Hope Predicts Long-Term Well-Being Following Cognitive Behavioral Therapy for Anxiety and Related Disorders

Abstract

Objective:

Prior research indicates that cognitive behavioral therapy (CBT) results in moderate-to-large increases in hope and well-being, though long-term treatment outcomes are less clear. This study examined whether gains in these outcomes were maintained up to 36 months after CBT and whether hope – an established change mechanism during treatment – was associated with sustained well-being during long-term follow-up when controlling for post-treatment anxiety and depression.

Method:

A subset of 69 participants [M_age= 30.29 (SD = 9.94) yrs, 62.32% female, 89.86% White, 89.85% Non-Hispanic/Latinx] with anxiety disorders from a randomized equivalence trial of transdiagnostic and single-diagnosis CBT protocols were re-consented into a long-term follow-up study and completed assessments every 6 months across 3 years.

Results:

Standardized mean gain scores and multilevel modeling (MLM) revealed no significant differences in well-being or hope between post-treatment and follow-up timepoints in either condition. MLM indicated that higher hope at the within-person (standardized γ₁₀ = 0.26, p < .001) and between-person levels (standardized γ₀₁ = 0.42, p < .001) were associated with greater well-being across long-term follow-up, whereas post-treatment anxiety and depression were not significant predictors.

Conclusion:

Findings suggest sustained increases in hope may help preserve gains in well-being up to three years after CBT.

Clinical or Methodological Significance of This Article:

These findings suggest that sustained hope may foster long-term well-being after CBT, even when considering post-treatment anxiety and depression. Focusing on therapeutic processes that promote well-being alongside symptom reduction can help researchers and clinicians address the full spectrum of patient functioning.

Cognitive-behavioral therapy (CBT) is an effective and efficacious intervention for anxiety and related disorders (Hofmann et al., 2012; Hoffmann & Smits, 2008). As such, it has been acknowledged as an empirically supported treatment for several anxiety-related disorders by the Society for Clinical Psychology, showing consistent evidence of clinical utility for panic disorder (PD), social anxiety disorder (SOC), generalized anxiety disorder (GAD), and obsessive-compulsive disorder (OCD; Society of Clinical Psychology, 2022). Nevertheless, research examining the long-term stability of treatment gains after CBT is relatively lacking, and few treatment studies include assessment periods beyond two years. Furthermore, there is increasing interest in understanding how CBT promotes positive functioning and well-being beyond the absence of mental illness symptoms. Thus, the present study aimed to examine whether improvements in well-being persist up to 36 months after CBT and to determine potential mechanisms through which these gains may be maintained.

Bolstering Well-being During Treatment

Research on psychotherapy outcomes has primarily focused on changes in symptoms of psychological disorders, consistent with the medical model of mental health, which tends to place less emphasis on positive functioning (Elkins, 2009). Though mental wellness is commonly equated with the absence of mental illness, mounting evidence suggests that mental illness and well-being are related but separate aspects of mental health rather than opposite ends of a single dimension (Keyes, 2005). In fact, evidence from a large nationally representative longitudinal study of well-being in the US found that Americans who did not meet criteria for a psychiatric disorder, but were nevertheless “languishing” in terms of their mental well-being (i.e., low in social, psychological, emotional, and well-being), generally showed comparable functional impairment to those who met criteria for mental illness alone (Keyes, 2005). This suggests a need to expand the purview of mental health care to include a greater emphasis on positive functioning. While efforts to operationalize positive functioning have lagged behind the development of taxonomies of mental illness, current conceptualizations of mental wellness incorporate aspects of emotional (e.g., positive affect and subjective well-being), psychological (e.g., a sense of meaning and purpose), and social (e.g., social connectedness and community involvement) well-being (Keyes, 2005).

Meta-analytic reviews have indicated that CBT is efficacious not only in reducing symptoms of anxiety and related disorders (i.e., emotional disorders; Bullis et al., 2019), but in improving aspects of positive functioning and quality of life (Kindred et al., 2022; Hofmann et al., 2014). CBT has been shown to bolster several forms of positive functioning, including savoring (upregulating/amplifying positive affect; Wilner Tirpak et al., 2019; Woodard et al., 2023), posttraumatic growth (Vrontaras et al, 2023), perceived control (Gallagher et al., 2013), optimism (Knaevelsrud et al., 2010), self-efficacy (Gallagher et al., 2013), and interpersonal skills like assertiveness (Dorstyn, Mathias, & Denson, 2011). Though less research has focused on how psychotherapy impacts well-being specifically, several studies have also found that CBT significantly improves well-being and quality of life for patients with anxiety disorders (Ginzburg et al., 2012; Miller et al., 2021; Telch et al., 1995).

Our own prior findings from a large randomized controlled trial (RCT) also suggest that CBT for anxiety and related disorders leads to broad improvement in mental well-being in addition to sustained symptom improvement. This clinical equivalence trial examined whether a transdiagnostic CBT, The Unified Protocol for The Transdiagnostic Treatment of Emotional Disorders, (UP; Barlow et al., 2017) was as efficacious as gold-standard, single-diagnosis CBT protocols (SDPs; Barlow et al., 2017). To develop the UP, common elements from established CBT protocols for individual anxiety disorders were distilled into a single unified protocol targeting their shared underlying mechanisms, allowing for simultaneous and efficient treatment. Initial findings supported the equivalence of the UP and SDPs, since both of these types of CBT produced comparable, statistically significant reductions in emotional disorder symptoms that were still present up to 36 months post-treatment (Barlow, et al 2017; Bullis et al, 2023). Furthermore, improvements in anxiety during CBT were associated with gains in well-being, which were greatest during the latter half of treatment following initial symptom reduction (Gallagher, Long, Richardson, et al., 2020). Across transdiagnostic and single-disorder CBTs, we observed comparable, moderate-to-large gains in emotional, psychological, and social well-being from pre- to post-treatment, which were maintained up to 6 months (Gallagher, Long, Richardson, et al., 2020). Despite this initial evidence that CBT can promote well-being alongside symptom reduction, more needs to be understood about how CBT impacts well-being and whether CBT-related gains in well-being are also maintained over longer periods of time.

The Role of Hope in Promoting Well-being

In order to understand how CBT impacts broader well-being, it is essential to identify mechanisms that drive and sustain treatment gains in this area. For instance, hopeful thinking is linked to superior mental health and may be a “common factor” that promotes positive outcomes across different psychotherapy modalities (Snyder & Taylor, 2000), including CBT. Snyder et al. (1991) described hope as a cognitive motivational resource involving positive expectations of achieving desired goals or outcomes, even in the face of obstacles. In this model, hope is defined as the reciprocal relationship between pathways thinking, the ability to generate routes towards goals, and agency thinking, the perceived willpower to utilize those pathways. Hope has been found to play an important role in various areas of psychological functioning, including mental wellness, resilience, and motivation (Lee & Gallagher, 2017). Whether a person aims to maintain fulfilling relationships, engage in meaningful activities, or maximize their occupational or academic performance, hopeful thinking predicts success in these endeavors (Lee & Gallagher, 2017).

Hopeful thinking is also associated with fewer symptoms of emotional disorders (Alarcon et al., 2013; Arnau, 2018) and may even protect against adverse outcomes of stressful or traumatic events, such as the development of PTSD (Gallagher, Long, & Phillips, 2020). Interventions designed to foster hope have been shown to increase hopeful thinking and reduce depression (Cheavens & Whitted, 2023; Salamanca-Balen et al., 2021; Sinaga et al., 2020). Further, there is growing evidence that hope acts as a change mechanism during CBT, particularly for patients with anxiety and trauma-related disorders (Long, 2022; Gallagher, Long, Richardson, et al., 2020). Several studies now indicate that these CBTs lead to greater levels of hope, which is associated with reductions in anxiety and PTSD symptoms, as well as better treatment outcomes (Long, 2022; Gallagher, Long, Richardson, et al., 2020). Thus, hopeful thinking may also help maintain treatment gains in well-being (Beck et al., 2021).

Present Study

Prior research (e.g., Chaves et al., 2017; Miller et al., 2021), including our own RCT (Gallagher, Long, Richardson, et al., 2020; Gallagher, Phillips, et al., 2020), suggests that CBT can also lead to improvements in and positive functioning, including hope and overall well-being (e.g., Chaves et al., 2017; Gallagher, Long, Richardson, et al., 2020; Miller et al., 2021). We previously observed moderate-to-large increases in hope and broader well-being across both single disorder and transdiagnostic CBT protocols in our RCT, which were maintained up to six months post-treatment (see Gallagher, Long, Richardson, et al., 2020; Gallagher, Phillips, et al., 2020). However, research on long-term outcomes of CBT for anxiety disorders is generally limited, and factors supporting the long-term durability of well-being following treatment are still not well understood.

The present study utilizes long-term follow-up data from our randomized clinical equivalence trial of CBTs for anxiety and related disorders, as described above. This study explores positive functioning following a course of CBT and examines whether higher hope predicts long-term well-being over and above the effects of residual anxiety and depression symptoms at post-treatment. Given our prior findings from this trial supporting the stability and equivalence of emotional disorder symptom reduction, we anticipated that gains in hope and well-being observed in the parent trial would be maintained across the three years following CBT. Furthermore, we hypothesized that greater hope – an established mechanism of change during treatment – would also act as a protective factor supporting sustained well-being across the follow-up period when controlling for post-treatment symptoms of anxiety and depression.

Method

Participants

The current sample (n = 69) is a subset of participants that were included in the parent randomized equivalence trial of CBTs for anxiety disorders and then enrolled in the long-term follow-up study. Participants were eligible for the parent trial if they were English-speaking, at least 18 years old, and met criteria for panic disorder with or without agoraphobia (PD/A), generalized anxiety disorder (GAD), obsessive-compulsive disorder (OCD), or social anxiety disorder (SOC) based on the Anxiety Disorders Interview Schedule (ADIS), a semi-structured clinical interview (Di Nardo et al., 1994). Please see Barlow et al., 2017 for full inclusion and exclusion criteria. To be eligible for the long-term follow-up phase of the study at the final assessment timepoint for the parent trial (12 months after treatment), participants had to meet additional criteria, including (1) randomization to one of the two treatment conditions (UP or SDP) (2) designation as a treatment completer after completing at least 75% of the total number of treatment sessions (i.e., 12 out of 16 sessions with the exception of those with PD, who were required to complete 9 out of 12 sessions), and (3) consenting to complete online assessments at four timepoints, as well as phone or in-person assessments at two timepoints during an additional two year follow-up period (i.e., two years after the final 12-month follow-up assessment within the parent trial, such that the final assessment would occur three years after treatment).

A total of 223 participants enrolled in the parent study and 179 were randomized to either the UP (n = 88) or a single-diagnosis protocol (n = 91). Eighty of those participants initially randomized to either the UP (n = 44) or a single-diagnosis protocol (SDP; n = 36) were later re-consented into the long-term follow-up study (Bullis et al., 2023). Of note, prior results from the long-term follow-up study (Bullis et al., 2023) found no significant differences between participants who enrolled in the extended follow-up study (n = 80) and the original intent-to-treat sample (n = 179) in terms of age, race, treatment condition, SDP received, or principal diagnosis, as well as any markers of clinical severity or symptomology assessed (including principal diagnosis clinical severity rating, both clinician-rated and self-reported symptoms of anxiety or depression, and functional impairment) at the point of enrollment (i.e., 12 months following treatment).

The current sample consists of 69 participants (UP: n = 37; SDP: n = 32) from the long-term follow-up study who completed measures of well-being and hope at post-treatment and at least 1 timepoint during the follow-up period (see table 1). Participants’ age averaged 30.29 years (SD = 9.94), 62.32% self-identified as female, and 89.86% identified as White. Participants generally reported high educational attainment, with 76.92% receiving at least a bachelor’s degree. Principal diagnoses for the current sample were as follows: 34.78% GAD, 33.33% SOC, 17.39% PD/A, and 14.49% OCD.

Table 1.

Descriptive Statistics and Standardized Mean Gain Scores for Treatment Outcomes

Variable	N	Mean	Std. Deviation	ESsg Post-to-Follow-up (95% CI)
Anxiety Post-Tx	69	4.80	3.13	--
Depression Post-Tx	69	3.04	3.49	--
Hope Post-Tx	69	34.25	9.24	--
Hope 6 MFU	59	34.02	10.11	−0.04 (−0.22, 0.14)
Hope 12 MFU	67	35.06	8.35	0.08 (−0.10, 0.27)
Hope 18 MFU	47	34.83	9.99	0.01 (−0.22, 0.25)
Hope 24 MFU	48	34.91	9.41	0.06 (−0.20, 0.33)
Hope 30 MFU	45	35.42	8.53	0.12 (−0.14, 0.38)
Hope 36 MFU	52	35.44	9.23	0.07 (−0.15, 0.30)
Well-being Post-Tx	69	45.70	14.30	--
Well-being 6 MFU	59	45.34	14.45	−0.05 (−0.22, 0.12)
Well-being 12 MFU	67	46.34	14.66	0.01 (−0.14, 0.17)
Well-being 18 MFU	47	45.02	14.50	−0.06 (−0.29, 0.17)
Well-being 24 MFU	48	47.35	13.45	−0.04 (−0.28, 0.20)
Well-being 30 MFU	45	44.89	14.94	−0.05 (−0.27, 0.17)
Well-being 36 MFU	52	45.44	15.82	−0.11 (−0.30, 0.08)

Note: Post-Tx = Post-treatment, MFU = month follow-up. ESsg refers to the standardized mean gain scores for post-treatment and follow-up differences.

Procedures

For a complete description of procedures from the parent study, please see (Barlow et al., 2017). Participants from the parent trial that were designated as treatment completers were informed of their eligibility for the long-term follow-up study either during or shortly after completing the 12-month follow-up. Participants were enrolled in the long-term follow-up study between September of 2013- September of 2016 and completed assessments at 6-month intervals across the course of three years. As such, data collection was completed by September of 2018. Approval for this study was obtained from the Boston University Charles River Campus Institutional Review Board.

Treatments

Participants in the SDP condition received individual therapy using the protocol that matched their primary diagnosis. Those with GAD were offered 16 sessions of Mastery of Your Anxiety and Worry, second edition (Zinbarg et al., 2006); those with SOC were offered 16 sessions of Managing Social Anxiety: A Cognitive-Behavioral Therapy Approach, second edition (Hope et al., 2010); those with PD/A were offered 12 sessions of Mastery of Your Anxiety and Panic, fourth edition (Craske & Barlow, 2007); and those with OCD were offered 16 sessions of Exposure and Response (Ritual) Prevention for Obsessive-Compulsive Disorder, second edition (Foa et al., 2012). These empirically-supported protocols feature cognitive behavioral strategies for addressing anxiety-related pathology, such as psychoeducation, goal-setting, cognitive restructuring/thinking skills, and behavioral strategies including exposure to fear, anxiety, and/or worry cues related to their primary diagnosis, along with weekly homework assignments focused on reinforcing learning and implementation of skills from treatment sessions.

Participants in the UP condition received 12–16 sessions of individual therapy with the UP (Barlow, 2018; Barlow et al., 2010) depending on their principal diagnosis so that the number of sessions offered matched the number associated with the corresponding SDP for their principal diagnosis. The UP follows a modular format and offers CBT strategies similar to those from the SDPs, but with greater focus on reactions to the experience of emotions themselves (e.g., autonomic arousal) as opposed to the situations or contexts that trigger them. In addition to earlier modules centered on motivational enhancement/goal-setting and psychoeducation for understanding emotions, the UP features 5 core modules focused on (1) mindful emotion awareness, (2) cognitive flexibility, (3) identifying and preventing patterns of emotion avoidance, (4) increasing awareness and tolerance of emotion-related physical sensations, and (5) interoceptive and situational emotion exposures.

Assessments

During the parent trial, participants completed various self-report measures, including measures for anxiety and depression symptoms (Overall Anxiety Severity and Impairment Scale, Norman et al., 2011; Overall Depression Severity and Impairment Scale, Bentley et al., 2014), overall well-being (Mental Health Continuum—Short Form; Keyes, 2005), and level of hope (State Hope Scale; Snyder et al., 1996) at baseline, after every four treatment sessions, at post-treatment, and 6 and 12 months. During the long-term follow-up study, participants were asked to complete periodic self-report assessments at 6-month intervals (i.e., at 18-, 24-, 30-, and 36- months) and were compensated $25 at each timepoint.

Measures

Hope

The State Hope Scale (SHS; Snyder et al., 1996) was used to assess levels of hope. The scale consists of 6 items (e.g., “I can think of many ways to reach my current goals”, “right now, I see myself as being pretty successful”), and responses are rated on an 8-point Likert-type scale, with scores ranging from 1 (Definitely False) to 8 (Definitely True). Total scores range from 6 to 48, with higher scores indicating higher hope. This measure has shown strong internal consistency and test-retest reliability (Snyder et al., 1996). Internal consistency in the present study was α = .93 at post-treatment.

Well-being

The Mental Health Continuum–Short Form (MHC-SF; Keyes, 2005) was used to assess emotional, psychological, and social well-being. The scale includes 14 items and participants are asked to rate how often they felt a certain way (e.g., “satisfied with life”, “that your life had a sense of direction or meaning to it”, “that you had warm and trusting relationships with others”) during the past month using a 6-point Likert-type scale ranging from 0 (Never) to 5 (Every Day). Total scores range from 0–75, with higher scores indicating higher well-being. This measure has shown high internal consistency and good test-retest reliability (Lamers et al., 2011). In the present sample, Cronbach’s alpha value for the total scale was α = .95 at post-treatment.

Anxiety

The Overall Anxiety Severity and Impairment Scale (OASIS; Norman et al., 2011) was used to assess anxiety severity and functional impairment. This scale consists of 5 items (e.g., “In the past week, how often have you felt anxious?”, “In the past week, how often did you avoid situations, places, objects, or activities because of anxiety or fear?”), and different response options are rated on a 5-point Likert scale ranging from 0 to 4. Total scores range from 0 to 20, with higher scores indicating higher anxiety. This measure has demonstrated strong internal consistency, test-retest reliability, and convergent and divergent validity (Norman et al., 2011). Internal consistency in the present study was α = .85 at post-treatment.

Depression

The Overall Depression Severity and Impairment Scale (ODSIS; Bentley et al., 2014) was used to assess depression severity and functional impairment. This scale consists of 5 items (e.g., “In the past week, how often have you felt depressed?”, “In the past week, how often did you have difficulty engaging in or being interested in activities you normally enjoy because of depression”) and different response options are rated on a 5-point Likert scale, with scores ranging from 0 to 4. Total scores range from 0 to 20, with higher scores indicting greater depression symptoms. This measure has also shown excellent internal consistency, test-retest reliability, and convergent and divergent validity (Bentley et al., 2014). Internal consistency in the present study was α = .95 at post-treatment.

Data Analysis

Preliminary analyses were conducted to evaluate the representativeness of the analytic sample and to inform the analytic strategy. To assess sample representativeness, we conducted independent-samples t-tests comparing participants included in the present analyses to those from the parent trial who completed post-treatment assessments but were not included due to missing follow-up data on the outcomes of interest (i.e., hope and well-being). Additionally, within the current analytic sample (n = 69), we evaluated whether post-treatment outcomes differed between participants originally randomized to the Unified Protocol (UP = 1) versus a single-disorder protocol (SDP = 0) using independent-samples t-tests. Because no significant between-group differences were found on post-treatment measures of self-reported anxiety, depression, well-being, or hope, participants from both conditions were combined into a single cognitive-behavioral therapy (CBT) group for subsequent analyses. All between-group comparisons and descriptive statistics were conducted using SPSS Version 29.

Standardized mean gain scores (ESsg) with 95% confidence were calculated with the assistance of MPlus version 7 to assess differences in mean hope and well-being total scores from post-treatment to 6, 12, 18, 24, 30, and 36 months (effect sizes are considered statistically significant if the 95% CI does not include zero). In addition, mixed effects multilevel modeling was conducted in R to examine changes in hope and well-being across the long-term follow-up period.¹ In these multilevel models, follow-up timepoints (Level 1) were nested within individuals (Level 2). Model building was carried out in the following manner. First, two null models (with a random intercept only) were fitted to determine the proportion of within-person versus between-person variance in hope (ICC = .70) and well-being (ICC = .75). Next, time was included in each model as a level 1 predictor with a fixed slope. Likelihood ratio tests comparing nested models indicated that including a random slope for time significantly improved model fit in both cases (hope: χ²(2) = 20.18, p < .001; well-being: χ²(2) = 13.46, p = .001), so the models with a random slope for time were retained:

$${\mathrm{Outcome}}_{\mathrm{ij}}={\gamma }_{00}+{\gamma }_{10}{\mathrm{Time}}_{\mathrm{ij}}\mathrm{}+{\mathrm{u}}_{0\mathrm{j}}\mathrm{}+{\mathrm{u}}_{1\mathrm{j}\mathrm{}}{\mathrm{Time}}_{\mathrm{ij}}\mathrm{}+{\mathrm{\epsilon }}_{\mathrm{ij}}$$

As a confirmatory step, treatment condition (level 2) and its interaction with time were added to evaluate whether trajectories of hope and well-being differed by treatment condition; no significant effects were expected based on prior findings of treatment equivalence.

$${\mathrm{Outcome}}_{\mathrm{ij}}\mathrm{}={\mathrm{\gamma }}_{00}\mathrm{}+{\mathrm{\gamma }}_{10}\mathrm{}{\mathrm{Time}}_{\mathrm{ij}\mathrm{}}+{\mathrm{\gamma }}_{01}\mathrm{}\mathrm{Treatment}{\mathrm{Condition}}_{\mathrm{j}\mathrm{}}+\mathrm{}{\mathrm{\gamma }}_{11\mathrm{}}({\mathrm{Time}}_{\mathrm{ij}}\times \mathrm{Treatment}{\mathrm{Condition}}_{\mathrm{j}})+{\mathrm{u}}_{0\mathrm{j}}\mathrm{}+{\mathrm{u}}_{1\mathrm{j}\mathrm{}}{\mathrm{Time}}_{\mathrm{ij}}\mathrm{}+{\mathrm{\epsilon }}_{\mathrm{ij}\mathrm{}}$$

A final mixed-effects multilevel model was constructed to examine the relationship between hope and well-being during long-term follow-up. The model included both within-person hope (person-centered to capture deviations from an individual’s mean, included at Level 1) and between-person hope (the individual’s average hope during follow-up, included at Level 2) as predictors. In this model, the outcome variable, well-being, was limited to follow-up timepoints to avoid redundancy, as post-treatment well-being was included as a covariate, along with post-treatment anxiety and depression. This approach allowed us to disentangle the effects of within-person variability in hope from between-person differences while accounting for initial differences in well-being and residual symptoms of emotional disorders at post-treatment.

$$\begin{gathered} \mathrm{Well}-{\mathrm{being}}_{\mathrm{ij}\mathrm{}}={\mathrm{\gamma }}_{00}\mathrm{}+{\mathrm{\gamma }}_{10}\mathrm{Hope}\_{\mathrm{Within}}_{\mathrm{ij}\mathrm{}}+{\mathrm{\gamma }}_{01}\mathrm{Hope}\_{\mathrm{Between}}_{\mathrm{j}}\mathrm{}+{\mathrm{\gamma }}_{02\mathrm{}}\mathrm{Post}-\mathrm{Treatment}\_\mathrm{Well}-{\mathrm{being}}_{\mathrm{j}} \\ +{\mathrm{\gamma }}_{03}\mathrm{}\mathrm{Post}-\mathrm{Treatment}\_{\mathrm{Anxiety}}_{\mathrm{j}}\mathrm{}+{\mathrm{\gamma }}_{04}\mathrm{}\mathrm{Post}-\mathrm{Treatment}\_{\mathrm{Depression}}_{\mathrm{j}\mathrm{}}+{\mathrm{u}}_{0\mathrm{j}}\mathrm{}+{\mathrm{\epsilon }}_{\mathrm{ij}\mathrm{}} \end{gathered}$$

Results

Preliminary Analyses

To assess the representativeness of the current analytic sample, we conducted exploratory comparisons with participants from the parent trial who completed post-treatment assessments but were not included in the present analyses due to missing corresponding follow-up data on the outcomes of interest (i.e., hope and well-being). Independent-samples t-tests indicated no significant differences between groups on post-treatment measures on anxiety (t(125) = 0.35, p = .724), depression (t₍₁₂₅₎ = 0.35, p = .729), hope (t₍₁₁₇₎ = –0.42, p = .674), or well-being (t₍₁₁₆₎ = –1.09, p = .276), suggesting the current analytic sample from the long-term follow-up study was broadly representative of the larger post-treatment sample from the parent trial.

Independent samples t-tests were then used to assess group differences in post-treatment well-being, hope, anxiety, and depression between participants who received the UP versus SDP in the current analytic sample. Post-treatment levels of well-being (t₍₆₇₎ = 0.69, p = .49) and hope (t₍₆₇₎ = 1.12, p = .27) did not significantly differ between the UP and SDP treatment groups. Furthermore, these groups did not show statistically significant differences in anxiety symptoms (t₍₆₇₎ = −1.52, p = .13) or depression (t₍₆₇₎ = −.46, p = .65) at post-treatment. Thus, the SDP and UP groups were collapsed into a single sample for the following analyses.

Long-term Maintenance of Post-Treatment Well-being and Hope Across Follow-up

Standardized mean gain scores (ESsg) with 95% confidence intervals were used to examine potential differences in hope and well-being at each timepoint during long-term follow-up to deduce whether treatment gains were maintained across the 36-month follow-up period. There were no significant differences observed in mean hope total scores between post-treatment and 6 month follow-up (ESsg = −0.04, 95% CI [−0.22, 0.14]), 12 month follow-up (ESsg = 0.08, 95% CI [−0.10, 0.27]), 18 month follow-up (ESsg = 0.01, 95% CI [−0.22, 0.25], 24 month follow-up (ESsg = 0.06, 95% CI [−0.20, 0.33]), 30 month follow-up (ESsg = 0.12, 95% CI [−0.14, 0.38]), or 36 month follow-up (ESsg = 0.07, 95% CI [−0.15, 0.30]), and all standardized effect sizes for observed differences were small (See table 1). Similarly, there were no significant differences observed in mean well-being total scores between post-treatment and 6 month follow-up (ESsg = −0.05, 95% CI [−0.22, 0.12]), 12 month follow-up (ESsg = 0.01, 95% CI [−0.14, 0.17]), 18 month follow-up (ESsg = −0.06, 95% CI [−0.29, 0.17], 24 month follow-up (ESsg = −0.04, 95% CI [−0.28, 0.20]), 30 month follow-up (ESsg = −0.05, 95% CI [−0.27, 0.17]), or 36 month follow-up (ESsg = −0.11, 95% CI [−0.30, 0.08]), and all standardized effect sizes for observed differences were also small. Thus, these findings suggest that gains in hope and well-being were maintained up to 3 years after CBT was completed.

In addition, mixed effects multilevel modeling was also used to examine long-term stability of hope and well-being following CBT treatment (See Table 2). A random intercept and random slope for time were specified in each model to account for individual variability in post-treatment levels and rates of change in these outcomes across the follow-up period. In the first model (1a) examining potential changes in hope during long-term follow-up, the fixed effect of time was not statistically significant (γ₁₀ = 0.20, 95% CI [− 0.14, 0.54], p = 0.26) and the standardized effect size bordered on zero (standardized γ₁₀ = 0.04, 95% CI of [- 0.03, 0.12], providing no evidence of linear change across the 36-month follow-up period. The random effects suggested more substantial between-person variability in post-treatment hope (as indicated by the variance of the intercept, τ₀₀=66.91) and smaller variability in the slope measuring change over the follow-up period (as indicated by the variance in the slope, τ₁₁= 1.02). The correlation between random intercepts and slopes was −0.35, suggesting that individuals with higher initial hope were less likely to show reductions in hope over time, with a moderate effect size. The ICC of .76 indicated that 76% of the variance in hope was attributable to between-person differences and 24% to within-person variability. The model explained 76% of the variance in hope when accounting for both fixed and random effects (conditional R² = .76), whereas the fixed effects alone explained a negligible amount of variance (marginal R² = .00).

Table 2.

Fixed Effects from Mixed Effects Models Examining Long-term Treatment Outcomes

Predictors	Estimate	95% CI	SE	t(df)	Std Est	95% CI	p
Hope Model (1a)
Intercept	34.19	32.10, 36.27	1.06	32.27 (69.67)			<.001
Time	0.20	−0.14, 0.54	0.17	1.14 (65.66)	0.04	−0.03, 0.12	.26
Hope Model (1b)
Intercept	32.40	29.40, 35.39	1.52	21.28 (68.90)			<.001
Time	0.37	−0.12, 0.86	0.25	1.49 (62.33)	0.04	−0.03, 0.12	.14
Treatment	3.35	−0.74, 7.45	2.08	1.61 (69.47)	0.13	−0.07, 0.34	.11
Time × Treatment	−0.33	−1.01, 0.35	0.34	−0.96 (65.01)	−0.04	−0.11 – 0.04	.34
Well-being Model (2a)
Intercept	45.73	42.50, 48.96	1.64	27.84 (69.23)			<.001
Time	−0.19	−0.67, 0.29	0.24	−0.76 (59.67)	−0.03	−0.09, 0.04	.45
Well-being Model (2b)
Intercept	43.92	39.23, 48.61	2.39	18.41 (68.52)			<.001
Time	0.06	−0.63, 0.75	0.35	0.17 (56.67)	−0.03	−0.09, 0.04	.86
Treatment	3.41	−3.00, 9.83	3.26	1.05 (69.03)	0.07	−0.14, 0.29	.30
Time × Treatment	−0.48	1.43, 0.48	0.49	−0.98 (59.31)	−0.03	−0.10, 0.03	.33
Well-being Model (3)
Intercept	−12.02	−22.19, −1.84	5.16	−2.32 (66.36)			.02
Within-person Hope	0.88	0.73, 1.03	0.08	11.70 (246.88)	0.26	0.22, 0.30	<.001
Between-person Hope	0.75	0.49, 1.01	0.13	5.69 (64.46)	0.42	0.28, 0.57	<.001
PT Well-being	0.59	0.43, 0.74	0.08	7.50 (65.01)	0.58	0.43, 0.73	<.001
PT Anxiety	0.57	−0.09 – 1.24	0.34	1.70 (64.40)	0.12	−0.02, 0.26	.09
PT Depression	0.44	−0.10 – 0.99	0.28	1.60 (63.03)	0.11	−0.02, 0.24	.11

Note. PT = post-treatment, Std Est = standardized estimate. Bolded p-values indicate statistical significance at p < .05. Time is modeled as a continuous predictor. Between-Person Hope reflects each individual’s average hope across the long-term follow-up period (i.e., person-mean hope). Within-Person Hope represents fluctuations in an individual’s hope at a given timepoint relative to their own average (i.e., person-centered hope). Model 1a marginal R² = 0.00, conditional R² = 0.76; Model 1b marginal R² = .02, conditional R² = 0.76; For Model 2a marginal R² = 0.00, conditional R² = 0.79; Model 2b marginal R² = .01, conditional R² = 0.79; Model 3 marginal R² = 0.70, conditional R² = 0.85.

Furthermore, the subsequent model examining change in hope as a function of treatment condition found no evidence that the type of CBT received influenced hope during long-term follow-up. The fixed effect of treatment condition confirmed that there were no significant differences in hope at post-treatment between participants who received the transdiagnostic versus single-disorder CBT protocols (γ₀₁ = 3.35, 95% CI [–0.74, 7.45], standardized γ₀₁ = 0.13, 95% CI [–0.07, 0.34], p = .11), and there were no significant differences in the trajectory of hope over time between treatment conditions, as indicated by the non-significant time × treatment interaction (γ₁₁ = –0.33, 95% CI [–1.01, 0.35], standardized γ₁₁ = –0.04, 95% CI [–0.11, 0.04], p = .34). Further, there was no meaningful change in the slope of time (γ₁₀ = 0.37, 95% CI [–0.12, 0.86], p = .14; standardized γ₁₀ = 0.04, 95% CI [–0.03, 0.12]) or in model R² (marginal R² = .02, conditional R² = 0.76), indicating no added explanatory value from including treatment condition in the model.

In the model examining potential changes in well-being following CBT, the fixed effect of time was not statistically significant (γ₁₀ = −0.19, 95% CI [− 0.67, 0.29], p = .45) and the standardized effect size bordered on zero (standardized γ₁₀ = −0.03, 95% CI of [- 0.09, 0.04], suggesting there was no linear change in well-being across the 36-month follow-up period. The random effects suggested substantial between-person variability in post-treatment well-being (as indicated by the variance of the intercept, τ₀₀=163.09) and smaller variability in the slope measuring change across the follow-up period (reflected by variance in the slope, τ₁₁= 1.81). The correlation between random intercepts and slopes was −0.15, suggesting that individuals with higher initial well-being were slightly less likely to show reductions in well-being over time. The ICC of .79 indicated that 79% of the variance in well-being could be attributed to between-person differences and 21% to within-person variability. The model explained 79% of the variance in well-being when accounting for both fixed and random effects (conditional R² = .79), whereas the fixed effects alone explained a negligible amount of variance (marginal R² = .00).

Further, the model examining change in well-being as a function of treatment condition reflected no notable differences in well-being between conditions during the follow-up period. The fixed effect of treatment condition confirmed that there were no initial differences in well-being at post-treatment between participants who received the transdiagnostic versus single-disorder CBT protocols (γ₀₁ = 3.41, 95% CI [−3.00, 9.83], standardized γ₀₁ = 0.07, 95% CI of [−0.14, 0.29], p = .30), and there were no significant differences in change in well-being across the follow-up period treatment between these groups (γ₁₁ = −0.48, 95% CI [−1.43, 0.48], standardized γ₀₁ = −0.03, 95% CI of [−0.10, 0.03], p = .33). Further, adding treatment condition to the model did not result in meaningful changes in R² (marginal R² = .01, conditional R² = .79) and the fixed effect of time was negligible (γ₁₀ = 0.06, 95% CI [−0.63, 0.75], standardized γ₁₀ = −0.03, 95% CI of [−0.10, 0.03], p = .33).

Predictors of Well-being During Follow-up

Afterward, both within-person hope (which was person-centered and included at level one) and between-person hope (i.e., the average for the individual during follow-up, which was included at level two) were examined as predictors of well-being during long-term follow-up when controlling for post-treatment individual differences in well-being, anxiety, and depression symptoms (See Table 2). Consistent with study hypotheses, between-person differences in hope emerged as a significant positive predictor (γ₀₁ = 0.75, 95% CI [0.49, 1.01], standardized γ₀₁ = 0.42, 95% CI [0.28, 0.57], p < .001), indicating that individuals with higher average hope during the long-term follow-up reported significantly higher well-being. Similarly, within-person fluctuations in hope significantly predicted well-being (γ₁₀ = 0.88, 95% CI [0.73, 1.03], standardized γ₁₀ = 0.26, 95% CI [0.22, 0.33], p < .001), suggesting that increases in hope relative to an individual’s mean were associated with improved well-being. Unsurprisingly, post-treatment well-being was also a strong positive predictor (γ₀₂ = 0.59, 95% CI [0.43, 0.74], standardized γ₀₂ = 0.58, 95% CI [0.43, 0.73], p < 0.001), reflecting the substantial stability in well-being from post-treatment to follow-up as previously observed, whereas post-treatment anxiety (γ₀₃ ​= 0.57, 95% CI [−0.09, 1.24], standardized γ₀₃​ = 0.12, 95% CI [−0.02, 0.26], p = 0.09) and depression (γ₀₄ = 0.44, 95% CI [−0.10, 0.99], standardized γ₀₄ = 0.11, 95% CI [−0.10, 0.99], p = 0.11) symptoms were not significant predictors of long-term well-being. The random effects revealed between-person variability in well-being (τ₀₀= 31.50) and some residual variance within individuals over timepoints (σ² = 33.27). The intraclass correlation coefficient (ICC) was .49, indicating that 49% of the total variance in well-being was attributable to differences between individuals, while the remaining 51% was due to within-person variability. Together, fixed and random effects explained 85% of the variance in well-being (conditional R² = .85), with the fixed effects alone accounting for 70% (marginal R² = .70). Thus, hope uniquely predicted well-being across the course of long-term follow-up, even when controlling for post-treatment well-being, anxiety, and depression symptoms, underscoring its robust and independent contribution to the long-term maintenance of gains in well-being.

Discussion

While CBT has been established as an empirically supported treatment for anxiety, depressive, and related disorders (or emotional disorders; Bullis et al., 2019), less is understood about how CBT impacts well-being and positive functioning, particularly in the long-term. Previously, we observed moderate-to-large improvements in well-being and hope during a large randomized equivalence trial of CBTs for emotional disorders, which were maintained during the initial months following treatment (Gallagher, Long, Richardson, et al., 2020; Gallagher, Phillips, et al., 2020). Our present findings suggest that maintenance of these gains extends up to 36 months after CBT is completed. No statistically significant differences were found in levels of hope or well-being between post-treatment or any of the follow-up timepoints, which occurred every six months over three years. Furthermore, residual symptoms of anxiety and depression following treatment were not predictive of long-term well-being independently of hope, which was a significant predictor of well-being three years after completing CBT. This suggests that hopeful thinking is an important factor to monitor and target for maximizing treatment gains in well-being.

Our prior work has indicated that reductions in anxiety during CBT are associated with gains in well-being, and when examining the timing of these changes, the majority of improvement in well-being appears to occur later in treatment, following initial improvements in anxiety symptoms (Gallagher, Long, & Phillips, 2020). Thus, while symptoms of emotional disorders are related to well-being, these facets of mental health should be considered distinct treatment outcomes. Our present findings also support this conclusion, given that post-treatment levels of anxiety and depression were not statistically significant predictors of long-term well-being after completing CBT.

In contrast, hope emerged as a predictor of long-term well-being following CBT. These results are consistent with the broader literature indicating that hope is robustly related to greater mental health and positive functioning (Alarcon et al., 2013; Gallagher, Long, & Phillips, 2020). Decades after hope was proposed as a “common factor,” or a shared mechanism of change across different treatment modalities (Frank, 1961; Snyder & Taylor, 2000), several empirical studies show that hopeful thinking increases during treatment and is associated with improvements in symptoms of anxiety and related disorders (Capone & Cameron, 2020; Gallagher, Long, Richardson, et al., 2020; Gallagher & Resick, 2012; Gilman et al., 2012; Irving et al., 2004). Our findings suggest that treatment gains in hopeful thinking are enduring, and may also help maintain well-being long after CBT is completed.

There are many ways through which hopeful thinking may lead to greater well-being as a consequence of treatment. CBT commonly begins with a discussion of patients’ goals for treatment, which promotes hopeful thinking (Snyder et al., 2000). Many patients enter treatment not only to “feel better” or less anxious but because they want to feel more fulfilled and are trying to cultivate a life worth living. Hope may also be an important motivating factor that helps patients approach aversive situations and/or emotional experiences during exposure-based CBT (Hayes et al., 2007). Pathways and agency thinking are intimately involved in building and progressing through an exposure hierarchy. Developing a tolerance of intense emotions by successfully confronting and withstanding anxiety-provoking situations during exposures engenders hopeful thinking that may generalize to other goals after treatment is completed, leading to broader well-being. In the termination stage of CBT, patients are typically encouraged to reflect on treatment gains, which reinforces agency thinking, and to consider ways to address challenges they may face in the future as a method of relapse prevention, which constitutes pathways thinking (Cheavens et al., 2006). These efforts may help position patients to maintain their treatment gains and continue to respond adaptively to challenges in a way conducive to their mental health more broadly.

This study’s strengths include the randomized clinical trial design with an extended follow-up period. Treatment outcome studies examining long-term outcomes of CBT for anxiety disorders are generally scarce and typically focus on symptoms of psychopathology. The present study incorporated a longer follow-up period with a larger sample than most (van Dis et al., 2020), so that well-being was regularly assessed every six months up to three years post-treatment. However, given the design of the original clinical trial, we could not compare the long-term stability of well-being after CBT to the natural temporal course of well-being in the absence of intervention since participants in the waitlist control condition initiated treatment immediately after their post-waitlist assessment. Furthermore, the present study focused on both intra-individual (i.e., within person) and interindividual (i.e., between persons) patterns of hope and well-being in response to CBT. This approach allows us to disentangle the effects of individual differences in stable, trait-like hope (i.e., between-person hope) from momentary, state-like fluctuations from an individual’s average (i.e., within person hope) on well-being, providing a nuanced understanding of how hope relates to long-term well-being following CBT. Future research should also examine other possible factors or mechanisms impacting hope and long-term well-being after treatment, such as physical health and resilience.

Furthermore, a limitation of the both the present study and parent trial is that the sample lacked racial/ethnic diversity and significant variation in education level. The present results supporting the long-term stability of gains in broader mental well-being and positive functioning following CBT are promising, but must be replicated in more diverse samples and contexts. Still, these findings add to growing evidence since the initial equivalence trial of the UP suggesting that this form of transdiagnostic CBT can flexibly applied to a range of clinical presentations in diverse global contexts (Barlow & Long, 2023). The UP has been translated into more than a dozen languages and implemented in numerous countries, with results supporting its efficacy for treating anxiety and depressive symptoms (Cassiello-Robbins et al., 2020). Nevertheless, expanding contextual and cultural adaptations of CBTs like the UP is a vital direction of future research, with promising initial efforts focused on exposure to armed conflict in Colombia (Castro-Camacho et al., 2019, 2023) tailoring group therapy for unhoused persons both in the US and abroad (Marín et al., 2021; Sauer-Zavala et al., 2019), and addressing minority stress related to sexual orientation (Pachankis et al., 2015).

Still, many do not have the privilege of accessing a full course of CBT. As such, large scale training (Unified Protocol Institute, 2025) and research efforts have aimed to support dissemination and implementation of the UP within hospitals and community healthcare centers (Farchione et al., 2024). A more recent line of work has focused on the development and evaluation of the internet-based UP, along with digital tools supporting delivery of the UP via telehealth within both private and public mental healthcare (Belanger et al, 2023; Long & Farchione, 2023; Timulak et al., 2022). Relatedly, an ongoing RCT is currently evaluating several digital CBTs, including the UP, within large US healthcare delivery systems, using machine learning to better understand for whom CBT works, with implications for personalization of CBT and expanding access to care (Clinical Trials Identifier: NCT05567640). Future research should prioritize evaluating and implementing CBTs such as the UP in routine clinical settings, with more diverse populations, and under circumstances more closely resembling real-world practice (Farchione et al., 2024).

Conclusion

In sum, the findings from the present study indicate that gains in hope and well-being are maintained over 36 months following CBT. The results also suggest that hopeful thinking may play a vital role in maintaining mental well-being. In contrast, residual emotional disorder symptoms following treatment have less impact on the stability of well-being during the three years following CBT. While broader well-being is a commonly neglected treatment outcome, there may be benefits to prioritizing well-being both during treatment and after treatment ends, in line with studies showing that directly targeting well-being may prevent relapse among individuals with remitted depression (Fava, 2016) and may be more effective at alleviating residual symptoms among individuals with remitted emotional disorders after a successful course of treatment compared to supplemental CBT (Fava et al., 1998). Attending to both changes in symptoms and adaptive processes that promote broader well-being will allow researchers and clinicians to better engage with the full spectrum of patient improvement and functioning.