Negative Results in Narcissism Interventions: A Meta-Analytic Review of Replication Failures in Social Media De-Addiction Programs

This study aim to assess the true effect size of digital addiction cessation intervention in reducing narcissism as well as sources of heterogeneity and publication bias in this field. Meta-analysis of all relevant research that followed the PRISMA 2020 guideline was performed on 25 studies (including 18 published studies and 7 unpublished studies) involving a total sample size of 1872 people. Search for literature was systematically done in six databases, which are PsycINFO, PubMed, Scopus, Web of Science, ProQuest, and Google Scholar, in 2025. The pooled effect size using random-effects model obtained a small and non-significant standardized mean difference (Hedges’ g) (g=0.18, 95% CI [-0.04, 0.40], p=0.112). There was high heterogeneity among studies (I² = 72%). Trim and fill analysis was done to correct publication bias, revealing six missing studies and adjusting the effect size to g = 0.04 (95% confidence interval [-0.21, 0.29]). Results from subgroup analysis showed that research employing a waitlist control group had a higher effect size (g = 0.41) compared to research employing an active control group (g = 0.02). Meta-regression further verified a significant negative correlation between sample size and effect size (p = 0.018). A significant difference was noted in the effect size between the published literature (g = 0.31) and grey literature (g = 0.01). These results indicate that the effectiveness of the interventions for digital addiction on narcissism is highly subject to publication bias and an inflated methodology, and the actual effect size is near zero.