|Home | About | Journals | Submit | Contact Us | Français|
Participants were 30 adult outpatients diagnosed with avoidant personality disorder or obsessive–compulsive personality disorder who enrolled in an open trial of cognitive therapy for personality disorders. Treatment consisted of up to 52 weekly sessions. Symptom evaluations were conducted at intake, at Sessions 17 and 34, and at the last session. Alliance variables were patients’ first alliance rating and “rupture-repair” episodes, which are disruptions in the therapeutic relationship that can provide corrective experiences and facilitate change. Stronger early alliances and rupture-repair episodes predicted more improvement in symptoms of personality disorder and depression. This work points to potentially important areas to target in treatment development for these personality disorders.
Empirical reviews (Horvath & Bedi, 2002; Orlinsky, Grawe, & Parks, 1994) and meta-analytic studies (Horvath & Symonds, 1991; Martin, Garske, & Davis, 2000) support a consistent relation between the quality of the therapeutic alliance and outcomes across a range of treatment orientations. The alliance has been studied primarily in the context of psychodynamic and existential therapies, and there is increasing evidence that it is also important in cognitive and cognitive–behavioral therapies (see Waddington, 2002, for a recent review). However, as Feeley, DeRubeis, and Gelfand (1999) emphasized, it is critical to examine whether the relations between alliance and outcome are due to prior symptom reduction.
The therapeutic relationship may be particularly relevant to the treatment of personality disorders (Beck, Davis, & Freeman, 2004; Benjamin, 1993; Young, Klosko, & Weishaar, 2003), but less empirical attention has been paid to the alliance in this population than in Axis I cohorts. Personality disorders are characterized by pervasive, debilitating interpersonal difficulties that make it difficult to establish and maintain a therapeutic alliance. Recent reviews of the few treatments available for personality disorders report dropout rates as high as 38% to 57%, with means that range from 15% to 22% (Leichsenring & Leibing, 2003; Perry, Banon, & Ianni, 1999). The strength of the alliance early in therapy is one factor that influences treatment engagement, retention, and outcomes. Ruptures in the alliance also occur and can be therapeutic or can be associated with early dropouts and worse outcomes, if not handled properly.
The present study examines the role of the alliance in a sample of 30 patients who participated in an open trial of cognitive therapy for personality disorders (CT-PD), for avoidant personality disorder (AVPD) and obsessive–compulsive personality disorder (OCPD; Beck, Freeman, and Associates, 1990). These two Cluster C personality disorders are highly comorbid with Axis I disorders (Mattia & Zimmerman, 2001) and often predict a poor response to treatment (Reich & Vasile, 1993). There is, however, some evidence that this may not be the case with cognitive therapy for depression in that those with and without comorbid personality disorders (primarily Cluster C) do not differ significantly in their treatment outcomes (for a review, see Mulder, 2002; Kuyken, Kurzer, DeRubeis, Beck, & Brown, 2001). Understanding how the alliance relates to treatment retention and to personality and depression symptom change can have important implications for treatment development.
A strong early alliance can influence outcomes by increasing treatment engagement, instilling hope, and providing a solid foundation for the course of therapy (Gaston, 1990; Horvath, Gaston, & Luborsky, 1993; Horvath & Luborsky, 1993; Whisman, 1993), especially in difficult-to-treat populations, such as those with personality disorders. Similarly, the strength of the alliance as early as Session 2 predicted the course of symptom change in those with chronic depression receiving cognitive–behavioral therapy (D. N. Klein et al., 2003). The early alliance remained a significant predictor of outcome, even after controlling for eight known predictors of outcome: prior and concurrent levels of depression, comorbid Axis I and II disorders, chronicity of depression, gender, social functioning, and childhood history of abuse and neglect. Early alliance (before Session 5) also strongly predicted symptom reduction in two other challenging populations, those receiving cognitive– behavioral therapies for childhood abuse-related posttraumatic stress disorder (PTSD; Cloitre, Chase Stovall-McClough, Miranda, & Chemtob, 2004) and bulimia nervosa (Constantino, Arnow, Blasey, & Agras, 2005).
The alliance can also vary across the course of therapy, and shifts in the alliance can influence outcomes (Gelso & Carter, 1994; Horvath, 1995; Safran, Muran, Samstag, & Stevens, 2002). One type of fluctuation is the alliance rupture, which is generally defined as difficulty maintaining the alliance or negative shifts in its quality. Ruptures can occur when patients’ core interpersonal schemata are activated in session or generalize to therapy from between-session experiences or when therapists do not attend to difficulties in the relationship (Muran, 2002). Working through and repairing ruptures can provide a potent opportunity to disconfirm maladaptive schemata and provide “corrective experiences” that can facilitate change. In contrast, ruptures that are not addressed adequately can increase avoidance and inhibit change.
Rupture-repair episodes can be quantified as a high-low-high (quadratic) pattern of alliance strength (Gelso & Carter, 1994). A quadratic alliance pattern was associated with better outcomes in a heterogeneous sample receiving short-term psychoanalytic counseling (Patton, Kivlighan, & Multon, 1997) and in a nonclinical sample of students seen by student counselors (Kivlighan & Shaughnessy, 2000). In the latter study, linear improvement and stable alliance patterns did not predict change. Stiles et al. (2004) compared different methods of measuring alliance fluctuations in cognitive– behavioral and psychodynamic–interpersonal therapy for depression. Only a V-shaped pattern of alliance scores predicted outcomes. This pattern consisted of a substantial decrease in alliance strength (rather than a minor fluctuation), followed by an improvement in alliance soon after. A more gradual, U-shaped function did not predict change, nor did measures of alliance slope or overall variability. Thus, rupture-repair episodes might be best captured by a steep, V-shaped pattern based on individual standard deviation scores.
Research to date suggests that the early alliance, as well as its disruption and repair, may herald positive change and that unresolved ruptures can be associated with poor outcomes. The first open trial of CT-PD for AVPD and OCPD provides a unique opportunity to study the alliance in these Axis II disorders, which are difficult to treat and associated with high rates of dropout. Early alliance ratings and rupture-repair episodes were examined as predictors of number of sessions completed and change in personality and depression symptoms.
All participants were recruited through newspaper advertisements and the general phone line of the University of Pennsylvania’s Department of Psychiatry between 1990 and 1995. Callers who described symptoms of chronic depression, generalized anxiety disorder, AVPD, or OCPD and were interested in participating in therapy studies at the clinic were invited for an extensive intake interview. All signed informed consent forms before the interview. Consent forms were approved by the department’s internal review board. Prior to enrollment, participants were administered the Structural Clinical Interview for DSM–III–R (SCID; Spitzer, Williams, Gibbons, & First, 1987) and the Structured Clinical Interview for DSM–III–R—Personality Disorders (SCID-II; Spitzer, Williams, Gibbons, & First, 1990). Those meeting criteria for a primary diagnosis of AVPD or OCPD without active suicidality, substance dependence in the past year, psychotic or bipolar disorder, schizotypal or borderline personality disorder, or organic dysfunction were invited to participate in this study.
Of the 40 patients enrolled (AVPD = 24; OCPD = 16), 2 attended therapy but did not complete pretreatment personality symptom assessments. Thirty completed at least one assessment of personality symptoms after intake and could be included in analyses of pretreatment to posttreatment change. Because depression symptoms were assessed at every session, depression data were available for all 40 patients. The mean number of sessions competed was 43.28 (SD = 9.89). Eight patients discontinued before the second assessment (Session 17) and were considered dropouts.1 This dropout rate of 21% (8/38) is similar to the mean rate of 22% reported in the Perry et al. (1999) review of therapies for personality disorders but somewhat higher than the rates in the Leichsenring and Leibing (2003) review (mean rates 15%–16%).
The mean age of the sample was 34.24 (SD = 9.32); 60% were single or divorced and 40% were married; 9% were ethnic minorities; and all but one had at least some college education. As many as 73% met criteria for a comorbid mood disorder,2 56% for a comorbid anxiety disorder, and 28% of those with a primary diagnosis of AVPD or OCPD also met criteria for the other personality disorder. The specific personality diagnosis of AVPD or OCPD (coded as 0 or 1) did not predict posttreatment outcome on the personality or depression measures of outcome, after controlling for pretreatment symptom scores. The two Cluster C groups were collapsed to increase statistical power. We also provide descriptive statistics by diagnostic group.
The alliance was assessed by patient self-report with the 24-item California Psychotherapy Alliance Scale (CALPAS; Marmar, Weiss, & Gaston, 1989). Patients’ alliance ratings have been shown to be better predictors of outcomes than therapists’ ratings (Horvath & Bedi, 2002; Horvath & Symonds, 1991; Martin et al., 2000). Total CALPAS scores (range = 24–168) were used to assess overall alliance strength. Therapists were blind to patients’ alliance ratings. Cronbach’s alpha coefficients across the eight assessments ranged from .77 to .93 (M =.87).
Personality disorder symptoms were assessed with the Wisconsin Personality Disorders Inventory (WISPI; M. J. Klein et al., 1993), a 240-item self-report measure of Axis II disorders (total score range = 240–2,400). The WISPI captures general patterns of interaction that correspond to Benjamin’s (1993, 1996) interpersonal theory of personality disorders, so items have remained largely constant across DSM–III and DSM–IV. Validation studies demonstrate excellent internal consistency and test–retest reliability (M. J. Klein et al., 1993) and good convergent and discriminant validity with the SCID-II, especially for diagnoses of AVPD and OCPD (Barber & Morse, 1994). Patients’ z scores for the AVPD or OCPD subscale were used in the present study.
Personality symptoms were also assessed by the SCID-II (Spitzer et al., 1990). Interviewers were postdoctoral psychologists with extensive training in structured interviewing, and they were blind to patients’ diagnosis and progress in therapy. Interviewers probed and rated the presence of each personality disorder symptom on a 3-point scale (0 = absent, 1 = sub-threshold, 2 = present). Unweighted kappa coefficients for interrater agreement for AVPD and OCPD diagnoses were .94 and .69, respectively, which fall in the good-to-excellent range (Landis & Koch, 1977). Personality disorder severity ratings were obtained by totaling the individual symptom scores for each disorder to yield dimensional scores corresponding to patients’ primary diagnosis (AVPD or OCPD).
The 21-item Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) was used to assess self-reported severity of depression symptoms over the past week (total score range = 0–63). The BDI has good construct validity and sufficient internal consistency and reliability (Beck & Steer, 1987; Beck, Steer, & Garbin, 1988).
The SCID, SCID-II, and WISPI were administered at intake, Session 17, Session 34, and termination. The CALPAS was administered up to eight times, at Sessions 2, 5, 10, 20, 30, 40, 50, and 52. The CALPAS was administered on the same schedule for all patients, but number of assessments varied as a function of the number of sessions attended. The mean number of alliance assessments was 5.76 (range = 3–8; mode = 6). The BDI was administered every week.
Therapists were 2 predoctoral and 12 doctoral-level clinicians. All were previously trained in cognitive therapy and received additional training in CT for personality disorders (Beck et al., 1990),3 which is similar to CT for Axis I disorders in its focus on dysfunctional schemata, cognitive–affective–behavioral connections, and teaching skills to modify schematic vulnerabilities. CT-PD places more emphasis on the historical roots of problems, interpersonal patterns, eliciting in-session affect, and the therapeutic alliance. Therapists received 1 hr of individual supervision for every 2 therapy hours, which included reviews of taped sessions. They attended weekly group supervision meetings and monthly case conferences. Treatment, billed on a sliding scale, included up to 52 weekly sessions that occurred within 12 to 16 months.
Outcome analyses were conducted for all patients with any personality and depression symptom data after intake. As seen in Table 1, CT-PD was associated with significant improvements in personality symptoms: For standardized WISPI, mean difference = 1.67, SD = 1.14, 95% confidence interval [CI] = 1.21–2.13, t(29) = 7.50, p < .001; for SCID-II, mean difference = 6.59, SD = 3.31, 95% CI = 5.27–7.91, t(29) = 10.16, p < .001; and for depression symptoms as measured with the BDI, mean difference = 8.60, SD = 8.45, 95% CI = 5.89–11.30; t(39) = 6.43, p < .001. To facilitate comparisons with recent reviews of therapies for personality disorders (Leichsenring & Leibing, 2003; Perry et al., 1999), we calculated within-group effect sizes as follows: (pretreatment score – posttreatment score)/pretreatment standard deviation. It is important to interpret these effect sizes with caution, as they are not based on a comparison with a control condition, and other studies have reported effect sizes calculated with the standard deviation of the pretreatment-to-posttreatment difference in the denominator.4
The effect sizes are larger than those reported by Perry et al. (1999) for active treatments (1.11 for self-report and 1.29 for interview-based measures) and control conditions (.25 for self-report and .50 for interview-based measures) and than those reported by Leichsenring and Leibing (2003) for cognitive–behavioral therapies (1.00) and psychodynamic therapies (1.46). It should be noted that the effect sizes in the previous reviews were based on measures of general functioning, whereas two of the outcome measures (SCID-II and WISPI) in this trial of CT-PD specifically assessed personality symptoms. In addition, all participants met SCID-II criteria for AVPD or OCPD at intake, but only 7% (2/30) met criteria at posttreatment. Although 73% (22/30) also met criteria for a comorbid mood disorder at intake, only 37% (11/30) met criteria at posttreatment. Using Jacobson and Truax’s (1991) methods for calculating clinically significant change, 57% (17/30) met criteria for significant change on the WISPI, 73% (22/30) on the SCID-II, and 60% (24/40) on the BDI (see Table 1).
Patients’ first CALPAS scores (Session 2 or 5) were used to measure early alliance. Scores were similar by diagnostic group: AVPD (n = 18), M = 139.07, SD = 16.63; OCPD (n = 12), M = 140.13, SD = 19.79; t(29) = −.16, p = .87. As seen in Table 2, higher early alliance scores were significantly associated with completing more sessions, whereas pretreatment symptom severity scores were not. To examine relations with early symptom change, early alliance scores were correlated with WISPI, SCID-II, and BDI scores at the second symptom assessment (Session 17), controlling for pretreatment scores. Early alliance scores were not significantly associated with early symptom change on the WISPI ( pr = .02, p = .93) or SCID-II ( pr= −.18, p = .45) but were significantly correlated with early BDI change ( pr = −.48, p = .007). Therefore, early BDI scores were controlled statistically in analyses of early alliance and posttreatment outcome.
As Stiles et al. (2004) noted, when assessing a rupture-repair episode, it is important to capture significant shifts in alliance ratings rather than minor fluctuations. We calculated the standard deviation of alliance scores across sessions for each patient. The mean of the individual standard deviation scores was 6.66 (SD = 3.35). Approximately half of the sample had individual standard deviation scores of less than 7 (range = 0.91–6.34), and half had standard deviation scores more than 7 (range = 7.39–12.94). Thus, 7 points seemed to represent a significant fluctuation and to differentiate patients. The mean number of alliance ratings for those with a rupture-repair episode (M = 5.57, SD = 1.91) and those without (M = 5.67, SD = 2.45) did not differ significantly, t(24) = .11, ns. The criteria for a rupture-repair episode were (a) at least three assessments of the alliance, (b) a decrease in alliance scores of ≤7 points, (c) a subsequent increase in alliance of at least 7 points, and (d) no subsequent decrease of ≤7 points that did not again increase ≤7 points. This method is intended to capture a high-low-high function of alliance and is consistent with previous methods of quantifying this pattern (Kivlighan & Shaughnessy, 2000; Patton et al., 1997; Stiles et al., 2004).
Of the 30 patients with pre- and posttreatment symptom data, 25 had at least three CALPAS ratings required to capture a high-low-high pattern of alliance. Using our classification criteria, rupture-repair episodes occurred for 56% (n = 14) of this sample of 25 and did not occur for 44% (n = 11). Of the 11 patients who did not meet our criteria for a rupture-repair episode, 27% (n = 3) experienced a rupture that was not repaired, 45% (n = 5) reported a linear decrease, and 27% (n = 3) reported a linear increase in the alliance. The rupture-repair pattern occurred for 8 of the 15 patients with AVPD (53%) and 6 of the 10 patients with OCPD (60%).
All of the patients who experienced a rupture-repair episode (n = 14) reported symptom reduction of 50% or greater on the SCID-II from the beginning to end of therapy. All but 1 of these patients (93%) reported the same amount of symptom reduction on the BDI, and 10 of the 14 patients (71%) reported this amount of improvement on the WISPI. In contrast, of the 11 patients who did not report a rupture-repair episode, only 5 (45%) reported a ≤50% symptom reduction on the SCID and BDI, and only 3 (27%) reported a ≤50% symptom reduction on the WISPI. Thus, rupture-repair episodes seemed to be associated with substantial symptom reduction.
The rupture-repair variable was coded as a 0 or 1 (occurred or did not, respectively), but this variable was analyzed in correlational analyses so that we could examine its associations with pretreatment symptoms severity, early symptom change, and the early alliance. As seen in Table 2, rupture-repair episodes were not significantly correlated with early alliance scores or number of sessions completed but were associated with lower WISPI scores at intake. To examine whether rupture-repair scores were associated with early symptom change (Session 17), we correlated rupture-repair scores and early symptom scores, controlling statistically for pretreatment symptom levels (WISPI, SCID-II, or BDI). Rupture-repair scores were not significantly associated with early symptom change on any measures (WISPI, pr = −.19; SCID-II, pr = −.21; BDI, pr = −.09; all ps = ns).
Early alliance and rupture-repair scores were examined in hierarchical regression analyses as predictors of posttreatment personality (WISPI, SCID-II) and depression (BDI) symptoms, controlling statistically for pretreatment severity and number of sessions completed. Pretreatment scores were entered in Step 1 of each equation, number of sessions in Step 2, and alliance predictors simultaneously in Step 3. As seen in Table 3, higher early alliance scores and rupture-repair episodes predicted improvement on personality symptoms (WISPI, SCID-II) and depression (BDI), even when the number of sessions completed was controlled statistically.
Because early alliance scores were associated with change in depression by the second assessment (Session 17), early alliance and rupture-repair scores were examined in hierarchical regression analyses as predictors of personality (WISPI, SCID-II) and depression (BDI) symptoms, controlling statistically for pretreatment severity and depression by the second assessment. Pretreatment personality or depression scores were entered in Step 1, early depression scores in Step 2, and alliance predictors simultaneously in Step 3. As seen in Table 4, early depression scores significantly predicted WISPI scores at posttreatment but did not predict outcome on other measures. Higher early alliance scores and rupture-repair episodes predicted improvement on all measures of outcome, even after controlling statistically for early change in depression.
In this investigation of the alliance in cognitive therapy for personality disorders, better early alliances and rupture-repair episodes contributed to change in self-reported (WISPI) and interview-rated (SCID-II) personality symptoms and depression symptoms (BDI), even when the number of sessions completed and early change in depression were statistically controlled. Although this was a small open trial (n = 30) and alliance rupture-repair episodes could be assessed in only 25 of these patients, these findings suggest that the alliance might function in more than one way over the course of CT-PD to influence improvement in personality and depression symptoms.
The significant relation between early alliance and treatment outcomes is consistent with Horvath and Symonds’s (1991) meta-analytic review and with Horvath and Luborsky’s (1993) assertion that establishing collaboration and trust early on is integral to the therapy process. Our findings are also consistent with research on the importance of the early alliance in other chronic problems, such as recurrent depression (D. N. Klein et al., 2003), childhood abuse-related PTSD (Cloitre et al., 2004), and bulimia nervosa (Constantino et al., 2005). We selected patients’ first alliance rating in an effort to assess the alliance before symptom change occurred, and we controlled statistically for early change in depression. Early alliance scores still predicted outcome. These findings differ from those of DeRubeis and Feeley (1990) and Feeley et al. (1999) in trials of CT for Axis I depression, where the alliance was no longer associated with treatment outcome when previous symptom reduction was controlled statistically. It is possible that the alliance is particularly important in Axis II populations, as hypothesized by Beck et al. (1990) and Beck, Freeman, and Associates (2003). Indeed, stronger early alliances were associated with the completion of more sessions and with more early improvement in depression (Session 17), as well as with more improvement in personality and depression symptoms at posttreatment. It is also important to note that those with higher pretreatment WISPI scores, which assess the interpersonal dysfunction associated with personality disorders, reported lower alliances and were less likely to experience a rupture-repair episode. Higher pretreatment WISPI scores might indicate that extra attention needs to be paid to the alliance to obtain maximal benefit from the therapeutic relationship.
Our findings also suggest that the alliance can worsen over the course of therapy and that, if handled properly, ruptures may be therapeutic (Horvath, 1995; Safran, 1993; Safran, Crocker, Mc-Main, & Murray, 1990; Safran & Muran, 1996, 2000). Most of those who reported rupture-repair episodes also reported pre- to posttreatment symptom reductions of 50% or greater on all measures. These findings are consistent with recent advances in CT that emphasize in-session transactions to reveal patients’ core interpersonal schemata (Alford & Beck, 1997; Newman, 1998; Robins & Hayes, 1993) and using the therapeutic relationship as a “corrective experience” to disconfirm maladaptive schemata (Beck et al., 1990, 2004; Safran, 1998; Safran & Segal, 1990; Young et al., 2003). In addition, our method of quantifying alliance ruptures, which is most similar to that of Stiles et al. (2004), provides another example of how the study of discontinuities in the course of therapy can reveal important change processes (Hayes, Laurenceau, Feldman, Strauss, & Cardaciotto, in press).
Both early alliance strength and the rupture-repair process appear to contribute to therapy outcomes, but a number of patients experienced an alliance rupture that was not repaired or a linear decrease in the quality of the alliance. These patterns should be studied as carefully as ruptures that are repaired, as they can perhaps reveal mistimed interventions, or instances where some therapists do not attend to ruptures or respond ineffectively, as Castonguay and colleagues (Castonguay, Goldfried, Wiser, Raue, & Hayes, 1996) noted in some sessions of CT for depression.
Although the results of this initial trial of CT for AVPD and OCPD are promising, there are important limitations to discuss. First, this is an open trial with no comparison condition, so within-group effects sizes were calculated. This method might yield overestimates of the effect sizes. Second, the sample size is small, but it is in line with the median sample sizes reported in recent reviews of therapies for personality disorders, which are typically of long duration and associated with significant rates of relapse (Perry et al., 1999: Mdn sample size = 25; Leichsenring and Leibing, 2003: for psychodynamic therapies, Mdn = 26; for cognitive–behavioral therapies, Mdn = 16). This study of the role of the alliance in change was meant to contribute to treatment development efforts, and the findings will need replication and further study. Third, personality symptoms were not assessed until Session 17 after the intake interview; therefore we can only draw conclusions about those who remained in therapy until at least that point. Fourth, the clinical utility and generalizability of our findings must be strengthened by similar studies with different clinical problems and treatment approaches. It may be that alliance rupture-repair episodes are not as useful in Axis I problems or with Cluster A and B personality disorders.
In addition, although these findings are consistent with alliance rupture-repair models of change, they are topographical in nature. We describe shifts in alliance but did not directly examine in-session transactions. We can only infer that ruptures were captured by our quantitative method, which was most similar to the methods of Stiles et al. (2004) and yielded similar findings. Another consideration is that the study included many therapists, but they treated too few patients to allow for analyses of therapist effects. However, therapist competence is an important variable, and in current work we are describing qualitatively and quantitatively the relations between alliance scores and ratings of therapist competence. We also note that the CALPAS (Marmar et al., 1989) is one measure of the alliance and that the Working Alliance Inventory (Horvath & Greenberg, 1994) might be more consistent with the format and goals of cognitive–behavioral therapies. The CALPAS, however, was an important predictor of symptom change, both as a measure of early alliance and of rupture-repair episodes.
The next phase of Adele M. Hayes’s treatment development research focuses on identifying therapist strategies associated with better and worse early alliances and rupture outcomes to improve treatment retention and treatment outcomes in this prevalent and challenging population. This work can complement the important work of Safran, Muran, and colleagues (reviewed in Muran, 2002; Safran & Muran, 2000) on markers of alliance ruptures and patient–therapist exchanges that inhibit and foster alliance repairs across types of therapy and clinical disorders. Together, such therapy process research can inform treatment development for the Cluster C personality disorders, which are among the most prevalent of the personality disorders in outpatient samples.
This study was supported in part by National Institute of Mental Health (NIMH) Grants P50-MH45178 and RO1-MH49902. Additional support was provided by NIMH Grant R21-MH62662 (awarded to Adele M. Hayes), Agency for Healthcare Research and Quality National Research Award Grant T32-HS00079 (awarded to Jennifer L. Strauss), and Department of Veteran Affairs Associate Investigator Award Grant AIA 04-025 (awarded to Jennifer L. Strauss). We thank Robert Gallup for his tireless data management support and Charles Carver, Ron Durán, Blanche Freund, Jeannie Beckham, and Hayden Bosworth for valuable feedback that enhanced the quality of this article. We also thank the therapists and patients for their participation.
1Reasons given for early discontinuation were financial constraints (n = 2), disagreement with/ambivalence about the cognitive model (n = 2), therapist referral to couple therapy (n = 1), and no reason given (n = 3).
2Of the 73% (22/30) who met criteria for a comorbid mood disorder, 27% (6/22) were diagnosed with major depression first episode, 36% (8/22) with major depression recurrent episodes, 5% (1/22) with depressive disorder not specified, and 32% (7/22) with dysthymia (which could co-occur with other categories).
3Three graduate students in clinical psychology used the Revised Cognitive Therapy Scale (CTS-R; Blackburn et al., 2001) to rate therapist competence. The CTS-R is a competency measure that extends the original CTS (Vallis, Shaw, & Dobson, 1986; Young & Beck, 1980) by including additional items to assess focus on emotion and the therapeutic relationship. These extensions are consistent with recent developments in CT and are especially relevant to treating personality disorders. The CTS-R consists of 12 items rated on a scale ranging from 0 to 6: agenda setting, feedback, collaboration, pacing and efficient use of time, interpersonal effectiveness, eliciting appropriate emotional expression, eliciting key cognitions, eliciting behaviors, guided discovery, conceptual integration, application of change methods, and homework setting. Competence is achieved at 36, equivalent to a rating of 3 on each item. The authors of the scale state that it can be used by expert therapists or by trained clinician raters who are not experts in cognitive therapy (Ian James, personal communication, August 2, 2005).
One session was randomly selected for coding before the second assessment and one from the sessions after that. Each session was rated independently by two raters, and raters were paired with each other an equal number of times. Raters were blind to diagnosis and symptom ratings. Interrater agreement was good on all coding categories (intraclass correlations =.82–.88). The mean competence ratings across therapists in this study (M = 38.30, SD = 6.36) were comparable with those in the Blackburn et al. (2001) study of therapists participating in a CT certification workshop. It is encouraging that competence ratings were adequate in this new application of cognitive therapy to personality disorders, as it can be more difficult to develop and maintain an alliance with this population, and the focus is more broad-based than in CT for Axis I disorders. Because all therapists were trained to criterion levels of competence, there was little overall variation, and mean competence scores were not associated with alliance scores or with measures of symptom severity at pretreatment or posttreatment.
4Although we divided the pretreatment–posttreatment difference by the standard deviation of the pretreatment score to allow for comparisons with previous studies on personality disorders, effect sizes reported from other open trials have used the following formula: (pretreatment score – post-treatment score)/standard deviation of the difference. Therefore, we present those effect sizes as well (WISPI standardized: d = 1.47; SCID-II: d = 1.98; BDI: d = 1.02).
Jennifer L. Strauss, Duke University Medical Center and Durham VA Medical Center, Durham, North Carolina.
Adele M. Hayes, Department of Psychology, University of Delaware.
Sheri L. Johnson, Department of Psychology, University of Miami.
Cory F. Newman, Department of Psychiatry, University of Pennsylvania School of Medicine.
Gregory K. Brown, Department of Psychiatry, University of Pennsylvania School of Medicine.
Jacques P. Barber, Department of Psychiatry, University of Pennsylvania School of Medicine.
Jean-Philippe Laurenceau, Department of Psychology, University of Delaware.
Aaron T. Beck, Beck Institute of Cognitive Therapy and Research, Bala Cynwyd, Pennsylvania, and Department of Psychiatry, University of Pennsylvania School of Medicine.