borderline
personality. New York: Wiley.
Cohen, J. (1992). A power primer. Psychology
Bulletin, 1(1), 155-159.
Cowdry, R. W., Pickar, D., & Davies, R.
(1985). Symptoms and EEG findings in
the borderline syndrome. International
Journal of Psychiatry in Medicine, 15,
Gabbard, G. O. (1991). Technical approaches
to transference hate in the analysis of
borderline patients. International Jour-
nal of Psychoanalysis, 72, 625-637.
Hoke, L. A., Lavori, P. W., & Perry, J. C.
(1992). Mood and global functioning in
borderline personality disorder: Indi-
vidual regression models for longitudi-
na l mea surem e nt s . Jo u rn a l of
Psychiatric Research, 26(1), 1-16.
Kernberg, O. F. (1984) Severe personality dis-
orders: Psychotherapeutic strategies.
New Haven: Yale University Press.
Kernberg, O. F. (1996). A psychoanalytic the-
ory of personality disorders. In J. F.
Clarkin & M. F. Lenzenweger (Eds.),
Major theories of personality disorder
(pp. 106-140). New York: Guilford
Linehan, M. M. (1987). Treatment History In-
terview (THI). Seattle: University of
Linehan, M. M., Armstrong, H. E., Suarez, A.,
Allmon, D., & Heard, H. L. (1991). Cog-
nitive-behavioral treatment of chroni-
cally parasuicidal borderline patients.
Archives of General Psychiatry, 48,
Linehan, M. M., Schmidt, H., Dimeff, L. A.,
Craft, J. C., Kanter, J., & Comtois, K. A.
(1999). Dialectical behavior therapy for
patients with borderline personality
disorder and drug-dependence. Ameri-
can Journal on Addictions, 8, 279-292.
Linehan, M. M., Wagner, A. W., & Cox, G.
(1989). Parasuicide History Interview:
Co mp re hen s i ve asse s sm ent of
parasuicidal behavior. Seattle: Univer-
sity of Washington.
McGlashan, T. H. (1986). The Chestnut Lodge
follow-up study: III. Long-term out-
come of borderline personalities. Ar-
chives of General Psychiatry, 43, 20-30.
Oldham, J. M., Gabbard, G. O., Goin, M. K.,
Gunderson, J., Soloff, P., Spiegel, D.,
Stone, M., & Phillips, K. A. (2001). Prac-
tice guideline for the treatment of pa-
ti en ts wi th bo r der lin e dis order.
Supplement to American Journal of
Psychiatry, 158, 1-52.
Piper, W. E., Azim, H. F. A., Joyce, A. S., &
McCallum, M. (1991). Transference in-
terpretations, therapeutic alliance, and
outcome in short-term individual psy-
chotherapy. Archives of General Psychi-
atry, 48, 946-953.
Shea, M. T., Benjamin, L. S., Clarkin, J. F., &
Magnavita, J. (1999). Personality disor-
ders: A discussion of current status
and future directions for research,
practice, and policy. In Session: Psy-
ch o th e rap y in P ra c ti ce , 5 5 (11),
Skodol, A. E., Gunderson, J. G., Livesley, W.
J., Pfohl, B., Siever, L. J., & Widiger, T.
A. (2000). The borderline diagnosis from
the perspectives of psychopathology,
comorbidity, personality structure, biol-
ogy, genetics, and course. Manuscript
submitted for publication.
PSYCHODYNAMIC TREATMENT FOR BPD 495
Stevenson, J., & Meares, R. (1992). An out-
come study of psychotherapy for pa-
tients with borderline personality
disorder. American Journal of Psychia-
try, 149, 358-362.
Stone, M. H. (1993). Long-term outcome in
personality disorders. British Journalof
Psychiatry, 162, 299-313.
Vaglum, P., Friis, S., Karterud, S., Mehlum,
L., & Vaglum, S. (1993). Stability of the
severe personality disorder diagnosis:
A 2- to 5-year prospective study. Jour-
nal of Personality Disorders, 7(4),
Weissman, M. M. (1993). The epidemiology of
personality disorders. In R. Michels, A.
M. Cooper, S. B. Guze, L. L. Judd, A. J.
Solni t, A. J. Stunkard, & M. M.
Weissman (Eds.), Psychiatry (vol. 1,
chapter 15.2, pp. 1-11). Philadelphia: