This study presents a large-scale scientometric analysis of 35,514 retracted publications indexed in Scopus between 2001 and 2024, focusing on temporal trends, citation dynamics and global disparities in retraction patterns. The primary goal was to investigate how retractions evolve over time and how retracted articles continue to influence academic discourse through ‘lost citations’ – citations to works no longer considered valid. Using generalised additive models, we identified two significant surges in retraction volume, in 2010–2011 and 2020–2022, and revealed that these were largely driven by bulk retractions in conference proceedings. Citation analysis showed that while early retracted papers were often highly cited, most retracted documents in the last decade have negligible scholarly impact. However, severe inequality persists: a small subset of retracted papers accumulates the majority of citations. At the author level, we found that citation influence is only weakly correlated with the number of retracted articles, indicating reputational distortions. Country-level trends revealed that China leads retraction counts, followed by India and the United States, while publisher-level data showed the dominance of conference-oriented outlets such as the IEEE Computer Society. Our findings raise important concerns about the persistence of invalidated knowledge in citation metrics and the structural vulnerabilities of certain publishing formats. This study calls for enhanced retraction labelling, improved transparency in databases and retraction-aware bibliometric evaluations to preserve academic integrity.
Lendvai et al. (Sat,) studied this question.