Behaviour Research and Therapy 74 (2015) 60e71

Contents lists available at ScienceDirect

Behaviour Research and Therapy

journal homepage: www.elsevier.com/locate/brat

Prediction of treatment discontinuation and recovery from Borderline

Personality Disorder: Results from an RCT comparing Schema Therapy
and Transference Focused Psychotherapy
Arnoud Arntz a, b, *, Sne

zana Stupar-Rutenfrans b, c, Josephine Bloo b, d, Richard van Dyck e,
f
Philip Spinhoven
a
Department of Clinical Psychology, University of Amsterdam, The Netherlands
b
Department of Clinical Psychological Science, Maastricht University, The Netherlands
c
International Media and Entertainment Management Academy for Digital Entertainment NHTV University of Applied Sciences, Breda, The Netherlands
d
Community Mental Health Center Virenze RIAGG, Vaals, The Netherlands
e
VU Medical Centre Amsterdam, The Netherlands
f
Institute of Psychology, Leiden University, Leiden, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: Knowing what predicts discontinuation or success of psychotherapies for Borderline Personality Disorder
Received 6 May 2015 (BPD) is important to improve treatments. Many variables have been reported in the literature, but
Received in revised form replication is needed and investigating what therapy process underlies the findings is necessary to un-
12 September 2015
derstand why variables predict outcome. Using data of an RCT comparing Schema Therapy and Trans-
Accepted 14 September 2015
Available online 21 September 2015
ference Focused Psychotherapy as treatments for BPD, variables derived from the literature were tested as
predictors of discontinuation and treatment success. Participants were 86 adult outpatients (80 women,
mean age 30.5 years) with a primary diagnosis of BPD who had on average received 3 previous treatment
Keywords:
Borderline Personality Disorder
modalities. First, single predictors were tested with logistic regression, controlling for treatment type (and
Outcome prediction medication use in case of treatment success). Next, with multivariate backward logistic regression essential
Treatment discontinuation predictors were detected. Baseline hostility and childhood physical abuse predicted treatment discon-
Dissociation tinuation. Baseline subjective burden of dissociation predicted a smaller chance of recovery. A second study
Schema Therapy demonstrated that in-session dissociation, assessed from session audiotapes, mediated the observed ef-
Transference Focused Psychotherapy fects of baseline dissociation on recovery, indicating that dissociation during sessions interferes with
treatment effectiveness. The results suggest that specifically addressing high hostility, childhood abuse,
and in-session dissociation might reduce dropout and lack of effectiveness of treatment.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction predictors of discontinuation and treatment success. Although

causality usually remains an issue to be further investigated,
Although the last decades showed a reduction in pessimism knowledge of predictors of treatment discontinuation and success
about the possibilities to treat Borderline Personality Disorder is a first step in the process to develop more acceptable and
(BPD), BPD remains one of the more challenging psychiatric dis- effective treatments. Unfortunately, the field of psychotherapy
orders, often needing specialized long-term treatment. Effective prediction studies is characterized by a lack of attempts to replicate
psychological treatments have been developed, but premature findings of previous studies, so that it is unclear whether findings
discontinuation of treatment and lack of effectiveness of treat- were accidental or not.
ments in a substantial proportion of patients remain areas where In the area of (B)PD treatment, a recent review (McMurran,
improvements can be achieved. Many studies have tried to detect Huband, & Overton, 2010) reported the following patient-related
pretreatment characteristics in single or multiple studies being
prognostic of treatment discontinuation, which we grouped in the
* Corresponding author. Department of Clinical Psychology, University of following categories.
Amsterdam, PO Box 19268, 1000 GG Amsterdam, The Netherlands. Biographical variables: younger age; lower education; lower
E-mail address: [email protected] (A. Arntz).

http://dx.doi.org/10.1016/j.brat.2015.09.002
0005-7967/© 2015 Elsevier Ltd. All rights reserved.
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 61

occupational level; unemployment; juvenile conviction; parents Focused Psychotherapy (TFP) for BPD (Giesen-Bloo et al., 2006), we
divorced before patient was 10; spending less time alone; being in a tested previously reported predictors (if included in our baseline
relationship for less than 6 months. assessment) of treatment discontinuation and outcome (recovery).
Treatment history: less previous experience with mental health In short, results showed that treatment dropout from TFP was
care; prior hospitalization. significantly higher than from ST (50.0% vs. 27.3%), and that ST was
BPD features: BPD-severity; impulsivity; fewer suicide attempts; superior to TFP in terms of recovery from BPD (45.5% vs. 23.8%),
current/previous substance abuse. changes in dimensional indices of BPD severity (including 6 of the 9
Defense mechanisms and personality organization: poor ego DSM-IV BPD-criteria), changes in secondary variables, and cost-
structure; defense level. effectiveness (Giesen-Bloo et al., 2006; van Asselt et al., 2008).
Axis-II comorbidity: more PD diagnoses/traits; dependent, Moreover, medication use at baseline, as well as during treatment,
obsessiveecompulsive, histrionic, or antisocial PD; Cluster-A or B predicted a smaller chance of recovery. Given the results of the RCT,
PDs; no specific PD; narcissism. we controlled for (significant) effects of condition and medication
Negative emotional problems: less depression (level/diagnosis); (the last for predicting recovery). Predictors that were (trend) sig-
high trait anxiety. nificant when tested singly were subsequently tested in a multi-
Childhood abuse and neglect: childhood emotional neglect. variate approach, so that their unique contribution could be
Others: pre-contemplation stage of change; lower general level assessed. The second study aimed at testing whether the effect of a
of functioning; low problem solving capacities; better social predictor of failure to recover was mediated by a process during
competence; lower persistence; problems in one area vs. several treatment sessions, based on ratings of recordings of treatment
areas; and higher avoidance (one study experiential avoidance, sessions.
another harm avoidance). Contradicting the last, avoidant PD was
associated with treatment completion. Conflicting results between 2. Study 1 method
studies were found as to interpersonal distress.
In addition, we detected in the literature the following pre- 2.1. Design and participants
dictors of treatment discontinuation: male gender (Arntz, 1999;
Links, Mitton, & Steiner, 1990; Lo €ffler-Stastka et al., 2003; This study used data from an RCT on the effectiveness of ST and
Thorma €hlen et al., 2003); being single/divorced (Links et al., TFP. In short, 86 patients with a primary diagnosis of BPD were
1990); and anger-hostility (Smith, Koenigsberg, Yeomans, Clarkin, randomized to these two treatments, and followed up during three
& Selzer, 1995; Rüsch et al., 2008 (trend)). years of treatment. Axis-1 and -2 diagnoses were based on SCID
As to the prediction of treatment response, Gunderson et al. interviews (First, Spitzer, Gibbon, & Williams, 1996, 1997;
(2006) and Barnicott et al. (2012) offer overviews of empirical Groenestijn, Akkerhuis, Kupka, Schneider, & Nolen, 1999;
findings. Gunderson et al. found similar predictors as for treatment Weertman, Arntz, & Kerkhofs, 2000). An extensive structured
discontinuation. Additional predictors for worse outcome included: interview, the Borderline Personality Disorder Severity Index
low IQ; affective instability; distractibility; family mental illness; version IV (BPDSI-IV; Giesen-Bloo, Wachters, Schouten, & Arntz,
early psychiatric contact; various forms of childhood maltreatment 2010) was used to assess severity of BPD-manifestations every 3
and loss; disability insurance benefits. In Barnicott et al.'s review it months. The BPDSI-IV has been recommended as a good outcome
is concluded that higher severity predicts larger improvements, measure for BPD treatment research by an international group of
and that demographic variables generally have no influence experts (Zanarini et al., 2010), and has been validated in various
(though some studies report effects of age, gender or employment), studies (Arntz et al., 2003; Giesen-Bloo et al., 2010; Kro € ger et al.,
as does social adjustment. Negative effects were found for suicide 2013; Leppa €nen, Lindeman, Arntz, & Hakko, 2013; Lepp€ anen,
attempts (two studies), and mixed effects for axis-1 symptom Hakko, Sintonen, & Lindeman, 2015) and used in various treatment
severity, self-harm (four studies no effect, two negative and one studies as outcome instrument (e.g., Bales et al., 2012; Bellino,
positive effect), medication use, and dissociation. As to dissociation, Paradiso, & Bogetto, 2006, 2015; Dickhaut & Arntz, 2014;
one study found that baseline level of dissociative symptoms pre- Lepp€ anen et al., 2015; Nadort et al., 2009; Schuppert et al., 2012;
dicted worse outcome (Kleindienst et al., 2011); however another Verheul et al., 2003). A statistically derived criterion based on the
(smaller) study did not find evidence for dissociation predicting BPDSI total score (when < 15) defined recovery from BPD with
less treatment response (Braakmann et al., 2007). Barnicott et al. specificity .97 and sensitivity 1.00 (Giesen-Bloo et al., 2006; 2010).
comment that mixed findings might be related to whether change Treatment discontinuation was also monitored during the study
or remission is taken as outcome. Gunderson et al.'s (2006) own (Giesen-Bloo et al., 2006). Inclusion criteria were BPD as primary
empirical findings indicated that BPD-severity, level of functioning, diagnosis, age 18e60 years, BPDSI-IV score >20, and Dutch literacy.
childhood trauma, and quality of current relationships (incl. BPD Exclusion criteria were psychotic disorders (except psychotic epi-
traits 1 and 2) predict outcome. In addition, we found two addi- sodes covered by BPD-criterion 9), bipolar disorder, dissociative
tional studies with contradictory findings as to self-injury (Chiesa & identity disorder, antisocial personality disorder, attention-deficit/
Fonagy, 2007; Plakun, 1991). hyperactivity disorder, addiction needing clinical detoxification
As said, one of the major problems with this overwhelming list of (after detox participation was allowed; addictions not needing
discontinuation and treatment success predictors is the lack of clinical detox were not excluded), psychiatric disorders secondary
replication. Moreover, usually only univariate analyses were done, to medical conditions, and mental retardation. When mental
obscuring which predictors are essential. Systematic replication of retardation was suspected, a formal IQ test was planned to be
previous findings is needed followed by meta-analytic studies to taken; however none of the candidates was suspected of mental
disentangle which predictors are robust, and which might have been retardation.
accidental findings. Moreover, there is a need for studies that inves- At baseline, the 86 participants (6 men) were on average 30.6
tigate how predictors affect the treatment. Without understanding years (SD 7.8), had a BPDSI-IV total score of 33.97 (SD 7.97), 2.7 (SD
“the how” little theoretical and clinical progress can be made. 1.6) Axis-1 and 2.1 (SD 1.2) PD-diagnoses, and had 2.9 (SD 1.3)
The present paper consists of two related studies. Study 1 aimed previous treatment modalities; 65 (75.6%) used psychotropic
to test previously reported predictors. Using data of a multicenter medication (58 (67.4%) antidepressants; 28 (32.6%) anxiolytics; 17
RCT on the effectiveness of Schema Therapy (ST) and Transference (19.8%) antipsychotics). Fifty-four (62.8%) didn't have a partner; 69
62 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71

(80.2) didn't have a paid job. More details about the study can be Schouten, & Bernstein, 2009). The interview yields dimensional
found in Giesen-Bloo et al. (2006). severity scores of emotional (Cronbach a ¼ .83), physical (a ¼ .88)
and sexual abuse (a ¼ .89), as well as emotional (a ¼ .75) and
2.2. Predictors physical neglect (a ¼ .58).
Personality Disorder comorbidity was assessed with the SCID-II
All predictors were assessed before randomization by self- total of PD-diagnoses and sum scores per personality disorder
report, independent research assistants, or clinicians that did in- (range of interrater reliability (ICC) per PD .69e.95, mean ICC ¼ .84;
takes and SCIDs. Lobbestael, Leurgans, & Arntz, 2011; range of internal consistency:
Biographical variables, treatment history, medication use, .55e.87, mean ¼ .73; data from Arntz et al., 2009). For narcissism,
employment status, disability compensation, educational level were an additional variable was explored, the entitlement score of the
derived from structured interviews taken by research assistants at Young Schema Questionnaire (YSQ; Rijkeboer, van den Bergh, & van
baseline assessment. In the Netherlands, the educational system den Bout, 2005; Cronbach alpha ¼ .76; Rijkeboer & van den Bergh,
has different levels of diplomas, which we ordered in categories 2006).
ranging from primary school to university. This is also a proxy for Level of depression, anxiety and hostility were assessed with
IQ, as research indicates correlations between IQ and educational subscales of the SCL-90 (Derogatis, Lipman, & Covi, 1973). Medians
level in the .50 - .85 range (Calvin, Fernandes, Smith, Visscher, & of internal consistencies over various Dutch samples are .90
Deary, 2010; Deary, Strand, Smith, & Fernandes, 2007; Gerritsen, (depression), .88 (anxiety) and .77 (hostility) (Arrindell & Ettema,
Berg, & Deelman, 2001; Luteijn & Barelds, 2004; Plassman et al., 2003).
1995; Tambs, Sundet, Magnus, & Berg, 1989; Wechsler, 2005).
BPD-severity was assessed with the total BPDSI-IV-score, rep- 2.3. Statistical analysis
resenting objectifiable frequency and extent of BPD-manifestations
during the last 3 months, and with the BPD-checklist, a self-report Predictors of discontinuation were first separately tested with
of the subjective burden of BPD-symptoms during the last month. logistic regression, controlling for treatment condition, given that
The internal consistency of the BPDSI-IV total score is .96 in a condition predicted treatment dropout, with ST having significantly
heterogeneous sample, and .85 in a BPD-sample; of the BPD- less dropout than TFP (Giesen-Bloo et al., 2006). Next, predictors
checklist .93 (see Giesen-Bloo et al., 2010 for both instruments). with a significance level <.10 were together entered and tested with
Level of recent suicidality was assessed by the sum of BPDSI-IV backward logistic regression, with a two-tailed significance level of
items 5.11e5.13 (plans, steps, attempts; Cronbach alpha ¼ .63 p ¼ .05 as criterion. Lastly, it was tested whether the resulting
(Giesen-Bloo et al., 2010)), and with BPD-checklist items 9 and 26 predictors were moderated by condition, by adding the predictor
(Cronbach alpha ¼ .84). by condition interactions to the regression equation. These
Level of recent self-injurious behavior was assessed with BPDSI-IV moderation tests were a second reason to include treatment con-
items 5.1e5.8 (Cronbach alpha ¼ .73), and with BPD-checklist item dition as covariate, as moderation tests should be done with main
6. effects as covariates forced into the equation. The same approach
Level of recent substance abuse was assessed with the sum of was used for predicting recovery, but now with both condition and
items BPDSI-IV 4.4e4.7, and with the sum of BPD-checklist items 12 baseline medication use as covariates given the findings of the
(alcohol) and 17 (drugs). The internal consistency of the first scale original RCT, where condition and medication use were significant
was .35, of the second .37. The low reliabilities point at the relatively predictors of recovery (with ST superior to TFP in recovery, and
low association between drugs and alcohol abuse. those using medication achieving less often recovery; Giesen-Bloo
Level of recent interpersonal instability was assessed with sum- et al., 2006). Analyses were based on intent-to-treat principle, that
med BPDSI-IV criterion 1 and 2 total scores (Cronbach alpha ¼ .85), is all available data was included. For the recovery analysis, the last
and a similar sum for the BPD-checklist (Cronbach alpha ¼ .91). available assessment was therefore used (in the 3 year period).
Level of recent impulsivity was assessed with BPDSI-IV criterion 4 Analyses were checked for robustness by redoing them with sur-
total score (Cronbach alpha ¼ .67), and similar for the BPD-checklist vival analysis, which yielded similar findings.
(Cronbach alpha ¼ .70). We did not correct for number of tests that were done, for
Level of recent anger regulation problems was assessed with instance by a Bonferroni correction, for the following reasons. First,
BPDSI-IV criterion 8 total score (Cronbach alpha ¼ .78), and similar we wanted to test whether each predictor reported previously in
for the BPD-checklist (Cronbach alpha ¼ .78). the literature would survive replication. Using a corrected p-level
Level of recent dissociation problems was assessed with BPDSI-IV (e.g., p < .0009 instead of p < .05, when a Bonferroni correction is
criterion 9 dissociation items sum score (with the paranoia items used for 55 predictors) would lead to an increased chance of
excluded; Cronbach alpha ¼ .77), and similar for the BPD-checklist rejecting predictors as not significant whilst they actually are valid
(Cronbach alpha ¼ .78). predictors. Thus, to give the suggested predictors a reasonable
Level of defense mechanisms was assessed with the 3 subscales chance to emerge, we used a significance level of .05 for each in-
(Mature, Neurotic and Immature Defenses) of the Defensive Style dividual predictor. Second, only unrealistically powerful predictors
Questionnaire (Andrews, Pollock, & Stewart, 1989). In the Giesen- would survive a corrected p-level with the current sample size, and
Bloo et al. (2010) data internal consistencies were .52 (Mature it seemed unlikely that such powerful predictors exist, as only a few
Defenses), .71 (Neurotic Defenses), and .88 (Immature Defenses). could exist and such obvious predictors would have already be
Level of personality organization (ego pathology) was assessed detected in previous research. Thus, instead of aiming to demon-
with the Inventory of Personality Organization (Lenzenweger, strate that (a) specific variable(s) predict treatment discontinuation
Clarkin, Kernberg, & Foelsch, 2001). Five subscales were assessed: or success, we aimed to contribute to building an empirical
Lower Level Defenses, Identity Diffusion, Pathological Object Re- knowledge base that requires replication and accumulation of
lations, Alterations in Reality Testing, Superego Pathology. Internal findings over studies to detect predictors.
consistencies range from .78 to .93 (Berghuis, Kamphuis, Boedijn, & We used a backward procedure to assess which predictors, that
Verheul, 2009). were (trend) significant when separately tested, would survive
Childhood abuse and neglect was assessed with the Interview for when controlled for each other. Stepwise procedures have been
Traumatic Events in Childhood (ITEC; Lobbestael, Arntz, Harkema- criticized for leading to overfitting of the model, including inflated
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 63

beta's, R2 values etc. This implies that replication with the same set ST, which reduces to 32% at the highest level of dissociation; for TFP
of predictors is necessary to come to unbiased estimates of pa- this is 47% respectively 14%.
rameters, R2, etc. Note however that the aim of the use of the
stepwise procedure was not to create a numerical prediction model 4. Discussion
to predict chances of dropout or recovery for new patients, but to
assess whether predictors that were (trend) significant when tested Evidence was found that treatment discontinuation was related
initially alone would survive when controlled for each other, that is to a history of childhood physical abuse and baseline level of hos-
to assess to what degree their contribution was unique. Given the tility, but also that discontinuation was not related to numerous
relatively small sample size, a multivariate model with all pre- other variables that have been mentioned in the literature. Spe-
dictors entered simultaneously was impossible, hence we first cifically, variables mentioned in the McMurran et al. (2010) review
tested the predictors separately. The relatively small sample size as most prominent over studies, i.e. younger age, lower education,
also makes any multivariate model with relatively many predictors lower occupational levels, and greater avoidance were not found to
problematic, which calls for a selection procedure. This implies that predict discontinuation. The association of childhood physical
the explanatory power of the resulting model after backwards abuse with discontinuation is at odds with a previous study that
elimination might be overestimated, and that replication in an in- failed to detect such a relationship, but found emotional neglect to
dependent dataset is essential to get unbiased estimation. Thus, the be a predictor (Perry, Bond, & Roy, 2007). However, the latter
current procedures should be seen as aiming to further select study's sample was mixed as to diagnoses and the study was con-
candidate predictors that need further testing in the future. Note on ducted in a natural (uncontrolled) context. Moreover, forms of
the other hand that, in contrast to most of the previous research, we childhood abuse tend to correlate (Lobbestael et al., 2009), making
choose variables that were reported in previous research to be the detection of essential predictors difficult. Hostility as predictor
significant predictors, thus our study did not use a simple shotgun of discontinuation has previously been reported by Smith et al.
approach. (1995) for TFP and Rüsch et al. (2008) for inpatient DBT (though
at trend level). The range of psychotherapy types of which hostility
3. Results is predictive of discontinuation is interesting and makes it a po-
tential candidate for a relatively universal discontinuation
3.1. Prediction of treatment discontinuation predictor.
As to treatment success, a quite different predictor was found:
Table 1 gives an overview of the tests of the predictors sepa- subjective burden of dissociation. This replicates a study where
rately, controlled for condition, as well as of the final model after dissociation was found to predict worse outcome of DBT
entering the predictors with p < .10 and stepwise deleting those (Kleindienst et al., 2011). Note however that another (smaller)
with p > .05. The final model contained three predictors: condition study did not find evidence for dissociation predicting less treat-
(TFP vs ST), hostility and childhood physical abuse. Patients with ment response of inpatient DBT (Braakmann et al., 2007). Still
higher levels of hostility (SCL-90) and with more severe childhood another study reported that pre-treatment dissociation severity did
physical abuse (ITEC) had a higher chance to discontinue treatment not predict remission from self-harm (Harned, Jackson, Comtois, &
prematurely. It was also explored whether conditions differed in Linehan, 2010). The last two studies were criticized as having low
predictive strength of hostility and childhood physical abuse, but quality (Barnicott et al., 2012). The finding that baseline dissocia-
predictor by condition interactions failed to reach significance, tion predicts a poorer response raised the hypothesis that perhaps
hostility (SCL-90) by condition: OR ¼ 1.033, p ¼ .75, 95%CI (.845; patients reporting to highly suffer from dissociation, also do so
1.263); childhood physical abuse: OR ¼ .944, p ¼ .77, 95%CI (.644; during psychotherapy sessions, which might interfere with profit-
1.383). ing from the session and thus explain the lack of success. Study 2
aimed to investigate this hypothesis.
3.2. Prediction of recovery
5. Study 2
Table 2 presents an overview of the tests of the predictors
separately, controlled for condition and medication, as well as of The prediction of recovery by self-reported dissociation at
the final model after entering the predictors with p < .10 and baseline raised the question whether perhaps this effect was
stepwise deleting those with p > .05. The tests of separate pre- caused by those reporting high levels of dissociation at baseline
dictors showed, controlled for condition and medication, burden of also suffering from dissociation during treatment sessions. If so,
total BPD-symptoms, burden of dissociative BPD-symptoms, and profiting from treatment could be compromised by in-session
burden of suicidality symptoms, all assessed with the BPD- dissociation, dissociation limiting the capacity to process the ex-
checklist, to be significant (Table 2). The backward regression periences and insights from the session and/or prohibiting the
procedure showed burden of dissociative symptoms to remain a therapist to use the full range of treatment techniques. This hy-
significant predictor, whilst other predictors with initial p < .10 pothesis was tested by having independent judges rate the level of
being removed. In sum, the final model showed that TFP (vs. ST), dissociation manifest during an audio-recorded session and testing
baseline medication use, and higher levels of subjective burden of whether the in-session level of dissociation statistically mediated
dissociative symptoms to be predictive of less chance of recovery. the relationship between baseline dissociation and recovery.
The moderation tests showed that treatment condition did not
significantly influence the association of subjective burden of 6. Method
dissociation with recovery: OR ¼ .851, p ¼ .37, 95%CI (.598; 1.211),
controlled for baseline medication. Fig. 1 illustrates the relationship 6.1. Tape ratings
between level of subjective burden of dissociation (grouped in
quintiles (each quintile represents 20% of sample) and chance of To investigate the possible mediation of negative effects of
recovery by condition, controlled for medication use (based on pretest dissociation on treatment outcome by dissociation during
results of the logistic regression). The influence of dissociation treatment sessions, a 15-item scale was developed to rate mani-
appears to be large. For example, at lowest level 73% recovers with festations of dissociation on the basis of session recordings, as a
64 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71

Table 1
Prediction of treatment discontinuation.a

Predictor b s.e. Wald p-value OR 95% Confidence

interval

Lower Upper

Condition (TFP vs ST) 981 .458 4.585 .032 2.667 1.087 6.544
Biographical characteristics
Gender (male vs female) 1.091 1.138 .918 .338 .336 .036 3.128
Age (in years) .014 .030 .233 .630 1.014 .957 1.076
Educational level (5 levels, low to high) .061 .182 .115 .735 1.063 .745 1.518
Highest employment level (7 levels, low to high) .096 .121 .638 .424 1.101 .869 1.395
Having a partner .469 .475 .977 .323 1.599 .631 4.056
Having a paid job .498 .565 .778 .378 1.645 .544 4.976
Receiving disability compensation .434 .626 .481 .488 1.544 .452 5.272
Treatment history
Medication use at baseline .129 .534 .058 .810 1.137 .399 3.238
Having had psychological therapy .382 .646 .351 .554 .682 .192 2.418
Number of previous treatments .016 .178 .008 .929 .984 .694 1.395
History of inpatient treatment .105 .489 .046 .830 1.111 .426 2.894
BPD features
BPD severity (BPDSI-IV total) .005 .029 .035 .851 .995 .940 1.052
BPD severity (burden of BPD symptoms) .008 .009 .844 .358 1.008 .991 1.027
Level of suicidality (BPDSI-IV) .813 .486 2.795 .095 2.255 .869 5.848
Burden of suicidality (BPD-checklist) .048 .118 .169 .681 .953 .757 1.200
Level of self-injury (BPDSI-IV) .168 .192 .770 .380 .845 .580 1.231
Burden of self-injury (BPD-checklist) .215 .119 1.159 .282 .807 .546 1.192
Level of substance abuse (BPDSI-IV) .036 .190 .036 .850 .965 .665 1.400
Burden of substance abuse (BPD-checklist) .077 .133 .331 .565 1.080 .831 1.403
Level of interpersonal instability (BPDSI-IV) .193 .181 1.137 .286 .824 .577 1.176
Burden of interpersonal instability (BPD-checklist) .004 .027 .026 .871 1.004 .952 1.059
Level of impulsivity (BPDSI-IV) .029 .210 .020 .889 1.030 .682 1.556
Burden of impulsivity (BPD-checklist) .045 .049 .852 .356 1.046 .950 1.152
Level of anger problems (BPDSI-IV) .007 .130 .003 .957 .993 .770 1.280
Burden of anger problems (BPD-checklist) .140 .072 3.777 .052 1.151 .999 1.326
Level of dissociation (BPDSI-IV) .018 .021 .749 .387 1.018 .977 1.061
Burden of dissociation (BPD-checklist) .007 .068 .011 .918 1.007 .881 1.151
Quality of current relations (WHOQOL) .141 .099 2.033 .154 .868 .715 1.054
Defense mechanisms & personality organization
Mature defense mechanisms (DSQ) .008 .021 .135 .713 1.008 .967 1.050
Neurotic defense mechanisms (DSQ) .004 .013 .107 .744 .966 .970 1.022
Immature defense mechanisms (DSQ) .012 .010 1.680 .195 1.012 .994 1.032
Lower level defenses (IPO) .009 .022 .173 .678 .991 .950 1.034
Identity diffusion (IPO) .006 .019 .113 .737 1.006 .970 1.045
Pathological object relations (IPO) .000 .013 .001 .981 1.000 .976 1.026
Alterations in reality testing (IPO) .005 .255 .000 .986 .995 .604 1.640
Superego pathology (IPO) .170 .307 .307 .579 1.186 .649 2.165
Axis-II comorbidity
Number of personality disorders (SCID-II) .180 .200 .813 .367 .835 .564 1.236
Avoidant traits (SCID-II) .014 .058 .063 .803 .986 .880 1.104
Dependent traits (SCID-II) .031 .050 .386 .534 .969 .879 1.069
Obsessive-compulsive traits (SCID-II) .033 .072 .211 .646 .968 .840 1.114
Paranoid traits (SCID-II) .046 .066 .482 .488 .955 .840 1.087
Schizotypal traits (SCID-II) .010 .096 .011 .917 .990 .820 1.195
Schizoid traits (SCID-II) .018 .152 .014 .907 .982 .729 1.324
Histrionic traits (SCID-II) .017 .091 .033 .855 .984 .823 1.175
Narcissistic traits (SCID-II) .095 .121 .608 .435 .910 .717 1.154
Antisocial traits (SCID-II) .119 .089 1.779 .182 1.126 .946 1.341
Entitlement (YSQ) .005 .027 .029 .865 1.005 .952 1.060
Negative emotional problems
Anxiety symptoms (SCL-90) .049 .031 2.511 .113 1.050 .989 1.115
Hostility (SCL-90) .121 .051 5.545 .019 1.128 1.020 1.247
Depressive symptoms (SCL-90) .018 .018 1.005 .316 1.018 .983 1.054
Childhood abuse and neglect
Physical abuse (ITEC) .221 .097 5.161 .023 1.247 1.031 1.509
Sexual abuse (ITEC) .054 .107 .259 .611 1.056 .857 1.301
Emotional abuse (ITEC) .365 .216 2.851 .091 1.441 .943 2.203
Physical neglect (ITEC) .165 1.446 .013 .909 .848 .050 14.430
Emotional neglect (ITEC) .192 .218 .772 .380 1.211 .790 1.858
Final multivariate model
Condition (TFP vs ST) 1.051 .504 4.353 .037 2.860 1.066 7.673
Hostility (SCL-90) .146 .057 6.491 .011 1.158 1.034 1.296
Childhood physical abuse (ITEC) .261 .103 6.491 .011 1.298 1.061 1.588

Bold values denote p < .05.

a
All predictors following condition controlled for condition.
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 65

Table 2
Prediction of recovery.a

Predictor b s.e. Wald p-value OR 95% Confidence interval

Lower Upper

Condition (TFP vs ST) 1.226 .516 5.640 .018 .294 .107 .807
Medication (use at baseline) 1.499 .565 7.042 .008 .223 .074 .676
Biographical characteristics
Gender (male vs female) .429 .964 .198 .656 .651 .098 4.304
Age (in years) .003 .031 .012 .914 1.003 .944 1.067
Educational level (5 levels, low to high) .047 .193 .058 .809 .955 .654 1.392
Highest employment level (7 levels, low to high) .001 .130 .000 .995 .999 .775 1.289
Having a partner .005 .508 .000 .992 1.005 .372 2.719
Having a paid job .595 .639 .867 .352 .552 .158 1.931
Receiving disability compensation .790 .664 1.414 .234 2.203 .599 8.101
Treatment history
Having had psychological therapy .001 .721 .000 .999 .999 .243 4.105
Number of previous treatments .326 .213 2.331 .127 .722 .475 1.097
History of inpatient treatment .091 .516 .031 .860 .913 .332 2.510
BPD features
BPD severity (BPDSI-IV total) .014 .032 .200 .655 .986 .926 1.049
BPD severity (burden of BPD symptoms) .026 .011 5.758 .016 .974 .953 .995
Level of suicidality (BPDSI-IV) .140 .497 .080 .778 .869 .328 2.301
Burden of suicidality (BPD-checklist) .339 .167 4.130 .042 .712 .514 .988
Level of self-injury (BPDSI-IV) .068 .212 .103 .749 .934 .616 1.416
Burden of self-injury (BPD-checklist) .373 .239 2.440 .118 .689 .431 1.100
Level of substance abuse (BPDSI-IV) .167 .202 .690 .406 1.182 .796 1.755
Burden of substance abuse (BPD-checklist) .004 .140 .001 .977 1.004 .763 1.320
Level of interpersonal instability (BPDSI-IV) .036 .187 .037 .848 1.037 .718 1.496
Burden of interpersonal instability (BPD-checklist) .030 .029 1.053 .305 .970 .916 1.028
Level of impulsivity (BPDSI-IV) .025 .226 .013 .910 1.026 .659 1.597
Burden of impulsivity (BPD-checklist) .098 .059 2.778 .096 .906 .807 1.071
Level of anger problems (BPDSI-IV) .010 .138 .005 .942 1.010 .771 1.323
Burden of anger problems (BPD-checklist) .162 .083 3.786 .052 .850 .722 1.001
Level of dissociation (BPDSI-IV) .030 .024 1.520 .218 .971 .926 1.018
Burden of dissociation (BPD-checklist) .230 .088 6.885 .009 .794 .669 .943
Quality of current relations (WHOQOL) .025 .101 .062 .803 .975 .800 1.188
Defense mechanisms & personality organization
Mature defense mechanisms (DSQ) .018 .023 .624 .430 1.018 .974 1.064
Neurotic defense mechanisms (DSQ) .013 .015 .833 .361 .987 .959 1.015
Immature defense mechanisms (DSQ) .005 .010 .237 .626 .995 .975 1.015
Lower level defenses (IPO) .015 .023 .455 .500 .985 .942 1.030
Identity diffusion (IPO) .010 .020 .235 .625 .990 .952 1.030
Pathological object relations (IPO) .013 .014 .863 .353 .988 .962 1.014
Alterations in reality testing (IPO) .094 .283 .110 .740 .910 .523 1.585
Superego pathology (IPO) .195 .329 .349 .555 .823 .432 1.570
Axis-II comorbidity
Number of personality disorders (SCID-II) .262 .223 1.377 .241 .770 .497 1.192
Avoidant traits (SCID-II) .058 .063 .832 .362 .944 .834 1.068
Dependent traits (SCID-II) .130 .072 3.256 .071 .878 .763 1.011
Obsessive-compulsive traits (SCID-II) .071 .076 .867 .352 .931 .802 1.082
Paranoid traits (SCID-II) .007 .069 .011 .917 .993 .867 1.137
Schizotypal traits (SCID-II) .002 .102 .000 .988 1.002 .819 1.224
Schizoid traits (SCID-II) .312 .224 1.949 .163 .732 .472 1.134
Histrionic traits (SCID-II) .073 .094 .610 .435 1.076 .895 1.293
Narcissistic traits (SCID-II) .086 .121 .511 .475 .917 .724 1.162
Antisocial traits (SCID-II) .020 .092 .046 .831 1.020 .852 1.220
Entitlement (YSQ) .020 .031 .408 .523 .981 .924 1.041
Negative emotional problems
Anxiety symptoms (SCL-90) .050 .033 2.359 .125 .951 .892 1.014
Hostility (SCL-90) .113 .059 3.742 .053 .893 .796 1.002
Depressive symptoms (SCL-90) .003 .019 .022 .882 .997 .960 1.035
Childhood abuse and neglect
Physical abuse (ITEC) .023 .094 .061 .805 .977 .813 1.174
Sexual abuse (ITEC) .096 .115 .696 .404 1.101 .878 1.380
Emotional abuse (ITEC) .166 .218 .578 .447 1.180 .770 1.811
Physical neglect (ITEC) 1.894 1.538 1.516 .218 6.648 .326 135.532
Emotional neglect (ITEC) .109 .230 .226 .634 1.115 .711 1.750
Final multivariate model
Condition (TFP vs ST) 1.241 .537 5.335 .021 .289 .101 .829
Medication use 1.259 .589 4.564 .033 .284 .089 .901
Burden of dissociation (BPD-checklist) .230 .088 6.885 .009 .794 .669 .943

Bold values denote p < .05.

a
All predictors following condition and medication controlled for these two variables.
66 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71

Fig. 1. Relationship between therapy success (recovery) and level of burden of dissociative symptoms at baseline (grouped by quintiles) by treatment condition, estimates from the
logistic regression, controlled for baseline medication use.

literature search did not yield a suitable observation rating scale of Hayes (2008a, b), which uses a bootstrap percentile approach to
dissociation. Items were derived from various sources, including estimate the 95% CI of the indirect path and is suitable for logistic
textbooks (e.g., Spiegel, 1994) and reviews (e.g., Korzekwa, Dell, regression with multiple covariates. Significance is concluded if the
Links, Thabane, & Fougere, 2009), self-report scales (Stiglmayr, 95% CI does not contain zero. To achieve high precision of the esti-
Shapiro, Stieglitz, Limberger, & Bohus, 2001; Vanderlinden, Van mation, the number of bootstrap samples was set at 50,000 (Koehle,
Dyck, Vandereycken, Vertommen, & Verkes, 1993), and interviews Brown, & Haneuse, 2009). Mediation is concluded when first, sec-
(Giesen-Bloo et al., 2010; Steinberg, 1993). Dissociative areas ond and last tests are significant, and there is a reduction of the
captured by items included depersonalization, derealization, predictive power of the original predictor in the third test. As tapes
amnesia, and disproportional anger, fear and distrust of the ther- were unavailable for 6 participants, analyses were repeated with
apist (see Appendix for the scale). The disproportional responses estimated in-session ratings for these missings: (1) by assigning the
towards the therapist items were added as expressions of more mean of the in-session dissociation factor scores to the missings;
severe forms of dissociation where these phenomena result from a and (2) by assigning a regression-based score to the missings, with
shift to a dissociated emotional state (usually viewed as related to baseline dissociation, condition and medication as predictors.
traumatic experiences) that drives them and makes them dispro-
portional. Three independent raters rated a random selection of 7. Results
tapes of months 2e12 of therapy (one tape per participant), per
rater 33e36 tapes were double rated. Four items (3,6,8,11) showed Table 3 presents the results of the mediation analyses. In all
very little variance, probably because raters were unable to infer analyses the relationship between baseline dissociation and in-
these rather internal experiences from the audiotapes, and were session dissociation was significant, as was the relationship be-
therefore disregarded for further analysis. The interrater agreement tween in-session dissociation and recovery. The association-
per item was high, mean ICC ¼ .82, median ICC ¼ .85. Item ratings strength between baseline dissociation and recovery reduced af-
were next averaged over raters. A principal component analysis ter adding in-session dissociation as predictor, to non-significant
supported a one factor solution, with mean factor loading .60, range levels in the two analyses of the full sample. The lowest row of
.45e.73. For further computations the factor scores were used. the table shows that the mediation path (the contribution of
Recordings were available for 80 of the 86 participants. baseline dissociation to recovery via in-session dissociation) was
significant in the full sample analyses, as the 95% CIs did not contain
6.2. Mediation test zero. For the reduced sample (N ¼ 80) the upper limit of the 95% CI
was zero.
Mediation was tested using the Preacher and Hayes approach
(2008a). Fig. 2 illustrates the approach. First, the association be- 8. Discussion
tween baseline dissociation and in-session dissociation was tested
with linear regression. Second, the association between in-session The findings, despite rating only a single session, at least
dissociation and recovery (controlled for condition and medica- partially supported the hypothesis that in-session dissociation
tion) was assessed with logistic regression. Third, the change of the might account for the effect of baseline dissociation on treatment
association between baseline dissociation and recovery, controlling success. For the reduced sample (N ¼ 80) the results of the medi-
for condition and medication, with adding in-session dissociation ation analysis were a bit ambiguous, as the 95%CI of the mediation
was assessed with logistic regression. Lastly, the significance of the path has zero as upper limit, instead of being smaller than zero. The
indirect path between baseline dissociation and recovery via in- effect of baseline dissociation reduced but not to a nonsignificant
session dissociation (the mediator), controlling for condition and level after adding the mediator to the model, suggesting what is
medication, was assessed with an SPSS macro by Preacher and called “partial mediation”. The ambiguous findings seem related to
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 67

A. Unmediated model

CondiƟon
(TFP vs ST)

Baseline Recovery
MedicaƟon from BPD

Baseline
Burden of c
DissociaƟon

B. MediaƟon model
In-session
DissociaƟon

a b

CondiƟon
(TFP vs ST)

Recovery
Baseline
from BPD
MedicaƟon

Baseline
c'
Burden of
DissociaƟon

Fig. 2. Illustration of the mediation analysis of the effect of baseline burden of dissociation on recovery by in-session dissociation. Mediation is concluded when the direct effect c
(in upper model A) reduces when controlled for the mediator (c0 (lower model B) should reduce compared to c), and effects a and b are significant, and the indirect effect (the
mediation path) ab is significant. When c0 becomes nonsignificant, “full mediation” is concluded. The mediation was tested controlling for condition and baseline medication as
covariates.

Table 3
Results of mediation tests of the dissociation effect.

N ¼ 80 N ¼ 86, missing N ¼ 86, missing replacement

replacement by mean by regression estimate

b s.e. p b s.e. p b s.e. p

Baseline dissociation / in-session dissociation .086 .031 .008 .085 .030 .005 .083 .030 .007
In-session dissociation / recoverya .946 .452 .036 .993 .473 .036 1.061 .486 .029
Baseline dissociation / recoverya .241 .089 .007 .230 .088 .009 .230 .088 .009
Baseline dissociation / recovery, controlled for in-session dissociationa .189 .093 .043 .175 .092 .057 .170 .092 .064
95% CI 95% CI 95% CI
Lower Upper Lower Upper Lower Upper
Mediation patha .2091 .0000 .2230 .0002 .2377 .0023
a
(also) controlled for condition and medication.

the fact that we did not have recordings of six participants, as with recovery became nonsignificant when controlled for mediation.
both models using estimated mediation scores for these six par- However, in the subsample with recordings available, the
ticipants, unequivocal evidence for full mediation was obtained. mediation 95% CI had exactly zero as upper limit and the direct path
remained significant, though reduced in strength, thus yielding
9. General discussion mixed evidence.
Hostility and childhood physical abuse were predictive of
The first study reported in this manuscript aimed to test pre- treatment discontinuation, irrespective of treatment condition.
dictors of dropout and treatment success mentioned in previous Previous studies found evidence for hostility as predicting discon-
studies on (B)PD in data from an RCT on ST and TFP as treatments tinuation (Rüsch et al., 2008; Smith et al., 1995) and we now have
for BPD. We found evidence for hostility and childhood physical evidence that it predicts discontinuation of DBT, ST and TFP. Pa-
abuse as predictors of treatment discontinuation, and for subjective tients struggling with high levels of hostility might find it difficult
burden of dissociation as predictor of failure to recover from BPD. to go through the frustrating process of treatment, demand a quick
The second study explored whether in-session dissociation ex- fix, and might easily feel hostile towards their therapist. Therapists
plains the predictive power of burden of baseline dissociation. on the other hand often find it difficult to deal with hostile patients,
Mediation tests confirmed at least partially this hypothesis, as in and this might weaken the collaboration in treatment. Studies
the full sample (where in-session mediation had to be estimated investigating such processes are needed to better understand why
because of absence of audio-recordings in 6 participants) media- hostility increases chance of discontinuation. A cautious conclusion
tion was significant and the direct effect of baseline dissociation on that can be drawn from our finding is that therapists should
68 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71

develop ways to better tolerate and deal with hostility, to prevent medication effect (Nadort et al., 2009). Moreover, medication effects
premature discontinuation of treatment. Childhood abuse has been were not found in other treatment types (Bateman & Fonagy, 1999;
reported previously as relevant predictor, but this was not always Black et al., 2009; Ryle & Golynkina, 2000), though samples sizes
specifically physical abuse (Gunderson et al., 2006; Perry et al., were considerably smaller in first and last studies, and Black et al.
2007). investigated a short-term program to be added to regular treatment.
As different forms of childhood abuse tend to correlate (e.g., However, if the medication effect exists, how to explain it? Medi-
physical abuse usually takes place in a context of emotional abuse cation use might be a marker of the more difficult patient, although
and neglect, Lobbestael et al., 2009), we should be cautious to baseline severity did not differ between medication users and non-
conclude it is specifically physical abuse that predicts discontinu- users. Another possibility is that medication interferes with the
ation. Indeed, emotional abuse showed a trend significant rela- psychological change processes used in ST and TFP. For example,
tionship with discontinuation. However, physical abuse might be medications like antidepressants and antipsychotics might flatten
an index of very severe abuse that damages trust in others, emotional arousal (Moncrieff & Cohen, 2009; Moncrieff, Cohen, &
including therapists. Clearly, future studies should try to replicate Mason, 2009; Price, Cole, & Goodwin, 2009), possibly interfering
whether childhood physical abuse is a specific predictor of treat- with the levels of emotional arousal necessary for psychological
ment discontinuation. treatments like ST and TFP (Price & Goodwin, 2009), e.g. by damp-
As hostility might also be related to a trauma history, the finding ening brain areas involved in (corrective) emotional learning like
that both hostility and childhood physical abuse predict treatment the amygdala (Harmer, 2008), and/or dampening cognitive biases
discontinuation suggests that it is important to effectively process and other processes that should be corrected through experiences in
trauma with empirically supported methods integrated in the treatment (Merens, Van der Does, & Spinhoven, 2007; Pringle,
offered treatment. This is actually part of the ST protocol, and recent Browning, Cowen, & Harmer, 2011). Or, patients might attribute
studies have demonstrated that trauma processing can also be symptom reduction to medication, instead of to (stable) psycho-
effectively integrated in DBT (Bohus et al., 2013; Harned, Korslund, logical changes and their own effort in therapy, a kind of attribution
& Linehan, 2014). The inclusion of (childhood) trauma processing in that is known to predict relapse in the long-term in the treatment of
ST might be one of the explanations why dropout was less and anxiety disorders and depression (Basoglu, Marks, Kilic, Brewin, &
recovery higher in ST than in TFP. Swinson, 1994; Moradveisi, Huibers, & Arntz, 2015; Powers, Smits,
The finding that baseline burden of dissociation was predictive of Whitley, Bystritsky, & Telch, 2008). Clearly, an RCT comparing
a diminished chance to recover was not only a replication of a similar specialized psychotherapy with vs. without medication is needed to
finding by Kleindienst et al. (2011), but was also further explained by test causality implied in the hypothesis that medication interferes
in-session dissociation accounting for this. One explanation is that with psychotherapy.
dissociation during therapy sessions reduces the impact of correc- At least as important are alleged predictors that turned out to
tive experiences and information, because dissociation interferes have no significant predictive power. Sociodemographic variables
with information processing. It should be noted that there are in- like age, gender, education level, unemployment, and disability
dications that the interfering effect of dissociation on information compensation were not significant. Moreover, indices of severity of
processing specifically concerns highly negative and traumatic BPD including level of self-injury and suicidality, did not predict
material (Olsen & Beck, 2012), which are of course topics of psy- dropout and recovery. These findings are in line with reviews
chotherapy. Another explanation is that therapists when confronted pointing out that most assumed predictors have in fact no predic-
with dissociation in their patients find it more difficult to use the full tive power (Barnicot et al. 2012; McMurran et al., 2010). This in-
range of techniques. The effects of dissociation were quite powerful dicates that we should be reluctant in using such variables in
with only a few of the highest dissociative patients recovering (see treatment allocation; more specifically we should not withhold
Fig. 1, highest 20%). This calls for revising treatments to better deal specialist treatment to patients with a long treatment history,
with in-session dissociation. Interestingly, one of the explanations specific age, lower educational level, or poor social functioning, etc.
Braakman et al. (2007) offered why in their study dissociation did Moreover, indices of “low-level” borderline organization (impul-
not predict worse outcome was their effectively addressing disso- sivity problems, substance abuse, self-injury, anger control prob-
ciation during treatment. Interestingly, the last decade ST-therapists lems) did not predict discontinuation or recovery, suggesting that
reported having become less afraid of dissociation during sessions, “low-level borderline” should not be used as exclusion criterion.
using methods to get patients out of dissociative states, and teaching Some of the predictors were based on the baseline BPDSI-IV,
patients to prevent a massive shift into a dissociative state. Future including the BPDSI-IV total score as one of the indices of baseline
studies will show us whether indeed the field has improved in better BPD-severity. The same instrument was used to define recovery.
handling dissociation during treatment so that treatment effects are Usually one sees a positive correlation between baseline and posttest
less affected. on the same instrument, which might be partially caused by the fact
The findings from the original RCT that ST had less dropout and that the very same instrument is used. Interestingly, no evidence for
more recovery than TFP, and medication had a negative influence on a correlation between predictors based on the baseline BPDSI-IV and
recovery, were maintained. Thus these effects turned out to be recovery was found, which indicates that treatment and/or time
robust for the detected predictors. Moreover, we failed to find in- caused fundamental changes so that the correlation between base-
teractions indicating that predictors had a different effect in the two line BPDSI-IV and recovery was nil (see also Giesen-Bloo et al., 2006).
treatments. It is not clear why medication use predicts poorer ef- A clinical implication is that recovery by specialized psychotherapies
fects. However, it should be noted that medication use during like ST and TFP is not predicted by baseline BPD-severity, indicating
treatment was reduced, notably in antidepressants (from N ¼ 58 that this should not be used to select patients.
(67.4%) to N ¼ 36 (41.9%); Giesen-Bloo et al., 2006). Nevertheless, the Several limitations should be mentioned. First, although we were
negative association between medication and recovery persisted able to investigate many of the alleged predictors mentioned in the
when the use of psychotropic medication was analyzed as a time- literature, there was a substantial set we couldn't assess due to lack
dependent covariate in a survival analysis of recovery (13 assess- of instruments assessing the pertinent construct. These included a
ments; Wald statistic ¼ 6.21; P ¼ .01; RR ¼ .38; 95% CI, .18e.81; wide range of variables, e.g. juvenile conviction, parent divorce
Giesen-Bloo et al., 2006). Another study of TFP found similar effects before the age of 10, trait anxiety, distractibility, and pre-
(Doering et al., 2010), though a study on ST failed to replicate the contemplation stage of change. Second, the large number of tests
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 69

and the use of uncorrected p-levels increase the chance of accidental symptoms, condition and medication resulted, which was the very
findings. However, it is important that we tested whether previously same as the initial test of burden of dissociative symptoms e hence
documented predictors survived replication. Instead of trying to no inflation took place. For dropout, the two single significant pre-
prove indisputable predictors in a single study, it seems better to dictors were selected (hostility and childhood physical abuse),
work on establishing an empirical data base over studies that can be explaining each by and large the same as they did when tested singly
used in meta-analyses to document the power of various predictors e although a bit of inflation took probably place, it was marginal
of treatment discontinuation and success. Third, although we compared to the initial tests and did not lead to different conclu-
explored whether the strength of the predictors found to be sig- sions. Taken together, replication in an independent dataset remains
nificant in the whole sample differed between treatments, our essential, and the current procedures should be seen as aiming to
sample size was insufficient to detect more subtle differences select candidate predictors that need further testing in the future.
(medium effect size and smaller). Such differences are of obvious In conclusion, this study found that most variables that were
importance as they inform us about the best treatment for a specific previously reported to predict treatment discontinuation or effec-
patient. Fourth, although with estimations of the six missing re- tiveness did not survive replication. However, we replicated hos-
cordings evidence for full mediation was obtained, this does not tility and childhood abuse (though only physical) as predictors of
prove that in-session dissociation causes the effect: in-session discontinuation, and dissociation as predictor of reduced treatment
dissociation might also be a proxy of another process (for success. Moreover we found evidence that in-session dissociation
example, continuous dissociation, poor sleep quality, etc.) that ac- explains the effect of baseline dissociation on recovery, suggesting
counts for the relationship. In other words, a third variable that we need to better deal with in-session dissociation to improve
explaining the association is still possible. Nevertheless, the medi- effects of treatment. Similarly, therapists might need to learn to
ation finding indicates that it is worthwhile to test whether a revised better deal with high levels of hostility in their patients, and with
treatment that reduces in-session dissociation would lead to better the aftermaths of severe childhood (physical) abuse to prevent
recovery in patients who suffer from high levels of dissociation. premature discontinuation.
Fifth, due to financial limitations, only one recording per patient was
rated, limiting the representativeness of the ratings per individual. Declarations
On the other hand, with only one recording we found evidence that
in-session dissociation mediated the predictive relationship of Financial support
baseline dissociation and recovery, which pleads for the power of
the effect. However, an additional limitation of having only one The RCT from which the data are derived that form the basis of
session ratings of dissociation is that the relationship between in- this manuscript was funded by grant OG-97.002 from the Dutch
dividual level of dissociation (and its change during treatment), and Health Care Insurance Board. The Dutch National Fund of Mental
treatment response, could not be studied on an individual level. For Health supported central training of the therapists. The grant or-
that repeated assessments of level of dissociation and treatment ganizations played no role in the data collection and analysis,
response are needed, and advanced multilevel tests distinguishing manuscript preparation, or authorization for publication.
intra- and inter-individual sources of (co-)variance. Sixth, in the
absence of validated dissociation observation rating scales that Conflict of interest
could be used to rate audio recordings, we had to develop one, of
which the validity is (except for the present findings) unknown. None.
Seventh, it is unclear why subjective burden of dissociation (BPD-
checklist) and not frequency of dissociation (BPDSI-IV) predicted
Ethics
recovery. Similarly, it is not clear why hostility assessed with the
SCL-90 predicted discontinuation, and not seemingly related vari-
The authors assert that all procedures contributing to this work
ables like anger control problems; and why childhood physical
comply with the ethical standards of the relevant national and
abuse predicted discontinuation, and not other types of abuse and
institutional committees on human experimentation and with the
neglect. Future studies are needed to shed light on these issues.
Helsinki Declaration of 1975, as revised in 2008. The medical ethics
Eighth, as the RCT was not planned as a prediction study some of the
committees of the participating centers approved the study (see
predictors had to be constructed from the available assessment in-
Giesen-Bloo et al., 2006).
struments, while for some perhaps more reliable or valid in-
struments are available. Some of the null findings might be related
to this, though reliabilities of many predictors were reasonable. Acknowledgments
Ninth, stepwise regression methods can lead to overfitting of the
model, with inflated beta's, R2 values etc. Replication with the same Thanks are due to Eva van den Hurk, Lauren van Litsenburg and
set of predictors is necessary to come to unbiased estimation. Note Rinske Zopfi for rating the sessions tapes. We acknowledge the
however that the aim of the stepwise procedure was not to create a contributions of the participating patients, therapists and consul-
numerical prediction model to predict chances of dropout or re- tants, the screening psychologists, and the research assistants.
covery for new patients, but to assess whether predictors that were
(trend) significant when tested initially alone would survive Appendix. Dissociation rating scale
multivariate tests. Interestingly the resulting models were not or
only in a very limited way different from the models of the single Indicate to which degree the phenomenon is shown by the
predictors: For recovery, the model with burden of dissociative patient during the session (not caused by medical condition or
substance abuse).
70 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71

1. Depersonalisation: Not feeling oneself anymore, like an (2012). Treatment outcome of 18-month, day hospital mentalization-based
treatment (MBT) in patients with severe borderline personality disorder in
outside observer of oneself, or experiencing oneself as in a
the Netherlands. Journal of Personality Disorders, 26(4), 568e582.
movie or dream. Barnicot, K., Katsakou, C., Bhatti, N., Savill, M., Fearns, N., & Priebe, S. (2012). Factors
2. Depersonalisation: Feeling of alienation or unreality that is predicting the outcome of psychotherapy for borderline personality disorder: a
usually frightening and is related to one's own thoughts, systematic review. Clinical Psychology Review, 32, 400e412.
Basoglu, M., Marks, I. M., Kilic, C., Brewin, C. R., & Swinson, R. P. (1994). Alprazolam
emotions, or body, during which reality testing is intact. and exposure for panic disorder with agoraphobia: attribution of improvement
3. Depersonalisation: The body feels different than normal, to medication predicts subsequent relapse. British Journal of Psychiatry, 164,
sometimes one feels nothing, or it seems as if body parts are 652e659.
Bateman, A., & Fonagy, P. (1999). The effectiveness of partial hospitalisation in the
not one's own. treatment of borderline personality disorder: a randomised controlled trial.
4. Derealisation: Observing or experiencing the world around American Journal of Psychiatry, 156, 1563e1569.
oneself totally differently so that it seems very odd or unreal Bellino, S., Bozzatello, P., & Bogetto, F. (2015). Combined treatment of borderline
personality disorder with interpersonal psychotherapy and pharmacotherapy:
(e.g. others look unfamiliar or like ‘robots’) predictors of response. Psychiatry Research, 226(1), 284e288.
5. Derealisation: Feeling of alienation or unreality that is Bellino, S., Paradiso, E., & Bogetto, F. (2006). Efficacy and tolerability of quetiapine in
usually frightening and relates to the environment, during the treatment of borderline personality disorder: a pilot study. Journal of Clinical
Psychiatry, 67(7), 1042e1046.
which reality testing is intact.
Berghuis, H., Kamphuis, J. H., Boedijn, G., & Verheul, R. (2009). Psychometric
6. Derealisation: Hearing or seeing things not good or they properties and validity of the Dutch inventory of personality organization (IPO-
sound or look different than normal. NL). Bulletin of the Menninger Clinic, 73(1), 44e60.
Black, D. W., Allen, J., St John, D., Pfhol, B., McCormick, B., & Blum, N. (2009). Pre-
7. Dissociative amnesia: Not being able to remember impor-
dictors of response to systems training for emotional predictability and prob-
tant things. lem solving (STEPPS) for borderline personality disorder: an exploratory model.
8. Dissociative amnesia: Inability to recall important personal Acta Psychiatrica Scandinavica, 120, 53e61.
information that is usually of a traumatic or stressful nature. Bohus, M., Dyer, A. S., Priebe, K., Krüger, A., Kleindienst, N., Schmahl, C., et al. (2013).
Dialectical behaviour therapy for post-traumatic stress disorder after childhood
9. Dissociative amnesia: Not knowing anymore what one has sexual abuse in patients with and without borderline personality disorder: a
done or where one is. randomised controlled trial. Psychotherapy and Psychosomatics, 82(4), 221e233.
10. Disproportional or even inexplicable anger toward the Braakmann, D., Ludewig, S., Milde, J., Stiglmayr, C. E., Hüppe, M., Sipos, V., et al.
(2007). Dissoziative symptome im Verlauf der Behandlung der Borderline-
therapist: Angry behavior toward the therapist, consisting of Perso€ nlichkeitssto € rung [Dissociative symptoms during treatment of borderline
raised voice, changes of timbre (e.g., higher harmonic, even personality disorder]. Psychotherapie Psychosomatik Medizinische Psychologie,
with the same undertone), and message content. Note: Ex- 57, 154e160.
Calvin, C. M., Fernandes, C., Smith, P., Visscher, P. M., & Deary, I. J. (2010). Sex, in-
ercises excluded. telligence and educational achievement in a national cohort of over 175,000 11-
11. Disproportional or even inexplicable anger toward the year-old schoolchildren in England. Intelligence, 38, 424e432.
therapist: Aggressive behavior such as yelling, scolding, Chiesa, M., & Fonagy, P. (2007). Prediction of medium-term outcome in cluster B
personality disorder following residential and outpatient psychosocial treat-
threatening, etc. or physically aggressive behavior. Note:
ment. Psychotherapy and Psychosomatics, 76, 347e353.
Exercises excluded. Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational
12. Disproportional mistrust of the therapist: Constantly achievement. Intelligence, 35, 13e21.
Derogatis, L. R., Lipman, R. S., & Covi, L. (1973). SCL-90: an outpatient psychiatric
questioning the good intentions of the therapist. Asking
rating scale: preliminary report. Psychopharmacological Bulletin, 9, 13e28.
controlling questions. Dickhaut, V., & Arntz, A. (2014). Combined group and individual schema therapy for
13. Disproportional fear of the therapist: Constantly asking for borderline personality disorder: a pilot study. Journal of Behavior Therapy and
confirmation to get constancy/certainty. Experimental Psychiatry, 45, 242e251.
Doering, S., Horz, S., Rentrop, M., Fischer-Kern, M., Schuster, P., Benecke, C., et al.
14. Disproportional mistrust of the therapist: Lying or con- (2010). Transference-focused psychotherapy versus treatment by community
cealing things to the therapist out of disproportionate psychotherapists for borderline personality disorder: randomised controlled
mistrust. trial. British Journal of Psychiatry, 196, 389e395.
First, M. B., Gibbon, M., Spitzer, R. L., Williams, J. B. W., & Benjamin, L. S. (1997).
15. Disproportional fear of the therapist: Telling little or User's guide for the structured clinical interview for DSM-IV axis II personality
nothing, or telling something very superficial. Remaining disorders (SCID-II). Washington, DC: American Psychiatric Press.
very correct. Constantly keeping an eye on what the therapist First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1996). Structured clinical
interview for DSM-IV axis I disordersepatient edition (SCID-I/P), version 2.0. New
does. York: Biometrics Research Department, New York State Psychiatric Institute.
Gerritsen, M., Berg, I., & Deelman, B. (2001). De Snijders-Oomen niet-verbale
intelligentietest: bruikbaar bij ouderen? [[The Snijders-Oomen nonverbal in-
telligence test: feasible for elderly?]] Tijdschrift voor Gerontologie en Geriatrie,
References 32, 24e28.
Giesen-Bloo, J., van Dyck, R., Spinhoven, P., van Tilburg, W., Dirksen, C., van
Andrews, G., Pollock, C., & Stewart, G. (1989). The determination of defense style by Asselt, T. … Arntz, A. (2006). Outpatient psychotherapy for borderline person-
questionnaire. Archives of General Psychiatry, 46, 455e460. ality disorder. Randomized trial of schema-focused therapy vs transference-
Arntz, A. (1999). Do personality disorders exist? On the validity of the concept and focused psychotherapy. Archives of General Psychiatry, 63(6), 649e658.
its cognitive-behavioral formulation and treatment. Behaviour Research and Giesen-Bloo, J. H., Wachters, L. M., Schouten, E., & Arntz, A. (2010). The borderline
Therapy, 37, 97e134. personality disorder severity index-IV: psychometric evaluation and dimen-
Arntz, A., Bernstein, D., Gielen, D., Nieuwenhuyzen van, M., Penders, K., Haslam, N., sional structure. Personality and Individual Differences, 49, 136e141.
et al. (2009). Taxometric evidence for the dimensional structure of cluster-c, Groenestijn, M. A. C., Akkerhuis, G. W., Kupka, R. W., Schneider, N., & Nolen, W. A.
paranoid, and borderline personality disorders. Journal of Personality Disor- (1999). Structured clinical interview for DSM-IV Axis I disorders. Lisse, the
ders, 23(6), 606e628. Netherlands: Swets & Zeitlinger BV.
Arntz, A., van den Hoorn, M., Cornelis, J., Verheul, R., van den Bosch, W. M., & de Gunderson, J. G., Daversa, M. T., Grilo, C. M., McGlashan, T. H., Zanarini, M. C.,
Bie, A. J. (2003). Reliability and validity of the borderline personality disorder Shea, M. T. … Stout, R. L. (2006). Predictors of 2-year outcome for patients with
severity index. Journal of Personality Disorders, 17(1), 45e59. borderline personality disorder. American Journal of Psychiatry, 163, 822e826.
Arrindell, W. A., & Ettema, J. H. (2003). SCL-90 Symptom checklist: Handleiding bij een Harmer, C. J. (2008). Serotonin and emotional processing: does it help explain
multidimensionele psychopathologie-indicator. Lisse: Swets Test Publishers [SCL- antidepressant drug action? Neuropharmacology, 55(6), 1023e1028.
90 Symptom checklist: manual of a multidimensional psychopathology Harned, M. S., Jackson, S. C., Comtois, K. A., & Linehan, M. M. (2010). Dialectical
indicator]. behavior therapy as a precursor to PTSD treatment for suicidal and/or self-
van Asselt, A. D. I., Dirksen, C. D., Arntz, A., Giesen-Bloo, J. H., Dyck van, R., injuring women with borderline personality disorder. Journal of Traumatic
Spinhoven, P., Tilburg van, W., Kremers, I. P., Nadort, M., & Severens, J. L. (2008). Stress, 23(4), 421e429.
Outpatient psychotherapy for borderline Personality disorder. Cost-effective- Harned, M. S., Korslund, K. E., & Linehan, M. M. (2014). A pilot randomized
ness of schema-focused therapy versus transference-focused psychotherapy. controlled trial of dialectical behavior therapy with and without the dialectical
British Journal of Psychiatry, 192(6), 450e457. behavior therapy prolonged exposure protocol for suicidal and self-injuring
Bales, D., van Beek, N., Smits, M., Willemsen, S., Busschbach, J. J., Verheul, R., et al. women with borderline personality disorder and PTSD. Behaviour Research
 A. Arntz et al. / Behaviour Research and Therapy 74 (2015) 60e71 71

and Therapy, 55, 7e17. Journal of Consulting and Clinical Psychology, 76(3), 478.
Kleindienst, N., Limberger, M. F., Ebner-Priemer, U. W., Keibel-Mauchnik, J., Dyer, A., Preacher, K. J., & Hayes, A. F. (2008a). SPSS macro for multiple mediation, written by
Berger, M. … Bohus, M. (2011). Dissociation predicts poor response to dialectical A.F. Hayes. The Ohio State University Accessed 27.03.15 http://www.afhayes.
behavioral therapy in female patients with borderline personality disorder. com.
Journal of Personality Disorders, 25(4), 432e447. Preacher, K. J., & Hayes, A. F. (2008b). Asymptotic and resampling strategies for
Koehle, E., Brown, E., & Haneuse, S. J. P. A. (2009). On the assessment of Monte Carlo assessing and comparing indirect effects in multiple mediator models. Behavior
error in simulation-based statistical analyses. The American Statistician, 63, Research Methods, 40, 879e891.
155e162. Price, J., Cole, V., & Goodwin, G. M. (2009). Emotional side-effects of selective se-
Korzekwa, M. I., Dell, P. F., Links, P. S., Thabane, L., & Fougere, P. (2009). Dissociation rotonin reuptake inhibitors: qualitative study. The British Journal of Psychiatry,
in borderline personality disorder: a detailed look. Journal of Trauma & Disso- 195(3), 211e217.
ciation, 10(3), 346e367. Price, J., & Goodwin, G. M. (2009). Emotional blunting or reduced reactivity
Kro€ger, C., Vonau, M., Kliem, S., Ro €pke, S., Kosfelder, J., & Arntz, A. (2013). Psycho- following remission of major depression. Medicographia, 31, 152e156.
metric properties of the German version of the borderline personality disorder Pringle, A., Browning, M., Cowen, P. J., & Harmer, C. J. (2011). A cognitive neuro-
severity INDEX e version IV. Psychopathology, 46, 396e403. psychological model of antidepressant drug action. Progress in Neuro-
Lenzenweger, M. F., Clarkin, J. F., Kernberg, O. F., & Foelsch, P. A. (2001). The in- Psychopharmacology and Biological Psychiatry, 35(7), 1586e1592.
ventory of personality organization: psychometric properties, factorial Rijkeboer, M. M., & van den Bergh, H. (2006). Multiple group confirmatory factor
composition, and criterion relations with affect, aggressive dyscontrol, psy- analysis of the young schema-questionnaire in a Dutch clinical versus non-
chosis proneness, and selfdomains in a non-clinical sample. Psychological clinical population. Cognitive Therapy and Research, 30(3), 263e278.
Assessment, 13, 577e591. Rijkeboer, M. M., van den Bergh, H., & van den Bout, J. (2005). Stability and
Leppa €nen, V., Hakko, H., Sintonen, H., & Lindeman, S. (2015). Comparing effec- discriminative power of the young schema-questionnaire in a Dutch clinical
tiveness of treatments for borderline personality disorder in communal mental versus a non-clinical population. Journal of Behavior Therapy and Experimental
health care: the Oulu BPD study. Community Mental Health Journal, 1e12. Psychiatry, 36, 129e144.
Leppa €nen, V., Lindeman, S., Arntz, A., & Hakko, H. (2013). Preliminary evaluation of Rüsch, N., Schiel, S., Corrigan, P. W., Leihener, F., Jacob, G. A., Olschewski, M., et al.
psychometric properties of the Finnish borderline personality disorder severity (2008). Predictors of dropout from inpatient dialectical behaviour therapy
index: Oulu-BPD-study. Nordic Journal of Psychiatry, 67, 312e319. among women with borderline personality disorder. Journal of Behavior Therapy
Links, P. S., Mitton, J. E., & Steiner, M. (1990). Predicting outcome for borderline and Experimental Psychiatry, 39, 497e503.
personality disorder. Comprehensive Psychiatry, 31(6), 490e498. Ryle, A., & Golynkina, K. (2000). Effectiveness of time-limited cognitive analytic
Lobbestael, J., Arntz, A., Harkema-Schouten, P., & Bernstein, D. (2009). Development therapy of borderline personality disorder: factors associated with outcome.
and psychometric evaluation of a new assessment method for childhood British Journal of Medical Psychology, 73, 197e210.
maltreatment experiences: the interview for traumatic events in childhood. Schuppert, H. M., Timmerman, M. E., Bloo, J., van Gemert, T. G., Wiersema, H. M.,
Child Abuse and Neglect, 33, 505e517. Minderaa, R. B. … Nauta, M. H. (2012). Emotion regulation training for adoles-
Lobbestael, J., Leurgans, M., & Arntz, A. (2011). Inter-rater reliability of the structural cents with borderline personality disorder traits: a randomized controlled trial.
and clinical interview for DSM-IV axis (SCID I) and axis 2 disorders (SCID II). Journal of the American Academy of Child & Adolescent Psychiatry, 51(12),
Clinical Psychology and Psychotherapy, 18, 75e79. 1314e1323.
€ ffler-Stastka, H., Voracek, M., Leithner, K., Fischer-Kern, M., Kunz, C., & Meissel, T.
Lo Smith, T. E., Koenigsberg, H. W., Yeomans, F. E., Clarkin, J. F., & Selzer, M. A. (1995).
(2003). Predicting psychotherapy utilization for patients with borderline per- Predictors of dropout in psychodynamic psychotherapy of borderline person-
sonality disorder. Psychotherapy Research, 13(2), 255e264. ality disorder. Journal of Psychotherapy Practice and Research, 4, 205e213.
Luteijn, F., & Barelds, D. P. H. (2004). Groninger Intelligentie test 2: Handleiding Spiegel, D. (1994). Dissociation: Culture, mind, and body. Washington, DC: American
[[Groninger intelligence test 2: manual]]. Amsterdam: Harcourt Assessment B.V. Psychiatric Press.
McMurran, M., Huband, N., & Overton, E. (2010). Non-completion of personality Steinberg, M. (1993). Structured clinical interview for DSM-IV dissociative disorders
disorder treatments: a systematic review of correlates, consequences, and in- (SCID-D). Arlington, VA, US: American Psychiatric Association.
terventions. Clinical Psychology Review, 30, 277e287. Stiglmayr, C. E., Shapiro, D. A., Stieglitz, R. D., Limberger, M. F., & Bohus, M. (2001).
Merens, W., Van der Does, A. W., & Spinhoven, P. (2007). The effects of serotonin Experience of aversive tension and dissociation in female patients with
manipulations on emotional information processing and mood. Journal of Af- borderline personality disorderda controlled study. Journal of Psychiatric
fective Disorders, 103(1), 43e62. Research, 35(2), 111e118.
Moncrieff, J., & Cohen, D. (2009). How do psychiatric drugs work? BMJ, 338. Tambs, K., Sundet, J. M., Magnus, P., & Berg, K. (1989). Genetic and environmental
Moncrieff, J., Cohen, D., & Mason, J. P. (2009). The subjective experience of taking contributions to the covariance between occupational status, educational
antipsychotic medication: a content analysis of internet data. Acta Psychiatrica attainment, and IQ: a study of twins. Behavior Genetics, 19, 209e222.
Scandinavica, 120(2), 102e111. Thorma €hlen, B., Weinryb, R. M., Nore n, K., Vinnars, B., Bågedahl-Strindlund, M., &
Moradveisi, L., Huibers, M. J., & Arntz, A. (2015). The influence of patients' attri- Barber, J. P. (2003). Patient factors predicting dropout from supportive-
butions of the immediate effects of treatment of depression on long-term eexpressive psychotherapy for patients with personality disorders. Psycho-
effectiveness of behavioural activation and antidepressant medication. Behav- therapy Research, 13, 493e509.
iour Research and Therapy, 69, 83e92. Vanderlinden, J., Van Dyck, R., Vandereycken, W., Vertommen, H., & Verkes, J. R.
Nadort, M., Arntz, A., Smit, J. H., Giesen-Bloo, J., Eikelenboom, M., Spinhoven, P., (1993). The dissociation questionnaire (DIS-Q): development and characteris-
et al.Dyck van, R. (2009). Implementation of outpatient schema therapy for tics of a new self-report questionnaire. Clinical Psychology & Psychotherapy, 1(1),
borderline personality disorder with versus without crisis support by the 21e27.
therapist outside office hours: a randomized trial. Behaviour Research and Verheul, R., van den Bosch, L. M., Koeter, M. W., De Ridder, M. A., Stijnen, T., & Van
Therapy, 47, 961e973. Den Brink, W. (2003). Dialectical behaviour therapy for women with borderline
Olsen, S. A., & Beck, J. G. (2012). The effects of dissociation on information pro- personality disorder 12-month, randomised clinical trial in the Netherlands.
cessing for analogue trauma and neutral stimuli: a laboratory study. Journal of British Journal of Psychiatry, 182(2), 135e140.
Anxiety Disorders, 26, 225e232. Wechsler, D. (2005). WAIS-III-NL Nederlandstalige bewerking. Technische handleiding
Perry, J. C., Bond, M., & Roy, C. (2007). Predictors of treatment duration and (herziene uitgave 2005) [[WAIS-III-NL Dutch adaptation. Technical Manual (revised
retention in a study of long-term dynamic psychotherapy: childhood adversity, edition 2005)]]. Amsterdam: Harcourt Test Publishers.
adult personality, and diagnosis. Journal of Psychiatric Practice, 13, 221e232. Weertman, A., Arntz, A., & Kerkhofs, M. L. M. (2000). Structured clinical interview for
Plakun, E. M. (1991). Prediction of outcome in borderline personality disorder. DSM- IVAxis II personality disorders. Lisse, the Netherlands: Swets & Zeitlinger
Journal of Personality Disorders, 5, 93e101. BV.
Plassman, B. L., Welsh, K. A., Helms, M., Brandt, J., Page, W. F., & Breitner, J. C. S. Zanarini, M. C., Stanley, B., Black, D. W., Markowitz, J. C., Goodman, M., Pilkonis, P.,
(1995). Intelligence and education as predictors of cognitive state in late life: a et al. (2010). Methodological considerations for treatment trials for persons
50-year follow-up. Neurology, 45, 1446e1450. with borderline personality disorder. Annals of Clinical Psychiatry: Official Jour-
Powers, M. B., Smits, J. A., Whitley, D., Bystritsky, A., & Telch, M. J. (2008). The effect nal of the American Academy of Clinical Psychiatrists, 22(2), 75.
of attributional processes concerning medication taking on return of fear.