In the last two decades, the number of studies focused on the mediators connecting insecure attachment with Eating Disorders (EDs), at both clinical and sub-clinical level, has considerably increased. However, there has not been a systematic synthesis of this literature to date. To fill this gap, the current meta-analytic review aimed at identifying and quantifying the extent to which mediators contribute to the explanation of this relationship. The present study was registered with PROSPERO (CRD42017076807). A comprehensive search process in seven different electronic databases retrieved 24 studies that examined how insecure attachment leads to ED symptoms through mediation analysis. Standardized regression coefficients of the indirect and total paths of 21 mediation models were pooled. Studies were coded and ranked for quality. We found evidence to show that maladaptive emotion regulation and depressive symptoms had the highest effect size for mediation (mediation ratio [P_M] = 0.71). Further, body dissatisfaction, neuroticism, perfectionism, mindfulness and social comparison had significant, but moderate to low mediating effects (P_M = 0.21–0.58). The methodological quality of these studies was mostly low to moderate and potential areas for development were highlighted. Our findings support the direct targeting of these psychological constructs in prevention programs and treatment of EDs. Future investigations addressing the time sequence between the variables will provide valuable clues to untangle the prospective contribution of each variable on the development and maintenance of eating pathology.

Objective: Despite findings from numerous cross‐sectional studies suggesting a substantial association between loneliness and different types of disordered eating, much remains unknown about the impact of confounding, the order of cause and effect, and gender differences in the relationship. Thus, this study followed a large, population‐based, mixed‐gender sample through adolescence and young adulthood, applying a random intercept cross‐lagged panel model (RI‐CLPM) approach to examine the bidirectional prospective associations between loneliness and disordered eating while ruling out the effect of unmeasured time‐invariant confounders.Method: A Norwegian sample of N = 2933 adolescents (54.2% female) was examined across four time points (T1, M_age = 15.44, grades 7–12; T2, M_age = 16.93; T3, M_age = 21.84; and T4, M_age = 28.33) from 1992 to 2005 using RI‐CLPMs for overall disordered eating and specific forms for disordered eating (dieting and bulimic symptoms). Multigroup structural equation models were used to assess gender differences.Results: For male participants, high levels of loneliness at T1 predicted more overall disordered eating and more dieting at T2. Meanwhile, among female participants, disordered eating and bulimic symptoms at T2 predicted more loneliness at T3, whereas loneliness at T3 predicted more disordered eating and bulimic symptoms at T4, and vice versa.Discussion: The findings suggest a pattern of bidirectional associations between loneliness and disordered eating that varies by time points, gender, and type of eating problem. Preventive interventions and treatment should consider social factors involved in the onset and maintenance of eating problems in male adolescents and young adult women.Public Significance: This study contributes to the existing knowledge by examining for the first time the dynamic nature of the association between loneliness and disordered eating while accounting for all time‐invariant confounding. Our findings reveal a pattern of bidirectional associations between loneliness and disordered eating that appears to vary by developmental period, gender, and type of eating problem. Our findings suggest that social factors have to be taken into account when designing prevention strategies aimed at disordered eating.