Individuals with BDD face elevated risks of suicidal ideation and attempts, as outlined in a meta-analysis by Angelakis et al. (2016). Moreover, BDD often coexists with major depressive disorder (MDD; Gunstad & Phillips, 2003), which is itself strongly associated with suicidality (Brown et al., 2000; Cai et al., 2021; Dong et al., 2018). While MDD is the most common comorbidity of BDD (Phillips et al., 2007a), most individuals with MDD do not concurrently experience BDD (Brawman-Mintzer et al., 1995). Previous findings suggest that BDD and comorbid major depression are associated with increased suicidality (e.g., Phillips et al., 2005c). Specifically, in their longitudinal study of 200 individuals with BDD, lifetime suicidal ideation, suicide attempts, comorbidities and functional impairment were assessed using structured interviews. Logistic regression analyses showed high lifetime rates of suicidal ideation (78%) and attempts (27.5%). Functional impairment due to BDD predicted suicidal ideation and attempts. Comorbid MDD was linked specifically to suicidal ideation. These results suggest that multiple factors, including MDD, contribute to suicidality in BDD. However, a mediation model was not tested in the study by Phillips et al. (2005c). Shaw et al. (2016) investigated potential mediators of the relationship between BDD symptoms and suicidality based on Joiner’s (2005) Interpersonal Theory of Suicide (IPTS). According to IPTS, suicidal desire emerges primarily from two psychological states: thwarted belongingness, the feeling of not being meaningfully connected to others and perceived burdensomeness, the belief that one is a burden to others. In their cross-sectional mediation model, Shaw and colleagues (2016) found that the association between BDD symptom severity and both key constructs was fully mediated by depressive symptoms. These findings imply that BDD symptoms may contribute to suicidality indirectly via increased depressive symptomatology, which in turn amplifies feelings of social disconnection and burden. In other words, based on their results, no direct link between BDD symptoms and suicidality would be expected once depression is accounted for. In contrast to this finding, several studies found a significant association between BDD symptoms and suicidality even after controlling for depressive symptoms, suggesting that BDD symptoms may independently contribute to heightened suicidality (Krebs et al., 2022; Snorrason et al., 2019).

While previous research has provided initial insights into the relationship between BDD, depressive symptoms and suicidality, findings remain inconsistent. Phillips et al. (2005c) identified MDD and functional impairment as significant independent predictors of suicidal ideation in BDD. Notably, their design and analyses did not assess mediation directly, leaving it unclear whether depressive symptoms explain the association between BDD severity and suicidality. Extending this line of research, Shaw et al. (2016) found evidence that the association between BDD symptoms and thwarted belongingness and perceived burdensomeness was fully mediated by depressive symptoms, suggesting an indirect pathway from BDD severity via depression to suicidal ideation. However, suicidality was not assessed directly in their study, limiting the conclusions regarding actual suicidal thoughts or behaviors. In contrast, other studies (e.g., Krebs et al., 2022; Snorrason et al., 2019) demonstrated that the association between BDD and suicidality persists even after controlling for depressive symptoms, pointing toward the existence of a direct effect in addition to any potential mediation through depression. These methodological and conceptual differences across studies likely contributed to the varying effects observed, highlighting the need for further research to clarify the interplay between BDD, depression, and suicidality. Thus, the present study seeks to clarify whether the association between BDD symptoms and suicidality is fully or only partially mediated by depressive symptoms. To address this, the total effect of BDD symptoms on suicidality will be decomposed into its direct and indirect components (via depressive symptoms), allowing for a more precise estimation of the magnitude of each effect. In contrast to Shaw et al. (2016), who examined risk factors for suicidality (thwarted belongingness and perceived burdensomeness), the aim of our study was to investigate suicidality, including suicidal thoughts and attempts. This deliberate distinction is essential for gaining a clearer understanding of the relationships between BDD symptomatology, depressive symptoms, and suicidality. Furthermore, these findings may hold clinical relevance by informing whether depressive symptoms should be a primary treatment target in reducing suicide risk in individuals with BDD — or whether BDD symptomatology itself constitutes an independent risk factor that requires direct therapeutic attention. To further explore the relationship between the three variables the following hypotheses are formulated:

Between February 2016 and May 2018, and from March 2019 to June 2021, a total of 15,273 individuals accessed the self-test, with 3,452 completing it. Exclusions were made for individuals who did not consent to data use (n = 808), participated repeatedly (n = 89), lacked sufficient language skills (n = 5), were under 18 years old (n = 78), or did not meet the diagnostic criteria for BDD (n = 1,216) according to the self-report measure BDD-5 (Möllmann & Buhlmann, 2024), which is based on DSM-5 criteria (APA, 2013). To differentiate BDD from eating disorder (ED) symptoms, participants completed the SCOFF questionnaire (Morgan et al., 1999). It consists of five yes/no items assessing disordered eating behaviors (e.g., “Do you worry you have lost control over how much you eat?”). A cut-off of two or more affirmative responses has shown good diagnostic accuracy, with pooled sensitivity of .80 and specificity of .93 (Botella et al., 2013). Participants who scored above this cut-off were excluded from the analyses. Consequently, the final sample size was n = 1,256, ranging from 18 to 71 years of age (M = 30.1, SD = 9.9). The majority of participants (72.3%, n = 908) were female (see Table 1). Participants’ appearance concerns referred to: Skin (55.6%, n = 698), nose (48.3%, n = 607), hair (41.9%, n = 526), breast (31.0%, n = 389), mouth (23.6%, n = 296), genitals (22.2%, n = 279), eyes (22.1%, n = 278), hands (12.9%, n = 162), muscles (12.3%, n = 155), ears (8.3%, n = 104), body characteristics related to ethnicity (3.0%, n = 38), and other (18.5%, n = 232). The five most prevalent diagnoses reported, based on prior reported clinical professional assessments, were depressive disorders (43.6%, n = 547), social anxiety disorder (17.4%, n = 219), BDD (12.4%, n = 156), generalized anxiety disorder (10.8%, n = 136), and ED (11.9%, n = 149). Nearly two-thirds (63.6%, n = 799) reported having received at least one diagnosis of a mental disorder.

Demographic data were collected on the variables age, gender, place of residence (country), highest educational degree, current occupation, and marital status (see Table 1).

Although this study did not include clinical interviews, we carefully assessed BDD using the BDD-5 (Möllmann & Buhlmann, 2024), a well-developed self-report questionnaire based on DSM-5 criteria, consisting of six dichotomous items (agree/disagree). It should be noted though that while this allows for an approximation of diagnostic criteria, it does not replace a formal clinical diagnosis. The severity of BDD symptoms within the past week was assessed using the Body Dysmorphic Symptoms Questionnaire (Fragebogen körperdysmorpher Symptome: FKS) developed by Buhlmann et al. (2009). It comprises two subscales: Item 1 and Items 4-15, which queries the extent of BDD symptoms and BDD-related dysfunctional behaviors, and Items 16-18, which includes items regarding aesthetic surgeries and suicidality. Items 2 and 3 do not contribute to the total score, and were therefore not considered in further analyses, as Item 2 assesses body areas relevant to the symptoms and Item 3 serves to differentiate BDD from EDs (Buhlmann et al., 2009). The remaining items refer to symptoms within the past week and are rated on a 5-point Likert scale from 0 (not at all/never/don’t think about it at all) to 4 (very much/more than 5 times/more than 8 hours per day).¹1

Furthermore, to assess BDD symptoms in our sample, Items FKS 17 and 18, which measure suicidality, were included in the total score. For the mediation analyses, however, these items were excluded to avoid confounding effects on the dependent variable Suicidality. Similarly, to assess depressive symptoms, PHQ Item 9, which measures suicidality, was excluded in the mediation analyses.

The Patient Health Questionnaire (PHQ-9; Löwe et al., 2002; Spitzer et al., 1999) measures depressive symptoms over the past two weeks. The scale consists of nine items that assess depressive symptoms over the past two weeks on a 4-point Likert scale (0 = not at all, 1 = several days, 2 = more than half the days, 3 = nearly every day). The PHQ-9 serves as a screening tool for MDD and is based on the criteria for a major depressive episode according to DSM-5 (APA, 2013; sensitivity = .95 and specificity = .86; Gräfe et al., 2004). The criteria for MDD are considered met if at least five of the Items 1-9, including either Item 1 or 2, are rated as occurring more than half of the days. In the present study, only Items 1-8 of the PHQ-9 were used to measure depressive symptoms¹. Internal consistency of the PHQ-9 in the current sample was α = .86.

Suicidal thoughts and actions were assessed using FKS Items 17 ("Have you ever thought about ending your life because you find parts of your physical appearance so ugly?"), and 18 ("Have you ever attempted suicide because you find parts of your physical appearance so ugly?"), as well as PHQ-9 Item 9 ("How often have you been bothered by the following problems over the past two weeks: Thoughts that you would be better off dead or of hurting yourself?"; (hereinafter referred to as FKS_17, FKS_18, and PHQ_9).

The data was analyzed using IBM SPSS Statistics (Version 29.0; IBM Corporation, 2022) for descriptive data analyses, and MPlus, version 8.7 (Muthén & Muthén, 2017) for factor analyses and structural equation modeling (SEM). These analyses were conducted using Item Response Theory (IRT) models for categorial items with a robust weighted least squares mean and variance (WLSMV) adjusted estimator (Brown, 2006). To test the factorial structure of the FKS and the PHQ-9 after extracting the items measuring suicidality, exploratory factor analyses (EFA) were conducted. To determine the number of factors, the Kaiser criterion (Guttman, 1954; Kaiser & Dickmann, 1959), and the interpretability of factor structure were considered. The model quality was assessed using the Root Mean Square Error of Approximation (RMSEA; Steiger, 1980), Comparative Fit Index (CFI; Bentler, 1990) and Tucker-Lewis Index (TLI; Tucker & Lewis, 1973). A RMSEA < .05 indicates a good model fit, < .08 an acceptable model fit, and > .10 a poor model fit (Browne & Cudeck, 1992). For CFI and TLI, acceptable model fit criteria were set as ≥ .95 (Hu & Bentler, 1999). Confirmatory factor analyses (CFA) were subsequently estimated for scale adjustment. First, items with factor-loading < .35 were excluded. Then, modification indices (MI) were inspected to detect model misspecifications (starting with the MI indicating the highest change in Chi-square-statistic). Suggested residual correlations were stepwise freely estimated only if they could be interpreted meaningfully in terms of content. Another CFA was performed to examine whether the three extracted items asking for suicidality (FKS_17, FKS_18, and PHQ_9) could also be represent as latent construct, considering the variable suicidality in mediation analyses both as manifest items, and as a latent construct.

Four structural equation models (SEMs; Models 1a, 2a, 3a, 4a) were estimated (Geiser, 2011; Haider et al., 2023) to test a possible mediation between BDD symptoms and suicidality through depressive symptoms. Suicidality was operationalized differently: By using the manifest items FKS_17 (Model 1a with FKS_17), FKS_18 (Model 2a with FKS_18), and PHQ_9 Model 3a with PHQ_9), and by using the latent factor suicidality, represented by the three items (Model 4a with SUI).

In a second step, all four models were re-estimated including gender and age as covariates to address potential confounding influences (Models 1b, 2b, 3b, 4b). Direct, indirect, and total effects were computed using the WLSMV estimation statistic via the Sobel test (Kleinke et al., 2017). Standardized effect sizes and model quality measures were reported. Bias-corrected bootstrap method with m = 10000 were employed to calculate confidence intervals for effects without assuming normal distribution (Kleinke et al., 2017; MacKinnon, 2012). A significance level of α = .05 was set for all analyses.

The average sum score of the FKS was M = 37.3 (SD = 8.7). Almost all participants achieved the cut-off value of 23 (94.7%, n = 1,189). The mean sum score of the PHQ-9 was M = 13.4 (SD = 6.2). Almost half (49.8%, n = 625) of the sample met the criteria for a current MDD according to the PHQ-9. Two thirds of the participants (67.0%, n = 842) reported appearance-related suicidal thoughts (FKS_17). Of these, 23.0% had strong suicidal thoughts, and 17.0% had very strong suicidal thoughts. In our sample, general suicidal thoughts occurred in 59.6% (n = 748) of the people (PHQ_9). Of these people, 15.2% suffered from suicidal thoughts on more than half of the days, and 17.1% almost every day. Every twelfth person (8.5%, n = 107) stated that they had attempted suicide in the past at least once because of their appearance (FKS_18; see Table 2 for the response frequencies of the items on suicidality).

The computed models are reported below (see Table 3 and in the Supplementary Materials, Appendix 4). The most comprehensive model, including the latent variable suicidality with gender and age as covariates, is shown in Figure 1.

The hypothesis that depression mediates the relationship between BDD symptoms and suicidality was supported by our data. Specifically, BDD symptoms were consistently positively associated with depressive symptoms across all models (β = .70), aligning with prior research linking BDD severity to increased depression (Phillips et al., 2007a) as well as with epidemiological evidence documenting the high prevalence of MDD among individuals with BDD (Gunstad & Phillips, 2003; Phillips et al., 2005a; 2007b). In turn, depressive symptoms showed a significant direct effect on suicidality, consistent with previous findings indicating that depression is a key correlate of suicidal ideation and behavior (Brown et al., 2000; Snorrason et al., 2019) and that there are elevated rates of suicidality in MDD (Cai et al., 2021; Dong et al., 2018). The association between depression and suicidality varied depending on the suicidality measure (Model 1a with FKS_17, Model 2a with FKS_18, Model 3a with PHQ_9, Model 4a with SUI), with the strongest effect observed for general suicidal thoughts (Model 3a, β = .62), comparable to the finding by Shaw et al. (2016). Regarding the mediation hypothesis, small indirect effects via depression were found between BDD symptoms and appearance-related suicidal thoughts (Model 1a, β = .20) and between BDD symptoms and appearance-related suicide attempts (Model 2a, β = .20). Interestingly, a moderate indirect effect via depression was found between BDD symptoms and general suicidal thoughts (Model 3a, β = .43). The finding of small indirect effects for appearance-related suicidality and a moderate indirect effect for general suicidality may point to the potential importance of distinguishing between appearance-related and general suicidal thoughts in individuals with BDD. However, it should be noted that a potential confounding factor is that the items used as dependent variables (FKS_17, FKS_18, PHQ_9) were extracted from questionnaires (FKS, PHQ-9) that also contributed to the independent variables or mediators in the models. This could potentially inflate effects between the affected constructs due to methodological artifacts. To address this, we constructed a composite suicidality variable (SUI; see Models 4a/4b in Figure 1, Table 3, and in the Supplementary Materials, Appendix 4) by combining the indicators of both questionnaires into a latent construct to reduces potential bias from item-level overlap by capturing the shared variance among the three indicators. This model revealed a moderate indirect effect of BDD symptoms on the latent variable suicidality (Model 4a, β = .35).

In sum, our results are partly consistent with prior research indicating that depressive symptoms partially (but not fully) mediate the association between BDD and suicidality, as reported by Shaw et al. (2016). However, unlike Shaw et al. (2016), who recruited participants based on self-reported appearance-related concerns without formally establishing a clinical diagnosis of BDD and investigated constructs theoretically linked to suicidal desire rather than suicidality directly, our study included individuals based on a validated self-report screening tool assessing established diagnostic criteria for BDD and employed latent measures of suicidality.

The second hypothesis was statistically supported, showing a direct association of BDD symptoms on suicidality across all eight models, consistent with previous research (Krebs et al., 2022; Phillips et al., 2005c; Snorrason et al., 2019). A strength of our study was that we estimated the magnitude of the total effects, as well as the direct and indirect effects. The direct association between BDD symptoms on appearance-related suicidal thoughts was stronger (Model 1a, β = .42) than the association with appearance-related suicide attempts (Model 2a, β = .28). This discrepancy might reflect the general difficulty in predicting suicide attempts compared to suicidal thoughts (Glenn & Nock, 2014; Ribeiro et al., 2019). Remarkably, a small but significant direct association was found between BDD symptoms and general suicidal thoughts (Model 3a, β = .10), diverging from findings by Shaw et al. (2016), who reported a full mediation via depression. However, Shaw et al. (2016) only examined the relationship between BDD symptoms and perceived burdensomeness and thwarted belongingness. These constructs may be more closely linked to depression than direct assessments of suicidality, potentially resulting in stronger predictability by depression. Overall, these findings suggest that depressive symptoms may not entirely account for suicidality in BDD.

Across all adjusted models (Models 1b–4b), adding gender and age as covariates did not substantially alter the size or significance of the direct and indirect effects (see Supplementary Materials, Appendix 4 and Figure 1).

BDD is characterized by perceived defects of one’s own physical appearance (Veale et al., 2016), which is often accompanied by rigid and sometimes delusional appearance beliefs. Affected individuals often feel misunderstood because they perceive something in a way that others do not (Brito et al., 2014). This may lead to increased hopelessness or increased risk of suicidality. Based on our findings, suicidality should be assessed in patients with BDD, regardless of comorbid depressive symptoms. In terms of clinical interventions, it is important to address BDD-typical risk factors (e.g., social withdrawal due to feelings of negative evaluation by others, not belonging to a social group or being a burden to others) into the treatment. General depression-focused interventions might not sufficiently capture the risk factors associated with suicidality in BDD.

This study has several limitations. First, BDD symptoms were assessed using self-report measures. This limits diagnostic accuracy and raises the risk of overlap with related disorders such as social anxiety disorder or obsessive-compulsive disorder (Schulte et al., 2020). Thus, future studies should incorporate structured clinical interviews to ensure more accurate diagnostic BDD assessment, including the assessment of possible comorbidities. It should be noted though that our sample had very comparable levels of BDD symptom severity and suicidality to those found in diagnosed clinical samples, which supports the idea that online studies can be a useful tool to particularly reach those individuals with BDD who might otherwise not be able to be reached (e.g., due to treatment barriers, being housebound). Second, the operationalization of suicidality was limited in parts of the study. In Models 1a/b, 2a/b, and 3a/b, suicidality was assessed using single items (FKS_17, FKS_18, PHQ_9), which may have led to skewed response distributions, potentially affecting the reliability and sensitivity of the measure. Moreover, single-item measurement is generally considered less reliable and more vulnerable to method effects, as items from the same questionnaire may correlate more strongly with each other than with conceptually similar items from different sources (Nunnally & Bernstein, 1994). However, in the SUI Model (4a/b), suicidality was modeled as a latent variable based on three indicators (FKS_17, FKS_18, PHQ_9), which partially reduces these limitations by allowing for more robust construct modeling. To further improve measurement precision, future studies should implement validated multi-item scales such as the Suicidal Ideation and Behavior Scale (SSEV; Teismann et al., 2021), which offer a more differentiated assessment of suicidal thoughts and behaviors. Third, the cross-sectional design of our study precludes conclusions about temporal or causal relations. Longitudinal research is needed to determine the directionality and potential bidirectionality of the relationships between BDD symptoms, depression, and suicidality. Lastly, the sample was predominantly female, limiting generalizability to male or gender-diverse individuals. To increase external validity, future studies should aim for more gender balanced samples.

Nonetheless, the present study includes a substantially larger sample size compared to earlier research (e.g., Phillips et al., 2005c, N = 200; Shaw et al., 2016, N = 235 participants), which enhances the statistical power and generalizability of our findings.

With respect to future research, additional psychosocial risk factors such as self-esteem (Kuck et al., 2021; Soto-Sanz et al., 2019), bullying and childhood trauma (e.g. abuse and neglect; Rück et al., 2024), or reduced quality of life (de Freitas Melo et al., 2022; Phillips et al., 2005b) should be considered as further explanatory variables in future models to better understand the complex mechanisms underlying suicidality in BDD.

The study underscores the elevated risk of suicidality among individuals with BDD, emphasizing the importance for clinical professionals to carefully assess suicidality when encountering signs of BDD. While depressive symptoms partially mediate the relationship between BDD symptoms and suicidality, the study suggests that BDD symptoms are associated with increased suicidality irrespective of depressive symptoms. This highlights the need for simultaneous treatment of both BDD-related symptoms and suicidality (Dozois & Rnic, 2015; Wilhelm et al., 2012). However, due to the cross-sectional design, causal conclusions cannot be drawn. Future research is needed to examine the association of BDD and suicidality using different methods and designs including clinician-administered assessment methods and longitudinal designs.