Single-case Experimental Design Evaluation of Repetitive Negative Thinking-Focused Acceptance and Commitment Therapy in GAD with Couple-related Worry

Full reference: Ruiz, F.J., García-Beltrán, D.M., Monroy-Cifuentes, A., & Suárez-Falcón, J.C. (2019). Single-case experimental design evaluation of repetitive negative thinking-focused acceptance and commitment therapy in generalized anxiety disorder with couple-related worry. International Journal of Psychology and Psychological Therapy, 19(3), 261-276.

Study type: Single-case experimental design (SCED) with delayed multiple baseline.

Background and objectives

Generalized Anxiety Disorder (GAD) is characterized by excessive, chronic, and generalized worry affecting multiple life domains (Barlow, 2002). Among young adults with GAD, couple-related worry is particularly frequent and constitutes a significant problem that tends to provoke dysfunctions in intimate relationships (Gould & Edelstein, 2010). Excessive worry about relationship stability and future is associated with lower levels of couple satisfaction and increased risk of relationship dissolution (Hunt, Issalidis, & Andrews, 2002).

In recent years, Acceptance and Commitment Therapy (ACT) has demonstrated efficacy in treating GAD (Lass et al., 2011; Wetherall et al., 2011). Recently, researchers have proposed that Repetitive Negative Thinking (RNT) functions as a central transdiagnostic mechanism, and that an ACT protocol specifically focused on reducing and altering the function of RNT in response to primary triggers could be particularly effective (Ruiz, Riaño-Hernández, Suárez-Falcón, & Luciano, 2016). The objective of the present study was to evaluate the efficacy of a 3-session RNT-focused ACT protocol in young adults suffering from severe GAD with couple-related worry as the primary worry domain. A single-case experimental design with delayed multiple baseline across three young adult participants was employed.

Method

Participants

The study included N = 3 participants (2 females, 1 male) recruited through social media advertisements. Inclusion criteria were: (a) being older than 18 years of age, (b) experiencing thoughts, memories, and worries for at least 12 months, (c) reporting significant interference in at least two life domains, with couple relationship being the most affected area, and (d) presenting extremely severe scores on depression or anxiety according to the Depression and Anxiety Stress Scale-21. Exclusion criteria were: (a) showing high suicidal ideation, (b) reporting substance abuse, and (c) reporting presence of intimate partner violence.

Demographic characteristics were as follows:

• P1: female, 20 years old, heterosexual orientation, undergraduate student, problem duration: 6 years, couple duration: 2 years, no previous psychological treatment. Diagnostic categories: GAD, depression (relapse).
• P2: male, 24 years old, homosexual orientation, undergraduate student, problem duration: 6 years, couple duration: 3 years, no previous psychological treatment. Diagnostic categories: GAD, depression, agoraphobia, obsessive-compulsive disorder.
• P3: female, 22 years old, heterosexual orientation, graduate, problem duration: 4 years, couple duration: 1.5 years, previous psychological treatment (Brief). Diagnostic categories: GAD, manic episode (past).

Design

A delayed multiple-baseline design across participants was implemented with three young adults with GAD. The independent variable was the staggered introduction of the 3-session RNT-focused ACT protocol. Following recent guidelines (Kratochwill & Levin, 2014), a minimum of 5 data points during baseline was used. The protocol was implemented once per week in an individual format on a weekly basis. Following each treatment session, baseline stability was assessed through the Theil-Sen slope evaluation (Sen, 1968). No statistically significant trends were found during baseline. Blinding procedures were not implemented because the study involved a single intervention, and dependent measures were obtained through automatic online emails. During the 12-week follow-up, participants provided data every two weeks.

Intervention

The protocol consisted of three sessions of variable duration: the first session lasted approximately 90 minutes, while the second and third sessions lasted approximately 60 minutes each. The protocol was based on the Relational Frame Theory (RFT) definition of psychological flexibility and the formation of self-as-context (Luciano et al., 2012; Törneke et al., 2016). Specific aims were: (a) to develop the ability to discriminate primary worry/rumination triggers through a hierarchical relationship; (b) to reduce their discriminative avoidance functions; (c) to promote experiencing engagement with worry and related experiential avoidance strategies; and (d) to identify the RNT process and defusion training.

Session 1 content (duration: ~90 minutes): (a) present intervention rationale, (b) identify hierarchical triggers for couple-related worry/rumination, (c) promote realization of the counterproductive nature of related experiential avoidance strategies, and (d) identify RNT process and defusion training.

Session 2 content (duration: ~60 minutes): (a) review advances in RNT disengagement and valued action engagement, (b) provide multiple examples of training in RNT trigger identification and defusion from them, and (c) identify valued actions in place of commitment to RNT.

Session 3 content (duration: ~60 minutes): (a) review advances in RNT disengagement and valued action engagement, (b) provide additional experiential exercises, and (c) plan committed action and select the intervention.

During intervention, participants received three audio files (approximately 30 minutes each) to practice during sessions on a daily basis. The audio file was designed to develop the skill of noticing differences between engaging in RNT and letting primary triggers drive behavior. Following the logic of the "go-around exercise," participants were encouraged to engage with RNT for several minutes while noticing the thinking process, thus allowing them to disengage from the process at each moment. Subsequently, participants were invited to repeat the exercise allowing triggers to activate during imagination of themselves engaging in valued actions. Participants were provided with three audio files at the end of the second session of exercise training: (a) the centering/defusion exercise, (b) the free exposure exercise, and (c) the "daydreaming" / worry exercise with a values exercise that summarized the content of that session.

Outcome measures

Primary outcome measures:

• Experiences in Close Relationships-Anxiety (ECR-A): a 7-point self-report scale (7 = strongly agree; 1 = strongly disagree) designed to assess two adult romantic attachment styles. Anxiety (i.e., fear of rejection and abandonment) was the primary focus. The Spanish version showed adequate internal consistency, with Cronbach's alpha between 0.82 and 0.88. In this study, the mean score on the ECR-A in a nonclinical sample was 65.28 (SD = 13.6).

• Penn State Worry Questionnaire-11 (PSWQ-11): an 11-item self-report instrument, scored on a 4-point Likert scale (1 = never true; 4 = always true) designed to assess the permanent and nonspecific degree of worry that characterizes GAD. A reduced 11-item version was used (as recommended by Sandín et al., 2009) because the original PSWQ includes reversed items that are difficult for Spanish speakers to understand. The internal consistency of the PSWQ-11 is excellent, and the test showed excellent content validity and discriminant validity. The PSWQ-11 possesses good psychometric properties. A score of 37/33 was deemed adequate for identifying severe GAD, whereas a score of 32/33 was adequate for moderate GAD.

Secondary outcome measures:

• General Health Questionnaire-12 (GHQ-12): a 12-item Likert-type scale (4 points) frequently used to screen for psychological disorders. Respondents were asked about the degree to which they have recently experienced psychological symptoms. The Likert scoring method was used in this study, with scores ranging from 1 to 3 for each response option. The Spanish version possesses good psychometric properties in Colombia (Ruiz, García-Beltrán, & Suárez-Falcón, 2017). Cronbach's alpha was 0.91 in both nonclinical and clinical samples from Colombia, with mean scores of 11.87 (SD = 7.47) and 16.54 (SD = 7.86) respectively.

• Cognitive Fusion Questionnaire (CFQ): a 7-item questionnaire, scored on a 7-point Likert scale (7 = always; 1 = never) consisting of sentences describing instances of cognitive fusion. The Spanish version by Ruiz et al. (2017) showed psychometric properties similar to the original version (alpha of 0.93 in general population). The mean score in a Colombian clinical sample was 27.80 (SD = 9.51).

• Valued Living Questionnaire (VQ): a 10-item questionnaire, scored on a 6-point Likert scale (6 = completely true; 0 = not at all true) designed to assess valued living during the past week. The VQ has two subscales: Progress (i.e., clarity of what is personally important and experiencing) and Obstruction (avoidance of valued living). The 10 items were slightly modified to specify values in the couple area. This version of the VQ can be obtained at https://osf.io/Servn/.

Data analysis

To explore the presence of statistically significant trends during baseline on outcome variables (ECR-A and PSWQ), the Theil-Sen slope was computed (Sen, 1968; Vannest, Parker, Davis, Soares, & Smith, 2012) at the end of five weeks of baseline. The Theil-Sen slope is a nonparametric estimator of the linear regression slope that does not assume any particular data distribution. Results were graphed, and subsequently statistical analyses for single-case experimental designs (SCED) were conducted.

Following a follow-up analysis of single-case experimental designs (Parker & Vannest, 2012), results were first graphed and subsequently statistical analyses of SCED were conducted. In general, data showed baselines with significantly different levels and/or improvement trends after intervention initiation. The statistical analysis was focused on all baseline data and the last three follow-up points (see similar rationale in Au et al., 2017; Parker & Vannest, 2012) to take advantage of facilitating comparison between the current study and the usual effects reached in typical clinical trials.

Two different but complementary statistical methods were selected and computed: (a) the Bayesian JZS+AR hypothesis testing for single-case experimental designs (de Vries & Morey, 2013, 2015); and (b) the standardized mean difference effect size statistic for single-case designs (Hedges, Pustejovsky, & Shadish, 2012, 2013). The Bayesian JZS+AR test models the serial dependence typical of single-case designs with an autoregressive [AR(1)] model. It provides a Bayes factor (BF), which quantifies the relative evidence in the data for the hypothesis of no intervention effect (i.e., the true mean in the baseline phase equals the true mean in the intervention phase: BF > 1) and for the hypothesis of intervention effect (i.e., the true means in both phases differ: BF < 1). All analyses regarding the JZS+AR model were conducted in the BayesSingleSub R package (de Vries & Morey, 2015) and correspond to follow-up group Cohen's d.

Results

A highly consistent pattern of results was observed. No statistically significant trends were found during baseline. Visual inspection showed that the protocol was effective in decreasing scores on both ECR-A and PSWQ measures in all participants.

Table 4. Results in the JZS+AR Standardized Mean Difference and Analysis for each Participant and Measure

Measure	P1	P2	P3
ECR-Anxiety	δ = 5.93; BF = 13.70; 95% CI [1.183, 10.04]
	(range: .00-.013)
PSWQ (pathological worry)	δ = 3.19; BF = .003; 95% CI [.027, -.112]
GHQ-12 (psychological distress)	δ = 2.90; BF = .078; 95% CI [.495, 2.118]
CFQ (cognitive fusion)	δ = 2.54; BF = 2.31; 95% CI [2.43, 6.09]
VQ-Progress (valued couple living)	δ = 1.36; BF = 3.34; 95% CI [.084, 1.03]
	δ = .036; BF = 2.18; 95% CI [.704, .704]
VQ-Obstruction (valued couple living)	δ = 2.74; BF = 1.32; 95% CI [1.50, .409]
	δ = .070; BF = .475; 95% CI [-.409, -]

Note: ECR = Experiences in Close Relationships; PSWQ = Penn State Worry Questionnaire; GHQ-12 = General Health Questionnaire-12; CFQ = Cognitive Fusion Questionnaire; VQ = Valuing Questionnaire.

Overall, participants reported to their therapists at the last follow-up that they were making better decisions in relation to their couples and other life areas. Table 5 shows that Hedges' g effect sizes for SCED were extremely large for primary outcome measures (ECR-Anxiety: g = 5.93; PSWQ: g = 3.19). Effect sizes were also very large for cognitive fusion (CFQ: g = 2.54) and for obstruction of couple-related values (VQ-Obstruction: g = 2.12), while for psychological distress (GHQ-12: g = 1.17) and couple-related values progress (VQ-Progress: g = 1.36), they were large.

Table 5. Results per Measure in the Hedges' g Standardized Mean Difference for SCED

Measure	g	df	95% CI
ECR-Anxiety	5.93	4.39	[1.183, 10.04]
PSWQ (pathological worry)	3.19	20	[1.29, 2.05]
GHQ-12 (psychological distress)	1.17	20	[1.29, 2.05]
CFQ (cognitive fusion)	2.54	65	[1.96, 4.12]
VQ-Progress (valued couple living)	1.36	21	[.47, .225]
VQ-Obstruction (valued couple living)	2.12	36	[1.05, 3.29]

Note: ECR = Experiences in Close Relationships; PSWQ = Penn State Worry Questionnaire; GHQ-12 = General Health Questionnaire-12; CFQ = Cognitive Fusion Questionnaire; VQ = Valuing Questionnaire.

Discussion and conclusions

Young adults with GAD typically present high levels of couple-related worry that tends to provoke dysfunctions in this and other areas such as work, friendships, and family. Previous research has shown that GAD is associated with couple problems, being the emotional disorder that most affects relationship quality (Whisman, 2007; Zaider, Heimberg, & Iida, 2010). Furthermore, research has shown that individuals with GAD and couple-related worries present greater treatment difficulties and poorer outcomes than other patients with GAD without couple problems (Durham et al., 1997; Priest, 2013; Whisman & Baucom, 2012).

Following previous studies that tested RNT-focused ACT protocols (Ruiz et al., 2016; Ruiz, Flórez, et al., 2018), the present study analyzed the effect of a 3-session RNT-focused ACT protocol in young adults suffering from severe GAD with couple-related worry as the main worry domain. A single-case experimental design with delayed multiple baseline was used with three young adults who participated in this study. Participants showed the following characteristics: (a) they were immersed in thoughts, memories, and worries for at least 12 months; (b) the entanglement interfered in at least 2 life domains, with couple being the most affected; (c) they presented extremely severe scores on the DASS-21 scale; (d) they presented GAD according to the MINI. Participants showed stable levels of couple-related worry and pathological worry during baseline. Subsequently, participants received a 3-session RNT-focused ACT protocol.

Tests of the SCED study indicated that there was evidence of intervention effect in all three participants on both outcome measures (i.e., ECR-Anxiety and PSWQ). Regarding process measures, all three participants showed evidence of intervention effect on cognitive fusion (CFQ) and couple-related values obstruction (VQ-Obstruction), where 2 of the 3 participants did so on psychological distress (P1 and P2) and couple-related values progress (P1 and P3), with the standardized mean difference effect size for SCED being extremely large for primary outcome measures (ECR-Anxiety: g = 5.93; PSWQ: g = 3.19) and smaller for process measures (CFQ: g = 2.54; VQ-Progress: g = 1.36; VQ-Obstruction: g = 2.12). Importantly, these effect sizes are in the same metric as group designs Cohen's d, which allows comparison of the current findings with empirical evidence provided by group treatment designs.

The effect sizes obtained by the 3-session ACT protocol are very large and are in the range of those obtained by RNT-focused ACT protocols in previous studies (Ruiz et al., 2016; Ruiz, Flórez, et al., 2018). This contrasts with data from a recent meta-analysis that yielded a weighted effect size of d = 0.80 in GAD treatment (Cuijpers, Cristea, Karyotaki, Reijnders, & Huibers, 2016) and d = 1.81 in GAD worry reduction according to the PSWQ (Hanrahan, Field, Jones, & Davey, 2013). Following Ruiz et al. (2016), the large effects found in this study could be due to three main features: (a) closed-loop feedback of the three angles was used to promote psychological flexibility; (b) emphasis on identifying and nonjudgmentally contacting RNT triggers and allowing triggers to be while choosing to commit to valued actions; (c) focus on disrupting the first and most pervasive reaction to triggers (i.e., worry/rumination), which usually leads to engagement in additional experiential avoidance strategies in an attempt to reduce distress; and (d) focus on disrupting the reinforcement cycle of this process. Although the study results are very promising, some limitations are worth noting.

Although significant advances have been made in recent years regarding the statistical analysis of SCED, there is still no consensus about what is the most adequate statistical test for this type of data. Indeed, influential authors (e.g., Manolov, Gast, Perdices, & Evans, 2014) have documented the outcome of several statistical procedures. Accordingly, two different but complementary statistical methods were selected: (a) the Bayesian JZS+AR hypothesis testing for single-case experimental designs (de Vries & Morey, 2013, 2015); and (b) the standardized mean difference effect size statistic for single-case designs (Hedges, Pustejovsky, & Shadish, 2012, 2013). The standardized mean difference effect size for SCED (g) proposed by Hedges et al. (2012) shares the same metric as the typical Cohen's d used in group designs, which has the advantage of facilitating direct comparison and integration of results obtained with both types of experimental design. This effect size measure has a formal mathematical development that requires at least three cases for computation, and assumes normal distributions of outcomes and lack of trends for single-case design. The Bayesian JZS+AR statistical test for SCED (de Vries & Morey, 2013, 2015) and the standardized mean difference effect size statistic for single-case designs was computed at the end of five weeks of baseline. Since no statistically significant trends were found during baseline, they represent baseline data that showed significant and important improvement levels after intervention initiation. The statistical analysis focused on all baseline data and the last three follow-up points (see similar rationale in Au et al., 2017; Parker & Vannest, 2012) to take advantage of facilitating comparison between the current study and the usual effects reached in typical clinical trials.

In conclusion, the present study constitutes an initial and promising step in the analysis of brief RNT-focused ACT protocols for the treatment of emotional disorders. Future studies need to conduct randomized controlled trials to compare the effect of the ACT protocol versus wait-list conditions or brief versions of empirically established treatments. Due to strict inclusion criteria, the number of participants recruited was minimal for conducting a multiple-baseline design. We wanted to test the efficacy of the brief ACT protocol in participants who, according to empirical evidence, would be relatively difficult to treat in view of the high level of emotional symptoms, the presence of comorbidity, and difficulties in the couple area. Accordingly, this study can be seen as a preliminary exploration of the potential efficacy of brief RNT-focused ACT protocols in individuals with GAD who present couple-related worry as the main problem. Future studies should test the efficacy of the intervention in a larger number of participants.

Significance and Contribution

This study contributes significantly to treatment literature for generalized anxiety disorder, demonstrating that a three-session protocol (approximately 210 minutes total) can produce clinically relevant changes and very large effect sizes in young adults with severe GAD, particularly when couple-related worry constitutes the primary worry domain. The systematic approach to hierarchical identification of repetitive negative thinking triggers, development of discrimination of avoidance functions, cognitive defusion, and engagement in valued committed actions provides empirical demonstration of brief clinical interventions for disorders where repetitive negative thinking is a central transdiagnostic mechanism.