An individual's mental health is fundamental to living a healthy and enjoyable lifestyle. Studies estimate that 1 in 3 people may suffer from a mental illness during their lifetime [1]. The World Health Organization (WHO) reports that depression is a “leading cause of disability worldwide and is a major contributor to the overall global burden of disease” affecting more than 264 million people of all ages globally [2]. Depression is a mental health disorder that the Sustainable Development Goals of the United Nations (2015) has listed among its 270 targets and 230 indicators. Depressive disorders account for most of the total disability-adjusted life years globally. Although depressive disorders are global, they particularly affect those living in high and upper-middle-income nations [3]. This heavy toll is exacerbated by the fact that up to one-half (50%) of the people living in high-income countries and 90% of those living in low-resource settings receive no treatment for depressive disorders [4].

Depressive disorders are a family of mental disorders ranging in severity from mild temporary episodes of sadness to more severe and persistent depression [5]. Depressive disorders include disruptive mood dysregulation disorder, premenstrual dysphoric disorder, substance- or medication-induced depressive disorder, and major depressive disorder [6]. Depressive disorders are characterized by disturbances in mood, thoughts, or behaviors. Furthermore, depressive disorders do not affect the mind alone but are reported to impact a person’s body [7,8]. Depression is known to be caused by a number of factors that interact in complex mechanisms: social, genetic, pathological, and chemical [9]. Treatments for depressive disorders are generally classified into either pharmacological or psychosocial (ie, nonpharmacological) interventions. Pharmacological treatments involve the use of drugs (eg, antidepressants) while examples of psychosocial treatments include cognitive behavioral therapy (CBT), exposure therapy, and exercise [10].

The use of serious games, defined as games that have a purpose other than entertainment, has seen a rise in recent years [11]. Serious games use elements unique to gaming in order to educate or influence change in experience or behaviors [12]. Several industries have adopted and continue to use serious games including health care, education, and airlines [13]. Among other things, serious games have been effectively utilized for education, prevention, and treatment of chronic conditions (eg, asthma and diabetes) [14], therapeutic rehabilitation [15], and educational resources for health care professionals [16]. Moreover, serious games have been used as a nonpharmacological therapeutic intervention for mental disorders [17]. Serious games have been utilized as a treatment for various mental disorders, including depression [18,19], anxiety [20], posttraumatic stress disorder [21,22], autism spectrum disorder [23,24], dementia [25,26], alcohol use disorder [27], attention deficit hyperactivity disorder [28], and obsessive-compulsive disorder [13,29].

Gaming as a therapeutic tool in mental health can potentially offer several specific advantages that may be missing from traditional forms of delivery. The gaming industry is, as ever, popular globally [30] and arguably easier to access than even basic mental health services [31]. Games by their very nature have the potential to engage the user in game play that can be rewarding through scoring points or following story arcs that can help improve user involvement and lower attrition rates [32,33]. Additionally, as the technology improves, gaming can utilize accessories to provide richer sensory environments and immersive user experiences that allow users to simulate real-life scenarios more safely and help in educating and achieving cognitive and behavioral changes through overlearning and repetition [33,34].

Many studies have assessed the effectiveness of serious games to alleviate depression. Aggregating the evidence from these studies is very important to draw more definitive conclusions about the effectiveness of serious games as viable therapeutic interventions in depressive disorders. Several published reviews have summarized the evidence about the effectiveness of serious games for depression [18,19,35-37]. However, these reviews are undermined by certain technical shortcomings that limit the generalization of the findings. Specifically, these reviews (1) focused on a certain type of serious games (ie, exergames) [19,37]; (2) focused on a certain age group (older adults) [37]; (3) included non-randomized controlled trials (RCTs) [19,35]; (4) did not search technical databases (such IEEE Xplore and the ACM Digital Library), thereby including only a few studies [35-37]; (5) did not assess the quality of the evidence [18,19,35-37]; and (6) were outdated publications [18,19,35]. Therefore, this study aimed to assess the effectiveness of serious games for alleviating depression by summarizing and pooling the results of previous studies and providing an up-to-date review.

This review included only RCTs that assessed the effectiveness of serious games for alleviating the severity of depressive symptoms. To be more precise, the intervention of interest in this review was serious games that were delivered on any digital platform such as computers, consoles (eg, Xbox, PlayStation), mobile phones, tablets, handheld devices, or any other computerized devices. The intervention had to utilize elements of gaming as an integral and primary method for therapeutic or prevention purposes. We did not consider nondigital games and those used for other purposes such as monitoring, screening, and diagnosis. We included RCTs whether they were parallel RCTs, cluster RCTs, crossover RCTs, or factorial RCTs but we excluded quasi-experiments, observational studies, and reviews. We focused on studies in which one of the measured outcomes was depression or depressive symptoms regardless of the outcome measures. Only trials in the English language were eligible for inclusion in this review. RCTs published as journal articles, conference proceedings, and dissertations were included, whereas we excluded conference abstracts and posters, commentaries, preprints, proposals, and editorials. We did not apply restrictions related to the population, year of publication, country of publication, comparator, and study settings.

We followed 3 steps to identify the relevant studies. In the first step, we exported the retrieved studies to EndNote to identify and remove duplicates. Then, 2 reviewers (EA and MA) independently screened the titles and abstracts of all retrieved studies. In the last step, the 2 reviewers independently screened the full texts of studies included from the second step. A third reviewer (AA) resolved any disagreements between the 2 reviewers in the second and third steps. Cohen κ in this review indicated a very good level of interrater agreement in the first (0.85) and second (0.90) steps [39].

Two reviewers (EA and MA) independently extracted data from the included reviews using Microsoft Excel (Microsoft Corporation, Redmond, WA). Multimedia Appendix 3 shows the data extraction form that was used by the 2 reviewers to extract the data precisely and systematically from the included studies. The form was pilot tested using 3 included studies. Any disagreements between the reviewers were resolved by consulting a third reviewer (AA). The interrater agreement between the reviewers was 0.87, indicating a very good level of agreement [39]. Some outcome data (eg, mean, standard deviation, sample size in each group) were missing in 10 studies. Therefore, we contacted their corresponding authors to get them, and 5 corresponding authors did not reply to our emails even after sending 2 reminders.

Two reviewers (EA and MA) independently assessed the risk of bias in the included studies using the Risk-of-Bias 2 (RoB 2) tool, which is recommended by the Cochrane Collaboration [40]. This tool appraises the risk of bias in 5 domains in RCTs: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result [40]. The risk of bias judgments in these domains is used to determine the overall risk of bias for each included study. A third reviewer (AA) resolved any disagreements in judgments between the 2 reviewers. Interrater agreement between the reviewers was very good (Cohen κ=0.93) [39].

We utilized narrative and statistical approaches to synthesize the extracted data. Specifically, in narrative synthesis, texts and tables were used to describe the characteristics of the included studies, population, intervention, comparator, and outcome measures. Then, we grouped and summarized the findings of the included studies according to the type of serious games (ie, exergames or computerized CBT games). A meta-analysis was conducted when at least 2 studies of the same type of serious game reported enough data for the analysis (ie, mean, standard deviation, number of participants in each intervention group). When this information was not reported in any included study, we contacted the first and corresponding authors to get the missing information.

Review Manager (RevMan 5.4) was used to conduct the meta-analysis. We measured the effect of each trial and the overall effect using the standardized mean difference (SMD; Cohen d) because the outcome data (severity of depressive symptoms) were continuous and tools used to measure the outcome were different between the included studies. The random effects model was used in the analysis given the clinical heterogeneity between the meta-analyzed studies in terms of serious game characteristics (eg, types, duration, frequency, and period), population characteristics (eg, sample size, mean age, and health condition), and outcome measures (ie, tools and follow-up period).

When the meta-analysis showed a statistically significant difference between groups, we examined whether this difference was clinically important. A minimal clinically important difference (MCID) is defined as the smallest change in a measured outcome that a patient would consider as worthy and significant and which mandates a change in a patient’s treatment. The MCID boundaries for an outcome were calculated as ±0.5 times the SMD of the meta-analyzed studies.

We checked the characteristics of participants, interventions, comparator, and outcomes in studies included in the meta-analysis to assess their clinical heterogeneity. We also examined the statistical heterogeneity of the meta-analyzed studies by calculating a Chi-square P value and I², which measures the statistical significance of heterogeneity and the degree of heterogeneity, respectively. A Chi-square P value ≤.05 indicates heterogeneous meta-analyzed studies [41]. The degree of heterogeneity was considered unimportant, moderate, substantial, or considerable when I² was 0%-40%, 30%-60%, 50%-90%, or 75%-100%, respectively [41].

We assessed the overall quality of evidence from the meta-analyses using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach, which assesses the quality of evidence based on 5 domains: risk of bias, inconsistency (ie, heterogeneity), indirectness, imprecision, and publication bias [42]. Two reviewers (CT and AA) independently assessed the overall quality of meta-analyzed evidence, and any disagreements were resolved through discussion and consensus. Interrater agreement between the reviewers was very good (Cohen κ=0.88) [39].

As shown in Figure 1, we retrieved 966 citations from searching the 7 electronic databases. Using the software EndNote, we identified and removed 225 duplicates of the retrieved citations. Screening titles and abstracts of the remaining 741 citations led to excluding 592 citations because (1) they did not use serious games (n=354); (2) the severity of depressive symptoms was not a measured outcome (n=69); (3) they were not RCTs (n=119); (4) they were not peer-reviewed articles, theses, nor conference proceedings (n=39); and (5) they were published in non-English languages (n=11). Reading the full text of the remaining 149 publications led to excluding 127 publications because (1) they did not use serious games (n=39), (2) the severity of depressive symptoms was not a measured outcome (n=32), (3) they were not RCTs (n=53), and (4) they were published in non-English languages (n=3). We identified 5 additional RCTs through backward and forward reference list checking. In total, 27 RCTs were included in the current review [43-69]. Of those, 16 RCTs were included in the meta-analyses [45-52,54,59-65].

The included studies were published between 2012 and 2021 (Table 1). The years that witnessed the largest number of included studies were 2018 (n=5) and 2020 (n=5). The included studies were carried out in 15 different countries, as shown in Table 1. The country that published the largest number of the included studies was Germany (n=5). All included papers were papers published in peer-reviewed journals. The trial type used in the most included studies was parallel RCTs (n=24).

The sample size in the included studies varied from 32 to 540, with an average of 104. The mean age of participants in the included studies ranged between 9.41 years and 84.5 years, with an average of 43.9 years. The percentage of the sample who were men reported in 26 studies ranged from 0% to 100%, with an average of 46.1%. Participants’ health conditions were varied between studies, and depression and stroke were the most common (n=4 each). Participants in most studies were recruited from clinical settings (n=20).

The intervention in the included studies was only serious games in 19 studies, serious games plus occupational therapy in 2 studies, and serious games plus psychotherapy in 1 study (Table 2). The most common games used in the included studies were SPARX (n=4) and Wii Fit (n=4). There were 5 types of serious games based on the therapeutic modality that they deliver: exergames (n=16), computerized CBT games (n=8), exposure therapy games (n=1), brain training games (n=1), rational emotive behavioral therapy (REBT) and rational emotive behavioral therapy education (REBE)–based game (n=1). Although games were designed with a “serious” purpose from the beginning (designed serious games) in 14 studies, they were not designed as serious games but were being used for a serious purpose (purpose-shifted games) in the remaining 13 studies. The most common platforms used for playing the games were computers (n=12) and video game consoles and their accessories (eg, balance board; n=12). The duration of the games in the included studies ranged between 5 minutes and 85 minutes, but it ranged between 20 minutes and 45 minutes in most studies (n=14). The frequency of playing the games varied between once a week and once a day, but it ranged between once a week and 3 times a week in 20 studies. The period of the intervention varied between 1 week and 24 weeks, but it ranged from 4 to 8 weeks in 19 studies.

As shown in Table 3, the comparison groups received inactive interventions in 15 studies, while they received active interventions in 18 studies (eg, conventional exercises, CBT programs, video games, and psychotherapy). Note that the numbers do not add up because 6 studies delivered both active and inactive interventions as comparators. The duration of the active comparators ranged between 12 minutes and 100 minutes. The frequency of playing the active comparators varied between once a week and once a day. The period of the active comparators varied between 1 week and 24 weeks. The outcome of interest (eg, severity of depressive symptoms) was measured using 18 different tools, but the most common tool used by the included studies was the Beck Depression Inventory (BDI; n=6), followed by the Hospital Anxiety and Depression Scale (HADS; n=4). The outcome of interest was measured immediately after the intervention in all included studies, and the most common follow-up period was 3 months (n=6). Participant attrition was reported in 24 studies and ranged from 0 to 134.

The random allocation sequence for the randomization process was appropriate in 23 included studies. However, only 10 studies concealed the allocation sequence until participants were enrolled and assigned to interventions, and groups were not comparable in 4 studies. Accordingly, the risk of bias due to the randomization process was rated as low for only 8 studies (Figure 2).

Participants and individuals delivering the interventions were aware of assigned interventions during the experiment in 22 and 20 studies, respectively. Deviation from the intended intervention occurred in 2 studies due to the experimental contexts. Only 14 studies used an appropriate analysis (intention-to-treat or modified intention-to-treat analyses) to estimate the effect of assignment to intervention. Therefore, the risk of bias due to the deviations from the intended interventions was judged as low in only 8 studies (Figure 2).

Outcome data were not available for all or nearly all participants in 21 studies, and there was evidence that the findings were not biased by missing outcome data in only 5 studies. The reasons for missing outcome data could not be related to the true value of the outcome in 18 studies. As a result, 17 studies were judged as having a low risk of bias in the “missing outcome data” domain.

All included studies assessed the outcome of interest (ie, depression level) using appropriate measures and used measurement methods comparable across intervention groups. However, the assessor of the outcome was blinded in only 9 studies. For this reason, only these studies were rated as low risk of bias in the “measuring the outcome” domain (Figure 2).

In 17 studies, a prespecified analysis plan (ie, protocol) was not published. Only 4 studies reported outcome measurements different from those specified in the analysis plan. There is no evidence that all included studies selected their results from many results produced from multiple eligible analyses of the data. Accordingly, the risk of bias due to the selection of the reported results was considered low in 4 studies (Figure 2).

In the last domain “overall bias,” the risk of bias was considered high in 20 studies as they were judged as having a high risk of bias in at least one domain; 6 studies were judged to have some concerns in the domain of overall bias as they had some concerns in at least one of the domains and were not at high risk for any domain. The remaining study was judged to be at low risk of bias for the domain of overall bias given that it was rated to be at low risk of bias for all domains. Reviewers’ judgments about each “risk of bias” domain for each included study are presented in Multimedia Appendix 4.

This review assessed the effectiveness of serious games on the severity of depressive symptoms as reported by RCTs. Although 27 RCTs were included in the current review, 16 studies were included in the meta-analysis. Very low-quality evidence from 7 RCTs showed no statistically significant effect of exergames on the severity of depressive symptoms as compared with conventional exercises. Furthermore, 3 studies that compared the effect of exergames with that of other active interventions (eg, occupational therapy and robot-assisted training) on the severity of depressive symptoms and were not included in the meta-analyses found no statistically significant difference between the groups. These findings indicate that exergames are as effective as active interventions, which are usually delivered and supervised by health care providers (eg, physiotherapists, occupational therapists, and psychologists).

Very low-quality evidence from 5 RCTs showed a statistically and clinically significant effect of exergames on the severity of depressive symptoms when compared with no intervention.

Findings in this review are comparable to other reviews. Specifically, a recently published meta-analysis of 5 RCTs conducted by Yen and Chiu [37] showed an overall statistically significant effect (P<.001) of exergames on depression. Additionally, another recent meta-analysis of 8 RCTs conducted by Li et al [19] showed a significant effect of exergames on depression. However, both reviews [19,37] compared the effect of exergames with the effects of different active and inactive interventions through one meta-analysis, while our review conducted 2 separate meta-analyses to compare exergames with conventional exercises and no intervention respecting the uniqueness of these 2 interventions. Further, in contrast to our review, 5 of the 8 studies included in the review by Li et al were not RCTs (quasi-experimental or pre-post one-group trials) [19].

Very low-quality evidence from 6 RCTs showed a statistically and clinically significant effect of computerized CBT games on the severity of depressive symptoms when compared with no intervention. In contrast, 3 studies that compared the effect of computerized CBT games with those of active interventions (eg, video games and conventional CBT) on depressive symptoms and were not included in the meta-analyses found no statistically significant difference between the groups. This insignificant effect can be attributed to the fact that conventional CBT is comparable to the active interventions, thereby comparing the effect of 2 comparable interventions usually produces no significant difference, which indicates that computerized CBT games are at least as effective as these active interventions. None of the previous reviews [18,19,35-37] assessed the effect of computerized CBT games on depression.

Overall, serious games can be better than no intervention in alleviating depression and as effective in alleviating depression as other active interventions (eg, conventional CBT, exposure therapy, conventional exercise). However, definitive conclusions regarding the effectiveness of serious games could not be drawn in this review because the overall risk of bias was high in most included studies, the quality of the meta-analyzed evidence was very low, and few studies recruited patients with depression. Therefore, we can only recommend health care providers consider offering serious games as an adjunct to existing interventions until further, more robust evidence is available. To have sufficient evidence, future studies should assess the effectiveness of serious games that are designed specifically to alleviate depression and deliver other therapeutic modalities, recruit participants with depression, and avoid biases by following recommended guidelines for conducting and reporting RCTs (eg, RoB 2).