1 INTRODUCTION

Anxiety disorders are highly prevalent, affecting 11.6% of the global population yearly (Baxter et al., 2013), women almost twice as often as men (Remes et al., 2016), and frequently presenting with comorbid other anxiety (48%–68%) or depressive disorders (63%; Lamers et al., 2011). The recommended psychotherapy, cognitive behavioral therapy (CBT), targets maladaptive cognitions and dysfunctional behavioral anxiety-related patterns, by challenging automatic thoughts about feared situations and by fear exposure. Numerous meta-analyses (Carpenter et al., 2018; Cuijpers et al., 2016; Hofmann & Smits, 2008; Stewart & Chambless, 2009) demonstrated CBT′s effectiveness for anxiety disorders in individual as well as group formats (Barkowski et al., 2016), effect sizes being typically medium to large compared to waitlist control conditions (Cuijpers et al., 2016). However, only 51% of anxiety disorder patients reach remission after receiving CBT (Springer et al., 2018), some patients refuse CBT (Goetter et al., 2020), and one in five discontinue therapy prematurely (Taylor et al., 2012). Fear of the therapy itself likely plays a role in treatment refusal (Cougle, 2012; Goetter et al., 2020). Most multidisciplinary guidelines list no, or only few, other effective evidence-based psychotherapeutic interventions for anxiety disorders (American Psychiatric Association, 2009; Baldwin et al., 2014; Katzman et al., 2014; National Institute for Health and Care Excellence, 2020; Trimbos Institute, 2013).

The present study therefore compared CBT to a novel approach to anxiety disorders, autonomy enhancing treatment (AET). AET is a transdiagnostic psychotherapy based on the rationale that deficits in autonomy-connectedness act as a vulnerability factor for the development and maintenance of anxiety- and other disorders (Bekker et al., 2016). Autonomy-connectedness is the capacity for self-governance in an interpersonal context (Bekker & van Assen, 2006). Anxiety patients may have difficulty identifying their own needs and wishes, tend to focus excessively on the needs of others, and are typically sub-assertive (Hawke & Provencher, 2012; Russell et al., 2011). Such autonomy deficits can be conceptualized using the three components of autonomy-connectedness: (i) “self-awareness,” the awareness of one's wishes, needs, and opinions, and the ability to communicate and pursue these in interpersonal relationships (Bekker & van Assen, 2006). Individuals high in self-awareness have a clear sense of identity and self-concept, are effective at communicating their needs and wishes (Bekker et al., 2008), and seem more resilient when facing interpersonal stressors (Kunst et al., 2019). Further, (ii) “sensitivity towards others” is the sensitivity towards the wishes, needs, and desires of other people, including the need and capacity for intimacy; and (iii) “capacity for managing new situations” reflects the tendency to explore new environments, flexibility, and independency of familiar structures (Bekker & van Assen, 2006).

Studies on internalizing disorders consistently showed lower levels of self-awareness and capacity for managing new situations, and higher sensitivity towards others, in individuals with anxiety or depression (symptoms or disorders) compared with healthy controls (Bekker & Belt, 2006; Bekker & Croon, 2010; Bekker & van Assen, 2017; Kunst et al., 2019; Maas, Laceulle, et al., 2019). Studies using different conceptualizations of autonomy deficits also show consistent and positive associations with anxiety symptoms across healthy and patient populations (e.g., Cámara & Calvete, 2012; Dunkley et al., 2006; Hallam et al., 2014; Hawke & Provencher, 2012). Maladaptive autonomy-connectedness patterns may predispose individuals to design one's life incongruently with own wishes and needs, raising chronic distress, and experiencing a chronic lack of control, thereby adopting avoidant and ruminative coping styles. AET therefore aims to strengthen self-awareness and capacity for managing new situations and to normalize one′s sensitivity to others (Bekker et al., 2016). A recent pilot indicated that AET was acceptable to patients and therapists, feasible to carry out in specialized mental health care, and indications were found for effectiveness compared with a waitlist control condition (Maas, van Balkom, et al., 2019). However, this study was underpowered and AET has never been compared directly to CBT.

The present study aimed to compare CBT versus AET in a mixed group of anxiety disorders. Because AET is primarily group-based psychotherapy, we compared AET to group-based CBT. We expected broader effects for AET than CBT (i.e., enhanced effects on general psychopathology, comorbid depressive symptoms, quality of life, autonomy-connectedness, self-esteem), and no significant differences for anxiety change. Additionally, treatment expectancies, adherence and evaluation, and long-term effects were explored.

2 MATERIALS AND METHODS 2.1 Design

The present study was a multicenter cluster-randomized clinical trial on the comparative effectiveness of 15-week group-AET and -CBT for anxiety disorders. Patients were randomly allocated to AET or CBT; outcome measures were assessed at six time points: pre-, mid-, and posttreatment, and at 3-, 6-, and 12-month follow-up. Eight treatment centers in the Netherlands participated in the study. The study was approved by the Brabant Medical Ethical Committee (#NL59064.028.16) and preregistered at the Netherlands Clinical Trial Registry (https://www.trialregister.nl/trial/6250)

2.2 Participants

Participants were 129 adult patients with DSM-5 anxiety disorders (panic disorder, social anxiety disorder, generalized anxiety disorder, agoraphobia, or specific phobia), aged M = 33.66 (SD = 12.57), 91 (70.5%) female (Table 1). Improving generalizability of our findings to the population of anxiety patients, comorbidity was permitted in our trial: 45.7% of participants had one or multiple comorbid anxiety disorder(s) and 34.1% had a comorbid depressive disorder. Exclusion criteria were: insufficient mastery of Dutch language, suspected intellectual disability, psychosis, addiction, severe personality pathology, acute bereavement, severe acute suicidality, and/or having undergone AET, CBT, or medication changes in the past 3 months. Stable use of psychopharmacological drugs (>3 months) was allowed.

Table 1. Patient characteristics, treatment expectancies, adherence and evaluation, by condition (n = 129) AET (n = 62) CBT (n = 67) AET (n = 62) CBT (n = 67) N (%) N (%) X2 (p) M (SD) M (SD) t (p) d Sex Age 34.59 (13.59) 32.81 (11.84) −.80 (.425) 0.14 Female 46 (50.5) 45 (49.5) .77 (.382) Expectancies Education Expected effectiveness 3.61 (0.94) 3.55 (0.78) .33 (.744) 0.07 Primary school 1 (50) 1 (50) 3.53 (.618) Motivation 4.59 (0.62) 4.63 (0.65) −.36 (.722) −0.066 Lower secondary 7 (36.8) 12 (63.2) Higher secondary 12 (48) 13 (52) Adherence (n = 97) Intermediate vocational 21 (58.3) 15 (41.7) Missed sessions 1.56 (1.25) 1.35 (1.22) .86 (.391) 0.18 Higher vocational 16 (47.1) 18 (52.9) Weekly minutes homework 89.69 (54.70) 150.61 (120.54) −3.22 (.002) −0.65 University degree 4 (33.3) 8 (66.7) Homework adherence 4.15 (1.19) 3.92 (1.038) 1.006 (.317) 0.20 Country of birth Evaluation (n = 97) The Netherlands 55 (46.2) 64 (53.8) 2.09 (.148) Treatment satisfaction 7.63 (1.68) 7.71 (1.099) −.31 (.758) −.056 Therapist alliance 4.27 (0.68) 4.33 (0.59) −.43 (.667) −.092 Primary diagnosis Group alliance 4.56 (0.49) 4.32 (0.60) 2.21 (.030) .45 Social anxiety 24 (47.1) 27 (52.9) 7.019 (.219) Too much homework 2.94 (0.38) 3.22 (0.51) −.31 (.002) −.62 Generalized anxiety dis. 18 (60) 12 (40) Therapy strain 3.50 (0.99) 3.67 (0.99) −.86 (.390) −.17 Panic disorder 11 (36.7) 19 (63.3) Person (vs. symptom) focused 3.31 (0.88) 2.65 (0.72) 4.039 (<.001) .82 Agoraphobia 0 3 (100) Focus on core of the problem 3.54 (1.41) 3.76 (1.15) −.82 (.417) −.17 Specific phobia 1 (50) 1 (50) Not otherwise specified 8 (61.5) 5 (38.5) N (%) N (%) □² (p) Previous treatments Comorbidity No 10 (37) 17 (63) 4.79 (.310) 1≤ other anxiety disorder 31 (52.5) 28 (47.5) .87 (.350) 1 treatment of 1 year or less 24 (53.3) 21 (46.7) Any depressive disorder 21 (47.7) 23 (52.3) .003 (.956) 1 treatment of 1–2 years or 2 short treatments 17 (54.8) 14 (45.2) 1 treatment of 3–7 years or 3–5 short treatments 10 (50) 10 (50) Psychotropic medication 1 treatment longer than 8 years or 6≤ treatments 1 (16.7) 5 (83.3) No 42 (47.7) 46 (52.3) .012 (.911) Benzodiazepines 10 (40) 15 (60) 1.98 (.160) Previous CBT for anxiety Antidepressants 10 (55.6) 8 (44.4) .59 (.443) Yes, probably 14 (46.7) 16 (53.3) Bèta blockers, 6 (54.5) 5 (45.5) .20 (.655) No, probably not 34 (48.6) 36 (51.4) antipsychotics or anticonvulsants Unclear 14 (48.3) 15 (51.7) 2.3 Procedure and randomization

Figure 1 displays patients' inclusion flow. Patients referred to one of the eight treatment centers underwent an intake procedure following Dutch multidisciplinary guidelines (inclusion: February 2017 to July 2019, follow-up stop: November 2019). Those diagnosed with a primary DSM-5 anxiety disorder and interested in research (n = 219) were contacted by author Laura E. Kunst by phone to assess eligibility and diagnoses (using the MINI International Neuropsychiatric Interview, version 5.0.0.; Sheehan et al., 1998; Vliet & de Beurs, 2007), which has excellent inter-rater reliability and good concurrence with more extensive clinical interviews (Sheehan et al., 1997). Written informed consent was obtained from all patients. Cluster-based randomization was applied (Teerenstra et al., 2006; Method s1). Each treatment center administered both AET and CBT groups in a predetermined, randomized, and balanced order, ensuring an equal distribution of both within each center. To limit potential selection bias associated with cluster-randomized procedures, we instructed treatment centers to blind intake staff for each upcoming condition, and during screening we verified that patients were unaware of treatment allocation.

CONSORT flow diagram

Included participants received questionnaires 1 week before the first group session (pre-test), after the seventh session (mid-test), and after the last session (post-test). Follow-up measurements were performed after 3, 6, and 12 months (see Method s2). Due to the project's time limit of 3 years, the first therapy groups received questionnaires at all six time points, and the final groups only at the pre-, mid-, and post-test. Patients were asked to refrain from additional treatment for 3 months after post-test.

2.4 Treatments

AET consisted of 15 weekly 2-h sessions following the protocol of Bekker et al. (2016), see Method s2 for details and a session-by-session overview. AET groups had an average of 8 participating patients at baseline (range 5–11). Sessions 1–5 consisted of psycho-education about autonomy-connectedness and its relation to psychopathology, and setting autonomy-related personal goals. From session 2 onward, patients took turns chairing the therapy sessions (opening and structuring the session, managing time), so that every patient led a total of one or two sessions, and so that therapy groups became increasingly “autonomous.” Sessions 6–12 featured different autonomy-related themes each week, serving as psycho-education as well as grounds for discussion over autonomy-related difficulties (e.g., relationships, boundaries, body and sexuality, emotions), and sessions 13–15 included relapse prevention. Key elements of CBT for anxiety disorders (e.g., behavioral experiments, interoceptive exposure, metacognitions, addressing self-monitoring and safety behaviors) were not applied within AET.

CBT also consisted of 15 weekly 2-h sessions. CBT groups had an average of 7 participating patients at baseline (range 5–9). CBT was administered following the evidence-based protocol of Keijsers et al. (2017), the gold standard CBT protocol used in the Netherlands. In terms of international literature, the protocol corresponds with treatment principles as described by Barlow and Craske (2007) for panic disorder; Clark and Wells (1995), and Hofmann and Otto (2008) for social anxiety disorder; gradual exposure in vivo for specific phobia (Wolitzky-Taylor et al., 2008); and cognitive therapy based on Beck et al. (1985) and the metacognitive model (Wells, 1995, 1997) for generalized anxiety disorder. To summarize, all sessions started with a relaxation exercise and homework discussion; sessions 1–3 focused on psycho-education following CBT anxiety models and formulating personal CBT-goals as well as anxiety registration and thought records, sessions 4–12 covered identifying and challenging maladaptive automatic thoughts, interoceptive and in vivo exposure, addressing meta-cognitions, behavioral experiments, and challenging core beliefs, and sessions 13–15 included relapse prevention (for a more elaborate description, see Method s2).

2.5 Therapists and training

Thirty-seven therapists participated in the study, aged M = 38.32 (SD = 11.66), 94.6% female, with on average 10.77 (SD = 9.39) years of experience in mental health care and ample experience with CBT for anxiety. Each AET and CBT therapy group was administered by two therapists, one of whom was always an experienced psychologist with 2<years of post-master clinical training. All AET therapists received a 1-day AET training. Most (89.2%) therapists administered either AET or CBT groups and four (10.8%) experienced therapists administered AET as well as CBT groups (see treatment integrity ratings). Regular supervision meetings by phone were organized, led by Brenda Kouwenhoven for AET and Kees Korrelboom for CBT (Method s3).

2.6 Treatment integrity ratings

Treatment integrity was assessed by rating 20% of the audio-recorded therapy sessions (Perepletchikova, 2011) for strict adherence to the AET and CBT protocols (scale from 0 to 10). Additionally, use of AET (e.g., focusing on self-acceptance, stimulating awareness and expression of needs) and CBT techniques (e.g., challenging thoughts, discussing symptom registration, planning behavioral experiments) were scored on a scale from 0 to 5 (see Method s3). Not all techniques were expected to be applied in every session: For instance, a perfect CBT session focusing on challenging thoughts would not receive a perfect score on use of CBT techniques, if behavioral experiments and symptom registration were not discussed in that session. We also expected some overlap in utilized techniques between conditions, because using basic therapeutic skills in response to behaviors initiated by patients can sometimes be rated as either CBT or AET techniques. For example, asking a patient within AET what exactly they fear, could be considered a CBT technique (challenging thoughts), and complimenting a patient who assertively expressed their needs in a CBT session, could be considered an AET technique (stimulating expression of needs). However, AET techniques were expected to be applied more in AET than CBT and vice versa.

Sessions were rated independently by two out of seven available raters and interrater reliabilities were computed using the intraclass correlation (ICC), based on two-way random effects model for absolute agreement and average measures (Gisev et al., 2013). Rater agreement was good for ratings of protocol adherence (ICC = .77) and moderate for ratings of therapeutic techniques (ICC = .69). Protocol adherence was high (M = 8.62 for AET [SD = 2.28] and M = 8.36 for CBT [SD = 2.43]), indicating that the sessions were provided as described in the protocols and that the treatments were clearly distinct. As expected, CBT techniques were applied more in CBT (M = 2.016, SD = 1.42) than in AET groups (M = 0.85, SD = 0.81, t(251.4) = 8.38, p < .001), and AET techniques were applied more in AET (M = 2.32, SD = 1.23) than CBT groups (M = 1.17, SD = 0.91, t(187.19) = −8.30, p < .001).

2.7 Primary outcome measures

Anxiety (10 items) and general psychopathology were measured using the Symptom Checklist (SCL-90), consisting of 90 5-point Likert items and nine (sub)scales: Agoraphobia, Anxiety, Depression, Somatization, Cognitive-performance deficits, Interpersonal sensitivity, Hostility, Sleep difficulties, and General psychopathology (total score; Arrindell & Ettema, 1975, 2005; Derogatis, 1994), asking if patients experienced symptoms in the past week. Cronbach's alphas for anxiety ranged from α = .87 (12-month follow-up) to α = .92 (mid-test); for General psychopathology the range was α = .97 (pre-test) to .98 (post-test).

Depressive symptoms were measured using the 21-item (four answering categories) Beck Depression Inventory (BDI-II-NL; Beck et al., 1996; Beck et al., 2002), which has excellent psychometric properties (Beck et al., 1988). Reliability in the present study ranged from α = .89 (mid-test) to α = .92 (3-month follow-up).

Quality of life was measured using the 26-item World Health Organization quality of life brief questionnaire (WHO-QoL-BREF), which assesses QoL in domains of physical health, psychological well-being, social relationships, and environment. It has good psychometric properties (World Health Organization, 1998), with reliability ranging from α = .90 (pre-test) to α = .93 (6-month follow-up).

Autonomy-connectedness was measured using the 30-item (five answering categories) Autonomy-Connectedness Scale (ACS-30), with subscales self-awareness (7 items), Sensitivity to others (17 items), and Capacity for managing new situations (6 items). Construct validity and internal reliability are good (Bekker & van Assen, 2006). Reliabilities in the present study ranged from α = .81 (6-month follow-up) to α = .87 (12-month follow-up) for self-awareness, from α = .83 to α = .87 (3-month follow-up) for Sensitivity to others, and from α = .74 (12-month follow-up) to α = .84 (6-month follow-up) for Capacity for managing new situations.

Self-esteem was assessed using the 10-item (four categories) Rosenberg Self-esteem Scale (RSES; Rosenberg, 1965), which has good psychometric qualities (Franck et al., 2008). Reliabilities in the present study ranged from α = .85 (pre-test) to α = .90 (3-month follow-up).

2.8 Patient characteristics and secondary outcome measures

Previous treatments, treatment expectancies, adherence, and evaluation were assessed using 5-point Likert scales and open-ended questions (Method s4).

2.9 Power considerations

A priori sample size calculation using G*Power 3.1.9.2, for independent t-tests on comparing change scores between conditions, medium effect size (d = .5), revealed that n = 128 participants (64 per group) were needed to achieve a power of .80.

2.10 Statistical analysis

A detailed description is included in Method s5. All analyses were conducted in SPSS (v.26) using an alpha of .05 and two-tailed testing. For the primary outcome analyses, an intention-to-treat (i.t.t.; n = 129) and completers (excluding drop-outs, n = 101) approach was used. Data of pre, mid, post-test, and 3-month follow-up were used; 6- and 12-month follow-up time points were examined exploratively due to their low completion rates. Primary outcome analyses were carried out using linear mixed modeling with an unstructured covariance matrix at the level of individuals (as MANOVA), and a random intercept for therapy group to control for dependency within therapy groups (level 2). Two models were estimated for each of the primary outcome measures and compared using the likelihood ratio test: (0) A null-model including fixed and random intercept and unstructured covariance matrix for pre-test to 3-month follow-up. Model (I) also included a main effect of condition (0 = CBT, 1 = AET) and of time (three time dummies reflecting the difference compared with baseline). In model (II) three interaction terms were added between conditions and each of the time dummies. Main hypotheses on a different change in the outcome measure between conditions were tested using the t-statistic of the interaction terms in model (II) for timepoints pre-test to post-test and pre-test to 3-month follow-up.

Although the presence of multiple outcome measures inflates the chance of Type-I error, Bonferroni correction would excessively inflate the Type-II error probability. We therefore chose to base our conclusions on the uncorrected tests, but also reported whether the outcomes would differ using a corrected alpha of .05/8 = .0063. Effect sizes were calculated for within-subject changes per condition as well as for between-subjects (AET vs. CBT) using Glass's delta (Lakens, 2013; Morris, 2008), which is more conservative than within-subject Cohen's drm. Pre-test standard deviations were pooled across conditions to obtain more reliable estimates of the population standard deviations and to facilitate comparison between conditions (Lakens, 2013; Tables S1–S8). Moderator analyses were performed using bootstrapped regression analyses (see Methods S5) to assess whether treatment response varied by demographic and clinical characteristics.

3 RESULTS

Participant characteristics are depicted in Table 1 and descriptive statistics are presented in Tables 2 and S1–S8. Intention-to-treat (Table 3, model I) as well as completers (Table S9) analyses showed that all outcomes improved from pre-test to all subsequent time points in the sample as a whole. Within-subject changes in anxiety, psychopathology, and depression at post-test and follow-up were large in both conditions (Tables 3 and S9). As expected, changes in anxiety were not significantly different between AET and CBT (shown in Figure 2). In contrast to the hypotheses, no differences between AET and CBT emerged for the broader outcomes: general psychopathology, comorbid depressive symptoms, quality of life, autonomy-connectedness, and self-esteem. These results did not vary by intention-to-treat vs. completer sample. Favorable treatment response (i.e., larger anxiety changes from pre- to post-test) and differential treatment response to AET versus CBT were not predicted by possible moderators sex, age, education level, prior CBT treatment, or primary diagnosis (social and generalized anxiety disorder vs. other anxiety disorders), ps > .05. Adopting a Bonferroni-corrected alpha of .05/8 = .0063 would eliminate the one *-marked correlations in Table 2 and the changes in Sensitivity to others and Capacity for managing new situations from pre-test to mid-test (Tables 3 and S9), not affecting the main conclusions.

Table 2. Means, standard deviations, and correlations between variables under study Means Correlations AET CBT M SD M SD