A systematic search was conducted on September 26, 2025, independently by 3 reviewers (QY, LCL, and VR) across 3 electronic databases (PubMed, Embase, and Web of Science) for studies published from January 1, 2011 (after which research incorporating smartphones began to emerge [20,51]).

Databases were searched for papers including in their titles and abstracts terms, such as “cell phone” or “smartphone” or “app” in combination with “addiction” or “problematic behavior” or “dependence” and “eating disorder” or “disordered eating” or “body dissatisfaction” or “emotional eating” (see Table S1 in Multimedia Appendix 1 for a full description of the search terms used per database). The searches were supplemented by manual screening of reference lists from included studies and citation tracking via Google Scholar to identify additional eligible studies not retrieved in the initial search.

Studies were included if they (1) examined the quantitative association between PSU and ED symptoms, using measures of effect size (eg, correlation coefficients and ß values), group differences (eg, t values and F values), and/or statistical significance (eg, P values); (2) assessed PSU using validated self-report measures that map onto at least 1 aspect of the criteria for behavioral addiction [20] (eg, Smartphone Addiction Scale [SAS] and Problematic Mobile Phone Use Questionnaire), or related indicators such as smartphone screen time (eg, hours per minutes of daily phone use); (3) assessed ED psychopathology or related symptoms using either validated self-report questionnaires (eg, Eating Disorder Examination-Questionnaire, Eating Disorder Inventory, Binge Eating Scale, Appearance Evaluation Scale, and Exercise Addiction Inventory) or structured diagnostic interviews (eg, SCID-5 and MINI); and (4) were published in English in peer-reviewed journals.

Studies were excluded if they (1) used PSU measures that were not specific to smartphone use, or that combined smartphone use with broader internet use or screen-based activity without reporting data separately; (2) were systematic reviews, scoping reviews, narrative reviews, meta-analyses, theses, commentaries, editorials, conference abstracts, or book chapters; (3) did not include validated measures for both PSU and ED symptoms; or (4) were published before January 1, 2011. Data from participants of any gender, age, or nationality were included.

Titles and abstracts were screened independently by 2 reviewers (QY and LCL) against basic eligibility criteria including language type, sample type, study focus, and paper type, and noneligible studies were excluded. Full-text papers of potentially eligible studies were then screened independently by the same 2 reviewers. Reasons for exclusion at the full-text stage were documented. Disagreements at any stage of the screening process were first discussed between the 2 reviewers, and unresolved cases were referred to the wider team (VRR, JLK, and BC).

Data extraction was carried out by 2 reviewers (QY and LCL) using a bespoke database and 50% of the extracted data were independently reviewed by another reviewer (JLK). Extracted variables included publication details (author, year), country, study design, setting, sample characteristics (eg, size, age, and gender), clinical characteristics (eg, ED diagnosis, BMI, and comorbidities), and the validated measures used to assess ED symptoms and PSU. Extracted data relating to results of the studies included a narrative summary of results, statistical results (eg, P values, r values, ß values, F values, and 95% CIs), covariates, and findings relating to studied moderators or mediators where applicable.

The quality assessment of included studies was carried out independently by 2 reviewers (QY and LCL) using a version of the Newcastle-Ottawa Scale (NOS) [52] modified for assessing cross-sectional studies and adapted to the topic of this review (Table S2 in Multimedia Appendix 1 [23,53-86]). The NOS is a widely used tool for assessing the quality of non–randomized observational studies according to 3 overarching categories: (1) the selection of the study groups, (2) the comparability of the groups, and (3) the ascertainment of the outcomes. Stars are awarded for each item, which determines the overall quality of the domain. Each study was then judged overall as good, fair, or poor quality. Studies were rated as “good” when all domains were rated good or only 1 was rated fair, with no poor ratings. They were rated as “fair” when 1 poor rating was present alongside otherwise fair or good domains, or when multiple domains were fair without any poor. Finally, studies were rated as “poor” when 2 or more domains were rated poor, or when a single poor rating was found in a critical domain such as outcome measure. Discrepancies were resolved by consensus and consultation with senior reviewers (JLK and BC).

 Following preregistration, the decision was made to include an assessment of the quality and certainty of evidence using the evidence-grading system developed by the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) collaboration (Table S3 in Multimedia Appendix 1) [87]. One senior author (JLK) initially applied the GRADE system, followed by an assessment of the evidence ratings for each outcome with another senior reviewer (BC). A final decision on the ratings was reached through discussion and consensus. As our review included only cross-sectional data (ie, observational studies), we downgraded the evidence for each outcome, starting from low quality, when there was a serious limitation in relation to a particular factor or by 2 levels if we considered that there was a very serious limitation. Evidence was only upgraded in the case of very large, consistent, and plausible associations, in line with GRADE guidance for observational studies. The GRADEpro (Evidence Prime) [88] software was used to determine the certainty of evidence and generate the Summary of Findings table (Table S3 in Multimedia Appendix 1).

Due to the contextual heterogeneity of the included studies in terms of population characteristics, outcome measures, and different statistical reporting formats, a narrative synthesis was conducted to summarize the findings. To explore patterns across studies, results were organized into key themes according to ED outcome, with the primary focus being results relating to ED symptomatology, with other themes including body image and broader altered or disordered eating behaviors (eg, food addiction and emotional eating). Within these themes, results were synthesized according to statistical approach, including between-group comparisons, linear associations (regression models and correlations), and mediation and moderation analyses. Results were also narratively synthesized for measures of PSU and smartphone screen time separately. Study characteristics and a brief summary of the key results, including statistical notation where available and appropriate, are presented per individual study in a summary table. No data conversions or sensitivity analyses were performed.

A total of 3002 papers were identified, of which the full texts of 340 were assessed against the full eligibility criteria. Of these, 28 met the criteria for inclusion and an additional 7 studies were identified through backward and forward citation tracking, resulting in a total of 35 studies included in the review (Figure 1).

Table 1 shows the methodological and sample characteristics of the included studies. The combined sample size across studies was 52,584, ranging from 86 [53] to 10,246 [54]. Studies were published between 2019 and 2025 and were conducted in Turkey (n=9), China (n=9), the United States (n=4), Germany (n=3), Spain (n=2), the United Arab Emirates (n=1), Bahrain (n=1), Italy (n=1), Austria (n=1), Brazil (n=1), South Korea (n=1), India (n=1), and Singapore (n=1). The vast majority of studies (n=32) was cross-sectional, apart from 3 studies that were prospective cohort, randomized controlled trials, and longitudinal studies, reporting cross-sectional data available for inclusion in this study.

Most samples were recruited from nonclinical settings, including universities (n=13), schools (n=11), both schools and universities (n=2), and the community (n=8). One study recruited the sample from a family medicine hospital polyclinic for weight loss [55]. Of the studies reporting age ranges (n=26), 11 studies were conducted in adolescent populations only (ie, those aged 18 years and younger) and 11 in adults only, of which 6 were in young adults (aged 18-35 years). Four studies included both adolescents and adults. The pooled mean (SD) age across studies was 17.0 (5.5) (reported by n=31 studies). Four studies were conducted in female-only samples, and in the rest, the percentage of female participants ranged from 44.4% to 85.6%.

In terms of the assessment of ED and ED-related symptoms, the most commonly used instrument was the Eating Attitudes Test (EAT [90]; n=9), used in its 26-item version in 5 studies, 40-item version in 3, and in 1 study the version was unclear. Other measures of ED symptoms included the Eating Disorder Examination Questionnaire (EDE-Q [91]; n=1) and in its 12-item form (Chinese version of Short Form of the Eating Disorder Examination Questionnaire [92]; n=1), and diagnostic or screening tools included the ICD-10 (International Statistical Classification of Diseases, Tenth Revision) Symptom Rating Scale (ICD-10 Symptom Rating Scale [ISR] [93]; n=1), the Minnesota Impulse Disorders Interview [94], the Kiddie Schedule for Affective Disorders and Schizophrenia-5 (n=1 [95]), and the Sick, Control, One Stone, Fat, Food Questionnaire (SCOFF [96]; n=2). In terms of measures of relevant wider eating behaviors, the Yale Food Addiction Scale [97] was used in 5 studies, with a version for children (Yale Food Addiction Scale for Children [98]) in 2 studies. The Dutch Eating Behavior Questionnaire [99] was used in 3 studies and the Three-Factor Eating Questionnaire (TFEQ) [100] in 2 studies. Body image dissatisfaction was measured in 12 studies, using a variety of measures (Table 1).

The most common measure of PSU was the SAS (n=4 [101]) and its short form (SAS-SV; n=14 [102]). Three studies used the Mobile Phone Addiction Index [103], 2 used the Nomophobia Questionnaire [104], and other questionnaires assessing PSU used in single studies included the Mobile Phone Involvement Questionnaire [105], Addictive Patterns of Use Scale [106], Problematic Mobile Phone Use Scale, Mobile Phone Addiction Tendency Scale [107], Smartphone Addiction Inventory—Italian version [108], Scale of Risk of Addiction to Social Media and the Internet for Adolescents [109], Mobile Phone Problematic Use Scale [110] and the Korean Smartphone Overdependence Scale for Adolescents. Smartphone screen time was measured in 8 studies. All of the aforementioned instruments addressed at least 1 domain of PSU aligning with the DSM-5 criteria for a behavioral addiction characterized by smartphone use.

Seven studies included samples that were deemed representative of the target population, whereas 18 studies included samples that were deemed somewhat representative of the average target population (eg, due to nonrandom sampling). Only 10 studies provided justification for their sample size. Most used a validated measurement tool for PSU to ascertain the exposure (n=31), and very few studies used a nonvalidated measurement tool (eg, self-reported smartphone use only) (n=4). In terms of comparability, most studies accounted for age as a confounding variable (n=23), and most also controlled for at least 1 additional potential confounding factor (eg, sex, BMI; n=28). All studies used a validated measurement tool for ED or ED-related outcomes, and almost all studies presented an appropriate and clearly described statistical test for the measurement of the association (n=34). Overall, most included studies were rated as good (n=28) and some were rated as fair (n=6), primarily due to not controlling for confounding variables (Table S2 in Multimedia Appendix 1). Two studies [55,56] received a rating of “poor” as they used a nonrepresentative sample and failed to account for any confounders to address potential sources of bias.

The full results of the included studies together with their characteristics are shown in Table S4 in Multimedia Appendix 1. Results were synthesized according to the type of ED or ED-related outcome, stratified according to the analytical approach taken. A total of 16 studies measured ED psychopathology, 5 studies measured food addiction symptoms, 6 measured other eating behaviors, and 12 measured body image–related outcomes.

Five studies assessed the association between PSU or PSU-related outcomes and food addiction scores as measured by the Yale Food Addiction Scale. One study found that South Korean adolescents with a high risk of PSU had greater food addiction scores than adolescents with a low risk of PSU [71]. On the contrary, a study in Turkish adults found no greater food addiction scores in people categorized as having a smartphone addiction compared with those without smartphone addiction, although this study received a rating of “poor” as per the modified NOS [55]. Additionally, 1 study found that food addiction scores were no different across groups stratified by presence, risk, or absence of nomophobia (the fear of being without a mobile phone) [72]. An additional study examining nomophobia found no linear association between nomophobia and food addiction scores before and after adjusting for age, sex, BMI, and insomnia [73]. Due to the cross-sectional nature of the studies and high suspicion of publication bias, the GRADE of evidence was rated as very low certainty for this outcome. Publication bias cannot be ruled out, as the majority of studies reporting positive associations had a combined sample size similar to the single study reporting no association, suggesting that studies with null findings may have been less likely to have been published.

Despite this, 3 of these studies, as well as 1 additional study, found that measures of smartphone addiction [55,85], “overdependence” [71], or nomophobia [72] were positively correlated with food addiction scores. Again, the grade of evidence for these findings was rated as very low certainty. In one of these studies, this was the case when adjusting for age, sex, BMI, and socioeconomic status [71], but in another, the association was not found when stratifying the analysis by gender [72]. One study examining mediators reported that the positive association between smartphone addiction and food addiction was mediated by greater emotion regulation difficulties [85].

Six studies examined the relationship between PSU-related outcomes and other eating behaviors such as emotional eating, cognitive restraint, and uncontrolled eating, as measured by the TFEQ (n=2), Dutch Eating Behavior Questionnaire (n=4), or Child Eating Behavior Questionnaire (n=1).

Both studies using the TFEQ found a positive correlation between smartphone addiction scores and the subscales of the TFEQ, including uncontrolled eating, cognitive restriction, and emotional eating [74,75]. In the former study, these associations remained significant when adding social media addiction scores and digital game addiction scores to a multiple regression model [74]. A positive association between smartphone addiction and emotional eating was also found in 2 additional studies [76,77], and in another study, it was found that adolescents with a higher risk of PSU had greater emotional overeating scores than adolescents with a lower risk of PSU [71]. Smartphone addiction was also positively associated with restrained eating [76,77,85], dysregulated eating [85], and external eating [76,77]. Similar to the previous outcomes, the GRADE evidence for the correlative findings pertaining to the relationship between PSU scores and emotional eating, as well as restrictive eating, was rated as low in both instances.

Three studies explored mediators [75,77,85]. It was found that emotion regulation difficulties significantly mediated the association between addictive phone use and dysregulated eating, restrained eating, and food addiction in 1 study [85]. In another, the relationship between body dissatisfaction and restrained eating, emotional eating, and external eating was mediated by smartphone addiction scores in 3 separate mediation models [77]. In this study, sequential mediation effects were found with smartphone addiction scores and depression symptoms as sequential mediators. Additionally, 1 study found that the relationship between smartphone addiction and higher BMI was positively mediated by all subscales of the TFEQ (higher uncontrolled eating, cognitive restriction, and emotional eating scores) in 3 separate mediation models [75].

This systematic review assessing the association between PSU and ED psychopathology included 35 studies, predominantly reporting cross-sectional data in nonclinical populations. The majority of studies were rated as good quality. Most samples were recruited from schools and universities, and approximately half were adolescents and half (mostly young) adults. Across the included studies, it was broadly found that individuals scoring highly on PSU measures scored more highly on measures of ED psychopathology, and there were weak-to-moderate positive linear associations found. These findings were largely consistent across age groups and countries. Additionally, higher daily smartphone use was consistently related to higher ED psychopathology, particularly where daily use exceeded 7 hours per day [63]. Measures of PSU were also related to greater food addiction symptoms, broader disordered eating behaviors (eg, uncontrolled eating, emotional overeating, and cognitive restriction), and body dissatisfaction in most studies. The findings assessing the relationship between smartphone screen time and body dissatisfaction were more mixed. Across outcomes, a GRADE assessment of the evidence led to ratings of “low certainty” in most instances due to the cross-sectional nature of the evidence, apart from the outcomes relating to the association between PSU and food addiction, which received ratings of “very low certainty” largely due to substantially inconsistent findings across studies or potential publication bias. Longitudinal and interventional studies are needed to establish temporal and causal relationships.

The most consistently identified mediators of the relationship between PSU and ED psychopathology related to increases in negative affect, including emotion regulation difficulties and anxiety and depressive symptomatology. Maladaptive smartphone use may actively (or passively) increase one’s exposure to content related to thinness and body shaping, which may lead to negative affect via self-objectification [111] and increase the risk of developing ED symptoms [112]. In addition, night-time or late-evening smartphone use may interfere with sleep quality. Sleep problems are reported transdiagnostically across the ED spectrum [29] and in subclinical contexts are related to several eating behaviors such as impulsive eating and emotional eating [113]. Indeed, it was found that negative affect as a result of PSU acted as sequential mediators from poor sleep quality to higher ED behaviors in one study [66] and from greater body dissatisfaction to greater disordered eating behaviors in another [77]. Alternatively, individuals with ED symptoms may use smartphones to relieve or avoid negative emotions and stress [114], eventually developing a dependence on smartphones. However, it was found in 1 study that baseline PSU scores positively predicted ED symptoms at a 1-year follow-up but not vice versa [23], which, tentatively, could be a preliminary indicator of causality.

Traits common in individuals with, or at risk of developing, an ED, such as low self-esteem [115,116], perfectionism [117], features of addiction such as greater reward sensitivity, and impulsive traits such as negative urgency [118,119] and lack of premeditation [120] that are associated with PSU [121], may increase the risk of PSU or reliance on smartphones. Smartphones, with their ubiquitous use in everyday life, represent a convenient tool to consistently facilitate exposure to both content reflecting idealized depictions of lives and bodies and the opportunity to self-monitor one’s eating behaviors and exercise in a gamified and engaging way. In people with EDs, this convenience may be more pernicious, and the use of non–social media smartphone apps such as calorie-tracking and exercise-tracking apps [122,123], with reminders and gamified features (eg, streaks), may additionally increase an individual’s tendency to engage with their smartphone and reinforce dependency on these apps [124]. It is possible that this may also reinforce the addictive nature of some ED behaviors, such as body checking, compulsive exercising, binge eating, calorie counting, and restricting food intake. Moreover, individuals prone to developing disordered eating behaviors may exhibit an attentional bias toward ED-related content [125,126]. For these individuals, engagement with algorithms that learn, adapt, and are personalized via attention-related metrics may increase exposure to ED-related content [127] and promote body dissatisfaction via appearance and/or lifestyle comparison.

On the other hand, it is apparent from the results of this study that even for people without diagnosed EDs, PSU is associated with poorer body dissatisfaction and altered eating behaviors, which are likely to cause distress. During adolescence, individuals increasingly form their self-concept through observing others and their surroundings, making social comparison a mechanism for self-evaluation [128]. Smartphones provide a more convenient way for such comparisons. Adolescents with PSU may be frequently exposed to idealized appearance images, making them more likely to compare their own appearance with these “standards,” which leads to body dissatisfaction and negative self-evaluation. For young people and individuals at risk of developing an ED, or who are exhibiting prodromal symptoms of EDs, prevention programs and psychoeducation that incorporate media literacy may benefit from specifically including content relating to PSU. Early intervention pathways may benefit from incorporating psychoeducational content that targets the quality and nature of smartphone use, rather than focusing solely on reducing smartphone screen time. Additionally, emphasizing the benefits of engaging in more offline social interactions and activities that can help bolster one’s social connectedness and support, emotion regulation, and general psychological well-being may also enhance the quality of their interpersonal relationships, interpersonal confidence, and emotion regulation skills and reduce dependence on smartphones.

To the authors’ knowledge, this review is the first to synthesize existing evidence on the relationship between PSU and EDs, as well as their related symptoms. The review adhered to a preregistered protocol and a comprehensive search was conducted across 3 databases, which was supplemented by manual screening of reference lists from included studies and citation tracking.

A major limitation pertaining to the included studies is that almost all included studies used cross-sectional designs. Given the cross-sectional nature of these data, it remains unclear whether PSU independently contributes to adverse psychological outcomes relating to ED symptoms, whether individuals with preexisting vulnerabilities are more likely to engage in PSU, or whether the relationship is bidirectional. Moreover, nearly all studies relied on self-report measures of both PSU and ED symptoms, which are subject to recall error and social desirability bias. In several studies, potential confounding variables such as gender, BMI, and age were considered; however, others that might influence the relationship between PSU and ED outcomes such as personality traits (eg, compulsivity and neuroticism [129]) and psychiatric comorbidity were not explored.

Such limitations underscore the need for future research using longitudinal or experimental designs, alongside objective measures and consideration of confounding factors, to strengthen the causal inference and reduce bias. As such, it is also important to acknowledge that ED outcomes were mostly assessed in nonclinical populations and utilized self-report questionnaires instead of formal diagnostic interviews, hospital admissions, or referrals to specialist services, meaning the findings cannot be fully generalized to clinical populations. Moreover, since patterns of social media overuse or maladaptive usage have been identified as plausible risk factors for the development of EDs cross-culturally [34], it is possible that problematic social media use contributed to our observed findings. Future studies should endeavor to simultaneously measure problematic social media usage and PSU, as well as specific app usage, to delineate pathways between these factors and specific ED symptoms and behaviors.

The studies included in this systematic review indicated a significant and consistent association between PSU and ED symptoms and related outcomes such as food addiction, emotional overeating, and body dissatisfaction. The identification of key mediators such as emotion regulation difficulties, anxiety, and depression underscores the need to adopt a transdiagnostic approach in ED prevention and intervention. While these results align with existing literature demonstrating an association between PSU and negative mental health outcomes, the cross-sectional nature of the included studies limits causal inferences. Nevertheless, the findings of this review underscore the importance of integrating psychoeducation on PSU into ED prevention strategies and early intervention frameworks to enhance the timely detection and treatment of ED symptoms.

Crucially, future longitudinal and experimental research should seek to elucidate the bidirectional nature of this relationship, include individuals with and without clinically diagnosed EDs, and evaluate whether psychoeducational interventions targeting PSU can reduce the onset and severity of ED symptoms in at-risk populations. Patterns of use of specific apps that could contribute to or explicate the nature of PSU within ED populations should be explored, alongside a collation of qualitative evidence (eg, a qualitative metasynthesis) to enrich our findings.