Video games for people with schizophrenia

Abstract

Background

Commercial video games are a vastly popular form of recreational activity. Whilst concerns persist regarding possible negative effects of video games, they have been suggested to provide cognitive benefits to users. They are also frequently employed as control interventions in comparisons of more complex cognitive or psychological interventions. If independently effective, video games ‐ being both engaging and relatively inexpensive ‐ could provide a much more cost‐effective add‐on intervention to standard treatment when compared to costly, cognitive interventions.

Objectives

To review the effects of video games (alone or as an additional intervention) compared to standard care alone or other interventions including, but not limited to, cognitive remediation or cognitive behavioural therapy for people with schizophrenia or schizophrenia‐like illnesses.

Search methods

We searched the Cochrane Schizophrenia Group's Study‐Based Register of Trials (March 2017, August 2018, August 2019).

Selection criteria

Randomised controlled trials focusing on video games for people with schizophrenia or schizophrenia‐like illnesses.

Data collection and analysis

Review authors extracted data independently. For binary outcomes we calculated risk ratio (RR) with its 95% confidence interval (CI) on an intention‐to‐treat basis. For continuous data we calculated the mean difference (MD) between groups and its CI. We employed a fixed‐effect model for analyses. We assessed risk of bias for the included studies and created a 'Summary of findings' table using GRADE.

Main results

This review includes seven trials conducted between 2009 and 2018 (total = 468 participants, range 32 to 121). Study duration varied from six weeks to twelve weeks. All interventions in the included trials were given in addition to standard care, including prescribed medication. In trials video games tend to be the control for testing efficacy of complex, cognitive therapies; only two small trials evaluated commercial video games as the intervention. We categorised video game interventions into 'non‐exergame' (played statically) and 'exergame' (the players use bodily movements to control the game). Our main outcomes of interest were clinically important changes in: general functioning, cognitive functioning, social functioning, mental state, quality of life, and physical fitness as well as clinically important adverse effects.

We found no clear difference between non‐exergames and cognitive remediation in general functioning scores (Strauss Carpenter Outcome Scale) (MD 0.42, 95% CI −0.62 to 1.46; participants = 86; studies = 1, very low‐quality evidence) or social functioning scores (Specific Levels of Functioning Scale) (MD −3.13, 95% CI ‐40.17 to 33.91; participants = 53; studies = 1, very low‐quality evidence). There was a clear difference favouring cognitive remediation for cognitive functioning (improved on at least one domain of MATRICS Consensus Cognitive Battery Test) (RR 0.58, 95% CI 0.34 to 0.99; participants = 42; studies = 1, low‐quality evidence). For mental state, Positive and Negative Syndrome Scale (PANSS) overall scores showed no clear difference between treatment groups (MD 0.20, 95% CI −3.89 to 4.28; participants = 269; studies = 4, low‐quality evidence). Quality of life ratings (Quality of Life Scale) similarly showed no clear intergroup difference (MD 0.01, 95% CI −0.40 to 0.42; participants = 87; studies = 1, very low‐quality evidence). Adverse effects were not reported; we chose leaving the study early as a proxy measure. The attrition rate by end of treatment was similar between treatment groups (RR 0.96, 95% CI 0.87 to 1.06; participants = 395; studies = 5, low‐quality evidence).

One small trial compared exergames with standard care, but few outcomes were reported. No clear difference between interventions was seen for cognitive functioning (measured by MATRICS Consensus Cognitive Battery Test) (MD 2.90, 95% CI ‐1.27 to 7.07; participants = 33; studies = 1, low‐quality evidence), however a benefit in favour of exergames was found for average change in physical fitness (aerobic fitness) (MD 3.82, 95% CI 1.75 to 5.89; participants = 33; studies = 1, low‐quality evidence). Adverse effects were not reported; we chose leaving the study early as a proxy measure. The attrition rate by end of treatment was similar between treatment groups (RR 1.06, 95% CI 0.75 to 1.51; participants = 33; studies = 1).

Another small trial compared exergames with non‐exergames. Only one of our main outcomes was reported ‐ physical fitness, which was measured by average time taken to walk 3 metres. No clear intergroup difference was identified at six‐week follow‐up (MD −0.50, 95% CI −1.17 to 0.17; participants = 28; studies = 1, very low‐quality evidence).

No trials reported adverse effects. We chose leaving the study early as a proxy outcome.

Authors' conclusions

Our results suggest that non‐exergames may have a less beneficial effect on cognitive functioning than cognitive remediation, but have comparable effects for all other outcomes. These data are from a small number of trials, and the evidence is graded as of low or very low quality and is very likely to change with more data. It is difficult to currently establish if the more sophisticated cognitive approaches do any more good ‐ or harm ‐ than 'static' video games for people with schizophrenia.

Where players use bodily movements to control the game (exergames), there is very limited evidence suggesting a possible benefit of exergames compared to standard care in terms of cognitive functioning and aerobic fitness. However, this finding must be replicated in trials with a larger sample size and that are conducted over a longer time frame.

We cannot draw any firm conclusions regarding the effects of video games until more high‐quality evidence is available. There are ongoing studies that may provide helpful data in the near future.

Plain language summary

Video games for schizophrenia

Review question

Are video games an effective treatment (on their own or as an add‐on) for improving the well‐being and functioning of people with schizophrenia or schizoaffective disorder?

Background

Schizophrenia is a severe mental illness that affects people worldwide. People with schizophrenia often have a distorted view of reality ‐ perceiving things that are not present (hallucinations) and believing things that are not true (delusions). People with schizophrenia may struggle to motivate themselves, experience anxiety and depression, and encounter cognitive symptoms, often struggling to stay focused on day‐to‐day activities and becoming disorientated. Hallucinations and delusions are usually treated with antipsychotic medications, whereas other symptoms can be difficult to manage with medication alone.

Psychological therapies are sometimes used alongside medication to help with some symptoms of schizophrenia, however these therapies can be complex and expensive. Video games are a relatively inexpensive, and, for many, engaging treatment. They have been suggested to help improve the cognitive impairments such as lack of focus or poor memory that people with schizophrenia often experience. If effective, video games could provide a simple and relatively low‐cost additional treatment for people with schizophrenia.

Searching

We searched for randomised controlled trials (a type of study in which participants are assigned to one of two or more treatment groups using a random method) involving people with schizophrenia receiving either a video game intervention or other type of treatment such as talking (cognitive) therapy or placebo (dummy treatment). We performed the searches in March 2017, August 2018, and August 2019.

Results

Seven trials met our inclusion criteria and provided useable data. Video game interventions were categorised into those that involved movements of the body ('exergames') and those that did not ('non‐exergames'). Non‐exergame trials compared the video game intervention to a form of "brain‐training" therapy known as cognitive remediation. One trial compared exergames to standard care, and another compared non‐exergames with exergames. All interventions in the included trials were given in addition to standard care.

The currently available evidence suggests that non‐exergames may not be as beneficial for cognitive functioning as cognitive remediation, but there were no other clear differences between non‐exergames and cognitive remediation for improving functioning in people with schizophrenia. The more exercise‐orientated video games may have some benefit compared to standard care for improving physical fitness. We cannot draw any firm conclusions regarding the effects of video games until higher‐quality evidence is available.

Summary of findings

Summary of findings 1. VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) compared to COGNITIVE REMEDIATION (plus standard care) for people with schizophrenia.

VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) compared to COGNITIVE REMEDIATION for people with schizophrenia (plus standard care) (all short term)
Patient or population: people with schizophrenia Setting: inpatient and outpatient Intervention: video games (commercial non‐exergames) Comparison: cognitive remediation
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	quality of the evidence (GRADE)	Comments
	Risk with cognitive remediation	Risk with video games (commercial non‐exergames)
Functioning: general ‐ average endpoint score (Strauss Carpenter Outcome Scale)* ‐ by end of treatment	‐	MD 0.42 higher (0.62 lower to 1.46 higher)	‐	86 (1 RCT)	⊕⊝⊝⊝ VERY LOW 1 2 3	*Clinically important change data not reported.
Functioning: specific ‐ cognitive ‐ improved (MCCB, at least 1 domain) ‐ 3‐month follow‐up	773 per 1000	448 per 1000 (263 to 765)	RR 0.58 (0.34 to 0.99)	42 (1 RCT)	⊕⊕⊝⊝ LOW 1 3
Functioning: specific ‐ social ‐ average endpoint score (SLOF, high = good)* ‐ by end of treatment	‐	MD 3.13 lower (40.17 lower to 33.91 higher)	‐	53 (1 RCT)	⊕⊝⊝⊝ VERY LOW 1 2 3	*Clinically important change data not reported.
Adverse effects*: leaving the study early ‐ did not leave ‐ by end of treatment	200 per 1000	232 per 1000 (156 to 342)	RR 0.96 (0.87 to 1.06)	395 (5 RCTs)	⊕⊕⊝⊝ LOW 1 4	*No studies directly reported on adverse effects.
Mental state: overall ‐ average endpoint score (PANSS, total, high = poor)* ‐ by end of treatment	‐	MD 0.20 higher (3.89 lower to 4.28 higher)	‐	269 (4 RCTs)	⊕⊕⊝⊝ LOW 1 2	*Clinically important change data not reported.
Physical fitness: clinically important change	‐	‐	‐	(0 RCTs)	‐	No studies reported on this important outcome.
Quality of life: overall ‐ average endpoint score (QLS, high = good) ‐ by end of treatment	‐	MD 0.01 higher (0.40 lower to 0.42 higher)	‐	87 (1 RCT)	⊕⊝⊝⊝ VERY LOW 1 2 3
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

¹Risk of bias: downgraded by 1 ‐ randomisation and concealment of allocation not well described.
²Indirectness: downgraded by 1 ‐ continuous data when binary stipulated in protocol.
³Imprecision: downgraded by 1 ‐ small trial, wide confidence intervals.
⁴Indirectness: downgraded by 1 ‐ leaving the study is proxy for adverse effects and may not even represent a negative event/effect.

MCCB: (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Consensus Cognitive Battery
PANSS: Positive and Negative Syndrome Scale
SLOF: Specific Levels of Functioning
QLS: Quality of Life Scale

Summary of findings 2. VIDEO GAMES (EXERGAMES) (plus standard care) compared to STANDARD CARE for people with schizophrenia.

VIDEO GAMES (EXERGAMES) (plus standard care) compared to STANDARD CARE for people with schizophrenia (all short term)
Patient or population: people with schizophrenia Setting: outpatient Intervention: video games (exergames) Comparison: standard care
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	quality of the evidence (GRADE)	Comments
	Risk with standard care	Risk with video games (exergames)
Functioning: general ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Functioning: specific ‐ cognitive ‐ average change score (MCCB, increase = good)*	‐	MD 2.90 higher (1.27 lower to 7.07 higher)	‐	33 (1 RCT)	⊕⊕⊝⊝ LOW 1 2	*Clinically important change data not reported.
Functioning: specific ‐ social ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Adverse effects*: leaving the study early ‐ did not leave	765 per 1000	811 per 1000 (574 to 1000)	RR 1.06 (0.75 to 1.51)	33 (1 RCT)	⊕⊕⊝⊝ LOW 2 3	*No studies directly reported on adverse effects.
Mental state: overall ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Physical fitness*: average change VO2 peak (AF, increase = good)	‐	MD 3.82 higher (1.75 higher to 5.89 higher)	‐	33 (1 RCT)	⊕⊕⊝⊝ LOW 2 4	*Clinically important change data not reported.
Quality of life: overall ‐ any change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

¹Indirectness: downgraded by 1 ‐ continuous data when binary stipulated in protocol.
²Imprecision: downgraded by 1 ‐ small trial, wide confidence intervals.
³Indirectness: downgraded by 1 ‐ leaving the study is proxy for adverse effects and may not even represent a negative event/effect.
⁴Indirectness: downgraded by 1 ‐ continuous data when binary stipulated in protocol; aerobic fitness is proxy for physical fitness and may not correspond to clinically important change.

AF: aerobic fitness
MCCB: (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Consensus Cognitive Battery
VO₂ peak: peak oxygen uptake

Summary of findings 3. VIDEO GAMES (EXERGAMES) (plus standard care) compared to VIDEO GAMES (NON‐EXERGAMES) (plus standard care) for people with schizophrenia.

VIDEO GAMES (EXERGAMES) (plus standard care) compared to VIDEO GAMES (NON‐EXERGAMES) (plus standard care) for people with schizophrenia (all short term)
Patient or population: people with schizophrenia Setting: outpatient Intervention: video games (exergames) (plus standard care) Comparison: video games (non‐exergames) (plus standard care)
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	quality of the evidence (GRADE)	Comments
	Risk with video games (non‐exergames)	Risk with video games (exergames)
Functioning: general ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Functioning: specific ‐ cognitive ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Functioning: specific ‐ social ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome. studies.
Adverse effect: any important adverse effect	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Mental state: overall ‐ clinically important change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
Physical fitness: specific ‐ walking speed ‐ average endpoint (seconds taken to walk 3 metres)*	‐	MD 0.50 seconds lower (1.17 lower to 0.17 higher)	‐	28 (1 RCT)	⊝⊝⊝⊝ VERY LOW 1 2 3	*Clinically important change data not reported.
Quality of life: overall ‐ any change	‐	‐	‐	(0 RCTs)	‐	No trials reported this important outcome.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; RCT: randomised controlled trial
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low quality: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low quality: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

¹Risk of bias: downgraded by 2 ‐ method of randomisation not described; allocation and blinding not described.
²Indirectness: downgraded by 1 ‐ continuous data when binary stipulated in protocol; walking speed is proxy for physical fitness and may not correspond to clinically important change.
³Imprecision: downgraded by 1 ‐ small trial, wide confidence intervals.

Background

Description of the condition

Schizophrenia is a psychiatric disorder that affects populations worldwide (Owen 2016). It is estimated that the median incidence for schizophrenia is 15.2 in every 100,000 people, with a prevalence ratio for males:females of 1.4:1 (McGrath 2008). Typical age of onset is late adolescence, which is the time when brain development can be susceptible to onset and development of psychosis (Gogtay 2011).

Compared with the general population, people with schizophrenia have a two‐ to three‐fold higher mortality rate, which translates to a 10‐ to 25‐year life expectancy reduction (Laursen 2012). Suicide is the most common cause of premature death for people with schizophrenia, with a meta‐analysis estimating that 4.9% of people with schizophrenia will commit suicide during their lifetime (Palmer 2005), whereas a lifetime prevalence cross‐nationally of suicide attempts in the general population is 2.7% (Nock 2008).

Adults with schizophrenia experience an array of symptoms. The 13 core symptoms may be grouped into five broad categories:

1. positive symptoms (delusion, unusual thought content, hallucinations);

2. negative symptoms (flat affect, emotional withdrawal, self‐neglect, impaired motivation);

3. depressive symptoms (depressive mood, anxiety);

4. cognitive symptoms (lack of attention, disorientation); and

5. behavioural symptoms (hostility, euphoria) (Bak 2001).

Antipsychotic medication is the mainstay treatment for the symptoms of schizophrenia, and these medications tend to have a beneficial effect on managing positive symptoms such as delusions, unusual thoughts, and hallucinations (Suenderhauf 2016). However, negative symptoms, such as anhedonia and social withdrawal, are usually more difficult to manage (Suenderhauf 2016). The number of people with chronic schizophrenia who do not get an adequate response with medication is around 50% to 60% (Yang 2015). As a result, people with schizophrenia continue to experience distressing, chronic symptoms (Bauer 2011).

Other interventions, such as cognitive remediation, are often used as an adjunctive treatment to antipsychotic drugs, incorporating computerised exercises to improve cognitive processes in people with schizophrenia (Barlati 2013). Video games, whilst being a form of computerised exercises, have also been linked to cognitive benefits (Granic 2014).

Description of the intervention

We will define video gaming as computerised,electronic manipulation of images forming an interactive, graphically interesting game played using a controller or keyboard on a display monitor. Video gaming is now one of the most popular recreational pastimes (Wittek 2016). It differs to other similar forms of recreational pastimes such as reading books and watching television due to its interactive elements. Players are not forced to adhere to a game's storyline, they are actively involved in the development of the game and, in turn, the game actively responds to the players' behaviour (Granic 2014). Games can be played alone, competitively against another person, or with people worldwide using online platforms. Games can be played using consoles (e.g. Xbox, Playstation), computers, or even mobile phones (Granic 2014). This is the largest entertainment industry in the UK (Hollingdale 2014); 25% of Europeans play video games at least once a week (Ipsos MediaCT 2012), and 59% of North Americans (Ipsos MediaCT 2014). Being a constantly improving technology, using serious games for training and educational purposes is also a burgeoning area. In people with psychosis, these types of games greatly improve the adherence to e‐interventions (O'Hanlon 2016). The games incorporate 'rewards' or 'extra lives' that can be redeemed daily, thus encouraging much needed engagement with services (O'Hanlon 2016). This improvement in adherence is not surprising given that by 21 years of age, the average adolescent is estimated to have played around 10,000 hours of video games (Kuhn 2014).

How the intervention might work

There exists a public perception that playing video games promotes intellectual laziness and sedentary lifestyles (Granic 2014; Hernandez 1999; Owen 2010), the latter of which is highly prevalent in people with schizophrenia (Kimhy 2016). However, it has been found that these games allow players to develop a wide range of cognitive skills (Granic 2014). Nevertheless, it has also been demonstrated that whilst video games can promote prosocial outcomes, violent video games can increase aggression (Greitemeyer 2014; Hollingdale 2014). Behavioural and magnetic resonance imaging studies have demonstrated that video games have the potential to impact brain plasticity (Suenderhauf 2016). Video games are easily accessible, and may counteract the limitations of other therapies (e.g. cognitive behavioural therapy). These include a reduction in the number of clinicians properly trained in the approach, limited patient resources, and people choosing not to access mental health services. About 50% of people with schizophrenia in the USA do not receive any treatment (Gottlieb 2013). Video games could potentially offer relief for people with schizophrenia without the need for professionals, and encourage a greater proportion of the population to engage with services. Furthermore, in discussions with academics in the field, it has been stated that people with schizophrenia have reportedly shown a reduction in delusional thinking and extrapyramidal symptoms following only eight weeks of Internet‐game play (Bavelier 2011; Han 2008).

One way in which video games may be beneficial in treating people with schizophrenia is through distraction, that is the holding or focusing of attention (Trygstad 2002). Due to the suspenseful and pleasurable nature of video games, they have the potential to fully engage a player's attention (Suenderhauf 2016). We will define distraction as a 'real‐life' action voluntarily taken, which then increases vulnerability to the involuntary distraction from symptoms.

The largest problem faced by clinicians is engagement with young people and adolescents. This may be due to young people and adolescents not engaging with the therapies or not recognising that they have a mental health problem (Granic 2014). Video games not only have the ability to engage adolescents, but also the wider population. In the UK, 37% of the population aged 16 to 49 years described themselves as 'active gamers' (Morris 2013). Furthermore, an age breakdown of gamers on average across Europe sees 51% of users below 35 years of age, with 49% of users 35 years or older (Ipsos MediaCT 2012). In addition to distraction from symptoms, video gaming may also increase positive attitudes, improve problem solving, and reduce abnormal behaviours (Fernandez‐Aranda 2012).

Video games are stereotypically perceived to be a static activity. However, 'exergames' are video games that look to promote physical fitness and rehabilitate motor function in various populations. These games allow play whilst incorporating exercise as the game responds to bodily movements. Physical activity is reduced in people with schizophrenia, and exergames improve endurance, motor co‐ordination, flexibility, and balance skills, and hence have the potential to be a robust intervention for people with schizophrenia (Campos 2015).

Due to a growing interest from several healthcare systems in increasing accessibility of treatment for mental health disorders, implementing these new technologies to treat people with schizophrenia is something that is desired by many (Fernandez‐Aranda 2012). If video gaming could be used successfully to provide targeted functional activation changes, it could be a beneficial intervention for people with severe psychiatric disorders such as schizophrenia. Given the majority of the games being targeted at young people, this could provide an added benefit, as it is consistent the typical age of onset of schizophrenia, in late adolescence (Gogtay 2011).

Why it is important to do this review

We are unsure if commercial games have an effect on the mental well‐being of people in general and of people with schizophrenia in particular. There is concern that they could be detrimental (Tortolero 2014). Notwithstanding, video gaming has not only shown possible physiological benefits but also a much needed improvement with engagement of services in people with psychosis (O'Hanlon 2016). However, we are also aware that video games can be used as an easy control group to more sophisticated ‐ and expensive ‐ interventions such as cognitive remediation. As with a previous Cochrane Review assessing supportive therapy for people with schizophrenia (Buckley 2015), the low‐grade 'control' to many more complex approaches is the main focus of this review. If, as for the 'Supportive therapy for schizophrenia' review (Buckley 2015), the control intervention is as effective as the more expensive approaches, simple commercial video games could save an enormous waste of finite resources on researching ostensibly convincing, but ultimately ineffective, interventions.

Objectives

To review the effects of video games (alone or as an additional intervention) compared to standard care alone or other interventions including, but not limited to, cognitive remediation or cognitive behavioural therapy for people with schizophrenia or schizophrenia‐like illnesses.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) meeting our inclusion criteria and reporting useable data. We considered trials in which randomisation was implied, which we included or excluded once we had carried out a sensitivity analysis (see Sensitivity analysis). We excluded quasi‐randomised studies, such as those that allocate intervention by alternate days of the week. Where people were given other treatments in addition to video games, we only included the studies if the adjunct treatment was evenly distributed between groups.

Types of participants

People with schizophrenia or schizophrenia‐like illnesses, including schizophreniform disorder, schizoaffective disorder, and delusional disorder, by any means of diagnosis. We included trials with participants with schizophrenia or related disorders alongside adults with schizophrenia. However, we excluded any trial that involved 'young people at risk of psychosis', as these people do not have psychosis.

To ensure that the information provided was relevant to the current care of people with schizophrenia, we aimed to highlight the current clinical state clearly (acute, early postacute, partial remission, remission), the stage (prodromal, first episode, early illness, persistent), and whether the studies primarily focused on people with particular problems (e.g. negative symptoms, treatment‐resistant illnesses).

Types of interventions

We compared video games versus standard care alone or other psychological treatments.

1. Video games

"A video game is an electronic game that involves interaction with a user interface to generate visual feedback to a video device such as TV screen or computer monitor [...] it implies any type of display device that can produce two‐ or three‐dimensional images" (en.wikipedia.org/wiki/Video_game), and an exergame "relies on technology that tracks body movement or reaction" (en.wikipedia.org/wiki/Exergaming).

We were interested in using commercial video games, played alone, co‐operatively, or competitively, using a console (e.g. Xbox, Playstation), computer, or mobile phone. To be considered 'video gaming', the games would have had to include an interactive element, but we also included games that were simply described as such and did not clearly specify the level of interaction. With players not adhering to a set storyline, but they must be actively engaged in the development of the story, and the video game in turn responds to the player's behaviour (Granic 2014).

We included any intervention that called itself a 'video' or 'computer' game. Additionally, we included 'exergames' as a form of video game if the game is interactive and responds to the player's actions. However, we did not include virtual reality as a form of video gaming, as this is the focus of another recent Cochrane Review (Moazzen 2015; Aali 2020). We also excluded 'serious games', being those designed for a primary purpose other than entertainment, such as a primary purpose of developing cognitive skills (Djaouti 2011).

We accepted studies that compared the effects of two video games against each other (e.g. exergames versus non‐exergames).

We envisaged that the video games would be given alongside the standard professional care that people with schizophrenia would receive.

2. Standard care alone

The standard care that the person normally would receive had they not participated in the trial, including, but not limited to, medication and hospitalisation.

3. Cognitive remediation

We considered a cognitive remediation to be an intervention that aims to improve any cognitive domains, such as: speed of processing, attention/vigilance, working memory, verbal learning and memory, visual learning and memory, reasoning and problem solving, social cognition (Nuechterlein 2004), and meta‐cognition.

See Differences between protocol and review.

4. Other psychological treatments

Including, but not limited to, cognitive behavioural therapy, psychodynamic psychotherapy, or problem‐solving therapy.

Types of outcome measures

We divided all outcomes into short term (less than six months), medium term (six to 12 months), and long term (over 12 months).

We sought to report binary outcomes recording clear and clinically meaningful degrees of change (e.g. global impression of much improved, or more than 50% improvement on a rating scale defined within the trials). Thereafter, we listed other binary outcomes and then those that were continuous.

Primary outcomes

1. Functioning

1.1 General: clinically important change in general functioning ‐ as defined by study, including working ability.
1.2 Specific ‐ cognitive: clinically important change in cognitive functioning ‐ as defined by study.
1.3 Specific ‐ social: clinically important change in social functioning ‐ as defined by study.

2. Adverse effects

2.1 At least one important adverse effect.

Secondary outcomes

1. Functioning

1.1 General: any change in general functioning ‐ as defined by study, including working ability.
1.2 Specific ‐ cognitive: any change in cognitive functioning ‐ as defined by study.
1.3 Specific ‐ social: mean endpoint or change score on social functioning scale.
1.4 Specific ‐ social: any change in social functioning ‐ as defined by study, e.g. social skills.

2. Adverse effects

2.1 Any change in adverse effects: general/specific.
2.2 Very serious event (e.g. death, serious self‐harm, suicide).
2.3 Any change in specific adverse effects (e.g. repetitive strain injury, aggression).

3. Mental state

3.1 Clinically important change in general mental state ‐ as defined by study.
3.2 Mean endpoint or change score on general mental state scale.
3.3 Clinically important change in specific symptoms ‐ as defined by study (positive, negative, affective, cognitive symptoms of schizophrenia).

4. Physical fitness

4.1 Clinically important change in overall physical fitness ‐ as defined by study.
4.2 Mean change score in overall physical fitness.
4.3 Clinically important change in specific aspects of physical fitness ‐ as defined by study.
4.4 Mean change scores in specific aspects of physical fitness.

5. Global state

5.1 Clinically important change in global state (clinical response as defined by study, e.g. global impression of much improved, or more than 50% improvement on a rating scale).
5.2 Relapse ‐ as defined by study.
5.3 Mean endpoint or change score on general global state scale.

6. Leaving the study early

6.1 For any reason.
6.2 Due to inefficacy.
6.3 Due to adverse effect.

7. Quality of life

7.1 Any change in overall quality of life ‐ as defined by study.
7.2 Mean endpoint or change score on quality of life scale.
7.3 Any change in specific aspects of quality of life ‐ as defined by study.
7.4 Mean endpoint or change score on specific aspects of quality of life scale.

8. Behaviour

8.1 Any change in general behaviour ‐ as defined by study.
8.2 Mean endpoint or change score on general behaviour scale.
8.3 Any change in specific aspects of behaviour ‐ as defined by study (e.g. aggression, violence).
8.4 Mean endpoint or change on specific aspects of behaviour scale.

9. Economic outcomes

9.1 Costs due to treatment, as defined by study.
9.2 Total direct and indirect costs.
9.3 Mean change in total cost of medical and mental health care.

'Summary of findings' table

We used the GRADE approach to interpret findings (Schünemann 2011), and GRADEpro GDT to export data from our review to create a 'Summary of findings' table. These tables provide outcome‐specific information concerning the overall quality of evidence from each included study in the comparison, the magnitude of effect of the interventions examined, and the sum of available data on all outcomes rated as important to patient care and decision making. We selected the following main outcomes for inclusion in the 'Summary of findings' table.

1. Functioning ‐ general: clinically important change in general functioning ‐ as defined by study, including working ability.

2. Functioning ‐ specific ‐ cognitive: clinically important change in cognitive functioning ‐ as defined by study.

3. Functioning ‐ specific ‐ social: clinically important change in social functioning ‐ as defined by study.

4. Adverse effect: at least one important adverse effect.

5. Mental state: overall ‐ clinically important change in mental state ‐ as defined by study.

6. Physical fitness: clinically important change in physical fitness ‐ as defined by study.

7. Quality of life: overall ‐ any change in quality of life ‐ as defined by study.

If data were not available for these prespecified outcomes but were available for ones that were similar, we presented the closest outcome available, but took this into account when grading the finding.

Search methods for identification of studies

Electronic searches

Cochrane Schizophrenia Group's Study‐Based Register of Trials

On 31 March 2017 and 23 August 2018 the Information Specialist searched the register using the following search strategy:

*Game* in Intervention Field of STUDY

In such study‐based register, searching the major concept retrieves all the synonyms and relevant studies because all the studies have already been organised based on their interventions and linked to the relevant topics (Shokraneh 2017; Shokraneh 2018).

This register is compiled by systematic searches of major resources (Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, Embase, AMED (Allied and Complementary Medicine), BIOSIS, CINAHL (Cumulative Index to Nursing and Allied Health Literature), PsycINFO, PubMed, US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov, World Health Organization International Clinical Trials Registry Platform) and their monthly updates, ProQuest Dissertations and Theses A&I and its quarterly update, Chinese databases (CBM (Chinese Biomedical Literature Database), CNKI (China National Knowledge Infrastructure), and Wanfang) and their annual updates, handsearches, grey literature, and conference proceedings (see Group's website). There is no language, date, document type, or publication status limitations for inclusion of records into the register.

The Information Specialist performed a further update search on 8 August 2019, employing the same strategy.

Searching other resources

1. Reference searching

We inspected references of all included studies for further relevant studies.

2. Personal contact

Where additional information regarding relevant studies was required, we contacted the first author.

Data collection and analysis

Selection of studies

Review authors (JL, GH, MF, AZB, MR) independently inspected the citations from the searches and identified relevant abstracts; MV independently re‐inspected a random 20% sample of these abstracts to ensure reliability of selection. Where disputes arose, the full report was acquired for a more detailed scrutiny. Review authors (JL, GH, MF, AZB) then obtained and inspected the full reports of the abstracts or reports meeting the review criteria. MV re‐inspected a random 20% of these full reports to ensure reliability of selection. MR inspected citations from the August 2019 update search, and subsequently re‐inspected all included and ongoing studies and ensured that they met inclusion criteria, amending as necessary. Where it was not possible to resolve disagreement by discussion, we would attempt to contact the authors of the study concerned for clarification.

Data extraction and management

1. Extraction

Review authors (JL, GH, MF, AZB) independently extracted data from all included studies. In addition, to ensure reliability, MV independently extracted data from a random sample of these studies, comprising 10% of the total. MR subsequently re‐checked data extraction from all included studies, and amended as necessary. We attempted to extract data presented only in graphs and figures whenever possible, but only included this information if two review authors independently obtained the same result. If studies were multicentre, then where possible, data were extracted relevant to each centre. Any disagreements were discussed by JL and GH and documented. If necessary, JL, GH, MF, and AZB would have attempted to contact authors through an open‐ended request to obtain missing information or for clarification. MV helped clarify issues regarding any remaining problems, and these final decisions were documented.

2. Management

2.1. Forms

We extracted data onto standard, pre‐designed, simple forms.

2.2. Scale‐derived data

We included continuous data from rating scales only if:

1. the psychometric properties of the measuring instrument had been described in a peer‐reviewed journal (Marshall 2000); and

2. the measuring instrument had not been written or modified by one of the trialists for that particular trial;

3. the instrument should be a global assessment of an area of functioning and not subscores which are not, in themselves, validated or shown to be reliable. However, there are exceptions: we included subscores from mental state scales measuring positive and negative symptoms of schizophrenia.

Ideally the measuring instrument should either be a self‐report or completed by an independent rater or relative (not the therapist). We realise that this is often not reported clearly (Description of studies).

We noted if this was the case or not.

2.3. Endpoint versus change data

There are advantages of both endpoint and change data: change data can remove a component of between‐person variability from the analysis; however, calculation of change needs two assessments (baseline and endpoint), which can be difficult to obtain in unstable and difficult‐to‐measure conditions such as schizophrenia. We primarily used endpoint data, only using change data if endpoint data were not available. If necessary, we would have combined endpoint and change data in the analysis, as we prefer using mean differences (MDs) to standardised mean differences (SMDs) (Deeks 2011).

2.4. Skewed data

Continuous data on clinical and social outcomes are often not normally distributed. To avoid the problems of applying parametric tests to non‐parametric data, we applied the following standards to relevant continuous data before inclusion.

For endpoint data from studies including fewer than 200 participants:

1. when a scale started from the finite number zero, we subtracted the lowest possible value from the mean, and divided this by the standard deviation (SD). If this value was lower than one, it strongly suggested that the data were skewed and we excluded these data. If this ratio was higher than one but less than two, it suggested that the data were skewed: we entered these data and tested whether their inclusion or exclusion would change the results substantially. Finally, if the ratio was larger than two, we included these data, because it was less likely that they were skewed (Altman 1996; Higgins 2011);

2. if a scale started from a positive value (such as the Positive and Negative Syndrome Scale (PANSS), which can have values from 30 to 210) (Kay 1986), we modified the calculation described above to take the scale starting point into account. In these cases, skewed data are present if 2 SD > (S ‐ S_min), where S is the mean score and S_min is the minimum score.

Please note: we would have entered all relevant data from studies of more than 200 participants in the analysis irrespective of the above rules, because skewed data pose less of a problem in large studies. However, we did not include any studies containing more than 200 participants. We also would have entered all relevant change data, as when continuous data are presented on a scale that includes a possibility of negative values (such as change data), it is difficult to determine whether data are skewed.

2.5. Common measurement

To facilitate comparison between trials, where relevant, we would have converted variables that can be reported in different metrics, such as days in hospital (mean days per year, per week, or per month) to a common metric (e.g. mean days per month).

2.6. Conversion of continuous to binary

Where possible, we would have converted outcome measures to dichotomous data. This can be done by identifying cut‐off points on rating scales and dividing participants accordingly into 'clinically improved' or 'not clinically improved'. It is generally assumed that if there is a 50% reduction in a scale‐derived score such as the Brief Psychiatric Rating Scale (BPRS) (Overall 1962), or the PANSS (Kay 1986), this could be considered as a clinically significant response (Leucht 2005; Leucht 2005a). If data based on these thresholds were not available, we would use the primary cut‐off presented by the original authors.

2.7. Direction of graphs

Where possible, we entered data so that the area to the right of the line of no effect indicates a favourable outcome for video games. Where keeping to this made it impossible to avoid outcome titles with clumsy double‐negatives (e.g. 'not unimproved'), we reported data where the right of the line indicates an unfavourable outcome and noted this in the relevant graphs.

Assessment of risk of bias in included studies

Review authors (JL, GH, MF, AZB, and MR) independently assessed risk of bias using criteria described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). This set of criteria is based on evidence of associations between potential overestimation of effect and the level of risk of bias of the article that may be due to aspects of sequence generation, allocation concealment, blinding, incomplete outcome data, and selective reporting, or the way in which these 'domains' are reported.

If the raters disagreed, we made the final rating by consensus with the involvement of MV. Where inadequate details of randomisation and other characteristics of trials were provided, we attempted to contact the authors of the studies to obtain the additional information. We reported non‐concurrence in risk of bias assessment, but any disputes arising regarding the category to which a trial was to be allocated were resolved by discussion.

We noted the level of risk of bias in the text of the review, 'Risk of bias' graph, 'Risk of bias' summary, and the 'Summary of findings' table/s.

Measures of treatment effect

1. Binary data

For binary outcomes, we calculated a standard estimation of the risk ratio (RR) and its 95% confidence interval (CI), as it has been shown that RR is more intuitive than odds ratios (Boissel 1999), and that odds ratios tend to be interpreted as RR by clinicians (Deeks 2000). Although the number needed to treat for an additional beneficial outcome (NNTB) and the number needed to treat for an additional harmful outcome (NNTH), with their CIs, are intuitively attractive to clinicians, they are problematic to calculate and interpret in meta‐analyses (Hutton 2009). For binary data presented in the 'Summary of findings' table/s, where possible, we would have calculated illustrative comparative risks.

2. Continuous data

For continuous outcomes, we estimated MD between groups. We preferred not to calculate effect size measures (SMD). However, if scales of very considerable similarity were used, we presumed there was a small difference in measurement, and calculated effect size and transformed the effect back to the units of one or more of the specific instruments.

Unit of analysis issues

1. Cluster trials

Studies increasingly employ 'cluster randomisation' (such as randomisation by clinician or practice), but analysis and pooling of clustered data poses problems. Authors often fail to account for intraclass correlation in clustered studies, leading to a unit of analysis error whereby P values are spuriously low, CIs unduly narrow, and statistical significance overestimated (Divine 1992). This causes type I errors (Bland 1997; Gulliford 1999).

Where clustering was incorporated into the analysis of primary studies, we presented these data as if from a non‐cluster‐randomised study, but adjusted for the clustering effect.

Where clustering was not accounted for in primary studies, we presented data in a table, with an * symbol to indicate the presence of a probable unit of analysis error. We attempted to contact first authors of studies to obtain intraclass correlation coefficients (ICC) for their clustered data, and adjusted for this using accepted methods (Gulliford 1999).

We have sought statistical advice and been advised that the binary data from cluster trials presented in a report should be divided by a 'design effect'. This is calculated using the mean number of participants per cluster (m) and the ICC: thus, design effect = 1 + (m ‐ 1) × ICC (Donner 2002). If the ICC was not reported, we assumed it to be 0.1 (Ukoumunne 1999).

If cluster studies were appropriately analysed and ICCs and relevant data documented in the report taken into account, synthesis with other studies would have been possible using the generic inverse variance technique.

2. Cross‐over trials

A major concern of cross‐over trials is the carry‐over effect. This occurs if an effect (e.g. pharmacological, physiological, or psychological) of the treatment in the first phase is carried over to the second phase. As a consequence, participants can differ significantly from their initial state at entry to the second phase, despite a washout phase. For the same reason, cross‐over trials are not appropriate if the condition of interest is unstable (Elbourne 2002). As both carry‐over and unstable conditions are very likely in severe mental illness, we would only use data from the first phase of cross‐over studies.

3. Studies with multiple treatment groups

Where a study involved more than two treatment arms, we would present the additional treatment arms in comparisons if relevant. If data were binary, we would simply add these and combine within the two‐by‐two table. If the data were continuous, we would combine data following the formula in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Where additional treatment arms were not relevant, we would not reproduce these data.

Dealing with missing data

1. Overall loss of credibility

At some degree of loss of follow‐up, data must lose credibility (Xia 2009). We chose that, for any particular outcome, should more than 50% of data be unaccounted for, we would not reproduce these data or use them within analyses. However, if more than 50% of participants in one arm of a study were lost, but the total loss was less than 50%, we would address this within the 'Summary of findings' table/s by downgrading quality. Finally, we also downgraded quality within the 'Summary of findings' table/s when the loss was 25% to 50% in total.

2. Binary

Where attrition for a binary outcome was between 0% and 50% and these data were not clearly described, we presented data on a 'once‐randomised‐always‐analyse' basis (an intention‐to‐treat (ITT) analysis). Participants leaving the study early were all assumed to have the same rates of negative outcome as participants who completed, with the exception of the outcome of death and adverse effects. For these outcomes, the rate of participants who stayed in the study (in that particular arm of the trial) would be used for participants who did not. We undertook a sensitivity analysis testing how prone the primary outcomes were to change when data only from people who completed the study to that point were compared to the ITT analysis using the above assumptions.

3. Continuous

3.1. Attrition

We used data where attrition for a continuous outcome was between 0% and 50%, and data only from people who completed the study to that point were reported.

3.2. Standard deviations

If SDs were not reported, we would try to obtain the missing values from the authors. If these were not available, where there were missing measures of variance for continuous data, but an exact standard error (SE) and CIs available for group means, and either P value or t value available for differences in mean, we could calculate SDs according to the rules described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). When only the SE was reported, we calculated SDs using the formula SD = SE × √(n). The Cochrane Handbook for Systematic Reviews of Interventions presents detailed formulae for estimating SDs from P, t, or F values; CIs; ranges; or other statistics (Higgins 2011). If these formulae did not apply, we would calculate the SDs according to a validated imputation method which is based on the SDs of the other included studies (Furukawa 2006). Although some of these imputation strategies can introduce error, the alternative would be to exclude a given study's outcome and thus to lose information. Nevertheless, we examined the validity of the imputations in a sensitivity analysis that excludes imputed values.

3.3. Assumptions about participants who left the trials early or were lost to follow‐up

Various methods are available to account for participants who left the trials early or were lost to follow‐up. Some trials just present the results of study completers; other trials use the method of last observation carried forward (LOCF); whilst more recently, methods such as multiple imputation or mixed‐effects models for repeated measurements (MMRM) have become more of a standard. While multiple imputation or mixed‐effects model methods seem to be somewhat better than LOCF (Leon 2006), we consider that the high percentage of participants leaving the studies early and differences between groups in their reasons for doing so is often the core problem in randomised schizophrenia trials. We would therefore not exclude studies based on the statistical approach used. However, by preference, we would use the more sophisticated approaches, that is we would prefer to use MMRM or multiple imputation to LOCF, and we would only present completer analyses if some type of ITT data were not available at all. Moreover, we would address this issue in the item 'incomplete outcome data' of the 'Risk of bias' tool.

Assessment of heterogeneity

1. Clinical heterogeneity

We considered all included studies initially, without seeing comparison data, to judge clinical heterogeneity. We simply inspected all studies for participants who were clearly outliers or situations that we had not predicted would arise and, where these were found, discussed such situations or participant groups.

2. Methodological heterogeneity

We considered all included studies initially, without seeing comparison data, to judge methodological heterogeneity. We simply inspected all studies for clearly outlying methods that we had not predicted would arise and discussed any such methodological outliers.

3. Statistical heterogeneity

3.1. Visual inspection

We inspected graphs visually to investigate the possibility of statistical heterogeneity.

3.2. Employing the I² statistic

We investigated heterogeneity between studies by considering the I² statistic alongside the Chi² P value. The I² statistic provides an estimate of the percentage of inconsistency thought to be due to chance (Higgins 2003). The importance of the observed value of the I² statistic depends on the magnitude and direction of effects and the strength of evidence for heterogeneity (e.g. P value from Chi² test, or a CI for the I² statistic). We would interpret an I² statistic estimate of 50% or greater and accompanied by a statistically significant Chi² statistic as evidence of substantial heterogeneity, as described in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). When we found substantial levels of heterogeneity in the primary outcome, we explored reasons for heterogeneity (Subgroup analysis and investigation of heterogeneity).

Assessment of reporting biases

Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results (Egger 1997), as described in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Sterne 2011). We are aware that funnel plots may be useful in investigating reporting biases, but are of limited power to detect small‐study effects. We did not use funnel plots for outcomes where there are 10 or fewer studies, or where all studies were of similar size. In future updates, where funnel plots are possible, we will seek statistical advice on their interpretation.

Data synthesis

We understand that there is no closed argument for preference for use of fixed‐effect or random‐effects models. The random‐effects method incorporates an assumption that the different studies are estimating different, yet related, intervention effects. This often seems to be true to us, and the random‐effects model takes into account differences between studies, even if there is no statistically significant heterogeneity. However, a disadvantage of the random‐effects model is that it puts added weight onto small studies, which are often the most biased studies. Depending on the direction of effect, these studies can either inflate or deflate the effect size. We used a fixed‐effect model for all analyses.

Subgroup analysis and investigation of heterogeneity

1. Subgroup analyses

1.1. Primary outcomes

We did not conduct a subgroup analysis as we did not anticipate sufficient power to carry it out.

2. Investigation of heterogeneity

We reported if inconsistency was high. We first investigated whether data had been entered correctly. Secondly, if data were correct, we inspected the graph visually and removed outlying studies successively to see if homogeneity was restored. We decided that should this occur with data contributing to the summary finding of no more than 10% of the total weighting, we would present data. If not, we would not pool these data and would discuss any issues. We know of no supporting research for this 10% cut‐off but investigated use of prediction intervals as an alternative to this unsatisfactory state.

When unanticipated clinical or methodological heterogeneity was obvious, we would simply state hypotheses regarding this for future reviews or versions of this review. We did not anticipate undertaking analyses relating to this.

Sensitivity analysis

1. Implication of randomisation

We had hoped to include trials in a sensitivity analysis if they were described in some way that implied randomisation. For primary outcomes, if the inclusion of these trials had not resulted in a substantive difference, they would have remained in the analyses. If their inclusion had resulted in clinical, but not necessarily statistically significant differences, we would have not added data from these lower‐quality studies to the results of the higher‐quality trials, but presented these data within a subcategory.

2. Assumptions for lost binary data

Where assumptions have to be made regarding people lost to follow‐up (see Dealing with missing data), we, for the primary outcomes, would have compared the findings of the primary outcomes when we used our assumption compared with completer data only. If there had been a substantial difference, we would have reported the results and discussed them, but continued to employ our assumption.

Should assumptions have had to be made regarding missing SD data (see Dealing with missing data), we would have compared the findings of primary outcomes when we used our assumption compared with completer data only. We would have undertaken a sensitivity analysis to test how prone results are to change when completer data only are compared to the imputed data using the above assumption. If there had been a substantial difference, we would have reported the results and discussed them, but continued to employ our assumption.

3. Risk of bias

We had hoped to analyse the effects of excluding trials that are at high risk of bias across one or more of the 'Risk of bias' domains (see Assessment of risk of bias in included studies) for the meta‐analysis of the primary outcomes. If exclusion of trials at high risk of bias had not altered the direction of effect or the precision of the effect estimates substantially, then we would have included relevant data from these trials.

4. Imputed values

We would have undertaken a sensitivity analysis to assess the effects of including data from trials where we use imputed values for ICC in calculating the design effect in cluster‐randomised trials.

5. Fixed‐effect and random‐effects models

We synthesised data using a fixed‐effect model; however, we also wished to synthesise data for the primary outcomes using a random‐effects model to evaluate whether this altered the significance of the results.

Results

Description of studies

For details see Characteristics of included studies, Characteristics of excluded studies, and Characteristics of ongoing studies.

Results of the search

The results of our searches are also shown in Figure 1. The original search produced 118 reports. Following inspection of abstracts and, where required, full papers, we excluded 60 reports of 41 studies with reasons, and included 49 reports of seven studies in the review. We also identified nine reports of seven ongoing studies that may provide further data in future updates of this review.

1.

PRISMA flow diagram for 2018 and 2019 searches.

Included studies

1. Methods

All included studies were parallel, randomised controlled trials. Five of the included studies disclosed their method of randomisation. Two studies used a random number generator (General CR: Ahmed 2015; General CR: Bryce 2018); two studies used stratified randomisation (Aud CR: Fisher 2014; Aud CR: Vinogradov 2014); and one study used a computer‐generated randomisation list drawn up by a statistician (Kimhy 2018). Two included studies did not state a method of randomisation (Aud CR: Keefe 2012; Leutwyler 2017). None of the included studies described a method of allocation concealment. Blinding was generally well‐reported, with one study double‐blind and five single‐blind. One study did not specify any method of blinding (Leutwyler 2017).

2. Study duration

Study duration across the seven included trials varied: the shortest was Leutwyler 2017, where across six weeks participants spent a total of three hours in the intervention, whilst the longest was Kimhy 2018, where over 12 weeks a total of 36 hours was spent in the intervention. The most common study duration was 12 weeks (Aud CR: Keefe 2012; Kimhy 2018).

3. Settings

Only one study reported that they were focused on inpatients (General CR: Ahmed 2015), and one study stated that participants from mental health treatment settings were involved (Aud CR: Fisher 2014). The remaining studies either involved clinically stable patients or outpatients, or did not specify. The included studies were conducted in Australia, China, Spain, and the USA.

4. Participants

The total number of participants in the included studies was 468 (7 different trials that took place between 2009 and 2018). Study sizes varied from 32 participants to 121, with only one study having over a hundred participants (Aud CR: Fisher 2014). Participants were people with schizophrenia or other schizophrenia‐like illnesses, that is schizoaffective and schizophreniform disorder. Participants were typically diagnosed using Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM‐IV) Axis I diagnosis. The majority of studies included people aged between 18 and 60 years. Four of the included studies did not specify an age range; however, all studies reported a mean age. All studies reported including males and females, from the demographics presented in six of the seven studies there were 290 males to 91 females. Kimhy 2018 did not describe the number of female and male participants in whole numbers. Aud CR: Fisher 2014 only described the number of male and female participants who did not withdraw from the study.

5. Interventions

We included interventions that were called 'video games' or 'computer games'; for full inclusion criteria see Types of interventions. Four studies referred to 'computer games', whilst three referred to 'video games' (Kantrowitz 2016; Kimhy 2018; Leutwyler 2017).

5.1 Video games

5.1.1 Non‐exergames

The descriptions of the non‐exergames were inexact. General CR: Bryce 2018 used commercially available arcade and puzzle video games, whilst Aud CR: Vinogradov 2014 and Aud CR: Fisher 2014 used 16 different graphics‐based, commercially available games (e.g. visuospatial puzzle games, clue‐gathering mystery games). Aud CR: Keefe 2012 employed 10 computerised games that were stated to be "enjoyable". We identified no studies for inclusion that compared non‐exergames as the intervention against a control.

5.1.2 Exergames

We included two studies that employed exergames. Kimhy 2018 compared exergames versus standard care, whilst Leutwyler 2017 compared exergames with non‐exergames. Both studies used Xbox 360 Kinect (en.wikipedia.org/wiki/Kinect#Kinect_for_Xbox_360_(2010)) (Figure 2) to track and encourage participant movements.

2.

Xbox 360 Kinect.

5.2 Cognitive remediation

Five studies used cognitive remediation as an intervention. Three studies used the Brain Fitness Program developed by Posit Science (Aud CR: Fisher 2014; Aud CR: Keefe 2012; Aud CR: Vinogradov 2014): "The Brain Fitness Program focuses on the auditory system of the brain. It speeds up and sharpens the ability to take in speech, so that the brain can hear and remember more details" (Brain Fitness Program). General CR: Bryce 2018 used COGPACK v8.91 (Marker Software 2015), which "offers exercises to target visual‐motor functioning, comprehension, attention, memory, language use, and skills training" (psyberguide.org/apps/cogpack/). General CR: Ahmed 2015 used a cognitive remediation programme, but provided no further detail.

6. Outcomes

An array of outcome measures were implemented to clinically assess participant response, yet we were unable to report on a select number of outcome measures due to poor reporting of data. We have presented details of the outcome measures we were able to include below.

6.1 Functioning

6.1.1 General

• Independent Living Skills Survey ‐ ILSS (ILSS 2000)

The ILSS is a questionnaire designed to assess the functioning of severely mentally ill individuals. It rates how often the individual performs the tasks needed to live a satisfying, independent life in the community. These tasks typically include taking care of one's personal appearance, money, possessions, residence, and health; finding and keeping a job; and interacting with others. The questionnaire asks how often the individual performed each of these tasks during the past 30 days by choosing the options of: always, usually, often, sometimes, never or no opportunity.

• Medication Adherence Rating Scale ‐ MARS (Zemmour 2016)

The MARS is a 10‐item self‐reported questionnaire that incorporates the Medical Adherence Questionnaire and the Drug Attitude Inventory, both which are validated for individuals with psychosis. A higher score in the questionnaire infers a better adherence to medication.

• Strauss Carpenter Outcome Scale (Nieman 2013)

This is a 21‐item scale referring to aspects such as: quantity and quality of useful work, social class, social and heterosexual relationships, medical family history, violent or suicidal tendencies, flat effect, delusions, hallucinations, hypomania or mania in the previous year. Each item is rated using a 5‐point severity scale, with higher scores indicating a more positive prognosis.

• University of California San Diego Performance Based Skills Assessment ‐ UPSA (Patterson 2001)

The UPSA measures the capability to implement everyday functioning, assessing skills in five areas: communication, finance, household tasks, transportation, and free‐time activity planning. Raw scores from each of these subscales are transformed into a 0‐to‐10 scale, and are then doubled, and scores from all subscales are added to give a total summary score from 0 to 100. A higher score indicates a more capable individual.

6.1.2 Cognitive

• Cognitive Assessment Interview ‐ CAI (Ventura 2013)

The CAI was produced from two parent interview assessments: the CGI‐CogS and the SCoRS. It uses 10 items to assess the majority of the MATRICS cognitive domains (6 of 7) and is rated using a 7‐point severity scale, with higher scores indicating worse cognitive functioning.

• MATRICS Consensus Cognitive Battery ‐ MCCB (McCleery 2014)

The MCCB is compiled of 10 tests that assess seven areas of cognition: processing speed, attention, working memory, verbal learning, reasoning and problem solving, visual learning, and social cognition. The MCCB also gives an overall composite score across all these domains. A higher score indicates better cognitive functioning. Four studies reported outcome data from this scale. Two further studies were conducted before the full MCCB was available (Aud CR: Fisher 2014; Aud CR: Vinogradov 2014), and instead used measures consistent with the MCCB test that had been published and recommended by MATRICS before the full release of the test.

6.1.3 Social

• Global Functioning: Role and Social Scales (Cornblatt 2007)

The Global Functioning Role and Social scales are clinician‐rated metrics that take into account age and stage of illness. The role scale assesses performance in school, work, or as a homemaker, depending on age. The social scale assesses quantity and quality of peer relationships, level of peer conflict, ageappropriate intimate relationships, and involvement with family members. One Study re[prted outcome data from these scales.

• Social and Occupational Functioning Assessment Scale ‐ SOFAS (Rybarczyk 2011)

The SOFAS focuses on the social and occupational functioning of the patient independent from the severity of the patient's illness. The score ranges from 0 to 100, with a lower score indicating more diminished level of functioning. One study reported outcome data from this scale.

• Specific Levels of Functioning Scale ‐ SLOF (Mucci 2014)

The SLOF consists of 43 items and does not focus on cognitive functioning or symptoms, but rather a person's skills, assets, and abilities. The scale measures physical functioning, personal care, social relationships, social acceptability, recreational activities, and work skills and is rated with a 5‐point Likert scale, with a higher score indicating higher function. One study reported outcome data from this scale.

6.2 Mental state

• Positive and Negative Syndrome Scale ‐ PANSS (Kay 1986)

The PANSS consists of 30 items and is rated using a seven‐point severity scale, from absent to severe. It measures positive symptoms, negative symptoms, and an overall total score of general psychopathology, with a higher score indicating more severe symptoms. Four studies reported an average total score from this scale, whilst two studies collected separate data for both positive and negative symptoms (General CR: Ahmed 2015; General CR: Bryce 2018).

6.3 Physical fitness

• Aerobic fitness ‐ AF (Kimhy 2018)

AF is a measure of various sustained exercises, such as jogging, rowing, swimming, or cycling, that stimulate and strengthen the heart and lungs, thereby improving the body's utilisation of oxygen. AF is measured as peak oxygen uptake (VO₂ peak) mL/kg/min.

• Brain‐derived neurotrophic factor ‐ BDNF (Kimhy 2018)

BDNF is the most abundant of the growth factor family and has a wide repertoire of neurotrophic and neuroprotective properties in the central nervous system. Amongst individuals with schizophrenia, reports point to lower serum BDNF being linked to poor memory and smaller hippocampal volumes.

• Short Physical Performance Battery ‐ SPPB (Leutwyler 2017)

The SPBB is a valid and reliable test for quantifying function and mobility and is useful in following change in walking speed over time. The test can also be useful for identifying individuals in a 'preclinical' stage of disability who may be ideal to target with disability‐delaying interventions.

6.4 Global state

• Clinical Global Impression Scale ‐ CGI (Guy 2000)

The CGI measures symptom severity, treatment response, and the efficacy of treatments in treatment studies of individuals with mental disorders.

6.5 Quality of life

• EUROHIS‐QOL (Power 2003)

Part of the WHO EUROHIS project to develop common survey instruments that could enhance the international comparability of national health data, the EUROHIS‐QOL scale is a shortened version of the World Health Organization Quality of Life Instrument‐Abbreviated Version (WHOQOL‐BREF). It is a survey using an eight‐item index to measure quality of life. One study reported data from this scale.

• Perceived Competence Scale ‐ PCS (http:// feelings of being able to stick with a treatment regimen and being able to master the material in a course.)

The PCS is a short, 4‐item questionnaire that evaluates the feelings of being able to stick with a treatment regimen and master the material in a course.

• Quality of Life Scale ‐ QLS (Heinrichs 1984)

The QLS consists of 21 items where participants provide information on symptoms and functioning through a semi‐structured interview. It is rated using a 7‐point scale, with 5 to 6 being relatively good functioning and 0 to 1 being relatively bad. One study reported outcome data from this scale. One study used an alternate version of this scale, the European Health Interview Survey (EUROHIS‐QOL), which is based on the same scale (General CR: Bryce 2018).

• Intrinsic Motivation Inventory‐Schizophrenia Research Intrinsic Motivation Inventory ‐ IMI‐SR (Monteiro 2015)

The IMI is a multidimensional measurement grounded on self‐determination theory that is used in assessing the subjective experiences of participants when developing an activity.

• Revised Self‐Efficacy Scale (Cardenas 2013)

This is a self‐report scale focusing on a participant’s confidence in their ability to complete everyday living and social behaviours (e.g. 'attend classes' and 'make friends') (Cardenas 2013). Items are rated on a 5‐point scale ranging from 1 (not at all confident) to 5 (extremely confident). Higher scores represent higher levels of self‐efficacy.

• Rosenberg Self‐Esteem Scale (The Betsi Cadwaladr University Health Board 2017)

This is a self‐esteem (en.wikipedia.org/wiki/Self-esteem) measure widely used in social science research. Individuals are rated on a scale of 0 to 30, where a score of less than 15 may indicate a problematic low self‐esteem.

7. Missing outcomes

Data were reported on the majority of primary outcomes. No data were reported for adverse effects, so we used study attrition rates as a poor‐quality metric for adverse effects. No economic or service use outcomes were reported.

Excluded studies

We excluded 41 studies from the review. We excluded four studies because of lack of randomisation (Han 2008; Heggelund 2011; Saleem 2013; Wu 2016). We excluded nine studies because participants did not have schizophrenia or not a large enough sample of participants had schizophrenia, or data for those with schizophrenia were not reported separately. We excluded 18 studies because the intervention was not a 'video game' or 'computer game'. We excluded one study because the intervention was a virtual reality game (Lopez 2016). Regrettably, we excluded six studies because no useable outcome data were reported. In Barnett 1978, both the participants were not diagnosed with schizophrenia and the intervention was not a video game.

Studies awaiting classification

There are currently no studies awaiting classification.

Ongoing studies

We have extracted some data from seven ongoing studies that were due to be completed by December 2019. However, three of these 11 studies (listed) did not specify an estimated completion date (Engh 2015; Kuehn 2018; Valimaki 2017). One of these (Valimaki 2017) has had its protocol recently accepted in the BMC, the study is still ongoing due to political and COVID‐19 related restrictions.

Risk of bias in included studies

Our overall response to the risk of bias assessment in the included studies is also shown in Figure 3 and Figure 4.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

4.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

All included studies reported randomisation. The method of randomisation used in the studies was generally well reported, with only two trials not fully disclosing their methods for randomisation (Aud CR: Keefe 2012; Leutwyler 2017). No study disclosed the method of allocation concealment, therefore we have rated all included studies as at unclear risk for allocation concealment.

Blinding

Five studies described single‐blinding (of assessors) without blinding of participants; one of these studies had conflicting documentation in different reports regarding blinding (Aud CR: Vinogradov 2014), but we assessed this trial as being single‐blinded. One study reported double‐blinding (General CR: Ahmed 2015), however given the nature of the intervention and control, it is not clear how this was achieved. One study did not document any blinding (Leutwyler 2017).

The necessary lack of full blinding in most of the studies led to an unclear risk of bias in this area ‐ most studies included assessor blinding, with mixed success (where effectiveness of blinding was reported), but not participant blinding. Where it was stated that participants were blinded, it was not clear how this was achieved given the nature of the intervention and control. We have therefore assessed all included trials as at unclear risk for blinding.

Incomplete outcome data

We assessed five studies that used an intention‐to‐treat (ITT) analysis as at low risk of attrition bias (Aud CR: Fisher 2014; Aud CR: Vinogradov 2014; General CR: Ahmed 2015; Kimhy 2018; Leutwyler 2017). We rated two studies that did not specify how they accounted for attrition within their final data as at unclear risk of attrition bias (Aud CR: Vinogradov 2014; General CR: Bryce 2018).

Selective reporting

Although we have not been able to obtain any protocols for the included studies, four of the seven included trials reported all useable outcome data and were therefore rated as at low risk of reporting bias (Aud CR: Keefe 2012; General CR: Ahmed 2015; Kimhy 2018; Leutwyler 2017). We assessed two studies as at high risk of reporting bias: General CR: Bryce 2018 stated "Three quarters (77%) of CR [cognitive remediation] completers showed a reliable improvement" with no mention of the proportion in the control group using only video games, whilst Aud CR: Vinogradov 2014 reported PANSS and quality of life at baseline and not postintervention.

Other potential sources of bias

We noted another potential source of bias for three studies that involved auditory‐focused cognitive remediation (Aud CR: Fisher 2014; Aud CR: Keefe 2012; Aud CR: Vinogradov 2014). All of these studies, whose authors overlap, used the same software, and involved at least one researcher with a financial connection to the company responsible for this software. The company in question provided the software free of charge in two of these studies (Aud CR: Fisher 2014; Aud CR: Vinogradov 2014). Whilst these conflicts of interest were properly and openly declared, we have still noted them as a potential source of bias.

Aud CR: Fisher 2014 used 'alternative forms' of tests from the MCCB and did not reassure the reader as to the validity of these tests, so we did not include these in the analysis.

Effects of interventions

See: Table 1; Table 2; Table 3

1. COMPARISON 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term)

1.1 Functioning: 1a. General ‐ average endpoint score (various scales, high = good) ‐ by end of treatment

Three studies measured general functioning using four different general functioning scales (Analysis 1.1).

1.1. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 1: Functioning: 1a. General ‐ average endpoint score (various scales, high = good) ‐ by end of treatment

1.1.1 ILSS

There was no clear difference between video games and cognitive remediation for this outcome (mean difference (MD) 0.03, 95% confidence interval (CI) −0.04 to 0.09; participants = 43; studies = 1).

1.1.2 MARS

There was no clear difference between video games and cognitive remediation for this outcome (MD −0.09, 95% CI −0.42 to 0.24; participants = 78; studies = 1).

1.1.3 Strauss Carpenter Outcome Scale

There was no clear difference between video games and cognitive remediation for this outcome (MD 0.42, CI −0.62 to 1.46; participants = 86; studies = 1, very low‐quality evidence).

1.1.4 UPSA ‐ total

There was no clear difference between video games and cognitive remediation for this outcome (MD 1.72, 95% CI −29.92 to 33.36; participants = 53; studies = 1).

1.1.5 UPSA ‐ communication

There was no clear difference between video games and cognitive remediation for this outcome (MD −0.95, 95% CI −3.33 to 1.43; participants = 78; studies = 1).

1.1.6 UPSA ‐ comprehension/planning

There was no clear difference between video games and cognitive remediation for this outcome (MD −1.90, 95% CI −3.97 to 0.17; participants = 78; studies = 1).

1.1.7 UPSA ‐ financial management

There was no clear difference between video games and cognitive remediation for this outcome (MD 0.35, 95% CI −1.71 to 2.41; participants = 78; studies = 1).

1.1.8 UPSA ‐ transportation (high = good)

There was no clear difference between video games and cognitive remediation for this outcome (MD −0.88, 95% CI −3.05 to 1.29; participants = 78; studies = 1).

1.2 Functioning: 1b. General ‐ average endpoint score (ILSS, high = good) ‐ 3‐month follow‐up

One study reported average ILSS endpoint scores at three‐month follow‐up. There was no clear difference between video games and cognitive remediation for this outcome (MD 0.04, 95% CI −0.03 to 0.12; participants = 33; studies = 1) (Analysis 1.2).

1.2. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 2: Functioning: 1b. General ‐ average endpoint score (ILSS, high = good) ‐ 3‐month follow‐up

1.3 Functioning: 2a. Specific ‐ cognitive ‐ improved (MCCB, at least one domain) ‐ by end of treatment

In one small study, a greater number of participants in the cognitive remediation group demonstrated improvement in at least one domain of the MCCB (risk ratio (RR) 0.58, 95% CI 0.34 to 0.99; participants = 42; studies = 1, low‐quality evidence) (Analysis 1.3).

1.3. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 3: Functioning: 2a. Specific ‐ cognitive ‐ improved (MCCB, at least one domain) ‐ by end of treatment

1.4 Functioning: 2b. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite score (high = good) ‐ by end of treatment

Three studies reported composite scores on the MCCB. There was a clear difference in scores by end of treatment favouring the cognitive remediation group (MD −3.60, 95% CI −6.43 to −0.76; participants = 174; studies = 3; I² = 44%) (Analysis 1.4).

1.4. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 4: Functioning 2b. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite, high = good) ‐ by end of treatment

1.5 Functioning: 2c. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite score) ‐ 3‐month follow‐up

One study reported on cognitive functioning at three‐month follow‐up. There was no clear difference in average endpoint MCCB‐composite scores between video game and cognitive remediation groups (MD −2.69, 95% CI −9.52 to 4.14; participants = 33; studies = 1) (Analysis 1.5).

1.5. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 5: Functioning 2c. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite, high = good) ‐ 3‐month follow‐up

1.6 Functioning: 2d. Specific ‐ cognitive ‐ average endpoint score (MATRICS ‐ Z score, high = good) ‐ by end of treatment

For Z scores on the MATRICS recommended measures that were available before the release of the full MCCB, there appeared to be no clear difference between video games and cognitive remediation, but data were skewed and have been reported in tabular form (participants = 173; studies = 2) (Analysis 1.6).

1.6. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 6: Functioning 2d. Specific ‐ cognitive ‐ average endpoint score (MATRICS ‐ z scores, high = good) ‐ by end of treatment (skewed data)

Functioning 2d. Specific ‐ cognitive ‐ average endpoint score (MATRICS ‐ z scores, high = good) ‐ by end of treatment (skewed data)
Study	Intervention	Mean	SD	N
Aud CR: Fisher 2014	Video games (commercial non‐exergame)	‐0.87	1.0	43
	Cognitive remediation	‐0.46	0.73	43
Aud CR: Vinogradov 2014	Video games (commercial non‐exergame)	‐0.89	0.75	41
	Cognitive remediation	‐0.84	0.77	46

1.7 Functioning: 2e. Specific ‐ cognitive ‐ average endpoint score (CAI, high = poor) ‐ by end of treatment

There was no clear difference between video games and cognitive remediation for this outcome (MD 0.24, 95% CI −0.25 to 0.73; participants = 53; studies = 1) (Analysis 1.7).

1.7. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 7: Functioning 2e. Specific ‐ cognitive ‐ average endpoint score (CAI, high = poor) ‐ by end of treatment

1.8 Functioning: 3a. Specific ‐ social ‐ average endpoint score (Global Functioning: Role and Social Scales, high = good) ‐ by end of treatment

One study measured social functioning using the Role and Social Scales (Analysis 1.8).

1.8. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 8: Functioning: 3a. Specific ‐ social ‐ average endpoint score (Global Functioning: Role and Social Scale, high = good) ‐ by end of treatment

1.8.1 Global functioning role

There was no clear difference between video games and cognitive remediation for this outcome (MD −0.08, 95% CI −1.15 to 0.99; participants = 86; studies = 1).

1.8.2 Global functioning social

There was no clear difference between video games and cognitive remediation for this outcome (MD 0.16, 95% CI −0.43 to 0.75; participants = 86; studies = 1).

1.9 Functioning: 3b. Specific ‐ social ‐ average endpoint score (SLOF, high = good) ‐ by end of treatment

One study measured social functioning using the SLOF. These data were skewed to some degree, with the video games group having moderate skew and the control group not exhibiting skew. We were not sure if we should remove these data and report them in tabular form (as per Data extraction and management), but in the end we left them in the analysis. There was no clear difference between video games and cognitive remediation for this outcome (MD −3.13, 95% CI −40.17 to 33.91; participants = 53; studies = 1, very low‐quality evidence) (Analysis 1.9).

1.9. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 9: Functioning: 3b. Specific ‐ social ‐ average endpoint score (SLOF, high = good) ‐ by end of treatment

1.10 Mental state: 1a. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ by end of treatment

Four studies reported mental state using the PANSS total. There was no clear difference between the average endpoint scores of participants in the video game groups and those of participants in the cognitive remediation groups (MD 0.20, 95% CI −3.89 to 4.28; participants = 269; studies = 4, low‐quality evidence) (Analysis 1.10). The Aud CR: Keefe 2012 data were very skewed. Their means were similar to the other studies, but the standard deviation (SD) denotes considerable skew. We checked these data, and they are the only tables that clearly state that the variance is reported as a standard error, and from this we have calculated the SD. The other studies clearly state that they are reporting the SD. Again, we were unsure how best to include these data, and again have left them in the analyses as their removal made no substantial difference to the final result (MD 0.15, 95% CI −3.95 to 4.25).

1.10. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 10: Mental state: 1a. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ by end of treatment

1.11 Mental state: 1b. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ 3‐month follow‐up

One trial reported endpoint PANSS total scores at three‐month follow‐up. There was still no clear difference between treatment groups (MD −5.58, 95% CI −15.31 to 4.15; participants = 33; studies = 1) (Analysis 1.11).

1.11. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 11: Mental state: 1b. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ 3‐month follow‐up

1.12 Mental state: 2a. Specific ‐ negative ‐ average endpoint (PANSS negative, high = poor) ‐ by end of treatment

Two studies reported on negative symptoms using the PANSS negative subscale. There was no clear difference between treatment groups for this outcome (MD 0.45, 95% CI −1.63 to 2.53; participants = 121; studies = 2) (Analysis 1.12).

1.12. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 12: Mental state: 2a. Specific ‐ negative ‐ average endpoint score (PANSS negative, high = poor) ‐ by end of treatment

1.13 Mental state: 2b. Specific ‐ positive ‐ average endpoint (PANSS positive, high = poor) ‐ by end of treatment

Two studies reported on positive symptoms using the PANSS positive subscale. There was no clear difference between treatment groups for this outcome (MD −0.34, 95% CI −2.21 to 1.54; participants = 121; studies = 2) (Analysis 1.13).

1.13. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 13: Mental state: 2b. Specific ‐ positive ‐ average endpoint score (PANSS positive, high = poor) ‐ by end of treatment

1.14 Global state: Average endpoint score (CGI ‐ severity, high = poor) ‐ by end of treatment

One study reported CGI‐severity average endpoint scores. There appeared to be no clear difference between video games and cognitive remediation, but data were very skewed and have been presented in tabular form (participants = 53; studies = 1) (Analysis 1.14).

1.14. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 14: Global state: Average endpoint score (CGI ‐ severity, high = poor) ‐ by end of treatment (skewed data)

Global state: Average endpoint score (CGI ‐ severity, high = poor) ‐ by end of treatment (skewed data)
Study	Intervention	Mean	SD	N
Aud CR: Keefe 2012	Video games (commercial non‐exergame)	3.61	4.79	26
	Cognitive remediation	3.12	16.21	27

1.15 Leaving the study early: Did not leave

Five of the seven studies provided useable attrition data (Analysis 1.15).

1.15. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 15: Leaving the study early: Did not leave

1.15.1 By end of treatment

Five studies reported numbers of participants who did not leave the studies before the end of treatment. There was no clear difference between groups in attrition (RR 0.96, 95% CI 0.87 to 1.06; participants = 395; studies = 5).

1.15.2 Follow‐up (3 to 9 months)

One study reported the number of participants available for follow‐up three to nine months after treatment. There was also no clear difference between groups in attrition (RR 1.29, 95% CI 0.83 to 2.00; participants = 56; studies = 1).

1.16 Quality of life: 1a. Overall ‐ average endpoint score (QLS, high = good) ‐ by end of treatment

Quality of life data reported by one study using the QLS showed no clear difference between video games and cognitive remediation (MD 0.01, CI −0.40 to 0.42; participants = 87; studies = 1, very low‐quality evidence) (Analysis 1.16).

1.16. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 16: Quality of life: 1a. Overall ‐ average endpoint score (QLS, high = good) ‐ by end of treatment

1.17 Quality of life: 1b. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ by end of treatment

Similarly, when using the EUROHIS‐QOL score to evaluate quality of life, there was no clear difference between video games and cognitive remediation (MD −0.32, 95% CI −3.34 to 2.70; participants = 43; studies = 1) (Analysis 1.17).

1.17. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 17: Quality of life: 1b. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ by end of treatment

1.18 Quality of life: 1c. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ 3‐month follow‐up

EUROHIS‐QOL quality of life data reported by one study at three‐month follow‐up showed no clear difference between video games and cognitive remediation (MD 0.58, 95% CI −2.69 to 3.85; participants = 33; studies = 1) (Analysis 1.18).

1.18. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 18: Quality of life: 1c. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ 3‐month follow‐up

1.19 Quality of life: 2a. Specific ‐ well‐being ‐ improved (IMI‐SR increase in score) ‐ by end of treatment

One study reported improvement in specific aspect of quality of life using increase in IMI‐SR scores; the numbers of participants showing improvement were similar between treatment groups (RR 1.50, 95% CI 0.70 to 3.20; participants = 43; studies = 1) (Analysis 1.19) .

1.19. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 19: Quality of life: 2a. Specific ‐ well‐being ‐ improved (IMI‐SR increase in score, high = good) ‐ by end of treatment

1.20 Quality of life: 2b. Specific ‐ well‐being ‐ average endpoint score (various scales, high = good) ‐ by end of treatment

Two studies used various scales measuring specific aspects of quality of life, reporting average endpoint scores at the end of treatment for intrinsic motivation (IMI‐SR), competency (PCS), self‐esteem, and self efficacy. There was no clear difference between video game and cognitive remediation groups for any of these measures (Analysis 1.20).

1.20. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 20: Quality of life: 2b. Specific ‐ well‐being ‐ average endpoint score (various scales, high = good) ‐ by end of treatment

1.20.1 IMI‐SR

(MD 5.16, 95% CI −1.22 to 11.54; participants = 96; studies = 2).

1.20.2 PCS

(MD 0.36, 95% CI −2.62 to 3.34; participants = 53; studies = 1).

1.20.3 Rosenberg Self‐Esteem Scale

(MD −0.69, 95% CI −3.34 to 1.96; participants = 53; studies = 1).

1.20.4 Revised Self‐Efficacy Scale

(MD 9.41, 95% CI −4.85 to 23.67; participants = 43; studies = 1).

1.21 Quality of life: 2c. Specific ‐ well‐being ‐ average endpoint score (Revised Self‐Efficacy Scale) ‐ 3‐month follow‐up

There was also no difference between treatment groups for self‐efficacy scores at three‐month follow‐up (MD 6.13, 95% CI −7.17 to 19.43; participants = 33; studies = 1) (Analysis 1.21).

1.21. Analysis.

Comparison 1: VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term), Outcome 21: Quality of life: 2c. Specific ‐ well‐being ‐ average endpoint score (Revised Self‐Efficacy Scale, high = good) ‐ 3‐month follow‐up

2. COMPARISON 2: VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term)

2.1 Functioning: 1a. Specific ‐ cognitive ‐ average change score (MCCB, increase = good)

Change scores on the MCCB reported by one study showed no clear difference between treatment groups (MD 2.90, 95% CI ‐1.27 to 7.07; participants = 33; studies = 1, low‐quality evidence) (Analysis 2.1).

2.1. Analysis.

Comparison 2: VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term), Outcome 1: Functioning: 1a. Specific ‐ cognitive ‐ average change score (MCCB, increase = good)

2.2 Physical fitness: 2a. Specific ‐ average change (BDNF, increase = good)

The same study measured change in BDNF, finding no clear difference between exergames and standard care for this outcome (MD ‐0.13, 95% CI −3.43 to 3.17; participants = 33; studies = 1) (Analysis 2.2).

2.2. Analysis.

Comparison 2: VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term), Outcome 2: Physical fitness: 2a. Specific ‐ average change (BDNF, increase = good)

2.3 Physical fitness 2b. Specific ‐ average change VO₂ peak (AF, increase = good)

There was a clear difference between treatment groups favouring exergames for aerobic fitness change (MD 3.82, 95% CI 1.75 to 5.89; participants = 33; studies = 1, low‐quality evidence) (Analysis 2.3).

2.3. Analysis.

Comparison 2: VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term), Outcome 3: Physical fitness: 2b. Specific ‐ average change VO₂ peak (AF, increase = good)

2.4 Leaving the study early: Did not leave

There was no clear difference in attrition rates (RR 1.06, 95% CI 0.75 to 1.51; participants = 33; studies = 1, low‐quality evidence) (Analysis 2.4).

2.4. Analysis.

Comparison 2: VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term), Outcome 4: Leaving the study early: Did not leave ‐ by end of treatment

3. COMPARISON 3: VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term)

3.1 Physical fitness: 1a. Specific ‐ walking speed ‐ average time (seconds taken to walk 3 metres, low = good)

There was no clear difference in average walking speeds of the participants (MD −0.50, 95% CI −1.17 to 0.17; participants = 28; studies = 1, very low‐quality evidence) (Analysis 3.1).

3.1. Analysis.

Comparison 3: VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term), Outcome 1: Physical fitness: 1a. Specific ‐ walking speed ‐ average time (seconds taken to walk 3 metres, low = good)

3.2 Physical fitness: 1b. Specific ‐ average endpoint score (SPPB, high = good)

Similarly, there was no clear difference between the two types of video games for physical fitness, as measured indirectly by the SPPB (MD −0.30, 95% CI −1.98 to 1.38; participants = 28; studies = 1) (Analysis 3.2).

3.2. Analysis.

Comparison 3: VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term), Outcome 2: Physical fitness: 1b. Specific ‐ average endpoint score (SPPB, high = good)

3.3 Behaviour: 2a. Specific ‐ average number of sessions attended (of 6 total)

There was no clear difference between exergames and non‐exergames in session attendance (MD −0.10, 95% CI −0.48 to 0.28; participants = 28; studies = 1) (Analysis 3.3).

3.3. Analysis.

Comparison 3: VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term), Outcome 3: Behaviour: 2a. Specific ‐ average number of sessions attended (of 6 in total)

3.4 Behaviour: 2b. Specific ‐ average time (minutes) at sessions (of 180 minutes in total)

There was likewise no clear difference between exergames and non‐exergames in session attendance when measured in minutes of sessions attended (MD −2.20, 95% CI −17.66 to 13.26; participants = 28; studies = 1) (Analysis 3.4).

3.4. Analysis.

Comparison 3: VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term), Outcome 4: Behaviour: 2b. Specific ‐ average time (minutes) at sessions (of 180 minutes in total)

4. Sensitivity analysis

4.1 Implication of randomisation

In the event no trial was included ‐ or excluded ‐ that implied randomisation. All clearly stated that they were, or were not, randomised. No sensitivity analysis was undertaken

4.2 Assumptions for lost binary data

Relatively small numbers of people were lost to follow up in these studies. However, not one of the studies across any of the three comparisons provided primary binary outcomes to allow us to keep to the directions of our pre‐published protocol (see Sensitivity analysis). The great majority of outcomes were continuous measures and so made our pre‐specified sensitivity analyses impossible. We do not have any impression that the modest loss to follow up in these studies would have made any substantive difference to any of the results.

4.3 Risk of bias

We rated three trials (General CR: Ahmed 2015; Kimhy 2018; Leutwyler 2017) to not be at high risk in one domain. We felt all others carried a high risk in one or more domains. The Kimhy 2018 and Leutwyler 2017 studies reported data for Comparison 2 and 3 and were the only studies to do so – so could not be analysed separately from trials which carried some clear risk of bias. General CR: Ahmed 2015, however, was part of several outcomes in Comparison 1. Although there were no outcomes reported that we had pre‐specified as being of primary importance (see Sensitivity analysis, outcome Analysis 1.4 was the one in which this trial was part of a meta‐analysis and removing its two less good quality companion studies did make the overall finding more clearly favour the cognitive remediation group (MD ‐9.20 CI‐15.84 to ‐2.56 vs MD ‐3.60 CI ‐6.43 to ‐0.76), but not statistically significantly so and the finding was generally the same – favouring cognitive remediation.

4.4 Imputed values

No trials used a cluster randomised technique.

4.5 Fixed‐effect and random‐effects models

Again – no primary outcomes were reported by any trial. We did undertake a sensity analysis on one continuous outcome (Analysis 1.4) to investigate if using the random model made a difference and it did. The original MD ‐3.60 CI ‐6.43 to ‐0.76 using the fixed model became MD ‐3.96 CI ‐8.12 to 0.20, further illustrating the instability of these results. We feel that this should not be over‐interpreted as this is not on a pre‐specified primary outcome and is opportunistic on what data happen to be reported across a few trials. Nevertheless it does cast some shadow across any confidence in being sure that video games are indeed worse than cognitive remediation when continuous measures of functioning are the outcome.

Discussion

Summary of main results

1. VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) for schizophrenia (all short term)

See Table 1.

Overall, we found little evidence of clear differences between commercial, non‐exergame video games and any of the complex cognitive remediation interventions. Only one outcome, cognitive functioning, showed a clear difference between groups. We recognise that we pre‐stated only a few outcomes to be of key importance, however few of the other outcomes showed any clear effect reaching conventional levels of statistical significance. We also recognise that several of the included studies were not designed to be anything more than exploratory. Nevertheless, such studies do still have potential, especially when in the company of all others, to reveal effects of clinical importance.

1.1 Functioning ‐ general ‐ average endpoint score (Strauss Carpenter Outcome Scale)

We felt this outcome to be of primary importance to this intervention. Only one small trial (n = 86) evaluated this outcome, and reported a continuous outcome, rather than the binary, clinically important change that we prespecified. Further undermining the value of these continuous data is that the scale used is not well‐explained in terms of clinical meaning. We rated these data to be of very low quality and identified no difference between the commercial games and the more complex and expensive cognitive interventions.

1.2 Functioning ‐ specific ‐ cognitive ‐ improved and endpoint score (MCCB)

Three studies evaluated this outcome, which was used to assess cognitive functioning. Two further studies reported z score data for MATRICS‐recommended measures that pre‐date the MCCB.

Only one study reported binary data, which was measured as improved on at least one domain. There was a clear difference between treatment groups favouring cognitive remediation. As this was the only source of binary data for this outcome, the data are presented in the 'Summary of findings' table but rated as of low quality due to the poor description of randomisation and allocation concealment provided, and the small and imprecise nature of the study. We are therefore uncertain about this estimate of effect, and the true effect of cognitive remediation on cognitive functioning may be substantially different.

Three studies reported continuous data for the MCCB, as opposed to the binary we prespecified in our protocol. We also found these data to be heterogeneous (I² = 44%). Nevertheless, there was weak evidence in these low‐quality, continuous data of a beneficial effect of cognitive remediation over video games.

1.3 Functioning ‐ specific ‐ social ‐ average endpoint score (SLOF)

There was no clear difference between video games and cognitive remediation for this outcome. Only one small trial used the scale measuring social functioning (n = 53). The SLOF itself is substantial, with 43 items assessing various domains of a person's skills, assets, and abilities (Mucci 2014). However, despite the use of 43 items, the clinical meaning of the results remains unclear; we have identified no explanation of what a three‐point difference in rating would mean clinically. We rated the data as of very low quality due to the indirectness of the continuous scale as well as imprecision seen by the width of the confidence intervals. Furthermore, we downgraded the quality of the evidence due to the poor description of randomisation and allocation concealment in the study itself which highlights potential risk of bias. We are very uncertain if this estimate of effect is a true effect.

1.4 Adverse effects

None of the included studies measured adverse effects directly.

We used leaving the study early as an indirect measure. Five of the included studies addressed loss of participants during the trial, with one study reporting loss of participants for up to a three‐month follow‐up (General CR: Bryce 2018). We have rated this evidence to be low quality due to the majority of studies not accounting for how the loss of participants affected the final result. Furthermore, loss of participants represents a poor metric for adverse effects, and the relevant studies had largely poor descriptions of randomisation and allocation concealment. Again, there was no clear difference between groups. We are uncertain if this estimate of effect is a true effect.

1.5 Mental state ‐ overall ‐ average endpoint score (PANSS)

PANSS data were well reported, with four studies giving an overall composite score. Despite being the outcome with the greatest number of participants, data showed no clear distinction between video games and cognitive interventions for effect on overall mental state. Although PANSS is a well‐established scale for assessing positive and negative symptoms for people with schizophrenia, we have rated the evidence as low quality because, once again, data were continuous, as opposed to a binary outcome as pre‐stated in our protocol. Furthermore, the three studies that made up roughly 90% of the weight are at high risk of reporting bias and provide no description of allocation concealment. We are uncertain if this estimate of effect is a true effect.

1.6 Physical fitness

No studies in this comparison reported an outcome measuring physical fitness. This is unsurprising as the interventions in this comparison do not involve bodily movement.

1.7 Quality of life ‐ overall ‐ average endpoint score (QLS) ‐ short term

One study with 87 participants reported relevant QLS data. There was no clear, statistically significant difference between cognitive remediation and video games for this outcome. We rated the evidence to be of very low quality. Once again, the scale was continuous, and not the binary scale we prespecified in our protocol. The study was also assessed as being at high risk of reporting bias and did not specify its allocation concealment. We are very uncertain if this estimate of effect is a true effect.

1.8 Unable to use

Data for aggression were reported using the Overt Aggression Scale (General CR: Ahmed 2015). However, there were only subscores given and not an overall rating for aggression; as per our protocol we did not analyse these subscore data but presented this information as additional data (Table 4). Numerous studies have investigated links between video games and an increased composite aggression score (Calvert 2017); it is therefore surprising that outcomes concerning adverse effects are so scarce. We propose that the reasoning for this here is that the video gaming intervention was the control intervention in the vast majority of studies included in this review, and therefore outcomes associated with video games were not the primary focus.

1. Subscale data ‐ aggression ‐ average endpoint score (Overt Aggression Scale, high = poor).

	Video games (non‐exergames)			Cognitive remediation
	Mean	SD	N	Mean	SD	N
Object	0.14	0.42	36	0.02	0.15	42
Physical	0.61	1.08	36	0.17	0.49	42
Self‐aggression	0.03	0.17	36	0.00	0.00	42
Verbal	1.00	1.43	36	0.50	1.29	42

SD: standard deviation

2. VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE for schizophrenia (all short term)

See Table 2.

2.1 Functioning ‐ general

Regrettably, no studies reported relevant data for this outcome.

2.2 Functioning ‐ specific ‐ cognitive ‐ average change score (MCCB, increase = good)

In one small (n = 33) study reported data for this outcome, but did not find any clear difference between treatment groups. Data were of low quality due to the small and imprecise nature of the study, and the use of continuous rather than binary data. We are uncertain if this is a true effect.

2.3 Functioning ‐ specific ‐ social

Regrettably, no studies reported relevant data for this outcome.

2.4 Adverse effects

Again, no direct measures of adverse effects were reported.

One small trial presented data regarding loss, with no statistically significant difference between groups. We rated this evidence as of low quality due to loss to follow‐up being a poor metric for adverse effects, and the small and imprecise nature of the trial in question. We are uncertain if this is a true effect.

2.5 Mental state

No studies reported relevant data for this outcome.

2.6 Physical fitness ‐ specific ‐ average change VO₂ peak (AF, increase = good)

One small trial reported data for aerobic fitness, with results favouring the exergame group. However, the sample size was very small and the evidence of low quality. We are uncertain if this is a true effect.

2.7 Missing outcomes

It is unclear how mental state, quality of life, and general or social functioning are affected by playing exergames. This is an active area of research, and our search found an ongoing study investigating exergames versus an exercise group control (Engh 2015). Their protocol stated that they will measure all of the outcomes that were missing from the one exergame study we did identify, with the exception of general functioning and adverse effects. With an estimated completion date of December 2019, this study could be included in updated versions of this review.

3. VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term)

See Table 3.

We identified only one outcome corresponding to our pre‐stated main outcomes for this comparison.

3.1 Physical fitness ‐ specific ‐ walking speed ‐ average time (seconds taken to walk 3 metres)

One small study reported relevant data for this outcome, demonstrating no clear difference in average walking speeds of the participants. We rated this evidence as of very low quality due to the lack of description of allocation and blinding and method of randomisation.

3.2 Missing outcomes

No studies reported data relating to general, social, or cognitive functioning or to adverse effects, mental state, or quality of life. More research is clearly needed in this area. We have not identified any currently ongoing studies that would address these gaps for this comparison. However, one study comparing different non‐exergames is ongoing (Valimaki 2017), which includes outcomes relating to general, cognitive, and social functioning, as well as to mental state.

Overall completeness and applicability of evidence

1. Completeness

The primary outcomes prespecified for this review were general and cognitive functioning, due to the potential cognitive benefits of video gaming (Granic 2014). This really applies to the comparisons where the focus was for a cognitive effect. However exergames, conversely, principally relate to physical fitness, hence cognitive benefits must be considered secondary. Typically, in such trials, only experimental groups are the focus of systematic reviews; changing this dynamic to measure the effect of a control group has been very interesting. The overall completeness of data for cognitive functioning was reasonable; however, there were very limited data for our third primary outcome of adverse effects. Considering the amount of literature condemning video games as related to increased violent behaviour, aggression, and depression (Calvert 2017; Tortolero 2014), good reporting of adverse effects was anticipated. Adverse effects are of importance in any study including people with schizophrenia. Cognitive interventions such as cognitive remediation are seeing increasing use, and are now the focus of several meta‐analyses investigating their effect (Kurtz 2001), and we would expect better reporting of presence or absence of adverse effects. Furthermore, economic outcomes were not reported at all. This is a major omission in this group of studies, particularly given the anticipated difference in real‐world price between commercial video games and cognitive remediation software.

The most disappointing aspect was the poor reporting of data in general. We excluded 32 studies, five of which could otherwise have been included but reported no useable data. Some authors have been most helpful (see Acknowledgements), but there are still data lost to this review ‐ and any other review that includes these studies. Many studies did measure our outcomes of interest, but did not report average scores or even report data across groups. Data were often presented in a way in which access to the outcome seems to favour access to the original cognitive training groups' numbers. Studies have found robust evidence that results of significance have a much higher chance of being reported compared to non‐significant results (Dwan 2008). We have an enduring concern that a bias operates in these data favouring more complete reporting of the cognitive groups' interventions. Furthermore, we had stipulated in the protocol that binary data were to be emphasised. This could be our bias operating, but we continue to argue that binary data eclipse continuous data in clinical usefulness. We are aware that some of the studies were in fact exploratory rather than clinical but, regardless, note that binary outcomes are easy to report, and that their omission is a problem, leaving more clinically orientated researchers using proxy measures and floundering for meaning.

2. Applicability

Most trials were conducted in the USA on participants that were clinically stable and being treated as outpatients. Also, studies tended to be carried out in hospital settings ‐ probably because the complex cognitive interventions are usually only available from centres of excellence. However, the majority of people with schizophrenia are treated in community settings ‐ and this is where people play video games. Should any benefit emerge in future studies of the non‐exer‐video games or the cognitive training, such studies should also be home‐based. Should there be any more exergame studies, it would seem important that these are undertaken in more real‐world settings.

Quality of the evidence

On the whole, the quality of the evidence was not good, and no trial was completely free of potential bias. None of the included trials specified their method of allocation concealment, and a number of trials had potential reporting bias. Blinding, however, was well reported by all studies. See Figure 3 for a visual representation of the quality of the included studies.

Potential biases in the review process

We are currently unable to detect any publication bias due to the small number of trials. Experiments with games could well be part of dissertations or theses that we have failed to identify. Two of the included studies were written in languages other than English (Mandarin and Spanish). There is always the possibility of searches incorporating a language bias, but we have no clear suggestion of this.

Another problem that could have arisen is that in all studies involving non‐exergame video games, these were the control groups for other better described interventions. It is possible that we failed to identify further studies because description of the control group (video game) was not been mentioned in the report.

Agreements and disagreements with other studies or reviews

We know of no other study systematically investigating the effect of commercial video games for people with schizophrenia.

Authors' conclusions

Implications for practice.

Given the low‐ and very low‐quality evidence from studies at high risk of bias for several ‘Risk of bias’ domains, it is difficult to draw any solid conclusions from the data.

1. For people with schizophrenia

For non‐exergames, we found no high‐quality evidence that the complex cognitive interventions examined in this review have any meaningful advantage over video games. A person with schizophrenia may prefer the thought of playing video games rather than receiving a cognitive intervention of equal benefit, although caution should be taken to moderate the quantity of time spent using such games.

If a person with schizophrenia wants to help control their weight or fitness, and likes video games, an exergame may be indicated given the ‐ admittedly weak ‐ data supporting aerobic fitness found in this review. Due to limitations with other forms of bodily movements, such as cost, time constraints, and disinterest, exergames games may provide an efficient alternative (Leutwyler 2012). However, more data are needed to be certain about this effect. Because more studies are needed, people with schizophrenia could help by volunteering to be part of an evaluative study in this area; we suggest that any such participation be done on the condition that all resulting data are made publicly available.

2. For clinicians

The data examined in this review do not establish that video games can provide substantial benefit for people with schizophrenia. There is low‐quality evidence suggesting that cognitive remediation may have an effect on cognitive functioning; however, there is insufficient high‐quality evidence to suggest that the cognitive remediation therapies examined in this review are any better than video games.

There is the suggestion of some improvement in physical fitness from exergames. This is important, as there are many therapeutic approaches for people with schizophrenia have shown no impact on this population. Exergames are experimental ‐ and more data are needed ‐ but limited evidence indicates that they may benefit the aerobic fitness level of people with schizophrenia.

3. For policymakers

This review suggests that non‐exergames are no more or less effective than cognitive remediation. One course of cognitive remediation (40 sessions) has been (conservatively) calculated to cost around GBP 632 (Patel 2010). This is considerably more expensive than the newest video games, plus their consoles. One role of a policymaker is to try to balance how much an intervention costs with the potential benefit of the intervention in people living with schizophrenia, thus such cost differentials must be taken into account.

All data in this review are limited, and we consider that no intervention examined herein is supported by data to encourage general implementation. However, should those funding research be considering evaluating the effects of video games, we would suggest supporting exergame research over standard video non‐exergames (see Implications for research).

Implications for research.

1. General

We excluded 41 trials that were fit for inclusion but fell short due to poor reporting (see Characteristics of excluded studies). With the exception of three trials, all included, excluded, and ongoing studies described in this review were published post‐1996, and therefore should all have adhered to the CONSORT Statement (CONSORT). Conforming to CONSORT would have improved reporting of clinical trials, increasing availability of data and easing interpretation for reviewers and consequently readers.

2. Specific

2.1 Reviews

In our protocol we specified that we would not include studies that involved participants at 'high risk of psychosis'. We excluded four different studies due to them being 'high risk' studies. We feel that 'high risk' should be its own, unique population and could be reviewed as a separate entity. This and other reviews suggested by the excluded studies are tabulated in Table 5. Furthermore, this review generated three different comparisons. The comparison with the most data is 'video games versus cognitive remediation', which again we feel could be reviewed separately given a more specific search is carried out. A separate review of cognitive remediation as a whole will be beneficial. This would also allow some conclusions to be drawn regarding whether the similar effect of video games to cognitive remediation, as seen in this review, stands to the credit of video games, or the detriment of cognitive remediation.

2.2 Further studies

Trials investigating the effects of video gaming that have been designed well and conducted and reported to a high quality are needed. Considering that there was no difference between video gaming and the complex cognitive interventions, we suggest that a trial where video gaming is the experimental group is warranted. Data were severely lacking in this area.

This review only included two studies with over 100 participants; if there was just one trial with 100 participants in each treatment arm reporting an unbiased result, this review would be of much greater use for people with schizophrenia, as well as for clinicians.

Future trials need to report any adverse effects, as this was very poorly reported, and of interest to policymakers, more economic data are pivotal. All outcomes in this review were reported in the short term (≤ 6 months), thus we propose that future trials investigate benefits over a longer period of time. We recognise that designing a good‐quality trial requires time, effort, and commitment.

Considering the poor quality of the trials included in this review, we have suggested our own outline for a trial in Table 6.

2. Reviews suggested by excluded trials.

	Participants	Intervention	Suggested title or link to existing review
Loewy 2013; Hooker 2019; Piskulic 2012; Cangas 2017; Gyllensten 2017; Koike 2018; Piskulic 2015	People with high risk of psychosis, not people with schizophrenia	Video game vs other treatment	Video games for people at high risk of psychosis
Bell 2015	People with schizophrenia	Posit Science vs Nintendo BrainAge	Cognitive training for people with schizophrenia
Bell 2018		Posit Science vs Nintendo BrainAge
Choi 2018		Cognitive remediation (computer assisted) with psychiatric rehabilitation vs psychiatric rehabilitation
Domen 2017		Cognitive training through playing AquaSnap video game vs treatment as usual
Fisher 2017		Targeted cognitive training vs social cognition training
Zhang 2017		Non‐auditory video games vs targeted cognitive training
Vinogradov 2016		Online auditory and social cognitive training vs free‐choice online cognitive exercises
Buchy 2018		Memory vs non‐memory training group
Vinogradov 2019		Online auditory and social cognitive training vs free‐choice online cognitive exercises
Zhang 2014		Psychological play therapy
Zhou 2014		Psychological games
Nieman 2015	OCD, schizophrenia/schizoaffective	Monster Valley cognition game vs TAU
Cavezian 2008; Demily 2009	People with schizophrenia	Chess playing vs TAU	Non‐video games for people with schizophrenia
Choi 2009		Motivational math game vs math game
Turvey 1985		Playing cards
Torres 2002		Train Game vs occupational therapy
Torres 2002		Train Game vs occupational therapy	Occupational therapy for people with schizophrenia
Kelly 1995	Nurse trainees	Trivia Psychotica game vs standard care, not video games	Bhoopathi 2006
Barnett 1978	Mentally disordered sex offenders	Group encounters vs hobby and game control	Group treatment for mentally disorder sex offenders; games for mentally disorder sex offenders

OCD: obsessive–compulsive disorder
TAU: treatment as usual

3. Suggested design of study.

Methods	Allocation: randomised ‐ method described. Allocation concealed ‐ method described. Blinding: none. Duration: 30 minutes, twice a week for 10 weeks with a 3‐, 6‐, and 9‐month follow‐up. Setting: community mental health setting.
Participants	Diagnosis: DSM‐IV Axis I diagnosis of schizophrenia or schizophrenia‐like illnesses. N = 300.* Age: 18 to 40. Sex: any.
Interventions	1. Exergame/non‐exergame** + treatment as usual. N = 150. 2. Treatment as usual. N = 150.
Outcomes	Functioning ‐ general: compliance with treatment, CGI ‐ clinically important changes. Functioning ‐ physical: SPPB. Functioning ‐ social: employed, relationship status. Adverse effects: relapse, hospitalisation, BPAQ, any recorded adverse effects. Mental state: PANSS. Quality of life: QOL. Economic outcomes: costs of intervention.
Notes	*For binary outcome, for 20% difference in groups, usually about 150 people in each group is needed to achieve adequate power. **We express no preference for which of the broad classes of games should be the focus of a new trial.

BPAQ: Buss‐Perry Aggression Questionnaire
CGI: Clinical Global Impression
DSM: Diagnostic and Statistical Manual of Mental Disorders
PANSS: Positive and Negative Syndrome Scale
QOL: Quality of Life Questionnaire
SPPB: Short Physical Performance Battery

What's new

Date	Event	Description
4 February 2021	Amended	One reference updated

History

Protocol first published: Issue 10, 2017
Review first published: Issue 2, 2021

Data and analyses

Comparison 1. VIDEO GAMES (COMMERCIAL NON‐EXERGAMES) (plus standard care) versus COGNITIVE REMEDIATION (plus standard care) (all short term).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Functioning: 1a. General ‐ average endpoint score (various scales, high = good) ‐ by end of treatment	4		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1.1 ILSS	1	43	Mean Difference (IV, Random, 95% CI)	0.03 [‐0.04, 0.09]
1.1.2 MARS	1	78	Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.42, 0.24]
1.1.3 Strauss Carpenter Outcome Scale	1	86	Mean Difference (IV, Random, 95% CI)	0.42 [‐0.62, 1.46]
1.1.4 UPSA ‐ total	1	53	Mean Difference (IV, Random, 95% CI)	1.72 [‐29.92, 33.36]
1.1.5 UPSA ‐ communication	1	78	Mean Difference (IV, Random, 95% CI)	‐0.95 [‐3.33, 1.43]
1.1.6 UPSA ‐ comprehension/planning	1	78	Mean Difference (IV, Random, 95% CI)	‐1.90 [‐3.97, 0.17]
1.1.7 UPSA ‐ financial management	1	78	Mean Difference (IV, Random, 95% CI)	0.35 [‐1.71, 2.41]
1.1.8 UPSA ‐ transportation	1	78	Mean Difference (IV, Random, 95% CI)	‐0.88 [‐3.05, 1.29]
1.2 Functioning: 1b. General ‐ average endpoint score (ILSS, high = good) ‐ 3‐month follow‐up	1	33	Mean Difference (IV, Fixed, 95% CI)	0.04 [‐0.03, 0.12]
1.3 Functioning: 2a. Specific ‐ cognitive ‐ improved (MCCB, at least one domain) ‐ by end of treatment	1	42	Risk Ratio (M‐H, Fixed, 95% CI)	0.58 [0.34, 0.99]
1.4 Functioning 2b. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite, high = good) ‐ by end of treatment	3	174	Mean Difference (IV, Fixed, 95% CI)	‐3.60 [‐6.43, ‐0.76]
1.5 Functioning 2c. Specific ‐ cognitive ‐ average endpoint score (MCCB ‐ composite, high = good) ‐ 3‐month follow‐up	1	33	Mean Difference (IV, Fixed, 95% CI)	‐2.69 [‐9.52, 4.14]
1.6 Functioning 2d. Specific ‐ cognitive ‐ average endpoint score (MATRICS ‐ z scores, high = good) ‐ by end of treatment (skewed data)	2		Other data	No numeric data
1.7 Functioning 2e. Specific ‐ cognitive ‐ average endpoint score (CAI, high = poor) ‐ by end of treatment	1	53	Mean Difference (IV, Fixed, 95% CI)	0.24 [‐0.25, 0.73]
1.8 Functioning: 3a. Specific ‐ social ‐ average endpoint score (Global Functioning: Role and Social Scale, high = good) ‐ by end of treatment	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.8.1 Global functioning role	1	86	Mean Difference (IV, Random, 95% CI)	‐0.08 [‐1.15, 0.99]
1.8.2 Global functioning social	1	86	Mean Difference (IV, Random, 95% CI)	0.16 [‐0.43, 0.75]
1.9 Functioning: 3b. Specific ‐ social ‐ average endpoint score (SLOF, high = good) ‐ by end of treatment	1	53	Mean Difference (IV, Random, 95% CI)	‐3.13 [‐40.17, 33.91]
1.10 Mental state: 1a. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ by end of treatment	4	269	Mean Difference (IV, Random, 95% CI)	0.20 [‐3.89, 4.28]
1.11 Mental state: 1b. Overall ‐ average endpoint score (PANSS total, high = poor) ‐ 3‐month follow‐up	1	33	Mean Difference (IV, Fixed, 95% CI)	‐5.58 [‐15.31, 4.15]
1.12 Mental state: 2a. Specific ‐ negative ‐ average endpoint score (PANSS negative, high = poor) ‐ by end of treatment	2	121	Mean Difference (IV, Random, 95% CI)	0.45 [‐1.63, 2.53]
1.13 Mental state: 2b. Specific ‐ positive ‐ average endpoint score (PANSS positive, high = poor) ‐ by end of treatment	2	121	Mean Difference (IV, Random, 95% CI)	‐0.34 [‐2.21, 1.54]
1.14 Global state: Average endpoint score (CGI ‐ severity, high = poor) ‐ by end of treatment (skewed data)	1		Other data	No numeric data
1.15 Leaving the study early: Did not leave	5		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.15.1 By end of treatment	5	395	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.87, 1.06]
1.15.2 Follow‐up (3 to 9 months)	1	56	Risk Ratio (M‐H, Fixed, 95% CI)	1.29 [0.83, 2.00]
1.16 Quality of life: 1a. Overall ‐ average endpoint score (QLS, high = good) ‐ by end of treatment	1	87	Mean Difference (IV, Random, 95% CI)	0.01 [‐0.40, 0.42]
1.17 Quality of life: 1b. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ by end of treatment	1	43	Mean Difference (IV, Random, 95% CI)	‐0.32 [‐3.34, 2.70]
1.18 Quality of life: 1c. Overall ‐ average endpoint score (EUROHIS‐QOL, high = good) ‐ 3‐month follow‐up	1	33	Mean Difference (IV, Fixed, 95% CI)	0.58 [‐2.69, 3.85]
1.19 Quality of life: 2a. Specific ‐ well‐being ‐ improved (IMI‐SR increase in score, high = good) ‐ by end of treatment	1	43	Risk Ratio (M‐H, Fixed, 95% CI)	1.50 [0.70, 3.20]
1.20 Quality of life: 2b. Specific ‐ well‐being ‐ average endpoint score (various scales, high = good) ‐ by end of treatment	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.20.1 IMI‐SR	2	96	Mean Difference (IV, Random, 95% CI)	5.16 [‐1.22, 11.54]
1.20.2 PCS	1	53	Mean Difference (IV, Random, 95% CI)	0.36 [‐2.62, 3.34]
1.20.3 Rosenberg Self‐Esteem Scale	1	53	Mean Difference (IV, Random, 95% CI)	‐0.69 [‐3.34, 1.96]
1.20.4 Revised Self‐Efficacy Scale	1	43	Mean Difference (IV, Random, 95% CI)	9.41 [‐4.85, 23.67]
1.21 Quality of life: 2c. Specific ‐ well‐being ‐ average endpoint score (Revised Self‐Efficacy Scale, high = good) ‐ 3‐month follow‐up	1	33	Mean Difference (IV, Fixed, 95% CI)	6.13 [‐7.17, 19.43]

Comparison 2. VIDEO GAMES (EXERGAMES) (plus standard care) versus STANDARD CARE (all short term).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Functioning: 1a. Specific ‐ cognitive ‐ average change score (MCCB, increase = good)	1	33	Mean Difference (IV, Fixed, 95% CI)	2.90 [‐1.27, 7.07]
2.2 Physical fitness: 2a. Specific ‐ average change (BDNF, increase = good)	1	33	Mean Difference (IV, Fixed, 95% CI)	‐0.13 [‐3.43, 3.17]
2.3 Physical fitness: 2b. Specific ‐ average change VO2 peak (AF, increase = good)	1	33	Mean Difference (IV, Fixed, 95% CI)	3.82 [1.75, 5.89]
2.4 Leaving the study early: Did not leave ‐ by end of treatment	1	33	Risk Ratio (M‐H, Fixed, 95% CI)	1.06 [0.75, 1.51]

Comparison 3. VIDEO GAMES (EXERGAMES) (plus standard care) versus NON‐EXERGAMES (plus standard care) (all short term).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Physical fitness: 1a. Specific ‐ walking speed ‐ average time (seconds taken to walk 3 metres, low = good)	1	28	Mean Difference (IV, Fixed, 95% CI)	‐0.50 [‐1.17, 0.17]
3.2 Physical fitness: 1b. Specific ‐ average endpoint score (SPPB, high = good)	1	28	Mean Difference (IV, Fixed, 95% CI)	‐0.30 [‐1.98, 1.38]
3.3 Behaviour: 2a. Specific ‐ average number of sessions attended (of 6 in total)	1	28	Mean Difference (IV, Fixed, 95% CI)	‐0.10 [‐0.48, 0.28]
3.4 Behaviour: 2b. Specific ‐ average time (minutes) at sessions (of 180 minutes in total)	1	28	Mean Difference (IV, Fixed, 95% CI)	‐2.20 [‐17.66, 13.26]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Aud CR: Fisher 2014.

Study characteristics
Methods	Allocation: randomised. Blinding: single (assessors).* Duration: 8 weeks. Setting: loaned laptops and participated at home, except 1 who participated in the lab. Design: parallel. Funding: The Stanley Medical Research Institute (06TAF‐972); the Laszlo Tauber Foundation; the National Institute of Mental Health (5R01MH081051); the San Francisco Department of Veterans Affairs Medical Center. Consent: > 18 years gave written informed consent, < 18 years provided assent with written parent/legal guardian consent.
Participants	Diagnosis: schizophrenia, schizophreniform or schizoaffective disorder. N = 121. Age: average ~21 years. Sex: 64 men, 22 women.** History: all achieved outpatient status for at least 3 months. 81 were taking psychiatric medication ‐ stable dose for > 1 month prior to participation; 5 did not take psychiatric medication. Ethnicity: not specified. Included: onset of first psychotic episode within past 5 years, good general physical health, age 14 to 30, fluent and proficient English, IQ ≥ 70. Exclude: neurological disorder or substance dependence.
Interventions	1. Video games: non‐exergames (16 different commercially available graphic‐based games), 40 h; play up to 4 to 5 games a day; 1 h/day, 5 days/week, 8 weeks plus standard care. N = 58. 2. Cognitive remediation: auditory training ‐ computerised exercises designed to improve speed and accuracy of auditory information processing whilst engaging auditory and verbal working memory; 1 h/day, 5 days/week, 8 weeks plus standard care. N = 63.
Outcomes	Functioning (general): Strauss Carpenter Outcome Scale. Functioning (specific): cognitive ‐ MATRICS score Functioning (specific): social ‐ Global Functioning: Role and Social Scales. Mental state: PANSS. Leaving the study early: participants excluded due to increased drug dose. Unable to use Functioning ‐ specific: cognitive: association between DNA variants in the catechol‐O‐methyltransferase (COMT) gene and the improvement in global cognition ‐ physiological measure, not outcomes prespecified in protocol.
Notes	*Report ‐ Biagianti 2017 ‐ states staff blinded, but no mention of participant blinding. **Report ‐ Fisher 2015 ‐ states that 121 participants were randomised, but 35 withdrew or were subsequently excluded. Demographic data are only presented for the 86 participants who were not excluded/withdrew.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "stratified by IQ, gender and symptoms severity and randomly assigned"
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Blinded: single (assessors). Quote: "All assessment staff were blinded to group assignment." Comment: no description of how this was achieved or how successful it was.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded: single (assessors).
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "We performed ... an intent‐to‐treat analysis using last observation carried forward on all randomized subjects" 19 auditory training group participants (of 63 total) were lost, 14 video game group participants (of 58 total) were lost.
Selective reporting (reporting bias)	Low risk	All outcomes reported.
Other bias	High risk	Quote: "Alternative forms of Hopkins Verbal Learning Test‐ Revised (HVLT‐R) and BVMT‐R were administered and counterbalanced at baseline and post training." Response: "Unsure the name or the validity of these 'alternative forms'" 2 researchers involved in this study are linked to the company that owns the cognitive training software used in this study (Posit Science). Bruno Biagianti is a postdoctoral research fellow partially funded by Posit Science. Sophia Vinogradov has worked as a paid consultant to the Brain Plasticity Institute, a division of Posit Science. Posit Science provided this software to the researchers free of charge. These conflicts of interest were declared.

Aud CR: Keefe 2012.

Study characteristics
Methods	Allocation: randomised. Blinding: single (assessors). Duration: up to 12 weeks. Setting: 9 sites from Schizophrenia Trials Network. Design: parallel. Funding: National Institute of Mental Health (grant number NO 1 MH90000 I). Consent: informed consent from all participants.
Participants	Diagnosis: schizophrenia (DSM‐IV). N = 53. Age: range 18 to 55 years, mean ~37 years (SD 10). Sex: 39 men, 14 women. Ethnicity: 30 "White", 18 "Black", 5 "Other". History: chronic schizophrenia. Included: PANSS score for hallucinatory behaviour, unusual thought content, and conceptual disorganisation ratings no greater than moderately severe (≤ 5); learnt English before age 12; able to complete MCCB at the baseline assessment; obtain raw score of 37 or greater on the Wide Range Achievement Test; minimum reading level of 6th grade; able to state specific goals relevant to intervention they want to achieve. Excluded: psychiatric hospitalisation within 8 weeks, current anticholinergic medication, DSM‐IV alcohol or substance abuse within last month or dependence within 6 months, intellectual/developmental disorders, neurological disorders.
Interventions	1. Video games: non‐exergames (10 computerised games described as "enjoyable"); 40 h within 12 weeks + 5 hours group sessions plus standard care. N = 26. 2. Cognitive remediation: auditory training: Brain Fitness developed by Posit Science, with "bridging groups" that met weekly; 40 h within 12 weeks + 5 hours group sessions plus standard care. N = 27.
Outcomes	Functioning (general): University of California San Diego Performance‐based Skills. Assessment (2nd edition). Functioning (specific): cognitive ‐ MCCB, CAI scores. Functioning (specific): social ‐ Specific Levels of Functioning Scale scores. Mental state: PANSS scores. Global state: CGI‐S, Rosenberg Self‐Esteem Scale, Intrinsic Motivation Inventory‐Schizophrenia Research scale, Perceived Competency Scale. Leaving the study early: numbers that did not leave early.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Each site randomly assigned 3 patients to cognitive remediation and 3 patients to the computer games control group" Response: method of randomisation not specified.
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Blinded: single (assessors). Quote: "cognitive testers were blinded to treatment assignment" See below for assessment of effectiveness of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "All cognitive testers completed a form asking them to estimate whether they were unblinded about patients' treatment condition. In 2 cases, 1 in each treatment condition, testers estimated 'definite' unblinding, and both reports were correct. In 15 additional cases, testers estimated 'possible' unblinding. However, less than 50% of those estimates were correct, which was less than chance. In 9 cases, testers were reasonably certain that a patient was receiving the cognitive remediation, but in only 4 of those cases was the tester correct. In 6 cases, the tester was reasonably certain that a patient was receiving the control intervention, but in only 3 of those cases was the tester correct."
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "Of the 6 patients that did not complete the study, 2 were not willing to make the time commitment and 4 withdrew for unspecified reasons"
Selective reporting (reporting bias)	Low risk	All outcomes stated were reported.
Other bias	High risk	Sophia Vinogradov has worked as a paid consultant to the Brain Plasticity Institute, a division of Posit Science, the company that owns the cognitive training software used in this study. These conflicts of interest were declared.

Aud CR: Vinogradov 2014.

Study characteristics
Methods	Allocation: randomised. Blinding: single (assessors).* Duration: 6‐month follow‐up.* Design: randomised controlled trial, parallel. Setting: San Francisco Department of Veterans Affairs Medical Center. Funding: supported by NIMH grant MH‐068725, NIMH Small Business Technology Transfer grant R42 MH‐073358, and the San Francisco VA Medical Center.*
Participants	Diagnosis: schizophrenia. N = 87** Age: auditory training mean = 40.7, video games mean = 43.2. Sex: auditory training (34 men, 12 women), video games (29 men, 12 women). History: clinically stable, chronically ill. Recruited from community mental health centres and outpatient clinics. Ethnicity: not specified. Excluded: enrolled in any psychiatric rehabilitation programme or prior cognitive remediation treatment.
Interventions	1. Video games: non‐exergames: 16 different graphics‐based, commercially available games (e.g. visuospatial puzzle games, clue‐gathering mystery games) playing 4 to 5 games on any given day for a total of 50 h (1 h/d, 1 d/week, 10 weeks) plus standard care. N = 41.*** 2. Cognitive remediation: auditory‐focused, Brain Fitness developed by Posit Science. 50 h (1 h/d, 1 d/week, 10 weeks) plus standard care. N = 46.
Outcomes	Functioning (cognitive): MATRICS recommended measures. Mental state: PANSS scores. Leaving the study early: numbers that did not leave the study early. Quality of life: QLS scores. Unable to use Mayer‐Salovey‐Caruso Emotional Intelligence Test scores ‐ data could not be included as it was measured for less than 50% of the study sample.
Notes	*Report ‐ Fisher 2009 ‐ states observer blinded, no mention of participants. **Many reports of this study use subsets of what seem to be a total of 87 people. ***Report ‐ Subramaniam 2012 ‐ states duration: 80 h over 16 weeks; report ‐ Fisher 2012 ‐ states duration: 100 h. ****Several outcomes could not be used due to lack of useable data, use of subscores, otherwise invalid.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The subjects were stratified by age, IQ, gender, and symptom severity and were randomly assigned to either the auditory training condition or a control condition of commercial computer games." Comment: how the list was generated is not specified, no indication that quasi‐random methods were used.
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "the assessment personnel were blind to the subjects’ group assignments" (Fisher 2009). Other reports state "double blind". Comment: conflicting accounts, possibly Blinded: Single.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "the assessment personnel were blind to the subjects’ group assignments" (Fisher 2009). Other reports state "double blind". Comment: conflicting accounts, possibly single‐blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "intent to treat analysis"; however, Fisher 2009: N = 59, N = 55 analysed; Dale 2016: "Missing VLM measures further reduced analysis to 15 AT and 15 CG participants"; Hooker 2012: "One participant in the CG group did not have pre MSCEIT data. This participant's post score was substituted for pre scores so the statistical power for the group analysis would not be lost."; Subramaniam 2014: "(1) two HC participants felt too claustrophobic to remain in the scanner." "(2) One SZ‐AT and 1 SZ‐CG were unavailable/unwilling to perform the fMRI N‐back task at the second time point and fMRI data from one HC was later excluded due to poor signal." Response: variable accounting for incomplete data.
Selective reporting (reporting bias)	High risk	Comment: PANSS and QOL scale reported at baseline, not after the treatment.
Other bias	High risk	Sophia Vinogradov has worked as a paid consultant to the Brain Plasticity Institute, a division of Posit Science, the company that owns the cognitive training software used in this study. Posit Science provided this software to the researchers free of charge. These conflicts of interest were declared.

General CR: Ahmed 2015.

Study characteristics
Methods	Allocation: randomised. Blinding: double.* Duration: 20 weeks. Setting: East Central Regional Hospital (ECRH), Augusta, Georgia, USA. Design: parallel.
Participants	Diagnosis: schizophrenia or schizoaffective disorder. N = 78. Age: ≥ 18 years. Sex: 68 men, 10 women. Ethnicity: 39 African‐American, 30 white, 5 Latino/Hispanic, 4 other. Included: schizophrenia, English speaking. Excluded: history of intellectual disability, neurological disorders, significant head trauma, or childhood antisocial behaviour.
Interventions	1. Video games: non‐exergames: 20 weeks, group, 3 x 60‐minute sessions/week (50‐minute video games + 10‐minute healthy behaviours discussion) plus standard care. N = 36. 2. Cognitive remediation: general CR: 20 weeks, group, delivered by masters‐ or doctoral‐level clinicians; 3 x 60‐minute sessions/week (50 min of computerised cognitive activities + 10‐minute "bridging" group discussion) plus standard care. N = 42.
Outcomes	Functioning (general): MARS, UPSA (baseline and post‐treatment) scores. Functioning (specific) cognitive: MCCB (baseline and post‐treatment) scores. Mental state: PANSS scores. Behaviour: aggression ‐ OAS scores. Unable to use Functioning ‐ general: WASI‐II scores, only taken at baseline. Behaviour: aggression ‐ OAS subscale scores only.
Notes	*Unclear how effective the method of referring to the computer games as 'cognitive remediation' would be in practice at blinding participants. No assessment of this completed during study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Assignment was completed using a Random Number Generator to create predetermined sequences."
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "Group leaders and study participants were blind to the active versus control of the study groups given that both the cognitive remediation and control groups were referred to as 'cognitive remediation'.” "Assessors were blind to participants' study assignment." Comment: unclear how effective the method of referring to the computer games as 'cognitive remediation' would be in practice at blinding participants. No assessment of this completed during study.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Assessors were blind to participants' study assignment."
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intent‐to‐treat analysis.
Selective reporting (reporting bias)	Low risk	All outcome measures reported.
Other bias	Low risk	No other bias detected.

General CR: Bryce 2018.

Study characteristics
Methods	Allocation: randomised. Blinding: single. Duration: 3 months. Setting: Victoria, Australia. Design: parallel.
Participants	Diagnosis: schizophrenia or schizoaffective disorder. N = 56 originally randomised, 43 completed intervention, 33 completed 3 months' follow‐up. Age: range 18 to 65 years; average 41 SD 10. Sex: 32 men, 17 women. History: duration ill average ~14 years SD 8; not admitted in last 8 weeks; not intellectual disability. Included: sufficient English skills. Exluded: psychiatric hospitalisations in the previous 2 months, intellectual/neurological impairment with cognitive sequelae, substance dependence, ECT in the previous 6 months. Consent: all participants gave written informed consent.
Interventions	1. Video games: non‐exergames ‐ commercially available arcade and puzzle video games: 20 x 1‐hour session twice/week, in groups of 2 to 5, over 10 weeks plus standard care. N = 24. 2. Cognitive remediation: general CR ‐ drill and strategy cognitive remediation: COGPACK software (Version 8.91, Marker Software): 20 x 1‐hour session twice/week, in groups of 2 to 5, over 10 weeks plus standard care. N = 25.
Outcomes	Functioning (specific): cognitive ‐ MCCB (improved); IMI‐SR, Revised Self‐Efficacy Scale, Independent Living Skills Survey scores. Mental state: PANSS scores. Global state (improved): IMI‐SR, ILSS, Revised Self‐Efficacy Scale scores. Leaving study early: numbers who did not leave early. Quality of life: EUROHIS‐QOL scores.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "participants were assigned to a group determined by a randomised sequence of CR or CG playing created using a random number generator"
Allocation concealment (selection bias)	Unclear risk	Comment: no details reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Blinded: single (assessors). Quote: "Assessors were blinded to group allocation"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded: single (assessors). Quote: "Assessors were blinded to group allocation"
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "56 people [....] were randomized; 49 analysed, and one more person did not provide final ratings and was dropped" Comment: completer analysis only, but for majority of participants.
Selective reporting (reporting bias)	High risk	Quote: "Three quarters (77%) of CR completers showed a reliable improvement" Comment: no details for control group, no data from some outcome scales.
Other bias	Low risk	Quote: "No conflict of interest"

Kimhy 2018.

Study characteristics
Methods	Allocation: randomised. Blinding: single (assessors). Duration: 12 weeks. Setting: greater New York City. Design: parallel. Funding: National Institute of Mental Health, Bethesda, MD (1R21MH096132 to DK). Consent: informed consent from all participants.
Participants	Diagnosis: people with schizophrenia (N = 26) or schizoaffective disorder (N = 7). N = 33. Age: range 18 to 55 years, average ~37 (SD 10). Sex: ~35% women, 65% men. Ethnicity: 72% Hispanic, 28% other. History: English‐speaking participants on stable dose of antipsychotic medication, fit to perform aerobic exercise, recruited from outpatient mental health clinics. Excluded: substance/alcohol abuse, cognitive sequelae from seizures/trauma, significant physical clinical abnormality/pathology, moderate disorganisation/depression, high suicide/homicide risk, previous involvement in neurocognitive studies in past 3 months.
Interventions	1. Video games: exergames ‐ aerobic exercise programme: active‐play video games including 2 active‐play video game systems (Xbox 360 Kinect, Microsoft) with whole‐body exercise software (Your Shape Fitness Evolved 2012, Ubisoft), 2 treadmill machines, a stationary bike, and elliptical machine; 60 minutes of aerobic exercise, 3/week for 12 weeks informed by American College of Sports Medicine and federal guidelines, in total 150 hours per week, plus standard care. N = 16. 2. Standard care: N = 17.
Outcomes	Functioning (specific) cognitive: MCCB scores Physical fitness (specific) aerobic: AF change score (indexed by VO2 peak mL/kg/min), serum‐BDNF change score. Leaving the study early: numbers who did not leave early. Unable to use Mental state: SAPS/SANS, BDI/BAI (no data reported).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Participants were randomized in the order they entered the study....participants to treatments based on an a prior computer‐generated randomization list"
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Blinded: single Quote: "The neurocognitive raters, as well as the technicians conducting the AF assessments and serum‐BDNF analyses were all blind to the participants’ treatment status and other collected data." See effectiveness of blinding below.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded: single Quote: "3 of the 33 randomized participants became un‐blinded to raters (9%) ‐ 1 in the TAU group (self‐disclosure) and 2 in the AE group (1 self‐disclosure, 1 accidental disclosure by clinical staff)"
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "In the AE group, all 3 non‐completers dropped‐out during the first week of the AE training after 0, 1, and 3 sessions, respectively due to loss of contact (n = 1), long commute‐time to AE site (n = 1), and not liking the AE program (n = 1). In the TAU group, 4 participants dropped‐out due to relocation (n = 1), hypomanic episode (n = 1), and protocol violations (n = 2)."
Selective reporting (reporting bias)	Low risk	Quote: "There were no adverse events associated with the AE training"
Other bias	Low risk	Quote: "None of the other authors had any conflict of interest relating to this project."

Leutwyler 2017.

Study characteristics
Methods	Allocation: randomised. Blinding: not specified. Duration: 6 weeks. Setting: not specified. Design: parallel.
Participants	Diagnosis: schizophrenia or schizoaffective disorder. N = 40 (28 included in analysis). Age: range 20 to 54 years; average 38 SD 9.5. Sex: 24 men, 4 women. History: sample of participants from 3 different facilities (an outpatient community treatment centre, a transitional residential facility, and a locked inpatient facility). Excluded: significant cardiovascular pathology.
Interventions	1. Video games: exergames ‐ using Kinect for Xbox 360 game system (Microsoft, Redmond, WA); each session began with bowling from Kinect Sports, and then participants were given option of 5 other games (Kinect Sports (baseball and skiing), Kinect Dance Central 2, Kinect Adventures, and Kinect Your Shape Fitness Evolved); 30 min, 1/week, groups 3 to 4, plus standard care. N = 13. 2. Video games: non‐exergames ‐ using sedentary video games (The Price is Right, Family Game Night, Wheel of Fortune, Sonic Hedgehog, Jeopardy) Xbox 360 system, plus standard care. N = 15.
Outcomes	Physical fitness (specific): walking speed ‐ SPPB (version 20). Behaviour (specific): sessions attended and total minutes attended.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "randomised controlled trial" Response: method of randomisation not specified, but no indication of use of quasi‐random methods.
Allocation concealment (selection bias)	Unclear risk	Not specified.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not specified.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not specified.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "Participants were excluded from the analyses if they did not complete the SPPB at enrolment and/or completion (n = 10) or if there was evidence of non‐compliance (n = 2)." Comment: explains why only 28 participants were included in analysis.
Selective reporting (reporting bias)	Low risk	All outcome measures reported.
Other bias	Low risk	Quote: "None of the authors have a potential conflict of interest."

AE: aerobic excercise
AT: auditory training
AF: aerobic fitness
BAI: Beck Anxiety Inventory
BAS: Barnes Akathisia Rating Scale
BDI: Beck Depression Inventory
BDNF: brain‐derived neurotrophic factor
BVMT‐R: Brief Visuospatial Memory Test‐ Revised
CACR: computer‐assisted cognitive remediation
CAI: Cognitive Assessment Interview
CG: computer games
CR: cognitive remediation
CGI‐S: Clinical Global Impression ‐ Severity
DSM: Diagnostic and Statistical Manual of Mental Disorders
ECT: electroconvulsive therapy
EUROHIS‐QOL: European Health Interview Survey ‐ Quality of Life
fMRI: functional magnetic resonance imaging
HC: healthy comparison participant
HVLT‐R: Hopkins Verbal Learning Test‐Revised
ILSS: Independent Living Skills Survey
IMI: Intrinsic Motivation Inventory
IMI‐SR: Intrinsic Motivation Inventory for Schizophrenia Research
IQ: Intelligence Quotient
MARS: Maryland Assessment of Recovery in Severe Mental Illness
MATRICS: Measurement and Treatment Research to Improve Cognition in Schizophrenia
MCCB: (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Consensus Cognitive Battery
MD: mean difference
MSCEIT: Mayer‐Salovey‐Caruso‐Emotional Intelligence Test
NAB: Neuropsychological Assessment Battery
OAS: Overt Aggression Scale
PANSS: Positive and Negative Symptom Scale
QLS: Quality of Life Scale
QOL: quality of life
SANS: Scale for the Assessment of Negative Symptoms
SAPS: Scale for the Assessment of Positive Symptoms
SAS: Simpson Angus Scale
SD: standard deviation
SPPB: Short Physical Performance Battery
TAU: treatment as usual
UPSA: University of California San Diego Performance‐based Skills Assessment
VLM: verbal learning and memory
VO₂ peak: peak oxygen uptake
WASI: Welchsler Abbreviated Scale of Intelligence

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Archie 2016	Allocation: randomised Participants: young people struggling with poverty, family conflict, homelessness, mental health issues, abuse and neglect, not people with schizophrenia.
Barnett 1978	Allocation: randomised. Participants: mentally disordered sex offenders; not people with schizophrenia. Intervention: group encounters vs hobby and game control; not video games.
Bell 2015	Allocation: randomised. Participants: people with schizophrenia. Intervention: compares 2 types of cognitive training ‐ Posit Science vs Nintendo BrainAge; not video games.
Bell 2018	Allocation: randomised. Participants: psychiatrically disabled participants in veterans affairs work service, 61% with schizophrenia and 39% other mental illness (depression, PTSD, bipolar). Intervention: compares 2 types of cognitive training ‐ Posit Science vs Nintendo BrainAge; not video games.
Boonstra 2013	Allocation: randomised. Participants: people with first‐episode psychosis. Intervention: serious game not designed to entertain (BiasBlaster); exclude serious games. Outcomes: trial terminated early with no reporting of outcomes.
Buchy 2018	Allocation: randomised. Participants: people with psychotic disorders. Intervention: memory vs non‐memory training group; not video games.
Cangas 2017	Allocation: randomised. Participants: students; not people with schizophrenia.
Cavezian 2008	Allocation: randomised. Participants: schizophrenia inpatients. Intervention: chess playing vs treatment as usual; not video games.
Choi 2009	Allocation: randomised. Participants: people with schizophrenia or schizoaffective disorder. Intervention: motivational maths game vs maths game; not interactive, not video game.
Choi 2018	Allocation: randomised. Participants: people with chronic schizophrenia. Intervention: cognitive remediation (computer assisted) with psychiatric rehabilitation vs psychiatric rehabilitation; not video games.
Choi 2018a	Allocation: randomised. Participants: young people at risk of developing schizophrenia, not people with schizophrenia.
Demily 2009	Allocation: randomised. Participants: people with schizophrenia. Intervention: chess playing vs treatment as usual; not video games.
Domen 2017	Allocation: randomised. Participants: OCD, schizophrenia, major depressive disorder. Intervention: cognitive training through playing AquaSnap video game vs treatment as usual; not commercial video game.
Fisher 2017	Allocation: randomised. Participants: people with schizophrenia, schizoaffective disorder, or psychosis not otherwise specified. Intervention: targeted cognitive training vs social cognition training, not video games.
Gyllensten 2017	Allocation: randomised clustered. Participants: people with schizophrenia (n = 14), neuropsychiatric disorder (n = 14), other psychosis (n = 4), bipolar disorder (n = 4), other diagnosis (n = 9), do not know (n = 28); less than 20% of participants had schizophrenia.
Han 2008	Allocation: not randomised, 32 video games vs 49 watching movies ‐ no randomisation mentioned.
Heggelund 2011	Allocation: not randomised, included consecutive patients first to high‐aerobic intensity training group and then to video games group; no randomisation mentioned.
Holzer 2014	Allocation: randomised. Participants: people with schizophrenia and people with high risk of psychosis; 37.5% of participants were at high risk of psychosis. Data not provided separately for people with schizophrenia.
Hooker 2019	Allocation: randomised. Participants: people with high risk of psychosis, not people with schizophrenia.
Iizuka 2014	Allocation: randomised. Participants: people with schizophrenia. Intervention: video games vs visuo‐spatial puzzle games. Outcomes: we have only identified an abstract of study protocol, no useable data.
Kantrowitz 2016	Allocation: randomised. Participants: people with schizophrenia, stabilised before start of trial on standard regimen of luratidine. Intervention: auditory‐focused cognitive remediation vs video games. Outcomes: data not reported separately for control and intervention groups, no useable data.
Kelly 1995	Allocation: randomised. Participants: nurse trainees. Intervention: Trivia Psychotica game vs standard care, not video games.
Koike 2018	Allocation: randomised. Participants: undergraduate and graduate students, not people with schizophrenia.
Loewy 2013	Allocation: randomised. Participants: people with high risk of psychosis, not people with schizophrenia.
Lopez 2016	Allocation: randomised. Participants: adults with schizophrenia. Intervention: virtual reality game vs standard care; exclude virtual reality games.
Mathalon 2013	Allocation: randomised. Participants: recent‐onset schizophrenia. Intervention: computerised cognitive training of auditory/verbal processing vs video games (8 weeks, 40 hours). Outcomes: fMRI, MATRICS‐recommended measures; no useable data.
Nieman 2015	Allocation: randomised. Participants: people with OCD and schizophrenia/schizoaffective disorder. Intervention: Monster Valley cognition game vs TAU; not commercial video games.
Olivet 2018	Allocation: randomised. Participants: youth and young adults with first‐episode psychosis. Intervention: prototype serious game (OnTrack), not commercially available video game.
Park 2017	Allocation: randomised. Participants: people with schizophrenia/schizoaffective individuals. Intervention: novel adaptive virtual reality low‐dose vs high‐dose virtual reality; exclude virtual reality games.
Piskulic 2012	Allocation: randomised. Participants: young people at high risk of psychosis; not people with schizophrenia.
Piskulic 2015	Allocation: randomised pilot trial. Participants: people at risk of schizophrenia; not people with schizophrenia.
Roberts 2015	Allocation: randomised. Participants: people with schizophrenia. Intervention: de‐biasing training vs video games control. Outcomes: measures of social cognitive capacity and bias ‐ no useable data.
Saleem 2013	Allocation: not randomised, cognitive training (n = 5) vs TAU (n = 6); no randomisation mentioned.
Sikich 2013	Allocation: randomised. Participants: young people with psychosis. Interventions: cognitive games vs control games. Outcomes: no useable data.
Torres 2002	Allocation: randomised. Participants: people with schizophrenia. Intervention: Train Game vs occupational therapy; not video games.
Turvey 1985	Allocation: randomised. Participants: people with schizophrenia. Intervention: playing cards; not video games.
Vinogradov 2019	Allocation: randomised. Participants: people with schizophrenia. Intervention: 40 hours online auditory and social cognitive training vs free‐choice online cognitive exercises; not video games.
Wu 2016	Allocation: allocated to intervention or control by ward, not fully randomised.
Zhang 2014	Allocation: randomised. Participants: people with schizophrenia. Intervention: psychological play therapy; not video games.
Zhang 2017	Allocation: randomised. Participants: people with schizophrenia. Intervention: non‐auditory video games (n = 11) vs targeted cognitive training (n = 18). Outcome: ASSR ‐ no useable data.
Zhou 2014	Allocation: randomised. Participants: people with schizophrenia. Intervention: psychological games; not video games.

ASSR: auditory steady‐state response
fMRI: functional magnetic resonance imaging
MATRICS: Measurement and Treatment Research to Improve Cognition in Schizophrenia
OCD: obsessive‐compulsive disorder
PTSD: post‐traumatic stress disorder
TAU: treatment as usual

Characteristics of ongoing studies [ordered by study ID]

Engh 2015.

Study name	Effects of high‐intensity aerobic exercise on psychotic symptoms and neuro cognition in outpatients with schizophrenia.
Methods	Allocation: randomised. Blinding: observer‐blinded. Duration: 12 weeks. Setting: Buskerud and Vestfold University College; video gaming ‐ local high school in Tønsberg. Design: randomised controlled trial, parallel. Consent: informed and written consent is a required for participation. Funding: the trial has received funding from Vestfold Hospital Trust, Norwegian Extra Foundation for Health and Rehabilitation through EXTRA funds, Norwegian Research network in Severe Mental Illness (NORSMI) NORMENT/KG Jebsen Centre for Psychosis Research, Torgeir Lindvik’s Trust, and Civitan International.
Participants	Diagnosis: schizophrenia spectrum disorder (DSM‐V). N = 126 (63 aerobic HIIT training; 63 video games skills). Age: will include ages between 18 and 67 years. Sex: not specified. Ethnicity: not specified. Include: understand and speak a Scandinavian language. Exclude: pregnancy, chest pain during exercise, unstable angina pectoris, recent myocardial infarction, uncontrollable cardiac arrhythmia, severe hypertension (> 180/110 mmHg), comorbid diagnosis of mild mental retardation, medical conditions incompatible with participation.
Interventions	1. Video games: computer simulated sports activities (Nintendo Wii Sports) N = 63. 2. Exercise group: HIIT training. Walking/ running twice a week for 12 weeks. 8‐minute warm‐up, 4 x 4 minutes 85% to 95% max HR, active pauses of walking/running at 70% max HR.
Outcomes	Mental state: PANSS, PSYRATS, BAVQ‐R, CDSS, PANAS, GAF. Global state: CGI, WHO‐5. Functioning ‐ cognitive: Emotional Biological Motion Test, BCIS. Physical fitness ‐ VO2 max, ActiGraph accelerometer, IPAQ. Behaviour: AUDIT, EuropASI, AUS, DUS. Functioning ‐ social: SCLOF. Leaving the study early: 5 left already.
Starting date	First received by ClinicalTrials.gov 29 July 2014. Estimated completion: December 2019
Contact information	Principal Investigator: John A Engh, MD, PhD, Division of Mental Health and Addiction, Vestfold Hospital Trust, Tønsberg, Norway.
Notes

Javitt 2016.

Study name	Remediation of auditory recognition in schizophrenia with tDCS.
Methods	Allocation: randomised. Blinding: quadruple (participant, care provider, investigator, outcomes assessor). Duration: 30 minutes, 2 to 3 sessions per week. Setting: New York State Psychiatric Institute. Design: randomised controlled trial, parallel. Funding: not specified.
Participants	Diagnosis: schizophrenia. Estimated N = 45. Sex: all. Age: 18 to 60. Ethnicity: not specified. History: not specified Included: English fluency, willing/capable to provide informed consent, auditory tone matching defects, stable doses of antipsychotic medication for at least 2 weeks. Excluded: serious neurological or medical condition known to affect the brain, current or past history (6 months) of substance abuse, pregnancy or breastfeeding, taking anticholinergic medication.
Interventions	1. Video games with active tDCS. 2. Auditory remediation with active tDCS. 3. Auditory remediation with sham tDCS.
Outcomes	Functioning ‐ cognitive: auditory tone matching task; auditory emotion recognition, Penn Emotion Recognition Test, Awareness of Social Inference Test, auditory mismatch negativity.
Starting date	Start: June 2016. Estimated completion: December 2018.
Contact information	Daniel Javitt, MD, PhD. Jaimie Gowatsky, MA.
Notes

Kuehn 2018.

Study name	Augmentation of neuronal network plasticity in schizophrenia.
Methods	Allocation: randomised. Duration: within 2 months. Setting: University Clinic Hamburg‐Eppendor. Funding: Universitätsklinikum Hamburg‐Eppendorf.
Participants	Diagnosis: schizophrenia. N = 150. Age: 18 to 45 years. Include: clinically stabilised. Exclude: clinically relevant anaemia, MRI contraindication, earlier electroconvulsive shock treatment, significant somatic or neurological disease, significant alcohol or substance abuse in the previous year, more than 1 h video games per day for more than 6 months before study start, simultaneous major psychiatric disease (if symptomatic in the foreground).
Interventions	1. Experimental group: participants intensively train with a 3D navigation video game (Super Mario DS). 2. Active comparator: participants train with a video game (Super Mario Bros) intervention without 3D navigation. 3. Control group: participants read on a Kindle device, no 3D navigation.
Outcomes	Mental state: PANSS. Functioning ‐ cognitive: MRI, MCCB, fMRI, DTI, BDNF.
Starting date	First received by: key record dates: May 2018. Estimated completion: December 2019.
Contact information	Principal Investigator: Maxi Becker, MSc (max.becker@uke.de), Simone Kuehn (Prof.s.kuehn@uke.de)
Notes

Ma 2018.

Study name	An intervention of Internet‐based cognitive training programs for mental disorders.
Methods	Allocation: randomised. Blinding: not specified. Duration: not specified. Setting: Beijing Anding Hospital, Capital Medical University, China. Design: randomised parallel controlled trial.
Participants	Diagnosis: those eligible for inclusion had an ICD‐10 diagnosis of schizophrenia, depressive disorder, or generalised anxiety disorder. Estimated N = 180. Age: 18 to 65 years. Sex: all. Ethnicity: not specified. Include: participants were treated with stable doses of medication throughout the interval between assessments, participants provided written informed consent. Exclude: combine other mental disorders, receipt of electroconvulsive therapy within 12 months prior to assessment, drug or alcohol abuse or dependence within the past 6 months, history of head injury resulting in loss of consciousness, overt sensorimotor disturbances lead to that computer cannot be used, history of neurological disease, identified learning disability, or diagnosis of attention deficit hyperactivity disorder.
Interventions	1. Games (N = 90). 2. Internet‐based cognitive training (N = 90).
Outcomes	Functioning ‐ cognitive: MCCB. Mental state: PANSS, Hamiltion Depression Scale, Hamiltion Anxiety Scale
Starting date	31 October 2017
Contact information	Study leader: Xin Ma; maxinanding@vip.163.com Applicant: Dan Wang; mydanfer@126.com
Notes

Rose 2015.

Study name	Randomized controlled trial of computer‐based treatment of social cognition in schizophrenia.
Methods	Allocation: randomised. Blinding: double. Duration: 12 to 14 weeks (40 sessions). Setting: home or in clinic; multisite study. Design: randomised controlled trial, parallel. Funding: Phase II SBIR Grant (R44MH091793) from the National Institute of Mental Health.
Participants	Diagnosis: schizophrenia (DSM‐V). N = 154 (target sample). Age: must be between 18 to 65 at the time of study screening. Sex: open to all genders. History: clinically stable and be stable on the doses of the psychiatric medications they are taking. Ethnicity: open to all races/ethnicity. Included: 18 to 65 years old, DSM‐IV diagnosis of schizophrenia. Adequate decisional capacity, in the judgement of the consenting study staff member, to make a choice about participating in this research study. Clinically stable (non‐acute) for 8 weeks prior to consent; in the judgement of the Site Principal Investigator. Stable treatment of antipsychotics and/or other psychotropic treatment for at least 6 weeks prior to consent. Learned English before the age of 12. Must have the visual, auditory, and motor capacity to use the computerised intervention in the judgement of the consenting study staff person. No more than a moderate severity rating on hallucinations and unusual thought content as shown by a score of ≤ 4 on PANSS. Exclude: psychiatric hospitalisation 8 weeks prior, intoxicated on any day of the experiment, mental retardation, IQ < 70, pervasive developmental disorder or other neurological disorder, cognitive training programme 5 years previous, concurrent clinical trials, more than 2 antipsychotics, suicidal behaviour from Columbia‐Suicide Severity Rating Scale.
Interventions	1. Video games: 13 conventional, progressive video games. 2. Experimental treatment programme (SocialVille): computerised cognitive remediation programme.
Outcomes	Functioning ‐ cognitive: MATRICS‐recommended measures. Functioning ‐ social: ER40, PROID, PMFT, UPSA‐2, SFS. Global state: MSCEIT, EA test. Mental state: PANSS. Functioning ‐ general: GFS, SLOF, VRFCAT. Quality of life: QLS.
Starting date	Start: March 2015. Estimated completion: August 2018.
Contact information	Dr Dana Frostig; dana.frostig@positscience.com Multicentre, USA.
Notes

Valimaki 2017.

Study name	The impact video games among people with schizophrenia.
Methods	Allocation: randomised. Blinding: single. Duration: 60 minutes of gaming 5 days a week, for 12 weeks (minimum of 50 hours of gaming must be completed). Setting: outpatient clinics. Design: randomised controlled trial, 3‐arm parallel‐group design. Funding: Hong Kong Polytechnic University.
Participants	Diagnosis: schizophrenia (DSM‐V). N = 144: 2 intervention groups, N = 48 each, and 1 active control group (N = 48). Age: must be between 18 and 44. Sex: open to all genders. Included: able to speak Cantonese, be unfamiliar with video games or at least non‐active game players (play < 5 h/week), able to provide written informed consent, be viewed as being able to safely take part and have the cognitive status deemed suitable for participation (assessed by a chief psychiatrist based on his/her clinical expertise). Exclude: meeting diagnostic criteria for a current major depressive, manic, or hypomanic episode (DSM‐IV), or mental retardation, having severe visual impairment, being an active game player (i.e. gaming > 5 h/week), displaying a lack of ability to decide their own participation, displaying substance abuse (other than nicotine dependence), having head injury, hemiplegia, or other neurological disorders, having had electroconvulsive therapy in the past 6 months, having a lack of MRI compatibility (e.g. patients with cardiac pacemakers, metallic implants, restless behaviour).
Interventions	1. Overwatch (Blizzard Entertainment) released for Windows, XBox One and PlayStation 4, it is a team‐based/multiplayer first‐person shooter game. 2. Project: Neural: multiplayer game, specifically designed to improve ecologically valid working memory functioning. Combines the task‐oriented immersion of first‐person action games with the targetable high‐level cognitive demand. 3. SimCity (control): non‐competitive game.
Outcomes	Functioning ‐ cognitive: WMS III (up to 6‐month follow‐up), SERT (up to 6‐month follow‐up), Wisconsin Card Sorting Task (up to 6‐month follow‐up), fMRI Functioning ‐ social: BSPS51. Mental state: TEPS, CAINS, ACP, CDS‐C, SAS, BARS, AIMS. Functioning ‐ general: GSE.
Starting date	Start: 1 May 2017. Completion date: 31 December 2018. Estimated primary completion date: 31 December 2018 (final data collection date for primary outcome measure).
Contact information	Maritta Välimäki, Professor 2766 6409; maritta.valimaki@polyu.edu.hk
Notes	As of 1 January 2000, study has recruited 34 participants.

Vinogradov 2016.

Study name	Cognitive remediation of schizophrenia in a community mental health setting (SECT)
Methods	Allocation: randomised. Blinding: single‐blind (participant). Duration: 80 hours. Setting: San Francisco, California, USA. Design: single‐group assignment.
Participants	Diagnosis: schizophrenia. Estimated N = 120. Age: 18 to 65 years. Sex: all. Ethnicity: not specified. Include: English primary language (learned before age 12), no major medical or neurological disorder. Exclude: substance abuse or dependence, those who miss 5 days due to intoxication will be dropped and replaced.
Interventions	1. Video games plus support employment ‐ 80 hours. 2. Targeted cognitive training plus support employment ‐ 80 hours.
Outcomes	Functioning ‐ cognitive: MCCB.
Starting date	Start: September 2009. Estimated completion: June 2016.
Contact information	Sophia Vinogradov, MD
Notes

ACP: Association of Child Psychology
AIMS: Abnormal Involuntary Movement Scale
AUDIT: Alcohol and Drug Use Disorder Identification Tests
AUS: Alcohol Use Scales
BARS: Behavioral Activity Rating Scale
BAVQ‐R: revised Beliefs About Voices Questionnaire
BCIS: Beck Cognitive Insight Scale
BDI: Beck Depression Inventory
BDNF: brain‐derived neurotrophic factor
CAINS: Clinical Assessment Interview for Negative Symptoms
CDSS: Calgary Depression Scale for Schizophrenia
CDS‐C: Calgary Depression Scale for Schizophrenia
CGI: Clinical Global Impression Scale
DSM: Diagnostic and Statistical Manual of Mental Disorders
DTI: diffusion tensor imaging
DUS: Drug Use Scales
EA: empathic accuracy
ER40: The Penn Emotion Recognition Test
EuropASI: European Addiction Severity Index
fMRI: functional magnetic resonance imaging
GAF: Global Assessment of Functioning
GEQ: Gaming Experience Questionnaire
GFS: Global Functioning Scale
GSE: General Self‐Efficacy Scale
HIIT: high‐intensity interval training
HHI: Herth Hope Index
HR: heart rate
ICD: International Classification of Diseases
IPAQ: International Physical Activity Questionnaire
ISMIS: Internalised Stigma of Mental Illness Scale
IQ: Intelligence Quotient
LSAS: Liebowitz Social Anxiety Scale
MATRICS: Measurement and Treatment Research to Improve Cognition in Schizophrenia
MCCB: (Measurement and Treatment Research to Improve Cognition in Schizophrenia) Consensus Cognitive Battery
MRI: magnetic resonance imaging
MSCEIT: Mayer‐Salovey‐Caruso Emotional Intelligence Test
PANAS: Positive and Negative Affect Schedule
PANSS: Positive and Negative Syndrome Scale
PMFT: Penn Faces Memory Test
PROID: prosody identification
PSYRATS: Psychotic Symptom Rating Scales
QLS: Quality of Life Scale
RAQ: Recovery Attitude Questionnaire
SAS: Simpson Angus Scale
SERT: serotonin transporter
SOLES: Singh O'Brien Level of Engagement Scale
SCLOF: Strauss Carpenter Level of Functioning Scale
SFS: Social Functioning Scale
SLOF: Specific Levels of Functioning
tDCS: Transcranial Direct Current Stimulation
TEPS: Temporal Experience of Pleasure Scale
UPSA‐2: University of California San Diego Performance‐based Skills Assessment
VRFCAT: Virtual Reality Functional Capacity Assessment Tool
VO₂ peak: peak oxygen uptake
WHO‐5: World Health Organization‐5 Well Being Index
WMS: Wechler Memory Scale

Differences between protocol and review

In the original Methods it was stated that only adult participants would be included in the review. One of the studies now included in the review, Aud CR: Fisher 2014, specified some adolescents in their inclusion criteria. We felt that due to the onset of schizophrenia being in late adolescence (Gogtay 2011), and because most of included participants were older than that, these studies had to be included in this review. However, there is an argument for a further review focusing on young people alone (see Implications for research).

Due to the significant proportion of data reported as subscores we decided to extract these data in order to provide a more comprehensive summary of the literature on this topic. Excluding all of these data would unfortunately have significantly limited the scope of this review. We continue to feel that these subscores are often not validated in themselves. There is a very low bar for inclusion of scale‐derived data (that the scale has been published in its own right). However, if the subscore has not been specifically validated, we feel that it is problematic to focus on, may well be biased (Marshall 2000), and leads to multiple statistical testing within the review and subsequent chance of false results.

In our protocol we stated we would compare (i) video games with (ii) standard care alone and (iii) other psychological treatments. We later felt that the category 'other psychological treatments' was unhelpfully vague, and that more specific categorisation would allow more useful comparisons to be made. In addressing this we created the new comparison group of cognitive remediation. As ultimately all of the studies we found that were initially grouped into the 'other psychological treatments' comparison were in fact studies into cognitive remediation, this change did not affect the numerical data of the comparisons in any way. Had we found any studies that did not involve cognitive remediation, but did involve other psychological therapies, we would have included them in an 'other psychological treatments' comparison separate to the cognitive remediation comparison.

In our protocol we did not initially state that we would include studies comparing video games against other video games (i.e. exergames versus non‐exergames), although we also did not state that we would exclude such studies. We added a sentence to the final review to clarify that we would include such studies, and the data for these comparisons are presented in this review.

In our protocol we described how we intended to manage skewed data as, at some point, these are ill‐advised to present in analyses (Data extraction and management). We have complied with this for Analysis 1.6 and Analysis 1.14. For Analysis 1.9, the skew was moderate and only applied to one of the two groups, so we have left these data in graphical form as, in any event, we noted them to be of very limited quality. Finally, for Analysis 1.10, data from Aud CR: Keefe 2012 were highly skewed, but we did leave them in as a sensitivity analysis showed that their inclusion made little difference to the overall result. Also, in Effects of interventions section 1.10, we have alerted the reader to the unusual data included in this outcome overall.

Subgroup analysis was not conducted as there was insufficient power to carry out such an analysis. Sensitivity analyses were to be undertaken on primary outcomes which, in the event, were not reported. In most cases, even if primary outcomes had been reported, however, such analyses were not warranted or indicated. In the two post hoc sensitivity analysis we did undertake (removing the clearly lower quality studies and using random effects models ‐ both for non‐primary outcome Analysis 1.4) only use of the random effects made a difference in the breadth of the confidence intervals and this is discussed in the text. We have tried to highlight that this analysis is a break from the protocol and not to over emphasise its findings.

We do not think these changes affect the validity of the review.

Contributions of authors

Matthew T Roberts: carried out the review; assessed updated list of trials for inclusion and reviewed all existing included trials; edited, corrected, and added to all previous work of selection, extraction, interpretation, and writing.

Jack Lloyd: carried out the review, formulated the protocol, searching, trial selection, data extraction and interpretation, report writing.

Maritta Välimäki: helped with the review, independently assessed the trials for inclusion.

Grace Ho: helped with the review, independently assessed the trials for inclusion.

Megan Freemantle: carried out the review, formulated the protocol, searching, trial selection, data extraction and interpretation, report writing.

Anna Zsófia Békefi: carried out the review, formulated the protocol, searching, trial selection, data extraction and interpretation, report writing.

Sources of support

Internal sources

• Cochrane Collaboration Schizophrenia Group, UK

  Employs review author Matthew Roberts.

• Nottingham University Hospitals NHS Trust, UK

  Employs review author Matthew Roberts.

• Royal Derby Hospital Medical School, UK

  Review author Jack Lloyd is a student at this university.

• Department of Nursing Science, University of Turku, Finland

  Employs review author Maritta Välimäki.

• Xiangya Nursing School, Central South University, China, Other

  Employs review author Professor Maritta Välimäki

• School of Nursing, The Hong Kong Polytechnic University, China

  Employs review authors Grace Ho

External sources

• No sources of support supplied

Declarations of interest

MTR: none known.

JL: none known.

MV: has an interest in the topic of gaming for people with schizophrenia. The Academy of Finland and Turku University Hospital has also offered a grant to conduct a series of systematic reviews related to people with serious mental disorders.

GH: none known.

MF: none known.

AZB: none known.

Edited (no change to conclusions)