Effectiveness of a transdiagnostic emotion regulation training in an at-risk sample: a randomized-controlled trial of group-based training versus self-help bibliotherapy
Abstract
Objectives
Subclinical psychotic, depression, and anxiety symptoms form a transdiagnostic ‘at-risk state’ for the development of mental disorders. Emotion regulation has been identified as a transdiagnostic factor relevant to the formation of these symptoms that can be successfully addressed in clinical interventions. Here, we tested whether a group-based emotion regulation training would be effective in reducing distress and at preventing the transition to mental disorders in an at-risk sample.
Methods
Participants with distressing subclinical psychotic, depression, or anxiety symptoms (n = 138) were randomly allocated to either the 8-week group-based affect regulation training (ART; Springer, New York) or an 8-week self-help bibliotherapy (BT). They underwent biweekly measurements during the intervention, as well as at a six- and 12-month follow-up. In an exploratory analysis, we tested whether the ART would be superior to BT in preventing the transition to any mental disorder at 12-month follow-up. We also tested for differences in trajectories of psychopathology and emotion regulation (via questionnaires) and emotion regulation in daily life (via the experience-sampling method).
Results
Participants in the ART condition showed a greater improvement of emotion regulation in daily life than those with BT, but the ART was not superior over BT in preventing the transition to mental disorders. There were significant longitudinal reductions from pre- to post-intervention for general psychopathology and symptoms but no superiority of the ART over BT.
Conclusions
Despite its efficacy in improving emotion regulation skills, the ART does not produce effects on psychopathology that justify its recommendation over self-help approaches.
Practitioner Points
- A group-based emotion regulation training effectively enhances emotion regulation skills.
- The training is not superior over a self-help bibliotherapy with regards to the prevention of mental disorders, the reduction of general psychopathology, and the reduction of symptoms of psychosis, depression, and anxiety.
- The results do not justify a recommendation of a group-based emotion regulation training over self-help approaches in at-risk populations.
- Flexible approaches, such as blended care, should be tested in prevention.
EFFICACY OF A TRANSDIAGNOSTIC EMOTION REGULATION TRAINING IN AN AT-RISK SAMPLE: A RANDOMIZED-CONTROLLED TRIAL OF GROUP-BASED TRAINING VERSUS SELF-HELP BIBLIOTHERAPY
There is compelling evidence that symptoms of psychosis, depression, and anxiety lie on a continuum from low frequency and minor distress beneath diagnostic threshold to high frequency and severe distress in mental disorders (Lewinsohn et al., 2000; Van Os et al., 2009). Symptoms below diagnostic threshold have been found to be common with an estimated lifetime prevalence of 6% for psychosis symptoms (McGrath et al., 2015), 17% for depression symptoms (Regeer et al., 2006), and a 3-year prevalence of 11% for anxiety symptoms (Bosman et al., 2019). Furthermore, these subthreshold symptoms have been shown to co-occur, influence each other (Stochl et al., 2015; van Os, 2013), and to be associated with a heightened risk to transition into various mental disorders (Hartmann et al., 2019). Subthreshold depression symptoms, for example, have been associated with a heightened risk for depression and psychotic disorders (adjusted prevalence ratios: 3.71 for depression and 1.74 for psychotic disorders). Beyond that, the likelihood to develop a mental disorder increases substantially if people experience additional subthreshold symptoms (Scott et al., 2021). This has led researchers to suggest that subclinical psychotic, depression, and anxiety symptoms form a transdiagnostic at-risk state (e.g., McGorry et al., 2018). Research also indicates that a range of common, transdiagnostic factors underlie the formation of this at-risk state (Lynch et al., 2021). Consequently, there is a compelling argument to develop broad and comprehensive early interventions that address these transdiagnostic factors and potentially prevent mental disorders (Guloksuz & van Os, 2018).
One transdiagnostic factor that has been studied extensively over the last decades is emotion regulation (cf. Cludius et al., 2020), which has been defined to comprise (1) identifying the need to regulate, (2) selecting a regulation strategy, and (3) implementing the selected strategy (Gross, 2015). Numerous studies point towards emotion dysregulation in individuals with mental disorders (Lincoln et al., 2022). For example, individuals with psychoses have been found to use rumination, self-blaming, and suppression more frequently and cognitive reappraisal less frequently than healthy controls (Ludwig et al., 2019). Similarly, major depression has been found to be associated with more rumination, avoidance, and suppression and less cognitive reappraisal, acceptance, and problem-solving (Visted et al., 2018). Finally, individuals with anxiety disorders have been reported to have less clarity about their emotions and to use less acceptance but more expressive suppression than healthy controls (Cisler & Olatunji, 2012; Dryman & Heimberg, 2018). Similar patterns of emotion dysregulation have been associated with subclinical psychotic (Chapman et al., 2019; Osborne et al., 2017), depression, and anxiety symptoms (Schäfer et al., 2017). Thus, there is sound evidence that emotion regulation represents a transdiagnostic factor associated with psychosis, depression, and anxiety.
Meanwhile, several interventions for different target groups have been developed to address emotion regulation, of which many have been found to show significant effects both on emotion regulation and on a broad range of psychopathological symptoms (Lincoln et al., 2022). One of these is the affect regulation training (ART; Berking & Whitley, 2014) that was designed as a transdiagnostic intervention that can be applied in subclinical populations (Berking & Lukas, 2015). The ART is delivered in a group-based format and aims to convey an adequate understanding of and a non-judgmental stance towards emotions, acceptance, self-compassion, and strategies to change emotions. As a complementary element to regular CBT, the ART was effective in improving emotion regulation and reducing depression in inpatients (Berking et al., 2013). As a stand-alone intervention, it was superior to a waitlist in reducing symptoms of depression, which was mediated via an improvement of emotion regulation skills (Berking et al., 2019). Finally, there is preliminary evidence that the ART improves emotion regulation skills in subclinical populations (Berking et al., 2010; Buruck et al., 2016). However, so far it has not been tested whether the ART reduces subclinical general psychopathology or symptoms of psychosis, depression, and anxiety or prevents transition to a mental disorder.
In the present study, we therefore tested the efficacy of the ART in an at-risk sample with distressing symptoms of psychosis, depression, or anxiety by comparing it with a control-condition in form of an unguided self-help bibliotherapy (BT) on emotion regulation. To date, health-care systems usually do not fund psychotherapy for individuals below diagnostic threshold. Individuals in at-risk states therefore mostly depend on self-help approaches, such as BT. Indeed, self-help BT represents a cost-effective, easily accessible intervention that has been shown to be effective in reducing symptoms of depression, anxiety, and psychosis (for a meta-analysis, see Den Boer et al., 2004; for a review, see Gualano et al., 2017; Hazell et al., 2018). Nevertheless, group-based approaches have been found to be more effective than self-help BT in individuals with depression (SMDs .34–.59; for a network meta-analysis, see Cuijpers et al., 2019). Thus, a BT on emotion regulation controls for a self-led engagement with emotion regulation and allows to test whether the additional effort associated with a group training on emotion regulation is also justified in prevention.
We hypothesized that the ART would be more effective than BT in reducing general psychopathology over the 8-week intervention period and at six- and 12-months follow-ups and in preventing the transition to mental disorders at 12-month follow-up (primary outcomes). We also expected the ART to be superior to BT in reducing symptoms of psychosis, depression, and anxiety over the intervention period and at six- and 12-months follow-ups and in improving emotion regulation skills in daily life (secondary outcomes).
METHOD
Design and procedure
We used a two-armed randomized active-controlled design comparing the 8-week, group-based ART to BT (see Figure 1). The study was carried out between February 2018 and March 2020 at Universität Hamburg, Germany. We recruited participants via online platforms, newsletters, leaflets, and posters in public spaces in Hamburg. Participants were pre-screened for eligibility via an online-questionnaire, followed by face-to-face diagnostic assessments of in- and exclusion criteria and provision of informed consent. The study design is depicted in Figure 2. Before the intervention (T1), participants answered a baseline online questionnaire on psychopathology and additional demographic information and were assessed for level of functioning. This was followed by biweekly online assessments during the intervention (T2–T4), an online post-intervention assessment (T5), and online follow-up assessments at six (FU6) and 12 months (FU12). T1 and T5 also involved a one-week experience-sampling assessment. The FU12 further included a diagnostic assessment, an assessment of level of functioning, life events, psychological and pharmacological support, and an evaluation of the intervention. All participants received a compensation of 40 €, participants with a minimum of 75% of T5 ESM questions received additional 10 €. The study was approved by the local ethics committee of the Universität Hamburg (2018_148). This study was not pre-registered.
Participants
The exclusion criteria contained currently receiving psychological therapy or a diagnosis of a current Axis I or II disorder as diagnosed with the Structured Clinical Interview for DSM-IV (SCID; German version: Wittchen et al., 1997). Note that specific phobia was not an exclusion criterion due to its high prevalence and relatively minor distress (Comer et al., 2011). The inclusion criteria contained sufficient German language skills, an age between 18 and 65 years and the presence of clinically relevant, subthreshold psychopathology. Subthreshold psychopathology was defined as deviation from the norm in symptom frequency and associated distress in the domain of either (a) positive psychotic, or (b) negative psychotic, or (c) depression symptoms as assessed with the Community Assessment of Psychic Experiences (CAPE; Stefanis et al., 2002), or (d) anxiety, or phobic symptoms as measured with the Brief Symptom Inventory (BSI; Franke, 2000). There was no upper limit for symptom frequency and distress. The magnitude of deviation from the norm is defined in the measures section.
The estimation of the necessary sample size was based on a simulation study for continuous multilevel data by Maas and Hox (2005) who found a sample size of n = 100 at level two to enable accurate estimation of level two variances for an expected medium effect size. Reckoning with a dropout rate of 20%, we aimed to randomize a minimum of n = 125 participants. Participants were continuously sampled and were able to start when a study cohort was complete (n = 12 in each condition).
Measures
Primary outcomes
General psychopathology was assessed with the Global Severity Index (GSI) of the BSI (German version: Franke, 2000). This questionnaire consists of 53 items, which are rated on a five-point Likert scale (0 = not at all; 4 = extremely) and refers to the past seven days. The GSI reflects overall psychopathological distress and is calculated as an average score across all 53 items. It showed excellent internal consistencies in validation studies (Franke, 2000). Cut-off scores on the anxiety and phobic subdomain of the BSI were calculated as one standard deviation above the mean in a healthy norm sample (anxiety symptoms ≥ .68 or phobic symptoms ≥ .39; Franke, 2000).
Diagnoses were assessed at FU12 as any current or remitted Axis I disorder that had occurred during the past 12 months with the SCID (German version: Wittchen et al., 1997) by trained research assistants, who were blind to the allocated condition.
Secondary outcomes
Psychotic symptoms were assessed with the CAPE (Stefanis et al., 2002). The scale consists of 42 items that participants rate on a 4-point Likert scale regarding their frequency (1 = never; 4 = [nearly] always) and distress (1 = not distressed; 4 = very distressed). It contains the subscales of positive, negative, and depression symptoms. As measurements (T1–T5) were two weeks apart, we adapted the reference period to two weeks. The CAPE has demonstrated good psychometric quality (Konings et al., 2006; German version: Schlier et al., 2015) and to be an appropriate instrument to detect psychosis risk (Mossaheb et al., 2012). Cut-off scores on the CAPE were calculated based on the highest quartile of a broad international community sample (Lincoln et al., 2017). As the reference period was adjusted from four to two weeks, original weighted values for the highest quartile were reduced by .1 and transformed back to sum scores: frequency ≥ 31 and distress ≥ 24 for positive psychotic symptoms, frequency ≥ 31 and distress ≥ 26 for negative psychotic symptoms, or frequency ≥ 17 and distress ≥ 18 for depression symptoms.
Depression Symptoms were assessed with the Beck Depression Inventory-II (BDI-II; original version: Beck et al., 1996; German version: Kühner et al., 2007). Participants rate 21 items that refer to the past two weeks on a 4-point Likert scale (0 = e.g., not feeling sad at all; 3 = e.g., feeling unbearably sad). The German version of the BDI-II has good psychometric quality (Alexandrowicz et al., 2014; Kühner et al., 2007).
Anxiety symptoms were assessed with the short version of the State Trait Anxiety Inventory (STAI; original version: Spielberger et al., 1971; German version: Laux et al., 1981). The STAI consists of two subscales measuring state anxiety (STAI-S) and trait anxiety (STAI-T). Both scales provide good psychometric quality (Laux et al., 1981).
Adaptive emotion regulation was assessed with the Emotion Regulation Skills Questionnaire-27 (ERSQ-27; Berking & Znoj, 2008). Participants indicated the frequency of emotion regulation skills used over the past week on a Likert scale (0 = not at all; 4 = [almost] always). We calculated a total score as an average across all 27 items that is indicative of a general emotion regulation ability and has good psychometric quality (Berking & Znoj, 2008).
Furthermore, we used the Emotion Regulation Skills Questionnaire-Emotion Specific version (ERSQ-ES; Ebert et al., 2013), which assesses regulation strategies applied in response to several emotions. Here, we adapted the ERSQ-ES for the experience-sampling method (ESM). The ESM represents an ecologically valid and fine-grained approach to assess processes in daily life and has been identified as a particularly change-sensitive approach to evaluate treatments effects (Myin-Germeys et al., 2018). It was administered via movisensXS Version 1.5.0 (movisens GmbH, Karlsruhe, Germany) on participants' private or university-owned smartphones. Participants were provided with a detailed explanation and examples of items along with an instruction sheet. Over a period of six consecutive days, five times per day between 9 AM and 10 PM, participants were notified with signal contingent, semirandom time sampling (Wheeler & Reis, 1991) with minimum time gaps of 1.5 hr in between notifications. At each notification, participants were first asked to rate the intensity of each of a pre-defined list of negative emotions experienced just before the notification, then to select the emotion with the highest intensity and to indicate the use of regulatory strategies on a five-point Likert scale (0 = ‘does not apply at all’; 4 = ‘fully applies’) specific to the selected emotion. We used the average score across all emotion regulation strategies, over all emotions, which has shown good psychometric properties (Ebert et al., 2013).
Additional outcomes
We included measures on well-being, general emotion regulation skills and specific regulation strategies, level of functioning, and ESM data on emotions and symptoms (i.e., paranoid ideation; see Appendix S1). They are not depicted in the main document given their limited additional value to the study. Time spent daily with training materials was assessed at T2–T5 via the item: ‘How long have you trained your emotion regulation skills on the basis of the material provided?’. Response options were: 1 (‘I didn't do any exercises’), 2 (‘less than 5 minutes per day’), 3 (less than ‘10 minutes per day’), 4 (‘less than 15 minutes per day’), and 5 (‘more than 15 minutes per day’). At FU12, participants evaluated both interventions (ART, BT) in an open question format. The baseline ESM data have also been used in a publication on associations between adaptive emotion regulation and paranoid ideation (Wittkamp et al., 2021).
Affect regulation training
The ART is a manualized, group-based training that begins with psychoeducation on emotions, emotion generation, the function of emotions, and their neuroscientific and psychophysiological background, which is followed by six modules on different emotion regulation skills: (1) muscle and breathing relaxation, (2) mindfulness awareness and labelling of emotions, (3) acceptance of emotions, (4) compassionate self-support, (5) analysing emotions, and (6) modifying emotions. We administered the ART in groups of six to ten participants over eight weeks with weekly sessions of 150 min in close accordance with the ART manual (Berking & Whitley, 2014). Participants who missed a session were given the opportunity to catch up via phone or face-to-face meetings with the trainer who administered the group sessions. Participants were additionally motivated to practice at home using practice materials, such as audio-files and pocket calendars including short exercises. The trainer (M.W.) was educated and supervised in the ART by its developer, Matthias Berking, and accompanied by a research assistant. All sessions were videotaped for supervision.
Self-help bibliotherapy
Participants in the active control condition (BT) were provided the self-help book ‘Gefühle im Griff’ (feelings under control) by Barnow (2018). The book contains a psychoeducational section on emotions and emotion regulation. In a practical section, an eight-week training program is provided that covers one emotion regulation strategy per week. Strategies covered include cognitive reappraisal, acceptance, and problem-solving (recommended as adaptive strategies) on the one hand and rumination, suppression, and avoidance (categorized as maladaptive) on the other. Each chapter contains self-tests, worksheets, and various exercises, such as breathing relaxation, mindfulness, awareness, and physical exercises. Readers are encouraged to practice 20 minutes daily. Participants were instructed to start the program on a specific date that corresponded to the first group session of the ART condition in the respective cohort and received no additional support from the research team.
Randomization
Participants were randomly allocated to the ART or BT via a computer-generated list (www.sealedenvelope.com) that was based on a permuted block design with block lengths of four to ensure an even distribution to both conditions (ART, BT). Research staff was blind to the allocation sequence before assignment.
Statistical analyses
Statistical analyses were conducted with IBM SPSS Statistics (version 28). To explore differences between the ART and BT in the development of mental disorders at FU12, we calculated chi square tests with the dependent variable diagnoses at FU12 and condition as a predictor. To test the effect on our primary and secondary outcomes, we calculated multilevel growth curve models with measurements at level 1 nested within participants at level 2. The time variable was coded as weeks since baseline. Time was included as a predictor at level 1. Condition was included as a level 2 predictor and moderator to test for the superior efficacy of the ART. All models were tested including random intercepts and slopes (cf. Barr et al., 2013), as well as a diagonal and heterogeneous autoregressive covariance structure. When models did not converge, we first alternated covariance structures and in a last step removed random slopes. In addition, we used models with the best fit after comparing restricted maximum likelihood (REML) against maximum likelihood (ML) parameter estimation. Intention-to-treat analyses were compared with complete case analyses. There was no missing data on independent variables (condition, time). For multilevel models with only missing data in the dependent variable, the ML solution and the restricted ML solution have been recommended as they do not assume an equal number of measurements and estimate the most likely values of parameters based on the observed data (van Buuren, 2018). To test for differences between the ART and BT regarding changes in emotion regulation in the ESM, we used multilevel models to account for the hierarchical structure of the data with measurements at level 1 nested within participants at level 2 including random intercepts. Time was included as a level 1 predictor and condition as a level 2 predictor and moderator.
RESULTS
Descriptive statistics
The flow of participants is depicted in Figure 1. As can be seen, n = 138 were allocated to the study conditions (ART, BT). Descriptive statistics for primary and secondary outcomes at measurement points are depicted in Table 2. Those participants in the ART who started the intervention attended on average 5.53 of eight possible sessions (SD = 2.12). At T2, participants in the ART had spent more daily time with the material than those with BT (MdnART = 4, less than 15 min/day; MdnBT = 3), exact Mann-Whitley-U-Test: U = 882.50, p < .001. Participants in the ART and BT did not differ in the daily time they had spent with training materials at T3–T5 (Mdn = 3, less than 10 min/day).
Preliminary analyses
There were no baseline differences between the ART and BT regarding demographics (see Table 1). We tested for possible confounding variables and found that the study cohort significantly moderated the association between time and STAI-T. Consequently, we included study cohort as a moderator in the main model for STAI-T. The average compliance rate in the experience-sampling assessment was 80% at T1 and 75% at T5, which corresponds to compliance rates found in other ESM studies (Vachon et al., 2019). A post-hoc categorization of the qualitative training evaluations revealed that all surveyed participants in the ART were satisfied/very satisfied with the conception of the training and with the ART group instruction. The majority (23 participants) were satisfied with the group atmosphere, four reported that the atmosphere improved over time and eight indicated both positive and negative aspects of the group atmosphere. In the BT, 23 participants were very satisfied/satisfied with the conception of the BT (nine neutral, two dissatisfied).
| Total (n = 138) | ART (n = 68) | BT (n = 70) | F/χ2 | p | |
|---|---|---|---|---|---|
| M (SD)/n (%) | M (SD)/n (%) | M (SD)/n (%) | |||
| Age (years) | 36.44 (12.08) | 35.58 (11.55) | 37.25 (12.59) | .559 | .456 |
| Gender (female) | 89 (75.4%) | 42 (73.7%) | 47 (77%) | .180 | .671 |
| Highest education | 4.591 | .468 | |||
| Middle school | 1 (.08%) | 0 (0%) | 1 (1.6%) | ||
| Secondary school | 4 (3.4%) | 3 (5.3%) | 1 (1.6%) | ||
| Upper second school | 24 (20.3%) | 12 (21.1%) | 12 (19.7%) | ||
| Apprenticeship | 22 (18.6%) | 9 (15.8%) | 13 (21.3%) | ||
| Bachelor's degree | 26 (22.0%) | 10 (17.5%) | 16 (26.2%) | ||
| Master's degree | 41 (34.7%) | 23 (40.4%) | 18 (29.5%) | ||
| Weekly working hours | 30.66 (14.17) | 32.93 (14.10) | 28.65 (14.07) | 2.296 | .133 |
| Income in Euro | 8.246 | .221 | |||
| Low (<1750) | 58 (49.2%) | 27 (47.4%) | 31 (50.8%) | ||
| Middle (1750–4000) | 45 (38.1%) | 25 (43.9%) | 20 (32.8%) | ||
| High (<4000) | 15 (12.7%) | 5 (8.8%) | 10 (16.4%) | ||
| Diagnosis lifetime | 30 (25.4%) | 15 (26.3%) | 15 (24.6%) | .046 | .830 |
| Parents diagnosis lifetime | 25 (21.2%) | 16 (28.1%) | 9 (14.8%) | 3.129 | .077 |
| Marital status | |||||
| Single/divorced/widowed | 56 (47.5%) | 26 (45.6%) | 30 (49.2%) | 1.333 | .514 |
| Married | 20 (16.9%) | 8 (14.0%) | 12 (19.7%) | ||
| In a relationship | 42 (35.6%) | 23 (40.4%) | 19 (31.1%) | ||
| Nationality (German) | 97 (82.9%) | 47 (83.9%) | 50 (82.0%) | 15.908 | .388 |
- Abbreviations: ART, Affect Regulation Training condition; BT, self-help bibliotherapy control condition.
Exploratory analysis: differences in rates of transition to mental disorders at FU12
As can be seen in Table 2, 19% in the ART as compared to 24% of participants in BT at FU12 fulfilled the diagnostic criteria for a current or a remitted disorder that had its onset during the past 12 months (primary outcome; n = 12 major depressive episode, n = 1 dysthymic disorder, n = 3 posttraumatic stress disorder, n = 1 generalized anxiety disorder, n = 1 delusional disorder). There were no significant differences between the ART and BT: Χ2 (2, 73) = .254, p = .614, φ = .06.
| Outcomes | Condition | T1 | T2 | T3 | T4 | T5 | FU6 | FU12 | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M (SD)/n | n | M (SD) | n | M (SD) | n | M (SD) | n | M (SD) | n | M (SD) | n | M (SD)/n | n | ||
| Primary outcomes | |||||||||||||||
| Current diagnosis | ART | n = 0 | n.a. | n.a. | n.a. | n.a. | n.a. | n = 5 | 36 | ||||||
| BT | n = 0 | n.a. | n.a. | n.a. | n.a. | n.a. | n = 6 | 37 | |||||||
| Diagnosis past 12 months | ART | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n = 2 | 36 | ||||||
| BT | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n = 3 | 37 | |||||||
| BSI | ART | .66 (.45) | 55 | .63 (.50) | 47 | .55 (.40) | 45 | .52 (.39) | 44 | .45 (.37) | 45 | .55 (.50) | 42 | .50 (.55) | 42 |
| BT | .74 (.40) | 61 | .70 (.46) | 60 | .58 (.40) | 56 | .49 (.35) | 52 | .51 (.40) | 54 | .65 (.61) | 42 | .54 (.41) | 42 | |
| Secondary outcomes | |||||||||||||||
| CAPE positive freq | ART | 28.32 (6.40) | 57 | 23.23 (3.74) | 47 | 22.91 (3.70) | 45 | 22.55 (3.83) | 44 | 22.24 (3.33) | 45 | 23.45 (4.44) | 42 | 23.19 (3.35) | 43 |
| BT | 28.34 (4.36) | 61 | 23.87 (3.60) | 60 | 23.43 (3.12) | 56 | 23.10 (3.21) | 52 | 23.05 (2.91) | 55 | 24.42 (3.35) | 43 | 23.86 (3.16) | 42 | |
| CAPE positive dis | ART | 27.11 (5.66) | 57 | 22.55 (3.60) | 47 | 21.89 (2.72) | 45 | 21.73 (3.20) | 44 | 21.51 (2.63) | 45 | 22.05 (3.95) | 42 | 21.81 (2.80) | 43 |
| BT | 26.79 (4.47) | 61 | 23.22 (3.51) | 60 | 22.36 (2.47) | 56 | 22.12 (2.56) | 52 | 22.07 (2.65) | 55 | 23.23 (3.46) | 43 | 22.33 (3.21) | 42 | |
| CAPE negative freq | ART | 30.39 (4.35) | 57 | 26.13 (6.81) | 47 | 24.71 (7.02) | 45 | 24.30 (6.93) | 44 | 22.80 (6.65) | 45 | 24.71 (6.82) | 42 | 23.49 (7.36) | 43 |
| BT | 31.46 (6.86) | 61 | 27.45 (7.39) | 60 | 25.18 (7.33) | 56 | 24.81 (7.68) | 52 | 24.96 (7.97) | 55 | 26.60 (8.36) | 43 | 24.95 (7.24) | 42 | |
| CAPE negative dis | ART | 32.25 (5.80) | 57 | 25.66 (8.36) | 47 | 24.64 (8.35) | 45 | 23.68 (8.21) | 44 | 22.16 (8.18) | 45 | 22.67 (8.07) | 42 | 21.37 (6.71) | 43 |
| BT | 32.44 (7.42) | 61 | 28.43 (8.88) | 25.27 (8.48) | 23.96 (8.04) | 52 | 24.20 (8.74) | 55 | 25.64 (9.28) | 43 | 24.40 (8.52) | 42 | |||
| BDI-II | ART | 12.25 (8.15) | 55 | n.a. | n.a. | n.a. | 8.56 (7.12) | 45 | 9.88 (8.35) | 42 | 9.19 (9.28) | 42 | |||
| BT | 13.79 (7.95) | 61 | n.a. | n.a. | n.a. | 9.65 (8.02) | 12.07 (12.09) | 42 | 8.55 (8.70) | 42 | |||||
| STAI-S | ART | 45.73 (9.43) | 55 | n.a. | n.a. | n.a. | 40.96 (9.03) | 45 | 41.33 (10.22) | 42 | 41.56 (12.14) | 43 | |||
| BT | 47.03 (9.80) | 61 | n.a. | n.a. | n.a. | 41.04 (8.57) | 42.40 (10.78) | 43 | 40.93 (9.56) | 42 | |||||
| STAI-T | ART | 47.90 (10.40) | 55 | n.a. | n.a. | n.a. | 42.00 (11.27) | 45 | 41.29 (11.56) | 42 | 40.88 (12.84) | 43 | |||
| BT | 49.72 (9.50) | 61 | n.a. | n.a. | n.a. | 44.24 (10.39) | 45.21 (13.54) | 43 | 43.81 (9.71) | 42 | |||||
| ERSQ-27 | ART | 2.49 (.65) | 56 | 2.62 (.60) | 47 | 2.87 (.49) | 45 | 2.89 (.49) | 44 | 2.98 (.55) | 45 | 2.94 (.64) | 42 | 2.95 (.65) | 43 |
| BT | 2.50 (.61) | 61 | 2.59 (.55) | 60 | 2.70 (.59) | 56 | 2.87 (.54) | 52 | 2.92 (.61) | 55 | 2.87 (.65) | 45 | 3.03 (.56) | 42 | |
| ERSQ-ES ESM | ART | 2.05 (1.00) | 69 | n.a. | n.a. | n.a. | n.a. | n.a. | 3.02 (.79) | 42 | |||||
| BT | 2.16 (.99) | 66 | n.a. | n.a. | n.a. | n.a. | n.a. | 2.56 (1.09) | 45 | ||||||
- Abbreviations: T, measurement, FU6/FU12, follow-ups at 6/12 months, n.a., not assessed; ART, Affect Regulation Training condition; BDI, Beck Depression Inventory; BT, self-help bibliotherapy control condition; BSI, Brief Symptom Inventory; CAPE, Community Assessment of Psychic Experiences; Dis, Distress; ERSQ, Emotion Regulation Skills Questionnaire; ERSQ-ES, Emotion Regulation Skills Questionnaire-Emotion Specific; ESM, Experience-Sampling Method; Freq, Frequency; STAI-S/STAI-T, State Trait Anxiety Inventory-State/Trait.
Differences in changes of general psychopathology
As can be seen in Table 3, there was no significant interaction effect between condition × time (T1–T5) on BSI (primary outcome). There was a significant effect of time (T1–T5) on BSI of medium size. However, time was no longer a significant predictor after including the measurements at FU6 and FU12 (see Table 4).
| Outcomes | Predictors | B | SE | 95% CI | t | p | R2 w/ba |
|---|---|---|---|---|---|---|---|
| T1/2/3/4/5 | |||||||
| Primary outcome | |||||||
| BSI | Time | −.030 | .005 | [−.040, −.020] | −5.858 | <.001*** | .32/.00 |
| Condition (ART) | −.070 | .073 | [−.215, .737] | −.965 | .336 | ||
| Time × condition | .005 | .008 | [−.010, .020] | .654 | .515 | .32/.00 | |
| Secondary outcomes | |||||||
| CAPE positive frequencyb | Time | −.183 | .039 | [−.259, .106] | −4.711 | <.001*** | .33/.00 |
| Condition (ART) | −.123 | .687 | [−1.478, 1.231] | −.179 | .858 | ||
| Time × condition | −.031 | .058 | [−.145, .083] | −.532 | .596 | .33/.00 | |
| CAPE positive distressb | Time | −.212 | .043 | [−.296, −.128] | −4.975 | <.001*** | .30/.00 |
| Condition (ART) | −.310 | .607 | [−1.506, .885] | −.512 | .609 | ||
| Time × condition | −.009 | .063 | [−.134, .116] | −.149 | .882 | .30/.00 | |
| CAPE negative freqc | Time | −.524 | .104 | [−.732, −.315] | −4.991 | <.001*** | .40/.00 |
| Condition (ART) | −.432 | 1.163 | [−2.736, 1.872] | −.372 | .711 | ||
| Time × condition | −.184 | .156 | [−.494, .126] | −1.180 | .241 | .40/.01 | |
| CAPE negative disc | Time | −.853 | .119 | [−1.088, −.617] | −7.177 | <.001*** | .40/.01 |
| Condition (ART) | −.544 | 1.280 | [−3.082, 1.994] | −.425 | .672 | ||
| Time × condition | −.090 | .177 | [−.442, .262] | −.507 | .613 | .40/.02 | |
| ERSQ-27 | Time | .055 | .009 | [.036, .073] | 5.928 | <.001 | .45/.00 |
| Condition (ART) | .022 | .108 | [−.192, .235] | .201 | .841 | ||
| Time × condition | .006 | .014 | [−.021, .033] | .425 | .672 | .45/.00 | |
| T1/T5 | |||||||
| BDI-IIb,c | Time | −.472 | .113 | [−.696, −.248] | −4.177 | <.001*** | .43/.00 |
| Condition (ART) | −1.703 | 1.490 | [−4.655, 1.249] | −1.143 | .255 | ||
| Time × condition | .072 | .168 | [−.261, .405] | .430 | .668 | .43/.00 | |
| STAI-Sb,c | Time | −.720 | .159 | [−1.036, −.403] | −4.513 | <.001*** | .19/.00 |
| Condition (ART) | −2.850 | 1.737 | [−6.292, .591] | −1.655 | .104 | ||
| Time × condition | .283 | .236 | [−.185, .750] | 1.198 | .234 | .19/.00 | |
| STAI-Tb | Time | −.995 | .217 | [−1.426, −.564] | −4.579 | <.001*** | .25/.00 |
| Condition (ART) | −3.436 | 1.846 | [−7.092, −.219] | −1.862 | .065 | ||
| Time × condition | .122 | .186 | [−.248, .492] | .655 | .514 | .25/.00 | |
| Study cohort | −1.584 | .561 | [−2.695, −.472] | −2.821 | .006** | ||
| Time × study cohort | .115 | .555 | [.005, .225] | 2.066 | .041* | ||
| ERSQ-ES ESM | Time | .437 | .090 | [.258, .615] | 4.855 | <.001* | .23/.34 |
| Condition (ART) | −1.076 | .119 | [−.344, .129] | −.901 | .369 | ||
| Time × condition | .390 | .131 | [−.129, .651] | 2.971 | .004** | .23/.36 | |
- Abbreviations: BDI, Beck Depression Inventory; BSI, Brief Symptom Inventory; CAPE, Community Assessment of Psychic Experiences; Dis, Distress; ERSQ, Emotion Regulation Skills Questionnaire; ERSQ-ES, Emotion Regulation Skills Questionnaire-Emotion Specific; ESM, Experience-Sampling Method; Freq, Frequency; STAI-S/-T, State Trait Anxiety Inventory-State/Trait.
- a R2 as effect size measure calculated based on centered predictors and standardized outcome variables to indicate (w = within)- and (b = between)-participant variance explained.
- b Multilevel models without random slopes.
- c Multilevel models with REML.
- *p < .05; **p < .01; ***p < .001.
| Outcomes | Predictors | B | SE | 95% CI | t | p |
|---|---|---|---|---|---|---|
| T1/2/3/4/5/FU6/12 | ||||||
| Primary outcome | ||||||
| BSI | Time | −.002 | .001 | [−.004, .000] | −1.967 | .052 |
| Condition (ART) | −.458 | .683 | [−.181, .089] | −.671 | .503 | |
| Time × condition | .001 | .001 | [−.002, .004] | .747 | .457 | |
| Secondary outcomes | ||||||
| CAPE positive freqa | Time | .009 | .006 | [−.002, .020] | 1.560 | .121 |
| Condition (ART) | −.311 | .604 | [−1.507, .885] | −.515 | .608 | |
| Time × condition | −.003 | .008 | [−.020, .013] | −.427 | .670 | |
| CAPE positive disa | Time | .001 | .006 | [−.011, .0129] | .112 | .911 |
| Condition (ART) | −.385 | .499 | [−1.372, .603] | −.771 | .442 | |
| Time × condition | −.005 | .009 | [−.022, .013] | −.528 | .599 | |
| CAPE negative freq | Time | −.020 | .014 | [−.049, .008] | −1.420 | .159 |
| Condition (ART) | −.796 | 1.217 | [−3.206, 1.613] | −.654 | .514 | |
| Time × condition | −.009 | .020 | [−.049, .032] | −.428 | .669 | |
| CAPE negative dis | Time | −.030 | .016 | [−.062, .002] | −1.891 | .062 |
| Condition (ART) | −.519 | 1.455 | [−3.402, 2.363] | −.357 | .722 | |
| Time × condition | −.036 | .023 | [−.081, .009] | −1.601 | .113 | |
| ERSQ-27 | Time | .006 | .001 | [.004, .009] | 4.897 | <.001*** |
| Condition (ART) | .070 | .093 | [−.115, .254] | .748 | .456 | |
| Time × condition | −.004 | .002 | [−.008, .000] | −2.043 | .044* | |
| T1/5/FU6/12 | ||||||
| BDI-II | Time | −.066 | .021 | [−.108, −.024] | −3.134 | .002** |
| Condition (ART) | −1.720 | 1.334 | [−4.360, −.920] | −1.289 | .200 | |
| Time × condition | .050 | .030 | [−.009, .110] | 1.672 | .098 | |
| STAI-Sb | Time | −.077 | .027 | [−.132, −.023] | −2.815 | .006** |
| Condition (ART) | −1.926 | 1.525 | [−4.943, 1.092] | −1.263 | .209 | |
| Time × condition | .057 | .039 | [−.021, .134] | 1.454 | .149 | |
| STAI-Tb | Time | −.077 | .023 | [−.122, −.032] | −3.372 | .001*** |
| Condition(ART) | −3.276 | 1.856 | [−6.952, .400] | −1.765 | .080 | |
| Time × condition | .020 | .032 | [−.044, .084] | .611 | .543 | |
- Abbreviations: T, measurement; FU6/FU12, follow-ups at 6/12 months; ART, Affect Regulation Training condition; BDI, Beck Depression Inventory; BSI, Brief Symptom Inventory; BT, self-help bibliotherapy control condition; CAPE, Community Assessment of Psychic Experiences; Dis, distress; ERSQ, Emotion Regulation Skills Questionnaire; Freq, frequency; STAI-S/STAI-T, State Trait Anxiety Inventory-State/Trait.
- a Multilevel models without random slopes.
- b Multilevel models with REML.
- *p < .05; **p < .01; ***p < .001.
Differences in changes of psychotic, depression, and anxiety symptoms
There was no significant interaction effect between condition × time (T1–T5) on CAPE, BDI-II and STAI-S, STAI-T (secondary outcomes; see Table 3), but again, there were significant effects of time with large effects on CAPE positive symptom frequency and distress, negative symptom frequency and distress; medium effects on STAI-T; and small effects on BDI-II and STAI-S. The effects of time remained significant at FU6 and FU12 on BDI-II, STAI-S, and STAI-T (see Table 4).
Differences in changes of emotion regulation
As can be seen in Table 3, there was a significant, medium interaction effect of condition × time showing higher improvements on ERSQ-ES total score as assessed with the ESM in the ART than in BT. There was no significant interaction effect of condition × time for ERSQ-27 as assessed with a retrospective questionnaire at (T1–T5), but this effect was significant in favour of BT including follow-ups FU6 and FU12 (Table 4). Furthermore, there was a significant, medium effect of time (T1–T5) and this effect was still significant at the follow-ups (T1–FU12).
Sensitivity analysis
The results of the complete case analysis confirm our main findings (see Appendix S1). Specifically, the results show effects of time for all primary and secondary outcomes and no interaction effects of time × group for primary and secondary outcomes except for ERSQ-ES as assessed with the ESM. A post-hoc power analysis with G*Power (Faul et al., 2007), 1 − β = .80 revealed that the chi-square test for our primary outcome of diagnoses with an estimated small effect size (φ = .10) had insufficient power (.01) to detect a clinically significant difference between the ART and BT. A total sample size of n = 964 would have been necessary to detect a small effect (i.e., a transition rate of 19% in the ART vs. 24% in BT).
DISCUSSION
We aimed to determine whether a group-based emotion regulation intervention would be superior to unguided BT in preventing the transition to mental disorders in a subclinical sample with distressing symptoms of psychosis, depression, or anxiety. The ART was superior to the BT in improving ESM-reported emotion regulation but was not superior in preventing transition to mental disorders at 12-months follow-up or in reducing general psychopathology and symptoms.
The efficacy of the ART for emotion regulation corroborates previous studies on the ART (Berking et al., 2010, 2019, 2022; Buruck et al., 2016). The ART did not show superiority over unguided BT, however, when emotion regulation was assessed via a retrospective questionnaire. This corroborates the marked discrepancy between retrospective (or trait) measures of emotion regulation and state emotion regulation as assessed with the experience-sampling assessments that has been reported in previous studies (e.g., Brockman et al., 2017; Ludwig et al., 2020). This discrepancy has been largely ascribed to a lack of validity of the retrospective questionnaires that are prone to recall bias and reporting biases due to present emotional states (Lincoln et al., 2022). Moreover, trait measures of emotion regulation have the disadvantage that they do not account for variations in strategy use across diverse contexts (Brockman et al., 2017). The ESM is seen as a major advancement of questionnaire assessments, as it has been shown to be more ecologically valid, less prone to recall bias, and—due to its change-sensitivity—a better indicator of intervention effects (Myin-Germeys et al., 2018). Therefore, we argue that it is justified to give more weight to the ESM findings and to conclude that our study corroborates previous findings that the ART in itself is effective in improving emotion regulation.
The ART could not prevent 19% of participants from transitioning to a mental disorder, which was comparable to 24% in BT. Among those who transitioned (n = 16), most fulfilled the criteria for a major depressive episode (n = 12). This corresponds to an incidence of 16%, which is higher than annual incidence rates for major depressive episodes in community samples (3%; for a review, see Ferrari et al., 2013) but lies in the range found in the intervention conditions of prevention trials for major depression (2%–21%; for a review, see Munoz et al., 2010). A post-hoc power analysis revealed that our test had insufficient power to detect a clinically significant difference between the ART and BT. Thus, the analysis on transition rates was exploratory and has to be interpreted with caution. However, considering the minimal difference between transitions to mental disorders, it is unlikely that a bigger sample would have led to a clinically significant difference. Interestingly, despite the fact that around one in five participants transitioned to a mental disorder in our study, the average general psychopathology, symptoms of psychosis, depression, and anxiety markedly decreased over the course of both interventions. Previous studies that tested BTs with a focus on facets of emotion regulation also found effects on symptoms of depression and anxiety (Hazlett-Stevens & Oren, 2017; Jeffcoat & Hayes, 2012). Indeed, most of our participants with BT used the training material and were satisfied with the intervention. Due to a lack of a control group receiving no intervention, we cannot clarify if both the ART and BT had an effect beyond unspecific effects. In the light of previous findings and our results, it can be speculated that both the ART and BT prevented an even higher transition rate. Given the lack of group difference, this interpretation also implies that the additional effort of a group-based ART is not required to prevent the transition to mental disorders.
The more elaborate, in-person ART was not superior over unguided BT in reducing psychopathology. One possible explanation could be that it needs stronger, possibly more comprehensive effects on emotion regulation than the medium effect on adaptive emotion regulation we found in the ESM. Comprehensive reviews indicate that psychopathology is more prominently associated with an overuse of ‘maladaptive’ strategies (e.g., suppression, rumination, and avoidance) than with an underuse of ‘adaptive’ strategies (e.g., cognitive reappraisal, problem-solving, and acceptance; see Lincoln et al., 2022—also for a discussion of the adaptive vs. maladaptive distinction). The ART has its main focus on conveying adaptive strategies (Berking & Whitley, 2014), whereas the BT used here also explicitly addresses maladaptive strategies (Barnow, 2018). Thus, it is conceivable that additional facets of emotion regulation need to be addressed, complementary to those addressed in the ART, to significantly reduce psychopathology and symptoms. Another disadvantage of the ART as compared to BT could have been its inflexibility and uncontrollability regarding pace of exposure, content focus, and session scheduling (cf. McKenna et al., 2010). In our study, attending ART group-sessions of 2.5 hr was challenging for participants with high average working hours (30 hr/week). Eight participants explained that scheduling problems prevented them from starting the ART and the average attendance of M = 5.5 (SD = 2.1) group sessions shows that some participants were unable to achieve a desirable session attendance. In contrast, participants in the BT condition could decide when to train and could focus on their prioritized aspects of emotion regulation. All participants in the ART were satisfied with the training conception as well as group instruction and a majority of n = 23 reported a pleasant group atmosphere. Nevertheless, eight participants also reported some negative aspects of the group atmosphere, some of which resulted from discomfort sharing personal experiences with strangers. Thus, for some participants the BT may have been more suitable.
Based on our findings, we therefore suggest that future research on prevention in an at-risk population could investigate an emotion regulation training with a stronger focus on identifying and preventing the overuse of suppression, rumination, and avoidance (cf. Barlow et al., 2011; Bullis et al., 2018; Linehan, 1993; Renna et al., 2017). Moreover, Arango et al. (2018) have recommended that while people in an at-risk state should be supported in strengthening certain skills (cf. emotion regulation) it is also relevant to target contextual risk factors, such as unhealthy nutrition, lack of exercise, bullying, and a stressful work environment. Furthermore, future research could examine whether participants would benefit from a more flexible, blended approach, including extended opportunities to review session content and to train regularly while underway, for example, with the support of an application on their mobile phone (cf. Böhme & Berking, 2021). Research also indicates the relevance of shared decision making and active choices for intervention formats with respect to clinical outcomes (Lindhiem et al., 2014). In clinical practice, participant preferences should therefore be taken into account in deciding whether to select a group-based intervention or a self-help BT.
Our findings should be considered in the light of some strengths and limitations. Self-help BT as an active control condition allowed to control for the stand-alone effect of a structured, autonomous engagement with emotion regulation content for an equal period of 8 weeks. This enabled us to evaluate the additional value of the ART as a more elaborate, group-based intervention. However, our design did not include a control group that enabled to draw strong conclusions in regard to emotion regulation as a mechanism of change. Furthermore, we did not include a control group receiving no intervention and thus had no control for unspecific effects, such as regression to the mean and spontaneous remission. Furthermore, our sample was 75% female. This might limit the generalizability of our findings as gender differences have been found for emotion regulation (Nolen-Hoeksema, 2012). Moreover, the sample was highly educated which might have led to a higher benefit from the BT than might be expected in a more representative sample. Finally, 16% of participants in the ART and 11% in the BT condition dropped out between T1–T5 and an additional 9% in the ART and 22% in BT dropped out until FU12. We addressed the problem of missing data by using multilevel analyses and by comparing intention-to-treat with complete case analyses.
We conclude that the ART was efficacious in improving emotion regulation in an at-risk sample, feasible, and accepted. However, it was not superior to a self-help manual in respect to the prevention of mental disorders or the reduction of psychopathology. Future research should test whether emotion regulation represents a relevant mechanism of change in the context of prevention. Furthermore, future studies could consider an extended and more flexible approach to reach a broad spectrum of at-risk participants, enhance compliance, and produce a stronger effect on psychopathology.
AUTHOR CONTRIBUTIONS
Martin F. Wittkamp: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; visualization; writing – original draft. Katarina Krkovic: Formal analysis; validation; writing – review and editing. Tania M. Lincoln: Funding acquisition; methodology; project administration; resources; supervision; validation; writing – review and editing.
ACKNOWLEDGEMENTS
We want to thank the students involved in data collection: Lea Hornung, Kristin Medel, Anna Möggenried, Joana Ohlmer, Franziska Sikorski, and Lena Yilmaz. This research received no specific grant from any funding agency, commercial, or not-for-profit sectors. Open Access funding enabled and organized by Projekt DEAL.
CONFLICT OF INTEREST STATEMENT
All authors declare no conflict of interest.
Open Research
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to ethical restrictions.
