Feasibility and Efficacy Estimate of an Emotion-Focused Version of Cognitive Behavior Therapy for Delusions in Comparison to Wait List in a Single-Blinded Randomized-Controlled Pilot Trial

Abstract

Background

Cognitive Behavior Therapy for psychosis (CBTp) effectively reduces positive symptoms, but the effects for delusions are small. Using approaches that focus on improving putative causal factors of delusions (emotion regulation [ER] and negative self-schemata) may be a way to improve CBTp for delusions.

Objectives

To pilot an emotion-focused CBTp intervention comprised of cognitive-behavioral and third-wave interventions targeting ER and negative self-schemata (CBT-E) to investigate its feasibility and its potential efficacy.

Methods

A single-blind 2-centered randomized wait-list controlled trial was performed. Patients with delusions in the context of a psychotic disorder (n = 64) were randomized to CBT-E (treatment group [TG]: n = 35) or wait list (WL: n = 29). TG received individual CBT-E in weekly sessions over 6 months, WL received CBT-E after a 6-month waiting time. Assessments were performed at baseline, after 6 months, and after 12 months.

Results

Rates of patients who consented and retained in CBT-E suggest acceptable feasibility and high acceptance of CBT-E. Multiple regression analysis in the intent-to-treat sample revealed no significant advantage in favor of the TG in the primary outcome delusions (PSYRATS delusions, PDI), no effect on the mediators (ER, negative self-schemata) and a significant advantage in only one of the 8 secondary outcome variables, namely general psychopathology.

Discussion

Although the intervention was feasible and acceptable, its effects on delusions did not hold up with the effects found for “targeted” approaches in previous research and it needs to be improved to produce a stronger effect on the targets ER and self-schemata before proceeding to a large-scale randomized-controlled trial. Clinicaltrials.gov-Identifier: NCT02787122.

Introduction

Delusions are present in the overwhelming majority of patients with schizophrenia-spectrum disorders.^1–3 Cognitive Behavior Therapy for psychosis (CBTp) has focused on changing these delusional beliefs. Most meta-analyses have found it to be more effective than routine care^4,5 in terms of reducing positive symptoms and overall psychopathology, whereas the evidence for its superiority over active control groups and other psychological interventions is less clear.⁶

Nevertheless, in 2 meta-analyses that focused on change in delusional beliefs specifically, the effect of CBTp in comparison to routine care (treatment as usual [TAU]) was small (d = 0.27⁷; g = 0.36⁸). However, some of the studies included in these meta-analyses did not focus on directly addressing delusional beliefs “via re-evaluating people’s perceptions, beliefs, or reasoning related to the target symptoms,” as is typical for CBTp (definition of CBTp by NICE⁹). Rather, these studies used a “causal interventionist” ¹⁰ or “targeted” approach¹¹ and focused on changing single putative causal factors such as worry and rumination^12,13 and negative self-schemata¹⁴ that were selected based on theoretical models of delusions.^15–17 Interestingly, a moderation analysis indicated these “targeted” interventions to be more effective in reducing delusional beliefs than the CBTp interventions that addressed delusional beliefs more directly (difference: d = 0.33⁷).

Thus, targeting putative causal factors of delusions could be the way forward to improving the unsatisfactory effect sizes in interventions for delusions. To intensify these interventions, a focus on the putative mediator emotion regulation (ER) seems worthwhile, as the 2 studies that targeted worry showed convincing effects on delusional beliefs^12,13 and worry is typically conceptualized as a (dysfunctional) ER strategy. ER strategies are defined as techniques used to increase, maintain, or decrease emotions.¹⁸ Targeting ER aligns with the strong evidence base promoting the pronounced self-reported difficulties in ER in patients with delusions¹⁹ and the increasing bulk of study findings showing increases in negative affect and arousal to precede increases in delusions.^20–24 In targeting ER, it also seems promising to broaden the focus and to also focus on patients’ ability to correctly identify and label their experienced emotions (emotional awareness), as problems in this area have a severe impact on patients’ daily functioning²⁵ and have been found to moderate the link between negative affect and delusions in patients daily life.²²

Although a “targeted” intervention that focused on negative self-schemata as a putative causal factors of delusions¹⁴ was less successful than the interventions with a focus on worrying,^12,13 the relevance of negative self-schemata to delusion formation and maintenance is also well backed up with studies repeatedly showing decreases in self-esteem to precede increases in delusions.^26–29 As the study that focused on negative self-schemata included a small number of sessions (6 sessions),¹⁴ it may be beneficial to prolong the intervention for negative self-schemata to achieve a more pronounced effect. Taken together, it appears worthwhile to focus on ER and negative self-schemata as putative causal factors of change in delusions, but to prolong the interventions in order to achieve a stronger overall intervention effect on delusions.

Interventions that focus on ER and self-schemata attempt to train new skills via self-practice. Several studies point to a dose-response relationship between self-practice and improved ER skills³⁰ and outcome variables.^31,32 As patients with psychosis often report problems with therapy homework,^33,34 another way to intensify the effect of existing approaches and help patients in training the new strategies could be to offer additional support from cotherapists in patients’ everyday life.

This study’s primary aim was to test the feasibility and acceptance of an emotion-oriented version of CBTp (CBT-E) for patients with psychosis and delusions that targets ER and negative self-schemata. A second aim was to probe for the efficacy of CBT-E to estimate an adequate sample size for a fully powered and more rigorous randomized-controlled trial (RCT). Specifically, we tested whether CBT-E would be more effective than routine care (wait list [WL]) in improving delusions (primary outcome) and in reducing positive symptoms, negative symptoms, general psychopathology, depression, and improving social functioning (main secondary outcomes). We also explored CBT-E’s effect on improving putative causal factors of change in delusions (ER including worrying, self-schemata) which we assumed to be putative mediators of change in delusions. In addition, we assessed the sustainability of primary and secondary outcome factors and putative mediators of delusions at a 12-month-follow-up (FU). A final aim was to test whether patients would welcome support from a cotherapist and whether adding support from a cotherapist can improve the interventions’ efficacy.

Methods

Ethical Approval and Registration

The ethics commission of the Faculty of Medicine in Marburg approved the study for both centers (Approval no. 150/3). The study was retrospectively registered at ClinicalTrials.gov (Identifier: NCT02787122). There were some changes between the ethics approval and the start of the study (detailed in Supplementary table S9). After the start of the study, no additional changes were made.

Study Design

The study design was a single-blind randomized wait-list controlled trial comparing individual CBT-E (treatment group [TG]) plus routine care (monthly psychiatric visits) with a WL condition receiving routine care only. The WL-group received CBT-E after 6 months of waiting time. Assessments were performed at the start of therapy, post 6 months waiting time (WL-group), post (approximately) 6 months of therapy, and after a 12-month-FU post-CBT-E. All patients receiving CBT-E were offered support by a cotherapist, but acceptance of this support was not mandatory for trial participation.

Participants, Inclusion and Exclusion Criteria

Participants were recruited at 2 outpatient clinics of psychological institutes (Marburg and Hamburg). Inclusion criteria were age between 18 and 70 years, fulfillment of diagnostic criteria for a psychotic disorder (schizophrenia, schizoaffective disorder, or delusional disorder) according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR³⁵) confirmed with the Structured Clinical Interview for DSM-IV (SCID³⁶), persistent delusions of at least mild degree as suggested by a score of at least 2 in the item P1 (general delusions) in the Positive and Negative Syndrome Scale (PANSS³⁷), an estimated verbal intelligence score of at least 70 in the German Mehrfachwahlwortschatztest (MWT-B³⁸) and no present suicidal tendencies as reported in the SCID interview. Exclusion criteria were a comorbid diagnosis of a borderline personality disorder or substance use disorder in the last 6 months, as these conditions would require different interventions or a transfer to a specialized patient unit in the German health care system.

Recruiting Procedure and Diagnostic Assessment

Recruitment began in January 2014 and ended in December 2016; FU assessments were completed in December 2018. Patients either directly contacted the outpatient clinics or were referred by their psychiatrist, general practitioner, or social worker who had been informed about the treatment option. Potentially eligible patients were offered a first session with their study therapist and were informed about the study, its duration, the focus of the interventions on ER and negative self-schemata and about the option to choose additional training sessions provided by a cotherapist.

If patients were interested in participation, an appointment for a 3–5-h diagnostic assessment that included interviews and questionnaires was arranged to assess eligibility. Patients brought back the written consent form at this appointment.

All primary and secondary outcome measures, as well as the putative mediators, were assessed at pre- and postassessment (and at postwaiting time in the WG). At 12-month-FU, an abbreviated assessment that included only the observer-rated measures and questionnaires assessing the CBT-E intervention’s perceived acceptance were administered.

Randomization Procedure

Participants were then randomized to the TG- or the WL-group (1:1). The random assignment was performed using random numbers created on the website www.random.org and was stratified by symptom severity (mild, moderate, or severe based on the total score in the PANSS: mild [0–53 points], moderate [54–74], or severe [>74]).

Adequate Sample Size and Power Analysis

Because the present study was a pilot trial, and CBT-E has never been tested previously, there was little basis for estimating an adequate effect size.

As a rough estimate, however, we also conducted a power calculation with the program G-power³⁹ based on the aggregated efficacy of the “targeted” study that detected an effect on delusions of moderate size (d = 0.52⁷). An adequate sample size for a multiple regression analysis using postscores of the primary outcome as criterion and prescores and group as predictors with an alpha score of .05 and power of .80 would be a total of 43 patients.

Intensified Emotion-Focused Cognitive Behavior Therapy for Patients With Delusions

CBT-E was offered in approximately 20 weekly therapy sessions (50 min) over 6 months. Therapists and patients decided on whether to arrange additional sessions after 6 months based on the achieved therapy goals, which is a routine procedure in the German health care insurance system,⁴⁰ which paid for the treatment sessions.

Selection of interventions was based on the interventions used in the “targeted” studies^12–14 and amplified with additional interventions focusing on a broader range of ER skills and negative self-schemata informed by various third wave CBT approaches (eg, Acceptance and Commitment Therapy,⁴¹ Emotion-Focused Therapy,⁴² Compassion-Focused Therapy,⁴³ and Schema Therapy^44,45).

CBT-E involved the following modules: (1) improving emotional stability including psychoeducation on lifestyle, activities, and sleep and their role in reducing the vulnerability for negative emotions, psychoeducation on the importance and planning of a regular daily schedule and implementation of an activity diary (mastery and pleasure technique); (2) improving emotion regulation including psychoeducation on emotions and how to identify and label them correctly and to become aware of the related thoughts, perceptions, and body sensations and psychoeducation on functional and dysfunctional ER strategies, challenging common dysfunctional beliefs about emotions, implementing an emotion diary, training ER strategies for specific emotions (eg, improving awareness, acceptance, reappraisal) and reducing worrying (ie, challenging positive and negative metacognitions on worrying, replacing worrying with alternative coping strategies [mindfulness]); (3) interventions to reduce negative self-schemata including psychoeducation on self-acceptance and the role of positive and negative self-schemata, implementing a diary to report self-schemata and related cognitions, challenging negative self-schemata and related cognitions using Socratic reasoning, chair work, imagination strategies, and helpful self-instructions. In the final sessions, interventions and helpful coping strategies were recapitulated.

CBT-E was designed to target the putative causal factors of delusional beliefs and not the delusional beliefs per se. Thus, therapists were instructed not to address delusional beliefs directly. All interventions are described in a manual (available from the first author).

Additional Sessions With a Student Cotherapist

Patients who accepted receiving support from a cotherapist met their cotherapist during the first therapy session and then in additional weekly sessions either at the patients’ home or at other locations in the patients’ everyday life. Interventions provided by cotherapists were based on a manual for cotherapists. Cotherapists recapitulated information provided in CBT-E sessions in additional sessions and supported patients in practicing any new interventions, eg, implementing activities in daily life or training and using functional ER strategies.

Training of Therapists and Cotherapists

Therapists were 20 clinical psychologists (MSc) and 1 psychiatrist. All therapists were enrolled in their second to third year of German postgraduate training to become certified CBT therapists (total duration 3–5 years). All psychologists had passed an examination that allowed them to treat outpatients with mental disorders supervised by a certified therapist and had been enrolled in approx. 200–300 h of theoretical postgraduate CBT training courses. They received an additional training course (20 h) consisting of lectures on the intervention and training exercises in form of role-plays supervised by 1 PI (SM). Also, they received biweekly manual supervision and case supervision after every fourth session by one of the study PIs (SM/TL). Cotherapists were 17 students of psychology who received 20 sessions of similar training as the therapists and individual weekly supervision by their therapists, and supervision every second week by one of the study PIs (SM). Also, they met with their fellow cotherapists in weekly collegial intervision sessions.

Measures

Assessment, Training, and Blinding Procedure

PANSS, PSYRATS, CDSS, and RFS were first rated by each patients’ study therapists who recorded the interviews to provide a videotape for the blinded assessment. Videotapes were later blinded and rated by 6 independent raters. Intraclass correlations (ICCs) between therapist raters and blinded raters were performed as an unadjusted analysis of variances.

Feasibility and Acceptance of Treatment

Feasibility and acceptance indicators included the number and proportion of referred patients that were eligible for an RCT, signed informed consent, completed the intervention as intended and accepted sessions with an additional cotherapist, as well as the mean number of sessions the patients attended, the number and ratio of patients who dropped out, and the number and ratio of adverse events.

Moreover, we assessed the perceived acceptance of CBT-E with a questionnaire at 12-month-FU that was also used in a previous study of CBTp⁴⁶ in a similar context. Acceptance was assessed with 4 items that measured (1) whether patients perceived CBT-E as helpful, (2) whether CBT-E was suited to their present situation, (3) whether they rated the strategies to cope with their problems as promising, and (4) whether they would recommend CBT-E to a friend. Items were rated on a 5-point Likert scale (range: −2: “strongly disagree” to +2: “strongly agree”).

Primary Outcome Measures Delusions

Psychotic Symptom Rating Scales delusions scale (PSYRATS⁴⁷). The PSYRATS delusions scale is based on a semistructured interview that assesses the amount of preoccupation, duration of preoccupation, conviction, disruption of daily life, amount of distress, and intensity of distress concerning delusions with 6 items that are answered on a 5-point Likert scale (range 0–24). Inter-rater reliability and validity were reported as high,⁴⁷ and ICCs in our sample were excellent (.95).

Peters et al. Delusions Inventory—short version (PDI-21⁴⁸; German version⁴⁹): The PDI-21 was used to assess delusional beliefs, frequency, conviction, preoccupation, and distress with 21 items rated on a 7-point Likert scale. Participants are first asked whether they endorse the belief and then asked to rate it concerning their frequency, preoccupation, distress, and conviction. The PDI-21 has demonstrated good internal consistency and validity.⁴⁸ The PDI grand total score⁴⁸ was used (sum of endorsed beliefs, preoccupation, distress, and conviction scales: range 0–336).

Secondary Outcome Measures

The Positive and Negative Syndrome Scale (PANSS⁵⁰) was used to assess positive and negative symptoms of schizophrenia and general psychopathology (anxiety, depression) based on a semistructured interview assessing 30 symptoms divided into 3 scales (PANSS positive scale (7 symptoms: range 7–49), PANSS negative scale (7 symptoms, range: 7–49), and PANSS general psychopathology scale (16 symptoms, range: 16–112) on a 7-point Likert scale. Internal consistency (.70–.85) and validity were acceptable^50,51 and ICCs were acceptable to good in our sample (.83, .72, .82).

The Calgary Depression Scale for Schizophrenia (CDSS⁵²) is a semistructured interview to assess observer-rated depressive symptoms (7 items) on a 4-point Likert scale. Reliability and validity were high⁵² and ICCs were good in the present sample (.83). The CDSS sum score was used (range 0–21).

The Beck Depression Inventory—Second version (BDI-II⁵³) was used to assess self-rated depressive symptoms with 21 items rated on a 4-point Likert scale. The BDI-II total score was used (range 0–44).

The Role Functioning Scale (RFS⁵⁴; German version⁴⁶) was used to assess functioning in the areas of working productivity, independent living, immediate social network relationships (friends and family), and extended social network relationships (other social contacts). Criterion validity was high⁵⁴ and ICC was acceptable in our sample (.66, .62, .70, .63). We used the mean score (range 12–48), with higher scores indicating higher social functioning.

The Paranoia Checklist⁵⁵ (German version⁵⁶) is a self-report questionnaire used to assess the frequency, distress, and conviction of persecutory delusions, with 18 items answered on a 5-point Likert scale. The Paranoia Checklist frequency scale and distress scale were used (ranges 0–72).

Putative Mediators of the Intervention Effect on Delusions

The Emotion Regulation Skills Questionnaire total score (ERSQ, German version⁵⁷) was used as a self-report of ER skills with 27 items rated on a 5-point Likert scale (range: 0–108).

The Perseverative Thinking Questionnaire total score (PTQ⁵⁸) was used to assess worrying and consists of 15 items that were self-rated on a 5-point Likert scale (range 0–60).

The Brief Core Schema Scale negative self-scale (BCSS⁵⁹) was used to assess negative self-schemata and consists of 6 items answered on a 5-point Likert scale (range 0–24).

Adverse Events Monitoring and Additional Assessments

Therapists were required to record any adverse events, including suicide attempts and hospital readmissions, along with the putative causes.

In addition, therapist raters assessed whether the delusions were acute vs nonacute and the type of the most prominent delusion (ie, persecutory delusions or other delusions) on the basis of the diagnostic interviews. Other additional outcome variables are reported in Supplementary table S1.

Statistical Analyses

Statistical analyses were conducted by an independent clinical statistician (C.H.) using SPSS (Version 26). Analyses were not controlled for multiple testing due to the pilot character of the study, in line with other studies on “targeted” interventions.^12–14

Missing data were present at all assessment times due to dropout (see figure 1 for more information), technical problems with the recording, and patients who refused to participate in the video recording or other parts of the assessment. Missing data varied between 0% and 31.3% at preassessment, 18.8% and 37.3% at postassessment (post-WL/post-TG), between 23.4% and 55.2% at post-treatment assessment in the WL-group and 34.5% and 41.4% at 12-month-FU assessment.

Fig. 1.

Flow chart of the study. Note: CBT-E, Emotion-focused version of Cognitive Behavior Therapy for delusions; FU, follow-up; TG, therapy group; WL, wait list group.

Missing data were replaced using the SPSS package for multiple imputations in the intent-to-treat (ITT) sample. As missing data were lower than 60%, we imputed a total of 40 datasets as recommended by Graham et al⁶⁰ to prevent loss of statistical power due to the imputation. All primary and secondary outcome variables were used as predictors. Missing data were replaced in 2 different datasets. Dataset 1 was used to assess the efficacy of primary and secondary outcome variables and mediators. It included the data points preassessment (TG: pretherapy; WL: prewaiting time) and postassessment (TG: post-therapy; WL: postwaiting time). Dataset 2 was used to investigate the sustainability and effects of cotherapists and consisted of the data points preassessment (TG: pretherapy; WL: postwaiting time/pretherapy), postassessment (TG and WL: post-therapy), and 12-month-FU assessment.

Analyses of Change in Primary and Secondary Outcome Variables and Putative Mediators

Change in primary and secondary outcome variables and change in putative mediators from pre- to postassessment was tested in dataset 1 using multiple regression analyses. Postscores of these variables were used as dependent variables and prescores of these variables and group as predictor variables. Analyses were repeated controlling for center and stratification (PANSS total score) by including these factors as additional predictors.

Sustainability of effects between pre- and posttherapy and between pre- and 12-month-FU was explored in the total sample (TG + WG) in dataset 2 using t tests for dependent variables.

The influence of cotherapists on change was analyzed using multiple regression analysis in dataset 2. Postscores/12-month-FU scores of all outcome variables were included as dependent variables and prescores of these variables and group (1 = presence of an additional cotherapist, 0 = absence) were included as predictors.

Sensitivity Analysis

As recommended by several guidelines for multiple imputation,^61,62 we performed an additional sensitivity analysis in the per-protocol sample without replacing missing data to control for the influence of multiple imputation. Furthermore, due to the lenient inclusion threshold for delusions (PANSS P1 score ≥2), we probed for efficacy in a subgroup of more acutely delusional patients using the PSYRATS delusional conviction item ≥3 in line with inclusion criteria of the “targeted” studies.^12–14

Results

Feasibility and Acceptance of CBT-E

As can be seen in the flow chart in figure 1, from the referred and screened 181 patients, 83 patients were fully assessed for eligibility. Of these patients, 78 were eligible (93.98%), and 14 of these declined to participate (17.95%). Sociodemographic and clinical variables of the eligible and consenting sample who were randomized to the intervention arms (ITT sample; n = 64) are provided in Table 1. All of the following results refer to this sample.

Table 1.

Sociodemographic and Clinical Characteristics of Patients With Psychotic Disorders Enrolled in the Study and Allocated to the Therapy Group (TG) or the Wait-List Group (WL) (Intention-to-Treat Sample; n = 64)

Variable	TG (n = 35) M (SD)/n (%)	WL (n = 29) M (SD)/n (%)
Sociodemographic variables
Age (years)	36.74 (10.49)	36.66 (10.69)
Gender (female)	17 (48.57%)	10 (39%)
Education gradea	No graduation: 2 9 school yearsa: 4 10 school yearsb: 15 High school equivalent: 14	No graduation: 0 9 school yearsa: 3 10 school yearsb: 9 High school equivalent: 17
Years of education	13.93 (3.17)	14.86 (3.92)
Family status	Single: 13 In relationship: 7 Married: 5 Divorced: 2	Single: 19 In relationship: 6 Married: 1 Divorced: 2 Widowed: 1
Living situation	Living alone: 13 With partner: 7 With parents: 4 Shared flat: 3 Assisted living: 8	Living alone: 9 With partner: 4 With parents: 6 Shared flat: 3 Assisted living: 7
Clinical variables
Main diagnosis: Schiz/schizoaff/del	27/5/3	24/2/3
Number of comorbid diagnoses	1: 14, 2: 3	1: 12, 2: 1, 3: 1
Medication: AP/AD/no medication	29/0/5	28/1/0
Chlorpromazine equivalent	734.25 (821.07)	605.11 (487.32)
Years of psychosis	12.71 (11.1)	12.31 (8.91)
PANSS P1 delusions	4.54 (1.02)	4.31 (1.12)
Acute delusions	29	27
Persecutory delusions	23	22
Delusions of reference	2	2
Delusions of grandeur	0	2
Delusions of guilt	1	0
Somatic delusions	1	0
Religious delusions	1	0
Delusions of love	1	0
Other delusions	0	1

Note: AD, antidepressive medication; AP, antipsychotic medication; M, mean; PANSS, Positive and Negative Syndrome Scale.

^aGerman Hauptschulabschluss (9 school years).

^bGerman Realschulabschluss (10 school years).

Most patients (51; 79.69%) completed the CBT-E intervention as intended (mean number of sessions in the 6-month period: 21.57, SD = 4.86, range 6–25 sessions) and 42 patients (65.63%) received additional therapy sessions after 6 months (mean: 10.5 sessions, SD = 8.55). On average, therapists rated that they spent 47.9% of the total sessions on interventions aiming at emotional stability, 35,2% on interventions aiming at emotion regulation and 16.9% on interventions aiming at negative self-schemata. The majority of patients (43; 67.19%) accepted additional sessions with a cotherapist.

The total percentage of attrition from randomization to post-CBT-E assessment was 20% (7 patients) in the TG and 20.7% (5 patients) in the WL-group (see Table 1 for more information). At the 12-month-FU assessment, the attrition rate was 34.4% (22 patients) in both groups (TG + WL).

In the TG-group, there were 2 adverse events during CBT-E: 2 patients were hospitalized, one of them committed suicide in hospital, the other continued CBT-E after hospitalization. In the WL-group, there were no adverse events during 6 months waiting time and 3 hospitalizations during CBT-E, but all of these patients continued CBT-E after hospitalization. The percentage of total adverse events for both groups was 7.8%.

The questionnaire on acceptance of CBT-E from the patients’ perspective was completed by 28 patients (43.75%) at 12-month-FU, who mostly rated CBT-E as helpful or very helpful (96.0%), the CBT-E strategies as suited or very suited to their present situation (93.7%), and the CBT-E strategies to cope with their personal problems as promising or very promising (82.2%). Also, the majority of patients (82.2%) would recommend or strongly recommend CBT-E to a friend.

Change in Primary and Secondary Outcome Variables and Putative Mediators in TG vs WL

Table 2 depicts the results of the multiple regression analysis in the ITT sample. There were no statistically significant group effects on the primary outcomes PSYRATS delusions scale and PDI grand total score. With regard to the secondary outcome variables and putative mediators, there was a significant effect of group on general psychopathology (PANSS GEN) in favor of TG (d = −0.49), but not on any of the other secondary outcome variables.

Table 2.

Change in Primary and Main Secondary Outcome Variables and Putative Mediators Between Pre- and Postassessment in the Therapy Group (TG) and Wait List Group (WL [Intent-to-Treat Sample: n = 64])

Measure	TG-Group (n = 35) Prescores	WL-Group (n = 29) Prescores	TG-Group (n = 35) Postscores	WL-Group (n = 29) Postscores	Group Differences			Effect Sized
	M (SD)	M (SD)	M (SD)	M (SD)	b unst	SE	P	d	95% CI
Primary outcome variables
PSYRATS delusions scale (BR)	12.43 (5.02)	10.69 (4.95)	8.69 (5.64)	9.66 (4.77)	−1.59	1.35	.236	−0.31	[−0.83, 0.21]
PDI-21 grand total score (P)	82.17 (66.58)	62.67 (44.37)	53.53 (42.04)	49.94 (29.94)	0.29	9.37	.975	0.01	[−0.31, 0.32]
Secondary outcome variables
PANSS POS (BR)	18.30 (5.16)	18.05 (4.85)	14.97 (3.95)	16.42 (3.60)	−1.54	0.95	.105	−0.30	[−0.67, 0.06]
PANSS NEG (BR)	14.90 (4.64)	14.02 (3.10)	13.74 (3.97)	13.68 (3.48)	−0.24	0.97	.806	−0.06	[−0.53, 0.41]
PANSS GEN (BR)	36.80 (7.15)	35.19 (4.23)	32.59 (6.84)	34.62 (6.20)	−3.00	1.53	.050	−0.49	[−0.99, 0.00]
Paranoia checklist frequency (P)	26.34 (17.77)	27.09 (18.18)	22.87 (16.27)	23.75 (15.36)	−0.60	3.72	.872	−0.03	[−0.43, 0.37]
Paranoia checklist distress (P)	30.06 (21.43)	31.80 (18.23)	23.64 (16.03)	24.75 (16.72)	−0.51	3.84	.894	−0.03	[−0.40, 0.35]
CDSS depression (BR)	7.23 (3.36)	6.61 (3.00)	6.38 (4.31)	5.48 (3.49)	0.58	1.00	.561	0.18	[−0.43, 0.79]
BDI-II sum score (P)	22.09 (11.85)	19.29 (11.10)	16.48 (10.58)	18.03 (8.73)	−2.70	2.25	.230	−0.23	[−0.61, 0.15]
RFS mean score (BR)	8.17 (1.73)	8.58 (2.08)	7.66 (1.77)	7.70 (1.45)	0.09	0.42	.838	0.04	[−0.38, 0.47]
Putative mediators
ERSQ total score (P)	45.54 (19.70)	47.88 (17.88)	57.07 (20.89)	51.62 (17.87)	6.44	4.68	.160	0.34	[−0.14, 0.82]
PTQ total score (P)	36.44 (12.80)	34.70 (11.0)	32.14 (13.48)	30.77 (10.20)	0.68	2.90	.820	0.06	[−0.41, 0.53]
BCSS Negative self (P)	10.68 (5.36)	8.74 (6.43)	8.31 (4.69)	9.22 (5.27)	−1.76	1.15	.160	−0.29	[−0.67, 0.08]

Note: BCSS, Brief Core Schema Scale; BDI-II, Beck Depression Inventory, revised version; BR, rating performed by blinded raters; CDSS, Calgary Depression Rating Scale; ERSQ, Emotion Regulation Style Questionnaire; M, mean; P, self-report questionnaires; PANSS, Positive and Negative Syndrome Scale; PANSS GEN, PANSS general psychopathology scale; PANSS NEG, PANSS negative scale; PANSS POS, PANSS positive scale; PDI, Peters et al. Delusions Inventory; PSYRATS, Psychotic Rating Symptom Scale; PTQ, Perseverative Thinking Questionnaire; RFS, Role Functioning Scale; RFS mean score = mean score of scales Social Network I, Social Network II, Work and Living.

Analyses were repeated, controlling for center and stratification (PANSS total score). The results were comparable (see Supplementary table S3). The results within the subgroup of patients with more acute delusions (PSYRATS delusional conviction score ≥3: n = 32) were comparable to results in the ITT sample (see Supplementary table S4).

Sustainability of Effects in the Combined Sample (TG and WL)

As depicted in Table 3, the primary outcome variable delusions (PSYRATS delusions scale/PDI grand total score) revealed a significant improvement from pre- to postassessment and from pre- to 12-month-FU assessment (only PSYRATS delusions scale).

Table 3.

Change in the Total Group (TG and WG) Between Pre- and Postscores and Between Pre- and 12-Month-Follow-up Scores in the Intent-to-Treat Sample (n = 64) in Primary and Main Secondary Outcome Variables and Putative Mediators

Measure	Total Group (TG + WL) Prescores M (SD)	Total Group (TG + WL) Postscores M (SD)	Total Group (TG + WL) Post 12-month-FU M (SD)	Group Differences at Post-treatment		Effect Size at Post-treatment		Group Differences at 12-Month-Follow-up		Effect Size at 12-Month-Follow-up
				t	P	d	95% CI	t	P	d	95% CI
Primary outcome variables
PSYRATS delusions (BR)	11.65 (5.00)	8.26 (5.20)	6.58 (4.92)	t(63) = 4.89	<.001	0.68	[0.38, 0.98]	t(63) = 6.34	<.001	1.01	[0.70, 1.33]
PDI-21 grand total score (P)	68.92 (55.33)	47.35 (36.49)	—	t(63) = 3.04	.002	0.39	[0.14, 0.64]	—	—	—	—
Secondary outcome variables
PANSS POS (BR)	17.74 (4.54)	15.29 (3.70)	14.31 (3.88)	t(63) = 4.03	<.001	0.54	[0.28, 0.80]	t(63) = 5.30	<.001	0.74	[0.48, 1.03]
PANSS NEG (BR)	14.45 (4.19)	13.28 (3.47)	12.63 (3.46)	t(63) = 1.82	.070	0.28	[−0.02, 0.58]	t(63) = 2.96	.003	0.44	[0.15, 0.73]
PANSS GEN (BR)	36.36 (6.77)	31.96 (5.90)	28.55 (6.45)	t(63) = 4.71	<.001	0.65	[0.38, 0.92]	t(63) = 7.15	<.001	1.15	[0.84, 1.47]
Paranoia checklist frequency (P)	25.36 (16.76)	21.43 (14.65)	—	t(63) = 1.96	.050	0.23	[0.01, 0.47]
Paranoia checklist distress (P)	27.87 (19.57)	25.30 (14.65)	—	t(63) = 1.09	.276	0.13	[−0.10, 0.37]
CDSS total score (BR)	6.50 (3.50)	5.70 (3.90)	5.25 (4.59)	t(63) = 1.45	.148	0.23	[−0.08, 0.54]	t(63) = 2.06	.039	0.36	[0.02, 0.70]
BDI-II sum score (P)	20.55 (10.70)	13.53 (9.11)	—	t(63) = 5.56	<.001	0.66	[0.42, 0.89]
RFS mean score (BR)	7.98 (1.61)	7.92 (1.56)	8.59 (1.46)	t(63) = 0.24	.811	−0.04	[−0.25, 0.32]	t(63) = −1.11	.014	−0.38	[−0.69, −0.08]
Putative mediators
ERSQ total score (P)	47.73 (18.90)	51.90 (22.39)	57.07 (20.89)	t(63) = −1.29	<.001	−0.15	[−0.23, −0.07]	—	—	—	—
PTQ total score (P)	33.86 (11.97)	29.33 (12.83)	32.14 (13.48)	t(63) = 2.66	.008	0.38	[0.10, 0.66]
BCSS Negative self (P)	10.17 (5.34)	7.60 (4.29)	8.31 (4.69)	t(63) = 3.78	<.001	0.48	[0.23, 0.73]	—	—	—	—

Note: BCSS, Brief Core Schema Scale; BDI-II, Beck Depression Inventory; BR, rating of blind raters; CDSS, Calgary Depression Rating Scale; ERSQ, Emotion Regulation Style Questionnaire; FU: follow-up; M, mean; Number of meetings/week, number of contacts with friends and family; P, self-rating questionnairess; PANSS, Positive and Negative Syndrome Scale; PANSS GEN, PANSS general psychopathology scale; PANSS NEG, PANSS negative scale; PANSS POS, PANSS positive scale; PDI, Peters et al. Delusions Inventory; PSYRATS, Psychotic Rating Symptom Scale; PTQ, Perseverative Thinking Questionnaire; RFS, Role Functioning Scale; RFS mean score, mean score of RFS scales Social Network I, Social Network II, Work and Living; SWLS, Satisfaction with life scale.

With regard to the observer-rated secondary outcome variables (assessed at pre- and posttherapy and 12-month-FU), there was significant improvement between pre- and both post- and 12-month-FU assessment in positive symptoms and general psychopathology (PANSS POS, PANSS GEN), whereas other outcomes improved only between pre- and 12-month-FU assessment (PANSS NEG, CDSS, RFS). Concerning self-rated outcome variables (assessed at pre- and posttherapy assessment), there was a significant improvement in most of the secondary outcome variables (ie, Paranoia Checklist frequency and BDI-II, but not Paranoia Checklist distress), and putative mediators (ERSQ, PTQ, BCSS negative self-schema).

Exploration of Receiving Additional Sessions by Cotherapists

There were no significant differences between the groups who accepted (n = 43) vs did not accept a cotherapist (n = 21) neither in sociodemographic and clinical baseline (see Supplementary table S2) nor in primary and secondary outcome variables and putative mediators at postassessment and 12-month-FU assessment (see Table 4).

Table 4.

Change Between Pre- and Postscores and Pre- and 12-Month-Follow-up (FU) Scores in Patients With and Patients Without a Cotherapist in the Intent-to-Treat (ITT) Sample (n = 64) in Primary and Main Secondary Outcome Variables and Putative Mediators

Measure	Patients with co-therapist Pre-scores (n = 43) M (SD)	Patients without co-therapist pre-scores (n = 21) M (SD)	Patients with co-therapist post-scores (n = 43) M (SD)	Patients without co-therapist post scores (n = 21) M (SD)	Patients with co-therapist 12-months- FU scores (n = 43) M (SD)	Patients without co-therapist 12-months-FU scores (n = 21) M (SD)	PT/ FU	Differences			Effect size
								b unst.	SE	p	d	CI
Primary outcome
PSYRATS del. (BR)	11.99 (4.81)	10.95 (5.32)	9.17 (5.05))	6.41 (4.99)	7.26 (5.13)	5.19 (4.13)	PT	2.36	1.39	.089	0.47	[− 0.07, 1.01]
							FU	1.83	1.40	.189	0.37	[−0.18, 0.91]
PDI-21	62.86 (36.55)	81.33 (79.82)	49.23 (41.77)	43.51 (21.54)	n. a.	n. a.	PT	9.74	0.08	.309	0.18	[− 0.16, 0.51]
Secondary outcomes
PANSS POS (BR)	17.40 (4.69)	18.42 (4.15)	15.28 (3.94)	15.30 (3.17)	14.61 (4.36)	13.72 (2.52)	PT	0.35	1.0	.727	.008	[− 0.36, 0.51]
							FU	1.23	1.06	.247	0.27	[− 0.19, 0.73
PANSS NEG (BR)	14.61 (3.97)	14.13 (4.59)	13.39 (3.61)	13.06 (3.15)	12.82 (3.77)	12.22 (2.66)	PT	0.24	1.0	.813	0.06	[− 0.41, 0.53]
							FU	0.50	1.01	.624	0.12	[− 0.36, 0.59]
PANSS GEN (BR)	36.58 (7.15)	35.89 (5.90)	32.37 (6.24)	31.13 (4.92)	29.05 (7.27)	27.51 (4.22)	PT	0.99	1.55	.522	0.15	[− 0.30, 0.59]
							FU	1.41	1.82	.439	0.21	[− 0.32, 0.73]
PCL frequency	26.00 (16.10)	24.05 (17.97)	22.57 (16.13)	19.07 (10.63)	n. a.	n. a.	PT	2.65	3.52	.452	0.16	[− 0.25, 0.57]
PCL distress	26.64 (18.32)	30.39 (21.70)	26.51 (18.02)	22.82 (11.53)	n. a.	n. a.	PT	5.22	3.91	.182	0.27	[−0.12, 0.66]
CDSS (BR)	6.91 (3.62)	5.67 (3.09)	5.76 (4.32)	5.57 (2.88)	5.61 (5.25)	4.51 (2.61)	PT	− 0.40	1.09	.710	− 0.12	[− 0.73, 0.49]
							FU	0.44	1.3	.733	0.13	[− 0.60, 0.86]
BDI-II	20.39 (10.59)	20.89 (10.93)	14.21 (10.19)	12.14 (6.12)	n. a.	n. a.	PT	2.30	2.15	.284	0.22	[− 0.18, 0.61]
RFS mean (BR)	7.77 (1.44)	8.39 (1.85)	7.77 (1.33)	8.22 (1.91)	8.38 (1.57)	9.03 (1.08)	PT	−0.24	0.42	.570	− 0.15	[− 0.66, 0.37]
							FU	− 0.55	0.42	.189	− 0.34	[− 0.84, 0.17]
Putative mediators
ERSQ total	47.37 (19.35)	48.49 (17.91)	50.30 (21.41)	55.06 (23.97)	n. a.	n. a.	PT	−1.03	6.20	.869	− 0.05	[− 0.70, 0.59]
PTQ total	34.42 (10.65)	32.68 (14.28)	30.76 (13.73)	26.40 (10.16)	n. a.	n. a.	PT	3.56	3.23	.270	0.30	[− 0.23, 0.83]
BCSS Neg. self	10.50 (5.20)	9.49 (5.54)	7.94 (4.36)	6.90 (4.05)	n. a.	n. a.	PT	0.70	1.13	.536	0.13	[− 0.28, 0.54]

Note: BCSS neg. self, Brief Core Schema Scale negative self scale; BDI-II, Beck Depression Inventory sum score; BR, rating of blind raters; CDSS,Calgary Depression Rating Scale total score; ERSQ total, Emotion Regulation Style Questionnaire total score; FU: comparison between pre- and 12-months-Follow-up scores; M, Mean; n. a., not assessed; PANSS, Positive and Negative Syndrome Scale; PANSS GEN, PANSS general psychopathology scale; PANSS NEG, PANSS negative scale; PANSS POS, PANSS positive scale; PCL, Paranoia checklist; PDI-21, Peters et al. Delusions Inventory short version grand total score; PSYRATS del., Psychotic Rating Symptom Scale delusions scale; PT, Comparison between pre- and post-scores; PTQ total, Perseverative Thinking Questionnaire total score; RFS mean, Role functioning scale; RFS mean score, mean scores of RFS scales Social Network I, Social Network II, Work and Living; SD, Standard deviation; unst., unstandardized.

Sensitivity Analysis and Other Additional Analysis

Results of the comparison between the TG and the WL-group at postassessment in the per-protocol sample are depicted in Supplementary table S5 and revealed comparable results to the results in the ITT sample. In addition, changes in pre- to postassessment and pre- to 12-month-FU assessment in both groups are depicted in Supplementary table S6 and also revealed comparable results. Analysis in other prespecified outcome variables are reported in Supplementary tables S7 and S8.

Discussion

The aim of the present study was to test the feasibility and acceptance of a CBT-E and to provide initial evidence for its efficacy. In general, the intervention proved to be feasible and the majority of the patients perceived CBT-E as a suitable approach to their problems. Despite significant changes from pre- to post-treatment and pre- to 12-month-FU on most outcome variables, there was no additional benefit of CBT-E over the WL-group regarding the primary outcome delusions, the majority of the secondary outcome variables or the putative mediators. The only outcome for which CBT-E showed a benefit was general psychopathology. Also, accepting additional support from a cotherapist did not result in any statistically significant advantage.

We will now briefly discuss the findings drawing on a comparison to 2 TAU-controlled trials (RCTs) that also targeted ER (worrying¹³) and negative self-schemata¹⁴ and thus similar factors to those targeted in our study. Given that our trial did not include a third arm with “classic” CBTp, we will also compare the findings with an earlier TAU-controlled RCT⁴⁶ that tested the effect of “classic” CBTp on overall psychopathology and shared several characteristics of the present trial (eg, recruitment pathways, number of sessions, qualification of therapists, and integration within the German health care insurance system), which renders it an interesting comparator.

Feasibility and Acceptance of CBT-E

Using somewhat more lenient criteria than the “targeted” studies,^13,14 that required a PSYRATS delusional conviction scale score of at least 3 and scores above predefined cutoff scores in the putative mediators, we were able to recruit a sufficient number of patients (64 patients in 2 centers), which is a positive foundation for a future more rigorous large-scale RCT.

Patients who stayed in CBT-E received slightly over 20 sessions on average, which is significantly more than the 5.5 sessions in the “targeted” studies,^13,14 but in the range of 15–25 sessions that have been identified as the minimal to optimal number of sessions required in CBTp.⁶³

Furthermore, most patients agreed to the additional support offered by cotherapists (67%). Many patients (66%) also accepted additional sessions after the postassessment at 6 months, which was similar to the CBTp study.⁴⁶ However, it needs noting that a slightly lower number of CBT-E patients rated the intervention as promising than in the CBTp trial (82.2% vs 95%⁴⁶) and less patients would recommend the treatment to a friend (82.2% vs 95%⁴⁶) which indicates that a focus on putative mediators rather than delusional beliefs does not necessarily increase acceptance, albeit this difference may be partly explicable by the lower efficacy of CBT-E in comparison to the CBTp trial.⁴⁶

Another indicator with relevance to feasibility and acceptance are the dropout rates, in particular those that occurred during the intervention phase. Among those randomized to the CBT-E intervention (n = 35), dropout occurred most frequently before the intervention had started. Reasons for this early dropout may be that the assessment phase with a total duration of 5 h (due to the necessity of trial-related assessments and assessments required by the health insurances) was perceived as distressing or unhelpful. The overall percentage of dropouts at postassessment (20%) was higher than both the “targeted” studies (0%¹⁴–8.3% dropouts at 8 weeks; 0%¹⁴–10.6% dropouts at 24-week-FU) and the CBTp trial⁴⁶ (8.75% after 6 months), but nevertheless in the range of what is found in CBTp (mean dropout rate: 18.8%⁴).

In the final indicator of feasibility and acceptance, adverse events, it needs to be noted that there was one very severe adverse event in the intervention group, namely a suicide during hospitalization. However, based on information concerning the timing and circumstances of the rehospitalization and the suicide, neither of these events were considered to be an adverse reaction to CBT-E. Other than that, the number and type of adverse events during the intervention in our study (7.8%) was overall comparable with both the “targeted” studies (10%¹⁴; 5.43%¹³) and the CBTp study (10%⁴⁶). Thus, taken together, CBT-E was feasible and acceptable to a similar, albeit slightly lesser extent than the “targeted” interventions and CBTp.

Efficacy of CBT-E

With regard to the efficacy of CBT-E, we did not find a statistically significant effect on the primary outcome delusions. Numerically, our effect size suggests a small effect on delusions (d = −0.31) that is comparable with average effects of CBTp on delusions reported in meta-analyses (d = 0.27⁷; g = 0.36⁸), but is clearly less pronounced than the effect sizes of the “targeted” studies (−0.91¹⁴; −0.49¹³). The lower effect size in comparison to the “targeted” studies may be partly explicable by the fact that we included patients with a lower threshold of delusions (PANSS item P1 score ≥2). Thus, our patient sample presented less delusional beliefs in comparison to the “targeted” studies that used a score of at least 3 in the PSYRATS delusional conviction scale as threshold and presented more pronounced scores in the PSYRATS delusions scale (our study: 12.43/WL: 10.69 vs 18.7/18.0¹⁴ and 18.4/18.1¹³). A subgroup analysis of the intervention effects on delusions in patients with more acute delusions (TG: n = 19, WL: n = 13) provided no indication for a stronger therapy effect in this group. However, a higher power is required for this type of analysis. Concluding, while our numerical effect sizes regarding change in delusions are positive, the small sample sizes might have prevented us from detecting statistically significant change in delusions.

The significant effect on the secondary outcome variable general psychopathology in our study was of moderate magnitude (d = −0.49; P = .05) and thus comparable to the effect on the PANSS general psychopathology subscale in the CBTp trial⁴⁶ and the “targeted” study on worrying.¹³ It also corroborates the consistent evidence for an effect of CBTp on general psychopathology across most meta-analyses of CBTp⁶ and fits in well with the overall effects for CBTp in meta-analysis. However, given the absence of effects in the other secondary outcomes and the fact that it verged on significance, this effect may also be a chance finding.

The absence of significant effects on all putative mediators (ER, negative self-schemata) was unexpected. Again, the “targeted” study¹³ that aimed at worrying did a better job at reducing it (d = −0.46¹³). However, the “targeted” study that aimed at self-esteem¹⁴ also did not produce a significant effect on the BCSS negative self-schemata, which appear to be notoriously difficult to change.

Taken together, the effect sizes produced by CBT-E were small for delusions and ranged from absent to moderate for the secondary outcomes and mediators. Thus, the hypothesis that this emotion-focused variant of CBTp would be more successful in reducing delusions and that this would be achieved by producing change in the putative mediators ER and negative self-schemata was not confirmed.

There are a number of reasons that could explain the absence of clear intervention effects. One is that our large set of interventions was too heterogeneous. The selection of CBT-E interventions took the patients’ individual goals regarding emotional stability and ER into account and was less focused on one precise therapy goal, which might have caused patients to learn a bit of everything, but nothing properly. In addition, based on therapists’ ratings, almost half of the total CBT-E sessions focused on increasing emotional stability (eg, daily structure, mastery and pleasure techniques), whereas fewer sessions focused on ER and negative self-schemata. Similar to recommendations for post-traumatic stress disorder where therapists are advised to focus less on stabilization, but rather to start exposure techniques more early during treatment,⁶⁴ our intervention might have been stronger if it had placed a stronger focus on ER strategies and self-esteem earlier in therapy.

The better effects found in the “targeted” interventions^13,14 indicate that selecting patients with a minimum level of pathology in the targeted area (worrying) and providing an intervention that focuses solely on that target could be the more promising approach for interventions focusing on ER. In addition, the “targeted” studies achieved this with fewer sessions. Possibly, brief and focused interventions are also beneficial in terms of therapists’ adherence to the manual.

Limitations

The study was powered to detect a moderate effect on delusions, but it lacked statistical power to test significance of the unexpectedly small effects. Thus, the nonsignificance of the small effects is hard to interpret.

It needs noting that our pilot study had some features of an effectiveness study, such as the interventions being offered by therapists with high case load, and the fact that patients were not paid for additional diagnostic appointments. These features were a consequence of the absence of external funding and might have had a negative impact on the study’s results, as effectiveness studies typically show smaller effects.^65–68 Nevertheless, most characteristics of the study support its categorization as a pilot efficacy study such as predefined inclusion criteria, randomization, a wait-list condition, blinded assessment, a manualized treatment and sophisticated training and supervision of the therapists and cotherapists.

Other limitations include the fairly high rate of missing values due to high dropout rates of patients at postassessment (20%) and FU (35%). We tried to remedy this problem by using an ITT analysis and an adequate number of imputations based on the ratio of missing data as recommended by Graham et al^60,69 and by performing sensitivity analysis in the per-protocol sample.

In addition, diagnostic assessments were performed by patients’ therapists and videotaped for the blinded rates which might have produced assessment bias. Moreover, therapists were trained and received extensive case and manual supervision, but their competency regarding CBT-E interventions was not tested and the study did not involve ratings of therapists’ adherence to the manual. Furthermore, receiving support by an additional cotherapist was based on patients’ choice and patients were not randomized to this condition. Thus, the lack of benefit of sessions with a cotherapist might be prone to selection bias. Finally, for reasons of comparability with other studies on CBTp, we used the 3 PANSS subscales as proposed in the manual.³⁷ This may have come at the disadvantage of lower construct validity, which has been found to be better in 5-factor solutions.^70,71

Implications of the Study Results on Future Studies

The fact that the results did not meet the expectations raises the question of whether pursuing the CBT-E route further is worthwhile. However, given that the rationale for the approach that we outlined in the introduction still stands, it seems a little bit too early to fully abandon the approach as such on the basis of this 1 pilot. Given that the interventions were not successful in changing the proposed mediators (ER and negative self-schemata) it needs pointing out that our results do not suggest that using a “targeted” approach as such is not worthwhile. Our results rather suggest that interventions were not closely focused enough on the intended “targets” ER and negative self-schemata.

Thus, it remains still an open question whether change in these targets might improve delusional beliefs. Concluding, it seems necessary to reflect on how the approach can be improved to achieve a stronger effect on the putative mediators (ER and self-esteem). This may be reached by reducing the number of different interventions in the manual, in favor of a clearer focus on ER and self-schemata and less focus on stabilization.

Given that the additional support provided by the cotherapists produced no further benefit and the organizational effort associated with training and supervising the cotherapists was high, further pursuing the cotherapist approach does not seem worthwhile. Alternative ways to support patients in implementing new ER skills could be to use additional online interventions or blended interventions^72,73 or to work more stringently with homework assignments. To meet this aim, it might also be necessary to intensify therapist training, as more intensive training has been found to lead to better therapist competence and outcomes^74,75 and also to assess manual adherence.

Conclusions

Taken together, the present pilot study of CBT-E revealed no significant effect on the primary outcome delusions and a significant effect in only one of the secondary outcome variables: general psychopathology and no statistically significant advantage of CBT-E on the putative mediators. However, the pattern of findings—overall acceptability, small effect sizes for delusions, moderate effect sizes on psychopathology—mirrors the pattern found for generic CBTp. Future work should test whether improvements in terms of selection of interventions and training can produce stronger effects on the putative mediators of delusions: ER and negative self-schemata.

Funding

This work was supported by a grant awarded by the cooperation contract between the universities of Marburg and Gießen (Germany) to S.M. (Project no. 33/2012 MR), as well as a travel grant by the Behring-Röntgen-Stiftung Marburg to S.M. The funders had no role in study design, data collection, and analysis and preparation of the manuscript.

Authors’ Contribution

S.M. and T.M. conceived the study, planned the trial’s design and are the chief investigators, S.M. and C.H. performed statistical analysis and C.H., C.N., W.R., and T.K. cowrote the manuscript. All authors read and approved the final manuscript.

Data Availability

Data of the study and all study material will be available from the corresponding author after a reasonable request.