CT, also called childhood maltreatment, adverse childhood experience (Zhang et al., 2020a,b), including physical abuse, emotional abuse, sexual abuse, neglect and other forms of harm that occurs in children under the age of 18 (World Health Organization, 2022). The research on CT originated from people’s concern about sexual abuse. Russell defined sexual abuse as one or more unwanted sexual experiences between an individual under 18 years of age and a related or unrelated person (Russell, 1983). Based on the definition of Sexual Abuse, Russell formulated the Sexual Abuse Questionnaire by asking 14 interview questions. Later, the meaning of CT became richer than before, including psychological abuse, physical abuse, neglect and other new connotations in addition to sexual abuse, resulting in a variety of tools for measuring. Childhood Trauma Questionnaire (CTQ) has a good effect among 52 CT questionnaires (Saini et al., 2019). The CTQ, including 70 items in four dimensions: physical, emotional abuse, emotional neglect, sexual abuse and physical neglect, was developed by Bernstein and colleagues to assess self-reported experiences of abuse and neglect during childhood and adolescence (Bernstein and Fink, 1998). In order to detect the CT condition of normal population and clinical population in a short time, Bernstein et al. reduced and adapted the initial CTQ items, divided the dimension “physical and emotional abuse” into “physical abuse” and “emotional abuse,” changed the original four rotated factors into five (Bernstein et al., 2003). The Childhood Trauma Questionnaire-Short Form (CTQ-SF) with only 28 questions has been developed, which has been widely verified in different groups (Bernstein et al., 1997; Garrusi and Nakhaee, 2009; Wingenfeld et al., 2010). Meng and Liu (1994) made the earliest research on CT in mainland China. They published a research review related to child abuse in 1994, and since then the issue of CT has been widely concerned by Chinese researchers (Deng, 2000; Li et al., 2004; Yang et al., 2004; Pan and Li, 2005). Among them, the earliest Questionnaire of CT was The Screen Questionnaire of Child Abuse, developed by Yang et al. (2004), according to the WHO definition of child abuse, the questionnaire included eight items, including verbal abuse, economic control, isolation, neglect, punching, kicking, biting and slapping, using knife and\or stick, and sexual assault. In the study of the relationship between CT and PIU, the Chinese version revised according to CTQ-SF by Zhao et al. is most commonly used by scholars in mainland China (Zhao et al., 2005). In addition, the Child Psychological Abuse and Neglect Scale (CPANS) compiled by Deng, a Chinese mainland scholar, was often used to detect CT (Liu, 2016; Yang et al., 2017; Du, 2020; Zeng, 2020). This scale named CPANS, based on the different definitions and types of psychological abuse and neglect in literature, contains psychological abuse sub-scales (scold, intimidation, interference) and ignore sub-scales (emotional neglect, education, body/supervision ignore). It is a five-point scale, from 0 (none) to 4 (always), Cronbach’s coefficient is above 0.8 (Deng et al., 2007).

Since the American psychiatrist Goldberg (1995) first proposed the term Internet addiction disorder (IAD) in 1995, scholars have carried out a series of exploratory studies in this field, and PIU measurement tools have been constantly emerging (Rial et al., 2015). In mainland China, the most widely used measurement tool for PIU research are Internet Addiction Test (IAT; Young, 2001), the Revised Chinese Internet Addiction Scale (CIAS-R; Chen et al., 2003), and Adolescent Pathological Internet Use Scale (APIUS; Lei and Yang, 2007). Young believed that the diagnostic criteria for pathological gambling were most closely related to the pathological features of PIU. Therefore, the DSM-IV pathological gambling criteria were revised to form IAT to measure PIU, including 20-item covering the degree to which their Internet use affects their daily routine, social life, productivity, sleeping pattern, and feelings, items are rated on a 5-point Likert scale, and the higher the score, the higher the degree of addiction (Widyanto and McMurran, 2004). CIAS-R is a self-rating scale, which was compiled by Chen et al. (2003), scholars from Taiwan, China, on the theoretical basis of the core symptoms of Internet overuse and the problems related to Internet overuse. The full scale contains five dimensions: compulsive Internet use, withdrawal behavior and withdrawal response, PIU tolerance, time management problems, interpersonal and health problems, a total of 26 items. The total score reflects the degree of personal Internet addiction, and the higher the total score, the higher the PIU tendency (Mak et al., 2014), the internal consistency coefficient of each factor in this measure is between 0.70 and 0.82, and the internal consistency coefficient of the whole scale is 0.92. APIUS was formulated by Lei and Yang, scholars from Mainland China, according to the definition and dimensions of PIU and with reference to the existing items of related scales. The full scale includes six dimensions (salient; tolerance; compulsive Internet use/withdrawal symptoms; mood change; social comfort; negative consequences) with 38 items. It is a five-point scale, the higher the score, the more serious the PIU. Additionally, some studies also use self-designed scales to measure PIU, such as Different Types of Internet Addiction Scale for Undergraduates, and An Internet Relationship Dependency Inventory (Qian et al., 2006; Zhou and Yang, 2006). Different tools have different theoretical bases and dimensions, so the use of different measurement tools to explore the relationship between CT and PIU may be prone to bias in a certain extent. Therefore, this study took measurement tools as moderating variables affecting the two for further analysis.

This study followed the requirements of Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) statement for literature screening (Moher D et al., 2009). We used databases to search the related studies on CT and PIU in Mainland China from January 2009 to November 2022, including CNKI, Wanfang Data, Chongqing VIP Information Co., Ltd. (VIP), Baidu scholar, ProQuest dissertations, SAGE Online Journals, Elsevier SDOL, Taylor & Francis, Springer, Web of Science, Google Scholar, EBSCO, Medline, Scopus Database, PubMed Central, Embase, The Cochrane Library, PsycINFO. The main search terms for CT are “childhood abuse” OR “childhood maltreatment” OR “childhood trauma” OR “childhood sexual abuse” OR “childhood physical abuse” OR “childhood psychological abuse” OR “childhood emotional abuse.” The search terms for PIU are “Internet overuse” OR “problematic Internet use” OR “Internet addiction” OR “Internet dependence” OR “excessive Internet use” OR “compulsive Internet use” OR “Internet addiction disorder” OR “pathological Internet use.”

Inclusion criteria: (1) Only cross-sectional studies were included, longitudinal studies were not included; (2) well-defined studies related to CT and PIU were included, complex trauma studies were not included; (3) only studies that use CT or PIU scales as research tools were included; (4) the empirical quantitative studies includes the Pearson product–moment correlation coefficient or the t-value and f-value that could be converted into r were reported; (5) accurate sample size was reported; (6) the subjects in the literature are Chinese mainland students, including primary school students, junior high school students, senior high school students, higher vocational students and university students, excluding students from Hong Kong, China; Macau, China; Taiwan, China. (7) self-assessment studies; (8) studies written in English or Chinese; (9) when the data is published repeatedly, only one professional academic journal is taken.

Exclusion criteria: (1) Incomplete reports of important data; (2) samples’ cognitive and judgment abilities were normal, and studies in which samples were patients. Excluding papers with no data, repeated publication and no clear sample size, a total of 31 papers met the meta-analysis criteria. Figure 1 depicts the PRISMA flow chart of the systematic search.

The features of collected literature were coded, including the author’s name, publication time, survey’s year, source of publication, regional distribution, grade subjects, sample size and the percentage of the total number of girls, measuring tools, correlation coefficient of CT and PIU (see Table 1). Effect values were extracted according to the following criteria: (1) The correlation coefficient between CT and PIU was included in the coding; (2) Independent samples were coded once. If multiple independent samples were reported at the same time, they were coded separately; (3) When calculating the effect size of each category, each original datum appeared only once in each category to ensure the independence of the effect value calculation.

The meta-analysis method of Person product difference correlation coefficient r was used to calculate the effect size in this study value (Borenstein et al., 2009). Fisher’s Z transformation was applied to r, and weights and 95% confidence intervals were calculated based on the sample size. Conversion formula: Zr = 0.5*ln[(1 + r)/(1 − r)], VZ = 1/n − 3, SEz = sqrt(1/n − 3), where Zr represents the converted value of the corresponding r, VZ is the variance, and SEz is the standard error.

For testing whether study result represents a estimated sample of the total effect size, a homogeneity test is required. Firstly, the homogeneity test provides a basis for whether the outcome adopts a fixed effect model or a random effect model. If the test results showed that the effect values were homogeneous, the fixed effect model was selected. If the heterogeneity is considerable, the random effect model should be used. Secondly, the homogeneity test also provides the basis for the analysis of the moderating effect, and the large heterogeneity indicates the existence of the moderating effect (Lipsey and Wilson, 2001).

Among the papers included in the meta-analysis in this study, 31 papers reflected the relationship between CT and PIU, involving 52,503 subjects, and the sample size ranged from 388 to 16,130. Table 2 shows the homogeneity test of CT and PIU in 31 independent samples, with Q statistic value of 1,499.887, p < 0.001, I² = 98.000, indicating the heterogeneity of the included literature. This may be due to the use of different measurement tools, sources of subjects and different sample sizes in the literature, that is, there may be a moderating effect. According to the method provided by Lipsey and Wilson (2001), the included papers are highly heterogeneous and must be analyzed by random models.

A random model was used to analyze the correlation between CT and PIU, and it was found that CT was significantly correlated with PIU, with a correlation coefficient of 0.281, 95%CI[0.221, 0.338]. This relationship between CT and PIU can be considered as a moderate correlation (Lipsey and Wilson, 2001). The Z-value of the relationship between CT and PIU was 8.846, p < 0.001, indicating that the relationship between CT and PIU was stable.

As mentioned above, random effects models should also be used in mediating effects analysis. Meta-ANOVA analysis is suitable for analyzing the moderating effects of categorical variables, such as the type of measurement tools, source of literature publication, regional distribution, and grade of subjects. In contrast, meta-regression analysis is suitable for analyzing the moderating effect of continuous variables, such as proportion of females and survey’s year.

In order to analyze the moderating effect of the relationship between CT and PIU, Meta-ANOVA analysis was used to analyze the moderating effect of categorical variables (Table 3). In terms of region, the results of homogeneity test (Q = 0.111, df = 1, p > 0.05) showed that region did not have a moderating effect on this correlation, and the relationship between CT and PIU was not affected by region.

In terms of grade, the results of homogeneity test (Q = 0.769, df = 1, p > 0.05) showed that grade did not have a moderating effect on this correlation, and the relationship between CT and PIU was not affected by grade.

As for the measurement tools of CT, the results of homogeneity test (Q = 2.482, df = 2, p > 0.05) showed that there was no moderating effect of the measurement tools of CT on this correlation, and the relationship between CT and PIU was not affected by the measurement tools of CT.

As for the measurement tools of PIU, the results of homogeneity test (Q = 8.044, df = 3, p < 0.05) showed that the measurement tools of PIU had moderating effect on the correlation, and the relationship between CT and PIU was affected by the measurement tools of PIU.

In terms of publication source, the results of homogeneity test (Q = 0.023, df = 1, p > 0.05) showed that there was no moderating effect of publication source on this correlation, and the relationship between CT and PIU was not affected by publication source.

To examine whether continuous variables (gender and survey’s year) moderate the effect sizes between CT and PIU, the r effect size was meta-regressed onto the percentage of female participants and survey’s year in each sample. In Table 4, meta-regression (Q_{Model [1, k = 31]} = 0.03, p > 0.05) demonstrated that there was no moderating effect of gender in the relationship between CT and PIU. Meta-regression (Q_{Model [1, k = 31]} = 4.43, p < 0.05) demonstrated that the relation between CT and PIU was moderated by survey’s year. It means with the increase of year, the correlation coefficient between CT and PIU also increases.

To examine whether the results were biased due to effect sizes from various sources, a funnel plot was drawn, indicating that the 31 effect sizes were symmetrically distributed on both sides of the average effect size, and an Egger’s regression (Egger et al., 1997) revealed no significant bias (t₍₂₉₎ = 1.173, p = 0.250 > 0.05). This result showed that in the overall correlation between CT and PIU was stable in this study (Figure 2).

Based on 52,503 students in mainland China from 31 articles, the results of a meta-analysis showed that a moderate Holmanpositive correlation between CT and PIU, consistent with previous research results (Hou et al., 2021a; Qin et al., 2022; Zhang, 2022), it means that students who have experienced CT are more likely to suffer from PIU (Sharifi, 2019; Li, 2020). It supports the Interaction of Person-Affect-Cognition-Execution (I-PACE) model (Brand et al., 2016), according to this model, person’s core characteristics, including biological genetic and further bio-psychological factors, are the predisposing factors for the formation of PIU. In addition, moderating variables (including the individual’s subjective perception of the situation, poor coping style, as well as cognitive biases on the Internet, etc.), will lead to the desire of individuals to use the Internet, and then weaken the inhibitory control and the executive function, resulting in the decision to use the Internet. When individuals gain satisfaction in this process, it will strengthen the Internet use behavior, and then lead to PIU (Brand et al., 2016). It can be said that CT belongs to further bio-psychological factors, which is a predisposing factor to the formation of PIU. Since CT groups are more likely to perceive stress (McLaughlin et al., 2009; Edalati et al., 2019), this makes CT students unable to perceive the passage of time in the process of using the Internet, resulting in temporal dissociation and cognitive absorption (Drakulić et al., 2018; Cannito et al., 2022; Holman et al., 2022), thus, the cognitive process changes resulting in dysfunction, impulsivity and other adverse coping styles (Young et al., 2007; Perlman et al., 2016; Wang S. et al., 2020). Additionally, various factors such as “Internet related cognitive bias” (i.e., happiness or escapism through using the Internet) generated in the process of using the Internet become the moderating variables for forming PIU. When the CT people have problems such as anxiety, depression, aggression, social skills deficit and poor interpersonal relationships (Romano et al., 2014), the characteristics of Anonymity, Convenience and Escape of the Internet (Young, 1997) will produce positive associations and trigger their desire to use the Internet. The emergence of desire will weaken their own executive power and control, and make the decision about using the Internet. When the use of the Internet is satisfied and compensated, this process will be strengthened, increasing the frequency of Internet use, and then PIU will emerge. Therefore, individuals with traumatic childhood experiences are more likely to suffer from PIU.