Treatment differences in the therapeutic relationship and introject during a 2-year randomized controlled trial of dialectical behavior therapy versus non-behavioral psychotherapy experts for borderline personality disorder

Abstract

Objective

The purpose of the present study was to explore the role of the therapeutic relationship and introject during the course of dialectical behavior therapy (DBT; Linehan, 1993) for the treatment of borderline personality disorder.

Method

Women meeting DSM-IV criteria for borderline personality disorder (N = 101) were randomized to receive DBT or community treatment by experts. The Structural Analysis of Social Behavior (SASB; Benjamin, 1974) was used to measure both the therapeutic relationship and introject.

Results

Using hierarchical linear modeling, DBT patients reported the development of a more positive introject including significantly greater self-affirmation, self-love, self-protection, and less self-attack during the course of treatment and one-year follow-up relative to community treatment by experts. The therapeutic relationship did not have an independent effect on intrapsychic or symptomatic outcome but did interact with treatment. DBT patients who perceived their therapist as affirming and protecting reported less frequent occurrences of non-suicidal self-injury.

Conclusions

The study showed positive intrapsychic change during DBT while emphasizing the importance of affirmation and control in the therapeutic relationship. Results are discussed in the context of understanding the mechanisms of change in DBT.

Evidence has continued to accumulate supporting the efficacy of dialectical behavior therapy (DBT; Linehan, 1993) for the treatment of borderline personality disorder (BPD; APA, 1994; Morey & Zanarini, 2000). DBT has been shown to be superior to treatment as usual and treatment by expert clinicians across a variety of domains of functioning including suicide attempts, emergency/inpatient treatment, intentional self-injury, anger, depression, and social and global adjustment among others (Kliem, Kroger, & Kosfelder, 2010; Lynch, Trost, Salsman, & Linehan, 2007). A critique of DBT, and these associated findings, has been DBT's perceived emphasis on symptomatic change as opposed to elements of patients’ experience that represent internal or intrapsychic change (e.g., Scheel, 2000). Similarly, behavioral approaches have historically been perceived as underemphasizing the use of the therapeutic relationship for the sake of technique (Lejuez, Hopko, Levine, Gholkar, & Collins, 2006). Broadly speaking these critiques speak to issues of the underlying mechanisms of change in DBT (e.g., Neacsiu, Rizvi, & Linehan, 2010). The current study sought to test these critiques by examining intrapsychic change and the role of the therapeutic relationship in DBT compared to a community treatment by experts (CTBE) control condition.

According to interpersonal theory (Sullivan, 1953), the concept of the introject can be defined as an aspect of an individual's personality that consists of self-directed actions including cognitive self-appraisals, and verbal and physical actions directed toward the self (Henry, 1996; Henry, Schacht, & Strupp, 1990). These internal, self-directed actions are thought to be fairly stable across the lifespan, reflective of the actions of early caregivers, and conceptually related to one's self-concept (Pincus, Gurtman, & Ruiz, 1998a). The concept of the introject has been particularly useful in furthering our understanding of BPD. Benjamin and Wonderlich (1994) found BPD patients to have an overall attacking and abandoning introject when compared to patients diagnosed with major depressive disorder and bipolar disorder. Additional studies have suggested that a hostile internal relationship with the self may act as a regulatory strategy for managing aversive experiences related to overall BPD symptom severity (Rosenthal, Cukrowicz, Cheavens, & Lynch, 2006) and that BPD patients value their negative self-concept as a desired part of their identity (Janis, Veague, & Driver-Linn, 2006).

Based on a biosocial model of BPD (Linehan, 1993), DBT treats the development of a positive self-concept as a primary target of therapy. Although not using the term introject per se, patients are taught methods of appropriate self-evaluation, methods of tolerating disapproval, and ways of decreasing self-invalidating behaviors that are antithetical to patient goals. In an exploratory test of introject change during DBT, Shearin and Linehan (1992) showed that patients who perceived their therapist as providing more control and instruction were in turn perceived by their therapist as having an introject that was more self-caring. Despite DBT's theoretical emphasis on patient self-concept, critics have suggested that behavioral treatments, such as DBT, fail to incorporate important personality factors, such as the introject (Benjamin, 1997; Scheel, 2000; Shedler, 2010). These same critiques suggest that the effectiveness of DBT in the treatment of BPD can be accounted for by symptom reduction rather than intrapsychic changes in patient factors considered more relevant to improvement when treating personality disorders. In light of the theoretical importance of the self-concept in DBT, and as a response to critiques, a goal of the present study was to examine the nature of introject change in DBT during the course of a randomized controlled trial of DBT.

One mechanism hypothesized to result in intrapsychic change during DBT is the therapeutic relationship (Lynch, Chapman, Rosenthal, Kuo, & Linehan, 2006). The therapeutic relationship is a common treatment factor that has been shown to have a consistent effect on outcome across a variety of therapeutic modalities (Castonguay, Constantino, & Grosse Holtforth, 2006). Although historically perceived as underemphasized in behavior therapy, recent work has begun to more fully explicate the role the therapeutic relationship in cognitive and behavioral interventions (Gilbert & Leahy, 2007; Grosse Holtforth & Castonguay, 2005; Lejuez et al., 2006). In DBT, a positive therapeutic relationship is considered essential to treatment and serves two functions (Linehan, 1993; Robins & Koons, 2000). First, the relationship between the therapist and patient is considered therapeutic as the therapist provides an accepting, nonjudgmental, and compassionate stance toward the patient. Second, the therapeutic relationship also provides an avenue for the therapist to help the patient regulate their affect and behavior. Thus, the relationship also allows the therapist to exert control and influence where the patient might be initially lacking the skills to do so themselves. During the course of treatment the therapist attempts to balance the two poles of acceptance and change as a method of simultaneously validating the patients’ experience and also correcting dysfunctional behaviors – a dialectic of DBT.

In a preliminary analysis of this dialectical style, Shearin and Linehan (1992) examined the therapeutic relationship in four therapist-patient dyads during DBT. Results showed patients’ perceptions of simultaneous control, autonomy, and warmth to be predictive of an overall decrease in suicidal behavior (i.e., non-suicidal self-injury, urges to self-harm, and suicide ideation) in two of the four cases. An overall significance test supported the dialectical hypothesis and results were maintained while controlling for non-dialectical hypotheses including therapists’ behavior that was purely autonomous or purely controlling. Although based on a few cases, these results provided preliminary data supporting the importance of acceptance and control in the interpersonal stance of the DBT therapist. Based on these initial impressions, the current study attempted to examine the therapeutic relationship in DBT using a much larger treatment sample in the context of a randomized controlled trial of DBT.

The present study had the following five hypotheses. First, we explored introject change during the course of DBT. Given DBT's emphasis on self-concept change, our first hypothesis predicted DBT patients to show a more affiliative introject during the course of treatment. In light of DBT's emphasis on therapists’ warmth and control, our second hypothesis predicted DBT therapists to be perceived by patients as emphasizing greater levels of affirmation, protection, and control during treatment. Our third and fourth hypotheses explored the association between therapist affiliation with introject and non-suicidal self-injury (NSSI) in DBT. We expected higher ratings of overall therapist affiliation to be related to increased introject affiliation (hypothesis three) and less frequent NSSI (hypothesis four) for DBT patients. Our fifth hypothesis predicted that the simultaneous use of emancipating, affirming, protecting, and controlling behavior, the DBT dialectic, to predict improved outcome. With each hypothesis, DBT was compared to CTBE. Although there were not specific hypotheses about how these two treatments would differ, CTBE practitioners were notably psychodynamic and eclectic in orientation and hence provides a meaningful comparator relative to our hypotheses about DBT.

Method

Participants

Participants (N = 101) were female and ranged in age from 18 to 45 years old. All participants were diagnosed using structured clinical interviews for Axis I and Axis II of the Diagnostic and Statistical Manual of Mental Disorders – Fourth edition (APA, 1994) and all received the diagnosis of borderline personality disorder. Average age for the entire sample was 29.3 (SD=7.5). The majority of participants reported their race as Caucasian (87%), marital status as single, divorced or separated (87%) and an annual income of less than $15,000 (75%). No significant differences in demographics were found between treatment conditions. All participants in the study were required to endorse a history of self-inflicted injury defined by at least two suicide attempts or NSSI in the past five years and a minimum of one incident in the past eight weeks. Exclusionary criteria included a lifetime history of schizophrenia, schizoaffective disorder, bipolar disorder, psychotic disorder not otherwise specified, and mental retardation. Individuals were also excluded under circumstances when treatment was mandated, a seizure disorder requiring medication was present, there was a primary need to treat another condition, or the patient was street homeless and participation interfered with access to housing services.

The participant coordinator randomly assigned participants to treatment conditions, either DBT (n = 51) or CTBE (n = 49; see Figure 1), using a computerized adaptive minimization randomization procedure based on five prognostic variables including: 1) number of psychiatric hospitalizations, 2) number of lifetime suicide attempts and NSSI, 3) a history of only suicide attempts, NSSI or both, 4) age, and 5) scores on the Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) or Global Assessment of Functioning (First, Spitzer, Gibbon, Williams, & Benjamin, 1995). Participants were enrolled between October 1994 through August 2003. Outcome measures were self-report and administered by blinded, independent clinical assessors. The protocol was approved by the University of Washington Institutional Review Board. Additional details about procedures as well as treatments can be found in Linehan et al. (2006b).

Figure 1.

Participant Flowchart. DBT = dialectical behavior therapy; CTBE = community treatment by experts

Treatments

Dialectical Behavior Therapy (DBT; Linehan, 1993)

DBT is a cognitive behavioral therapy originally developed for the treatment of individuals exhibiting suicidal behavior and later expanded to those meeting criteria for borderline personality disorder. DBT consists of a blend of behavioral problem solving techniques (e.g., functional analyses, behavioral skills training, exposure/response prevention, contingency management, cognitive restructuring) with acceptance strategies (e.g., validation, interpersonal reciprocity). Treatment interventions are organized based on dialectical processes, where the primary dialectic is acceptance and change. Sixteen therapists were recruited for the DBT treatment condition based on recommendations by colleagues. Of these therapists, eight had no previous DBT experience and eight had a range of exposure to DBT. Study therapists received 45-hours of DBT training including supervised practice. DBT therapists were assigned patients after being successfully rated as adherent to DBT practice in six of eight consecutive therapy sessions.

Community nominated Treatment by Experts (CTBE)

The CTBE condition consisted of therapists nominated by leaders in the mental health community as experts in the treatment of difficult patients. Thirty-eight therapists were selected from a larger pool of ninety-four therapists. Of the thirty-eight therapists, twenty-five were assigned patients. Treatment provided by CTBE therapists was uncontrolled by the research team. Selected therapists described their theoretical orientation as “eclectic” or “mostly psychodynamic” with no cognitive behavioral therapists in the CTBE condition. Therapists were encouraged to provide the type of therapy and dose they would normally prescribe with a minimum of one individual session scheduled per week and additional treatment prescribed as needed.

Measures

Therapeutic Relationship and Patient Introject

The current study used Benjamin's Structural Analysis of Social Behavior (SASB) Intrex short form (Benjamin, 1983) to measure the therapeutic relationship and patient introject (see Figure 2). The SASB (Benjamin, 1974) is a three-surface circumplex model of interpersonal and intrapsychic behavior rooted in the theoretical traditions of Sullivan (1953) and Leary (1957). The SASB model in Figure 2 differentiates each surface by bold, underlined, or italicized labels. The first two surfaces are interpersonal in nature. Surface one, with emboldened descriptors, represents actions that are transitive in nature where the focus of behavior is directed towards another person. Surface two, with underlined descriptors, represents intransitive reactions where the focus of behavior is the self. The third surface, or introject, is defined by italicized descriptors and represents transitive actions directed towards the self. Each surface is a two-dimensional model based on the dimensions of affiliation and interdependence. The horizontal axis for each surface reflects the affiliation dimension and ranges from hate (left) to love (right). The vertical axis for each of the interpersonal surfaces reflects the interdependence dimension and captures the degree of differentiation (emancipate/separate; top) and enmeshment (control/submit; bottom) in the relationship. The vertical axis for the introject captures the degree of self-emancipating versus self-controlling behavior. Surrounding each surface are eight behavioral descriptors referred to as clusters. Each cluster is a blend of the affiliation and interdependence dimensions. The simplified model presented in Figure 2 defines each cluster by a single word for each of the aforementioned surfaces.

Figure 2.

Structural Analysis of Social Behavior

The SASB was rated by participants using the Intrex Short Form (Benjamin, 1983). Each item on the Intrex short form measures one SASB cluster score for each of three SASB surfaces. Patients participating in the present study rated eight items from surface one to measure their therapists’ actions directed towards them and eight items from surface three to rate their introject. Each of the sixteen items was rated on a scale from 0 (never, not at all) to 100 (always, perfectly) in 10-point increments. SASB ratings of the therapists’ behavior were assessed during the active phase of both treatments at 4-months, 8-months, and 12-months. Introject ratings were made at two weeks into treatment and then again during treatment at 4-months, 8-months, and 12-months and also during post-therapy follow-up at 16-months, 20-months, and 24-months. At the two-week assessment for the introject, participants were asked to “indicate how well each question describes yourself in the past four months.” Follow-up instructions for both therapist and introject ratings asked participants to indicate how well each question described their therapist and themselves “since your last assessment.”

There are several available methods for summarizing SASB Intrex data (Benjamin, 2000; Pincus, Newes, Dickinson, & Ruiz, 1998b). Of these methods, the most general summary indices are affiliation and interdependence dimension ratings. Both of these indices are weighted transformations of SASB cluster scores where clusters on the horizontal axis are weighted positively for affiliative behavior and weighted negatively for disaffiliative behavior. SASB clusters on the vertical axis are weighted positively for independence or autonomy (i.e., emancipate-separate) and negatively for enmeshment (i.e., control-submit). These summary scores can be used to capture overall levels of affiliation and interdependence for each of the SASB surfaces and are recommended for large, nomothetic databases as they typically assume a normal distribution (Pincus et al., 1998b). In the present study, affiliation and interdependence dimension ratings were used to provide a general summary of the degree of affiliation and autonomy-control in the transitive actions of the therapist towards the patient as rated by patients. In order to test hypotheses related to the introject, we calculated an affiliation dimension rating for patients’ self-reported introject. In a validation study of the SASB, Pincus et al. (1998a) found introject affiliation to be reflective an individual's self-esteem, sense of agency, self-acceptance and the ability to “more confidently engage and adapt in the social world and take charge of things when it is required” (p. 1642).

Pattern coefficients are a second method of summarizing Intrex data and are the product-moment correlations between participant's eight SASB cluster scores with 21 theoretical patterns centered on various clusters of the circumplex model (Benjamin & Wonderlich, 1994). These theoretical curves define a specific interpersonal pattern taking the form of a single cosine curve, linear polynomial curve, or nonlinear polynomial curve. In the present study, we were interested in a theoretical pattern of therapist behavior consistent with the dialectic of DBT where therapists are perceived as simultaneously emancipating, affirming, protecting, and controlling defined as the DBT pattern coefficient (Schmidt, 2001; Shearin & Linehan, 1992).

The final and most specific index of SASB Intrex ratings are the individual raw cluster scores. Based on our hypothesis we examined four SASB cluster scores for patients’ ratings of therapists’ actions on surface one including affirm, active love, protect, and control. Based on hypotheses, we also examined four SASB cluster scores for introject including self-affirm, active self-love, self-protect, and self-attack. In summary, we used SASB dimension ratings of therapist affiliation, therapist autonomy-control, introject affiliation, the DBT pattern coefficient, and cluster scores of therapist behavior and introject in our analyses.

Non-Suicidal Self-Injury

The Suicide Attempt Self-Injury Interview (Linehan, Comtois, Brown, Heard, & Wagner, 2006a) was used to measure suicide intent, non-suicidal self-injury, and medical severity of each suicide attempt. Interrater reliabilities ranged from 0.88 to 0.94. In the current study we used non-suicidal self-injury as our measure of self-harm, as opposed to suicide attempts, due to its greater frequency throughout treatment allowing for more accurate estimates for statistical modeling across time. The use of non-suicidal self-injury is also consistent with prior studies examining the association between the therapeutic relationship and suicidal behavior in DBT (Shearin & Linehan, 1992).

Data Analysis

Hierarchical Linear Modeling (HLM; Raudenbush & Bryk, 2002), also known as mixed effects or multilevel modeling (Pinheiro & Bates, 2000), was the primary data analytic tool on the intent-to-treat sample. Compared to traditional methods, HLM is more flexible in its treatment of time as a continuous factor, allowing for variability in the actual time of assessment for each participant. In addition, HLM can model incomplete data across time making analyses more powerful due to the inclusion of more data points. Lastly, HLM allows for time-varying and time-invariant covariates as a method of exploring the cross-time association and change between two factors. HLM assumes normality, homoskedasticity, and independence of the error terms. All assumptions were met unless otherwise noted.

In order to test the first two hypotheses, our basic model for treatment differences for introject and therapist behavior across time, is shown in Equation 1:

$$\begin{array}{cc}\hfill \text{Level}1:& SAS{B}_{ti}={\pi }_{0i}+{\pi }_1{\left(Time\right)}_{ti}+{\pi }_2{\left(Tim{e}^2\right)}_{ti}+e_{ti}\hfill \\ \hfill \text{Level}2:& {\pi }_{0i}={\beta }_{00}+{\beta }_{01}Tx+r_{0i}\hfill \\ \hfill & {\pi }_1={\beta }_{10}+{\beta }_{11}Tx\hfill \\ \hfill & {\pi }_2={\beta }_{20}+{\beta }_{21}Tx\hfill \end{array}$$ (1)

where t indexes time and I indexes individuals. Time measured in weeks from the start of treatment was modeled using linear and quadratic terms, allowing for nonlinear change. Tx is a treatment dummy-variable, coded 0 for DBT and 1 for CTBE. The cross-level interactions between Tx and Time (i.e., β₁₁ and β₂₁) model treatment differences in the trajectories across time. The Level 2 variance term (i.e., r_0i) models heterogeneity in individual participant trajectories. Additional random effects were also examined (e.g., slope, quadratic effect), but deviance tests revealed they did not improve the fit of the model.

Hypotheses three, four and five examined the association between: 1) therapist behavior and introject and 2) therapist behavior and NSSI across both treatments. In order to test these hypotheses we followed the recommendations of Singer and Willett (2003) in using time-varying covariates (TVCs) in which the TVCs were added as an additional fixed effect at level 1 of the model, as shown in equation 2:

$$\begin{array}{cc}\hfill \text{Level}1:& D{V}_{ti}={\pi }_{0i}+{\pi }_1{\left(Time\right)}_{ti}+{\pi }_2{\left(Tim{e}^2\right)}_{ti}+{\pi }_3{\left(TVC\right)}_{ti}+e_{ti}\hfill \\ \hfill \text{Level}2:& {\pi }_{0i}={\beta }_{00}+{\beta }_{01}Tx+r_{0i}\hfill \\ \hfill & {\pi }_1={\beta }_{10}+{\beta }_{11}Tx\hfill \\ \hfill & {\pi }_2={\beta }_{20}+{\beta }_{21}Tx\hfill \\ \hfill & {\pi }_3={\beta }_{30}+{\beta }_{31}Tx\hfill \end{array}$$

where DV indicates either NSSI or SASB introject; t indexes time and I indexes individuals. Time measures the assessment period in weeks. Equation 2 captures the initial DV score for DBT (β₀₀ due to the coding of Tx), change in the DV over time for DBT (β₁₀ , β₂₀), and the cross-time association between DV and SASB rated therapist behavior. The Tx terms describe the difference between DBT and CTBE for each parameter. In addition to the same time analyses we also conducted a series of lagged analyses to further examine the cross-time association between perceived therapist behavior with introject and NSSI. Using the model in equation two, these analyses provided us with a general sense of order of change across these variables (e.g., Are prior assessment period ratings of therapist behavior associated with patients’ self-reported introject in the next assessment period?).

Given that NSSI is a count variable and positively skewed, a Poisson mixed-effects regression was used for all analyses involving NSSI as the outcome (Raudenbush & Bryk, 2002). Poisson mixed effects models for HLM assume that level 1 data follows a Poisson distribution, which is often appropriate for count data such as number of NSSI instances (Atkins & Gallop, 2007). There is one additional, critical aspect of the Poisson mixed-effects model for interpretation. The Poisson model uses a natural logarithm link function, and the coefficients are typically exponentiated (i.e., e^B) to be interpreted, sometimes called the rate-ratio and somewhat similar to interpreting odds-ratios for logistic regression.

Plots of predicted regressions were used to facilitate interpretation of all models. Sensitivity analyses to examine the effect of differential rates of dropout on hypotheses were conducted and did not alter results. All data analyses were conducted in R v2.6.1 (R Development Core Team, 2009).

Results

Hypothesis One: DBT patients will show a more affiliative introject during the course of treatment

Our initial step in examining the SASB introject was to explore changes in the overall dimension ratings for introject affiliation across treatment and time, where time was assessed as number of weeks in treatment through one-year follow-up. The model in equation 1 was fit for introject affiliation as the outcome. Treatment (DBT = 0 and CTBE = 1) differences at the initial assessment point were not significant. The effect for change in DBT across treatment was significant, B = 1.06, SE = .09, t(366) = 11.60, p < 0.00, CI = 0.88, 1.24 (all CIs are 95% CI), where DBT patients reported an increase in introject affiliation through treatment and follow-up. A treatment by weeks in treatment interaction was found for introject affiliation, B = -.44, SE = .14, t(366) = -3.23, p < 0.01, CI = -.72, -0.17, where participants in DBT reported a significantly greater increase in introject affiliation during treatment and follow-up compared to CTBE (see Figure 3). In Figure 3 it is important to note that the zero point on the vertical axis is meaningful in that it indicates a change in overall introject from hostile to affiliative. As can be seen in Figure 3, the DBT group reaches this zero point prior to the end of treatment while the CTBE group begins to approach this value towards the end of one-year follow-up.

Figure 3.

Patient ratings of introject affiliation across treatment and time.

SASB introject clusters scores provided a more detailed, exploratory, analysis of change in introject affiliation during the course of treatment. Results showed a significant effect of change in DBT for SASB introject clusters including self-affirm, B = 0.22, SE = 0.03, t(370) = 7.28, p < 0.00, CI = 0.17, 0.29, active self-love, B = 0.26, SE = 0.03, t(371) = 9.10, p < 0.00, CI = 0.21, 0.32, self-protect, B = 0.25, SE = 0.03, t(370) = 8.73, p < 0.00, CI = 0.19, 0.31, and self-attack, B = -0.31, SE = 0.03, t(370) = -8.92, p < 0.00, CI = -0.38, -0.24, where DBT patients reported perceiving themselves as significantly more self-affirming, self-loving, self-protecting, and less self-attacking across treatment and through follow-up. In comparison to CTBE, results showed a significant interaction where DBT patients reported more self-affirm, active self-love, self-protect, and less self-attack across treatment and through follow-up.¹

Hypothesis Two: DBT therapists will be perceived by patients as emphasizing greater levels of affirmation, protection, and control during treatment

A total of 76 (43=DBT; 33=CTBE) participants in our sample completed ratings of the therapeutic relationship. Results did not reveal significant differences in outcome measures for participants completing versus not completing ratings of the therapeutic relationship. Of the participants completing ratings of therapeutic relationship, three in the DBT condition and four in the CTBE condition reported rating a different therapist during the course of treatment with all of the transitions occurring prior to the second assessment of the therapeutic relationship. Sensitivity analyses were conducted to examine the effect of therapist switching on the study hypotheses and found no effect of therapist switching on the results presented below. We fitted the model in equation 1 for each of the SASB clusters across treatment and time, where time was assessed as number of weeks in treatment starting at approximately four months into treatment through termination. Results supported a significant quadratic effect for therapist affirm during DBT, B = -.02, SE = 0.01, t(102) = -3.13, p < 0.00, CI = -.04, -.01, where DBT therapists were perceived as increasingly more affirming early in treatment with a plateau and deceleration towards the end of treatment. Results were not significant for change in therapist protect during treatment. Ratings of perceived therapist control, showed strong positive skews with a high frequency of zero endorsement. Given the nature of the distribution, a Poisson distribution was selected as the most appropriate model for hypothesis testing. Results supported a quadratic effect for therapist control during DBT, B = -0.00, SE = 0.00, z = -2.81, p < 0.00, where DBT therapists were perceived as increasingly controlling early in treatment and less controlling as treatment approached termination. In comparison to CTBE, results supported a significant interaction between treatment condition and the quadratic effect of weeks in treatment for therapist affirm, therapist protect, and therapist control where DBT patients reported their therapists as increasingly more affirming, protecting, and controlling during treatment compared to CTBE therapists.

We conducted two exploratory analyses to examine change in the dimensional ratings for therapist affiliation and therapist autonomy-control. Dimensional ratings are composite indices of all SASB clusters and provide an overall sense of therapist behavior. The HLM models showed a significant linear increase for therapist affiliation, B = 0.60, SE = 0.27, t(102) = 2.22, p < 0.05, CI = 0.07, 1.13, and significant quadratic effect for therapist autonomy-control, B = 0.04, SE = 0.02, t(100) = 2.41, p < 0.05, CI = 0.01, 0.08, for DBT patients where patients perceived their therapist as increasingly affiliative and less autonomy granting during the course of treatment. The quadratic effect for autonomy-control suggested that the tendency for DBT therapists to be perceived as less autonomy granting showed a plateau towards the end of treatment. In comparison to CTBE, the interaction between treatment condition and the quadratic effect of weeks in treatment for therapist affiliation and therapist autonomy-control were both significant. In these models the DBT therapists were perceived as increasingly more affiliative and less autonomy granting compared to CTBE therapists (see Figure 4). In contrast, CTBE patients perceived their therapist as autonomy granting throughout treatment and perceived a deceleration of therapist affiliation in the middle phase of treatment and an increase towards the end of treatment, although never reaching levels comparable to DBT therapists.

Figure 4.

Patient affiliation and autonomy ratings of the therapeutic relationship across treatment and time.

Hypothesis Three: Increased therapist affiliation will be associated with increased introject affiliation in DBT

In order to test the third hypothesis we applied equation 2 to examine the association between perceived therapist affiliation and introject affiliation across weeks in treatment and treatment group. Our primary variables included: 1) the dimensional ratings of therapist affiliation with introject affiliation and 2) the SASB cluster ratings of affiliative behavior including therapist affirm with introject self-affirm, therapist active love with introject active self-love, and therapist protect with introject self-protect. The four HLM analyses examining the association between these variables in DBT, when rated at the same assessment period, were not significant nor were the independent effects of therapist affiliation on introject affiliation.

Using the same variables, we ran a series of four lagged HLM analyses examining the association between ratings of the therapist behavior and next period ratings of introject. Dimensional ratings of therapist affiliation on next period introject affiliation in DBT were not significant nor were the independent effects of therapist affiliation on next period introject affiliation. Analysis of the lagged association between the SASB clusters revealed two significant findings. DBT patients who reported greater ratings of therapist active love reported more introject self-love in the following assessment period, B = 0.23, SE = 0.08, t(97) = 2.99, p < 0.00, CI = 0.08, 0.39. Similarly DBT patients who reported greater ratings of therapist protect reported an increase in introject self-protect in the following assessment period, B = 0.18, SE = 0.09, t(97) = 2.11, p < 0.04, CI = 0.01, 0.35. Results were not significant for an increased association between therapist affirm and next period ratings of introject self-affirm in DBT.

In comparison to CTBE, DBT patients reported a stronger, positive association between therapist affirm and next period ratings of introject self-affirm, B = -0.37, SE = 0.21, t(96) = -2.25, p < 0.03, CI = -0.69, -0.04. In contrast, CTBE patients reported a tendency for the opposite pattern where higher ratings of therapist affirm predicted less introject self-affirm in the following assessment period. DBT patients also reported a stronger, more positive association between therapist active love and next period ratings of introject self-love compared to CTBE, B = -0.26, SE = 0.11, t(97) = -2.32, p < 0.03, CI = -0.47, -0.04 (See Figure 5). Results were not significant when comparing treatment differences in the lagged association between therapist protect and introject self-protect or dimensional ratings of therapist affiliation and introject affiliation.

Figure 5.

The association between patient ratings of therapist active love with next period patient ratings of introject self-love across treatment and time.

Hypothesis Four: Increased therapist affiliation will be associated with less frequent NSSI in DBT

The fourth hypothesis examined the association between SASB rated therapist affiliation and NSSI during DBT. Poisson HLM models showed no significant effect for the dimensional rating of therapist affiliation on NSSI apart from treatment. In the DBT condition, patients who perceived their therapists as more affiliative also reported less NSSI, B = -0.87, SE = 0.45, z = -1.94, p < 0.05, regardless of time in treatment. In comparison to CTBE, DBT therapists reported a significantly greater association between increased therapist affiliation and less NSSI, B = 0.01, SE = 0.00, z = 2.36, p < 0.02.

Supplementary analyses examined the specific SASB therapist clusters contributing to this overall effect. Analyses resulted in three significant effects where higher levels of therapist affirm, B = -0.01, SE = 0.00, z = -2.37, p < 0.05, higher levels of therapist active love, B = -0.01, SE = 0.00, z = -2.56, p < 0.05, and higher levels of therapist protect, B = -0.01, SE = 0.00, z = -2.70, p < 0.05, were associated with fewer occurrences of NSSI for DBT patients. In comparison to CTBE, results showed a significant treatment interaction for therapist affirm and therapist protect where DBT patients reported a stronger association between increased affirmation and protection with decreased NSSI. In contrast, CTBE showed the opposite pattern where patients who reported higher levels of therapist affirm showed more frequent NSSI while therapist protect showed little association with NSSI. Lagged analyses examining the order of change between patient perception of therapist behavior and next period ratings of NSSI were not significant.

Hypothesis Five: The simultaneous use of emancipating, affirming, protecting, and controlling behavior, the DBT dialectic, will predict improved outcome in DBT

The DBT pattern coefficient for therapist behavior assessed the degree to which patients perceived their therapists as acting in a way consistent with a theoretically defined pattern of behavior where therapists were seen as simultaneously emancipating, affirming, protecting, and controlling. The average correlation between patient Intrex ratings of therapist behavior with the DBT dialectic pattern of behavior was 0.67, regardless of treatment and assessment period. The DBT dialectic pattern was not significantly different across treatment, regardless of time, B = 0.00, SE = 0.01, z = 0.08, p = 0.94. We found no significant independent main effects for patient ratings of the DBT pattern of therapist behavior on introject or NSSI. However, using the Poisson distribution in an HLM model, we found a significant effect for the DBT pattern of behavior on NSSI for DBT patients, B = -0.01, SE = 0.00, z = -2.66, p < 0.05, where higher patient ratings of the DBT pattern were associated with a decrease in NSSI. The results for the DBT condition remained significant regardless of levels of overall perceived therapist affiliation, B = -0.01, SE = 0.00, z = -2.46, p < 0.05, and perceived therapist autonomy, B = -0.01, SE = 0.00, z = -2.44, p < 0.05. In comparison to CTBE, results supported a significant interaction between treatment and time where DBT patients reported a stronger association between an increase in the DBT pattern and a decrease in NSSI. Results showed the opposite pattern for CTBE patients where higher patients ratings of the DBT pattern were associated with an increase in NSSI.

Discussion

The current study provided us with the opportunity to respond to two criticisms of dialectical behavior therapy (DBT) stating that 1) DBT is symptom focused with little to no impact on intrapsychic or personality related factors and 2) behavioral therapies underemphasize the therapeutic relationship. The current findings also revealed potential mechanisms of change during the course of DBT for the treatment of borderline personality disorder. Using Benjamin's Structural Analysis of Social Behavior (Benjamin, 1974) we explored the nature of change for both the introject and the therapeutic relationship in a randomized controlled trial of DBT versus CTBE for BPD.

As predicted, individuals assigned to DBT reported significantly greater increases in introject affiliation including self-affirmation, self-love, self-protection, and greater decreases in self-attack during the course of treatment and one-year follow-up. These results also surpassed the CTBE condition. Consistent with prior literature (Janis et al., 2006; Rosenthal et al., 2006), participants in both conditions started therapy with an overall hostile, critical, and punishing introject. By the end of treatment and through one-year follow-up, patients in DBT moved from an overall hostile and self-punishing introject to an overall affiliative and protective introject that could best be characterized as a tendency to work hard in order to take care of oneself. CTBE patients, although showing less self-directed hostility over time in treatment, showed significantly less improvement in their introject relative to the DBT group. These data suggest that the benefits of DBT go beyond symptom reduction and extend to elements of patients’ personality that have been shown to be reflective of an individual's self-esteem and sense of social competence.

The therapeutic relationship in DBT, as perceived by the patient, appeared to involve complex elements of both affirmation and control that varied as a function of time in treatment. During the course of treatment, DBT therapists were perceived as increasingly more accepting and managing of the patient. Towards the end of the treatment the interpersonal stance of the DBT therapist appeared to shift as they were perceived as somewhat less affiliative and less controlling although scores suggested they maintained an overall affiliative and instructional stance. These results are consistent with a DBT model of the therapeutic relationship that combines warmth with management and structure (Linehan, 1993). In contrast to DBT therapists, CTBE therapists were perceived by their patients as affiliative and autonomy granting. As treatment approached termination, CTBE therapists appeared to exert less autonomy and more control. Overall, these results show two very different trajectories of the therapeutic relationship including: 1) a DBT style that is initially very warm, accepting, and managing while becoming less controlling as treatment approached termination and 2) a CTBE control style that that was generally warm and autonomous.

An additional objective of the study was to explore the association between the therapeutic relationship and outcome. By itself the therapeutic relationship did not predict introject or NSSI outcomes across treatments. Our results did, however, show a significant interaction between the therapeutic relationship and treatment condition for both introject and NSSI. In examining the effect of the therapeutic relationship on introject, our findings showed that DBT patients who reported their therapists as actively loving and protecting also reported greater self-love and self-care in the following assessment period. In comparison to CTBE, DBT therapists showed a stronger, positive association between therapist affirmation and therapist active love with following period ratings of introject self-affirm and self-love, respectively. CTBE therapists showed a tendency for the reverse effect where therapist affirmation was related to less patient self-acceptance in the following assessment period. These results could be interpreted as further support for the complex role of validation and affirmation in psychotherapy in general (e.g., Karpiak & Benjamin, 2004; Linehan, 1997). The present findings lend support to the hypothesis that DBT therapists may utilize validation strategies as a method of confirming or modeling a positive self-view for their patients (Lynch et al., 2006). The opposite effect for CTBE therapists may be indicative of a less precise use of validation that may, unintentionally, reinforce a hostile view of self (e.g., Swann, 1997).

We found a similar interaction when testing the association between perceived therapist behavior and NSSI by treatment condition. During the same time period, DBT patients who perceived higher levels of therapist warmth, including affirmation, love, and protection, reported less NSSI. These results might reflect the tendency for DBT therapists to use the therapeutic relationship, including affirmation and guidance, as a method of reinforcing reductions in NSSI. Interestingly, our data showed an opposite pattern for CTBE patients. In CTBE, patients who perceived higher levels of therapist affirmation also reported increased NSSI. These results may be indicative of a tendency for CTBE therapists to respond to high levels of NSSI with affirmation that, from a behavioral standpoint, would be reinforcing of dysfunctional behavior as therapists attempt to manage a difficult situation.

The therapeutic relationship in DBT is also unique as therapists attempt to balance strategies of acceptance and change. In balancing these strategies, DBT therapists also balance levels of autonomy and control in their interpersonal stance towards the patient. Based on prior results (Shearin & Linehan, 1992), we created an index of dialectical behavior where therapists are perceived as concurrently emancipating, affirming, protecting, and controlling. Similar to our results examining general levels of warmth and autonomy-control in the therapeutic relationship, the DBT pattern did not have a significant impact on introject or NSSI apart from its interaction with treatment. The DBT pattern of behavior was also not seen as occurring significantly more frequently for DBT therapists than CTBE therapists. Our results did, however, show that the effectiveness of the DBT pattern on NSSI was solely associated with its use by DBT therapists relative to CTBE therapists. For DBT patients, higher levels of the perceived dialectical pattern were associated with fewer instances of NSSI whereas higher levels of the DBT pattern were associated with an increase in NSSI for CTBE patients. Furthermore, the effect of the DBT pattern, for DBT patients only, held even after controlling for general levels of warmth and autonomy-control in the therapeutic relationship.

These results confirmed and extended prior results that showed perceptions of DBT therapists as simultaneously instructing, controlling, and granting of autonomy to be associated with decreased suicidal behavior (Shearin & Linehan, 1992). Our data also suggest that the DBT defined pattern, although not unique to DBT therapists, may only be an effective interpersonal stance for DBT clinicians who are skilled and trained in the use of a complicated and multifaceted interpersonal stance. An alternative hypothesis, based on the fact that the DBT dialectical stance is mandated in the treatment manual under specified conditions, is that therapists who rigorously follow a treatment manual have other characteristics that make them more competent therapists in enacting such a complex therapeutic stance. Although patients with BPD may pull for therapists to respond in a manner consistent with the defined DBT dialectic, CTBE therapists, who were more likely to be eclectic rather than manual-based, may have lacked a rationale for responding in a dialectical manner, which in turn could have iatrogenic effects.

Limitations and future directions

The current study has a number of strengths including the use of multiple time-points for the assessment of both introject, symptomatic change, and the therapeutic relationship all within the context of a larger randomized controlled trial. There are also a number of limitations to our study. In terms of measurement, our assessment of core BPD symptoms in this study was limited to NSSI. Further research is needed to extend these results to other domains relevant to BPD such as emotion dysregulation and interpersonal problems. While we used several time-points for the assessment of introject and NSSI, we had only three time-points for the assessment of the therapeutic relationship which limited our ability to test hypotheses predicting sequential change in behaviors. In addition, our first assessment of the therapeutic relationship asked participants to rate the behavior of their therapist during the first four months of treatment. Such a method, while capturing the initial phase of treatment, does not capture the earliest perceptions of the therapeutic relationship that have been shown to be a predictor of certain therapeutic outcomes (Klein et al., 2003).

An additional strength of this study was in the use of well-validated instruments for the assessment of intrapsychic outcome and the therapeutic relationship. Despite this strength our study was limited by the fact that all measurements were taken from the patient's perspective. Our data could be extended by including multiple perspectives of outcome and the therapeutic relationship including those of the therapist and independent observers. Additionally, the therapeutic relationship is a multifaceted construct that can be operationally defined in numerous ways (Castonguay et al., 2006). Our conceptualization of the therapeutic relationship was limited to the interpersonal relationship between patient and therapist and does not speak to other constructs related to the therapeutic relationship (e.g., agreement on goals, tasks, etc.).

An additional limitation was the present study's use of quarterly, self-reported, ratings of the therapeutic relationship that, although beneficial, provided a limited perspective on the DBT dialectic of acceptance and change. The therapeutic relationship in DBT is one that is fluid and dynamic where clinicians are required to be flexible enough to take on a variety of interpersonal stances dependent upon the dominant dialectic presented in session. One such dialectic is that of acceptance and change. Given the fluid nature of the therapeutic relationship, a natural extension of the current study would be to examine the moment-to-moment interpersonal process in individual DBT sessions using observational ratings. Such a methodology would likely provide a more refined lens to better capture the dynamic movement, speed, and flow of a DBT session including therapists’ contingent use of affirmation and control in relation to patient behavior.

Our statistical tests limited our ability to make causal associations and instead are suggestive of possible causal pathways linking the therapeutic relationship, introject, and NSSI. The number of analyses that examined the order of change in NSSI, introject, and therapist behavior did not correct for a possible increase in Type I error. These concerns, however, are partly abated by the consistent findings regarding introject change and the association between therapist behavior and NSSI and introject in expected directions. Nevertheless, replication of these findings is needed. Lastly, a limitation of the present study was the lack of detailed understanding of the interventions provided by the CTBE control therapists. Our study is therefore limited to speaking to introject change and the therapeutic relationship in DBT and not specific, alternative interventions that comprised the CTBE control condition.

This was the first study to examine intrapsychic factors in the outcome of DBT and the first study to examine the therapeutic relationship in a randomized controlled trial of DBT. The study demonstrated the beneficial impact of DBT on personality factors and NSSI and the importance of considering the therapeutic relationship in DBT. Future research is needed to study this phenomenon in more depth and with a wider range of outcomes.