The PRISMA guidelines for systematic reviews and meta-analyses were adhered to in the conduct of this study (41). The research protocol was officially registered in PROSPERO under the registration number CRD42024577593. It was noted that the PRISMA guidelines, recognized internationally as a standard, were provided as a comprehensive and rigorous framework for systematic reviews and meta-analyses. Meanwhile, the PROSPERO registration platform was utilized to ensure the transparency, reproducibility, and scientific integrity of the research protocol. This platform was used to help reduce research bias and to enable other researchers to track and evaluate the progress and quality of the study more effectively. Through PROSPERO registration, the design and methods of this study were clearly documented and made publicly accessible, thereby laying a solid foundation for subsequent research communication and application.

To explore the relationship between PIU and suicidal behavior, we searched seven databases (PubMed, Web of Science, EMBASE, CNKI, Wanfang Data, VIP Database, and CBM) from their establishment to July 1, 2024. In the search, we used detailed methods. In titles, abstracts, and subject terms, we carefully selected keywords divided into three groups: (“Suicide”, “Problematic Internet Use”, and “Adolescents”). For “Suicide”, we included common and specific terms like “suicide”, “self-harm”, “attempted suicide”, “suicidal behav*”, “suicidal ideation”, etc. For “Problematic Internet Use”, we covered expressions like “excessive internet use”, “internet addiction”, “problematic online behavior”, “compulsive internet use”, “maladaptive internet usage”, etc. For “Adolescents”, we included common and specific terms like “adolescents”, “teenagers”, “young people”, “youth”, “middle school students”, “college students”, etc. To ensure thoroughness, reference lists of the included studies were manually reviewed to identify any additional relevant articles. Additionally, manual cross-referencing of citations was performed to ensure thoroughness. The entire search strategy was meticulously documented to guarantee reproducibility. Further details regarding the search methodology can be found in Supplementary Material .

Inclusion criteria for the meta-analysis were as follows (1): Population: Studies that focused on Chinese adolescents aged 10 to 24 years (42) (2). Exposure: Research that specifically examined the association between PIU and suicidal behavior (3). Outcomes: Studies that reported relevant outcomes, including SI, SP, SA, with corresponding statistical measures such as Odds Ratio (OR) (4). Study Design: Peer-reviewed articles employing cross-sectional, case-control, or cohort study designs (5). Language: Articles published in either Chinese or English (6). Methodological Rigor: Studies that provided sufficient detail on their methodology, allowing for assessment of bias and overall study quality.

Exclusion criteria were defined as (1): Population: Studies that involved non-Chinese adolescents or included participants outside the age range of 10 to 24 years (2). Exposure: Research that focused on general internet use, other behavioral addictions, or any exposure not directly related to PIU and its association with suicidal behavior (3). Outcomes: Articles that did not report extractable data on suicidal behaviors or provided insufficient statistical information to calculate effect sizes (4). Study Design: Research formats such as reviews, systematic reviews, meta-analyses, conference abstracts, editorials, letters, comments, or case reports were excluded. Studies with a high risk of bias or those lacking peer review were also excluded.

Data extraction was conducted systematically to ensure consistency and accuracy across all included studies. A standardized data extraction form was developed and utilized by two independent reviewers (XBH, CS) to collect essential information, including study characteristics (author(s), year of publication, and study location), population details (sample size, age range, gender distribution, and specific demographic characteristics), exposure variables (definitions and measures of PIU, including criteria, scales, or questionnaires used), and outcome variables related to suicidal behavior (SI, SP, and SA), with corresponding OR and 95% CI where available. Adjustments for potential confounding factors, such as depression, were also noted, and the key findings of each study, particularly those related to the association between PIU and suicidal behavior, were summarized. The initial collection of literature, as well as the identification and removal of duplicate records, was performed using NoteExpress software. To ensure accuracy, all extracted data were cross-checked by a third reviewer (PY), with any discrepancies between the two primary reviewers resolved through discussion and consensus. In cases where consensus could not be reached, a senior researcher (BY) was consulted. The extracted data were then compiled into a comprehensive database for subsequent analysis.

The quality of the included studies was evaluated using two established tools: the Agency for Healthcare Research and Quality (AHRQ) methodology checklist and the Newcastle-Ottawa Scale (NOS). For cross-sectional studies, the AHRQ tool assessed 11 items, including population definition, sampling method, response rate, variable measurement, and data analysis. Each item was scored as “yes” (1 point), “no” (0 points), or “unclear” (0 points), with a maximum score of 11. Studies scoring 8 or above were rated as high quality, 4 to 7 as moderate quality, and below 4 as low quality. For this meta-analysis, only studies with a score of 4 or higher were included (43). For cohort and case-control studies, the NOS was used, which evaluated studies on three criteria: selection (up to 4 stars), comparability (up to 2 stars), and outcome/exposure ascertainment (up to 3 stars). The NOS scores ranged from 0 to 9 stars, with 7 to 9 stars indicating high quality, 4 to 6 stars moderate quality, and below 4 stars low quality. This analysis included only studies with at least 4 stars to ensure methodological robustness (44). Two reviewers, XBH and SC, independently assessed each study, with any discrepancies resolved through discussion or consultation with a third reviewer, PY. The quality assessment ensured that only reliable, high-quality studies contributed to the meta-analysis.

To examine the relationship between PIU and suicidal behavior, including SI, SP, and SA, pooled OR with 95% CI were estimated using random-effects models. STATA software (version 16.0) was employed to conduct all statistical analyses. Heterogeneity across the included studies was assessed through the I² statistic and Cochran’s Q test. Significant heterogeneity was considered present when the I² statistic exceeded 50% or when Cochran’s Q test yielded a p-value of less than 0.05, warranting the application of a random-effects model (45). To assess the robustness of the overall results, sensitivity analyses were conducted. Specifically, the meta-analysis was re-evaluated by excluding studies with a high risk of bias and by using alternative statistical models (46). Furthermore, Publication bias was initially assessed through funnel plot analysis (47) and Egger’s test (48). If publication bias was detected, the trim-and-fill method was subsequently applied to adjust for it (49). A p-value of less than 0.05 was considered statistically significant. Subgroup analyses were conducted to explore the association between PIU and SI. These analyses were stratified by age groups, measurement tools, geographical regions, control variables including depression, and gender distribution to investigate potential sources of heterogeneity and identify variations in the strength of association across different subgroups. The fixed-effects model was employed to assess differences between these subgroups. All findings were reported in compliance with PRISMA guidelines, ensuring that the analysis was both transparent and reproducible.

According to the predefined search strategy, we initially identified 1,474 relevant articles. After initial screening with NoteExpress, 522 papers remained. Then, after reviewing titles and abstracts, the number reduced to 55. A comprehensive full-text analysis was done, and eventually 23 articles meeting the predefined inclusion criteria (33, 50–71) were selected. All studies were conducted as cross-sectional research. Longitudinal studies were excluded because they were either unavailable within the scope of the search results or did not meet the specific inclusion criteria. These selected studies covered 353,904 participants. Regarding assessment tools for SI, 16 studies (50–57, 59, 61, 63, 65, 68–71) used specialized and validated instruments for measuring SI. For example, they might include professional scales for measuring the frequency, intensity, and duration of suicidal thoughts. Meanwhile, 7 studies (33, 58, 60, 62, 64, 66, 67) used self-designed tools. These might be custom-developed based on specific research questions and populations. In terms of the assessment of PIU, 7 studies (33, 51, 55–57, 59, 64) adopted the Young Internet Addiction Test by Young et al. (72), which might cover multiple dimensions like internet usage time, frequency, behavior patterns, and dependence degree. Additionally, 5 studies (50, 60, 62, 69, 71) chose the Chen Internet Addiction Scale by Chen et al. (73), which might focus on aspects such as loss of control, interference with daily life, and withdrawal reactions. Furthermore, 3 studies (66, 68, 70) used the Mobile Phone Addiction Index translated into Chinese by Liu et al. (74), concentrating on evaluating addiction degree, impact on social and daily life, etc. The remaining 8 studies (52–54, 58, 61, 63, 65, 67) adopted other assessment tools based on specific scenarios and requirements. The sample sizes of the included studies varied significantly. The smallest had 439 participants, while the largest covered 136,266. As detailed in the Supplementary Materials , we conducted a very detailed quality assessment. 18 studies were classified as high-quality for excellent performance in research design, methods, data quality, and result reliability. Another 5 met certain standards in various aspects and were categorized as medium-quality. Table 1 summarizes the main characteristics of these literatures, and Figure 1 outlines each inclusion stage.

A homogeneity test conducted on 6 independent samples examining the link between PIU and SP revealed significant heterogeneity across the studies (I² = 96.0%, p < 0.001). Given this high level of variability, a random-effects model was employed to estimate the overall effect. The model produced a pooled effect size of OR = (1.50, 95% CI: 1.02, 2.20), indicating that individuals with PIU were 1.50 times more likely to engage in suicidal planning compared to those without PIU. This association was statistically significant (Z = 2.08, P = 0.037) (see Figure 8 ).

A homogeneity test conducted on 10 independent samples examining the link between PIU and SA indicated substantial heterogeneity across the studies (I² = 91.6%; p < 0.001). Given this considerable level of heterogeneity, a random-effects model was utilized to provide a more accurate estimation of the overall effect. The model produced a pooled effect size of OR = 1.48 (95% CI: 1.16, 1.89), indicating that individuals with PIU were 1.48 times more likely to attempt suicide compared to those without PIU. This association was statistically significant (Z = 3.20, P = 0.001) (see Figure 9 ).

To evaluate the robustness of the research findings, sensitivity analyses were conducted on the correlations between PIU and SI (see Figure 10 ), SP (see Supplementary Materials ), and SA (see Supplementary Materials ). The results indicated that the overall correlation coefficients exhibited only minor variations, thereby highlighting the stability and reliability of these findings.

The correlations between PIU and SI ( Figure 11 ), SP (see Supplementary Materials ), and SA (see Supplementary Materials ) were evaluated using funnel plots. The plots demonstrated a balanced distribution on both sides. Additionally, Egger’s regression tests were conducted for PIU and SI (t = 1.12, p = 0.95) ( Figure 12 ), SP (t = 0.37, p = 0.72) (see Supplementary Materials ), and SA (t = 0.83, p = 0.45) (see Supplementary Materials ). None of these tests revealed statistically significant differences, further indicating that no substantial publication bias was present.

The findings of this study revealed a significant moderating effect of regional differences on the relationship between PIU and SI among adolescents. Specifically, adolescents in the central and western regions of China with PIU faced a significantly higher risk of SI, whereas those in the eastern regions were at a lower risk. Several potential factors might have contributed to these regional differences. Firstly, economic development in China was uneven, with the eastern regions being more developed compared to the central and western regions. This economic disparity influenced various aspects of life, including access to resources that could have mitigated the impact of PIU. As a result, adolescents in economically disadvantaged areas might have experienced higher levels of stress due to poverty and limited opportunities, which could have exacerbated the effects of PIU on SI (96). Moreover, the level of urbanization in the eastern regions provided more social and recreational opportunities, potentially reducing the negative impact of PIU. Additionally, social support systems varied across regions, with the eastern regions potentially offering stronger and more diverse networks that could have helped adolescents cope with the challenges associated with PIU. In contrast, weaker social support structures in the central and western regions might have led to greater isolation and increased vulnerability to SI among adolescents with PIU (97, 98). Furthermore, cultural attitudes towards mental health might have differed, with more traditional views in the less-developed regions possibly leading to stigmatization and reduced help-seeking behavior (99). Educational disparities also played a significant role. It is likely that the eastern regions had better mental health education and awareness programs, which could have helped adolescents and their families recognize and address PIU early on. Conversely, lower levels of mental health education in the central and western regions might have contributed to a lack of awareness about the dangers of PIU, thereby increasing the risk of severe outcomes like SI (100, 101). Lastly, higher digital literacy in the eastern regions may have equipped adolescents with better coping strategies for managing internet use, thereby reducing the likelihood of PIU.

Moreover, the analysis demonstrated a notable moderating effect of age on the PIU-SI association, revealing that university students were more vulnerable to SI than middle school students when afflicted by PIU. Several key factors may explain the significant age moderation in the PIU-SI relationship. Firstly, developmental differences between university students and middle school students were considered pivotal. University students, being at a more advanced stage of psychological development, faced heightened levels of academic pressure, career uncertainty, and the challenges associated with independent living (102, 103). These factors were likely to exacerbate the effects of PIU, thereby increasing the risk of SI. In contrast, middle school students were generally in more protected environments, characterized by lower social pressures and less psychological maturity, which may have buffered them against the severe consequences of PIU. Secondly, there was a distinct difference in how these age groups engaged with social networks and experienced loneliness. University students were more inclined to rely on virtual social interactions, which could have intensified feelings of loneliness and isolation—both strongly correlated with SI. Conversely, middle school students typically maintained more in-person social interactions, which may have mitigated the psychological impact of PIU (83). Lastly, the availability and accessibility of mental health services differed significantly between these age groups. University students, often residing away from home, might have experienced limited access to support systems, thereby worsening the mental health issues associated with PIU. On the other hand, middle school students generally had more immediate access to parental and school-based support, which could have mitigated the risk of SI (104). These findings underscored the importance of developing age-specific intervention strategies. Educational institutions played a crucial role in implementing programs tailored to the distinct challenges faced by different age groups. Specifically, school-based interventions for middle school students and college wellness initiatives might have proven particularly effective in addressing the unique challenges these populations encountered in managing internet use and mitigating its potential harms.

Furthermore, this study revealed that controlling for depression significantly moderated the relationship between PIU and SI. This finding suggested that the elevated risk of SI among adolescents experiencing PIU could be partially attributed to underlying depressive symptoms, although other variables also played critical roles. This result was consistent with the findings of Wang et al. (52), who noted that depressive symptoms intensified the negative effects associated with PIU. A plausible explanation for this phenomenon was the close association between depressive disorders and PIU. Adolescents suffering from depression might have resorted to excessive internet use as a means of escaping psychological distress. However, such reliance on the internet often escalated into PIU, which, in turn, could have exacerbated depressive symptoms, creating a self-perpetuating cycle that amplified the risk of SI (28, 105). Moreover, adolescents with depression frequently experienced severe social isolation and a lack of effective coping strategies, which further increased their susceptibility to excessive internet use. While engaging with the online environment might have provided temporary relief, it often led to deeper psychological issues, thereby raising the likelihood of SI (106, 107). Given the complexities involved, it was deemed essential that the risk of SI in adolescents with PIU be assessed, irrespective of the presence of depression. Thus, a multifaceted approach—including comprehensive mental health evaluations, early identification of risk factors, and tailored interventions—was considered crucial for reducing the risk of SI in this vulnerable population.

In addition, the type of measurement tool used to assess SI was found to moderate the relationship between PIU and SI. Specifically, the correlation between PIU and SI was observed to be lower when SI was measured using validated tools compared to unvalidated ones. This finding can be plausibly explained by differences in the validity of the measurement instruments. Validated tools have undergone rigorous validation processes, ensuring higher reliability and validity, which allows for a more accurate assessment of SI. In contrast, unvalidated tools may have introduced measurement errors or biases, potentially inflating the observed correlations between PIU and SI (108). Consequently, when validated tools were employed, the positive correlation between PIU and SI was more likely to reflect the true relationship, rather than being exaggerated by the limitations of the measurement tool. Additionally, validated measures were often better designed to reduce social desirability bias in participants’ responses to SI-related questions. For instance, validated tools may have assessed SI more discreetly, thereby reducing the likelihood of participants concealing their true thoughts, which could have contributed to the observed weaker correlation when these tools were used (109).

Similarly, the type of measurement tool used to assess PIU was found to significantly moderate the relationship between PIU and SI. When tools designed to measure mobile phone addiction were employed, the correlation between PIU and SI was comparatively weaker, suggesting that the risk of SI among Chinese adolescents addicted to mobile phones may have been lower than that associated with other forms of internet addiction. Several plausible explanations can account for this observation: To begin with, the nature of mobile phone addiction, which typically involves excessive engagement in social media, instant messaging, and recreational activities, might have provided adolescents with social support and emotional regulation, potentially mitigating the onset of SI (110). In contrast, other forms of internet addiction, such as online gambling or pornography, were likely to result in greater psychological distress and isolation, thereby heightening the risk of SI. Moreover, cultural factors were considered to play a significant role; within the Chinese context, mobile phone use for social interaction and entertainment was widely accepted as a normal aspect of adolescent life. This acceptance might have reduced the negative psychological consequences of excessive use, thereby lowering the risk of SI (111). Lastly, mobile phone addiction may have served as a functional coping strategy for adolescents to escape real-life problems, such as academic pressure or family conflicts. While not a healthy coping mechanism, it might have temporarily diminished the likelihood of developing SI (112–114). Additionally, the specificity of mobile phone addiction measurement tools, which tended to focus on the frequency and duration of use rather than on deeper psychological distress, could have led to an underestimation of the correlation between PIU and SI (115). These considerations suggest that mobile phone addiction among Chinese adolescents may have been associated with a lower risk of SI compared to other types of internet addiction.

Nevertheless, the study results indicated that the relationship between PIU and SI was not significantly moderated by gender. This finding suggested that the impact of PIU on SI remained generally consistent across both males and females. This phenomenon could be attributed to two primary factors: On the one hand, it is plausible that PIU exerted broad and consistent negative effects on mental health that transcended gender differences. Irrespective of gender, excessive or unhealthy internet use could have led to social isolation, sleep disturbances, and emotional instability, all of which might have increased the risk of SI (103, 116, 117). On the other hand, it should be acknowledged that the subgroup analysis in this study included only three articles, which might have limited the generalizability of the findings and the ability to detect a gender moderation effect.

By systematically reviewing studies, this analysis investigated the PIU-suicidal behavior connection in Chinese adolescents. The use of large, varied samples strengthened the conclusions’ reliability. The findings may inform new prevention and management strategies, establishing a solid basis for future interventions.

Nevertheless, certain limitations should be recognized. First, the inclusion of only cross-sectional observational studies meant that the results reflected a correlation between PIU and suicidal behavior, without establishing causality. At the same time, the ongoing debate in the literature regarding the directionality (or bidirectionality) of the relationship between PIU and suicidal behavior was acknowledged, and the incorporation of longitudinal designs in future studies was identified as essential for better understanding these temporal dynamics and causal pathways. Secondly, the significant effects were observed in only a limited number of moderating variables, leaving much of the heterogeneity unaddressed. This limitation indicated that other potential moderating factors influencing the relationship between PIU and suicidal behavior were likely overlooked in the current analysis. Consequently, the full complexity of this relationship may not have been fully captured. It was recommended that future meta-analyses expand their scope to include a wider range of moderating variables, which could lead to a more comprehensive understanding of the factors contributing to variability in outcomes and clarify under which conditions PIU is most strongly associated with suicidal behavior. Thirdly, our analysis was constrained to studies published in Chinese and English, thereby excluding unpublished works, which may have introduced publication bias. Furthermore, by focusing exclusively on Chinese adolescents, the research scope is inherently limited, which restricts the applicability of the findings to broader populations. The exclusion of other age groups, such as children, adults, or elderly individuals, means that the results may not be generalizable beyond the specific demographic studied. Consequently, the insights gained from this research may not fully capture the complexities of PIU and suicidal behavior in different age groups, potentially overlooking important age-related differences that could influence the relationship between these variables. In addition, the reliance on self-report questionnaires across all studies, as opposed to clinical assessments, could have influenced the robustness of the findings. Lastly, a limitation of this study is that subgroup analysis was conducted solely on the relationship between adolescent PIU and SI. This leaves the connections between PIU and SP, as well as SA, underexplored. Future research should conduct subgroup analyses on the relationships between adolescent PIU and SP, as well as SA, to provide a more comprehensive understanding of PIU’s impact on different types of suicidal behavior. Such refined analyses would enable researchers to more accurately identify the underlying mechanisms of various suicidal behaviors, thereby offering stronger evidence for prevention and intervention strategies.