Background
Internet addiction (IA), initially reported by Young [
1], is considered as a new psychiatric disorder, but IA was still not listed as a clinical entity in the fifth edition of Diagnostic and Statistical Manual of Mental Disorders (DSM-5). People using Internet excessively and pathologically might suffer from adverse consequences, including arguments, fatigue, lying, poor grading in school or vocational achievement during working, social isolation and even functional problems such as school failure, job loss and marriage failure [
2]. The pathway from adaptive Internet use to IA is very complicate and ambiguous, which could be affected by many different factors including both individual and environment. It’s reported that IA was prevalent in both eastern and western countries. Because of the different questionnaires, diagnostic criteria used, the prevalence of IA in different areas with different culture is different. It’s reported that the prevalence of IA ranges from 1% to 36.7% in a literature review [
3]. Given the large scale of Internet usage and so many negative consequences, it is important to untangle the potential risks associated with IA.
In a previously published systematic review about the association between IA and Attention Deficit/Hyperactivity Disorder (ADHD), positive correlations were confirmed after controlling covariates [
4]. A 2-year prospective study found that adolescents diagnosed as ADHD were the most likely to be addicted to the Internet than other psychiatric symptoms such as hostility and social phobia [
5]. However, it remains a matter of debate that if there were indeed any causalities between IA and ADHD., and the association could be explained from different aspects. For instance, in the biopsychosocial model, “being easily bored” and “having an aversion for delayed rewards” are two main ADHD symptoms [
6,
7]. Internet use provides multiple windows with a variety of activities at the same time and immediate reward may decrease the boredom feeling and reward quickly, which makes people with ADHD addicted to Internet more easily. Furthermore, some researchers also found that subjects with ADHD have abnormal brain activities that would lead to impaired inhibition, which results in lack of self-control ability, so that Internet users would become more unable to restrain themselves and vulnerable to IA. Thus, ADHD could be a possible risk factor that may lead to IA.
Two well-established systematic reviews have summarized relevant articles on the relation of IA and psychiatric comorbidities, but their conclusions regarding the association between IA and ADHD were hampered by the some methodological deficiencies and paucity of included studies, only five and four observational studies that reported odds ratio (OR) were included in two aforementioned studies respectively [
4,
8]. The drawn conclusions were based upon ORs of unadjusted results, which could weaken the reliability of pooled results. As new evidence is emerging in recent years, it is necessary to perform an updated meta-analysis to reevaluate the association between IA and ADHD. Furthermore, our present study aims to assess the influence of IA on symptoms of ADHD, and to clarify the relationships between IA and demographic characteristics of enrolled participants.
Methods
This systematic review was conducted in accordance to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [
9].
Literature search
A comprehensive electronic literature search was conducted by using following databases: CENTRAL, EMBASE, PubMed and PsychINFO. Relevant articles published from inception to June 2016 were searched in databases above by two reviewers (B.Q. Wang and N.Q. Yao) independently and no language restriction was imposed. Free text words and Medical Subject Headings (MeSH) were employed as search terms independently or in combination of according to specifications of each database.
The following searching strategy was utilized: (Internet addiction or problematic Internet use or Internet addiction disorder or pathological Internet use or Internet game addiction or excessive Internet use or compulsive Internet use or Internet dependency or computer addiction) and (“Attention Deficit Disorder with Hyperactivity”[Mesh] or ADHD or ADDH or Attention Deficit Disorder with Hyperactivity or Attention Deficit-Hyperactivity Disorder or Hyperkinetic Syndrome or Attention Deficit Hyperactivity Disorder or Attention Deficit Disorder). The bibliographies of relevant systematic reviews and clinical guidelines were manually searched. References from each retrieved papers were also manually searched.
Types of participants
Patients diagnosed with IA by a standard criterion were recruited in IA group. The tools employed for the assessment of IA included CIAS [
10], IAT [
11], IAS [
12], along with other well-established tools. No restriction on age, race and gender was imposed.
Types of control
Subjects without IA were included without other restrictions.
Outcome measures
The primary outcome was adjusted odds ratio (AOR), secondary outcome measures included crude odds ratio (COR) and parameters assessing the severity of symptoms of ADHD. COR should be reported by included studies or could be calculated based upon raw data.
Types of included studies
Observational studies including cohort studies, case-control studies and cross-sectional studies without restrictions on geographic area or sample size were included.
Exclusion criteria
Case series, case repots, book chapters, editorials and papers of conferences were excluded. Studies failed to report the diagnostic criteria of IA were excluded. Studies on pathologic internet use such as spending time and time to sleep but without a specific definition of IA were also excluded. We also excluded case series, case reports and articles only studied the brain imaging, electroencephalogram (EEG), treatments, intervention or other related symptoms such as impulsivity, lifestyle and sexual attitude but not studying the relationship with ADHD. Studies with abstracts written in English language but with full-text in other languages were excluded. In addition, articles with only abstracts were also removed because detailed data could not be obtained so the methodologic quality of them could not be assessed.
Two investigators (B.Q. Wang and N.Q. Yao) individually reviewed each article and were blinded to the process and outcomes of each other. According to the inclusion criteria defined above, we implemented a strict screening to include articles with the eligibility. Data was also collected independently from these selected articles using the same collection form including first author, country, year of the publication, study design, source of cases, sample size in each group, mean age of all the enrolled subjects, definition of IA and definition of ADHD, prevalence of ADHD in each group and scales used to assess the symptoms of ADHD. Any disagreement between the two reviewers was resolved through discussion until a consensus was reached. The third review author (Z.T. Lv) was consulted if an agreement could not be achieved.
Methodological quality assessment
The Newcastle-Ottawa Scale (NOS) [
13] and an adapted form of NOS [
14] were utilized to assess the methodological quality of non-randomized studies in this systematic review. Two reviewers assessed the methodological qualities of each study independently, the results were compared afterwards.
Data synthesis and analysis
OR and the associated 95% confidence interval (CI) in each included studies were combined in order to assess the possible association between IA and ADHD. The standardized mean difference (SMD) as well as the associated 95% CI was both calculated for severity of ADHD and combined using the same method. Prevalence of ADHD in IA groups was also combined, and stratified analysis was made by two age groups. As included studies measured the outcomes using different scales, the random-effect model was used to conduct the statistical analysis. Heterogeneity between studies was assessed by the Higgins I2 test (P > 0.1 and I2 < 50% indicate acceptable heterogeneity) and a standard chi-square test. And the heterogeneity outcomes showing P > 0.1 and I2 < 50% could be acceptable.
Meta-regression analyses on age (≥18 years and <18 years), ethnicity (Asian and European) and risk of bias (high, medium or low risk of bias) was implemented using Stata version 12.0 (Stata Corp LP, USA) to identify the probable cause of heterogeneity. Sensitivity analysis by removing each related study at a time was also made to evaluate the impact of each study on the pooled OR and the severity of ADHD symptoms. Begg’s rank correlation test and Egger’s linear regression test were used to evaluating the publication bias. The forest plot was made by RevMan 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).
The effect size of association was expressed as small, moderate and large according to Cohen. OR was converted into these groups according to Chinn [
15]. Cohen’s d was calculated based on original data from studies that did not provide OR. Thus, the effect sizes were explained as follow: small (Cohen’s d = 0.2, OR = 1.44), moderate (Cohen’s d = 0.5, OR = 2.48) and large (Cohen’s d = 0.8, OR = 4.27).
Discussion
In summary, the finding of our present study suggested a positive association between IA and ADHD even after controlling confounding factors, symptoms of ADHD in IA groups were more severe than control groups. Male adolescents and young adults were more likely to be diagnosed with IA, but age pattern was not positively associated with IA in our included studies. Evidence support a causal relation between IA and ADHD is still lacking.
To the best of our knowledge, this is the first meta-analysis to individually investigate the association between IA and ADHD with consideration of heterogeneity while the previously published articles were systematic reviews about IA and several psychiatric co-morbidities [
4,
8] or narrative literature review about the association between Internet gaming disorder and ADHD [
40]. In two previously published systematic reviews, only few studies reporting OR as outcome were included for meta-analysis. The extracted ORs were pooled without differentiating COR and AOR, which might lead to an exaggeration or underestimation of the correlations between IA and ADHD. Thus, the conclusions drawn by aforementioned studies should be interpreted with caution. The results of our current work were powered by sufficient number of included studies and rigorous methodological quality assessment by independent reviewers. Both dichotomous variables and continuous variables were taken into consideration by our study, which greatly fill the blank of now-existing literature. In addition, the magnitude of effect size across studies was compared according to Cohen.
ADHD patients have poor self-control ability, so they’re more easily to sustain an addiction to substances as well as Internet. But studies have reported that striatal dopamine could help game users focus and gain better performance while playing Internet games [
41], which let ADHD patients compensate for the failure in real-life and prefer into the virtual world. Compared with real life, Internet users would get response, reward and establish interpersonal relationships more easily online. Our results demonstrated that patients with IA were present with more severe symptoms of ADHD than healthy control, so that IA may also have influence on ADHD. Ko and colleagues reported that ADHD could predict the occurrence of IA in the 2-year follow-up. Chen et al. [
16] also reported that high ADHD symptoms were significantly associated with the occurrence of IA. In summary, IA and ADHD may interact with each other. However, evidence supporting a causality between IA and ADHD is still lacking, only two included studies were based on a prospective design. The causality between both entities is still a matter of debate.
The majority of our included studies suggested moderate and strong associations between IA and ADHD. The obviously observed heterogeneity (I2 = 68%) in combination of COR suggested that demographic factors and other social or family factors could possibly affect the association, this hypothesis was partially verified by the low heterogeneity in pooled AOR (I2 = 8%). We further undertook meta-regression by age, ethnicity and methodological quality to determine whether they contributed to the heterogeneity in symptoms score, methodological quality of included studies was found to be associated with heterogeneity across studies. Thus, prospective cohort studies with high methodological quality are required.
The risk of addiction to Internet was higher among males than females, which might be explained by two reasons. Firstly, more males than females tend to seek self-esteem feelings and make social contacts online. Secondly, girls may receive more close supervision regarding internet use than boys in a family. Studies found that inattention was the most associated symptoms of ADHD among young adults and more significant in female adults.
Except for gender difference, other factors for IA such as low family support, protective parenting style, poor grading in school, bad interpersonal ability [
42] might predict IA. These predicting factors found could be specifically targeted when designing the prevention program for IA among the children and adolescent population. On the other hand, these factors might be confounders that could affect the association between IA and ADHD, which should be controlled in more prospective cohort studies in the future. Studies showed that 65% children having ADHD during childhood had persistent ADHD symptoms till their adulthood [
43]. Adult ADHD brings many negative effects but is seldom known by public [
44]. Thus, ADHD patients should be very significant target group for the prevention for IA.
An interesting phenomenon observed in our study is that most included studies were performed in Asian countries. It was reported in a study of Zhang et al. [
45] that IA was more prevalent in some Asian countries than in the United States. A possible reason might be the differences in sociocultural background [
46]. Unlike in Asia, where Internet cafés are easily accessible and frequently used, in the US games and virtual sex are accessed from the home. Furthermore, attempts to evaluate the phenomenon are impeded by shame, denial, and minimization [
47]. However, this explanation should be confirmed in further studies.
Given the results found in the current study, problems of IA, ADHD as well as the comorbidity conditions are required to get more concern of public health. Government should look for effective prevention policies and strategies to reduce the related health risks and negative outcomes. First, IA was identified as an emerging public health issue in both South Korea and China, as well as Taiwan. But until now, the scope of a universal definition of IA is still absent. A standard terminology as well as diagnostic criterion should be established and the cross-cultural validity should be examined to enable international comparisons. Secondly, according to the results of our current study, age was not associated with IA, which suggested that both adolescents and young adults should be targeted for the IA prevention. Last but not least, until now, although the significant relationship between IA and ADHD was proved, whether the ADHD is the risk factor of IA or a comorbidity is still unknown. It’s suggested that the comorbidity disease should be cautiously screened if diagnosed one of them. ADHD symptoms should be carefully prevented and early identified among at-risk subjects and their families by effective strategies.
There were several limitations in our study. Firstly, only two prospective cohort studies conducted in Taiwan were included in our study [
16], definite conclusion about the causal relationship between ADHD and IA could not be drawn. Secondly, homogeneous geographic distribution and lack of a universal diagnostic criterion for IA are two primary problems still remaining. The prevalence of IA is inconsistent across included studies, except for cultural reasons and sample selection, the varying questionnaires and thresholds employed may contribute to the variations in these results. Seven of our included studies used CIAS criteria to diagnose IA with the cut-off point of 64 (accuracy 87.6%, specificity 92.6%). But to what extend could different questionnaires and thresholds employed influence the prevalence of IA could not be addressed by our current study. More studies need to be carried out in other geographic areas of the world for comparisons among different cultures, when using unified diagnostic criteria. Lastly, many of our included studies had recruitment bias because the method of sampling was highly selective, the conclusions could not be generalized to community population. Within the fifteen included studies, eight were evaluated with high risk of bias and seven had medium risk of bias. Thus the results should be interpreted cautiously.