The relationship between social media addiction and metacognitions in patients with social anxiety disorder

Abstract

Background

Mental health problems including depression, eating disorders and anxiety disorders are associated with extensive social media use. The literature has examined the relationship between social anxiety and social media addiction; however, there is a lack of studies conducted in clinical patient populations. This study aimed to investigate the role of metacognitions related to social media addiction in individuals diagnosed with social anxiety disorder (SAD).

Methods

In this cross-sectional case–control study, 99 participants aged 18–65 years with social anxiety disorder and 113 control participants were recruited. One control participant with missing BSMAS data was excluded from all analyses, including sociodemographic comparisons; therefore, the final analytic sample included 99 SAD participants and 112 control participants without current psychiatric disorder. The SCID-5 was administered to participants in the SAD group to confirm the diagnosis of social anxiety disorder and to control participants to screen for current psychiatric disorders. No current psychiatric disorder was identified among control participants at the time of assessment. After written informed consent was obtained, all participants completed the Bergen Social Media Addiction Scale (BSMAS), Liebowitz Social Anxiety Scale (LSAS), and Metacognitions Questionnaire-30 (MCQ-30).

Results

SMA prevalence did not differ significantly between participants with SAD and control participants without current psychiatric disorder (52.5% vs. 46.4%; χ²=0.781, p = 0.377; φ = 0.06). Within the SAD group, SMA was associated with higher negative metacognitive beliefs, particularly uncontrollability of worry and the need to control thoughts, and BSMAS severity correlated with these metacognitive domains. Among participants meeting the operational SMA criterion, those with SAD showed higher MCQ-30 uncontrollability of worry scores than control participants. Sensitivity analyses using an alternative BSMAS threshold showed the same overall pattern (see Table 8).

Conclusion

SMA prevalence was comparable between participants with SAD and control participants without current psychiatric disorder; however, within SAD, problematic social media use was associated with negative metacognitive beliefs. These cross-sectional findings suggest that metacognitive beliefs may be treatment-relevant correlates of problematic social media use in SAD; longitudinal and interventional studies are needed.

Clinical trial number

Not applicable.

Introduction

With technological advances and the increasing availability of the internet, social media use has increased worldwide. It was anticipated that 5.04 billion people used social media in 2024, and by 2027, there may be 5.85 billion users worldwide [1]. Social networking platforms are used for interpersonal interaction, rapid access to information, work-related activities, and entertainment. Although social media may facilitate communication and access to information, excessive or problematic use has been associated with several adverse outcomes [2]. Terms such as social media addiction (SMA), social media use disorder (SMUD), and compulsive social media use are used to describe problematic social media use. Excessive social media use has been associated with addiction-like features, including mood regulation difficulties, withdrawal, tolerance, and relapse [3]. Studies have shown that SMA is associated with many psychiatric disorders including major depression, anxiety disorder, eating disorder, social anxiety disorder and insomnia [4].

Social anxiety disorder (SAD) is characterized by marked fear or anxiety in social or performance situations, accompanied by concerns about being humiliated, negatively evaluated, or behaving in an embarrassing way [5]. Individuals with SAD may feel physical symptoms such as sweating, flushing, trembling and palpitations in social environments due to autonomic nervous system activation. Beyond subjective distress, SAD can cause impairment in social, occupational, and academic functioning, leading to reduced quality of life [6].

For individuals with SAD, social media may provide an alternative social environment in which face-to-face anxiety cues are less salient [7]. The results of a recent meta-analysis confirm that individuals with social anxiety prefer computer-mediated communication rather than face-to-face communication. Individuals with high social anxiety may perceive internet-mediated interactions as less threatening because visible anxiety symptoms, such as blushing or trembling, are less apparent to others [8]. However, online communication does not help individuals overcome their catastrophic beliefs and dysfunctions. Previous studies have suggested that frequent social comparison on social media may be associated with loneliness and social anxiety [9].

Metacognition refers to higher-order cognitive knowledge and processes involved in monitoring, controlling, and evaluating cognition [10]. Recent theories suggest that metacognitions have an important role in the onset and maintenance of psychiatric disorders. According to metacognitive theory, metacognitive processes play a central role in the adaptive regulation of cognition. Dysregulation in these processes has been proposed to contribute to various forms of psychopathology [11]. For example, the role of metacognitions has been shown in many disorders such as depression, obsessive-compulsive disorder, alcohol use disorder and social anxiety disorder [12]. In behavioral addictions, dysfunctional coping behaviors have also been proposed to be activated and maintained by maladaptive metacognitive beliefs [13]. In this framework, positive metacognitive beliefs (e.g., that engaging in a behaviour helps regulate mood or reduce distress) may increase the tendency to use social media as a coping strategy, whereas negative metacognitive beliefs (e.g., beliefs about uncontrollability, cognitive harm, or the need to control thoughts) may contribute to the persistence of repetitive use despite negative consequences. Accordingly, metacognitive beliefs may shape how individuals respond to internal experiences such as anxiety, worry, and intrusive thoughts, and may guide coping choices toward repetitive behaviours. When social media use is appraised as helpful for managing distress while simultaneously perceived as difficult to control, a self-reinforcing cycle may emerge in which short-term relief strengthens continued use, thereby potentially contributing to problematic engagement.

Although there are studies on the relationship between social anxiety and social media addiction in the literature, this issue has not, to our knowledge, been studied from a metacognitive perspective and in patients with social anxiety disorder. Previous studies have often focused on symptom correlations; however, considering metacognition as a potential transdiagnostic pathway may help clarify shared cognitive–affective mechanisms relevant to both anxiety and problematic social media use. This may also inform future treatment-oriented work, as identifying metacognitive domains associated with SMA could help refine clinical targets. In this study, we aimed to examine the relationship between metacognitive beliefs and problematic social media use in participants with DSM-5 social anxiety disorder recruited from a psychiatry outpatient clinic. We also examined whether metacognitive profiles differed between participants with SAD and control participants. We hypothesized that SMA prevalence might differ between participants with SAD and control participants without current psychiatric disorder. We also hypothesized that, within the SAD group, greater SMA severity (BSMAS total score) would be associated with more dysfunctional metacognitive beliefs (MCQ-30), particularly negative metacognitions such as uncontrollability of worry and the need to control thoughts.

Materials and methods

Study design and participants

This observational cross-sectional case–control study was conducted at the psychiatry outpatient clinic of Dr. Abdurrahman Yurtaslan Oncology Training and Research Hospital in Ankara, Turkey. A large national survey reported Bergen Social Media Addiction Scale scores [14]; the overall variability can be approximated (overall SD ≈ 4.78). Because an established minimal clinically important difference (MCID) for the BSMAS is not available, we a priori assumed a 2-point difference as a pragmatic small-to-medium effect for planning purposes. Under these variance assumptions, a 2-point difference corresponds to an effect size of approximately d ≈ 0.44 that we used for the a priori power calculation. Accordingly, it was calculated that a total of 166 participants (minimum 83 per group) were required to detect this effect size with 80% power and a two-sided alpha of 0.05.

The study included 99 participants with a confirmed diagnosis of SAD according to DSM-5 criteria. Additionally, 113 control participants were recruited and frequency-matched with the SAD group in terms of age and gender. One control participant had missing BSMAS data and was excluded from all analyses, including sociodemographic comparisons; thus, the analytic control sample was 112. Inclusion criteria for the SAD group were a diagnosis of SAD, age between 18 and 65 years, and provision of informed consent. Exclusion criteria included the presence of any major neurological disorder, secondary psychiatric disorder or intellectual disability [15].

Procedure

Participants completed standardized self-report questionnaires assessing problematic social media use, social anxiety symptoms, and metacognitive beliefs. Before data collection, the study obtained approval from the hospital’s ethical committee. All participants provided written informed consent. The diagnosis of SAD was verified by using Structured Clinical Interview for DSM-5 (SCID-5). The SCID-5 was also administered to all control participants to screen for current psychiatric disorders; no active psychiatric diagnoses were identified at the time of study.

Measures

Bergen Social Media Addiction Scale (BSMAS)

Participants’ levels of social media addiction were evaluated using BSMAS. This scale has six items with a 5-point Likert scale. These items assess salience, mood modification, tolerance, withdrawal, conflict, and relapse, which are commonly described components of addictive behaviors. Higher scores indicate greater problematic social media use [14]. The scale was translated into Turkish by [16], demonstrating an acceptable reliability coefficient and item-total correlations. The results indicated that the Bergen Social Media Addiction Scale is applicable in research conducted in Turkey. In the present study, participants scoring 3 or above on at least 4 of the 6 BSMAS items were classified as meeting the operational SMA criterion, as used in previous research. The same operational criterion for SMA (≥ 4 of 6 items rated ≥ 3) was applied identically in both the SAD and control groups. Sensitivity analyses using a stricter operational SMA definition showed the same overall pattern (Table 1, Panel A and Panel B).

Table 1.

Sensitivity analysis using a stricter operational definition of social media addiction (SMA) Panel A. SMA prevalence by group using primary vs. stricter cut-off (6/6 items ≥ 3)

Panel A. SMA prevalence by group using primary vs. stricter cut-off (6/6 items ≥3)
SMA definition	SAD (n = 99) n (%)	Control (n = 112) n (%)	χ² (df = 1)	p	φ
Primary (≥ 4 of 6 items rated ≥ 3)	52 (52.5)	52 (46.4)	0.781	0.377	0.060
Stricter (6 of 6 items rated ≥ 3)	21 (21.2)	19 (17.0)	0.617	0.432	0.054

Panel B. Within-SAD associations: directional consistency of key correlations (Table 7)
Association (SAD group)	Spearman r (main analysis)	95% CI	Sensitivity (direction)
BSMAS total × MCQ Uncontrollability of worry	0.289	0.097–0.460	Positive (unchanged)
BSMAS total × MCQ Need to control thoughts	0.368	0.184–0.527	Positive (unchanged)
BSMAS total × MCQ Cognitive awareness	0.250	0.055–0.426	Positive (unchanged)
BSMAS total × MCQ Total score	0.319	0.130–0.486	Positive (unchanged)

Note: φ denotes the phi coefficient (equivalent to Cramer’s V for 2 × 2 tables)

Note: Correlations were computed using BSMAS total score as a continuous measure; therefore, altering the categorical SMA threshold affects prevalence classification (Panel A) but does not change the direction of BSMAS–MCQ correlations

Liebowitz social anxiety scale (LSAS)

A widely used self-assessment instrument, the LSAS measures the severity of symptoms and distress associated with social anxiety. This assessment measures the levels of fear and avoidance exhibited in both social interactions and performance situations. Elevated scores on the LSAS suggest greater degrees of social anxiety [17]. The Turkish validity and reliability study was performed by Nesrin Dilbaz in 2001. The scale, primarily utilized for the SAD group, is also applicable for screening healthy individuals. The instrument has a four-point Likert scale comprising 48 items, with 24 items designated for evaluating anxiety and 24 for assessing avoidance. Subscale scores are obtained by separately summing the individual’s scores in the anxiety and avoidance sections. Each item on the scale is assigned a score ranging from 1 to 4 points. The Turkish variant of the scale lacks a cut-off score [18].

Metacognition Questionnaire-30 (MCQ-30)

The MCQ-30 is a 30-item self-report questionnaire designed to assess metacognitive beliefs and processes. These encompass positive beliefs about worry, negative beliefs about lack of control and potential harm, cognitive confidence, desire for control, and cognitive self-awareness. Higher scores indicate more dysfunctional metacognitive beliefs [19]. The Turkish validity and reliability assessment of the MCQ-30 scale was performed by Tosun and Irak in 2008 [20].

Internal consistency in the current sample was acceptable to excellent (BSMAS α = 0.864; LSAS α = 0.956; MCQ-30 α = 0.879).

Statistical analysis

Continuous variables were summarized using mean, standard deviation, median, minimum, and maximum values. Categorical variables were summarized using frequencies and percentages. The Shapiro-Wilk statistical test was implemented to assess the normality of the distribution of continuous data. Analyses were conducted with two complementary aims: (i) between-group comparisons (SAD vs. control group) and (ii) within-group association analyses in the SAD sample. Social media addiction (SMA) was operationalized using the Bergen Social Media Addiction Scale (BSMAS) in two ways: Outcome 1 (binary): SMA status, defined as meeting versus not meeting the predefined operational BSMAS criterion (≥ 4 of 6 items rated ≥ 3), and Outcome 2 (continuous): BSMAS total score as an index of problematic social media use severity. Metacognitions were assessed using the Metacognitions Questionnaire-30 (MCQ-30) and were evaluated as the main predictors (MCQ-30 total score and subscale scores, where applicable).

Between-group comparisons were performed using independent-samples t-tests for normally distributed continuous variables and Mann–Whitney U tests for non-normally distributed variables. An analysis of nominal variables was conducted by cross-tabulations employing the Chi-Square test and Fisher’s exact test. To compare the scores of the Bergen Social Media Addiction Scale (BSMAS) across two categories of patient characteristics, the Mann-Whitney U test was employed. The Kruskal-Wallis Analysis of Variance was employed for comparisons that involved more than two categories of patient characteristics. An examination of the associations among scale scores was performed using Spearman’s correlation coefficient. Effect sizes were calculated as Cohen’s d for independent-samples t-tests, r for Mann–Whitney U tests (r = |Z|/√N), and Cramer’s V (or Phi for 2 × 2 tables) for chi-square tests, and are reported alongside p-values. Within the SAD group, we examined the association between SMA severity (BSMAS) and metacognitions (MCQ-30). Univariate and multivariate logistic regression analyses were conducted to identify factors associated with meeting the operational SMA criterion (outcome: meeting vs. not meeting the operational SMA criterion based on the predefined BSMAS threshold). MCQ-30 subscales were entered simultaneously in a single block together with education level, income level, and employment status as categorical covariates. Reference categories were literate only for education level, < 5,500 TL for income level, and full-time employment for employment status. Multicollinearity was assessed using variance inflation factors (VIFs) and tolerance values. VIF values ranged from 1.11 to 2.59 and tolerance values ranged from 0.386 to 0.904, indicating no evidence of problematic multicollinearity among the variables included in the multivariate logistic regression model. The full model specification and results are reported in Table 2. Sensitivity analyses were conducted using alternative SMA thresholds on the BSMAS to evaluate the robustness of the main findings. Statistical analyses were performed using IBM SPSS Statistics version 20, and p values < 0.05 were considered statistically significant.

Table 2.

Multivariate binary logistic regression: factors associated with meeting the operational SMA criterion within the SAD Group (n = 99)

Predictor	B	SE	OR	95% CI	p
MCQ-30 Subscales
Positive beliefs about worry	−0.131	0.068	0.877	0.767–1.003	0.054
Uncontrollability of worry	−0.063	0.087	0.939	0.792–1.113	0.469
Cognitive insecurity	0.019	0.053	1.020	0.920–1.130	0.713
Need to control thoughts	0.173	0.088	1.189	1.001–1.413	0.049*
Cognitive awareness	0.184	0.101	1.202	0.986–1.465	0.069
Education (ref: literate only)
Primary school graduate	−0.611	2.356	0.543	0.005–55.1	0.795
High school graduate	1.011	1.604	2.748	0.118–63.9	0.528
University graduate	0.526	1.611	1.692	0.072–39.7	0.744
Master’s/doctoral degree	0.994	1.878	2.702	0.068–107.3	0.597
Income (ref: <5,500 TL)
5,500–10,000 TL	0.737	0.939	2.089	0.332–13.16	0.433
10,000–15,000 TL	−2.741	1.345	0.065	0.005–0.901	0.042*
>15,000 TL	0.095	0.957	1.099	0.168–7.178	0.921
Employment status (ref: full-time employed)
Part-time	1.127	1.833	3.087	0.085–112.0	0.539
Not working	0.771	1.021	2.162	0.292–15.97	0.450
Student	−0.001	0.931	0.999	0.161–6.199	0.999

Note. OR = odds ratio; CI = confidence interval. The dependent variable was SMA status based on the operational BSMAS criterion, coded as 1 = meeting the criterion and 0 = not meeting the criterion. All MCQ-30 subscales and categorical covariates were entered simultaneously. Reference categories were literate only for education level, < 5,500 TL for income level, and full-time employment for employment status. Because some categorical covariates contained small cell counts, estimates for these categories should be interpreted cautiously. VIF values ranged from 1.11 to 2.59, indicating no problematic multicollinearity. *p < 0.05

Results

Given that the SAD and control groups differed significantly on several sociodemographic characteristics (Table 3), all between-group comparisons in clinical and metacognitive scales should be interpreted with caution; residual confounding cannot be excluded. The within-SAD analyses of the association between BSMAS severity and MCQ-30 subscale scores represent the most methodologically robust findings of the present study.

Table 3.

Comparison of sociodemographic characteristics in SAD and control groups

Sociodemographic Characteristics	SAD (n = 99)	Control (n = 112)	Test Statistic and Effect Size	p
Age (years)	Mean ± SD: 26.69 ± 7.82 Median (Min-Max): 18–57	Mean ± SD: 28.26 ± 5.87 Median (Min-Max): 17–57	t=-1.63 d = 0.22	0.104
Gender	n (%)	n (%)
Female	70 (70.7%)	86 (76.8%)	χ²= 1.008 V = 0.07	0.315
Male	29 (29.3%)	26 (23.2%)
Marital Status
Single	76 (76.8%)	64 (57.1%)	χ²= 13.816 V = 0.26	0.001**
Married	18 (18.2%)	46 (41.1%)
Other (Divorced/Widowed)	5 (5.0%)	2 (1.8%)
Education Level
High School or Below	38 (38.4%)	15 (13.4%)	χ²= 17.449 V = 0.29	< 0.001**
University or Above	61 (61.6%)	97 (86.6%)
Income Level
Below 5500 TL	44 (44.4%)	19 (17%)	χ²=35.261 V = 0.41	< 0.001**
5500–10,000 TL	11 (11.1%)	5 (4.5%)
10,000–15,000 TL	10 (10.1%)	4 (3.6%)
Above 15,000 TL	34 (34.3%)	84 (75.0%)
Smoking
No	69 (69.7%)	90 (80.4%)	χ²= 3.216 V = 0.12	0.073
Yes	30 (30.3%)	22 (19.6%)
Alcohol
No	70 (70.7%)	86 (76.8%)	χ²= 1.008 V = 0.07	0.315
Yes	29 (29.3%)	26 (23.2%)
Employment Status
Employed	44 (44.4%)	89 (76.8%)	χ²= 27.658 V = 0.36	< 0.001**
Unemployed	55 (55.6%)	23 (20.4%)
Physical Illness
No	83 (83.8%)	105 (93.8%)	χ²=5.315 V = 0.16	0.021*
Yes	16 (16.2%)	7 (6.2%)
Previous Psychiatric Diagnosis
No	46 (46.5%)	97 (86.6%)	χ²= 38.770 V = 0.39	< 0.001**
Yes	53 (53.5%)	15 (13.4%)
Family History of Psychiatric Illness
No	66 (66.7%)	84 (75%)	χ²=1.776 V = 0.09	0.183
Yes	33 (33.3%)	28 (25%)

Sociodemographic characteristics

The sociodemographic characteristics of the study participants are presented in Table 3 (p.37). The mean age was 26.69 years (SD = 7.82) in the SAD group and 28.26 years (SD = 5.87) in the control group (t=-1.63, p = 0.104). The proportion of female participants was 70.7% in the SAD group and 76.8% in the control group (χ²=1.008, p = 0.315). Significant between-group differences were observed for marital status, education level, income level, employment status, physical illness, and previous psychiatric diagnosis (Table 3).

Prevalence of meeting the operational SMA criterion

The proportion of participants meeting the operational SMA criterion was 52.5% in the SAD group and 46.4% in the control group. The difference was not statistically significant (χ² = 0.781, p = 0.377, φ = 0.06; Table 4). Using the stricter operational SMA definition (6/6 items ≥ 3), the between-group comparison similarly showed no significant difference (χ² = 0.617, p = 0.432, φ = 0.054; Table 1, Panel A).

Table 4.

Comparison of meeting SMA criterion based on bergen social media addiction scale (BSMAS) between the SAD and Control Groups

	SAD	Control	Test Statistic	p	Effect size
	n	n*
According to BSMAS
Not meeting SMA criterion	47	60	χ² = 0.781	0.377	φ = 0.06
Meeting SMA criterion	52	52

χ² : Chi-square test,

*analysis-specific n

Note: One recruited control participant had missing BSMAS data and was excluded from all analyses, including sociodemographic comparisons. Therefore, all control-group analyses are based on the final analytic sample of 112 control participants

Metacognitive beliefs associated with problematic social media use

Participants with SAD who met the operational SMA criterion scored higher on MCQ-30 uncontrollability of worry than control participants who met the same criterion (U = 943.0, p = 0.008, r = 0.26) (Table 5, p.30). Within the SAD group, participants who met the operational SMA criterion had higher MCQ-30 uncontrollability of worry (t=-2.265, p = 0.026, d = 0.46), need to control thoughts (t=-2.929, p = 0.004, d = 0.59), cognitive awareness (t=-2.458, p = 0.016, d = 0.49), and MCQ-30 total score (t=-2.241, p = 0.027, d = 0.45) compared with those not meeting the criterion (Table 6, p.26). In the SAD group, BSMAS total score was positively correlated with MCQ-30 uncontrollability of worry (r = 0.289, p = 0.004, 95% CI [0.097, 0.460]), need to control thoughts (r = 0.368, p < 0.001, 95% CI [0.184, 0.527]), cognitive awareness (r = 0.250, p = 0.013, 95% CI [0.055, 0.426]), and MCQ-30 total score (r = 0.319, p = 0.001, 95% CI [0.130, 0.486]) (Table 7). To examine whether metacognitive beliefs were associated with meeting the operational SMA criterion independently of sociodemographic differences, a multivariate binary logistic regression analysis was conducted within the SAD group. The model included SMA status (meeting vs. not meeting the operational SMA criterion) as the dependent variable. Education level, income level, and employment status were entered as covariates to control for their potential confounding effects. In the adjusted model, the MCQ-30 need to control thoughts subscale was significantly associated with meeting the operational SMA criterion (B = 0.173, SE = 0.088, OR = 1.19, 95% CI: 1.00–1.41, p = 0.049). Among the sociodemographic covariates, only the income category of 10,000–15,000 TL compared with < 5,500 TL was significant (B = − 2.741, SE = 1.345, OR = 0.06, 95% CI: 0.005–0.90, p = 0.042). All other MCQ-30 subscales and sociodemographic covariates were not statistically significant. Because of the cross-sectional design and small cell counts in some categorical covariates, these findings should be interpreted as associations rather than causal effects.

Table 5.

Comparison of MCQ-30 scores between participants with SAD and control participants meeting the operational SMA criterion

	SAD Participants Meeting the Operational SMA Criterion (n = 52)		Control Group Participants Meeting the Operational SMA Criterion (n = 52)*		Test Statistic p	Effect Size
	Mean ± SD	Median (Min-Max)	Mean ± SD	Median (Min-Max)
MCQ - Positive Beliefs About Worry	12.13 ± 3.92	12 (6–20)	12.82 ± 3.67	12.5 (6–22)	U = 1208.0 0.348	r = 0.09
MCQ - Uncontrollability of Worry	18.73 ± 3.47	19 (10–24)	16.42 ± 4.48	16.5 (8–24)	U = 943.0 0.008**	r = 0.26
MCQ - Cognitive Insecurity	14.86 ± 4.96	15 (6–24)	13.80 ± 4.56	15 (6–24)	U = 1200.5 0.323	r = 0.10
MCQ - Need to Control Thoughts	16.59 ± 3.94	17 (7–23)	14.98 ± 4.14	15.5 (6–23)	U = 1064.0 0.060	r = 0.18
MCQ - Cognitive Awareness	18.46 ± 2.69	18 (11–24)	17.30 ± 3.07	18 (10–23)	U = 1060.5 0.056	r = 0.19
MCQ - Total	80.78 ± 12.13	82.5 (44–104)	75.34 ± 13.95	75.5 (43–109)	t = 2.122 0.036*	d = 0.42

t = Independent Samples t-test

U = Mann-Whitney U test

**analysis-specific n

Table 6.

Comparison of liebowitz and MCQ-30 scores between participants meeting and not meeting the operational SMA criterion in the SAD group

	Not meeting criterion (n = 47)		Meeting criterion (n = 52)		Test Statistic p	Effect Size
	Mean ± SD	Median (Min-Max)	Mean ± SD	Median (Min-Max)
Liebowitz Total Anxiety Score	57.93 ± 15.04	54 (31–89)	60.07 ± 13.44	58 (36–95)	t=-0.747 0.457	d = 0.15
Liebowitz Total Avoidance Score	53.40 ± 14.99	51 (31–88)	55.51 ± 12.20	54 (35–95)	t=-0.773 0.442	d = 0.16
Liebowitz Total Score	111.34 ± 29.09	106 (62–177)	115.59 ± 24.88	115 (73–190)	t=-0.784 0.435	d = 0.16
MCQ - Positive Beliefs About Worry	12.86 ± 4.49	13 (6–22)	12.13 ± 3.92	12 (6–20)	t = 0.847 0.399	d = 0.17
MCQ - Uncontrollability of Worry	16.95 ± 4.30	17 (6–24)	18.73 ± 3.47	19 (10–24)	t=-2.265 0.026*	d = 0.46
MCQ - Cognitive Insecurity	13.65 ± 5.24	13 (5–24)	14.86 ± 4.96	15 (6–24)	t=-1.175 0.243	d = 0.24
MCQ - Need to Control Thoughts	14.29 ± 3.84	14 (6–22)	16.59 ± 3.94	17 (7–23)	t=-2.929 0.004**	d = 0.59
MCQ - Cognitive Awareness	17.02 ± 3.13	17 (8–24)	18.46 ± 2.69	18 (11–24)	t=-2.458 0.016*	d = 0.49
MCQ - Total	74.78 ± 14.48	76 (46–109)	80.78 ± 12.13	82 (44–104)	t=-2.241 0.027*	d = 0.45

t = Independent Samples t-test. Effect size calculated by Cohen’s d

Table 7.

Correlations between BSMAS total score, metacognitions (MCQ-30), and social anxiety symptoms (LSAS) in the SAD Group

Variables	Bergen SMAS Total Score
	R*	p	Effect Size/Cl
Liebowitz Total Anxiety Score	0.155	0.126	95% CI [-0.044, 0.342]
Liebowitz Total Avoidance Score	0.190	0.059	95% CI [-0.008, 0.373]
Liebowitz Total Score	0.179	0.076	95% CI [-0.019, 0.364]
MCQ - Positive Beliefs About Worry	-0.020	0.842	95% CI [-0.217, 0.178]
MCQ - Uncontrollability of Worry	0.289	0.004**	95% CI [0.097, 0.460]
MCQ - Cognitive Insecurity	0.154	0.128	95% CI [-0.045, 0.341]
MCQ - Need to Control Thoughts	0.368	< 0.001**	95% CI [0.184, 0.527]
MCQ - Cognitive Awareness	0.250	0.013*	95% CI [0.055, 0.426]
MCQ - TOTAL	0.319	0.001**	95% CI [0.130, 0.486]

95% CIs were calculated using Fisher’s z transformation

Discussion

In this cross-sectional case–control study, the prevalence of SMA did not differ significantly between participants with SAD and control participants without current psychiatric disorder (52.5% vs. 46.4%; χ²=0.781, p = 0.377), with a small effect size (φ = 0.06) (Table 4). Within the SAD group, SMA was associated with higher negative metacognitive beliefs, particularly uncontrollability of worry and the need to control thoughts (Table 6), and BSMAS severity correlated with these metacognitive domains (Table 7). Among participants meeting the operational SMA criterion, participants with SAD showed higher MCQ-30 uncontrollability of worry scores than control participants (Table 5). Sensitivity analyses using an alternative BSMAS threshold showed the same overall pattern (Table 1).

Our prevalence findings are broadly consistent with previous studies conducted in Türkiye and in other cultural settings. A meta-analysis reported that the prevalence of SMA ranged between 0% and 82%. However, the prevalence of SMA was found to be between 18 and 31% in studies in which those who scored 3 or more points from at least 4 of the 6 items suggested by Andreassen et al., were evaluated as SMA [21]. The prevalence of SMA in the SAD and control groups participating in our study was higher than the average of the studies using a similar cut-off score (24%). Cultural and social differences, access to technology, inclusion criteria, and the timing of data collection may partly explain this difference.

Although the proportion of participants meeting the operational SMA criterion was numerically higher in the SAD group than in the control group, this difference was not statistically significant (Table 4, p. 25). One possible explanation is that social anxiety and social media use patterns may be related in different directions depending on how social media is used. While active use of social media is negatively associated with social anxiety, more passive use of social media is thought to be positively associated with social anxiety [22]. Active social media use, such as sharing thoughts and interacting with others online, may reduce loneliness and social anxiety for some individuals [23]. Passive social media use, such as viewing others’ updates or photos, may be associated with lower well-being, reduced social interaction, and body image concerns (Burke et al., 2010). In addition, studies have suggested that active social media use may benefit interpersonal relationships and positively affect communication capacity in university students, whereas passive social media use may reduce the individual’s communication capacity by replacing interactions in daily life [9]. In our study, participants meeting the operational SMA criterion more often reported using social media to relieve discomfort in social settings, reduce boredom, or listen to music, rather than to meet new people or engage in interpersonal communication (Table 8, p.28). In line with the literature, these findings, based on self-reported usage purposes, may indicate that participants with SAD meeting the operational SMA criterion use social media less for interpersonal engagement; however, active versus passive use patterns were not directly measured.

Table 8.

Comparison of social media usage purposes between participants meeting and not meeting the operational SMA criterion in the SAD group

	Not meeting criterion (n = 47)		Meeting criterion (n = 52)	Test Statistic p	Effect Size
	n	%	n	%
Communicating	24	51.1	36	69.2 χ² = 3.413 0.065	V = 0.19
Passing Time	22	46.8	43	82.7 χ² = 14.097  < 0.001**	V = 0.38
Wanting to be Alone	4	8.5	10	19.2 χ² = 2.337 0.128	V = 0.15
Shopping	8	17.0	17	32.7 χ² = 3.212 0.073	V = 0.18
Listening to Music	17	36.2	35	67.3 χ² = 9.598 0.002**	V = 0.31
Entertainment/Relaxing	21	44.7	40	76.9 χ² = 10.851 0.001**	V = 0.33
Meeting New People	3	6.4	6	11.5 χ² = 0.794 0.492	V = 0.09
Following the News	22	46.8	33	63.5 χ² = 2.773 0.096	V = 0.17
Easy Access to Information	15	31.9	25	48.1 χ² = 2.678 0.102	V = 0.16
Exchanging Ideas	6	12.8	10	19.2 χ² = 0.761 0.383	V = 0.09
Feeling Uncomfortable in Social Settings	3	6.4	12	23.1 χ² = 5.352 0.021*	V = 0.23

χ² : Chi-square test/Fisher’s Exact test. Effect size calculated by Cramer’s V

In behavioural addictions, it has been suggested that positive metacognitions are involved in the onset of addiction by increasing the motivation of individuals towards behaviour, while negative metacognitions play a role in the continuity of these behaviours ([24], Sun et al., 2017). In a study published by [25], all negative and positive metacognitions were found to be associated with problematic social media use. Anxious feelings and thoughts about the lack of social skills may activate positive metacognitions that social media use helps to distract the individual’s mind from distress or to control such disturbing thoughts, thus making social media use a self-regulation tool [26]. In our study, no relationship was found between positive beliefs about worry and SMA. This may be because the patients were in the maintenance phase of behavioural addiction. However, it is also important that especially negative metacognitions are associated with psychopathology [26]. Our findings are consistent with previous studies suggesting that negative metacognitive beliefs may be related to more problematic patterns of addictive behaviors.

Among participants meeting the operational SMA criterion, participants with SAD showed higher MCQ-30 uncontrollability of worry scores than control participants. However, this between-group finding should be interpreted cautiously because the SAD and control groups differed on several sociodemographic characteristics that may influence both social media use and metacognitive beliefs (Table 5, p.30). In social anxiety, beliefs about being “inadequate” or “powerless,” together with perceived difficulty controlling anxiety, may be related to negative metacognitive beliefs about uncontrollability. Similar to other behavioral addictions, these beliefs may be associated with repetitive social media use as a perceived strategy for regulating distress, despite possible negative consequences [27]. When the relationship between Internet gaming disorder and SAD was analyzed, it was suggested that the fear of appearing inadequate leading to the persistence of SAD symptoms may be uncontrollable and the individual tries to prevent these thoughts by distancing from reality through gaming [26]. In addition, metacognitions related to the uncontrollability of thoughts in these individuals may lead to limited efforts to regulate one’s thoughts about gaming disorder and inadequacy in social areas [12]. Similar to the relationship between Internet gaming disorder and SAD, our findings suggest that negative metacognitions may represent a clinically relevant associated factor in the context of SMA within SAD. However, given the cross-sectional design, these associations do not establish temporal direction or causality, and longitudinal confirmation is required before any causal inference can be drawn.

The most consistent findings concerned negative metacognitive beliefs, particularly beliefs about the uncontrollability of worry and the perceived need to control thoughts. Importantly, BSMAS severity within SAD appears to be associated less with the overall severity of social anxiety symptoms (as indexed by LSAS scores, which did not differ between participants meeting and not meeting the SMA criterion) and more with metacognitive processes relating to the uncontrollability or harmfulness of internal experiences and the perceived imperative to regulate one’s own thoughts. These specific metacognitive processes may be involved in a self-reinforcing cycle in which social media use temporarily reduces distress associated with perceived thought uncontrollability, but the relief is transient and continued use may itself come to be appraised as difficult to control, thereby deepening the cycle. This interpretation aligns with the metacognitive model [28] but requires prospective and experimental confirmation.

Several clinical and demographic characteristics also differed according to SMA status within the SAD group. When the sociodemographic data of the SAD group were analyzed in our study, we found that participants meeting the operational SMA criterion were younger and more frequently unemployed, and differed in income level; marital status showed only a borderline difference (Table 9, p.32). In the literature, it has been shown that SMA is more likely to be seen in young individuals, low income groups and low education levels, but it has been reported that the absence of a romantic relationship has a low effect on SMA even if it is statistically significant [14]. Among participants with SAD, social withdrawal may be associated with greater reliance on online environments. These sociodemographic patterns may reflect differential opportunities, daily routines, or psychosocial stressors associated with problematic social media use; however, these interpretations should be considered exploratory.

Table 9.

Sociodemographic characteristics of participants meeting and not meeting the operational SMA criterion in the SAD group

Sociodemographic Characteristics	Not meeting criterion (n = 47)	Meeting Criterion (n = 52)	Test Statistic and Effect Size	p
Age (years)	Mean ± SD: 28.55 ± 8.75 Median (Min-Max): 19–57	Mean ± SD: 25.01 ± 6.51 Median (Min-Max): 18–51	t = 2.291 d = 0.46	0.024*
Gender	n (%)	n (%)
Female	30 (63.8%)	40 (76.9%)	χ²=2.043 V = 0.14	0.153
Male	17 (36.2%)	12 (23.1%)
Marital Status
Single	31 (66.0%)	45 (86.5%)	χ²=5.921 V = 0.24	0.051
Married	12 (25.5%)	6 (11.5%)
Other (Divorced/Widowed)	4 (8.5%)	1 (1.9%)
Education Level
High School or Below	16 (34.0%)	22 (42.3%)	χ²=0.713 V = 0.08	0.398
University or Above	31 (66.0%)	30 (57.7%)
Income Level
Below 5500 TL	18 (38.3%)	26 (50.0%)	χ²=9.900 V = 0.32	0.019*
5500–10,000 TL	3 (6.4%)	8 (15.4%)
10,000–15,000 TL	9 (19.1%)	1 (1.9%)
Above 15,000 TL	17 (36.2%)	17 (32.4%)
Smoking
No	30 (63.8%)	39 (75.0%)	χ²=1.458 V = 0.12	0.227
Yes	17 (36.2%)	13 (25.0%)
Alcohol Use
No	36 (76.6%)	34 (65.4%)	χ²=1.498 V = 0.12	0.221
Yes	11 (23.4%)	18 (34.6%)
Employment Status
Employed	26 (55.3%)	18 (34.6%)	χ²=4.286 V = 0.21	0.038*
Unemployed	21 (44.7%)	34 (65.4%)
Physical Illness
No	41 (87.2%)	42 (80.8%)	χ²=0.761 V = 0.09	0.383
Yes	6 (12.8%)	10 (19.2%)
Previous Psychiatric Diagnosis
No	26 (55.3%)	27 (51.9%)	χ²=0.114 V = 0.03	0.735
Yes	21 (44.7%)	25 (48.1%)
Family History of Psychiatric Illness
No	15 (31.9%)	18 (34.6%)	χ²=0.081 V = 0.03	0.776
Yes	32 (68.1%)	34 (65.4%)

In our sample, no difference was found between the severity of social anxiety symptoms assessed by the LSAS when participants meeting and not meeting the operational SMA criterion were compared within the SAD group (Table 7, p.35). To the best of our knowledge, SMA has not been analyzed in individuals diagnosed with SAD in the literature. In studies conducted without a diagnosis, an increase in LSAS scores was associated with SMA [29]. However, there are also studies suggesting that social anxiety symptom severity does not directly lead to SMA and problematic social media use may be indirectly related to social anxiety through interpersonal difficulties and experiential avoidance [30]. Taken together, these findings suggest that problematic social media use in SAD may be more closely related to beliefs about the uncontrollability or harmfulness of internal experiences and the perceived need to regulate thoughts than to the overall severity of social anxiety symptoms alone. Nevertheless, constructs such as experiential avoidance were not directly assessed in this study, and this interpretation should therefore be considered preliminary.

Overall, the proportion of participants meeting the operational SMA criterion did not differ significantly between the SAD and control groups in our sample. However, this proportion was higher than that reported in several studies using similar BSMAS thresholds, suggesting that problematic social media use may warrant clinical and public health attention.

Limitations

It is important to consider several limitations when interpreting the findings of this study. First, the cross-sectional design limits the ability to draw causal conclusions regarding the relationships between social anxiety, metacognitions, and social media addiction. Accordingly, the directionality of these associations cannot be determined. Future research should prioritize longitudinal designs to clarify temporal pathways and to examine whether metacognitive interventions may reduce problematic social media use among individuals with SAD.

Second, we did not directly assess active versus passive social media use patterns or specific mechanisms such as avoidance motives or experiential avoidance. Therefore, interpretations involving “avoidance” should be regarded as tentative, and future studies incorporating validated avoidance measures and objective usage indicators (e.g., screen-time logs) would help clarify these mechanisms. In addition, SMA status was defined using a self-report operational criterion; although we conducted sensitivity analyses using a stricter definition, some degree of misclassification may still be possible.

Third, the dependence on self-report measures might potentially introduce response biases, such as social desirability or recall bias. Such biases may affect the accuracy of symptom and behavior reporting. In order to enhance the validity of self-report questionnaires, future research should focus on integrating objective indicators of social media usage, such as digital usage data.

Fourth, the SAD and control groups differed on some sociodemographic characteristics (e.g., education, income, and employment), which may confound between-group comparisons; thus, residual confounding cannot be ruled out. Moreover, because the sample was recruited from a single outpatient clinic at one hospital, generalizability to other clinical settings, community samples, and cultural contexts may be limited.

Finally, the assumed 2-point difference used for the a priori power calculation was a pragmatic planning assumption, as a validated MCID for the BSMAS is not currently established.

Conclusion

This study is, to our knowledge, among the first to examine problematic social media use and its metacognitive correlates in a clinical sample of individuals with social anxiety disorder (SAD). While the proportion of participants meeting the operational SMA criterion was similar between the SAD and control groups, the most robust findings were observed within the SAD group: BSMAS severity was positively associated with negative metacognitive domains.

These findings suggest that specific negative metacognitive beliefs, particularly beliefs about the uncontrollability of worry and the need to control thoughts, may be treatment-relevant correlates of problematic social media use in SAD. However, given the cross-sectional design, causal interpretations cannot be made, and mechanisms such as avoidance or experiential avoidance were not directly assessed. Future longitudinal and interventional studies incorporating measures of avoidance and objective usage patterns are needed to clarify directionality and mechanisms.

Abbreviations

SMA

Social Media Addiction

SAD

Social Anxiety Disorder

BSMAS

Bergen Social Media Addiction Scale

MCQ-30

Metacognitions Questionnaire-30

LSAS

Liebowitz Social Anxiety Scale

DSM-5

Diagnostic and Statistical Manual of Mental Disorders 5th Edition

Author contribution

Authors (Oğuzhan Uğur, Ali Çayköylü) confirm sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.

Funding

This research was not awarded any dedicated grant from any public, commercial, or not-for-profit funding agency.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was carried out in compliance with prevailing legal and ethical norms. The study was approved by the Dr. Abdurrahman Yurtaslan Ankara Oncology Health Practice and Research Center Ethics Committee on 25 January 2023, approval number 2023-01/16. In compliance with the Declaration of Helsinki, all volunteers involved in the study were provided with information regarding the study and their written informed consent was acquired.

Consent for publication

Each individual participant included in the study provided informed consent about publication.

Competing interests

The authors declare no competing interests.