Intolerance of uncertainty as a mediator of reductions in worry in a cognitive behavioral treatment program for generalized anxiety disorder

Intolerance of uncertainty (IU)—the tendency to react negatively to the unknown—is theorized to play a central role in the development and maintenance of anxiety, particularly generalized anxiety disorder (GAD). GAD is characterized by pervasive, uncontrollable worry about potential negative events. Cognitive models posit that high IU triggers maladaptive cognitive and behavioral responses (e.g., catastrophic interpretations, excessive information seeking, cognitive avoidance) that maintain worry and anxiety. While CBT can reduce IU, it is unclear whether decreases in IU drive decreases in worry during treatment, or whether reductions in worry lead to lower IU. The present study addresses this question by testing whether reductions in IU mediate subsequent reductions in worry during a CBT program for individuals with GAD, thereby evaluating IU as a mechanism of symptom change.

Prior work has linked IU to multiple anxiety disorders, with the most developed model focused on GAD. Treatment studies have shown that CBT programs—both IU-focused protocols and broader transdiagnostic approaches—reduce IU and symptoms such as worry and depression. Case-controlled and analogue studies suggest that reductions in IU may precede or predict reductions in worry, and experimental manipulations of IU impact worry levels, supporting a potential causal role. However, longitudinal mediation across the course of treatment, establishing temporal precedence of IU change leading to worry reduction, had not been firmly demonstrated prior to this study.

Design: Data were drawn from an open trial examining neural markers and treatment response to 10 sessions of CBT (NCT00947570). Symptom measures were collected at baseline and bi-weekly during treatment (sessions 2, 4, 6, 8, 10). All procedures were IRB-approved and participants consented. Participants: Twenty-eight adults (ages 18–55) from the community met DSM criteria for primary GAD (MINI interview). Inclusion required at least a high school education. Exclusions: psychotic disorder, bipolar I, organic mental disorder, recent substance dependence (past 12 months) or abuse (past month), psychotropic/anti-epileptic use within 6 weeks, heavy caffeine use (>6/day) or cigarette use, MRI contraindications. Comorbid depression allowed if not severe per QIDS-6 criteria. Intervention: Ten 1-hour individual sessions over 10–12 weeks using modules adapted from the CALM program. Generic modules included self-monitoring, psychoeducation, breathing retraining, relapse prevention; disorder-specific modules included exposures (including imaginal exposure to worry themes) and cognitive restructuring. IU was not an explicit target, though exercises could implicitly affect IU. Treatment fidelity was monitored; therapists (PhD-level) received top adherence and competence ratings. Assessments: Demographics; Diagnostic status via MINI. Worry: PSWQ-A (8 items; strong reliability, Cronbach’s α = 0.84–0.92). Intolerance of Uncertainty: IUS-12 derived from the 27-item IUS (Cronbach’s α = 0.92–0.94). Descriptive session means reported for IUS-12 and PSWQ-A at all timepoints. Analytic strategy: Repeated measures (Level 1) were nested within participants (Level 2). Intent-to-treat multilevel models in SPSS 18 handled missingness. Mediation was tested using lower-level (within-person) multilevel mediation with temporal precedence by lagging the mediator: IU at time t predicting worry at time t+1. The reverse mediation (worry mediating IU) was also tested. Indirect effects and 95% CIs were estimated via Prodclin; percent mediation calculated per Kenny et al. (2003).

- Symptom change over time: Both worry and IU decreased during treatment. • PSWQ-A: F(1,122) = 8.23, p < .01, ηρ² = .12. • IUS-12: F(1,114) = 51.69, p < .001, ηρ² = .41. - Mediation (IU mediating subsequent worry reduction): • Time → Worry (path c): B = −0.34, SE = 0.10, t = −3.36, p < .01. • Time → IU (path a): B = −1.68, SE = 0.37, t = −4.57, p = .001. • IU → Worry controlling for Time (path b): B = 0.12, SE = 0.03, t = 4.56, p < .001. • Time → Worry controlling for IU (path c′): B = −0.14, SE = 0.11, t = −1.22, p = .23. • Indirect effect (ab): −0.20; 95% CI [−0.35, −0.09]; significant. • Percent mediation: 59% of the Time → Worry effect accounted for by IU change. - Reverse mediation (Worry mediating IU reduction): • Time → IU (path c): B = −2.17, SE = 0.40, t = −5.35, p < .001. • Time → Worry (path a): B = −0.16, SE = 0.13, t = −0.92, p = .37. • Worry → IU controlling for Time (path b): B = −0.11, SE = 0.25, t = −0.46, p = .65. • Time → IU controlling for Worry (path c′): B = −2.20, SE = 0.41, t = −5.32, p < .001. • Indirect effect (ab): −0.16; 95% CI [−0.06, 0.12]; not significant; <1% mediation. - Descriptive means suggest steady declines in IUS-12 and modest declines in PSWQ-A across sessions.

The study tested whether reductions in intolerance of uncertainty (IU) drive decreases in worry during CBT for GAD. Longitudinal multilevel mediation with temporal precedence supported the hypothesized mechanism: decreases in IU significantly mediated subsequent reductions in worry, accounting for a majority of the time-related improvement in worry. The reverse pathway—worry mediating IU reductions—was not supported, suggesting IU changes are not simply concomitants of worry change. These findings align with cognitive models positioning IU as a central vulnerability that sustains worry, and extend prior analogue and experimental work by demonstrating within-treatment temporal precedence and mediation in a clinical GAD sample. Clinically, even a transdiagnostic CBT protocol not explicitly targeting IU produced meaningful IU reductions, implying that exposure to uncertainty and cognitive restructuring may naturally decrease negative beliefs about uncertainty. Incorporating explicit IU-focused strategies (e.g., targeted exposure to uncertain situations, challenging beliefs that uncertainty is inherently threatening) may further enhance outcomes for worry-related disorders.

This study provides longitudinal evidence that reductions in intolerance of uncertainty mediate subsequent reductions in worry during CBT for individuals with GAD, supporting IU as a key mechanism of change. The results suggest that targeting IU within CBT may be an effective strategy to reduce pathological worry. Future research should compare IU-specific versus transdiagnostic approaches on process of change, identify patient-level moderators (e.g., baseline IU), include larger and more diverse samples with control conditions, evaluate durability beyond acute treatment, and examine whether IU reductions mediate improvements across other anxiety-related conditions.

- Sample selection constrained by neuroimaging trial criteria (e.g., exclusions for medications/substances), limiting generalizability. - Modest effect size for PSWQ-A change; the transdiagnostic protocol may have suboptimally targeted worry relative to GAD-specific CBT. - Use of abbreviated PSWQ-A may have reduced sensitivity to change compared to full PSWQ. - Small sample size (N=28) and open trial design without a control group; cannot attribute changes solely to treatment components versus nonspecific factors or time. - Assessments occurred every other session only during the acute phase; finer-grained temporal dynamics and post-acute trajectories are unknown. - Findings may not generalize to populations with comorbidities or medication use excluded here; demographic composition further limits external validity.