Reward‐related decision‐making deficits in internet gaming disorder: a systematic review and meta‐analysis

Introduction

Internet gaming disorder (IGD) has received much attention with the rapid expansion of on-line gaming use during the last two decades. There were more than 930 million active internet gamers around the world in 2020 (https://www.statista.com/outlook/212/online-games) and the prevalence of IGD was approximately 4.6%, according to a recent meta-analysis of 16 survey studies [1]. IGD may lead to a variety of dysfunctions related to physical health, work performance and social interactions [2, 3]. Considering the large number of affected individuals and its negative impact on both personal life and social productivity, IGD has been included in both the 11th revision of the International Classification of Diseases (ICD-11; https://icd.who.int/browse11/l-m/en) and the appendix of the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) [4] as a world-wide condition warranting further study [5]. Despite much effort, the psychopathology of IGD remains to be elucidated.

Individuals with IGD are characterized by persistent gaming despite potential negative consequences [2, 4], which may be attributed to impaired risk evaluation and reward processing during decision-making [6]. Indeed, substance use disorders (SUDs) and gambling disorder [7-10] have been frequently associated with impaired reward-related decision-making, a pervasive process when individuals need to make a choice from several options based on subjective values [10, 11]. Reward-related decision-making dysfunction was included in the Research Domain Criteria (RDoC) framework as an important transdiagnosic construct [12]. Most theories of IGD also highlight its role in the development and maintenance of this condition [6, 13, 14]. For example, both the Interaction of Person–Affect–Cognition–Execution (I-PACE) [6] and tripartite neurocognitive models [14] propose that the imbalance of the impulsive (or affect) and reflective (or cognitive) systems lead to poor decision-making in IGD. These theoretical models received some empirical support [15-17]; however, a few studies also showed normal or even better decision-making performance in individuals with IGD [18-20]. Although a previous meta-analysis has partly addressed the relationship between decision-making alterations and internet addiction, it focused upon general cognitive deficits and included only several reward-related decision-making studies [8]. Thus, it is not yet clear to what extent individuals with IGD may be impaired on reward-related decision-making.

A possible factor that might influence the relationship between IGD and reward-related decision-making performance is the decision-making situation. Multiple reward-related decision-making tasks have been investigated in IGD studies. These tasks may be broadly categorized into three situational types: ambiguous, risky and inter-temporal decision-making. Specifically, ambiguous decision-making involves choices in which the outcome information (e.g. risk, reward) is unclear beforehand. Therefore, to achieve better performance, participants need to learn the information related to each option from trial and error [21]. A representative example of this situation is the Iowa Gambling Task (IGT) [22], in which different monetary outcomes and winning probabilities are associated with four decks of cards but unknown to participants at the beginning. In order to earn more money, they should learn more information via feedback and try to select the decks with the highest expected values. Conversely, risky decision-making presents the outcomes and probabilities of each option explicitly. Thus, trial-and-error learning is typically not necessary for this situation [23]. This category includes a few classic tasks, including the Game of Dice Task (GDT) [24], Cambridge Gambling Task (CGT) [25] and Cups Task [26]. Finally, inter-temporal decision-making focuses upon delays to gain the reward (instead of probabilities described in the above two situations) as a discounting factor [27], so tasks assessing this form of decision-making have also been termed delay-discounting tasks [7].

Although these three decision-making situations have been widely examined, few studies have investigated if individuals with IGD may be particularly impaired on a certain domain or show comparable impairments among these situations. Different decision-making situations may depend upon overlapping but distinct cognitive processes [11]. For example, valuation and action selection are commonly involved in all these three situations [11], whereas some other processes, such as time evaluation for inter-temporal decision-making [27], are uniquely required in certain situations. Therefore, a direct comparison between decision-making situations may provide insight into which basic processes that individuals with IGD show impairments. Tailored treatments targeting these deficits may achieve better therapeutic outcomes [28].

Besides decision-making situation, valence of outcomes (e.g. gains or losses) may also contribute importantly to decision-making and have different impacts upon healthy individuals and those with IGD. Individuals are typically more sensitive to the prospect of losses than gains during decision-making [29, 30]. Reduced loss aversion has been observed in SUDs and gambling disorder [31-33]. In the field of IGD, the findings are largely mixed. Although a few studies showed worse loss-related decision-making in this population [23, 34], opposite findings have also been reported [35]. Thus, whether individuals with IGD behave differently when seeking gains and avoiding losses remains an open question.

Taken together, the existing literature suggests a close association between IGD and decision-making deficits. However, most studies only included a small number of participants and used heterogeneous tasks, which may yield inconsistent results. To clarify the situation, we conducted a systemic review and comprehensive meta-analysis to estimate the aggregated effect size (ES) for decision-making alterations in IGD and examine the moderating effects of decision-making situation and valence. Based on the above-mentioned evidence in IGD and other addictive disorders, we hypothesized that individuals with IGD would exhibit dysfunction related to reward-related decision-making in general, and such deficits might be particularly associated with certain types of decision-making situations (e.g. inter-temporal decision-making) [36, 37] and valence (e.g. loss-related decision-making) [31, 34]. As studies based upon clinical and community samples were included, we also examined systematic differences between these studies and hypothesized that clinical samples would perform worse. Moreover, some [e.g. functional magnetic resonance imaging (fMRI) or electroencephalography (EEG)] studies conducted the decision-making tasks when measuring neural activity; we thus explored the impact of the testing environments on the results. Although studies have suggested that the additional equipment may lead to motor slowing and less attentional focus [38], neither is critical for most reward-related decision-making tasks [11]. Therefore, we hypothesized that behavioral and neuroimaging studies would yield similar results. Finally, as reward-related decision-making deficits have been proposed to relate to the development of IGD [6, 13, 14] and may overlap with impulsivity [39], another core component of addictions [40], we hypothesized that decision-making alterations in IGD would be associated with both IGD severity and impulsivity differences at the study level.

Methods Study selection

We followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to conduct the study selection [41]. Relevant studies (January 1995–June 2020) were identified via searches of PubMed, Web of Science and ProQuest databases using the following search terms: (decision OR choice OR risk OR ambiguity OR uncertainty OR gambl* OR ‘Game of Dice Task’ OR GDT OR ‘delay discounting’ OR DDT) AND (‘internet gaming’ OR ‘computer gaming’ OR ‘online gaming’ OR ‘video gaming’ OR ‘gaming addiction’ OR ‘gaming disorder’ OR ‘excessive gaming’).

Studies were included if they: (1) were peer-reviewed original articles on humans; (2) included at least one reward-related decision-making task; (3) included an IGD group diagnosed by predefined criteria; (4) included a healthy control group; (5) compared reward-related decision-making performance between IGD and control groups; and (6) reported sufficient information for the calculation of effect sizes (ESs). Studies that focused on off-line gaming disorder or determined IGD status after participant enrollment (e.g. based on median scores) were excluded. Moreover, as the current study focused upon reward-related decision-making, studies using perceptual and social decision-making tasks were also excluded. The study selection was independently conducted by two researchers (Y.W.Y. and L.L.). Kappa statistics showed a high agreement between reviewers for study selection (k = 0.71, P < 0.01). Discrepancies were resolved by discussion.

Quality assessment

We used the critical appraisal checklist of case–control studies from Specialist Unit for Review Evidence (SURE; http://www.cardiff.ac.uk/insrv/libraries/sure/checklists.html) to assess the quality of included studies. This is a 11-item checklist that covers key potential sources of bias proposed by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline [42].

Data extraction

Data extraction included: (1) basic demographics of participants (e.g. sample sizes and age information); (2) diagnostic tools for assessing IGD; and (3) decision-making tasks and related dependent variables. If multiple dependent variables were reported for a task, one widely used measure (e.g. net score for IGT) was selected for that task based on the available data. The selected variables for the included studies are listed in Table 1.

Table 1. Studies included in the meta-analysis. Study Sample ID Sample size Mean age (SD) IGD diagnosis Task Situation DVs Valence Study type Sample type Comorbidity control Deleuze et al. 2017 [43] 1 IGD = 32 IGD = 21.84 (3.21) DSM-5 GDT Risky Net score Mixed Behavioral Community No HC = 65 HC = 22.38 (3.97) Dong et al. 2016 [44] 2 IGD = 20 IGD = 21.33 (2.18) DSM-5 + YIAT Risk DM Risky Disadvantageous choice % Mixed fMRI Community Yes (MINI) HC = 16 HC = 21.9 (2.33) Jiang et al. 2020 [15] 3 IGD = 30 IGD = 22.00 (5.00) DSM-5 IGT Ambiguous Net score Mixed Behavioral Clinical Yes (DSM-5) HC = 30 HC = 22.00 (6.00) Kim et al. 2018 [45] 4 IGD = 18 IGD = 22.17 (2.00) YIAT RL Ambiguous Advantageous choice % Gain + loss fMRI Community Yes (self-report) HC = 20 HC = 21.20 (2.20) Ko et al. 2017 [35] 5 IGD = 87 IGD = 23.29 (2.38) DSM-5 Risk DM Risky Risky choice % Gain + loss Behavioral Community Yes (MINI) HC = 87 HC = 23.38 (2.40) Li et al. 2020 [46] 6 IGD = 31 IGD = 15.81 (1.68) DSM-5 Risk DM Risky Risky choice % Mixed EEG Clinical Yes (DSM-4, BDI, BAI) HC = 32 HC = 15.91 (1.73) Tian et al. 2018 [37] 6 IGD = 35 IGD = 15.57 (1.17) DSM-5 1. DDT 1. Inter-temporal 1. AUC 1. Gain + loss Behavioral Clinical Yes (DSM-4, BDI, BAI) HC = 38 HC = 15.78 (0.94) 2. PDT 2. Risky 2. AUC 2. Gain + loss Wang et al. 2020 [34] 6 IGD = 45 IGD = 15.58 (1.14) DSM-5 + YDQ + gaming time Mixed gambles task Risky Loss aversion (lambda) Mixed Behavioral Clinical Yes (MINI, BDI, BAI) HC = 43 HC = 15.72 (0.96) Lin et al. 2019 [18] 7 IGD = 23 IGD = 25.39 (2.04) DSM-5 IGT Ambiguous Net score Mixed Behavioral Community No HC = 38 HC = 25.66 (2.22) Lin et al. 2015 [47] 8 IGD = 19 IGD = 22.20 (3.08) YIAT PDT Risky Logged h Gain fMRI Community Yes (MINI) HC = 21 HC = 22.80 (2.35) Wang et al. 2017a [48] 8 IGD = 18 IGD = 22.10 (3.20) YIAT + gaming time DDT Inter-temporal Logged k Gain fMRI Community Yes (MINI) HC = 21 HC = 23.10 (2.00) Liu et al. 2017 [49] 9 IGD = 41 IGD = 21.93 (1.88) CIAS + gaming time Cups Task Risky Risky choice % Gain + loss fMRI Community Yes (BDI, BAI) HC = 27 HC = 22.74 (2.35) Metcalf et al. 2014 [19] 10 IGD = 10 NA AEQ IGT Ambiguous Net score Mixed Behavioral Community No HC = 13 Park et al. 2020 [50] 11 IGD = 34 IGD = 25.90 (6.00) DSM-5 + gaming time CGT Risky Risky choice % Mixed Behavioral Clinical Yes (DSM-5) HC = 34 HC = 25.50 (4.30) Pawlikowski et al. 2011 [16] 12 IGD = 19 IGD = 23.47 (3.88) YIAT GDT Risky Net score Mixed Behavioral Community Yes (self-report) HC = 19 HC = 24.32 (3.62) Qi et al. 2016 [51] 13 IGD = 24 IGD = 17.17 (3.51) DSM-5 + YIAT + gaming time BART Ambiguous Adjusted pumps Mixed fMRI Community Yes (MINI) HC = 24 HC = 17.42 (3.05) Wang et al. 2017b [52] 14 IGD = 20 IGD = 20.95 (2.44) DSM-5 + YIAT + gaming time 1. DDT 1. Inter-temporal 1. Logged k 1. Gain fMRI Community Yes (self-report) HC = 20 HC = 21.95 (2.37) 2. PDT 2. Risky 2. Logged h 2. Gain Wölfling et al. 2020 [53] 15 IGD = 30 IGD = 26.90 (5.97) DSM-5 1. DDT 1. Inter-temporal 1. AUC 1. Gain Behavioral Clinical Yes (clinical interview) HC = 27 HC = 25.60 (3.25) 2. IGT 2. Ambiguous 2. Total balance 2. Mixed Wu et al. 2018 [54] 16 IGD = 22 IGD = 21.40 (1.30) DSM-5 + YIAT + gaming time Slot machine task Ambiguous Risky choice in persistence phase Gain Behavioral Community Yes (self-report) HC = 22 HC = 22.00 (1.70) Yao et al. 2017 [36] 17 IGD = 25 IGD = 22.28 (1.62) DSM-5 + gaming time 1. DDT 1. Inter-temporal 1. Logged k 1. Gain Behavioral Community Yes (MINI) HC = 21 HC = 22.00 (2.26) 2. BART 2. Ambiguous 2. Adjusted pumps 2. Mixed Yao et al. 2014 [17] 18 IGD = 26 IGD = 22.54 (2.10) CIAS + gaming time 1. GDT 1. Risky 1. Net score 1. Mixed Behavioral Community Yes (self-report) HC = 26 HC = 22.00 (2.15) 2. Modified GDT 2. Risky 2. Net score 2. Mixed Yao et al. 2015a [20] 19 IGD = 34 IGD = 22.29 (2.07) CIAS + gaming time IGT Ambiguous Net score Mixed Behavioral Community Yes (self-report) HC = 32 HC = 22.47 (2.08) Yao et al. 2015b [23] 19 IGD = 60 IGD = 22.40 (2.07) CIAS + gaming time Cups Task Risky Risky choice % Gain + loss Behavioral Community Yes (self-report) HC = 42 HC = 22.38 (2.10) Zha et al. 2019 [55] 20 IGD = 19 IGD = 20.50 (3.10) DSM-5 DDT Inter-temporal Logged k Gain fMRI Community Yes (self-report) HC = 46 HC = 23.80 (1.70) AEQ = Addiction Engagement Questionnaire; AUC = area under the curve; BAI = Beck Anxiety Inventory; BART = Balloon Analog Risk Task; BDI = Beck Depression Inventory; CGT = Cambridge Gambling Task; CIAS = Chen Internet Addiction Scale; DDT = Delay Discounting Task; DM = decision-making; DSM = Diagnostic and Statistical Manual of Mental Disorders; DV = dependent variable; EEG = electroencephalogram; fMRI = functional magnetic resonance imaging; GDT = Game of Dice Task; HC = healthy control; IGD = Internet Gaming Disorder; IGT = Iowa Gambling Task; MINI = Mini-International Neuropsychiatric Interview; PDT = Probabilistic Discounting Task; RL = Reinforcement learning; SD = standard deviation; YIAT = Young's Internet Addiction Test.

The standardized ES for each variable was calculated based on Hedge's g method, which is considered to be

留言 (0)

沒有登入
gif