Framework for assessing the risk to a field from fraudulent research

Key results

1. ShownNo individual author can reach most of the AD literature through direct citations. Most authors reach a tiny share, though the most-cited researchers reach a sizable minority of papers.

   Of 22,941 authors with multiple first- or last-author AD papers, the mean fraction of literature citing an author was 0.3%; under 1% of authors were cited by over 5% of 112,720 papers; the top 150 authors averaged 7.8% reach (92% of papers do not cite them); the single most-cited author reached around 21%. · The risk of citing individual authors; Discussion and Conclusions

2. ShownInfluence in the AD literature is highly concentrated: a small share of papers and authors receives most citations and produces most output, following approximate power-law distributions.

   10% of AD publications received 57% of AD citations; among 22,941 authors, about 10% contributed 51% of papers, and about 70% of 2,051,603 citations went to 10% of authors; papers per author follow an approximate power law. · Overview of the Alzheimer’s research enterprise; Risk analysis of individual papers; The risk of citing individual authors

3. ShownMost AD papers cite both a recent work and that work's antecedent (a citation triangle) rather than forming linear chains. The authors take this to mean indirect, second-order citations are unlikely to be the main route by which flawed findings spread.

   65% of papers cited the immediately preceding work plus at least one antecedent, vs 35% citing only the preceding work; triangles were significantly more common than in a randomized network (p < 2.2e-16, n = 2,051,603, Fisher's two-sided exact test). · Risk analysis of individual papers; Discussion and Conclusions; Figure legends

4. ShownA single questioned paper touches only a small share of the field. The 2006 study under misconduct investigation, though heavily cited, is cited by a minor fraction of the AD corpus, and direct links between papers are rare.

   The investigated paper's 2,300 citations equal about 2% of 112,720 AD papers; average shortest citation path between papers is 4.4; only 0.01% of pathway links (n = 2,051,603) are direct (shortest path=1). · Introduction; Risk analysis of individual papers; Discussion and Conclusions

5. ShownAn author's citation reach rises closely with their citation volume, so the integrity of the most highly cited researchers matters most for protecting the field.

   Fraction of AD literature citing an author vs number of citations: r = 0.84, p < 2.2e-16, Pearson's correlation test, n = 22,941 authors. · The risk of citing individual authors; Figure legends