A Randomized Clinical Trial of Cognitive-Behavioral Therapy and Applied Relaxation for Adults With Generalized Anxiety Disorder

Generalized anxiety disorder (GAD) has been reconceptualized from a mild, broad anxiety syndrome to a disabling disorder characterized by excessive, uncontrollable worry. Treatments have evolved from general anxiety-reduction approaches (e.g., applied relaxation) to interventions targeting mechanisms of pathological worry (e.g., CBT protocols). The authors developed a CBT protocol grounded in a cognitive model in which intolerance of uncertainty, positive beliefs about worry, negative problem orientation, and cognitive avoidance maintain GAD. Prior trials showed this CBT superior to wait-list and nondirective therapy, with maintenance and even continued improvement in worry over long-term follow-up. The present study aimed to compare this GAD-specific CBT to an active, directive comparison—applied relaxation (AR)—and to replicate superiority of both active treatments over a wait-list control. Hypotheses: (1) Both CBT and AR would be superior to WL at posttest; (2) CBT would be superior to AR over follow-up; (3) CBT (and not AR) would lead to continued improvement over follow-up. To mitigate allegiance effects, independent assessors were blinded, a therapist without CBT training treated most participants, and experts supervised the therapist in each modality.

Applied relaxation (AR) has longstanding empirical support for GAD and is considered an empirically supported treatment, with evidence for reducing diagnostic severity, worry, anxiety, depression, and general psychopathology. Few direct comparisons exist between GAD-specific CBT and AR alone; Borkovec and Costello (1993) found CBT and AR equivalent at posttreatment, with maintenance and some evidence of further gains in CBT at follow-up. The authors’ CBT targets intolerance of uncertainty and related cognitive factors, showing superiority to wait-list and nondirective therapy in prior trials, with sustained or continued improvements in worry. The literature suggests AR is effective, but whether GAD-specific CBT offers advantages over AR, especially long-term, remains to be clarified.

Design: Randomized clinical trial with three conditions: CBT, AR, and wait-list (WL). Treatments: 12 weekly 1-hour manualized sessions. Assessments: pretest (and pretest-wait-list for WL), posttreatment, and 6-, 12-, and 24-month follow-ups. Blinded independent assessors administered diagnostic interviews and measures at all time points. Participants: N=65 (43 women, 22 men), Francophone adults with primary GAD; mean age 38.5 (SD=12.0), education 15.3 years (SD=3.4). Ethnicity: 91% White/European, 5% Middle Eastern, 2% Hispanic, 2% Asian. Employment: 62.5% employed, 10.9% students, 26.6% unemployed. Mean GAD duration 13.9 years (SD=16.7); mean ADIS-IV Clinician’s Severity Rating (CSR) 5.7 (SD=1.2). Comorbidity in 58.5% (panic disorder n=27; specific phobia n=13; social anxiety disorder n=9; dysthymic disorder n=8; major depressive disorder n=5; OCD n=3; hypochondriasis n=1). Medication: 55.4% on anxiolytics or antidepressants; 43.1% had prior CBT. Recruitment/Screening: Patients referred to the Anxiety Disorders Clinic or by local providers. Initial screening with MINI (Version 4.4); candidates meeting primary GAD were consented and assessed with ADIS-IV and questionnaires. Inclusion criteria: Primary GAD with CSR ≥4; ≥2-point CSR difference between GAD and any comorbid conditions; age 18–64; stable medication (≥4 weeks for benzodiazepines; ≥12 weeks for antidepressants/hypnotics) prior to assessment and during treatment; no suicidal intent; no current substance abuse; no current/past schizophrenia, bipolar disorder, or organic mental disorder. Randomization and allocation: Of 83 assessed, 65 eligible participants randomized to CBT (n=23), AR (n=22), or WL (n=20). Allocation implemented by research coordinator after team meeting; psychiatrist informed patients of allocation. Therapists: Main therapist (Renée Leblanc), a licensed psychologist with psychodynamic training, treated 61/65 participants; received ~8 hours of formal training in each modality and weekly supervision (~1 hour per modality) from study authors. The first author treated 4 participants early for training purposes. Treatment integrity assessed via checklists on randomly selected cases. Interventions: CBT (12 sessions): (1) Psychoeducation and worry monitoring; distinguishing current-problem vs hypothetical worries. (2) Uncertainty recognition and behavioral exposure to uncertainty-inducing situations. (3) Reevaluation of positive beliefs about worry. (4) Problem-solving training for current problems. (5) Imaginal exposure for hypothetical worries (downward arrow technique; looped recording for daily exposure). Relapse prevention in final session; written summaries and homework forms throughout. AR (12 sessions): (1) Psychoeducation and tension monitoring. (2) Progressive tension-release training across muscle groups. (3) Relaxation by recall. (4) Relaxation by counting. (5) Conditioned relaxation applied in real-life via graded hierarchies; relapse prevention and written materials similar to CBT. WL: 12-week wait-list, monitored via phone every three weeks by the psychiatrist. Measures: - Diagnostic: MINI; ADIS-IV with CSR (0–8 scale). Interrater reliability assessed across interviews. - GAD/worry/somatic: Penn State Worry Questionnaire (PSWQ); Worry and Anxiety Questionnaire Somatic subscale (WAQ-Som). - Ancillary: State-Trait Anxiety Inventory–Trait (STAI-T); Beck Depression Inventory-II (BDI-II); Clinical Global Impression–Improvement (CGI-I). - Common therapy factors: Credibility and Expectancy Scale for GAD (CES-GAD); Nijmegen Motivation List (NML); Therapist Rating Scale (TRS). Analysis: Intent-to-treat for pretest–posttest with missing posttest replaced by pretest for dropouts. Group×Time ANOVAs comparing each treatment to WL (Bonferroni p<.008) and between-group posttest CGI-I ANOVA. Within-group one-way ANOVAs for change. For long-term outcomes, hierarchical linear modeling (HLM) for growth curves from posttreatment to 24 months; slope comparisons between groups and against zero. Diagnostic remission defined as ADIS-IV CSR ≤3. Medication use analyzed dichotomously over time via HLM.

Sample and integrity: Interrater diagnostic agreement κ=.66 (all 83) and κ=.70 (final 65). Treatment integrity high: CBT 90.1%, AR 93.1%. Dropouts in first 12 weeks: CBT 2, AR 5, WL 0. Posttest vs WL: CBT superior to WL on CSR (F(1,41)=24.67, p<.001, partial η²=.38), PSWQ (F(1,41)=25.30, p<.001, partial η²=.38), WAQ-Som (F(1,40)=8.87, p=.005, partial η²=.18), and CGI-I (F(1,41)=13.87, p=.001, partial η²=.25). AR superior to WL only on CSR (F(1,40)=8.27, p=.006, partial η²=.17). Neither treatment outperformed WL on STAI-T or BDI-II. Within-group changes: Significant decreases on all measures in CBT and AR; WL showed significant decreases on CSR and WAQ-Som. Pretest–posttest partial η² effect sizes: CBT (CSR .76; PSWQ .74; WAQ-Som .61; STAI-T .55; BDI-II .55); AR (CSR .62; PSWQ .34; WAQ-Som .37; STAI-T .36; BDI-II .49); WL (CSR .39; PSWQ .03; WAQ-Som .23; STAI-T .16; BDI-II .10). CBT vs AR at posttreatment: Repeated-measures ANOVAs showed similar improvements across specific symptom measures; CGI-I favored CBT (F(1,62)=6.05, p<.05, partial η²=.09). Long-term (post to 24 months): Between-group comparisons of slopes were nonsignificant across measures (CBT and AR equivalent). Against zero slope: CBT showed continued improvement on PSWQ (slope −1.98, t(30)=−3.99, p<.001), STAI-T (slope −1.33, t(30)=−2.64, p<.05), and CGI-I (slope −0.14, t(30)=−2.28, p<.05). AR showed no slopes significantly different from zero (maintenance of gains without clear further improvement). Diagnostic remission (CSR ≤3): CBT: 70% post, 76% at 6 months, 84% at 12 months, 77% at 24 months. AR: 55% post, 70% at 6 months, 68% at 12 months, 61% at 24 months. Chi-square comparisons at each time point were nonsignificant. Medication: Percent using medication decreased over time in CBT (58% pretreatment to 36% at 24 months) and fluctuated in AR (58% pretreatment to 46% at 24 months); HLM slopes did not differ between groups and were not significantly different from zero within groups. Common therapy factors: CES-GAD, NML, TRS showed no between-group differences after session 3, indicating comparable credibility, expectancy, motivation, and therapist characteristics across modalities.

Findings partially supported the hypotheses. Both treatments were intended to outperform WL; CBT clearly did so on multiple domains (GAD severity, pathological worry, somatic symptoms, global improvement), whereas AR showed limited superiority over WL (GAD severity only), potentially influenced by unusually large WL improvements. Direct CBT vs AR comparisons suggested broadly similar short-term and long-term outcomes across specific symptom measures, with CBT showing greater global improvement at posttreatment. Over follow-up, CBT uniquely demonstrated continued gains in worry, trait anxiety, and global improvement, consistent with a model positing lasting changes as patients increasingly engage with uncertainty across real-world contexts. AR participants maintained gains without evidence of relapse. Overall, direct comparisons indicate equivalence, but comparisons to WL and to no-change trajectories provide stronger support for the efficacy and sustained benefits of CBT. Clinically, results endorse both modalities for GAD, with CBT offering potential added value in continued posttreatment improvement.

This trial demonstrates that both a GAD-specific CBT targeting intolerance of uncertainty and applied relaxation are efficacious for adults with primary GAD. In direct comparisons, outcomes are largely similar; however, CBT shows broader superiority over wait-list and uniquely supports continued improvement in worry, trait anxiety, and global clinical status over two years. The study contributes evidence for the sustainability and breadth of benefits of mechanism-focused CBT. Future research should refine CBT to address interactions between intolerance of uncertainty and fear of anxiety, test with larger samples, include multiple therapists, and assess therapist competency to enhance generalizability and interpretability of treatment effects.

Potential allegiance effects given the development of the CBT by the first author; efforts to mitigate (blinded assessors, non-CBT-trained therapist, dual supervision) may not eliminate bias. Single therapist treated most participants, limiting generalizability and preventing assessment of therapist effects; therapist competency was not measured (only integrity), which constrains interpretation. Diagnostic reliability concerns: ADIS-IV administered only to patients with GAD on the MINI may introduce demand characteristics; assessors knew inclusion criteria (≥2-point CSR difference), potentially influencing ratings. Sample size relatively small for some analyses. Unusually large improvements in the wait-list group may have reduced the ability to detect superiority of active treatments on certain outcomes.