News

This study compared the effects of a higher dose of cognitive behavioral therapy (CBT) for panic disorder versus CBT for panic disorder combined with “straying” to CBT for comorbid disorders in individuals with a principal diagnosis of panic disorder with or without agoraphobia. Sixty-five participants were randomly assigned to one of two treatment conditions, either CBT focused solely upon panic disorder and agoraphobia or CBT that simultaneously addressed panic disorder and agoraphobia and, to a lesser degree, the most severe comorbid condition. Results indicated a significant reduction in panic disorder severity and a decline in severity of comorbid diagnoses across both treatment conditions. However, individuals receiving CBT focused only on panic disorder were more likely to meet high end-state functioning at post-treatment, even in intent-to-treat analyses, and report zero panic attacks at the I-year follow-up, although this effect was not retained in intent-to-treat analyses. At follow-up, CBT focused only on panic disorder yielded more substantial improvement in the most severe baseline comorbid condition, although not in intent-to-treat analyses, and a greater proportion of individuals in this treatment condition were rated as having no comorbid diagnoses, even in intent­to-treat analyses. These findings raise the possibility that remaining focused on CBT for panic disorder may be more beneficial for both principal and comorbid diagnoses than combining CBT for panic disorder with ‘straying’ to CBT for comorbid disorders.

Michelle G. Craske^a,*, Todd J. Farchione^b, Laura B. Allen^b, Velma Barrios^c, Milena Stoyanova^a, Raphael Rose^a

a Department of Psychology, UCLA., 1285 Franz Hall, Box 951563, Los Angeles, CA 90095-1563, USA

b Center for Anxiety and Related Disorders, Boston University, 648 Beacon Street, Sixth floor, Kenmore Square, Boston, MA 02215, USA CDepartment of Psychology, SUNY, Albany, Park Hall, Room 206, Buffalo, NY 14260-4110, USA

Received 11 July 2006; received in revised form 17 August 2006; accepted 7 September 2006

Published by Elsevier Ltd.

Keywords: Comorbidity; Cognitive behavioral therapy; Panic disorder; Agoraphobia

The efficacy of cognitive behavioral treatments has been well established for most if not all of the major anxiety disorders (e.g., Chambless, 1996; Otto, Smits, & Reese, 2004). In addition, there is now evidence indicating that cognitive behavioral therapy (CBT) for a targeted anxiety disorder yields positive benefits upon comorbid disorders, most often demonstrated with CBT for panic disorder (e.g., Brown & Barlow, 1995; Tsao, Lewin, & Craske, 1998; Tsao, Mystkowski, Zucker, & Craske, 2002) but also with CBT for generalized anxiety disorder (Borkovec, Abel, & Newman, 1995; Stanley, Beck, & Glassco, 1996). This finding is especially important since 50-80% of those with an anxiety disorder meet diagnostic criteria for at least one other diagnosis, most commonly other anxiety disorders, mood disorders, substance use related disorders, and somatoform disorders (e.g., Brown & Barlow, 1992; Brown, Chorpita, & Barlow, 1998; Kessler et aI., 1996).

Furthermore, reductions in comorbidity are unlikely to be due to the passage of time alone, since a panic disorder group waitlisted for ten weeks (albeit small in size, n = 11) showed no change in rates of comorbidity (Tsao, Mystkowski, Zucker, & Craske, 2005; Craske, Lang, Aikins, & Mystkowski, 2005).1

These outcomes derive from CBT comprised of core therapeutic strategies (such as cognitive restructuring and exposure therapy) that are tailored specifically to the features unique to the principal anxiety disorder (i.e., panic disorder or generalized anxiety disorder). Potential mechanisms for the positive effects upon comorbidity include reductions in disorders that are functionally secondary to the principal disorder, application of core CBT therapeutic strategies to comorbid symptoms, or down regulation of emotional processes shared across emotional disorders, such as negative affectivity and perceived lack of emotional control (Tsao et aI., 2005).

This type of single focus CBT, tailored to the features unique to the principal anxiety disorder, is characteristic of all CBT efficacy studies since the advent of specific treatment manuals for specific emotional disorders. However, there has been no direct investigation of the importance of a ‘single focus’. There is only indirect evidence which suggests that a single focus is more effective than multiple foci. Specifically, standardized in vivo exposure therapy was found to be superior to individualized treatments planned by the therapist for phobic participants (Schulte, Kunzel, Pepping, & Schulte-Bahrenberg, 1992); in the former condition, therapists were assumed to remain more focused on the targeted phobic disorder. In addition, Frank et ai. (1991) evaluated monthly maintenance interpersonal therapy sessions for prevention of recurrent depression, and found that therapy sessions rated as more specific to interpersonal issues were associated with significantly less relapse; conversely, patients of therapists who “strayed” from interpersonal issues had more relapse. Furthermore, extensive reviews of the child literature have identified core differences between community based interventions and research based interventions, including a broad, multi-problem focus and flexibility in responding to issues that the client brings to each session (in an effort to treat the “whole” client) in the former case compared to a clearly delineated narrow focus, and imposing of an agenda and a pre­planned structured manual in the latter case (Weisz, Donenberg, Han, & Weiss, 1995; Weisz, Weiss, & Donenberg, 1992; Weisz, Jensen, & McLeod, 2005). Furthermore, Weisz et ai. (1995) suggested that this, along with core differences, may account for the substantially lower rates of effectiveness in community based settings compared to research studies, at least for childhood disorders.²

We aimed to evaluate whether retaining a single focused of CBT tailored specifically to panic disorder was more effective than introducing an additional secondary focus upon a comorbid disorder. That is, in this preliminary investigation, we provided CBT for the principal anxiety disorder of panic disorder/agoraphobia to all participants and compared the effects of additional CBT focused solely upon panic disorder/ agoraphobia (i.e., ‘more of the same’) to additional CBT that targeted a comorbid condition (i.e., ‘less of more’, or “straying” from, or deviating from, the main focus). Based on the earlier findings, we hypothesized that diversified CBT would have less positive influences than more of the same CBT. Thus, we hypothesized larger and longer lasting reductions in the principal diagnosis of panic disorder and agoraphobia in the group receiving ‘more of the same’ CBT. In addition, we hypothesized that, by virtue of greater improvement in the targeted disorder, the single focused CBT condition also would lead to larger and longer lasting reductions in comorbidity. This hypothesis was supported by the evidence for greater reductions in comorbidity amongst individuals who responded most positively to CBT for generalized anxiety disorder (Borkovec et al., 1995). Furthermore, it was assumed that reductions in the targeted anxiety disorder would be associated with

IThe report by Woody, Taylor, McLean and Koch (1999) that improvements in comorbid depression after CBT for panic disorder were comparable to a minimal intervention group was confounded since the group given CBT presented with panic disorder was comorbid with depression whereas the minimal intervention group presented with depression without panic disorder.

2Although Weisz et al. (1995) did not find a separate effect when subsequently comparing studies with a clear problem focus vs a vague or diffuse focus, the categorization of ‘a clear problem focus’ broad terms such as phobias-anxiety, which is broader than the focus of most research based intervention manuals for anxiety disorders.

increases in emotional control and decreases in neuroticism, and that these variables would predict decreases in comorbidity by follow-up.

Method

Participants

Potential participants were recruited through advertisements and referrals. Sixty-five individuals, averaging 36.8 years of age (SD = 9.1) and 15.3 years (SD = 2.5) of schooling participated. Sixty percent were female, 84% were Caucasian, 43% were married or cohabiting, and 74% were employed. All met DSM-IV (American Psychiatric Association, 1994) criteria for a principal diagnosis of panic disorder with or without agoraphobia, and at least one comorbid anxiety or mood disorder. Exclusion criteria included a history of bipolar disorder, psychosis, current substance abuse/dependence,3 suicidality (active attempts in past two years, and current plans or intent), and neurological, cardiovascular, renal, and thyroid diseases. Diagnoses were ascertained from the Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV; Brown, DiNardo, & Barlow, 1994), conducted by clinical psychology doctoral students and postgraduate fellows trained to meet reliability standards. Final diagnoses were reviewed in case conference with the principal author (MGC). Participants who used psychotropic medication either withdrew from medications⁴ or were stabilized (58%) (1 month for l3-blockers and benzodiazepines and 3 months for all other psychotropic medications) prior to evaluation of diagnostic eligibility. Those who were participating in other psychotherapies were required to have been stable in those other psychotherapies for at least 6 months. Participants who received CBT in the past were considered eligible only if the CBT was received at least 2 years prior to baseline assessment. Treatment fees were either free or on a sliding scale.

Design

Participants were randomized to one of two treatment conditions, PDA (group and individual treatment sessions targeting panic disorder/agoraphobia) or PDA + C (group sessions targeting panic disorder/ agoraphobia, and individual sessions targeted the most severe comorbid disorder), and were assessed at baseline, post-treatment and 6 and 12 months later. Due to several missing data points at the 12-month assessment (n = 7), analyses were conducted comparing baseline, post-treatment, and the final data point available (i.e., 6–12 months).

Dependent variables

Panic disorder psychopathology

(1) The ADIS-IV diagnostic interview was repeated at each assessment by interviewers who were blind to the participant’s treatment assignment. Variables extracted from the interview for purposes of outcome analyses included a clinical rating of severity of distress and disablement (CSR; 0 = not at all, 8 = extreme) for the diagnosis of panic disorder with or without agoraphobia. Diagnoses receiving ratings of “4” or higher meet diagnostic criteria and are considered clinically significant. Adequate reliabilities have been demonstrated for diagnoses and CSRs from the ADIS-IV interview (Brown, DiNardo, Lehman, & Campbell, 2001). Inter­rater reliability from our clinic (including n = 16 from the current sample) was established from independent ratings by a primary interviewer and a second rater who listened to audio-taped interviews (n = 38). A K coefficient of 0.93 indicated excellent inter-rater agreement regarding the diagnosis of panic disorder (n = 27); CSR ratings for panic disorder diagnoses were also reliable between raters (pearson r = 0.86). Other variables extracted from the ADIS- IV included categorization oflevel of agoraphobia as either none/mild or moderate/ severe and number of panic attacks over the last 2 weeks.

3S_ubstance use disorders were a reason for exclusion given the lack of evidence that CBT is effective for panic disorder when comorbid with such disorders.

4Diagnostic status was re-evaluated after one month of withdrawal and washout from benzodiazepines and three months for anti­depressants.

(2) Standardized self-report scales included the 16-item Anxiety Sensitivity Index (ASI; Reiss, Peterson, Gursky, & McNally, 1986), which measures discomfort with anxiety-related symptoms. Extensive research has shown the ASI to have high internal consistency, test-retest reliability and good validity (e.g., McNally, 1989; Peterson & Reiss, 1992; Rapee & Medoro, 1994). Internal consistency was high in the current sample (ex = 0.86). The Agoraphobia subscale of the Fear Questionnaire (FQ-A; Marks & Mathews, 1979) was included as a measure of agoraphobia avoidance; the scale possesses excellent internal consistency and test-retest reliability, and good validity (e.g., Cox, Swinson, Parker, Kuch, & Reichman, 1993; Michelson & Mavissakalian, 1983). Again, internal consistency was high in the current sample (ex = 0.83).

(3) Participants completed two separate behavioral approach tasks. The first was comprised of three interoceptive tasks in counterbalanced order; hyperventilation for up to 90 s (i.e., breathing rapidly and deeply), spinning for up to 2 min (i.e., standing and turning once every 3 s), and restricted breathing through a small tube for up to 2 min. In each case, participants were first informed about likely sensations such as dizziness and shortness of breath. In the second behavioral approach task, participants were asked to sit in a small enclosed space (2′ x 3′ x 6′) that could not be opened from the inside for up to 10 min in the dark (i.e., booth task), while the experimenter waited approximately 20ft away. This booth task has been used in prior studies (e.g., Craske, deCola, Sachs, & Pontillo, 2003), and was moderately anxiety provoking for the current sample, without differences between conditions (see Table 3). Ratings of anticipatory (0-8) and maximum (0-8) anxiety and duration were recorded for each task.

Comorbidity psychopathology

(1) Comorbid diagnoses and accompanying CSRs were obtained from the ADIS-IV interview, conducted at each assessment by interviewers blind to the participant’s treatment assignment. K coefficients from our reliability sample indicated good to excellent inter-rater agreement for the diagnoses with sufficient n’s in our reliability sample: major depression = 1.0 (n = 7); generalized anxiety disorder = 0.70 (n = 13); social phobia = 0.66 (n = 6); and specific phobia = 0.69 (n = 14). Pearson correlation coefficients for CSR ratings were as follows: major depression, r = 0.90, generalized anxiety disorder, r = 0.64, social phobia, r = 0.90, and specific phobia, r = 0.76.

At each assessment point, comorbid diagnoses were categorized as clinically severe (CSRs of 4 or higher) or subclinical (CSRs of less than 4). In this sample, all participants were required to have at least one comorbid diagnosis at baseline with a CSR of 3 or higher; CSRs of 3 were assigned either because one of the secondary diagnostic criteria was not met (e.g., 2 vs. 3 of the 6 physical symptoms of the GAD diagnosis) or because all diagnostic criteria were met but the level of distress and disablement was judged to be just below clinical severity. A CSR level of 3 or higher was selected to increase sample size. Also, it was assumed that a level of 3 or higher would be relevant to real world settings where clinicians may be inclined to provide CBT to multiple emotional problems including those that are distressing and disabling although just below clinical severity or diagnostic criteria levels. Furthermore, it was presumed that patients with comorbid disorders just below clinical threshold are symptomatic enough to benefit from an intervention. The number of clinically severe comorbid diagnoses was summed at each assessment point, as was the total number (clinical and subclinical) of comorbid diagnoses.⁵ In addition, the CSR for the comorbid diagnosis rated as most severe at baseline was evaluated across each assessment point.

(2) The brief symptom inventory, total score (BSI; Derogatis, 1993) was used as a standardized self-report measure of anxious and depressed mood symptoms; the inventory possesses adequate reliability and validity (Derogatis, 1993). Internal consistency in the current sample was high (ex = 0.96). In addition, participants completed the subjective symptoms scale (SSS; a modification of a scale introduced by Hafner & Marks, 1976) to provide a measure of interference with daily functioning in six different domains, each rated on a 0-8 point scale (e.g., work, home life, leisure activities). Again, internal consistency in the current sample was high (ex = 0.80).

5Specific phobias of different subtypes (i.e., environmental, animal, situational, blood/injury/injection, and other) were summed as individual diagnoses, if CSRs were 3 or higher. The number with more than one specific phobia did not differ between conditions (PDA; 2 phobias, n = 5; 3 phobias, n = 1: PDA + C; 2 phobias, n = 6; 3 phobias, n = 1).

(3) Standardized self-report measures of vulnerability to emotional disorders included seven items selected from the 12-item Neuroticism scale of the NEO Personality Inventory (NEO; Costa & McCrae, 1985); these seven items have been shown to be internally consistent and predictive of persistence of depression after controlling for initial depression severity (Katon et aI., 1994). Internal consistency in the current sample was high (a. = 0.84). Also, the anxiety control questionnaire (ACQ; Rapee, Craske, Brown, & Barlow, 1996) was included to measure perceived control over internal and external events. The ACQ has good internal and retest reliability and validity (Rapee et aI., 1996). Internal consistency in the current sample was high (a. = 0.88). These measures were included to evaluate the degree to which CBT influenced emotional vulnerability, and the degree to which measures of emotional vulnerability influence reductions in comorbidity.

Treatment

Following from our prior research, we provided CBT for panic disorder/agoraphobia in small group sessions to all participants, and then evaluated the relative efficacy of additional individual sessions of CBT tailored to panic disorder (as an extension and refinement of the groups sessions) to CBT tailored to comorbid conditions.

Group cognitive behavioral therapy for panic disorder/agoraphobia

All participants received 12 weekly, group CBT sessions, each lasting 1¹/²_2 h, that targeted symptoms related to panic disorder/agoraphobia. Groups were comprised of 3 to 6 patients and 2 therapists. CBT was based on Panic Control Therapy (Barlow & Craske, 1988), which has been shown to be highly effective for panic disorder and agoraphobia after 12 treatment sessions (Mitte, 2005). Treatment began with educational information and cognitive restructuring to address misappraisals of panic attacks and associated bodily sensations. Breathing retraining was taught as a coping technique especially for symptoms associated with overbreathing. During interoceptive exposure, participants repeatedly induced physical sensations associated with panic attacks (e.g., spinning, hyperventilating) to weaken conditioned emotional reactions and obtain corrective information designed to disconfirm fearful misappraisals. Interoceptive exercises were extended to naturalistic activities that induced feared bodily sensations (e.g., sports, caffeine) and were incorporated into structured in vivo exposure to agoraphobic situations. Group CBT for panic disorder/agoraphobia was manuaIized, but therapists were encouraged to adapt the general principles to each participant’s presenting problems.

Individual cognitive behavioral therapy

Once every 2 weeks, participants received an additional hour long, individual CBT session with a therapist from the group. For participants assigned to PDA, the content of the individual sessions reiterated material presented in the group treatment sessions. For participants assigned to PDA + C, the content of the individual sessions was tailored to the most severe comorbid anxiety or mood disorder, as defined by baseline CSR values. Therapists followed manualized guidelines for each comorbid anxiety or mood disorder (e.g., obsessive compulsive disorder~response prevention; generalized anxiety disorder–eatastrophic imagery exposure; specific phobia of blood, injury, injections~applied tension; posttraumatic stress disorder~trauma memory exposure; social phobia~in vivo exposure to social situations; depression~behavioral activation and problem solving),6 but again were encouraged to adapt the general principles to each participant’s presenting problems.

In order to mimic the notion of ‘straying’ from the main topic, as was found to be detrimental by Frank et aI. (1991), we chose biweekly (i.e., six) instead of weekly individual sessions. In other words, we modeled a treatment that was primarily focused on a principal disorder with ‘less of more’ CBT focused on a comorbid disorder.

Therapists and treatment integrity

Therapists were senior clinical psychology doctoral students and postdoctoral fellows trained by the principal author through a one-year supervision program for CBT for anxiety disorders. Each group

6A complete list of the cognitive behavioral strategies specific to each comorbid anxiety disorder and major depression/dysthymia is available from the principal author.

treatment session was audio-taped and 11 % (n = 27) were selected randomly for independent adherence ratings of each content item of each session (1 = none, 7 = complete adherence).7 Also, clinic case notes (n = 34, 17 per treatment condition)8 of individual treatment sessions were reviewed by an independent rater blind to participants’ treatment condition assignment, and case note descriptions of CBT strategies relevant to panic disorder/agoraphobia and to comorbid conditions were coded using a standardized checklist of strategies. In addition, participants completed a treatment credibility questionnaire (Borkovec & Nau, 1973) after the completion of their second group and first individual visit.

Results

Attrition

Of 65 participants, 33 were randomized to PDA and 32 to PDA +c. Ten participants (15%) withdrew during treatment; 4 (12.5%) from PDA +C and 6 (18.2%) from PDA, ns. Reasons for withdrawal are not known, but there were no differences between the 10 withdrawals and remaining participants on any demographic, diagnostic, or psychological measure.

CBT integrity

Average adherence ratings (1-7) across the group CBT sessions ranged from 6.2 to 6.9, with a total average of 6.5 (SD = 0.3), indicating excellent adherence overall. Also, there were no differences in group CBT adherence ratings between PDA + C (M = 6.53, SD = 0.33) and PDA (M = 6.52, SD = 0.45). As expected, the average number of CBT strategies per session targeting panic disorder in the individual treatment case notes was significantly higher in PDA (M = 2.14, SD = 0.97) than PDA + C (M = 0.31, SD = 0.98), t(32) = 5.44, p < 0.001. Consistent with our design, the average number of CBT strategies per session targeting comorbid disorders in the individual treatment case notes was significantly higher in PDA + C (M = 1.80, SD = 1.23) than PDA (M = 0.20, SD = 0.59), t(32) = 4.8, pThe two treatment conditions did not differ in averaged credibility ratings (0-8 point scale): PDA + C M = 6.1 (SD = 1.5), PDA M = 6.2 (SD = 1.0), t(33) = -0.3 ns.

Baseline

PDA + C and PDA groups did not differ on any demographic, diagnostic, self-report or medication variables (see Tables 1 and 2). The average CSR rating for panic disorder/agoraphobia was 5.5 (SD = 1.0) and 51 % of the entire completer sample was diagnosed with moderate to severe levels of agoraphobia. Overall, participants averaged 4.2 (SD = 7.1) panic attacks in the last 2 weeks, and 5.4 years (SD = 6.7) since panic attacks became problematic.

Rates of comorbid diagnoses with CSR values of 4 or higher for the entire sample were as follows (rates for diagnoses with CSR values of 3 or higher are in parentheses): generalized anxiety disorder, 38.5% (49.2%); specific phobia, 43.1 % (47.7%); major depression, 32.3% (33.8%); social phobia, 20% (33.8%); dysthymia, 9.2% (16.9%); hypochondriasis, 9.2% (13.8%); obsessive compulsive disorder, 6.2% (13.8%); posttraumatic stress disorder, 4.6% (7.7%); and somatization, 1.5% (1.5%). Rates did not differ between the treatment conditions for any comorbid diagnosis (see Table 2). Overall, 86.2% met criteria for at least one anxiety or mood comorbid disorder with a CSR of 4 or higher, as did 100% with a CSR of 3 or higher. Four (12.5%) PDA + C participants and five (15.2%) PDA participants had baseline comorbid diagnoses with CSR values of 3, ns. The average number of comorbid diagnoses with CSRs of 4 or higher was 1.8 (SD = 1.6) and the average total number of comorbid diagnoses, including those with CSR = 3, was 2.4 (SD = 1.5); these mean values did not differ between treatment conditions. The average CSR of the most severe comorbid diagnosis at

7Each of the 12 treatment sessions was equally likely to be rated for adherence. The number of content items ranged from 6-10 per session.

8Not all case notes were sufficiently detailed to be independently rated for CBT content.

Table 2

Comorbidity at baseline in PDA + C and PDA conditions

% Generalized anxiety disorder, CSR 4 + CSR 3+

% Specific phobia, CSR 4+ CSR 3+

% Major depression, CSR 4+ CSR 3+

% Social phobia, CSR 4+ CSR 3+

% Dysthymia, CSR 4 + CSR 3+

% Hypochondriasis, CSR 4 + CSR 3+

% Obsessive-compulsive disorder, CSR 4+ CSR 3+

% Post-traumatic stress disorder, CSR 4+ CSR 3+

% Somatization disorder, CSR 4 + CSR 3+

% With I + diagnosis, CSR 4+ % With I + diagnosis, CSR 3 + # Comorbid diagnosis, CSR 4 + # Comorbid diagnosis, CSR 3 +

CSR most severe baseline comorbid diagnosis

baseline was 4.7 (SD = 1.2) and again there were no differences between conditions. Which particular comorbid diagnosis was most severe at baseline did not differ between the groups; generalized anxiety disorder (PD+C = 11, PDA = 12), specific phobia (pDA +C = 9, PDA = 9), depression/dysthymia (PDA +C = 10, PDA = 10), social phobia (PDA +C = 4, PDA = 6), hypochondriasis (PDA +C = 3, PDA = 3), obsessive compulsive disorder (PDA+C = 2, PDA = 1) and posttraumatic stress disorder (pDA+C = 1, PDA = 1).9

9There were 8 ties in PDA + C and 9 ties in PDA for the most severe comorbid diagnosis.

Effect of treatment upon panic disorder/agoraphobia

The overall statistical approach involved repeated measures ANOVAs, followed by linear and quadratic within subject contrasts to evaluate patterns of change from baseline, to post-treatment and follow-up between the two treatment conditions. Categorical data were analyzed using Fisher exact tests. Bonferroni corrections were not applied due to the risk of Type II error in small sample sizes (e.g., Legendre & Legendre, 1998).

Three participants completed treatment but failed to complete post assessment (1, PDA +C; 2, PDA) and seven failed to complete follow-up assessments (5, PDA + C; 2, PDA). Hence, final sample sizes were 22 for PDA + C and 23 for PDA. Additional participants failed to complete the self-report battery across all three assessments; final sample sizes were 17 for PDA + C and18 for PDA for questionnaire data. Means and standard deviations are shown in Table 3. Those who completed treatment but failed to complete either post

Table 3

Effect of treatment upon panic disorder in PDA + C and PDA conditions

PDA +C (n = 13-22)	PDA (n = 13-23)
CSR
Baseline	5.4 (0.9)	5.3 (0.8)
Post	3.5 (1.4)	2.5 (1.8)
Follow-up	2.2 (1.7)	1.8 (1.7)
% Moderate-severe agoraphobia
Baseline	63.6	47.8
Post	27.3	26.1
Follow-up	13.6	8.7
% Zero panic attacks past two weeks
Baseline	27.3	34.8
Post	50.0	69.6
Follow-up	68.2	91.3
ASI scores
Baseline	31.6 (10.4)	31.6 (10.7)
Post	21.8 (15.2)	17.3 (6.8)
Follow-up	17.4 (14.8)	16.2 (10.6)
Fear Questionnaire-agoraphobia subscale
Baseline	15.2 (11.1)	13.8 (8.6)
Post	9.0 (9.5)	7.4 (7.8)
Follow-up	6.4 (8.5)	5.0 (8.2)
Interoceptive tasks-Maximwn anxiety
Baseline	3.4 (1.6)	3.4 (1.9)
Post	2.4 (1.7)	1.7 (1.3)
Follow-up	1.7 (1.5)	1.6 (1.5)
Booth task-Anticipatory anxiety
Baseline	4.5 (1.8)	3.4 (1.9)
Post	2.7 (2.3)	1.7 (2.2)
Follow-up	1.3 (1.7)	1.3 (1.8)
Booth task-Maximwn anxiety
Baseline	4.7 (2.8)	4.1 (1.8)
Post	3.0 (2.4)	2.2 (2.4)
Follow-up	2.0 (2.0)	1.5 (2.0)
Booth task-% completing entire task
Baseline	86.2	80.0
Post	90.9	86.4
Follow-up	94.4	94.4

or follow-up assessments did not differ from those who completed treatment and all three assessments on any diagnostic or questionnaire measure.

CSRs associated with panic disorder declined significantly over time, F(2,86) = 109.5, pES = 0.72, although the effects for Condition and Condition x time were non-significant (ES = 0.05). The percentage of participants rated as moderately to severely agoraphobic also decreased from 55.6% to 26.7% at post and 11.1 % at follow-up, and there were no differences between treatment conditions. The percentage of those with zero panic attacks in the last two weeks increased over time (see Table 3), with a trend for proportionately more ofPDA reporting zero panic attacks at follow-up, X²(1) = 3.8, p = 0.05. High end-state functioning was operationalized as panic disorder/agoraphobia CSR of 0-2, none to mild agoraphobia, and zero panic attacks. Proportionately more of PDA (47.8%) reached high end-state at post-treatment than PDA +C (18.2%), X²(1) = 4.5, p < 0.05, although proportions did not differ at follow-up, at which time 61 % of the PDA and 55% of PDA + C reached high end-state status. The same pattern of results was obtained when analyses were restricted to participants with comorbid diagnoses associated with CSR values of 4 or higher at baseline (PDA, n = 18; PDA +C, n = 18). Scores on the ASI and FQ-A each declined significantly over time, F(2,68) = 34.9,p<0.001, ES = 0.51, and F(2,68) = 50.66, pES = 0.60, but the effects for condition (ES = 0.00 and 0.01) and Condi­tion x Time (ES = 0.00) were non-significant. Means and standard deviations are shown in Table 3. Anticipatory and maximum anxiety ratings averaged across the three interoceptive tasks correlated strongly (coefficients ranged from r = 0.75 to 0.87, and averaged r = 0.81). Hence, only the maximum anxiety ratings were analyzed. In contrast, anticipatory and maximum anxiety ratings were less strongly correlated for the booth task (coefficients ranged from 0.51 to 0.84, and averaged 0.65). Hence, booth task anticipatory and maximum anxiety ratings were analyzed separately. Behavioral approach test data were obtained from 13 PDA + C and 17 PDA participants across all three assessments. The effect of Time was significant in each case; interoceptive tasks maximum anxiety, F(2,56) = 28.39, pES = 0.50; booth task anticipatory anxiety, F(2,56) = 24.10, pES = 0.46, and booth task maximum anxiety, F(2,56) = 31.97, pES = 0.53). Means and standard deviations are shown in Table 3. However, effects were not significant for Condition (ES = 0.01—0.05) or Condition x Time (ES = 0.01—0.06). Almost all participants completed each interoceptive task at baseline and therefore duration was not analyzed. However, the proportion who endured the full ten minutes in the booth increased over time, from 85.3% at baseline to 94.4% at follow-up, without differences between conditions. Intent-to-treat: Intent-to-treat analyses were conducted by carrying forward baseline diagnostic data for participants who dropped during treatment or who completed treatment but failed to complete post assessments (n = 13), and carrying forward post-treatment data for participants who failed to complete follow-up assessments (n = 7), yielding sample sizes of PDA + C = 32 and PDA = 33. The CSR values declined over time, F(2, 128) = 76.4, pES = 0.54. There was no effect of condition (ES = 0.02) or Condition x Time (ES = 0.02), although the quadratic effect of the interaction was significant, albeit yielding a small effect size, F(l ,64) = 4.3, p < 0.05, ES = 0.06, due to a trend towards greater reductions in CSR values by post-treatment in the PDA condition, t(64) = 1.8, p<0.08. The percentage who remained moderately to severely agoraphobic in the entire intent-to-treat sample was 29.2% at post and 18.5% at follow-up. Corresponding numbers for zero panic attacks were 60% at post and 73.8% at follow-up; there were no differences between conditions. However, significantly more PDA (36.4%) than PDA + C (12.5%) participants met criteria for high end-state at post-treatment, X²(1) = 3.9, p<0.05, although no differences were apparent at follow-up, at which time 44.6% of the entire sample met criteria for high end-state functioning. Again, the pattern of results did not differ when analyses were restricted to participants with baseline comorbid diagnoses associated with CSR values of 4 or higher.

Effect of treatment upon comorbid conditions

The number of comorbid conditions with CSRs of 4 or higher declined over time, F(2,86) = 72.0, pES = 0.63, but the effects of condition (ES = 0.03) and Condition x Time (ES = 0.01) were not significant. Means and standard deviations are shown in Table 4. The total number of comorbid conditions (including diagnoses with CSRs = 1-3) also declined over time, F(2,86) = 31.6,p<-0.001, ES = 0.42, but again with no

Table 4

Effect of treatment upon comorbidity in PDA + C and PDA conditions

PDA+C (n = 22)	PDA (n = 23)
# Comorbid diagnoses, CSR 4+
Baseline	2.1 (1.5)	1.9 (1.0)
Post	0.6 (0.9)	0.3 (0.5)
Follow-up	0.2 (0.5)	0.1 (0.3)
# Comorbid diagnoses, CSR 1 +
Baseline	3.3 (1.9)	3.0 (1.6)
Post	1.9 (1.4)	1.7 (1.4)
Follow-up	1.7 (1.1)	0.7 (1.4)
CSR for most severe baseline comorbid diagnosis
Baseline	4.6 (1.3)	4.6 (1.0)
Post	1.7 (1.7)	1.3 (1.7)
Follow-up	1.7 (1.6)	0.4 (1.1)
% I + diagnosis, CSR 4 +
Baseline	81.8	87.0
Post	36.4	21.7
Follow-up	18.2	13.0
% I + diagnosis, CSR 1+
Baseline	100.0	100.0
Post	77.3	82.6
Follow-up	81.8	34.8

effects of condition (ES = 0.05) or Condition x Time (ES = 0.03). The proportion of participants with any comorbid diagnosis with CSRs of 4 or higher reduced from 84.4% at baseline to 28.9% at post-treatment, and 15.6% at follow-up, with no differences between conditions. However, the proportion of participants with any comorbid diagnosis of clinical (CSR 4 +) or subclinical (CSR 1-3) severity reduced from 100% at baseline to 80% at post, and 57.8% at follow-up, with proportionately fewer PDA participants having any such diagnoses at follow-up (34.8%) than PDA + C (81.8%), X²(1) = 10.2,p<0.01. Most importantly, CSR values for the most severe baseline comorbid diagnosis not only declined over Time, F(2,86) = 99.1, pES = 0.70, but in a way that tended to differ between the conditions, F(2,86) = 3.0, p<0.06, ES = 0.07, and specifically due to a significant Condition x Time linear interaction effect, F(1,43) = 6.1, p<0.05, ES = 0.11. Whereas both conditions decreased from baseline to post-treatment, PDA showed a pattern of continuing reduction from post to follow-up, t(22) = 2.5, p <0.05, while PDA + C showed no further improvement from post to follow-up. The pattern of results did not differ when analyses were restricted to participants with comorbid diagnoses associated with CSR values of 4 or higher at baseline. Scores on the BSI declined over time, F(2,66) = 29.0, pES = 0.47, as did scores on the SSS, F(2,66) = 33.7,p<0.001, ES = 0.51. Although the Condition x Time interaction terms were not significant for SSS (ES = 0.02) or BSI (ES = 0.06), the quadratic interaction effect was significant for BSI, F(1,33) = 4.0, p = 0.05, ES = 0.10), due to lower severities at post-treatment in PDA vs. PDA + C, t(19.9) = 2.23, p<0.05. Means and standard deviations are shown in Table 5. Scores on the NEO increased significantly over time, F(2,64) = 11.5, pES = 0.23, as did scores on the ACQ, F(2,64) = 32.9, pES = 0.51. Although trends indicated greater increases in ACQ by post and follow-up in the PDA group, neither the interaction terms nor the linear and quadratic contrasts were significant (ES = 0.04-0.06) (means and SD’s are shown in Table 5). The intent-to-treat analyses replicated the overall reductions in the number of comorbid diagnoses with CSR values of 4 or higher, F(2,126) = 77.7, pES = 0.55, and with CSR values of 1 or higher, F(2,126) = 29.7,p<0.001, ES = 0.32, without interaction effects (ES = 0.02 and 0.03). In addition, the intent­to-treat analyses rendered the Condition x Time interaction effect for the CSR for the most severe baseline

M. G. Craske et al. / Behaviour Research and Therapy 45 (2007) 1095-1109

Table 5

Effect of treatment upon general symptoms and emotional vulnerabilities in PDA + C and PDA conditions

 

PDA+C (n = 17)	PDA (n = 18)
Brief symptom inventory total
Baseline	65.8 (37.5)	59.8 (31.1)
Post	47.2 (48.0)	23.3 (15.6)
Follow-up	33.7 (41.9)	17.6 (16.8)
Subjective symptoms scale
Baseline	19.1 (10.6)	18.1 (5.8)
Post	11.4 (9.7)	7.7 (5.6)
Follow-up	9.1 (9.0)	7.7(8.4)
Neuroticism
Baseline	19.7 (6.3)	20.1 (5.2)
Post	21.8 (6.1)	21.7 (5.9)
Follow-up	23.1 (5.7)	24.7 (7.1)
Anxiety control questionnaire
Baseline	73.5 (20.6)	71.1 (16.3)
Post	91.4 (22.3)	102.4 (18.0)
Follow-up	96.9 (22.9)	103.6 (22.6)

comorbid diagnosis a trend only, with a very small effect size, F(2,126) = 2.4, p = 0.1, ES = 0.04. However, the linear contrast effect for the interaction term showed a trend to significance, F(1,63) = 3.9, p = 0.05, ES = 0.07, due to a lower CSR value at post-treatment in the PDA group, t(64) = 2.5,p<0.05, and due to the CSR value continuing to decrease from post to follow-up within the PDA group, t(32) = 2.4, p < 0.05, but not within the PDA + C group. The percentage of the entire intent-to-treat sample with at least one comorbid diagnosis with CSR of 4 + was 43.1 % at post and 35.4% at follow-up, with no differences between conditions. However, the percentage with at least one comorbid diagnosis of any severity at follow-up remained significantly different between conditions in the intent-to-treat sample (PDA +C = 87.5%, PDA = 51.5%), X²(l) = 9.9, P <0.01. The pattern of results was replicated when analyses were restricted to participants with CSRs of 4 or higher for baseline comorbid diagnoses.

Additional treatment

The groups did not differ in the proportions recelVlng medications at baseline (pDA + C = 54.5%; PDA = 39.1 %), at post-treatment (PDA + C = 40%; PDA = 45.8%) or at follow-up (PDA + C = 43.5%; PDA = 54.5%). Similarly, no differences were found in the proportions receiving additional psychotherapy at baseline and at post-treatment (pDA +C = 24%; PDA = 18.5%) or at follow-up (PDA +C = 26.1 %; PDA = 17.4%).

Predictors of comorbidity at follow-up

Hierarchical linear regressions were used to evaluate predictors of CSR values of the most severe baseline comorbid diagnosis at follow-up. All variables were first centered. Severity of the most severe comorbid diagnosis at baseline was entered on the first step; changes in panic disorder CSR, BSI and SSS values from pre- to post-treatment were entered on the second step using a forward stepwise approach; and post-treatment measures of emotional vulnerability (NEO, ACQ) were entered on the third step, again using a forward stepwise approach. After controlling for CSRs for the most severe comorbidity at baseline, R² = 0.002, P = 0.047, ns, only scores on the ACQ at post-treatment, R² = 0.15, P = -0.38, p<0.05, F(1,38) = 6.4, P < 0.05, added significantly to the CSR value of the comorbid diagnosis at follow-up. The same result was obtained when the sum of CSR values of all comorbidity at baseline was entered on step 1, to predict the sum of CSR values of all comorbidity at follow-up.

Discussion

The results from this preliminary investigation provide partial support for the primary study hypothesis.

Specifically, more of the same CBT focused on a principal diagnosis of panic disorder/agoraphobia yielded more positive impact upon the principal diagnosis and upon comorbid conditions than did less of more CBT that targeted a comorbid condition. However, the results were limited to a restricted number of measures, some with small effect sizes, and some of which were not retained in, albeit very conservative, intent-to-treat analyses.

Nonetheless, the results raise the interesting possibility that staying focused is superior to straying.

Specifically, PDA treatment resulted in higher rates of high end-state at post-treatment, a finding that was retained in the conservative intent-to-treat analyses, and a higher proportion reporting zero panic attacks at follow-up, six to twelve months following treatment completion, than did the PDA + C treatment. Notably, the rates of high end-state status are slightly lower than other studies of CBT for panic disorder (e.g., Clark et aI., 1994), possibly due to the relatively high rates of moderate to severe agoraphobia, which tend to be associated with lower rates of high-end state (Williams & Falbo, 1996). On the other hand, the two groups did not differ in terms of standardized self-report measures of panic disorder/agoraphobia symptomatology, including the anxiety sensitivity index and the agoraphobia subscale of the fear questionnaire. Similarly, rates of improvement were equivalent between the two conditions for the behavioral approach tasks involving interoceptive inductions and an enclosed booth.

Our sample was selected to have at least one comorbid anxiety or mood diagnosis with a CSR value of 3 or higher. Notably, the pattern of results did not differ when analyses were restricted to those with a CSR value of 4 or higher for baseline comorbid anxiety or mood diagnoses. As is typical, the most frequently co­occurring disorders were generalized anxiety disorder, specific phobias, major depression, and social phobia (Brown, Campbell, Lehman, Grishman, & Mancill, 2001; Tsao et aI., 2002). The most severe comorbid diagnosis at baseline was tracked over time for each participant by an independent clinical interviewer who was blind to the participant’s treatment assignment. This clinical severity rating reduced more so in more of the same, single focused CBT compared to diversified CBT that addressed comorbidity, with reductions continuing from post to follow-up assessment, although these effects diminished in the intent-to-treat analyses. However, a greater proportion of those receiving more of the same CBT were rated as having no comorbid diagnosis of clinical or subclinical severity at follow-up; a finding which was retained in the intent­to-treat analyses. On the other hand, overall numbers of comorbid diagnoses did not differ between the groups over time, although there was a trend in the standardized self-report questionnaires of general emotional distress for greater improvements by post-treatment in the more of the same, single focused CBT condition.

Notably, the purpose of this study was to directly evaluate the impact of straying to another disorder focus.

The different question of the effectiveness of equal attention given to principal and comorbid diagnoses would involve a comparison between an equal dosage of CBT for a principal diagnosis combined with either 12 sessions of CBT for comorbid diagnoses or 12 sessions of an active placebo.

We hypothesized that those who responded most positively to CBT for panic disorder/agoraphobia would show the greatest declines in comorbidity. Interestingly, after controlling for baseline values of the most severe comorbidity at baseline, the CSR for that same diagnosis at follow-up was predicted only by post-treatment scores on the anxiety control questionnaire, with those reporting more perceived control having less comorbidity severity. The same finding occurred when predicting sum of CSRs for all comorbid diagnoses at follow-up. This finding is consistent with the notion that comorbidity decreases when CBT effectively down regulates shared emotional processes such as emotional control. Interestingly, our other measure of a shared emotional process, neuroticism, was not a significant predictor. Nonetheless, the findings regarding emotional control are consistent with experimental evidence for the role of uncontrollability in the generation of anxiety (e.g., Mineka & Cook, 1986) and models that emphasize the role of perceived uncontrollability over events and emotions as a critical risk factor for proneness to anxiety (e.g., Barlow, 2002; Chorpita & Barlow, 1998).

Clearly, the current findings are in need of replication in larger sample sizes before strong clinical implications can be drawn. There also are a number of limitations to the current study. For example, comorbidity was limited to other anxiety disorders and depression, and the sample was selected so as to exclude those with another common comorbid diagnosis, that being substance use related disorders. Other exclusions, such as suicidality and bipolar disorders, also limited the generalizability of the sample. Second, evaluation of the content of the individual treatment sessions was limited to case note reviews; independent rating of audio-taped treatment sessions would have provided more detail and a check on the quality with which CBT was implemented in these individual sessions. Third, as already noted, small sample sizes, especially for questionnaire and behavioral measures, limited the effects. Fourth, lengthier follow-up intervals are needed, especially given the evidence that comorbidity may resurge when assessed two years following CBT for panic disorder (Brown & Barlow, 1995). Fifth, the use of group treatment may limit the generalizability of the findings, although group treatments for panic disorder/agoraphobia generally yield similar effects to individual treatments (Neron, Lacroix, & Chaput, 1995; Lidren, Watkins, Gould, Clum, Asterino, & Tulloch, 1994). Sixth, although diagnostic reliability for the anxiety disorders is usually satisfactory, at least when using fully or semi-structured interviews (e.g., DiNardo, Moras, Barlow, Rapee, & Brown, 1993; Brown et aI., 2001), reliability is more problematic for comorbid conditions of depression. In addition, reliability is compromised for all diagnoses when attempting to differentiate between the subclinical and clinicalleve1s of severity. Consequently, unreliability may lead to apparent shifts in diagnostic status when none really exist.

In summary, group CBT targeting panic disorder and agoraphobia combined with biweekly individual CBT sessions similarly targeting panic disorder and agoraphobia yielded more positive influences both in terms of the principal diagnosis of panic disorder and agoraphobia, and comorbidity by follow-up. The results are limited but nonetheless raise the possibility that doing more of the same may yield better overall results than doing less of more.

References

Barlow, D. H. (2002). Anxiety and its disorders: The nature and treatment of anxiety and panic (Second Ed). New York, US: The Guilford Press.

Barlow, D. H., & Craske, M. G. (1988). Mastery of your anxiety and panic. Albany, New York: Graywind Publications.

Borkovec, T. D., Abel, J. L., & Newman, H. (1995). Effects of psychotherapy on comorbid conditions in generalized anxiety disorder.

Journal of Consulting & Clinical Psychology, 63(3), 479-483.

Brown, T. A., Campbell, L. A., Lehman, C. L., Grishman, J. R., & Mancill, R. B. (2001). Current and lifetime comorbidity of the DSM-IV anxiety and mood disorders in a large clinical sample. Journal of Abnormal Psychology, 110(4), 585-599.

Brown, T. A., & Barlow, D. H. (1992)