The systematic review and meta-analysis were conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines [25] and were registered on the International Prospective Register of Systematic Reviews (PROSPERO) with registration number CRD42024544081.

We searched 6 databases (ie, PubMed, Embase, Cochrane Library, Web of Science, CINAHL, and SPORTDiscus) in English for relevant RCTs published from the inception of each database to March 21, 2024. The search strategy was developed with the assistance of a research librarian based on the PICOS (Participants, Interventions, Comparisons, Outcomes, and Study Design) framework. MeSH (Medical Subject Headings) and key search terms related to “exergaming,” “exergame*,” “active-video gam*,” “exercise,” “physical activit*,” and “randomized controlled trial” were included. Additionally, to ensure comprehensive retrieval, we manually searched the references of previously published relevant reviews (Multimedia Appendix 1).

The inclusion criteria for this systematic review and meta-analysis were as follows: (1) participants: any human population, irrespective of health condition; (2) interventions: exergame used either as the primary intervention or as a supplementary component; (3) comparisons: either active controls (eg, PA education programs) or inactive controls (eg, no intervention and waitlist control); (4) outcomes: PA outcomes measured both objectively, including minutes of various types of PA, and subjectively through assessment tools such as the International Physical Activity Questionnaire; in addition, BMI was also included as an indirect measure of PA effects; and (5) studies: RCTs of all forms, including parallel, cluster, and crossover designs. Conference papers and abstracts were excluded to ensure data quality, and full texts lacking available data were also excluded.

All studies retrieved were imported into EndNote 20 (Clarivate), and duplicates were automatically removed. The results were then exported to the Rayyan web-based platform [26]. Two authors (SJL and LQZ) independently screened the titles and abstracts based on the eligibility criteria. Next, full texts of potentially relevant articles were retrieved and assessed for inclusion, with reasons for exclusion recorded. Any discrepancies were resolved through discussions with a third investigator (HMM) to reach a consensus.

Study characteristics (including author, year, country, and study arms), population details (including population type, sample size, average years, sex difference, and baseline BMI), intervention specifics (including theory, site, frequency, type of training, duration, and period), exergame characteristics (including game name, primary device, and content), brief comparison content, and outcome data were extracted into a predesigned Microsoft Excel format by one author (SJL) and cross-checked by another reviewer (LQZ). The mean plus SD format was used to represent outcome data, and other statistics were converted to mean (SD) format [27]. Median values were transformed into means if mean values were not reported in the original articles [28]. Additionally, other data formats (such as 95% CI or standard errors) were also converted into SD (Cochrane Handbook, version 6.4, Chapter 5) [27]. If outcome indicators had different units, such as using metabolic equivalent of task to measure PA, we applied conversion formulas mentioned in the methods section of the articles to convert them into minutes per day, ensuring consistency in units. If a study involved multiple study arms, each individual intervention was treated as a separate study, and the number of participants in the shared control group was divided by the number of study arms to avoid duplicate inclusion of participants in the meta-analysis (Cochrane Handbook, version 6.4, Chapter 23) [27]. Michie’s BCT taxonomy (BCTTv1) was used to identify BCTs used in the original studies (Multimedia Appendix 2) [21]. If intervention descriptors in original RCTs matched the definitions of BCTs, we assigned a “√.” One reviewer (SJL) initially identified the BCTs, which were then double-checked and confirmed by another reviewer (HMM).

Two independent reviewers (SJL and LQZ) evaluated the methodological quality of the included RCTs using version 1 of the Cochrane risk of bias tool [29]. This tool consisted of 7 categories: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other potential biases such as conflicts of interest. Each domain was assessed and rated as having low, unclear, or high risk of bias. Any discrepancies during the evaluation process were resolved by a third investigator (HMM).

The meta-analysis aimed to assess the effectiveness of exergame-based behavior change interventions on promoting PA. Total PA, defined as the sum of LPA, MPA, and VPA measured either by a scale or an accelerometer, was the primary outcome, with MVPA, LPA, MPA, VPA, SED, step count, and BMI as secondary outcomes. Given the diverse measurement methods for outcomes, we calculated standard mean differences (SMDs), estimated using adjusted Hedges g, to generate effect sizes [30,31]. SMD values of 0.2, 0.5, and 0.8 corresponded to small, medium, and large effects, respectively [32]. Data heterogeneity was evaluated using the I² statistic, categorizing values of 25%, 50%, or 75% as low, moderate, and high, respectively [33]. A random-effects model was used for data analysis to obtain conservative results. Publication bias was assessed through visual evaluation of funnel plots asymmetry and quantified using Egger tests. Publication bias assessments were conducted only for outcome measures that included more than 10 original studies (Cochrane Handbook, version 6.4, Chapter 13) [27]. Subgroup analyses were conducted to explore influencing factors of exergame-based behavior change interventions and potential sources of heterogeneity. Review Manager software (version 5.3; Cochrane Collaboration) and Stata software (version 16; StataCorp) were used to analyze data. In all analyses, P<.05 (2-sided) was considered statistically significant.

The literature search initially yielded 4452 records. Additionally, 7 studies were identified by hand-searching the references of relevant reviews. After the removal of 2222 duplicates using EndNote 20, titles and abstracts of 2237 studies were screened. Subsequently, 85 studies were included for full text screening, with 65 being excluded due to wrong interventions (n=1), wrong comparisons (n=3), wrong outcomes (n=41), wrong study designs (n=2), lack of outcome data (n=4), conference abstracts (n=13), and 2 protocols (see Multimedia Appendix 3 for a detailed list of exclusions with corresponding reasons). Finally, 20 RCTs [12-18,34-46] were included in the systematic review and meta-analysis (see Figure 1).

The studies included in the review were published between 2008 and 2023. A total of 20 studies were performed in Canada [15,34,37], the United States [14,17,18,35,36,42,43], Turkey [16,44], Spain [13], Australia [38,45], China (Hong Kong) [40], the Netherlands [46], Germany [39], Singapore [41], and Finland [12]. Two studies [15,17] used a 3-arm RCT design, while the remaining studies [12-14,16,18,34-46] used a 2-arm RCT design. Among them, 5 studies [15,18,36,44,46] focused on adults and 15 studies [12-14,16,17,34,35,37-43,45] targeted children. The sample sizes varied from 16 to 1112, with participants’ mean ages ranging from 7.5 to 79 years. The review included 2211 participants for meta-analysis, with 48.5% (n=1073) being female. BMI was also recorded to provide a general assessment of the participants’ body composition. Detailed characteristics of the studies and populations are listed in Table 1.

The design of some exergame-based interventions incorporated theories (ie, self-determination theory, social cognitive theory, and the theory of planned behavior) or theoretical frameworks (reserve capacity model and family ecological model). Interventions took place either at home or outside the home, such as in a hospital. The type of training ranged from simple activities such as walking and cycling to complex multicomponent exercises. The intervention period spanned from 1 to 24 weeks, with frequencies ranging from 1 to 7 times per week, each lasting between 20 and 60 minutes. Measured outcomes included PA, MVPA, LPA, MPA, VPA, SED, step count, and BMI, as detailed in Table 2. Regarding the characteristics of exergames, a total of 50 exergames (eg, Game Bike, Dance Dance Revolution, and Wii Sports) were identified across the 20 RCTs. Sony PlayStation, Nintendo Wii, and Microsoft Xbox were the primary devices used in these exergames. The game content included a diverse range of activities such as cycling, sports, dance, aerobics, strength training, yoga, adventure, walking, mini-games, pet simulation, and rhythm-based games. Detailed information about the exergames can be found in Multimedia Appendix 4 [12-18,34-46].

The application of BCTs in the included RCTs varied, with each study incorporating between 1 and 15 different techniques. The most frequently used BCTs, which appeared in more than half of the studies, included “1.4 action planning” (n=15), “1.1 goal setting” (n=13), “12.5 adding objects to the environment” (n=13), “2.3 self-monitoring of behavior” (n=11), and “4.1 instruction on how to perform the behavior” (n=11) (shown in Table 3 and Multimedia Appendix 5).

A vast majority (15/20, 75%) of the studies explicitly described the generation of random sequences, while those that mentioned randomization without detailing the process were categorized as unclear. Only 4 studies provided the specific details on the methods of allocation concealment, such as using opaque envelopes to conceal the random sequences. Due to the nature of behavior change intervention studies, it is unlikely to blind participants. Therefore, all studies were rated as high risk in this category. Information regarding the blinding of outcome assessment was missing in 60% (12/20) of the studies. Only one study was assessed as unclear in the incomplete outcome data section due to an insufficient explanation for the missing data. Most studies were categorized as low risk in the selective reporting and other potential biases. Detailed risk of bias assessment results can be seen in Figure 2.

The systematic review and meta-analysis assessed the effectiveness of exergame-based interventions on primary outcomes, specifically PA, as well as on secondary outcomes, including MVPA, LPA, MPA, VPA, SED, step count, and BMI. Additionally, subgroup analyses were performed to identify potential factors influencing the effectiveness of the interventions.

A total of 11 studies [12-14,36,38,40-44,46], including 828 participants in the intervention groups and 775 participants in the control groups, were analyzed to investigate the effects of exergame-based interventions on PA. The pooled effect size of PA was 0.19 (95% CI 0.05-0.33), indicating a statistically significant small effect. Additionally, a low level of heterogeneity was observed (I²=14%) (see Figure 3). No publication bias was detected by the Egger test (P=.12). The funnel plot is displayed in Figure 4.

Sixteen subgroup analyses were conducted to evaluate the effects of exergame-based interventions: population (age group [children vs adults] and health status [healthy participants vs participants with health conditions]), intervention design (theoretical basis [theory-based vs nontheory-based], site [school vs home vs hospital], frequency [4x/week vs ≥4x/week], session duration [<40 minutes vs ≥40 minutes], intervention period [<3 months vs ≥3 months], family involvement [yes vs no], technology support [entirely vs partially]), and intervention implementer [research assistants vs other implementers]), exergame design (game primary device [console vs smartphone vs cycling simulator], game console [Xbox, PlayStation, and Nintendo Wii], game participation type [individual game vs nonindividual game], and game content [single-category vs multicategory]), PA measurement (subjective vs objective), and BCT number used (n<7 vs 7≤n<10 vs n≥10).

Intervention implementer, game console, game participation type, measurement methods, and the number of BCTs used in each study significantly influenced the effects of exergame-based interventions on PA. Compared to other study implementers (eg, physical therapists), studies implemented by research assistants exhibited statistically significant higher effect sizes (SMD 0.36 vs 0.09, P=.04); similarly, the Xbox group and the nonindividual game group achieved significantly higher effect sizes than the PlayStation and Nintendo Wii group (SMD 0.72 vs 0.21 vs 0.09, P=.01), and the individual game group (SMD 0.46 vs 0.10, P=.02), respectively. Additionally, objectively measured PA demonstrated a larger effect size (SMD 0.41) compared to subjectively measured PA (SMD 0.09, P=.03). Similarly, studies using 10 or more BCTs showed a greater effect size (SMD 0.55) than those in the other groups.

Other important variables, such as age groups, intervention period, and game content, warrant our attention. Although they did not reveal significant subgroup differences, children showed significantly greater improvements in PA compared to adults (SMD 0.26, 95% CI 0.03 to 0.49 vs SMD 0.13, 95% CI −0.16 to 0.42). Interventions lasting longer than 3 months demonstrated significantly better PA outcomes than those shorter than 3 months (SMD 0.29, 95% CI 0.03 to 0.56 vs SMD 0.10, 95% CI −0.01 to 0.21). Single-category game content had significantly better PA outcomes than multicategory game content, such as combining dance and sports (SMD 0.29, 95% CI 0.04 to 0.54 vs SMD 0.13, 95% CI −0.02 to 0.28). Detailed subgroup analyses are listed in Table 4.

Exergame-based behavior change interventions significantly increased MVPA (SMD 0.48, 95% CI 0.12-0.85, I²=65%) and step counts (SMD 0.54, 95% CI 0.13-0.94, I²=62%). However, the interventions did not show significant effects on LPA, MPA, VPA, SED, or BMI. Forest plots for these secondary outcomes are available in Multimedia Appendix 6.

To our knowledge, this is the first meta-analysis of RCTs that evaluates the effects of exergame-based behavior change interventions for promoting PA and identifies the BCTs used in these interventions. This systematic review and meta-analysis included a total of 20 studies. Eight outcomes (ie, PA, MVPA, LPA, MPA, VPA, SED, step count, and BMI) were explored. The results demonstrated that exergame-based behavior change interventions significantly increased PA, MVPA, and step counts. Furthermore, subgroup analyses showed that intervention implementer (research assistants vs other implementers), game console (Xbox vs PlayStation vs Nintendo Wii), game participation type (individual game vs nonindividual game), measurement method (subjective vs objective), and number of BCTs (n<7 vs 7≤n<10 vs n≥10) significantly influenced the effectiveness of these interventions.

In our study, the pooled SMD for PA was 0.19 (95% CI 0.05-0.33), indicating a small effect with statistical significance. This finding is partially consistent with a previous meta-analysis conducted by Moller et al [8] in 2023, which reported a moderate effect size (0.53, 95% CI 0.32-0.73) of exergame interventions on PA behaviors. The main difference between our study and that of Moller et al [8] is that our analysis was restricted to RCTs, whereas Moller’s study included both RCTs and other types of experimental studies. This inclusion may account for the slight differences between the results. However, compared to other study designs, RCTs provide stronger evidence. Overall, exergame-based interventions have shown effectiveness in promoting PA. Flow experience offers a potential explanation for the positive changes in PA behavior observed in exergame-based interventions [47]. Exergames represent an innovative way to combine exercise and gaming, enhancing participants’ pleasure and enjoyment during physical activities [48,49]. Therefore, participants are more likely to experience flow while engaging in exergames, leading to positive behavioral changes [50,51]. The relationship between positive behavior changes in PA and improved health outcomes merits further discussion. Children and adolescents are recommended to engage in at least an average of 60 minutes/day of MVPA, while adults should do a minimum of 150‐300 minutes of MPA, or 75‐150 minutes of VPA per week [2]. Given the variations among different PA metrics, total PA is not sufficient to infer changes in health outcomes resulting from PA. Therefore, we performed meta-analyses to examine the effects on various types of PA, including MVPA, LPA, MPA, VPA, and SED. Exergame-based interventions increased MVPA (SMD 0.48, 95% CI 0.12-0.85) and step counts (SMD 0.54, 95% CI 0.13-0.94), representing moderate effect sizes. However, effect sizes seen in different fields of research vary [52]. We recommend that future studies standardize the units of measurement for PA, such as minutes/day, to directly observe whether behavior change interventions lead to significant health improvements. The effects of exergame-based interventions on LPA, MPA, VPA, SED, and BMI were not statistically significant. Therefore, there is a need to explore better-designed exergame-based interventions.

RCTs included in the meta-analysis involved 2 types of participants: children and adults. Compared to adults, children exhibited greater improvements in PA behavior. This may be because exergame-based interventions are more attractive for children [53]. Exergames should be designed to meet the needs of various participant types to maximize their effectiveness [54,55]. Standardized design processes for exergame-based interventions can also be used in future studies [56]. Theory plays a crucial role in the design of interventions [57,58]. Only 3 RCTs used theory to guide exergame-based intervention design, and no significant subgroup differences in effect sizes were found between using and not using a theory. In the design of exergame-based interventions, the application of theory is not sufficiently emphasized, and theory-based interventions do not fully apply these theories.

The intervention site is also important in designing exergame-based interventions. In our study, we compared the effectiveness of interventions conducted in 3 different settings: school versus home versus hospital. Home-based interventions help participants overcome barriers associated with playing exergames, such as transportation and costs. In contrast, outside-home interventions (ie, school and hospital) are often supervised by professional staff, potentially enhancing the effectiveness of the intervention [59,60]. However, our study did not find significant differences between the 2 intervention sites, indicating that the choice of site may not be a main factor influencing the effectiveness of exergame-based interventions. Nevertheless, given the variability in the populations included in our study, the result should be treated with caution. We recommend that the selection of an intervention site should be based on the study population and research aims. Considering participants’ preferences regarding intervention sites is also important as it can influence their engagement and adherence to exergame-based interventions [61,62].

In order to achieve the recommended PA goals and transform behavior change interventions to significant health benefits, the intervention frequency and session duration should be designed and standardized. Although a slight difference in intervention frequency was observed, the sample sizes of included studies were limited, and our analysis failed to reveal significant differences between these subgroups, and the optimal intervention frequency and session duration require future exploration. The interventions performed by research assistants demonstrated significantly larger effect sizes compared to those implemented by other practitioners (eg, physical therapists). The reason behind this phenomenon might be that exergame-based interventions mainly require implementers to encourage and support participants in engaging with the interventions, rather than relying on specialized skills. Research assistants often have a higher sense of responsibility and motivation to adhere strictly to the study protocol, resulting in better intervention results.

Additionally, to our surprise, the game console brand acted as a significant factor in influencing health benefits effects. This may be related to the specific goals and development strategies used by the game developers in designing these games, particularly those aimed at maximizing either PA promotion or entertainment, depending on the focus of the game. According to our results, Xbox showed better PA promotion outcomes than PlayStation and Nintendo Wii. We advocate for collaborations between behavior change researchers and game companies to achieve the goal of both entertainment and health promotion. Nonindividual games (eg, team games) showed significantly better results than individual games, emphasizing the importance of integrating social support in intervention design, which is also one of the essential BCTs.

Another finding from our subgroup analysis is that interventions lasting 3 months or longer exhibited enhanced effectiveness, although no significant subgroup difference was observed. This may be explained by habit formation, which occurs when behavior is consistently repeated in a regular environment [63,64]. The more frequent the repetitions, the more likely it is that a behavior will become a habit. The link between habit formation and PA behavior change may be a valuable future research direction. The measurement of PA is important when it is the primary outcome of a study [65,66]. We have classified the measurement methods into subjective and objective measurements. Our results indicate that, compared to subjective methods, objective measurement demonstrated greater effectiveness in assessing PA. Using objective measurements in future studies is advised to enhance the reliability of the outcomes. Additionally, combining objective and subjective measurements can complement each other and provide rich result data.

A comprehensive examination of the effectiveness of exergame-based interventions for promoting PA was conducted. Only RCTs were included in our systematic review and meta-analysis, ensuring the rigor of the study design. We explored 8 PA-related outcomes, providing a comprehensive overview of the results from exergame-based interventions. Additionally, subgroup analyses were performed to identify factors influencing effectiveness, offering valuable insights for the design of future interventions. However, the meta-analysis still has some limitations. First, although our systematic review and meta-analysis included 20 RCTs, only 11 of these, which provided total PA as an outcome, were included in subgroup analyses. The limited number of studies and sample sizes might restrict the ability to detect significant differences between subgroups, and the criteria for subgroup categorization were somewhat subjective, which might have overlooked other influencing factors due to classification issues. Therefore, the corresponding subgroup results should be treated with caution. Second, while the BCT coding was reviewed by another author (HMM) and showed good consistency, it still retained some degree of subjectivity. Additionally, differences in the reporting by the original study authors might have led to variations between the identified BCTs and those actually implemented in the interventions. Furthermore, the quality and intensity of the BCT used were more important than their labeling. For instance, both daily and weekly feedback were coded as “feedback” in the BCT taxonomy, but their effects were substantially different. This is why we do not conduct meta-regression in our study to identify effective BCTs. Although significant effects may be observed in some BCTs, the results were not reliable and could be misguiding. We recommend using BCT as a guide when designing interventions rather than inferring the effectiveness of using specific BCTs. Third, due to PA measurement differences across the original RCTs, we used SMD to aggregate effect sizes. However, this approach prevented direct comparisons with recommended PA dosage, thus limiting our ability to infer the health benefits from behavior changes. We suggest that standardized PA measurement methods should be used in future studies to enhance direct comparisons and interpretations.

In conclusion, our systematic review and meta-analysis have demonstrated that exergame-based interventions are effective in promoting PA. Furthermore, we explored how different factors and BCTs influence the effectiveness of these interventions. Future trials are needed to further validate the insights proposed in our studies and to assess the long-term effects on PA.