Ghostwriting in biomedicine: a review of the published literature

Abstract

Introduction: The systematic review of biomedical ghostwriting has proven challenging due to problems in consistency and in study design. Moreover, authorship guidelines established by the International Committee of Medical Journal Editors (ICMJE) may have inadvertently created opportunities to potentiate ghostwriting. Given continued interest in ghostwriting by the International Society of Medical Publication Professionals (ISMPP) and other organizations, we undertook an analysis of ghostwriting in the biomedical literature.Methods: We searched PubMed (search terms: ghost writ*, ghostwrit*, ghost writer, ghostwriter, ghostwriting and ghost writing). Results, including abstracts, were reviewed for relevance (relationship to ghostwriting in biomedical journals) to aid in removal of inapplicable work and duplicate publications. After review, we consolidated expert opinions for publication professionals.Results: Overlap was poor across search terms; of 181 unique papers identified, most (112/181) were opinion pieces. An increasing number of papers are using the term “ghostwriting” to describe genetics as well as diverse phenomena of misattributed authorship, including “ghost authorship”. Eight primary studies and 1 systematic review of ghostwriting incidence were identified, reporting prevalence ranging from <1% to 91%, in varied settings using differing methods and definitions of ghostwriting. Suggestions for avoiding ghostwriting include early consensus building and better definitions of authorship among manuscript teams.Discussion: The prevalence and definition of ghostwriting remain unclear. Increased transparency and auditable authorship practices that align with specific guidelines may aid in the avoidance of ghostwriting. In addition, MeSH or clearer indexing terms may be helpful to separate usages of ghostwriting in scientific settings (e.g. genetic research) versus biomedical publishing.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.