Zombie trials are often monocentric and cluster within fabricated evidence “factories”: a cohort study of 236 retracted randomized controlled trials with confirmed data fabrication

Le résumé fourni par la source

Abstract Background Randomized controlled trials (RCTs) that rely on fabricated or fictitious data, commonly known as “zombie” trials, threaten evidence-based medicine by contaminating systematic reviews and clinical guidelines. We describe a cohort of zombie trials with confirmed data fabrication. Methods The cohort was derived from the Retraction Watch database by updating and extending the dataset used in the VITALITY study to include all retracted zombie trials identified from database inception through December 11, 2025. To ensure that included RCTs were identified as zombie trials rather than retracted for other reasons, retraction notices were text-mined for terms indicative of data fabrication. Trial- and author-level characteristics were extracted using a combination of automated and manual methods by one reviewer. We applied descriptive statistics, co-authorship network analysis and cascade effect mapping. Results Among 236 retracted RCTs published between 1983 and 2024 and identified as zombie trials with confirmed data fabrication, 197 (83.5%) were associated with authors who had a history of repeated misconduct. These trials involved 450 authors overall, including only 66 unique first authors, with one author alone accounting for 104 trials (43.8% of the cohort). Authors associated with multiple zombie trials experienced substantially longer delays to retraction, with a median of 13.6 years (IQR 8.0–15.3), compared with 2.75 years for authors linked to a single zombie trial. Most retracted zombie RCTs were monocentric (91.9%), involved a median of 3.5 authors per trial, and predominantly evaluated pharmacological interventions. They were largely concentrated in anesthesiology (58.1%) and in Japan (56.4%). Adjusted for the volume of trials produced by country, Japan, Tunisia, and Egypt had the highest rates. Co-authorship networks formed fragmented, largely disconnected clusters, suggesting localized “factories” of fabricated evidence. Retraction cascades were common, with the identification of a single fraudulent trial often leading to chains of retractions that exposed misconduct spanning decades. Conclusion Confirmed zombie trials were usually generated from single centers that published many of such trials.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.