We conducted electronic searching on 3 September 2018 and identified 5,367 potentially relevant studies and removed 1,380 duplicates. Of the resulting 3,987 studies, 3,948 studies were excluded by title and abstract review. After the remaining 39 full-text studies were read; we excluded 35 full-text studies because of differing inclusion criteria. Finally, we included four studies for this review (27-30), with two of these studies included in the quantitative analysis (28, 29) (Fig. 1). We resolved all disagreements by discussion with each screening author during the screening process.
The results of the risk of bias assessment are displayed in Fig. 2 and Fig. 3. Both were randomized control trials. However, due to the nature of the intervention, it was not possible to blind patients, and all the outcomes were evaluated by self-assessment questionnaire. Hence we assessed both trials as a high risk for performance bias and detection bias. Fazelnya et al. (2017) did not provide enough description to evaluate bias, and we assessed as unclear for selection bias, attrition bias, and reporting bias (28). The summary of findings table (Table 1) shows the GRADE assessment. We included 439 participants who were undergoing treatment for cancer and were between eight and 29 years of age, of these 388 were children who met our inclusion criteria and 47 young adults (over 19 years old). Although there were four excluded participants after random assignment, we could not identify the age. The age group was different from each study. Kato et al. (2008) included between 18 years and 29 years of age and Fazelniya et al. (2017) included between eight years and 12 years of age. Settings were one hospital in Iran (28), and 34 cancer treatment centers in the United States, Canada, Australia (29). Types of interventions were different computer games, compared with standard care or an alternative non-cancer video game. We summarized the characteristics of included studies in Table 2. We conducted a meta-analysis about Quality of life (QOL). However, we could not conduct a meta-analysis for the outcomes of (a) knowledge, (b) perceived stress, (c) self-efficacy, and (d) health locus of control because only one trial reported on them. Thus, we synthesized the results narratively.
Primary outcomes
Knowledge, psychological, and physical outcomes were described. Psychological outcomes reported were: (a) perceived stress, (b) self-efficacy, and (c) health locus of control. QOL was included in both psychological and physical health outcomes. There were no descriptions of social outcomes.
Knowledge
Kato et al. (2008) reported on knowledge. They developed an original cancer knowledge scale for their study. The cancer knowledge scale is a questionnaire with 18 multiple-choice items; total scores ranged from 0% to 100%, and higher scores indicate adequate cancer knowledge. Cancer-related knowledge in the audiovisual intervention group had a greater increase than in the control group for cancer knowledge based on a mixed-effect liner model analyses (group × time interaction, p = 0.035) (very low certainty of the evidence) (29).
Quality of life
There were two RCT studies on QOL (28, 29). Fazelniya et al. (2017) measured QOL using the Pediatric Quality of Life Inventory (PedsQL) 3.0 Cancer Module child self-report, which was developed to measure health-related QOL, specifically for children aged 8-12 years old, who were diagnosed with cancer (31). Kato et al. (2008) measured QOL using the Pediatric Quality of Life self-report instrument (PQL) for children and adolescents under 18 years old (32). We conducted a meta-analysis using a random effect model and showed SMD. There was no significant difference between the audiovisual intervention group and the control group (SMD 0.73, 95%[CI -0.51-1.97]), and heterogeneity was high (X 2 = 17.11, df = 1, I2 = 94%). However, two studies showed a positive direction for the effect (Fig. 4) (very low certainty of the evidence). In terms of study results, Fazelniya et al. (2017) reported that the intervention increased participants’ QOL (post hoc Least Significant Difference (LSD) test [after intervention: p = 0.030, after one month: p < 0.001]) (28). Kato et al. (2008) reported that the intervention did not increase participants QOL (between group × time interaction; PQL, p = 0.112) (29).
Perceived stress
Kato et al. (2008) reported on perceived stress. They measured perceived stress using the Perceived Stress Scale 10 with ten items and a total range from 10 to 50 (33). There was no significant difference between group × time interaction by mixed-effect liner model analyses (p = 0.931) (very low certainty of evidence) (29).
Self-efficacy
Kato et al. (2008) measured self-efficacy, using the self-efficacy scale developed for the trial. The intervention increased self-efficacy by mixed-effect liner model analyses (group × time interaction, p = 0.011) (very low certainty of the evidence) (29).
Health locus of control
Kato et al. (2008) measured health locus of control, using the Multidimensional Health Locus of Control Scale Form with 18 items including five subscales and total range from three to 36 (34). We did not show those results on the summary of findings table (Table 1), because they were reported as the results of five subscales and there were no descriptions of the total scores. There was no significant difference between group × time interaction by mixed-effect liner model analyses (p = 0.608) (29). We assessed the certainty of evidence as a very low certainty of evidence because of the high risk of bias, small sample size, and inclusion of young adults.