---

---

---
title: "Gøtzsche's 1987 BMJ report is likely the first documented demonstration of citation bias in healthcare, an independent decades-old clinical-epidemiology lineage (Ravnskov 1992, Duyx et al. 2017) naming the same legitimation-selection mechanism Greenberg later found"
type: claim
status: seedling
source_url: "https://www.jameslindlibrary.org/articles/citation-bias-questionable-research-practice-or-scientific-misconduct/"
source_title: "Citation bias: questionable research practice or scientific misconduct?"
source_author: "Peter C. Gøtzsche"
source_date: "2021"
source_venue: "JLL Bulletin: Commentaries on the history of treatment evaluation, The James Lind Library (republished Journal of the Royal Society of Medicine 2022;115:31-35)"
source_tier: 2
source_quote: "The reference bias shown in this study seems to be real. Such a finding has important implications, since there is no reason to believe that rheumatologists are more biased than others in selecting references. A reader tracing the literature on any new drug using the reference lists given in the articles might risk obtaining a biased sample. Reference bias has another serious implication: it may render the conclusion of the individual article less reliable."
source_sha: "1f4f39b13cce6aca33a18613a43ecd8b4dd3fa31eb1120277e890158f43520a2"
source_quote_2: "Articles with a positive conclusion that supported the investigators' conclusion were cited 2.7 (2.0 to 3.7) times more often than others."
source_url_2: "https://www.jameslindlibrary.org/articles/citation-bias-questionable-research-practice-or-scientific-misconduct/"
source_sha_2: "1f4f39b13cce6aca33a18613a43ecd8b4dd3fa31eb1120277e890158f43520a2"
source_tier_2: 2
note: "source_quote is Gøtzsche's own first-person account of his own 1987 BMJ finding (Tier 1 for that claim, as the originating author describing his own work in his own venue, though recorded here via his 2021 retrospective rather than a direct 1987 primary read). source_quote_2 is Gøtzsche's own secondary report, as a named expert in his own venue, of Duyx et al. 2017's pooled figure -- Tier 2 for that number, one layer removed from the Duyx primary, which 403'd on both extract_pdf and archive_page via cris.maastrichtuniversity.nl and sciencedirect.com (known-blocked per sources.md's publisher-platform entries). Both quotes passed quote_check against the archived JLL page (sha256 above)."
provenance: "Promotion from 10-inbox/raw/2026-11-03-does-citation-studies-or-sociology-of-science-literature.md, 2026-11-03 (headless)"
origin: batch
derived_from: ["10-inbox/raw/2026-11-03-does-citation-studies-or-sociology-of-science-literature.md"]
date_created: 2026-11-03
writer_model: claude-sonnet-5
tags: [citation-bias, reference-bias, evidence-based-medicine, bibliometrics, legitimation, gotzsche, ravnskov, duyx]
seek_code_commit: d40ee39
quote_sweep: "PASS 2026-11-04 — source_quote verbatim (normalized) in the capture-time archive (sha256 1f4f39b13cce…), same-night model-free sweep"
audit_status: "2026-11-04 scheduled cross-model audit (claude-opus-5 auditing claude-sonnet-5): source re-fetched live and all pointers resolve. source_quote and source_quote_2 both confirmed verbatim against the live JLL page; Ravnskov 1992 (BMJ 305:15-9) and Gøtzsche 1987 (Br Med J 295:654-6) confirmed against the article's own reference list, as is Duyx et al. 2017 (J Clin Epidemiol 88:92-101) and its 52-study scope. CORRECTED, two items: (1) the commentary had attributed Duyx et al.'s omission of Gøtzsche 1987 to a search-term mismatch (\"reference bias\" vs. \"citation bias\"); the source states the opposite — one of their own search terms was \"referenc* bias*\", that term retrieves the 1987 paper from PubMed, and three of their 36 included biomedical studies had themselves quoted it. Old wording preserved in this audit's log entry. (2) the body's quoted P value read \"(P<0.01)\" where the source prints \"(P < 0.01; sign test)\", and lowercased the sentence-initial \"Ten\" inside the quotation marks; both corrected to print. No change to the note's central claim."

# Gøtzsche's 1987 BMJ report is likely the first documented demonstration of citation bias in healthcare, an independent decades-old clinical-epidemiology lineage naming the same legitimation-selection mechanism Greenberg later found

Peter Gøtzsche's 1987 *BMJ* report — by his own later account "what seems likely to be the first demonstration of citation bias in a systematic review of trials in healthcare" — found that among reports citing earlier trials comparing the same drugs, "[t]en articles had a neutral selection of references, 22 a negative selection, and 44 a positive selection (P < 0.01; sign test)," a pattern he found was not explained by the cited articles' methodological quality or by which journals were most cited. His own conclusion: "The reference bias shown in this study seems to be real... A reader tracing the literature on any new drug using the reference lists given in the articles might risk obtaining a biased sample. Reference bias has another serious implication: it may render the conclusion of the individual article less reliable." Uffe Ravnskov's 1992 analysis of cholesterol-lowering trials independently found supportive trials "cited almost six times more often" than unsupportive ones, with unsupportive trials "not cited at all after 1970." A 2017 systematic review and meta-analysis of 52 studies across disciplines (Duyx, Urlings, Swaen, Bouter & Zeegers, *Journal of Clinical Epidemiology* 88:92-101), pooling the evidence, found: "Articles with a positive conclusion that supported the investigators' conclusion were cited 2.7 (2.0 to 3.7) times more often than others," and — distinguishing this from popularity-driven selection — "Research quality was not related to the number of citations, whereas journal impact factor was."

This is the clinical-epidemiology field's own name — "citation bias" or "reference bias" — for exactly the legitimation-selection mechanism the vault's [[moc-legitimation-not-origination|Song Jian/Kwakernaak reading]] documents: citation selected or withheld by fit to a predetermined conclusion, independent of the cited work's actual methodological merit. It is a different scholarly lineage from [[claim-greenberg-2009-citation-bias-names-selective-ignoring-of-contrary-data|Greenberg's 2009 citation-network finding]] (clinical-trial citation-network epidemiology vs. a single disease-belief citation-network study) and from the bibliometric [[entity-matthew-effect|Matthew-effect]] lineage ([[claim-merton-1968-coined-the-matthew-effect-in-science]]) — three separate research traditions converging on variants of the same mechanism, two of them under the identical label "citation bias," without appearing to cite one another.

> [!note] Seek's commentary:
> Gøtzsche's own essay is itself a small lesson in the mechanism it describes: Duyx et al.'s 2017 review claimed to have done "the first systematic review of citation bias" and to have taken account of "all available evidence," while missing his own 1987 paper — the foundational paper in the field it was reviewing. The omission is *not* explained by a search-term mismatch. One of Duyx et al.'s own search terms was "referenc\* bias\*," and Gøtzsche reports that searching PubMed with that term retrieved both his 1987 paper and a later systematic review from his group (Schmidt and Gøtzsche 2005), "even though both papers had 'reference bias' in their titles" — and that three of the 36 biomedical studies they *did* include "had quoted my 1987 report." A literature about selective citation turns out to have selectively under-cited its own founder, by a route its own stated method should have closed.