Evaluating Three John F. Clarkin, Ph.D.

Kenneth N. Levy, Ph.D.

Mark F. Lenzenweger, Ph.D.
Otto F. Kernberg, M.D.

Treatments for
Borderline Personality
Disorder: A Multiwave Study
Objective: The authors examined three yearlong outpatient treatments for borderline personality disorder: dialectical
behavior therapy, transference-focused psychotherapy, and a dynamic supportive treatment. Method: Ninety patients
who were diagnosed with borderline personality disorder were randomly assigned to transference-focused psychotherapy,
dialectical behavior therapy, or supportive treatment and received medication when indicated. Prior to treatment and at
4-month intervals during a 1-year period, blind raters assessed the domains of suicidal behavior, aggression, impulsivity,
anxiety, depression, and social adjustment in a multiwave study design. Results: Individual growth curve analysis
revealed that patients in all three treatment groups showed significant positive change in depression, anxiety,
global functioning, and social adjustment across 1 year of treatment. Both transference-focused psychotherapy and
dialectical behavior therapy were significantly associated with improvement in suicidality. Only transference-focused
psychotherapy and supportive treatment were associated with improvement in anger. Transference-focused

PUBLICATIONS
INFLUENTIAL
psychotherapy and supportive treatment were each associated with improvement in facets of impulsivity. Only
transference-focused psychotherapy was significantly predictive of change in irritability and verbal and direct assault.
Conclusions: Patients with borderline personality disorder respond to structured treatments in an outpatient setting
with change in multiple domains of outcome. A structured dynamic treatment, transference-focused psychotherapy was
associated with change in multiple constructs across six domains; dialectical behavior therapy and supportive treatment
were associated with fewer changes. Future research is needed to examine the specific mechanisms of change in these
treatments beyond common structures.

(Reprinted with permission from The American Journal of Psychiatry 2007;164:922–928)

Impulsivity, diminished nonaffective constraint, Other therapeutic approaches, such as psycho-

negative affectivity, and emotional dysregulation are dynamic treatments, continue to be prominent in
core characteristics of borderline personality disorder the treatment of borderline personality disorder, as
(1–3). The prevalence of borderline personality supported by the APA Practice Guideline (8) and
disorder in the community is approximately 1.3% prior research (12). A promising psychodynamic
to 1.4% (4, 5). This chronic and debilitating syn- treatment approach is an object relations approach
drome is associated with high rates of medical and called transference-focused psychotherapy (13).
psychiatric utilization of services (6, 7). Psychophar- Transference-focused psychotherapy is an effective
macology notwithstanding, psychotherapy repre- treatment using patients as their own comparisons
sents the recommended primary technique for (14) and has demonstrated superiority over treat-
treating borderline personality disorder (8). Di- ment as usual (unpublished data by KN Levy et al.
alectical behavior therapy (9) has demonstrated available from the authors).
superiority over treatment as usual (10) and ther- A necessary and first step in illuminating effective
apy by community experts (11). treatments for borderline personality disorder is to

focus.psychiatryonline.org FOCUS Spring 2013, Vol. XI, No. 2 269

 CLARKIN ET AL.

personality disorder. Individuals with comorbid

Measures and Related
Table 1. psychotic disorders, bipolar I disorder, delusional
Outcome Domains disorder, delirium, dementia, and/or amnestic as
Measure Outcome Domain well as other cognitive disorders were excluded.
Overt Aggression Scale- Suicidality Those with active substance dependence were also
Modified (34) excluded, although patients with past substance
Anger, Irritability, and Assault Aggression dependence and past and current substance abuse
Questionnaire (35) were included. Patients were recruited from New
Barratt Impulsiveness Impulsivity York City and the adjacent tri-state area.
Scale-II (29)
Brief Symptom Inventory (36) Anxiety
ASSESSMENTS AND PROCEDURES
Beck Depression Inventory (37) Depression
Global Assessment of Functioning Social Adjustment Patients were screened for age and location in
Scale (18) telephone interviews. Suitable individuals were
Social Adjustment Scale (38) Social Adjustment assessed in face-to-face evaluations by trained eval-
uators prior to their assignment to treatment. Each
participant received a diagnosis of borderline per-
sonality disorder based on DSM-IV criteria as
show that a given treatment is associated with sig- assessed by the International Personality Disorders
nificant improvement in the disorder—improvement Examination (17). High levels of reliability were
in relevant dimensions of pathology beyond self- obtained for the number of DSM-IV borderline
damaging behaviors. Empirical evidence should personality disorder criteria met by each subject
show that candidate treatments, such as dialectical (single rater intraclass correlation coefficient [1, 1]5
behavior therapy and psychodynamic approaches, 0.83). An acceptable level of reliability for border-
are systematically related to change in a number of line personality disorder diagnosis was obtained
substantive domains of clinical significance. A recent (kappa50.64). Exclusion diagnoses were based on
influential review reported that existing therapies for the Structured Clinical Interview for DSM-IV
borderline personality disorder remain experimental, (SCID) (18).
and more “real-world” studies are necessary (15). Domains of outcome. A priori, we chose suici-
We examined patients who were taken from the dality, aggression, and impulsivity as primary out-
community and reliably diagnosed with border- come domains and anxiety, depression, and social
line personality disorder. Patients were randomly adjustment as secondary outcome domains. The
assigned to transference-focused psychotherapy, variables of interest were assessed at baseline and at
dialectical behavior therapy, or supportive treatment 4, 8, and 12 months (termination of treatment).
for 1 year. Our study has characteristics of both ef- Thus, each study participant was measured on the
ficacy and effectiveness studies. Similar to an efficacy same variables at approximately the same intervals at
study, we used random assignment to treatments, four points in time; however, the assessment intervals
manualized treatments, blind raters, therapists blind varied slightly. Measures and outcome domains are
to all baseline assessments, and specific and reliably provided in Table 1.
measured outcome variables. Similar to an effective- Treatment and therapists. After initial assess-
ness study, however, we included a range of border- ments, patients were randomly assigned to one of the
line personality disorder patients. Therapists provided three outpatient treatment conditions for 1 year.
treatment in their private offices in the community Additionally, all patients were evaluated for phar-
rather than in a university or hospital setting. Med- macotherapy at entry into the study by one of three
ication treatmentwas decided onan individualbasisso study psychiatrists who were blind to psychotherapy
that the use or nonuse of medication and medication assignment. A medication algorithm (19) was used
type and amount were not standardized. to guide the pharmacological treatment.
The three psychotherapies were delivered with
METHOD attention to preserving their integrity and ecological
validity (20). Therefore, dialectical behavior therapy
consisted of a weekly individual and group session
and available telephone consultation. Transference-
PARTICIPANTS focused psychotherapy consisted of two individual
The patients were men and women between the weekly sessions, and supportive treatment consisted
ages of 18 and 50 (mean age: 30.9 [SD57.85] years) of one weekly session, which could be supplemented
who met DSM-IV criteria (16) for borderline with additional sessions as needed.

270 Spring 2013, Vol. XI, No. 2 FOCUS THE JOURNAL OF LIFELONG LEARNING IN PSYCHIATRY
 CLARKIN ET AL.

The primary focus of transference-focused psy- linear modeling (full maximum likelihood) (27).
chotherapy is on the dominant affect-laden themes First, we conducted a set of “unconditional” growth
that emerge in the relationship between patient and analyses (24, 26), in which we posited a linear
therapist. It has been hypothesized that dialectical individual-change trajectory at level 1. Such un-
behavior therapy operates through the learning of conditional analyses partition the outcome var-
emotion regulation skills in the validating environ- iation into variance components that describe
ment provided by the treatment (9). Supportive the net variation in slope and intercept across
treatment (21) provides emotional support—advice individuals.
on the daily problems facing the patient with bor- We fitted two sets of “conditional” (level 2)
derline personality disorder. The therapist follows analyses that examined systematic interindividual
and manages the transference but explicitly does not differences in the intercept and slope values from
use interpretations. level 1. The first set of level 2 conditional analyses
In the present study, each of the three psycho- examined whether age at entry into the study pre-
therapies was administered and supervised by a dicted change. The second set of conditional anal-
treatment condition leader. Dialectical behavior yses represented the central analyses for the study,
therapy was supervised by Barbara Stanley, Ph.D., namely the use of the between-subjects factor of the
transference-focused psychotherapy by Frank treatment group (transference-focused psychother-
Yeomans, M.D., and supportive treatment by Ann apy, supportive treatment, dialectical behavior ther-
Appelbaum, M.D., all of whom are acknowledged apy) to explain differences in level 1 intercept and
experts. A total of 19 therapists were selected by the slope values. Importantly, the conditional model for
treatment condition leaders based on prior dem- these analyses deleted the intercepts in each of the
onstration of competence in their respective mo- two level 2 equations. Deletion of the intercepts from
dality. All therapists had advanced degrees in social the level 2 equations allows for easy inspection of the
work, psychology, or psychiatry, with at least 2 years fixed effects associated with each treatment. It also
of prior experience treating patients with borderline allows for straightforward evaluation of the null hy-
personality disorder. All therapists were monitored pothesis that membership in a given treatment group
and supervised weekly by treatment condition leaders is not associated with an intercept or slope that
who were available to observe videotaped sessions, differs from zero. The level 2 models yield “fixed
provide feedback, and rate therapists for adherence effects” in the prediction of the slope and intercept
and competence. values at level 1. Additionally, level 2 models yield

PUBLICATIONS
INFLUENTIAL
estimates of residual variance that describe re-
maining interindividual variability in the indi-
STATISTICAL ANALYSIS OF CHANGE
vidual slopes and intercepts after accounting for
We used individual growth curve analysis to in- the fixed effects (available upon request from the
vestigate change in the dimensions of symptoms and authors).
functioning over time (22–26). The individual Using each subject’s date of birth and exact as-
growth curve approach hypothesizes that, for each sessment dates, the time between assessments for
individual, the continuous outcome variable is each participant was calculated in months and then
a specified function of time called the individual centered on age at entry into the study for each
growth trajectory, plus error. This trajectory is participant. Centering the assessment intervals on age
specified as a simple linear function of time con- at entry and including age at entry as a predictor at
taining two important unknown individual growth level 2 account for each participant’s unique chro-
parameters—an intercept and a slope—that de- nological age when he or she began the study and
termine the shape of individual true growth over cause the individual level 1 intercepts to represent
time. The individual intercept parameter represents the true value of the first (time 1) assessments as the
the net “elevation” of the trajectory over time. The participants’ “initial status.” Individual growth curves
individual slope parameter represents the rate of were derived only for those participants who had
change over time and in this study is the within- completed three or four assessment waves.
person rate of change in the dependent variable over
time. Individual growth trajectories are specified at RESULTS
level 1 and capture individual change over time. A
level 2 model is then used to investigate the way that
the individual growth parameters at level 1 are re-
lated to between-subjects factors.
PATIENT CHARACTERISTICS

The hypothesized level 1 and level 2 statistical Between 1998 and 2003, 336 patients were re-
models were fitted simultaneously using hierarchical ferred to our project. Of the 336 referrals, 129 either

focus.psychiatryonline.org FOCUS Spring 2013, Vol. XI, No. 2 271

 CLARKIN ET AL.

to treatment and those who were not. Patient

Table 2. Patient Characteristics characteristics are provided in Table 2 (number of
Characteristic N % lifetime axis II disorders: mean52.49 [SD51.13]).
Gender More detail regarding participant referral and se-
Women 83 92.2
lection, rater and participant characteristics, and
assessment of reliability is available elsewhere (28).
Men 7 7.8
The present analyses are based on the patients for
Marital status
whom we obtained three or more data points,
Married 7 7.7 which included 23 transference-focused psycho-
Divorced 40 44.4 therapy patients, 17 dialectical behavior therapy
Living with partner 11 12.2 patients, and 22 supportive treatment patients,
In relationship 21 23.3 indicating continuation into the 9- to 12-month
Education period.
Less than high school 3 3.3
High school graduate 7 7.8 MEDICATION TREATMENT
Some college 28 31.1
Associate’s degree 6 6.7 At treatment onset, 70% of dialectical behavior
College degree 29 32.2
therapy, 65% of supportive treatment, and 52% of
transference-focused psychotherapy patients were
Graduate training 17 18.9
placed on medication. The percentage of patients
Employment
receiving medication remained relatively constant
Full-time 30 33.3 throughout the 1-year treatment period. Any differ-
Part-time 23 25.6 ence in the percentage of patients receiving medi-
Ethnicity cation in the three treatment cells cannot be attributed
Caucasian 61 67.8 to symptom severity, since there were no significant
African American 9 10.0 differences between the three groups of patients at
Hispanic 8 8.9 time 1 on the domain measures.
Asian 5 5.6
Other 7 7.8 INDIVIDUAL GROWTH CURVE ANALYSES
Lifetime axis I disorders
Any mood disorder 69 76.7
An unconditional growth model was fitted for all
domains, providing estimates of the average eleva-
Any anxiety spectrum disorder 43 47.8
tion and rate of change parameters and their natural
Any eating disorder 30 33.3
variation across all participants at entry into the
Drug/alcohol abuse/dependence 34 37.8 study. The estimated average elevation of the in-
Suicidal behavior dividual growth trajectories at entry into the study
Prior suicidal behavior 51 56.7 (intercepts) differed significantly from zero for all
Prior parasuicidal behavior 56 62.2 domain dimensions (all p,0.001). These results
No history of suicidal/ 15 16.7 were expected, since the borderline personality
parasuicidal behavior disorder patients were relatively impaired. Notably,
the estimated average rates of change (i.e., slopes)
also differed significantly from zero for all of the
did not meet criteria or decided not to schedule an domain dimensions, except for the Barratt Factor 3
intake interview. We interviewed 207 individuals impulsivity and anxiety dimensions (29), indicating
for at least one evaluation session; of these, 109 that much change over time was evident in the data
were eligible for randomization. Exclusions were (all p5#0.05).
because of the following reasons: did not meet The first set of level 2 conditional analyses ex-
criteria for borderline personality disorder (N534), amined whether the age at which a participant en-
age (N530), current substance dependence (N59), tered the study was related to change on the various
psychotic disorder (N58), bipolar I disorder (N56), domain variables, irrespective of treatment group.
IQ below 80 (N52), scheduling conflict (N51), Statistically, age at entry into the study was not
and dropouts (N58). Of the 109 participants el- significantly related to intercept (level) values for
igible for randomization, 90 were randomized to any of the domain dimensions. Age at entry into the
treatment. There were no differences in terms of study was also essentially unrelated to the slope (rate
demographics, diagnostic data, or severity of psy- of change) values for the domain dimensions, ex-
chopathology between those who were randomized cept for global functioning (p,0.02), suicidality

272 Spring 2013, Vol. XI, No. 2 FOCUS THE JOURNAL OF LIFELONG LEARNING IN PSYCHIATRY
 CLARKIN ET AL.

(past week) (p,0.05), and Barratt Factor 1 im- predicted significant improvement in 10 of the
pulsivity (p,0.02). Thus, differing developmental 12 variables, dialectical behavior therapy in five of
levels reflected, albeit crudely, through chrono- the 12 variables, and supportive treatment in six of
logical age were not related to elevation and rate of the 12 variables.
change in the domain dimensions. Therefore, age We also refit the level 2 models examining treat-
at entry was not included in the second set of level ment effects for the global functioning, suicidality,
2 conditional analyses. Additionally, age at entry and Barratt Factor 1 domains, including both the
did not differ significantly across the three treat- treatment group and age at entry variables. For both
ment groups (F51.26, df52, 59, p50.30). the global functioning and suicidality domains, in-
The second set of level 2 conditional analyses clusion of the age at entry variable made the sig-
investigated the impact of the three treatments on nificant associations between the three treatment
the level and rate of change (slope) in the participants. types and improvement in these domains stronger.
As noted previously, the intercept was deleted from Inclusion of age at entry for Barratt Factor 1 did not
each of the two level 2 equations in order to facilitate alter the nonsignificant relationships between the
direct examination of the effect of each treatment treatment types and this variable as found in the
group on the level and rate of change. In these an- initial conditional analyses (Table 2).
alyses, we determined whether the level and rate of The hierarchical linear model approach allows for
change observed for any of the treatment groups hypothesis testing using the method of contrast
differed significantly from zero. With respect to the analysis. We conducted one contrast analysis for each
rate of change (or slope), the null hypothesis would domain of analysis based on theoretical predictions
be that a treatment has no impact on change, and contained in the treatment manuals (15, 27) and the
therefore it should not be statistically related to results of prior research. In short, to conserve space,
change (or growth) over time. Table 3 contains the we found that only one contrast was close to statis-
fixed effects and variance components associated tically reliable, namely a contrast of slope coefficients
with each level 2 predictor (i.e., treatment group), that posited that transference-focused psychotherapy
the approximate p value for testing that these ef- and dialectical behavior therapy should show stron-
fects are zero in the population, and an estimate of ger relationships with decreases in suicidal behavior
the “effect size r” (30). than supportive treatment (and transference-focused
With respect to elevation, each of the three treat- psychotherapy and dialectical behavior therapy
ment groups was a statistically significant predictor of should not differ [i.e., transference-focused psy-

PUBLICATIONS
INFLUENTIAL
individual elevation parameters in the domain vari- chotherapy5dialectical behavior therapy.support-
ables (all p,0.001, “large” effect sizes). Given that ive treatment]). This contrast of slope coefficients
the cohort consisted of subjects diagnosed with approached significance (x252.17, df51, p,0.07;
borderline personality disorder, it was expected that one-tailed).
the elevations would all differ from zero. We also conducted an intent-to-treat analysis.
The critical growth parameter for this study is the Such an analysis is used to determine whether those
individual slope parameter as it directly indexes the participants dropping out of the various treatment
rate and direction of individual change over time. cells have affected the pattern of findings across the
The prediction of slope (change) at level 2 by each of treatments compared with when the analyses are
the three treatments was significant for depression, restricted to only those completing the study. We
anxiety, global functioning, and social adjustment carried forward the last known value for a variable for
(all p,0.05); the direction of effects was toward any participant who was assessed once or twice in the
symptom improvement. Both transference-focused protocol. For those participants who dropped out of
psychotherapy and dialectical behavior therapy were the study, we assumed assessments at 4, 8, or 12
significantly associated with improvement in sui- months as time intervals. Therefore, the intent-to-
cidality over time, and both transference-focused treat analysis database contained all those partic-
psychotherapy and supportive treatment were sig- ipants analyzed in the “completer” analyses as well as
nificantly associated with improvement in anger those who had data carried forward as described
over time. Only transference-focused psychotherapy above. In the intent-to-treat analysis, the same var-
was significantly predictive of symptom improve- iables of interest were analyzed using the identical
ment in Barratt Factor 2 impulsivity, irritability, hierarchical linear model procedure described pre-
verbal assault, and direct assault. Supportive treat- viously. The results for the intent-to-treat analysis
ment alone was predictive of improvement in Barratt did not differ in terms of the pattern of findings
Factor 3 impulsivity. None of the three treatments from those obtained through the “completer” anal-
was associated with improvement in Barratt Factor 1 yses. The p values attached to each of the signifi-
impulsivity. Thus, transference-focused psychotherapy cance tests for the coefficients for the intercept and

focus.psychiatryonline.org FOCUS Spring 2013, Vol. XI, No. 2 273

 CLARKIN ET AL.

Predicting Interindividual Differences in Change of Borderline Personality

Table 3.
Disorder Patients Across Three Treatments in Multiwave Perspectivea
Elevation (Intercept) of Rate of Change (Slope) of
Individual Trajectory (B0i) Individual Trajectory (B1i)
Fixed-Effect Fixed-Effect
Measure and Treatment Type Coefficient p Effect Size r Coefficient p Effect Size r
Depression (Beck Depression Inventory [37])
Transference-focused psychotherapy 33.41 0.001 0.93 20.55 0.001 0.50
Dialectical behavior therapy 35.91 0.001 0.91 20.59 0.003 0.38
Supportive treatment 34.91 0.001 0.93 20.69 0.001 0.49
Anxiety (Brief Symptom Inventory [36])
Transference-focused psychotherapy 50.03 0.001 0.98 20.34 0.004 0.37
Dialectical behavior therapy 53.23 0.001 0.97 20.57 0.001 0.50
Supportive treatment 49.16 0.001 0.98 20.40 0.001 0.48
Functioning (Global Assessment of
Functioning Scale [18])
Transference-focused psychotherapy 52.84 0.001 0.95 0.59 0.001 0.44
Dialectical behavior therapy 52.21 0.001 0.96 0.67 0.004 0.36
Supportive treatment 50.22 0.001 0.96 0.62 0.001 0.43
Suicidality (Overt Aggression Scale-Modified
[34])
Transference-focused psychotherapy 1.20 0.001 0.57 20.05 0.01 0.33
Dialectical behavior therapy 2.17 0.001 0.59 20.09 0.01 0.34
Supportive treatment 0.84 0.001 0.44 20.03 b
0.18
Social adjustment (Social Adjustment Scale
[38])
Transference-focused psychotherapy 4.47 0.001 0.92 20.04 0.03 0.28
Dialectical behavior therapy 4.75 0.001 0.92 20.09 0.001 0.44
Supportive treatment 5.04 0.001 0.98 20.09 0.001 0.59
Barratt Factor 1
Transference-focused psychotherapy 29.90 0.001 0.97 20.01 b
0.01
Dialectical behavior therapy 31.92 0.001 0.97 20.17 b
0.18
Supportive treatment 29.87 0.001 0.93 20.09 b
0.17
Barratt Factor 2
Transference-focused psychotherapy 13.41 0.001 0.96 20.09 0.005 0.36
Dialectical behavior therapy 13.73 0.001 0.91 20.05 b
0.14
Supportive treatment 13.32 0.001 0.96 20.03 b
0.12
Barratt Factor 3
Transference-focused psychotherapy 21.02 0.001 0.96 20.06 b
0.16
Dialectical behavior therapy 19.62 0.001 0.93 20.02 b
0.05
Supportive treatment 20.55 0.001 0.97 20.10 0.02 0.31
Irritability
Transference-focused psychotherapy 1.92 0.001 0.94 20.03 0.01 0.33
Dialectical behavior therapy 1.61 0.001 0.89 20.01 b
0.11
Supportive treatment 1.63 0.001 0.86 20.02 b
0.16
Anger
Transference-focused psychotherapy 1.74 0.001 0.87 20.06 0.001 0.44
Dialectical behavior therapy 1.52 0.001 0.80 20.03 b
0.25
Supportive treatment 1.34 0.001 0.82 20.03 0.05 0.28
Verbal assault
Transference-focused psychotherapy 1.80 0.001 0.94 20.04 0.001 0.43
Dialectical behavior therapy 1.55 0.001 0.87 20.02 b
0.21
Supportive treatment 1.49 0.001 0.88 20.02 b
0.19
Direct assault
Transference-focused psychotherapy 0.82 0.001 0.72 20.02 0.05 0.26
Dialectical behavior therapy 0.73 0.001 0.56 20.002 b
0.01
b
Supportive treatment 0.72 0.001 0.64 0.001 0.01
a
Level 2 analysis detected variability in change across individuals and determined the relationship between predictors and the elevation (intercept) and rate-of-change (slope)
components of each patient’s individual growth trajectory from the level 1 analysis. All components of the level 1 and 2 models were estimated simultaneously. Treatment types
were coded as yes51 and no50. Intercepts were deleted from the level 2 equations to enable entry of each treatment type without dummy coding. Barratt Factor 2 and 3
dimensions were scored so that increases in scores over time reflect decreasing impulsivity. Values represent the final estimates of the fixed effects with robust standard
errors. Fixed effects were tested to determine whether they differ from zero. Effect size was interpreted as 0.105small; 0.245medium; and 0.375large (30).
b
Not significant (p,0.05).

274 Spring 2013, Vol. XI, No. 2 FOCUS THE JOURNAL OF LIFELONG LEARNING IN PSYCHIATRY
 CLARKIN ET AL.

slope variables were, as one would expect, higher, the reduction of suicidal behaviors in borderline
given the greater degrees of freedom. These results personality disorder patients, were more effective
suggest that participant attrition did not substan- than a general supportive treatment.
tially alter the findings obtained in the completer Four design issues in this study deserve comment.
analyses. To our knowledge, this is the first randomized
One must consider the possible impact of medi- controlled trial design that examines three well-
cation on change, specifically whether medication described (manualized) treatments for borderline
interacted differentially with any of the three treat- personality disorder, one of which (dialectical be-
ments to amplify change for patients in one or more havior therapy) is considered by many to be a stan-
of the treatments. To examine this possibility, we dard treatment in the field, which represents a design
conducted a parallel set of growth curve analyses on of the highest level of control (32). Second, although
only those subjects who had been medicated at the transference-focused psychotherapy and supportive
study entry and through at least the third assessment. treatment share many of the same basic techniques,
In short, the pattern of findings (direction of effects, they differ in a way that allows for a component of
effect sizes) across the 12 dependent variables for the therapeutic control (transference-focused psycho-
three treatments in this restricted cohort where therapy employs transference interpretation as a
medication was held constant was highly similar to possible mechanism of change, whereas supportive
that reported in Table 3 for the entire cohort. treatment does not). Third, our design combines
features of both efficacy and effectiveness studies.
DISCUSSION Patients from a large tri-state metropolitan area were
referred to us, and inclusion/exclusion criteria were
The major finding of this randomized controlled based on criteria used in clinical practice. Patients
trial was that transference-focused psychotherapy, were assessed at a community-serving, university-
dialectical behavior therapy, and supportive treat- affiliated hospital, and the treatments were deliv-
ment showed some significant relation to positive ered by community practitioners in their private
change in multiple domains across 1 year of out- offices. One might expect attenuated outcomes with
patient treatment. This pattern suggests that these the treatments delivered in the community, without
structured treatments for borderline personality dis- the structure and the expectancies provided by a
order are generally equivalent with respect to broad university setting in which many randomized con-
positive change in borderline personality disorder trolled studies of this kind are conducted. Finally, we

PUBLICATIONS
INFLUENTIAL
(31). Nonetheless, some differences emerged across report both completer and intent-to-treat analyses
the three treatments in relation to change. For and thus address both the internal and external val-
the primary outcome variables, both transference- idity of the study. Completer analyses speak to the
focused psychotherapy and dialectical behavior issue of sufficient dose, whereas intent-to-treat anal-
therapy were significantly associated with improve- yses provide generalizability to the community.
ment in suicidality, whereas transference-focused There are three areas of limitation in the present
psychotherapy and supportive treatment were study. Similar to other statistical methods, statistical
associated with improvement in anger. Only power in multilevel modeling of growth curves is a
transference-focused psychotherapy was signifi- function of four factors: cohort size, variance in the
cantly predictive of symptom improvement in cohort, size of the effect being studied, and the
Barratt Factor 2 impulsivity, irritability, verbal as- number of data waves available. Although reliable
sault, and direct assault. Supportive treatment alone a priori effect sizes were not available for our three
was predictive of improvement in Barratt Factor 3 treatments and we did not wish to constrain the
impulsivity. Regarding secondary outcome vari- patient characteristics in an artificial manner to re-
ables, each of the three treatments was significantly duce variability, we feel that a larger cohort and more
predictive of the rate of change in a positive di- data waves might have yielded more significant
rection for depression, anxiety, global functioning, results because of increased power. Ethical con-
and social adjustment. Overall, transference-focused straints did not allow a no-treatment comparison
psychotherapy was predictive of significant improve- group for self-destructive and highly suicidal bor-
ment in 10 of the 12 variables across the six domains, derline personality disorder patients. In a similar
dialectical behavior therapy in five of the 12, and fashion, it was not possible to standardize medica-
supportive treatment in six of the 12. tions across these patients whose pathology called
Only one contrast analysis yielded a tendency for individualized medication treatments. However,
toward significance, which suggests that transference- we note that the pattern of results observed for the
focused psychotherapy and dialectical behavior ther- medicated subgroup of patients was highly similar to
apy, the two treatments with a specific focus on that observed for the full cohort.

focus.psychiatryonline.org FOCUS Spring 2013, Vol. XI, No. 2 275

 CLARKIN ET AL.

The general equivalence of outcome across the 13. Clarkin JF, Yeomans F, Kernberg OF: Psychotherapy of Borderline Person-
ality. New York, John Wiley and Sons, 1999
three treatments studied suggests that there may be 14. Clarkin JF, Foelsch PA, Levy KN, Hull JW, Delaney JC, Kernberg OF: The
different routes to symptom change in patients with development of a psychodynamic treatment for patients with borderline
personality disorder: a preliminary study of behavioral change. J Personal
borderline personality disorder. In contrast to di- Disord 2001; 15:487–495
alectical behavior therapy, which focuses directly on 15. Binks CA, Fenton M, McCarthy L, Lee T, Adams CE, Duggan C: Psychological
skills to help the patient regulate emotion and reduce therapies for people with borderline personality disorder. Cochrane Data-
base Syst Rev 2006; 25:CD005652
symptoms, transference-focused psychotherapy fo- 16. American Psychiatric Association: Diagnostic and Statistical Manual of
cuses on developing greater self-control through Mental Disorders, 4th ed. Washington, DC, 1994
17. Loranger A: International Personality Disorder Examination (IPDE) Manual.
the integration of representations of self and other Odessa, Fla, Psychological Assessment Resources, Inc., 1999
as they are activated in the relationship with the 18. First MB, Gibbon M, Spitzer RL, Williams JBW: User’s Guide for the Struc-
therapists. In this study, transference-focused psy- tured Clinical Interview for DSM-IV Axis I Research Version (SCID-1), version
2.0, New York, New York State Psychiatric Institute, Biometrics Research,
chotherapy effects were observed across a broader 1996
range of outcome domains than dialectical be- 19. Soloff PH: Psychopharmacology of borderline personality disorder. Psy-
chiatr Clin North Am 2000; 23:169–192
havior therapy and supportive treatment. Given 20. Elkin J, Pilkonis P, Docherty JP, Stosky SM: Conceptual and methodological
the pervasiveness and chronicity of borderline per- issues in comparative studies of psychotherapy and psychopharmacology,
I: active ingredients and mechanisms of change. Am J Psychiatry 1988;
sonality disorder, future research will need to ex- 145:909–917
amine the maintenance of these symptom changes 21. Rockland LH: Supportive Therapy for Borderline Patients: A Psychodynamic
and specification of the mechanisms through which Approach. New York, Guilford, 1992
22. Rogosa D, Brandt D, Zimowski M: A growth curve approach to the mea-
treatments relate to outcome (33). surement of change. Psychol Bull 1982; 92:726–748
23. Rogosa DR, Willett JB: Understanding correlates of change by modeling
individual differences in growth. Psychometrika 1985; 50:203–228
24. Singer JD, Willett JB: Applied Longitudinal Data Analysis: Modeling Change
REFERENCES and Event Occurrence. New York, Oxford University Press, 2003
25. Raudenbush SW, Bryk AS: Hierarchical Linear Models: Applications and
1. Depue RA, Lenzenweger MF: A neurobehavioral model of personality dis- Data Analysis Methods, 2nd ed. Thousand Oaks, Calif, Sage Publications,
turbance, in Major Theories of Personality Disorder, 2nd ed. Edited by 2002
Lenzenweger MF, Clarkin JF. New York, Guilford, 2005, pp 391–453 26. Lenzenweger MF, Johnson MD, Willett JB: Individual growth curve analysis
2. Paris J: The classification of personality disorders should be rooted in bi- illuminates stability and change in personality disorder features: the longi-
ology. J Personal Disord 2000; 14:127–136 tudinal study of personality disorders. Arch Gen Psychiatry 2004; 61:1015–
3. Trull TJ: Structural relations between borderline personality disorder features 1024
and putative etiological correlates. J Abnorm Psychol 2001; 110:471–481 27. Raudenbush SW, Bryk AS, Congdon R: HLM-6: Hierarchical Linear and Non-
4. Lenzenweger MF, Loranger AW, Korfine L, Neff C: Detecting personality linear Modeling. Lincolnwood, Ill, Scientific Software International, 2005
disorders in a nonclinical population: application of a two-stage procedure 28. Critchfield KL, Levy KN, Clarkin JF: The Personality Disorders Institute/
for case identification. Arch Gen Psychiatry 1997; 54:345–351 Borderline Personality Disorder Research Foundation randomized control
5. Lenzenweger MF, Lane M, Loranger AW, Kessler RC: DSM-IV personality trial for borderline personality disorder: reliability of axis I and II diagnoses.
disorders in the National Comorbidity Survey Replication (NCS-R). Biol Psychiatry Q 2007; 78:15–24
Psychiatry (in press) 29. Patton JH, Stanford MS, Barratt ES: Factor structure of the Barratt Impul-
6. Bender DS, Dolan RT, Skodol AE, Sanislow CA, Dyck IR, McGlashan TH, siveness Scale. J Clin Psychol 1995; 51:768–774
Shea MT, Zanarini MC, Oldham JM, Gunderson JG: Treatment utilization 30. Rosenthal R, Rosnow RL: Essentials of Behavioral Research: Methods and
by patients with personality disorders. Am J Psychiatry 2001; 158:295–302 Data Analysis, 2nd ed. New York, McGraw-Hill, 1991
7. Zanarini MC, Frankenburg FR, Hennen J, Silk KR: Mental health service 31. Roth A, Fonagy P: What Works for Whom? A Critical Review of Psychother-
utilization by borderline personality disorder patients and axis II comparison apy Research, 2nd ed. New York, Guilford, 2005
subjects followed prospectively for 6 years. J Clin Psychiatry 2004; 65: 32. Gabbard GO, Gunderson JG, Fonagy P: The place of psychoanalytic treat-
28–36 ments within psychiatry. Arch Gen Psychiatry 2002; 59:505–510
8. Oldham JM, Gabbard GO, Goin MK, Gunderson J, Soloff P, Spiegel D, Stone 33. Clarkin JF, Levy KN: Special issue: putative mechanisms of action in the
M, Phillips KA: Practice guideline for the treatment of patients with border- psychotherapy treatment of borderline personality disorder. J Clin Psychol
line disorder. Am J Psychiatry 2001; 158(suppl):1–52 2006; 62:405–516
9. Linehan MM: Cognitive-Behavioral Treatment of Borderline Personality Dis- 34. Coccaro EF, Harvey PD, Kupsaw-Lawerence E, Herbert JL, Bernstein DP:
order. New York, Guilford, 1993 Development of neuropharmacologically based behavioral assessments
10. Linehan MM, Armstrong HE, Suarez A, Allmon D, Heard H: Cognitive- of impulsive aggressive behavior. J Neuropsychiatry Clin Neurosci 1991;
behavioral treatment of chronically parasuicidal borderline patients. Arch 3(suppl):s44–s51
Gen Psychiatry 1991; 48:1060–1064 35. Coccaro EF, Kavoussi RJ: Fluoxetine and impulsive-aggressive behavior in
11. Linehan MM, Comtois KA, Murray AM, Brown MZ, Gallop RJ, Heard HL, personality disordered subjects. Arch Gen Psychiatry 1997; 54:1081–
Korslund KE, Tutek DA, Reynolds SK, Lindenboim N: Two-year randomized 1088
controlled trial and follow-up of dialectical behavior theapy vs. therapy by 36. Derogatis LR: Brief Symptom Inventory (BSI): Administration, Scoring, and
experts for suicidal behaviors and borderline personality disorder. Arch Gen Procedures Manual, 3rd ed. Minneapolis, National Computer Systems, 1993
Psychiatry 2006; 63:757–766 37. Beck AT, Steer RA, Brown GK: Manual for the Beck Depression Inventory,
12. Bateman A, Fonagy P: Effectiveness of partial hospitalization in the treat- 2nd ed. San Antonio, Tex, Psychological Corp, 1996
ment of borderline personality disorder: a randomized controlled trial. Am J 38. Weissman MM, Bothwell S: Assessment of social adjustment by patient
Psychiatry 1999; 156:1563–1569 self-report. Arch Gen Psychiatry 1976; 33:1111–1115

276 Spring 2013, Vol. XI, No. 2 FOCUS THE JOURNAL OF LIFELONG LEARNING IN PSYCHIATRY