Introduction

Schizophrenia is a severe mental disorder with relevant consequences for the individual and society, being ranked as one of the most debilitating disorders worldwide [2]. The disease burden for patients, relatives and society is dramatic [3, 4].

Since the first symptoms of schizophrenia typically appear in the age between 20 and 30 years, this has big impact on the life-perspectives of the young adult patients, who often do not complete their education, have difficulties in finding an occupation as well as to form relationships [1]. It is estimated that 80–90% of patients are unemployed [5]. These high rates of loss of productivity and unemployment lead to high costs for the society; with estimated total costs of more than 93 billion Euros per year, schizophrenia is among the most expensive illnesses in the EU [4].

To support patients and their families to face such challenges, it is important to address not only the symptoms of the disorder, but also the functioning of the individuals and their ability to be active members of the society. The concept of functioning is not limited to employment and economical contribution, but includes social behavior, participation and activities of daily living and self-care [6]. These aspects are included in most of the rating scales to measure functioning such as Global Assessment of Functioning (GAF) [7], Personal and Social Performance scale (PSP) [8], or the Social Functioning Scale (SFS) [9].

The importance of functioning as a therapeutic goal in schizophrenia is recognized in the scientific community [10], as well as explicitly expressed in clinical guidelines from National Institute for Care and Health Excellence (NICE) [11], Scottish Intercollegiate Guidelines Network (SIGN) [12], German Association for Psychiatry, Psychotherapy and Psychosomatics (DGPPN) [13] and other national and international guidelines.

Nevertheless, the evidence on the effects of psychological interventions on functioning in schizophrenia is very limited, and mostly focused on cognitive behavioral therapy (CBT). A network meta-analysis investigating psychological interventions in the acute phase of schizophrenia found that, on 53 included studies, 40 were focused on CBT, and only 20 had measured functioning [14]. Laws et al. conducted a meta-analysis investigating the effects of CBT on functioning, distress and quality of life [15]. Based on 25 RCTs, they found an SMD of 0.25 (95% CI 0.14–0.33) for CBT compared to control conditions [treatment as usual (TAU) or other psychological interventions]. Two Cochrane reviews by Jones et al. investigated the effects of CBT compared to TAU and compared to other psychosocial interventions and considered functioning among other outcomes. However, they provided effects for each rating scale separately and for different time points separately, resulting in analyses that include very few studies each and do not inform on the general picture [10, 16].

The evidence on other psychological interventions such as creative therapies is limited to Cochrane reviews that investigated their effect in many outcomes, but present only scattered data, separating data measured with different rating scales and at different time points [17, 18].

Randomized controlled trials have been conducted investigating other therapeutic approaches, such as third-wave cognitive therapies. After a first wave of strictly behavioral approaches, and a second characterized by the implementation of a cognitive model, the third wave of cognitive behavioral therapies includes interventions in which an emphasis is put on metacognition and how the patient relates to thoughts and emotions, such as acceptance and commitment therapy (ACT), mindfulness-based treatments and metacognitive training [19]. Integrated approaches, combining multiple fundamentally different therapeutic strategies, have been also developed and investigated [20,21,22].

A network meta-analysis investigated the effects of different psychological interventions in patients with schizophrenia, but was focused on patients in the acute phase, which presented positive symptoms [14]. When investigating functioning, it is important to consider also chronic patients and patients with predominant or prominent negative symptoms; in the present analysis, we included all subgroups of patients with schizophrenia.

Specific therapeutic approaches have shown different effects in patients with schizophrenia, so that it is meaningful to investigate them separately [14, 23]. On the other side, an overall picture about the efficacy of psychological interventions is missing from the literature.

The aim of the present systematic review and meta-analysis of randomized clinical trials is to provide a comprehensive overview about the efficacy of psychological interventions in improving functioning in patients with schizophrenia regardless of the comparator, time point and rating scale used. In this way, we want to answer the research question: are psychological interventions efficacious for improving functioning in patients with schizophrenia?

Methods

Study design and inclusion criteria

The methods of the present work were adapted from the protocol, which was registered in PROSPERO with the number CRD42017067795 and published in a peer-reviewed journal [24]. The methods have been developed according to the PRISMA statement [25]. We included studies conducted in adults with a diagnosis of schizophrenia, schizophreniform or schizoaffective disorder, with no restrictions on setting, gender or ethnicity. We excluded studies that, based on their inclusion criteria, recruited only patients with concomitant somatic or psychiatric comorbidity, or only patients with first episode psychosis. Studies were included if at least 80% of the participants had schizophrenia or related disorders (schizoaffective disorder, schizophreniform disorder, delusional disorder or non-affective psychotic disorder). We included studies regardless of the diagnostic criteria used.

Studies investigating psychological interventions were included. We considered for inclusion the interventions described in the list of psychological therapies of the Cochrane Common Mental Disorders Group (CCMD) (formerly Cochrane Collaboration Depression, Anxiety and Neurosis Group [CCDAN]) [26], such as cognitive behavioral therapy, acceptance and commitment therapy, mindfulness, art therapy and music therapy. Psychosocial and community interventions such as case management or assertive community treatment were not included, as well as family interventions. The psychological intervention was usually provided in addition to the standard care, which typically includes medication with antipsychotics [27]. We accepted as comparator another psychological intervention, inactive control, defined as interventions intended to control for non-specific aspects of the therapy (for example activity groups, befriending), treatment as usual (TAU) and waiting list.

Studies were included in the analysis if they provided data for functioning measured with a validated rating scale, such as the Global Assessment of Functioning (GAF) scale or the Social Functioning Scale (SFS) [9, 28].

Search strategy

We searched EMBASE, MEDLINE, PsycINFO, PubMed, BIOSIS, Cochrane Library, World Health Organization’s International Clinical Trials Registry Platform and ClinicalTrials.gov for RCTs published up to January 2020 and the Study register of the Cochrane Schizophrenia Group from January 2020 up to September 2021, investigating the efficacy of psychological interventions in people with schizophrenia [29]. No time limit on how old the articles could be and no language restrictions were applied (Table 1).

Screening and data extraction

Two reviewers among IB, SW, CR and FS screened independently all abstracts (first phase) and full texts (second phase) identified in the search for eligibility. Results of the update search from January 2018 to September 2021 were screened by IB; NHS independently re-inspected 25% of these results, to ensure reliability of selection. Disagreements were resolved by discussion, and in case of doubt, the full paper was retrieved for further inspection. Two of IB, SW, CR, FS and NHS extracted relevant data independently in a Microsoft Access database explicitly created for this study and assessed the different domains of risk of bias using the Cochrane Risk of Bias tool [30]. We also rated an overall risk of bias for each study, following the approach described by Furukawa et al. [31]. Disagreements were resolved by discussion, by involving the senior author and, in case of need, by asking the study authors. Authors of the studies were contacted via e-mail and asked if they could provide additional data relevant for the analysis.

Data analysis

We performed random-effects pairwise meta-analyses using Review Manager version 5.3 and R Studio version 1.3.959, package meta [32, 33]. We calculated standardized mean differences (SMDs) and 95% confidence intervals (CIs). We planned different levels of analysis: (i) all psychological interventions compared to all control conditions (primary analysis); (ii) groups of psychological interventions compared to control conditions (e.g., third wave cognitive behavioral therapy, creative therapies); (iii) specific psychological interventions considered separately. The decision which studies to consider for each treatment comparison was made by two independent reviewers and then discussed, not solely based on the name the study authors gave to the intervention, but based on the description, they provided about the treatment and control conditions (Table 3).

Table 1 Search strategy for PsycINFO. (Created with Microsoft Office)
Full size table

Effect sizes are described according to Cohen, considering an effect size of 0.20 small, 0.50 moderate and 0.80 large [34].

We evaluated heterogeneity using I2, and considered heterogeneity probably not important for an I2 of up to 40%, moderate for an I2 from 30 to 60%, substantial for an I2 from 50 to 90% and considerable if over 75% according to the Cochrane Handbook for Systematic Reviews [35].

To explore potential sources for heterogeneity, we conducted subgroup and meta-regression analyses for the primary analysis, for the following potential effect modifiers: treatment setting (individual vs group), therapist expertise (trainee therapist allowed vs only expert therapists), treatment duration, age, percentage males, number of sessions, and baseline severity. Sensitivity analyses were conducted excluding studies that did not employ a blind outcome assessor, studies with researcher’s allegiance, studies focused on treatment resistant patients and studies with high overall risk of bias [31]. Subgroup, meta-regression and sensitivity analyses were considered only exploratory; therefore, we did not adjust for multiple hypotheses testing.

For the primary analysis, we assessed small trial effect, potentially associated with publication bias, by visual inspection of the funnel plot and by applying Egger’s test for funnel plot asymmetry [36]. The trim-and-fill method by Duval and Tweedie was used to give an estimate of the effect size after correcting for publication bias [37].

For the primary analysis, we assessed confidence in the estimate with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach [38].

Changes from protocol

Participants. For the present review, it was not required that patients have current positive symptoms. Studies focused on patients with predominant negative symptoms were not excluded.

Interventions with a primary aim different from positive symptoms were not excluded. The present review focuses on the outcome functioning. Pairwise meta-analyses were performed as data analysis method, and the GRADE approach was used to evaluate the confidence in the estimate [38].

Results

Characteristics of included studies

The search identified 28,420 records, of which 3570 were considered eligible and retrieved in full. 253 studies met the inclusion criteria, of which 58 had usable data and were included in the meta-analysis [22, 39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95]. The study selection process is illustrated in Fig. 1, and the included studies are described in Table 2.

Fig. 1
figure 1

Study selection. (Created with Microsoft Office)

Full size image
Table 2 Characteristics of included studies. Arrows indicate under which term the intervention was considered for the analysis when applicable. (Adapted from [1], created with Microsoft Office)
Full size table

The different psychological interventions investigated in the studies are described in detail in Table 3.

Table 3 Description of interventions. (Adapted from [1], created with Microsoft Office)
Full size table

Risk-of-bias assessment

Six, 25 and 27 studies were judged to be at low, moderate and high overall risk of bias, respectively (Table 2).

Concerning random sequence generation, the risk of bias was low in 40 (69%) studies; concerning allocation concealment, it was low in 25 (43%) studies; concerning blinding of participants and personnel, the risk of bias was never low; concerning blinding of outcome assessment in 29 (50%) studies; concerning attrition bias in 15 (26%) studies; concerning selective reporting in 15 (26%) studies; concerning researchers’ allegiance in 12 (21%) studies; and in 53 (91%) studies concerning other bias (Fig. 2).

Fig. 2
figure 2

Risk-of-bias judgements for the included studies. Reviewers’ judgements about each risk of bias item for each included study. (Created with Review Manager 5.3)

Full size image

All psychological interventions compared to all control conditions (primary analysis)

58 studies with 5048 participants provided data for this analysis. Psychological treatments were associated with a greater improvement in participants’ functioning scores in comparison to control conditions (SMD =  – 0.37, 95% CI  – 0.49 to  – 0.25), with substantial heterogeneity (I2 = 76%) (Fig. 3). The confidence in the estimate assessed with the GRADE approach was judged to be very low, due to the presence of studies at high risk of bias, substantial heterogeneity and suspected publication bias (Table 4).

Fig. 3
figure 3

Forest plot all psychological interventions versus control. (Created with Review Manager 5.3)

Full size image
Table 4 GRADE evidence profile. (Created with GRADEpro)
Full size table

Groups of psychological interventions and specific psychological interventions compared to control conditions

CBT versus control

30 studies with 2657 participants provided data for this analysis. Overall, CBT was associated with a greater improvement in functioning (SMD =  – 0.26, 95% CI  – 0.39 to  – 0.12), with substantial heterogeneity (I2 = 62%) (Fig. 4). The benefit was clear in comparison with TAU (SMD =  – 0.36, 95% CI  – 0.55 to  – 0.16), supportive therapy (SMD =  – 0.26, 95% CI  – 0.50 to  – 0.01) and psychoeducation (SMD =  – 0.95, 95% CI  – 1.74 to  – 0.16), while for the comparisons with inactive control, cognitive remediation, wait-list, family intervention and psychodynamic therapy, the confidence intervals include the possibility of no difference.

Fig. 4
figure 4

Forest plot CBT versus control. (Created with Review Manager 5.3)

Full size image

Third-wave cognitive behavior therapies versus control

15 studies with 1391 participants were included in this analysis. Third-wave CBT interventions were associated with an improvement in functioning (SMD =  – 0.60, 95% CI  – 0.83 to  – 0.37), with substantial heterogeneity (I2 = 73%) (Fig. 5).

Fig. 5
figure 5

Third-wave, creative, multiple and other therapies versus control. (Created with Review Manager 5.3 and Microsoft Power Point)

Full size image

Of these, seven studies investigated mindfulness (SMD =  – 0.72, 95% CI  – 0.98 to  – 0.46), one study investigated ACT (SMD = 0.05, 95% CI  – 0.35 to 0.45) and seven studies investigated metacognitive training (SMD =  – 0.47, 95% CI  – 0.92 to  – 0.02).

Creative therapies versus control

In this analysis, eight studies on art therapy, music therapy and movement therapy provided data for 750 participants. No difference was found between creative therapies and the control group (SMD = 0.01, 95% CI  – 0.38 to 0.39), with considerable heterogeneity (I2 = 81) (Fig. 5).

Four studies investigated art therapy (SMD =  – 0.04, 95% CI  – 0.27 to 0.36), three studies music therapy (SMD =  – 0.60, 95% CI  – 2.21 to 1.01) and one study movement therapy (SMD =  – 0.04, 95% CI  – 0.54 to 0.46).

Integrated therapies versus control

Four studies with 182 participants were included for this comparison. No difference between integrated therapies and control group was found (SMD =  – 0.21, 95% CI  – 1.20 to 0.78), with considerable heterogeneity (I2 = 88%) (Fig. 5).

Two studies investigated CBT combined with family intervention (SMD =  – 1.15, 95% CI  – 2.14 to  – 0.15), one study hallucination focused integrated treatment (SMD = 0.58, 95% CI  – 1.09 to  – 0.08) and one study multiple therapies (SMD = 0.95, 95% CI  – 0.90 to 2.81).

Other therapies versus control

Among other therapies, we included positive psychotherapy, psychosocial therapy, narrative therapy and not further specified psychotherapy. Based on four studies with 197 participants, these interventions were not associated with an improvement in functioning (SMD =  – 0.74, 95% CI  – 1.52 to 0.04), with considerable heterogeneity (I2 = 82%) (Fig. 5).

One study investigated positive psychotherapy (SMD = 0.08, 95% CI  – 0.34 to 0.51), one study investigated psychosocial therapy (SMD =  – 1.87, 95% CI  – 2.64 to  – 1.10), one study psychotherapy, without further specification (SMD = -0.69, 95% CI  – 1.35 to  – 0.03), and one study narrative therapy (SMD = 0.17, 95% CI  – 0.57 to 0.91).

Subgroup analyses

Treatment setting: individual versus group

In 28 studies, the psychological intervention was delivered in a group setting (SMD =  – 0.38, 95% CI  – 0.57 to  – 0.20, I2 = 78%) and in 25 studies in an individual setting (SMD =  – 0.31, 95% CI  – 0.48 to  – 0.14, I2 = 70%).

Test for subgroup difference did not find a difference between these two subgroups (p = 0.56).

Therapist expertise: trainee therapist allowed vs only expert therapists

In 32 studies, only expert therapists conducted therapy (SMD =  – 0.39, 95% CI  – 0.55 to  – 0.22, I2 = 79%), in seven studies’ therapists in training conducted treatment, as well (SMD =  – 0.16, 95% CI  – 0.34 to 0.03, I2 = 17%). Test for subgroup difference did not find a difference between these two subgroups (p = 0.07).

Metaregression analyses

The effect of psychological interventions on functioning was not found to be associated with number of sessions (p = 0.4347), study duration (p = 0.0901), male percentage (p = 0.1636), or baseline severity (p = 0.1244).

Age was found to have a role in moderating treatment effect on functioning, with a possible bigger treatment effect for younger patients (p = 0.0072) (Table 5).

Table 5 Results of meta-regression analyses. (Adapted from [1], created with Microsoft Office)
Full size table

Sensitivity analyses

Excluding 14 open label studies did not substantially change the results of the analysis (SMD =   –  0.38, CI  – 0.52 to  – 0.25). Heterogeneity remained similar to the original analysis (I2 = 77%).

Excluding 34 studies with high researcher allegiance, the confidence interval includes the possibility of no difference between the psychological interventions and the control condition (SMD =   –  0.21, CI  – 0.42 to 0.00). Heterogeneity remained similar compared to the original analysis (I2 = 75%).

Excluding 27 studies with high overall risk of bias did not change the results of the analysis substantially (SMD =   –  0.44, CI  – 0.58 to  – 0.25). Heterogeneity remained similar (I2 = 79%).

Excluding 11 studies focused on treatment resistant patients led to a slight decrease of effect size (SMD =   – 0.42, CI  – 0.60 to  – 0.27). Heterogeneity remained similar (I2 = 80%).

Publication bias

Visual inspection of the funnel plot reveals some asymmetry, suggesting that small studies favoring the control condition could have remained unpublished (Fig. 6a). Egger’s test for funnel plot asymmetry confirmed this (p = 0.0097) [36]. By applying the trim-and-fill method by Duval and Tweedie 16 studies were added to the funnel plot (represented in white in Fig. 6b) confidence intervals included the possibility of no effect (SMD =  – 0.13, 95% CI  – 0.27 to 0.01) [37].

Fig. 6
figure 6

Funnel plot. (Created with R Studio version 1.3.959 and Microsoft Power Point): a shows the funnel plot for the comparison all psychological interventions versus all control conditions. In b, the trim-and-fill method by Duval and Tweedie was applied; 16 studies added are represented as white dots

Full size image

Discussion

Summary of main results

We conducted the first systematic review and meta-analysis investigating the effect of psychological interventions for functioning in patients with schizophrenia. After a thorough literature search, we were able to include 58 studies with usable outcome data.

We found that psychological interventions overall showed a benefit in improving functioning compared to control conditions. According to Cohen (0.2 = small, 0.5 = medium, 0.8 = large effect size), this effect size can be considered small to medium [34]. CBT compared to any control condition had also a small effect in improving functioning, third-wave CBT interventions produced a medium benefit. Creative therapies, integrated therapies and other therapies were not associated with an improvement in functioning.

Discussion in light of previous literature

In a previous network meta-analysis by our group focused on acute patients with positive symptoms, we found an effect of  – 0.25 (95% CI  – 0.48 to  – 0.03) indicating a benefit for CBT compared to treatment as usual [14], while in the present work on the general population of patients with schizophrenia, the effect of CBT vs TAU was larger ( – 0.43). It can be argued that, after acute symptoms of schizophrenia are treated, patients are more receptive for interventions aimed at improving functioning. Results of Bighelli et al. on CBT compared with supportive therapy, family intervention, inactive control and wait-list are in line with the ones of the present work, not showing a difference between these interventions [14]. It must be noted, however, that evidence for these interventions is based on only 1–2 studies each.

Our results are also in line with the findings of Laws et al. that found an effect of 0.25 (95% CI 0.10–0.39) showing a benefit for CBT compared with control conditions [15].

Contradicting the findings of the present meta-analysis, Jones found no evidence supporting the use of CBT for functioning in people with schizophrenia [10]. This conclusion is, however, strongly affected by the different analysis that the authors conducted, analyzing different rating scales and different time points separately, so that our results cannot be compared with the ones of the Cochrane review.

In the present analysis, findings on mindfulness, metacognitive training, hallucination focused integrated treatment and psychosocial therapy are promising, but based only on a small number of trials. A recent review focused on metacognitive training, including randomized and non-randomized evidence, found a similar effect on functioning (SMD 0.41, 95% CI 0.12 to 0–69) [96].

Results of subgroup, meta-regression and sensitivity analyses did not find a role for the investigated variables in moderating the effect of psychological interventions on functioning, with exception of a possible moderating role for age (with bigger treatment effects associated with younger patients) and for researchers’ allegiance.

Younger patients might be more open for change and to engage in a psychotherapy. A systematic review reported larger effect sizes for psychotherapy in young adults with depression (up to 24 years) than in older adults [97].

Excluding studies that were conducted by the same authors who developed the treatment manual, the effect of the interventions on functioning was not so clear anymore, suggesting that effects might be inflated by allegiance of the authors to the investigated interventions. It must be noted that heterogeneity remained high in all subgroup analysis, confirming that the variables investigated in the subgroup analyses did not account for heterogeneity. A possible further explanation for heterogeneity in the investigated studies could be the use of different control conditions, that we pooled together. A network meta-analysis approach could help to disentangle this issue, analyzing also control conditions as different nodes of the network.

Limitations

First, the pooling of different rating scales is a problematic issue. As the concept of functioning developed through the years, the rating scales changed as well, including and giving a focus on different aspects like living skills, disability, social and occupational role [1, 6]. Moreover, some scales, such as the Global Assessment Scale (GAS), include psychopathology and some do not, for example the Social and Occupational Functioning Assessment Scale (SOFAS) [98, 99]. To account for this difficulty, we decided to include only published scales, for which it is possible to check the original reference and therefore description and metric properties, and we applied a statistical correction, by calculating SMDs. Still, many of the comparisons present a considerable heterogeneity, and one of the possible causes could be the use of different rating scales. On the contrary, the strategy in Cochrane reviews is to keep every measure separate, but paying the price of losing the overall picture [10, 16]. An ideal solution does not exist; an agreement on one functioning measure would make research results more comparable. Currently, in the Diagnostic and Statistical Manual 5, the American Psychiatric Association suggests using the World Health Organization Disability Assessment Schedule (WHODAS) 2.0. [100]

Second, pooling and classifying psychological treatments are not straightforward. We adopted a transparent approach, classifying the interventions according to the description given in each study, and presenting the assumptions made in Table 3. We also present different level of analyses for treatment grouped and taken singularly, so that an appraisal of the evidence is provided also independently from our classification of the psychological interventions.

Third, of 253 studies that met our eligibility criteria, only 58 reported data on functioning as an outcome. Most of all, there was scarcity of evidence for interventions other than CBT, and for some interventions, the evidence is based on few studies only. Results need therefore to be interpreted with caution.

A further limitation of the current analysis, and of studies on psychological interventions in patients with schizophrenia in general, is that participants of the studies are generally also receiving antipsychotic medication. Insufficient details on the medication were provided in the studies, so that it was not possible to disentangle the effect of psychological and pharmacological therapy. Randomization ensures that the observed effect sizes refer to the presence of the psychological intervention.

Finally, the certainty of the evidence was evaluated as very low with the GRADE approach. This evaluation is motivated by three aspects; (i) the studies providing data are mostly at overall moderate or high risk of bias; (ii) there was substantial heterogeneity. This may be due to the fact that we analyzed the results of studies with different duration together. The meta-regression analysis investigating the role of study duration was of borderline significance, so this aspect remains unclear; (iii) the results are potentially affected by small study effect, that can be associated with publication bias. Even if we conducted a thorough literature search, including study registries and gray literature, it is possible that some small studies favoring the control condition remained unpublished and were not possible to detect.

Implications for future research and practice

Despite limitations, the present data suggest that psychological interventions can improve functioning in people with schizophrenia. In particular, CBT and third-wave CBT interventions seem to have a positive effect on functioning.

To increase the amount of evidence on other treatments, future trials investigating psychological interventions for schizophrenia should address functioning among outcomes, not only psychopathology.