See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6508904

The Therapeutic Alliance in Schema-Focused Therapy and Transference-

Focused Psychotherapy for Borderline Personality Disorder

Article in Journal of Consulting and Clinical Psychology · March 2007

DOI: 10.1037/0022-006X.75.1.104 · Source: PubMed

CITATIONS READS
130 3,465

5 authors, including:

Philip Spinhoven Richard van Dyck

Leiden University Vrije Universiteit Amsterdam
504 PUBLICATIONS 23,950 CITATIONS 330 PUBLICATIONS 17,066 CITATIONS

SEE PROFILE SEE PROFILE

Arnoud Arntz
University of Amsterdam
437 PUBLICATIONS 15,596 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development and validation of the Patient Doctor Relationship Questionnaire (PDRQ9) View project

Effects of a team-based intervention based on Schema Therapy on group climate and aggression in closed youth care View project

All content following this page was uploaded by Arnoud Arntz on 08 January 2014.

The user has requested enhancement of the downloaded file.

Journal of Consulting and Clinical Psychology Copyright 2007 by the American Psychological Association
2007, Vol. 75, No. 1, 104 –115 0022-006X/07/$12.00 DOI: 10.1037/0022-006X.75.1.104

The Therapeutic Alliance in Schema-Focused Therapy and Transference-

Focused Psychotherapy for Borderline Personality Disorder

Philip Spinhoven Josephine Giesen-Bloo

Leiden University University Maastricht

Richard van Dyck Kees Kooiman

Vrije Universiteit Amsterdam Leiden University Medical Center

Arnoud Arntz
University Maastricht

This study investigated the quality and development of the therapeutic alliance as a mediator of change
in schema-focused therapy (SFT) and transference-focused psychotherapy (TFP) for borderline person-
ality disorder. Seventy-eight patients were randomly allocated to 3 years of biweekly SFT or TFP. Scores
of both therapists and patients for the therapeutic alliance were higher in SFT than in TFP. Negative
ratings of therapists and patients at early treatment were predictive of dropout, whereas increasingly
positive ratings of patients in the 1st half of treatment predicted subsequent clinical improvement.
Dissimilarity between therapist and patients in pathological personality characteristics had a direct effect
on growth of the therapeutic alliance but showed no relationship with clinical improvement. The authors
conclude that the therapeutic alliance and specific techniques interact with and influence one another and
may serve to facilitate change processes underlying clinical improvement in patients with borderline
personality disorder.

Keywords: borderline personality disorder, therapeutic alliance, therapeutic relationship, schema-focused

therapy, transference-focused psychotherapy

The therapeutic alliance can be defined as the quality of involve- therapeutic alliance is necessary to detect possible similarities and
ment between therapist and patient as reflected in their task team- differences across orientations.
work and personal rapport (Orlinsky, Ronnestad, & Willutzki, The results of studies investigating the therapeutic alliance
2004). The quality of the alliance proves to be consistently asso- across different treatment modalities in mostly heterogeneous
ciated with a positive outcome across different forms of psycho- groups of patients with an Axis I disorder have been equivocal
therapy. Especially the therapist’s contribution to the alliance as (e.g., Brunink & Schroeder, 1979; Marmar, Gaston, Gallagher, &
rated by patients shows a consistent although modest relationship Thompson, 1989; Raue, Goldfried, & Barkham, 1997; Sloane,
with outcome (Martin, Garske, & Davis, 2000; Orlinsky et al., Staples, Cristol, Yorkston, & Whipple, 1975). Nevertheless, the
2004). Although the therapeutic alliance seems to be a common available evidence has suggested that compared with psychody-
therapy factor crucial to the change process across different ther- namic psychotherapy the alliance in cognitive behavior therapy is
apeutic orientations, comparative research of the quality of the more characterized by supportive communication, expressed sym-
pathy, and interpersonal contact. It is conceivable that there are
more pronounced differences in the therapeutic alliance in the
treatment of personality disorders instead of Axis I disorders. From
Philip Spinhoven, Department of Psychology and Department of Psy-
chiatry, Leiden University, Leiden, the Netherlands; Josephine Giesen- a cognitive– behavioral perspective on the treatment of personality
Bloo and Arnoud Arntz, Department of Medical, Clinical, and Experimen- disorders, a closer, warmer therapeutic alliance is deemed more
tal Psychology, University Maastricht, Maastricht, the Netherlands; necessary than in the treatment of an acute Axis I disorder such as
Richard van Dyck, Department of Psychiatry, Institute for Research in anxiety or depression (Beck, Freeman, & Associates, 1990; Beck,
Extramural Medicine, University Medical Center, Vrije Universiteit Am- Freeman, Davis, & Associates, 2004). Much of the therapist’s role
sterdam, Amsterdam, the Netherlands; Kees Kooiman, Leiden University consists of a process of reeducation, and in the course of time the
Medical Center, Leiden, the Netherlands. therapist even becomes a role model for the patient. In schema-
This research was funded by Grant OG-97.002 from the Dutch Health
focused therapy (SFT), therapists even try to provide “limited
Council.
Correspondence concerning this article should be addressed to Philip reparenting” to meet partially the unmet emotional needs in an
Spinhoven, Department of Psychology and Department of Psychiatry, effort to develop more healthy schemas (Young, Klosko, &
Leiden University, Wassenaarseweg 52, 2333 AK Leiden, the Netherlands. Weishaar, 2003). This is in contradiction to transference-focused
E-mail: spinhoven@fsw.leidenuniv.nl psychotherapy (TFP), in which an active supportive relationship

104
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 105

with the patient is considered contraproductive. In TFP, the pre- In psychotherapy outcome research, the major focus is on com-
dominance of negative transference reactions distorting the real paring the effectiveness of various theoretical and technical ap-
relationship, including the one with the therapist, is seen as the proaches while trying to control, reduce, or eliminate the influence
core pathology of the borderline patient. Supportive interventions of therapist factors. Therapist factors are controlled by construct-
are considered to interfere with the development of the negative ing treatment manuals that ideally can be applied identically by
transference or—in a less harsh scenario—to blur the negative any therapist to all patients within a particular diagnostic category
transference, creating an as-if world and making the negative (Lambert, 1989). As a result, relatively little attention has been
transference less amenable to therapeutic interventions. The de- given to the therapeutic impact of discrete therapist variables such
structive aggression of the patient, as manifested in the transfer- as age, experience and training, ethnicity, and gender, and only a
ence, however, is addressed early in treatment to protect the few studies have examined therapist variables in a relational or
treatment, enhance reality testing, and foster the development of a interpersonal context (Beutler et al., 2004).
more differentiated, realistic representation of important others as, Particularly with respect to personality traits, research has tried
for example, the therapist (Clarkin, Yeomans, & Kernberg, 1999). to define ways in which therapist and patient qualities mutually
Until now, no longitudinal research has been available in dif- interact, resulting in patterns of match and mismatch (Beutler et
ferent alliance qualities between psychodynamic versus cognitive– al., 2004). Two opposing viewpoints have been developed. The
behavioral therapy in the treatment of personality-disordered indi- conventional presupposition that similarity in personality traits
viduals. This study tries to advance earlier comparative research of increases the attachment and participation in treatment and conse-
the therapeutic alliance (a) by investigating a homogeneous group quently promotes outcome has been confirmed in some studies
of patients with a borderline personality disorder (BPD), (b) by (e.g., Herman, 1998), but support for the opposite presupposition
using two well-defined forms of cognitive– behavioral and psy- has also been found (e.g., Berry & Sipps, 1991). In this view, it is
chodynamic therapy with dissimilar therapeutic alliance qualities, argued that if the therapist’s and the patient’s dysfunctional per-
and (c) by studying the development of the therapeutic alliance sonality characteristics are harmoniously blended, this will result
during treatment. in a therapist’s blind spot severely hampering an adequate under-
The consistent relationship of the quality of the therapeutic standing of his or her emotional reactions to the patient and
alliance with outcome can be interpreted in different ways. Out- controlling these in the therapeutic relationship.
come and alliance measures may be confounded, and the quality of It is well known that the therapeutic alliance can be extremely
the alliance may even be a mere epiphenomenon of positive complicated in the treatment of BPD (Beck et al., 1990; Yeomans,
treatment change. However, in many studies, evidence for the Clarkin, & Kernberg, 2002; Young, 1994). Both within a cognitive
therapeutic alliance as a mediator of change has been collected and psychodynamic perspective, personal qualities of the thera-
showing that outcome can be predicted from early alliance ratings pists in their reciprocal relationship with the personal qualities of
(e.g., Gaston, Marmar, Gallagher, & Thompson, 1991; Salvio, the patient are assumed to be responsible for the quality and
Beutler, Wood, & Engle, 1992). Although it may be expected that development of the therapeutic alliance. Activation of therapists’
in the treatment of personality-disordered individuals early rating cognitive schemas by the expression of similar early maladaptive
of the therapeutic alliance will predict premature termination and schemas in patients (Beck et al., 1990; Young et al., 2003) or
outcome, it is conceivable that, in addition, growth of the thera- unconscious countertransference reactions elicited by the patient’s
peutic alliance during the first phase of treatment will facilitate primitive projective mechanisms (Clarkin et al., 1999) may pre-
later outcome. Especially in the treatment of BPD patients, the clude the functional processing of transference reactions in the
establishment of a therapeutic relationship is not readily accom- therapeutic relationship. Therefore, it is to be expected that the
plished because most of these patients’ problems are generally match of pathological personality characteristics of therapists and
manifested in the interpersonal realms. As a result, the develop- patients will impede the development of the therapeutic alliance.
ment and maintenance of a collaborative therapeutic alliance dur- By examining dissimilarity in pathological personality charac-
ing the first year of treatment as a prerequisite for further treatment teristics between therapists and patients, the development of the
is seen as one of the central issues of intensive long-term treatment therapeutic alliance, and the clinical outcome within one compre-
across different psychotherapeutic orientations (Beck et al., 1990; hensive analysis, the present study tries to improve on earlier
Clarkin et al., 1999). research on the impact of therapists’ personality traits on outcome
The investigation of the intertwined and sequential relationship (a) by assessing pathological personality traits of relevance for
between alliance and client improvement during treatment is seen borderline personality disorder, (b) by studying these traits in an
as an advancement compared with the research into early alliance interpersonal context, and (c) by investigating whether therapist–
scores as predictors of later outcome (Barber, Connolly, Crits- patient similarity in pathological personality characteristics indi-
Christoph, Gladis, & Siqueland, 2000; Klein et al., 2003). Until rectly impacts outcome by its direct effect on the therapeutic
now, no systematic research has been reported on the predictive alliance.
relationships among changes in the therapeutic alliance and out- In sum, the purpose of the present study was to investigate the
come in different phases of long-term treatment of personality following predictions: (a) The quality of the therapeutic alliance is
disorder. Consequently, the purposes of the present study are not rated higher in SFT than in TFP; (b) a lower quality of the
only to analyze (a) whether the quality of the therapeutic alliance therapeutic alliance at early treatment predicts premature treatment
at early treatment predicts dropout and outcome but also (b) termination and a worse clinical outcome; (c) growth of the ther-
whether there is any support of a causal role of growth of the apeutic alliance during the first year of therapy facilitates later
therapeutic alliance during the first phase of treatment in facilitat- clinical improvement; and (d) dissimilarity in pathological person-
ing later outcomes. ality characteristics between therapists and patients facilitates the
106 SPINHOVEN ET AL.

development of the therapeutic alliance and indirectly affects ther- reported elsewhere (Giesen-Bloo et al., 2006) and are summarized
apy outcome. in Figure 1.

Method Design
The present study was conducted as part of a multicenter- Patients’ first assessment was made after inclusion and before
randomized two-group design. Randomization to SFT or TFP was random allocation to treatment conditions. Then, assessments were
stratified over four community mental health centers, was carried made every 3 months for 3 years. Primary outcome measures were
out by a study-independent person, and was performed following administered at each assessment. The measurements for the quality
the adaptive biased urn procedure (Schouten, 1995). The study was of the therapeutic alliance were collected only after 3 months (i.e.,
conducted between September 1, 1999, and April 30, 2004. Details early treatment), after 15 months (i.e., midtreatment), and after 33
about participants, method, and results of this trial have been months (i.e., late treatment). Personality assessments of patients

Assessed for eligibility

N = 173

Excluded N = 85
Not meeting inclusion criteria
N = 24
Enrollment
Meeting exclusion criteria
N = 19
Randomized N = 88 Refused to participate
N = 40
Other reasons
N=2

Allocated to SFT Allocated to TFP

N = 45 N = 43
Received allocated intervention Received allocated intervention
N = 44 Allocation N = 42
Did not receive allocated intervention Did not receive allocated intervention
N=1 N=1
Give reasons: Give reasons:
Increased blindness Untraceable after randomization

Lost to follow-up N = 11 Lost to follow-up N = 22

N = 0 between 0 and 3 months N = 8 between 0 and 3 months
N = 5 between 3 and 15 months N = 7 between 3 and 15 months
N = 6 between 15 and 33 months Follow-Up N = 7 between 15 and 33 months

Completed N = 33 Completed N = 21
Analyzed (ITT) N = 44 Analysis Analyzed (ITT) N = 34
Excluded from analysis (N = 0) Excluded from analysis (N = 8)
(No 3-months alliance ratings)

Figure 1. Flow diagram of participant progress through the phases of the randomized trial. SFT ⫽ schema-
focused therapy; TFP ⫽ transference-focused psychotherapy; ITT ⫽ intention-to-treat.
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 107

took place at baseline, whereas therapists answered the personality tence Scale for BPD (Young, Arntz, & Giesen-Bloo, 2006). Both
questionnaires 3 months after the start of therapy. instruments consist of VAS- and Likert-scale items and have an
Independent trained research assistants assessed patients. Pa- identical cutoff score ⱖ 60. Fifty-six TFP tapes and 77 SFT tapes
tients’ diagnoses were assessed with the semistructured clinical of the second and/or sixth trimester were rated (intraclass corre-
interviews for the Diagnostic and Statistical Manual of Mental lation coefficients [ICCs] over 21 TFP and 20 SFT tapes that were
Disorders, (4th ed.; DSM-IV; First, Spitzer, Gibbon, & Williams, rated twice). Only an average of 7.5% of time (median 4%) was
1994, 1997). Patients were also screened with a semistructured spent on non-TFP techniques in TFP, and in SFT no non-SFT
clinical interview, the Borderline Personality Disorder Severity techniques were observed. The median competence/quality level
Index (4th version; BPDSI-IV; Arntz et al., 2003). A BPDSI-IV of applying TFP was 65.60 (ICC ⫽ 0.73) and of applying SFT was
cutoff score of 20 (range ⫽ 0 –90) discriminates BPD patients 85.67 (ICC ⫽ 0.69). The global competence rating median of the
from other personality pathology patients. Signed informed con- TFP therapists was 65.00 (ICC ⫽ 0.70) and of the SFT therapists
sent was obtained after full explanation of procedures and of both 73.00 (ICC ⫽ 0.78).
therapies but before the first assessment and randomization. Study
researchers, therapists, and research assistants had no foreknowl- Measures
edge of treatment allocation. The study protocol was approved by
the medical ethical committees of the four participating centers. Borderline Personality Disorder Severity Index (BPDSI-IV).
The primary outcome measure, the BPDSI-IV, is a DSM-IV BPD
Participants and Settings criteria-based semistructured interview and forms a quantitative
index of the current severity and frequency of specific BPD
Patients were referred by mental health institutes. Inclusion symptoms (Arntz et al., 2003). The interview covers a period of 3
criteria were a main diagnosis of BPD, age between 18 and 60 months, is suitable for use as a treatment outcome measure, and
years, a BPDSI-IV score above 20, and Dutch literacy. General shows excellent (interrater) reliability, validity, and sensitivity to
exclusion criteria were psychotic disorders (except short, reactive change. The internal consistency of the BPDSI-IV in the current
psychotic episodes), bipolar disorder, dissociative identity disor- study was .83.
der, antisocial personality disorder, attention-deficit/hyperactivity Working Alliance Inventory (WAI). The WAI (Horvath &
disorder, addiction of such severity that clinical detoxification was Greenberg, 1989) is one of the most commonly used and exten-
indicated (after which entering treatment was possible), psychiatric sively validated measures of the alliance. It is pantheoretical,
disorders secondary to medical conditions, and mental retardation. moderately correlated with other measures of the alliance, and has
been found to predict therapy outcome in numerous studies (Mar-
Treatment Conditions and Therapists tin et al., 2000; Orlinsky et al., 2004). The Dutch version of the
WAI consists of three subscales of 12 items each, rated on a
Both treatments were administered in biweekly 50-min sessions. 5-point instead of a 7-point Likert-type scale ranging from 1
Treatment protocols addressed the theoretical model, treatment (never) to 5 (always). The subscales based on Bordin’s (1979)
frame, different phases, and the use of strategies and techniques of working alliance theory address agreement about the goals of
SFT (Young, 1994; Young et al., 2003) and TFP (Clarkin et al., therapy, agreement about the tasks of therapy, and the bond
1999; Yeomans et al., 2002). Jeffrey Young (SFT) and Frank between the client and therapist. Patients completed the patient
Yeomans (TFP) trained the participating therapists before the start form (WAI-P) measuring the contribution of the therapist to the
of the study in SFT and TFP, respectively. Essential to both alliance as perceived by the patient, and therapists completed the
treatments is supervision. Weekly local peer supervision with 4 –5 therapist form (WAI-T), in which they rated the contribution of the
SFT or TFP therapists as well as 4-monthly 1-day central super- patient to the alliance. Because of the high intercorrelations among
vision and 9-monthly 2-day central supervision by Jeffrey Young subscales (WAI-P range ⫽ .69 –.88; WAI-T range ⫽ .67–.89),
(SFT) or Frank Yeomans (TFP) was provided in the study. subscale mean scores were added together to derive a global score.
A higher score on the WAI indicates a higher quality of the
Treatment Integrity Check working alliance. In the present study, the internal consistency of
the WAI-P was .94 and of the WAI-T was .95.
Randomly selected audio tapes of Sessions 1– 6 (for the TFP- Difficult Doctor–Patient Relationship Questionnaire—Ten Item
contract phase), and of each quarter, were saved for evaluation. All Version (DDPRQ-10). The DDPRQ (Hahn, Thompson, Stern,
raters were study independent and blinded for treatment outcome. Budner, & Wills, 1990) is a self-report questionnaire that aims to
One psychologist listened to one randomly selected tape of each measure the extent to which patients are experienced as frustrating
patient then stated the treatment administered and classified all or difficult in the therapeutic relationship by their doctor or ther-
tapes except one SFT tape correctly. Twenty-one (partial) TFP- apist and as provoking levels of distress that transcend the ex-
contract phases were rated by trained graduate students in psy- pected and accepted level of difficulty. Of the DDPRQ-10, five
chology on the Contract Rating Scale (Yeomans, Selzer, & Clar- items are about the therapist’s subjective experience (e.g., “Do you
kin, 1993). The contract setting adherence and competence had an find yourself secretly hoping that this patient will not return?”),
average rating of 3.22 (range ⫽ 2.86 –3.54), whereas a predeter- four are quasi-objective questions about the patient’s behavior
mined rating of 3 was considered adequate (Yeomans, Selzer, & (e.g., “How time consuming is caring for this patient?”), and one
Clarkin, 1989, 1993). Other trained therapists for each orientation item about symptoms combines elements of the patient’s behavior
assessed the TFP Rating of Adherence and Competence Scale and the therapist’s subjective response (i.e., “To what extent are
(Clarkin et al., 1999) or the SFT Therapy Adherence and Compe- you frustrated by this patient’s vague complaints?”). The items are
108 SPINHOVEN ET AL.

answered on a 6-point Likert-type scale ranging from 1 (not at all) scores were calculated for each subscale. A higher score on a scale
to 6 (a great deal). The DDPRQ was shown to be a reliable and indicates a higher level of this pathological personality character-
practical instrument in the physician–patient relationship. Difficult istic. Subscale mean scores were added together to derive a global
patients have been found to be characterized by psychosomatic score for impairments in personality organization. Internal consis-
symptoms, personality disorder, and Axis I (major) psychopathol- tencies of the IPO subscales in the current study varied from .72 to
ogy, and most had more than one of these characteristics (Hahn et .88 in the therapist sample and from .76 to .93 in the patient
al., 1996; Hahn, Thompson, Wills, Stern, & Budner, 1994). The sample.
total score of the DDPRQ equals the mean of the 10 items. A
higher score indicates a higher level of therapist frustration. The Statistical Analyses
internal consistency of the DDPRQ in the current study was .79.
Young Schema Questionnaire (YSQ). The YSQ is a 205-item A BPDSI-IV-based power analysis indicated that 45 patients per
self-report questionnaire developed to measure 16 core beliefs or group are needed to detect a 22% versus 50% recovery difference
early maladaptive schemas (Young, 1994). The items are answered between two groups by means of survival analysis (two-sided
on a 6-point Likert-type scale ranging from 1 (totally inapplicable significance level of 5% and a power of 80%; see Giesen-Bloo et
to me) to 6 (describes me perfectly). The 16 core beliefs are (1) al., 2006, for more details). Differences in the quality and devel-
abandonment/instability, (2) defectiveness/shame, (3) emotional opment of the therapeutic alliance between treatment conditions
deprivation, (4) mistrust/abuse, (5) social isolation, (6) depen- were analyzed with a 2 (group) ⫻ 3 (time) mixed factorial design
dence/incompetence, (7) vulnerability to harm and illness, (8) with repeated measures on the second factor.
enmeshment, (9) failure to achieve, (10) social undesirability, (11) Because in previous research a clinical cutoff score of 15
entitlement/grandiosity, (12) insufficient self-control/self- discriminated between BPD patients and nonpatient control par-
discipline, (13) self-sacrifice, (14) subjugation, (15) emotional ticipants (Arntz et al., 2003), with a sensitivity of 1 and specificity
inhibition, and (16) unrelenting standards. Schmidt, Joiner, Young, of 0.97, the recovery criterion was therefore defined as a
and Telch (1995) studied the YSQ in American patient and student BPDSI-IV score of less than 15 and maintenance of this score until
samples, and in the patient sample a 15-factor solution closely the last assessment. A second success criterion was the Jacobson
matching the rationally derived scales was found. The internal and Truax reliable change index (Jacobson & Truax, 1991). For
consistency of the subscales is sufficient to good (Stopa, Thorne, the BPDSI-IV, reliable change was achieved when a reduction of
Waters, & Preston, 2001; Waller, Meyer, & Ohanian, 2001), and at least 11.70 was achieved. By using logistic regression analyses,
findings support the discriminant validity of the YSQ, suggesting we investigated to what extent early treatment WAI-P, WAI-T,
that patients with different psychiatric diagnoses can be differen- and DDPRQ scores predicted these two outcome criteria above
tiated on the basis of their core beliefs (Stopa et al., 2001; Waller and independent of pretreatment BPDSI scores and treatment
et al., 2001; Waller, Shah, Ohanian, & Elliott, 2001). Moreover, condition. To detect time to dropout, survival analyses were con-
Rijkeboer, van den Bergh, and van den Bout (2005) reported ducted by using a proportional hazard approach to survival anal-
adequate rank-order stability and a high sensitivity of the Dutch ysis (Cox regression) with dropout as the dependent variable and
YSQ and its subscales in predicting the presence or absence of pretreatment BPDSI, treatment condition, and early treatment pro-
psychopathology in a clinical and nonclinical sample. Item mean cess variables as independent variables.
scores were calculated for each scale. A higher score on a scale To determine whether early to midtreatment changes in process
indicates a higher endorsement of dysfunctional core beliefs. Sub- variables predicted mid- to late treatment changes in outcome,
scale mean scores were summed up to derive a global score for cross-lagged correlations among residualized change scores were
dysfunctional core beliefs. Internal consistencies of the YSQ sub- calculated (Finkel, 1995). When a correlation between early pro-
scales in the present study varied from .77 to .94 in the therapist cess changes and later outcome changes was statistically signifi-
sample and from .78 to .93 in the patient sample. cant, hierarchical regression analyses were performed to test
Inventory of Personality Organization (IPO). The 90-item whether early process changes still predicted later outcome
IPO consists of three primary clinical and two secondary interper- changes after controlling for autocorrelations (i.e., the correlations
sonal relations scales. The IPO items have a 5-point Likert-type between early and late process changes) and synchronous corre-
scale ranging from 1 (never true) to 5 (always true). The three lations (i.e., the correlations between early process and early
primary clinical scales relevant to the central dimensions of Kern- outcome changes; cf., Burns, Kubilus, Bruehl, & Harden, 2003;
berg’s personality organization model (i.e., Reality Testing, 13 Evon & Burns, 2004). Regressions were also used to determine the
items; Identity Diffusion, 17 items; and Primitive Psychological inverse association.
Defenses, 14 items) have been psychometrically investigated. Differences between therapists and patients in the profile of
These scales display adequate internal consistency and good test– cognitive schemas or personality organization were analyzed ac-
retest reliability. Each of the scales is associated with increased cording to Cronbach and Gleser (1953). Therapist–patient differ-
negative affect, aggressive dyscontrol, and dysphoria as well as ence values were computed by using YSQ or IPO subscale scores
lower levels of positive affect consistent with Kernberg’s model of corrected for elevation and scatter. The D2 statistic represents the
borderline personality organization. Moreover, the Reality Testing sum of the squared differences on the subscales of the YSQ or IPO.
scale is closely related to various measures of psychotic-like The larger the obtained value, the greater the degree of dissimi-
phenomena (Lenzenweger, Clarkin, Kernberg, & Foelsch, 2001). larity between therapist and patient. To test the causal model
The two secondary interpersonal relations scales are also relevant linking dissimilarity in pathological personality characteristics,
for borderline character pathology (i.e., Pathological Object Rela- development of the therapeutic alliance, and outcome, we used the
tions, 38 items; and Superego Pathology, 12 items). Item mean analytic strategy as recommended by Baron and Kenny (1986).
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 109

Statistical analyses were performed in the intention-to-treat and Table 1

the completers sample. An intention-to-treat approach was applied, Biographical and Clinical Characteristics (Theoretical Ranges
either by using the last clinical endpoint during the 3-year treat- Between Parentheses)
ment period or by using the last clinical endpoint carried forward
for trend analyses. Missing data were substituted with a last- Schema- Transference-
focused focused
observation-carried-forward method. In addition, a completers therapy psychotherapy
analysis was performed. All tests were interpreted with a signifi- (n ⫽ 44) (n ⫽ 34)
cance level of 5%. The Statistical Package for Social Sciences,
Version 11.0 for Windows, was used for all analyses. Variable M SD M SD

Age 31.7 8.9 29.4 6.5

Results BPDSI (0–90) 33.1 7.1 35.4 9.1
n Axis I diagnoses 2.9 1.5 2.4 1.7
n Axis II diagnoses included BPD 2.1 1.2 2.1 1.3
Characteristics of Participants YSQ global score (16–96) 53.5 11.9 58.8 12.2
IPO global score (5–25) 12.8 3.1 13.8 2.6
Of the 88 randomized patients, 86 patients were included in the
analysis of the randomized controlled trial comparing SFT and Variable n % n %
TFP (Giesen-Bloo et al., 2006). Eight patients in TFP compared
with none in SFT terminated treatment prematurely between pre- Gender
Women 40 90.9 32 94.1
treatment and the first repeated assessment after 3 months (Fish- Men 4 9.1 2 5.9
er’s exact test, p ⬍ .01). In 2 patients, it did not prove to be Education
possible to agree on a treatment contract, which in TFP is consid- Graduate/professional 6 13.6 2 9.5
ered to be a prerequisite for starting treatment, and in 1 patient TFP College graduate 3 6.8 6 16.7
was contraindicated according to the subjective evaluation of the Some college 17 38.6 14 33.4
High school graduate 5 11.4 6 23.7
therapist. Five other patients dropped out in the first 3 months of Grades 7–11 13 29.6 6 16.7
treatment after completing treatment contracting. Although no
patients’ WAI-P scores of the 5 dropouts from TFP in this period Note. BPDSI ⫽ Borderline Personality Disorder (BPD) Severity Index;
are available, therapists’ WAI-T and DDPRQ have been collected. YSQ ⫽ Young Schema Questionnaire; IPO ⫽ Inventory of Personality.
Compared with patients still in treatment at early treatment (n ⫽
57), the quality of the therapeutic alliance on the WAI-T was rated
as significantly lower (M ⫽ 8.4, SD ⫽ 1.1 vs. M ⫽ 10.7, SD ⫽ respect to the patients they were treating. Chi-square analyses
1.3), t(60) ⫽ 3.67, p ⬍ .01, and these patients were also rated as revealed no significant differences in gender, professional back-
significantly more irritating and frustrating in the therapeutic re- ground, therapy condition, or number of patients in treatment (all
lationship on the DDPRQ (M ⫽ 3.5, SD ⫽ 0.9 vs. M ⫽ 2.8, SD ⫽ ps ⬎ .1) between these therapists and the remaining 14 therapists.
0.7), t(60) ⫽ 2.14, p ⬍ .05. These 30 therapists rated the therapeutic alliance of the 57
Table 1 gives an overview of the biographical and clinical patients who they treated with SFT or TFP (73.1% of the intention-
characteristics of the 78 patients in early treatment that constitute to-treat sample of 78 patients). Chi-square analyses and t tests for
the intention-to-treat sample of the present study. Chi-square anal- independent samples revealed no significant differences with re-
yses and t tests for independent samples revealed no significant spect to biographical and clinical characteristics, WAI-P scores,
differences with respect to biographical and clinical characteristics and BPDSI scores between these 57 patients versus the 29 patients
between the 34 patients in the TFP and 44 patients in the SFT who were treated by the other 14 therapists and for whom no
condition (all ps ⬎ .1). therapist ratings are available (all ps ⬎ .1).
Forty-four therapists (21 in TFP and 23 in SFT) participated in Of the 78 patients with early WAI-P scores and still in treatment
the randomized controlled trial. Three therapists held doctoral at mid- and late treatment, data on the middle and late alliance
degrees, 37 therapists held master’s degrees, and 4 therapists held were available for 66 (100.0%) and 52 (98.1%) patients, respec-
bachelor’s degrees with postgraduate training. All therapists had tively. Of the 57 patients with early alliance ratings by the therapist
prior therapy experience in the associated therapeutic orientation and still in treatment at mid- and late treatment, WAI-T and
and clinical experience in treating patients with a borderline per- DDPRQ scores of the middle and late alliance were available for
sonality disorder. The ratio of male to female therapists was 1:1. 42 (85.7%) and 33 (86.8%) of the patients at mid- and late
Thirty-two therapists were clinical psychologists, 7 were psychi- treatment.
atrists and 5 were social workers or psychiatric nurses with an
advanced psychotherapeutic training. Thirteen therapists treated 1
Early to Mid- to Late Treatment Changes
patient, 28 therapists treated 2 patients, and 3 therapists treated 3
patients of the present intention-to-treat sample (n ⫽ 78). Of the 44 Table 2 summarizes early, mid-, and late treatment values for
therapists participating in the multicenter-randomized trial, 30 the process and outcome variables. In analyzing BPDSI scores
(68.2%; 13 in TFP and 17 in SFT) also gave their informed consent (n ⫽ 78), a significant effect for time, F(1, 76) ⫽ 41.54, p ⬍ .001,
to participate in the present subsidiary study on the therapeutic was found. Planned comparisons revealed that BPDSI scores
alliance in the treatment of borderline personality disorder and changed significantly from early to midtreatment and also from
were willing to answer both the YSQ and IPO concerning their mid- to late treatment. No significant effect for group or significant
own personality characteristics and the WAI-T and DDPRQ with Group ⫻ Time interaction was observed.
110 SPINHOVEN ET AL.

Table 2
Early-, Mid-, and Late Treatment Values for Process and Outcome Variables (Theoretical Ranges Between Parentheses)

Schema-focused therapy (SFT) Transference-focused psychotherapy (TFP)

Early Mid Late Early Mid Late

Variable M SD M SD M SD M SD M SD M SD
a
BPDSI (0–90) 27.4 9.5 21.3 11.1 18.5 11.6 29.2 7.9 24.5 9.7 22.7 12.1
WAI-Pa (3–15) 11.0 1.5 11.5 1.6 11.5 1.7 10.2 1.9 10.4 1.8 10.7 2.0
WAI-Tb (3–15) 11.1 1.2 11.0 1.2 11.2 1.4 10.1 1.4 10.1 1.5 10.0 1.5
DDPRQb (1–6) 2.6 0.6 2.6 0.7 2.4 0.8 3.0 0.7 3.3 0.7 3.3 0.8

Note. Early ⫽ early treatment; Mid ⫽ midtreatment; Late ⫽ late treatment; BPDSI ⫽ Borderline Personality Disorder Severity Index; WAI-P ⫽ Working
Alliance Inventory–Patient Version; WAI-T ⫽ Working Alliance Inventory–Therapist Version; DDPRQ ⫽ Difficult Doctor–Patient Relationship
Questionnaire.
a
n ⫽ 78 (44 in SFT and 34 in TFP).
b
n ⫽ 57 (35 in SFT and 22 in TFP).

In analyzing WAI-P scores (n ⫽ 78), a significant effect for criterion after 3 years of treatment. Chi-square analyses revealed
group, F(1, 76) ⫽ 6.00, p ⬍ .05, and time, F(1, 76) ⫽ 7.15, p ⬍ no significant association of treatment condition with either the
.01, was found. Planned comparisons showed that WAI-P scores BPDSI recovery or reliable change criterion (all p values ⬎ .10).
changed significantly from early to late treatment. No significant Table 3 gives an overview of the WAI-P, WAI-T, and DDPRQ
effect for the Group ⫻ Time interaction was observed. In analyz- scores at early treatment of patients divided on the basis of reach-
ing WAI-T scores (n ⫽ 57), only a significant effect for group, ing the BPDSI recovery or reliable change criterion.
F(1, 55) ⫽ 9.23, p ⬍ .01, was observed. No significant main effect With separate logistic regression analyses using the BPDSI
for time or Group ⫻ Time interaction was observed. In analyzing recovery or reliable change criterion as dependent variables and
DDPRQ scores (n ⫽ 57), a significant effect for group, F(1, 55) ⫽ BPDSI pretreatment scores, treatment condition, and early treat-
15.25, p ⬍ .001, and significant Group ⫻ Time interaction, F(1, ment process scores as independent variables, neither early treat-
55) ⫽ 6.32, p ⬍ .05, was observed. No significant main effect for ment patients’ ratings (WAI-P scores) nor early treatment thera-
time was observed. pists’ ratings of the therapeutic alliance (WAI-T and DDPRQ
Overall, these results suggested that the quality of the therapeu- scores) were predictive of clinical improvement after 3 years of
tic alliance is rated higher in SFT than in TFP by therapists as well treatment. However, WAI-P scores were predictive of both the
as by patients. Moreover, the quality of the therapeutic alliance as BPDSI recovery criterion, Wald ⫽ 4.489, p ⬍ .05, odds ratio ⫽
rated by the patient increases in the course of treatment irrespec- 1.386, 95% confidence interval (CI) ⫽ 1.025–1.874; and the
tive of treatment condition, whereas therapist frustration decreased reliable change criterion, Wald ⫽ 3.886, p ⬍ .05, odds ratio ⫽
in SFT but increased in TFP.1 1.359, 95% CI ⫽ 1.002–1.843, while controlling only for BPDSI
To determine whether the differences in WAI-P, WAI-T, and pretreatment scores. Apparently, patients’ early treatment alliance
DDPRQ scores between treatment conditions may be due to clin- ratings are no longer predictive of clinical improvement after
ical improvement as achieved in the first 3 months of treatment, inclusion of treatment condition into the prediction model because
first baseline BPDSI scores at the start of treatment were regressed of the significant association of WAI-P scores with treatment
to 3-months early treatment scores to form residualized change condition.
scores. Next, early treatment WAI and DDPRQ scores were ana- Furthermore, it was investigated whether premature treatment
lyzed with one-way analyses of variance with residualized BPDSI termination was related to the quality of the therapeutic alliance at
change scores as a covariate. Also, after statistically controlling for early treatment. From early treatment (after 3 months) to midtreat-
any changes in BPDSI scores between baseline and early treat- ment (after 15 months), 12 patients dropped out of treatment (7 in
ment, the differences between conditions in WAI-P scores, F(1, TFP and 5 in SFT). In the period between midtreatment and late
75) ⫽ 3.94, p ⬍ .05; WAI-T scores, F(1, 54) ⫽ 8.54, p ⬍ .01; and treatment (after 33 months), a further 13 patients dropped out (6 in
DDPRQ scores, F(1, 54) ⫽ 5.57, p ⬍ .05, between SFT and TFP
remained statistically significant.
1
Because in the present study most of the therapists had more than 1
patient, any differences between scores of patients from different therapist
Prediction of 3-Year Outcome by Early Treatment pools may be due in part to differences between therapists (Crits-Christoph
Process Variables & Mintz, 1991). For this reason, we conducted separate analyses at the
group level and the individual level (Kenny & La Voie, 1985). Adjusted
At the 13th assessment 3 years after the start of treatment, 20 of
scores were calculated for each treatment condition separately. Because in
the 44 patients in SFT (45.5%) and 10 of the 34 patients in TFP the present study only 31 of the 44 therapists had more than 1 patient in
(29.4%) reached the BPDSI recovery criterion (a score on the treatment, data of only 65 patients could be partitioned into these two
BPDSI of less than 15). With respect to the reliable change index components. Subsequent analyses yielded the same results except that in
(a change of at least 11.70 on the BPDSI), 29 of the 44 patients in the analyses of WAI-T scores at the individual level also a significant
SFT (65.9%) and 17 of the 34 patients in TFP (50.0%) reached this effect for time ( p ⬍ .01) emerged.
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 111

Table 3 also in many cases nonsignificantly related, with only one signif-
Ratings of the Therapeutic Alliance at Early Treatment of icant association of early to midtreatment changes in DDPRQ
Recovered Versus Not-Recovered and Reliably Changed Versus scores with early to midtreatment changes in BPDSI scores,
Not-Reliably Changed Patients (Theoretical Ranges Between r(57) ⫽ .31, p ⬍ .05. These results suggest that variations because
Parentheses) of autocorrelation or synchronous correlations will not substan-
tially affect cross-lagged associations.
Not-recovered Recovered Cross-lagged correlations demonstrated that early to mid-
Variable M SD M SD
WAI-P was significantly related to mid- to late BPDSI, r(78) ⫽
–.34, p ⬍ .01, whereas early to mid-BPDSI was not significantly
WAI-Pa (3–15) 10.4 1.8 11.2 1.4 related to mid- to late WAI-P, r(78) ⫽ .11, ns. On the other hand,
WAI-Tb (3–15) 10.7 1.4 10.7 1.3 early to mid-WAI-T and DDPRQ were not related to mid- to late
DDPRQb (1–6) 2.8 1.7 2.7 0.6
BPDSI, whereas the converse correlations were also nonsignifi-
Non-Reliably cant.
Changed Reliably Changed
Testing Cross-Lagged Associations With Hierarchical
M SD M SD
Multiple Regressions
WAI-Pc (3–15) 10.2 1.8 11.0 1.6
WAI–Td (3–15) 10.5 1.3 10.9 1.4 Hierarchical regressions were performed to analyze whether
DDPRQd (1–6) 2.8 0.7 2.7 0.6 early to mid-WAI-P change scores remained a significant predictor
of mid- to late treatment BPDSI change scores when variance
Note. WAI-P ⫽ Working Alliance Inventory–Patient Version; WAI-T ⫽ because of early to midtreatment changes on the BPDSI and mid-
Working Alliance Inventory–Therapist Version; DDPRQ ⫽ Difficult
Doctor–Patient Relationship Questionnaire.
to late treatment changes on the WAI-P were controlled (Finkel,
a
n ⫽ 78 (48 not-recovered vs. 30 recovered). 1995). For mid- to late BPDSI, pre- to mid-WAI-P did emerge as
b
n ⫽ 57 (35 not-recovered vs. 22 recovered). a significant predictor also after controlling for the other change
c
n ⫽ 78 (32 not-reliably changed vs. 46 reliably changed). scores and treatment condition, Fchange(1, 73) ⫽ 11.701, p ⬍ .001.
d
n ⫽ 57 (26 not-reliably changed vs. 31 reliably changed). In testing the converse lagged association, no significant associa-
tions were observed (see Table 4).
TFP and 7 in SFT). No patient committed suicide. There was no
significant association of treatment condition with dropout rate in Relationship of Patient–Therapist Dissimilarity to
this particular treatment period. Separate Cox regression analyses Changes on Process and Outcome Variables
with BPDSI pretreatment scores, treatment condition, and early
By using t tests for independent groups, no significant differ-
treatment process scores as independent variables revealed a sig-
ences in early maladaptive schema (YSQ) and personality organi-
nificant effect for early treatment WAI-P scores (n ⫽ 78), Wald ⫽
4.379, p ⬍ .05, Hazard ratio ⫽ 0.775, 95% CI ⫽ 0.610 – 0.984.
Table 4
Moreover, especially therapists’ early treatment WAI-T and
Summary of Hierarchical Regression Analyses: Cross-Lagged
DDPRQ scores (n ⫽ 57) proved to be predictive for time to
Regressions for Working Alliance Inventory–Patient Version
dropout: WAI-T, Wald ⫽ 8.171, p ⬍ .01, Hazard ratio ⫽ 0.551,
(n ⫽ 78)
95% CI ⫽ 0.367– 0.829; and DDPRQ, Wald ⫽ 11.134, p ⬍ .001,
Hazard ratio ⫽ 3.133, 95% CI ⫽ 1.602– 6.129. Overall, these Variable B SE B R2 ⌬ R2 of step
results suggest that time to dropout is dependent on the quality of
the therapeutic alliance as perceived by patient or therapist as early Mid- to late BPDSI
as 3 months after the start of therapy over and above the effect of
Step 1
pretreatment BPDSI scores and treatment condition. Condition ⫺.080 .217
Early to mid-BPDSI ⫺.120 .109
Zero-Order Correlations for Early to Midtreatment and Mid- to late WAI-P ⫺.201 .107
Step 2 .044 .044
Mid- to Late Treatment Change Scores Early to mid-WAI-P ⫺.376 .110 .176* .132*
For each process (WAI-P, WAI-T, and DDPRQ) and outcome Mid- to late WAI-P
variable (BPDSI), early treatment scores were regressed on
midtreatment scores, and midtreatment scores were regressed on Step 1
late treatment scores to form early to midtreatment and mid- to late Condition ⫺.080 .231
treatment residualized chance scores. As in previous studies of Early to mid-WAI-P ⫺.166 .125
Mid- to late BPDSI ⫺.228 .122 .062 .062
process and outcome changes in panel designs (Burns et al., 2003; Step 2
Evon & Burns, 2004), changes among the process and outcome Early to mid-BPDSI .060 .116 .066 .004
variables at different time periods (i.e., autocorrelations) were
nonsignificant, suggesting that early treatment changes with re- Note. Variables are residualized change scores. Early to mid ⫽ early to
mid-treatment; Mid- to late ⫽ mid- to late treatment; BPDSI ⫽ Borderline
spect to these factors were unrelated to their corresponding late Personality Disorder Severity Index; WAI-P ⫽ Working Alliance Inven-
treatment changes. Synchronous correlations showed that changes tory—Patient Version.
in process and outcome factors during the same time period were *
p ⬍ .01.
112 SPINHOVEN ET AL.

zation (IPO) global scores between therapists (n ⫽ 17) from the early to mid- to late treatment changes, cross-lagged associations
SFT condition and TFP condition (n ⫽ 13) were found. Also, D2 of changes on process and outcome measures, and the influence of
YSQ and IPO dissimilarity scores of therapist–patient dyads in patient–therapist similarity to changes on process and outcome
SFT (n ⫽ 35) did not differ from those in the TFP condition (n ⫽ variables essentially yielded the same significant results.
22; all p values ⬎ .1).
Descriptively, therapist–patient YSQ dissimilarity scores were Discussion
found to range from 15.6 to 309.0 (M ⫽ 55.9; SD ⫽ 47.3) and IPO
dissimilarity scores from 2.7 to 90.1 (M ⫽ 17.5; SD ⫽ 17.0). This study had four aims. Firstly, it was hypothesized that the
Because of positive skewness, both dissimilarity measures were quality of the therapeutic alliance would be rated higher in SFT
log transformed for further statistical analyses, resulting in a quasi- than in TFP. Consistent with previous studies in Axis I disorders
normal distribution of both dissimilarity measures with adequate (e.g., Raue et al., 1997) this hypothesis was supported both with
skewness (⬍ 1). respect to patients’ and therapists’ ratings of the therapeutic alli-
First, we investigated whether dissimilarity between therapist ance. Furthermore, it was observed that the quality of the thera-
and patient in YSQ or IPO predicted the development of the peutic alliance as rated by the patient increased in the course of
therapeutic alliance between early to midtreatment by using simple treatment irrespective of treatment condition, whereas therapist
regression coefficients of D2 YSQ or IPO dissimilarity scores on frustration decreased in SFT but increased in TFP. These results
early to mid-WAI and DDPRQ scores. Both YSQ dissimilarity indicate that the rating of the alliance reflecting the overall quality
scores, r(57) ⫽ .26, p ⬍ .05, and IPO dissimilarity scores, r(57) ⫽ of experiences and feelings during a large number of therapy
.30, p ⬍ .05, were positively and significantly associated with sessions clearly differs between treatment conditions. The higher
early to midtreatment changes on the WAI-P. The associations of ratings in SFT possibly reflect the effort in SFT to connect to the
YSQ and IPO dissimilarity scores with early to mid- changes on patient by adapting an unthreatening and supportive attitude and to
the WAI-T and DDPRQ were all nonsignificant (all p values ⬎ .1). develop mutual trust and positive regard (Beck et al., 1990, 2004;
Next, it was investigated whether D2 YSQ or IPO dissimilarity Young et al., 2003). Using a schema mode model might help to
scores predicted early to mid- changes on the WAI and DDPRQ in increase sympathy with the BPD patient, as most dysfunctional
addition to and independent of treatment condition and the abso- behaviors are understood as stemming from unfortunate early life
lute levels of endorsement of early maladaptive schemas by pa- experiences (Young et al., 2003). In contrast, TFP with a contract
tients or the patients’ personality organization. By using multiple phase that by its working out might introduce an unnecessarily
regression analyses with patients’ total YSQ scores and condition defensive and adverse tone to the therapy and in which (negative)
forced into the equation in the first step, we found that YSQ transference manifestations are interpreted without the use of ex-
dissimilarity scores had a positive and significant semipartial cor- plicit supportive interventions possibly pressurizes the therapeutic
relation, r(57) ⫽ .30, p ⬍ .05, with early to mid-WAI-P scores, alliance (Gunderson, 2000) and even results in growing therapist
Fchange(1, 53) ⫽ 5.33, p ⬍ .05. IPO dissimilarity scores also had frustration in the course of therapy.
a positive and significant correlation, r(57) ⫽ .32, p ⬍ .05, with Secondly, it was expected that a lower quality of the therapeutic
early to mid-WAI-P, Fchange(1, 53) ⫽ 5.934, p ⬍ .05. The asso- alliance at early treatment would predict premature treatment ter-
ciations of YSQ and IPO dissimilarity scores with early to mid- mination and outcome. This expectation was partly corroborated as
changes on the WAI-T and DDPRQ were all nonsignificant (all p available therapists’ ratings of the alliance were associated with
value ⬎ .1). early dropout in TFP during the first 3 months of treatment and
Next, evidence for a direct effect of therapist–patient dissimi- both patients’ and therapists’ ratings of the therapeutic alliance
larity on early to midtreatment and mid- to late treatment changes after 3 months of treatment were found to predict time to dropout
on outcome measures was investigated. None of the simple regres- in the remaining treatment period in TFP and SFT. These results
sion coefficients of D2 YSQ or IPO dissimilarity on early to support the contention that compared with TFP the therapeutic
midtreatment and mid- to late BPDSI scores proved to be signif- alliance in SFT may not only be different but also more therapeutic
icant. t tests for independent groups also failed to show any at least in the first phase of therapy. The finding that, compared
significant differences between dropouts (n ⫽ 19) and treatment with SFT, significantly more patients terminated TFP prematurely
completers (n ⫽ 38) with respect to YSQ and IPO dissimilarity (Giesen-Bloo et al., 2006) may partly be accounted for by the
scores ( p ⬎ .1). quality of the therapeutic alliance. As has also been stressed by
These results indicate that a higher degree of dissimilarity in Linehan and colleagues, part of the dropout rate arises from the
maladaptive schemas or personality organization between thera- effect on the therapist of working with difficult patients. Both
pists and patients is associated with the development of a better therapist and patient distress possibly can be reduced by shifting
therapeutic alliance from the patient’s and not the therapist’s point therapist interpretations of patient behavior from hostile to friendly
of view irrespective of treatment condition or the patients’ absolute (Shearin & Linehan, 1992). Possibly, the first stages of TFP in
endorsement of maladaptive schemas or global level of personality which fragmented and partial aggressive self and object represen-
organization. tations are activated and interpreted by the therapist pose too high
demands on the beginning therapeutic alliance for a substantial
Completers Analyses proportion of patients with a borderline personality disorder.
Thirdly, it was hypothesized that growth of the therapeutic
It was investigated whether the results of the analyses in the alliance during the first year of therapy represents an important
completers sample (n ⫽ 53) were comparable with those in the therapeutic mechanism by which a later reduction of borderline
intention-to-treat sample (n ⫽ 78). The results of the analyses of personality disorder pathology is facilitated. The study results
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 113

corroborated this hypothesis with respect to an enhanced quality of that the quality of the alliance is affected by the nature of treat-
the therapeutic alliance as experienced by the patient. This result ment. Apparently, factors specific to a particular approach influ-
enlarges our understanding of the causal role of the therapeutic ence the quality of the alliance, and SFT with its emphasis on the
alliance in the treatment of different psychiatric disorders. Some “necessary and sufficient conditions” as identified by the client-
previous studies investigating relationships among alliance and centered school produces a better alliance according to the ratings
outcome changes in different periods of therapy seem to suggest of both therapists and patients. In addition, the magnitude of the
that growth of the alliance may be a mere epiphenomenon of consistent and positive association of alliance with outcome is
treatment gain (e.g., Evon & Burns, 2004; Klein et al., 2003; Tang relatively modest as in most previous studies (Martin et al., 2000;
& DeRubeis, 1999). However, it is conceivable that in structured, Orlinsky et al., 2004). It is extremely unlikely that any single
manualized, and short-term treatments of many Axis I disorders a process or mechanism will adequately explain most of the variance
collaborative therapeutic relationship is readily accomplished with in outcome. So, the causal role of the alliance does not rule out the
relatively few complications masking the fact that such a relation- possible role of other common factors (such as mitigation of
ship is a necessary precondition to treatment. Of note is that in isolation) or variables unique for a particular treatment approach
most treatment studies in Axis I disorders, ratings of the quality of (such as developing more healthy schemas or enhancing reality
the alliance with the WAI are generally high, possibly restricting orientation).
associations with other variables because of restriction of range. There are at least two reasons to think that the current data
Although the somewhat different answering format of the WAI in deserve serious consideration. First, we used a DSM-IV criteria-
the present study precludes a direct comparison with previous based semistructured interview for BPD with good psychometric
studies, the ratings of the alliance in the present study in patients properties as a primary outcome measure instead of relying on a
with borderline personality disorder, although positive, seem self-report measure (Arntz et al., 2003). Second, data on the
somewhat lower, possibly allowing the discovery of relationships quality and development of the therapeutic alliance were collected
with other variables, such as treatment condition, premature ter- in the course of a randomized clinical trial of intensive long-term
mination, and subsequent improvement. In the more semistruc- SFT and TFP in a rather large and clinically representative group
tured and long-term treatment of Axis II disorders, the develop- of BPD patients (Giesen-Bloo et al., 2006).
ment and maintenance of the therapeutic alliance constitutes a At least three limitations of this study merit consideration. First,
central issue of therapy and may constitute a central curing mech- the 12- and 18-months treatment lags probably are too extended to
anism (Orlinsky et al., 2004). Consequently, attaining this goal accurately detect the timing of process and outcome changes.
will affect therapy outcome. It seems worthwhile to continue Collapsing across a 12-month treatment period and then trying to
research into the intertwined and sequential relationships between relate process and outcome changes to changes collapsed across
alliance and client’s improvement during treatment in patient the following 18 months of treatment may have diluted or ob-
samples in which the establishment of the therapeutic alliance is scured the actual effects of the therapeutic alliance on outcome or
not always readily accomplished (such as Axis II disorders but also vice versa. In future research, it may be advisable to adapt shorter
Axis I disorders like addictive behavior or medically unexplained time lags to analyze the precise sequence of process and outcome
somatic symptoms). changes (Orlinsky et al., 2004; Tang & DeRubeis, 1999). Second,
The last aim of this study was to test whether dissimilarity in also the results of cross-lagged panel design analyses remain
pathological personality characteristics between therapists and pa- correlational and preclude definitive conclusions about the direc-
tients would facilitate the development of the therapeutic alliance tion of causality (Finkel, 1995). Third, because 8 of the 86 patients
and indirectly affect therapy outcome. Although dissimilarity in included in the randomized controlled trial prematurely terminated
pathological personality characteristics directly influenced the TFP treatment between pretreatment and the first assessment of the
growth of the therapeutic alliance as rated by the patient, it showed therapeutic relationship after 3 months, the significantly greater
no relationship with outcome. efficacy of SFT compared with TFP (Giesen-Bloo et al., 2006)
These results show that it is fruitful to study the impact of could not be demonstrated in the present study in 78 patients
therapists’ variables in a relational or interpersonal context (Beu- because of reduced statistical power and noninclusion of these 8
tler et al., 2004) and also that differentiating between (in)direct nonresponders all from the TFP condition. Consequently, no com-
effects on process or outcome can yield a more balanced view of plete mediational analysis (Baron & Kenny, 1986) could be per-
the causal network in which the alliance– outcome link is embed- formed to investigate whether the superior efficacy of SFT is
ded (cf., Hilliard, Henry, & Strup, 2000). The present findings are mediated by differences in the quality of the therapeutic relation-
in accordance with the presupposition of clinicians from various ship between treatment conditions.
theoretical orientations that therapists must be able to preserve a In sum, results provide support for the assertion that type of
neutral part of their mind that is able to accurately monitor and treatment differentially affects the quality and development of the
analyze their reactions provoked by schema activation or projec- therapeutic alliance and that dropout rate and clinical outcome can
tive identification. be partly accounted for by the quality of the therapeutic alliance.
Taken together, our study results seem to suggest that the Especially ratings of the therapeutic alliance by therapists at early
therapeutic alliance constitutes an important common factor in the treatment are predictive of dropout, whereas growth of the thera-
psychotherapeutic change process partly influenced by therapist– peutic alliance as experienced by patients in the first part of
patient dissimilarity in personality profile. However, this does not therapy seems to predict subsequent symptom reduction. These
imply that the therapeutic alliance is a “necessary and sufficient” results suggest that the therapeutic relationship and specific tech-
component of change in the treatment of borderline personality niques interact with and influence one another and may serve to
disorder. Clear alliance differences between treatments indicate facilitate change processes underlying clinical improvement
114 SPINHOVEN ET AL.

(Goldfried & Davila, 2005). Further research is needed to help us Giesen-Bloo, J., van Dyck, R., Spinhoven, P., van Tilburg, W., Dirksen, C.,
better understand the parameters associated with the role of tech- van Asselt, T., et al. (2006). Outpatient psychotherapy for borderline
nique and relationship in fostering general principles of change in personality disorder: Randomized trial of schema-focused therapy vs
the treatment of patients with borderline personality disorder. transference-focused psychotherapy. Archives of General Psychiatry,
63, 649 – 658.
Goldfried, M. R., & Davila, J. (2005). The role of relationship and
References technique in therapeutic change. Psychotherapy: Theory, Research,
Practice, Training, 42, 421– 430.
Arntz, A., van den Hoorn, M., Cornelis, J., Verheul, R., van den Bosch, Gunderson, J. G. (2000). Psychodynamic psychotherapy for borderline
W. M. C., & van de Bie, A. J. H. T. (2003). Reliability and validity of personality disorder. In J. G. Gunderson & G. O. Gabbard (Eds.),
the Borderline Personality Disorder Severity Index. Journal of Person- Psychotherapy for personality disorders (pp. 33– 64). Washington, DC:
ality Disorders, 17, 45–59. American Psychiatric Press.
Barber, J. P., Connolly, M. B., Crits-Christoph, P., Gladis, L., & Siqueland, Hahn, S. R., Kroenke, K., Spitzer, R. L., Brody, D., Williams, J. B. W.,
L. (2000). Alliance predicts clients’ outcome beyond in-treatment Linzer, M., & de Gruy, F. V. (1996). The difficult patient: Prevalence,
change in symptoms. Journal of Consulting and Clinical Psychology, psychopathology, and functional impairment. Journal of General Inter-
68, 1027–1032. nal Medicine, 11, 1– 8.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable Hahn, S. R., Thompson, K. S., Stern, V., Budner, N. S., & Wills, T. A.
distinction in social psychological research: Conceptual, strategic, and (1990). The Difficult Doctor–Patient-Relationship Questionnaire—
statistical considerations. Journal of Personality and Social Psychology, Assessing the problem patient. Clinical Research, 38, A739.
51, 1173–1182.
Hahn, S. R., Thompson, K. S., Wills, T. A., Stern, V., & Budner, N. S.
Beck, A. T., Freeman, A., & Associates. (1990). Cognitive therapy of
(1994). The difficult doctor–patient relationship: Somatization, person-
personality disorders. New York: Guilford Press.
ality and psychopathology. Journal of Clinical Epidemiology, 47, 647–
Beck, A. T., Freeman, A., Davis, D. D., & Associates. (2004). Cognitive
657.
therapy of personality disorders (2nd ed.). New York: Guilford Press.
Herman, S. M. (1998). The relationship between therapist– client modality
Berry, W. G., & Sipps, G. J. (1991). Interactive effects of counselor-client
similarity and psychotherapy outcome. Journal of Psychotherapy Prac-
similarity and client self-esteem on termination type and number of
tice and Research, 7, 56 – 64.
sessions. Journal of Counseling Psychology, 38, 120 –125.
Hilliard, R. B., Henry, W. P., & Strup, H. H. (2000). An interpersonal
Beutler, L. E., Malik, M., Alimohamed, S., Harwood, T. M., Talebi, H.,
model of psychotherapy: Linking patient and therapist developmental
Noble, S., & Wong, E. (2004). Therapist variables. In M. J. Lambert
history, therapeutic process, and types of outcome. Journal of Consult-
(Ed.), Bergin and Garfield’s handbook of psychotherapy and behavioral
ing and Clinical Psychology, 68, 125–133.
change (pp. 227–306). New York: Wiley.
Horvath, A. O., & Greenberg, L. S. (1989). Development and validation of
Bordin, E. (1979). The generalizability of the psychoanalytic concept of the
the Working Alliance Inventory. Journal of Counseling Psychology, 36,
working alliance. Psychotherapy: Theory, Research and Practice, 16,
223–233.
252–260.
Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical
Brunink, S. A., & Schroeder, H. E. (1979). Verbal therapeutic behavior of
expert psychoanalytically oriented, gestalt, and behavior therapists. approach to defining meaningful change in psychotherapy research.
Journal of Consulting and Clinical Psychology, 47, 567–574. Journal of Consulting and Clinical Psychology, 59, 12–19.
Burns, J. W., Kubilus, A., Bruehl, S., & Harden, R. N. (2003). Do changes Kenny, D. A., & La Voie, L. (1985). Separating individual and group
in cognitive factors influence outcome following multidisciplinary treat- effects. Journal of Personality and Social Psychology, 38, 249 –256.
ment for chronic pain: A cross-lagged panel analysis. Journal of Con- Klein, D. N., Schwartz, J. E., Santiago, N. J., Vivuan, D., Vocisano, C.,
sulting and Clinical Psychology, 71, 81–91. Castonguay, L. G., et al. (2003). Therapeutic alliance in depression
Clarkin, J. F., Yeomans, F. E., & Kernberg, O. F. (1999). Psychotherapy treatment: Controlling for prior change and patient characteristics. Jour-
for borderline personality. New York: Wiley. nal of Consulting and Clinical Psychology, 71, 997–1006.
Crits-Christoph, P., & Mintz, J. (1991). Implications of therapist effects for Lambert, M. J. (1989). The individual therapist’s contribution to psycho-
the design and analysis of comparative studies of psychotherapies. therapy process and outcome. Clinical Psychology Review, 9, 469 – 485.
Journal of Consulting and Clinical Psychology, 55, 584 –594. Lenzenweger, M. F., Clarkin, J. F., Kernberg, O. F., & Foelsch, P. A.
Cronbach, L. J., & Gleser, G. C. (1953). Assessing similarity between (2001). The Inventory of Personality Organization: Psychometric prop-
profiles. Psychological Bulletin, 50, 456 – 473. erties, factorial composition, and criterion relations with affect, aggres-
Evon, D. M., & Burns, J. W. (2004). Process and outcome in cardiac sive dyscontrol, psychosis proneness, and self-domains in a non-clinical
rehabilitation: An examination of cross-lagged effects. Journal of Con- sample. Psychological Assessment, 13, 577–591.
sulting and Clinical Psychology, 72, 605– 616. Marmar, C. R., Gaston, L., Gallagher, D., & Thompson, L. W. (1989).
Finkel, S. E. (1995). Causal analysis with panel data. Thousands Oaks, Alliance and outcome in late-life depression. Journal of Nervous and
CA: Sage. Mental Disease, 177, 464 – 472.
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1994). Martin, D. J., Garske, J. P., & Davis, M. K. (2000). Relation of the
Structured Clinical Interview for DSM-IV Personality Disorders (SCID- therapeutic alliance with outcome and other variables: A meta-analytical
II). New York: Biometrics Research Department, New York State Psy- review. Journal of Consulting and Clinical Psychology, 68, 438 – 450.
chiatric Institute. Orlinsky, D. E., Ronnestad, M. H., & Willutzki, U. (2004). Fifty years of
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1997). psychotherapy process-outcome research: Continuity and change. In
Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I/P). M. J. Lambert (Ed.), Bergin and Garfield’s handbook of psychotherapy
New York: Biometrics Research Department, New York State Psychi- and behavioral change (5th ed., pp. 307–390). New York: Wiley.
atric Institute. Raue, P. J., Goldfried, M. R., & Barkham, M. (1997). The therapeutic
Gaston, L., Marmar, C. R., Gallagher, D., & Thompson, L. W. (1991). alliance in psychodynamic-interpersonal and cognitive-behavioral ther-
Alliance prediction of outcome beyond in-treatment symptomatic apy. Journal of Consulting and Clinical Psychology, 65, 582–587.
change as psychotherapy progresses. Psychotherapy Research, 1, 104 – Rijkeboer, M. M., van den Bergh, H., & van den Bout, J. (2005). Stability
112. and discriminative power of the Young Schema Questionnaire in a
 ALLIANCE AND BORDERLINE PERSONALITY DISORDER 115

Dutch clinical versus non-clinical population. Journal of Behavior Ther- beliefs among bulimic and comparison women. Cognitive Therapy and
apy and Experimental Psychiatry, 36, 129 –144. Research, 25, 137–147.
Salvio, M. A., Beutler, L. E., Wood, J. M., & Engle, D. (1992). The Waller, G., Shah, R., Ohanian, V., & Elliott, P. (2001). Core beliefs in
strength of the therapeutic alliance in three treatments of depression. bulimia nervosa and depression: The discriminant validity of Young’s
Psychotherapy Research, 2, 31–36. Schema Questionnaire. Behavior Therapy, 32, 139 –153.
Schmidt, N. B., Joiner, T. E., Young, J. E., & Telch, M. J. (1995). The Yeomans, F. E., Clarkin, J. F., & Kernberg, O. F. (2002). A primer for
Schema Questionnaire: Investigation of psychometric properties and the transference focused psychotherapy for the borderline patient. North-
hierarchical structure of a measure of maladaptive schemas. Cognitive vale, NJ: Jason Aronson.
Therapy and Research, 19, 295–321. Yeomans, F. E., Selzer, M., & Clarkin, J. (1989). Treating the borderline
Schouten, H. J. A. (1995). Adaptive biased urn randomization in small patient: A contract-based approach. New York: Basic Books.
strata when blinding is impossible. Biometrics, 51, 1529 –1535. Yeomans, F. E., Selzer, M., & Clarkin, J. (1993). Studying the treatment
Shearin, E. N., & Linehan, M. M. (1992). Patient-therapist ratings and contract in intensive psychotherapy with borderline patients. Psychiatry:
relationship to progress in dialectical behavior therapy for borderline Interpersonal and Biological Processes, 56, 254 –263.
personality disorder. Behavior Therapy, 23, 730 –741. Young, J. E. (1994). Cognitive therapy for personality disorders: A
Sloane, R. B., Staples, F. R., Cristol, A. H., Yorkston, N. J., & Whipple, K. schema-focused approach. (Revised ed.). Sarasota, FL: Professional
(1975). Psychotherapy versus behavior therapy. Cambridge, MA: Har- Resource Press.
vard University Press. Young, J. E., Arntz, A., & Giesen-Bloo, J. (2006). Therapy Adherence and
Stopa, L., Thorne, P., Waters, A., & Preston, J. (2001). Are the short and Competence Scale. Retrieved May 1, 2006, from http://www.unimaas.
long forms of the Young Schema Questionnaire comparable and how epp.nl
well does each version predict psychopathology scores? Journal of Young, J. E., Klosko, J., & Weishaar, M. E. (2003). Schema therapy: A
Cognitive Psychotherapy, 15, 253–272. practitioner’s guide. New York: Guilford Press.
Tang, T. Z., & DeRubeis, R. J. (1999). Sudden gains and critical sessions
in cognitive-behavioral therapy for depression. Journal of Consulting
and Clinical Psychology, 67, 894 –904. Received January 9, 2006
Waller, G., Meyer, C., & Ohanian, V. (2001). Psychometric properties of Revision received September 7, 2006
the long and short versions of the Young Schema Questionnaire: Core Accepted September 15, 2006 䡲

View publication stats