1.1 Psychotic symptoms

Psychotic symptoms are characterised by delusions, hallucinations, and paranoia (positive symptoms) as well as difficulties with thinking, blunted emotions, and low motivation (negative symptoms) (Jablensky, 2010). Psychotic symptoms can occur in a range of disorders including schizophrenia spectrum disorders, bipolar disorder, major depressive disorder, dissociative disorders and borderline personality disorder and can therefore be considered a transdiagnostic phenomenon (Buckley et al., 2008). The life expectancy of people with psychotic symptoms is significantly reduced (Saha et al., 2007) due to suicide and greater health problems (Hannerz et al., 2001; Yuen et al., 2014).

1.2 Psychosis and trauma

Exposure to psychological trauma has consistently been associated with increased risk of psychotic experiences. The evidence fulfils the Bradford Hill criteria (Hill, 1965) supporting the hypothesis of a causative relationship between trauma and psychosis, including temporal relationships (Kelleher et al., 2013), dose-response relationships (Croft et al., 2019) and plausible biological mechanisms (Howes & Murray, 2014).

Traumatic experiences can also give rise to post-traumatic stress disorder (PTSD), and there is a growing body of evidence showing a relationship between PTSD and psychotic symptoms in people with psychosis who have experienced trauma (Bloomfield et al., 2021). PTSD is a risk factor for the development of psychotic symptoms (Okkels et al., 2016), and around 39% of people with psychosis experience concurrent PTSD (Mueser et al., 2010). Within trauma survivors, auditory hallucinations have been proposed to be a type of posttraumatic intrusion related to a traumatic memory (Peach et al., 2018; Steel, 2015). This is consistent with studies that have shown that hallucinatory content is often linked to experiences of trauma (Hardy et al., 2005; Onyeama et al., 2011; Peach et al., 2020). Within this framework, an intrusive trauma memory may not be experienced as a memory and is instead misattributed in a psychotic way (e.g. as a voice). Indeed, there are broad similarities between dominant cognitive theories of PTSD (Ehlers & Clark, 2000) and trauma-induced psychotic symptoms (Morrison et al., 2003).

1.3 Trauma-focused psychotherapies

Given the links between traumatic experiences, psychotic symptoms and PTSD, there is growing interest in trauma-focused psychotherapies (TFPT) for psychotic symptoms. TFPT are a family of therapies developed to treat PTSD that explicitly focus on reprocessing memories of traumatic experiences (Schnurr, 2017). Some TFPT may utilise cognitive techniques only, some may use exposure, and some may use a combination of the two. Broadly, TFPT that utilise exposure are thought to work by promoting emotional habituation and reprocessing of traumatic memories via repeated exposure to the traumatic event and related cues. Cognitive therapies such as trauma-focused CBT (TFCBT) additionally explicitly focus on restructuring peri-traumatic (such as ‘I'm going to die’) and post-traumatic (such as ‘I should have coped better’) appraisals, often utilising learnings gained from exposure. In the UK, National Institute for Clinical Excellence guidelines recommend the use of TFCBT and eye movement desensitisation and reprocessing (EMDR) in the treatment of PTSD (National Institute for Health and Clinical Excellence, 2018). Though the literature remains in its infancy, two recent reviews found that TFPT can safely reduce PTSD symptoms in those with psychosis (Sin & Spain, 2016; Swan et al., 2017).

1.4 The need for improved research and outcomes

There is little research into whether TFPT can reduce psychotic symptoms themselves. The only review to date (Brand et al., 2018) found small effects for TFPT on positive symptoms of psychosis. There remains therefore a pressing need for more research to improve treatments and outcomes for people with psychotic symptoms and trauma histories. We sought to address this by systematically reviewing the current evidence for the efficacy, tolerability, and acceptability of TFPT on psychotic symptoms. To differentiate this review from previously published work and to draw together disparate research, we have not limited the review to any specific type of TFPT. Additionally, as psychotic symptoms in the context of trauma are likely a transdiagnostic phenomenon (Buckley et al., 2008), we have not limited the review to any specific diagnosis.

2.1 Search strategy and selection criteria

We pre-registered our review with PROSPERO (CRD42020202135). We followed the preferred reporting items for systematic reviews and meta-analysis guidelines (PRISMA) (Moher et al., 2009). We included any study of the efficacy of a TFPT with a quantitative outcome measure for psychotic symptoms. We defined TFPTs as any psychological intervention from any modality of psychotherapy that had an explicit focus on past traumatic memories and/or was described as ‘trauma-focused’. We distinguished between TFCBT that would be in-line with the principles of Ehlers & Clark (2000) by facilitating memory reprocessing through exposure, and trauma-informed CBT (TICBT) that did not include exposure. Interventions could be delivered in any setting in a group or individual format. We defined psychotic symptoms as hallucinations, delusions and/or paranoia.

We included studies on adults over the age of 18. Otherwise, we placed no limits on the population to be included. We included any study design that offered a TFPT, including case reports, case series and randomised controlled trials. We did not limit our criteria to any specific diagnoses. For example, dissociative identity disorder was included due to research highlighting elevated levels of psychotic symptoms (Foote & Park, 2008). We excluded studies that were not reported in English, were not published in peer-reviewed journals, and studies of non-clinical populations.

We used Ovid to systematically search medical and psychological databases (MEDLINE and PsycINFO) and ProQuest to search PTSDPubs from the earliest possible date to the date of the search. We searched a psychotic symptoms term (e.g. hallucinat* OR delus*) AND a trauma-focused term (e.g. trauma* OR neglect*) AND a psychotherapy term (e.g. ‘Exposure therapy’ OR EMDR). See Supporting Information S1 for the full list of search terms by database and screening methodology.

Reviewer 1 (R1) imported articles generated from the search into a reference management software (EndNote 20) (The EndNote Team, 2013) and checked for duplicates which were removed from the review. R1 and reviewer 2 (R2) then screened for inclusion by title and abstract first, then by reading the full text, with any disagreements resolved by discussion, facilitated through the Rayyan platform (Ouzzani et al., 2016).

2.2 Data analysis

We defined the primary outcome a priori as quantitatively measured psychotic symptoms, including global measures of psychotic symptoms and measures of hallucinations, delusions, or paranoia. Our secondary outcomes were measures of other domains of psychopathology (depression, anxiety and PTSD) and social functioning. We assigned a level of evidence (OCEBM Levels of Evidence Working Group, 2011) with case reports assigned a level of evidence of 5. For randomised controlled trials (RCT), we assessed risk of bias using the Cochrane Risk of Bias 2 tool (Sterne et al., 2019). For case series, we used the Quality Appraisal Tool for Case Series Studies (Institute of Health Economics, 2014). For case reports, we used the Joanna Briggs Institute Checklist for Case Reports (Moola et al., 2017). R1 and reviewer 3 (R3) undertook a risk of bias and quality assessment. Any discrepancies between reviewers were resolved through discussion and by consensus. If agreement could not be reached, reviewer 5 (R5) was consulted to resolve this.

We extracted data manually from each study paper. R1 extracted the following data for each study: study design, n, participant characteristics and clinical presentation, location and setting of the treatment, the type of trauma participants had experienced, intervention type and dose, medications prescribed to participants, control or comparison, primary outcomes, secondary outcomes, adverse events (identified as adverse events reported by the authors and/or symptom exacerbation) and dropout rates. We grouped studies by therapeutic modality.

We considered findings statistically significant when the p-value was below 0.05. Where possible, we calculated an effect size (Hedges' g; see Supporting Information S1 for formulae) for each study at each time point for each outcome of interest. We chose Hedges' g as it has superior properties to Cohen's d with small sample sizes (Cumming, 2012).

We undertook a narrative synthesis of quantitative outcomes following SWiM (synthesis without meta-analysis) guidance (Campbell et al., 2020), grouping by modality of psychotherapy, including any relevant information about tolerability or acceptability of these interventions. We followed the GRADE guidance for clinical recommendations (Guyatt et al., 2008).

3.1 Trauma-informed CBT

Two of three TICBT studies (Mueser et al., 2015; Steel et al., 2016) were RCTs following the same protocol teaching cognitive restructuring to challenge trauma-related thoughts and beliefs, meaning no exposure techniques were used (Ehlers & Clark, 2008). TICBT was not superior to the respective control groups in reducing positive psychotic symptoms. The third study (Trappler & Newville, 2007) was a case series using Cloitre's Skill Training in Affect Regulation preparatory work (Cloitre et al., 2002). It found significant decreases in measures of overall psychotic symptoms, delusions, and paranoia. Only paranoia also significantly decreased in a matched group undergoing supportive psychotherapy.

Regarding secondary outcomes, one of the two controlled studies (Mueser et al., 2015) found a small but statistically significant decrease in PTSD symptoms as compared to control. The non-controlled study (Trappler & Newville, 2007) found a significant effect for its treatment programme on PTSD symptoms, with no effect found in its matched supportive psychotherapy group. There were no significant effects in these studies on measures of depression, anxiety nor quality of life.

3.2 Trauma-focused CBT

The TFCBT studies utilised a variety of exposure techniques, such as imagery rescripting and reliving. To qualify as TFCBT, it is necessary that therapeutic techniques intended to process and modify unhelpful peri- or post-traumatic thoughts and feelings, such as cognitive distortions, guilt, and shame, are included. Each included TFCBT study was non-controlled. Due to lack of comprehensive raw data and statistical analysis conducted by the authors, effect sizes were not able to be calculated for these studies. One (Keen et al., 2017) found decreases in auditory hallucinations and delusions scores, 29% and 43% respectively of which remained clinically significant at 6-month follow-up. Of the case reports, two (Callcott et al., 2004; Ward-Brown et al., 2018) found decreases in their measures of psychotic symptoms. The final included case report (McCartney et al., 2019) found a reliable but not clinically significant decrease in auditory hallucinations.

Regarding secondary outcomes, PTSD symptoms decreased in all TFCBT studies. Consistent decreases in depression and anxiety measures were also found. Only one study (Ward-Brown et al., 2018) included a quality-of-life scale, reporting a substantial improvement.

3.3 Eye movement desensitisation and reprocessing

All the EMDR studies included in this review used the standard eight-phase protocol (Shapiro, 2001). This protocol focuses mostly on exposure, although one phase focuses on ‘installing’ more helpful thoughts and cognitions. Whilst EMDR had the greatest number of included studies, the evidence of effectiveness was mixed. One RCT (de Bont et al., 2016) found EMDR was superior to the control in reducing paranoia following treatment as well as across all time points. EMDR was also associated with greater likelihood of remission from a psychotic disorder following treatment, but this was no longer statistically significant at 6-month follow-up. This contrasts with the other included RCT (Kim et al., 2010), which found no statistically significant differences on psychotic symptoms between EMDR and either of its two control conditions. The case series and case reports provide mixed evidence of small effects on specific symptom domains, particularly hallucinations (Slotema et al., 2019; van den Berg & van der Gaag, 2012) and overall symptoms (van den Berg & van der Gaag, 2012; Yasar et al., 2017). No study found a specific effect on paranoia.

Regarding secondary outcomes, no controlled study reported on PTSD symptoms. Four studies found large decreases on PTSD symptom measures (de Bont et al., 2013; Slotema et al., 2019; van den Berg & van der Gaag, 2012; Yasar et al., 2017). Regarding depression, the two controlled studies found no superior effect for EMDR over the control following treatment (de Bont et al., 2016; Kim et al., 2010). Uncontrolled studies found decreases in depression measures (van den Berg & van der Gaag, 2012; Yasar et al., 2017). One controlled study measured anxiety finding no effect for EMDR over the controls (Kim et al., 2010). Two uncontrolled studies reported decreases in anxiety measures, reaching significance when significance testing was possible (van den Berg & van der Gaag, 2012). For quality-of-life measures, one controlled study found no superior effect for EMDR over control following intervention or 6-month follow-up (de Bont et al., 2016), whilst one uncontrolled study found a significant improvement following treatment and 3-month follow-up (de Bont et al., 2013).

3.4 Prolonged Exposure

Two studies used PE (Foa et al., 2007), a protocol which focuses solely on exposure without a cognitive component. One RCT (de Bont et al., 2016) found a superior effect for PE over control in reducing paranoia following treatment, which sustained at 6 and 12-month follow-up (van den Berg et al., 2018). In addition, PE participants were significantly more likely to be in remission from a psychotic disorder at follow-up. No significant impact for PE over the control for auditory hallucinations was found. A study that combined participants that had undergone PE or EMDR as one treatment group has been reported above (de Bont et al., 2013).

Regarding secondary outcomes, PE resulted in statistically significant decreases on depression scores following treatment and 6-month follow-up compared to the control, but not on social functioning scores (de Bont et al., 2016).

3.5 Other interventions

One case series investigated the effects of ‘phasic trauma treatment’ for Dissociative Identity Disorder (Brand & Loewenstein, 2013) (incorporating an exposure component), described in the Guidelines for Treating Dissociative Identity Disorder in Adults (International Society for the Study of Trauma and Dissociation, 2011). They found this treatment had a significant impact on a hearing voices item, with effect sizes growing larger the longer the person was in treatment.

There were no secondary outcomes reported relevant to this review within this study.

One case series investigated the effects of imagery rescripting (Paulik et al., 2019). This intervention had a beneficial impact on voice hearing, with auditory hallucination distress and frequency, and belief in voice malevolence seeing significant decreases following treatment, sustained at 3-month follow-up. However, belief in voice omnipotence did not significantly decrease following treatment.

This study reported on PTSD symptomology, finding a significant decrease following treatment. No significant effect was found on depression or anxiety measures (Paulik et al., 2019).

One case series investigated the effects of a one-off video interview regarding a personal experience of the Holocaust, with a focus on details about losses and grief experienced (Strous et al., 2005). There was no effect on psychotic nor PTSD symptoms.

One case report (Arens, 2014) looked at the effectiveness of trauma management therapy (Turner et al., 2005) (TMT) in a combat veteran. TMT incorporates in vivo and virtual reality assisted imaginal exposure sessions together with group social and emotional coping skills. The number of self-reported auditory and visual hallucinations declined from 7 to 1, and 2 to 0 respectively by 3-month follow-up. Two measures of PTSD symptomology showed a decrease at 3-month follow-up. Anxiety scores had also decreased at 3-month follow-up.

3.6 Tolerability and acceptability

Regarding tolerability, 41% (7/17) of all studies and 100% (3/3) of the non-exposure based protocols did not report on adverse events or harm. Of the exposure-based studies, 29% (4/14) did not report on harm, 36% (5/14) reported that there were no instances of adverse events or harm, with one (de Bont et al., 2016) reporting that fewer participants in the active condition (EMDR) experienced symptom exacerbation or adverse events than in the control condition. 29% (4/14) reported a brief symptom exacerbation early in treatment that resolved later, sometimes with brief (i.e. one session or one conversation) additional support provided (e.g. psychoeducation about increased arousal when starting a new intervention). Only one reported exacerbation that may not have resolved; in this case stress and an increase in instability leading to dropouts in 4/47 (9%) of their participants, with only 32/47 (68%) completing EMDR treatment (Slotema et al., 2019).

Regarding acceptability, 17% (2/12) of studies did not report on dropouts (five studies have not been included in this calculation due to being case reports). We amalgamated dropout rates within treatment modality (excluding again case reports) and found a dropout rate of 22% (26/119 participants) within TICBT studies, 0% (0/9) within TFCBT studies, 23% (34/147) within EMDR studies, 24% (14/58) within PE studies and 3% for other interventions (1/36 participants). In total, this results in a dropout rate of 22% (26/119 participants) for non exposure-based protocols and 20% (49/250 participants) for exposure-based protocols.