A systematic review and meta-analysis on the clinical implications of probability discounting among individuals with Internet gaming disorder

Abstract

The significance of probability discounting (PD) among individuals with Internet gaming disorder (IGD) remains unclear. Following the PRISMA guidelines, we systematically searched the PubMed, Embase, and ScienceDirect databases for English articles on Internet addiction that included comparison between individuals with and without IGD as well as probabilistic discounting task as the main outcome from January 1970 to July 2020 using the appropriate keyword strings. The primary outcome was the overall difference in rate of PD, while the secondary outcomes included the difference in PD with magnitude of probabilistic reward and response time of the PD task. Effect size (ES) was calculated through dividing the group means (e.g., h value or AUC) by the pooled standard deviations of the two groups. A total of five studies with 300 participants (i.e., IGD group, n = 150, mean age = 20.27 ± 2.68; healthy controls, n = 150, mean age = 20.70 ± 2.81) were analyzed. The IGD group was more willing to take risks in probabilistic gains but performances on probabilistic losses were similar between the two groups. The IGD group also exhibited a shorter response time (Hedge's g = - 0.51; 95%CI = - 0.87 to - 0.15). Meta-regression demonstrated a positive correlation between maximum reward magnitude and PD rate (p < 0.04). However, significant publication bias was noted among the included studies (Egger's test, p < 0.01). In conclusion, individuals with IGD seemed more impulsive in making risky decisions, especially when the potential gains were expected. Our findings not only supported the use of PD for assessing individuals with IGD but may also provide new insights into appropriate interventions.

Figure 1

Probabilistic discounting functions. (A) Probability discounting of gains, and (B) Probability discounting of losses.

Figure 2

PRISMA flowchart for identifying eligible studies.

Figure 3

Forest plot of the effect sizes for the difference in probability discounting between subjects with Internet gaming disorder and healthy controls. The bars with squares in the middle denote 95% confidence intervals (95% CIs) and effect sizes (ESs). The diamond represents the pooled effect size (ES).

Figure 4

Leave-one-out sensitivity analysis. The bars with squares in the middle denote 95% confidence intervals (95% CIs) and effect sizes (ESs). The diamond represents the pooled effect size (ES).

Figure 5

Forest plot of the effect sizes for the difference in probabilistic (A) gain, and (B) loss between subjects with Internet gaming disorder and healthy controls. The bars with squares in the middle denote 95% confidence intervals (95% CIs) and effect sizes (ESs). The vertical dotted line indicates the ESs for null hypothesis. The diamond represents the pooled effect size (ES).

Figure 6

Random-effects funnel plot detailing publication bias in the studies reporting probability discounting between subjects with Internet gaming disorder and healthy controls.