Professional Documents
Culture Documents
CONTRIBUTORSHIP
W
HAT HAPPENS WHEN SCI- Methods Purposive sampling of 10 research articles published in The Lancet; quali-
entists disagree? Most tative analysis of answers to 6 questions about the meaning of the study put to con-
times, readers of re- tributors who were listed on the byline of these articles. Fifty-four contributors listed
search papers never on the bylines of the 10 articles were evaluated, and answers to questions were com-
know. However, in 1995, a dispute pared between contributors within research groups and against the published re-
among the writing committee of the search report.
Italian Multicentre Acute Stroke Trial Results A total of 36 (67%) of 54 contributors replied to this survey. Important weak-
spilled out onto the pages of The Lan- nesses were often admitted on direct questioning but were not included in the published
cet.1,2 During peer review, it became article. Contributors frequently disagreed about the importance of their findings, impli-
clear that 2 committee members inter- cations, and directions for future research. I could find no effort to study systematically
preted the results of the trial very dif- past evidence relating to the investigators’ own findings in either survey responses or the
published article. Overall, the diversity of contributor opinion was commonly excluded
ferently from their colleagues. They
from the published report. I found that discussion sections were haphazardly organized
had, for the good of the collaboration, and did not deal systematically with important questions about the study.
self-censored their own views. How-
Conclusions A research paper rarely represents the opinions of those scientists whose
ever, this fragile truce broke down once
work it reports. The findings described herein reveal evidence of (self-)censored criti-
an editor asked for signatures confirm- cism, obscured meanings, confused assessment of implications, and failures to indi-
ing each contributor’s assent for the pa- cate directions for future research. There is now empirical support for the introduction
per to be published. Tognoni and Ron- of structured discussion sections in research papers. Editors might also explore ways
caglioni2 described their disagreement to recover the plurality of contributors’ opinions.
as “unfortunate.” JAMA. 2002;287:2775-2778 www.jama.com
Such disagreements have come to
light before. For example, divided in- ence that is most relevant to publica- contact contributors on the article’s by-
terpretations about a polio outbreak in tion because its disclosure can iden- line and to explain the background and
Israel led to separate commentaries in tify who is accountable for what part nature of the study. Once permission
The Lancet. 3 In an even more pro- of the research and allows the reader had been granted, I wrote to all con-
tracted dispute, a competing manu- to assign credit fairly.” tributors and asked 6 questions about
script based on the same study was However, important as contributor- their work (BOX 1). Contributors were
eventually published 4 years after the ship is, this mechanism of disclosure written to twice after that and tele-
original article appeared.4 Harmony does not take account of the ideas or phoned once to obtain replies.
among authors cannot be relied on. interpretations contributed to the re- Once available replies had been col-
A crisis over definitions of author- search being reported. I wanted to know lected, individual answers to ques-
ship during recent years has led sev- whether the views expressed in a re- tions were compared with one an-
eral medical journals to discard rigid search paper are accurate representa- other among contributors for each
rules for who can or who cannot be an tions of contributors’ opinions. research paper. These answers were also
author. Instead, the idea of the con- compared with the contents of the pub-
tributor has emerged.5 In place of the METHODS lished article. Finally, contributor sec-
assumption that scientists named on the I selected 10 articles published in The tions were examined to discover if there
byline of an article are true authors, Lancet during 2000 (TABLE). This study were any identifiable connections
journals such as The Lancet, BMJ, and had a qualitative design: articles were between stated contributions and the
Annals of Internal Medicine now re- selected purposively with varying num- answers to these 6 questions.
quire contributors to state explicitly bers of contributors, across a range of
Author Affiliation: Dr Horton is Editor of The Lan-
what part they played in the research subject areas, and including a spec- cet, London, England.
being reported. This conceptual shift trum of research methods. I wrote to Corresponding Author and Reprints: Richard Hor-
ton, FRCP, The Lancet, 32 Jamestown Rd, London
was best summarized by Rennie et al5: the corresponding author of each re- NW1 78Y, England (e-mail: richard.horton@lancet
“Contribution is the activity of sci- search article to secure permission to .com).
©2002 American Medical Association. All rights reserved. (Reprinted) JAMA, June 5, 2002—Vol 287, No. 21 2775
portant implication” of the trial was that Confusion was also common when sociations could be made, mostly be-
the results “would help remove the nega- implications of new research were being cause contributor statements lacked
tive social connotations associated with considered. For example, in the folate sufficient descriptive detail.
this disorder.” Nowhere was this impli- and vitamin B6 randomized trial, one con- In reviewing these 10 published ar-
cation mentioned in the article pub- tributor concluded that the results “pro- ticles, the most frustrating aspect of
lished in The Lancet. vide some justification” for treating pa- comparing survey responses with pub-
Finally, in considering lines for tients at high risk of atherothrombotic lished reports was the chaotic nature of
future research, several possibilities disease with folic acid. Another contribu- discussion sections. There was no clear
were identified: the physiology of tor simply considered that this “first or consistent approach by contribu-
bulimia (4 contributors), comparisons (little) piece of evidence . . . should be tors to the discussion of their results.
of ondansetron with other treatments seen as an encouragement for other tri- Limitations were frequently omitted,
(3), the inclusion of patients with less als” only. In a systematic review of stress clinical interpretations were often
severe conditions in subsequent trials hyperglycemia and risk of death after mixed with mechanistic reflections, and
(2), and a longer study duration (2). myocardial infarction, one contributor repetition of key results was common.
None of these ideas was discussed in drew a diagnostic conclusion: “a simple,
the published article. early available, and cheap plasma glu- COMMENT
Many of these omissions and pat- cose identifies patients at a high risk for The results of this qualitative study
terns of reporting were found in the in hospital complications and death.” An- show that a research paper rarely rep-
other articles studied (data not shown). other drew a treatment lesson: “clini- resents the full range of opinions of
However, there were exceptions. For ex- cians should recognise hyperglycaemia those scientists whose work it claims
ample, in a randomized trial of folic acid as an important prognostic marker and to report. I have found evidence of cen-
plus vitamin B6 to lower plasma homo- take an aggressive therapeutic ap- sored criticism; obscured views about
cysteine concentrations and perhaps to proach for patients who have elevated the meaning of research findings; in-
ameliorate atherosclerosis, the weak- blood glucose readings at the time of complete, confused, and sometimes bi-
nesses cited in the survey responses [myocardial infarction].” The article it- ased assessment of the implications of
(small size, use of surrogate measures, self does not explore this range of opin- a study; and frequent failure to indi-
and short study duration) were all dis- ion and takes a more conservative line, cate directions for future research. Some
cussed in the published article. The strik- emphasizing glucose as a risk factor only. papers have more complete evalua-
ing fact, therefore, was the inconsis- Contributors differed in their views tions of findings than others. What was
tency across this sample of articles. For about future research. In response to a striking was the inconsistency in pub-
instance, in a study of how El Niño af- direct question, contributors to a pa- lished evaluations, especially regard-
fects diarrheal diseases in Peruvian chil- per on risk factors for suicide sug- ing weaknesses. The strengths of this
dren, several contributors pointed out gested looking at sex, family history of work are its qualitative design, which
that only one El Niño event had been illness, age, socioeconomic back- produced a rich data set, and a purpo-
studied. This weakness was not dis- ground, psychiatric diagnosis, and life sive sampling technique that con-
cussed in the article. Similarly, al- events. Yet none of these ideas was dis- firmed the findings across a range of
though wide confidence intervals were cussed in the published report. In a subject areas and study designs.
cited as a weakness in a study of cancer study of Helicobacter pylori transmis- This work also had several limita-
in individuals with Down syndrome, this sion among siblings, readers were given tions. First, these data are preliminary.
weakness was not highlighted in the ar- no direction about where future re- The sample of articles was small and
ticle published in The Lancet. search might be directed. However, in came from one journal only. The risk of
The question that yielded the most their survey responses, the contribu- bias is substantial. Second, since I sur-
uniformly disappointing response con- tors suggested work on the natural his- veyed contributors after publication of
cerned interpretation in the context of tory of the infection, factors (espe- their studies, I could not rule out the pos-
the totality of available evidence. In nei- cially those in the family) that influence sibility that contributors discussed their
ther the survey responses nor the pub- the dynamics of infection, and a focus responses with one another before re-
lished articles were any efforts made to on early childhood years. plying. Third, variance of opinion was
describe systematically evidence that re- The Lancet introduced contributors’ determined by one person (R.H.) using
lated to the investigators’ own find- descriptions of the parts each person a nonvalidated survey instrument. Mul-
ings. Anecdotal reporting of other work played in the research being reported tiple independent assessments, perhaps
was the norm in both settings. The con- in 1997. These descriptions are writ- adopting a quantitative scale, could im-
sistent failure by scientists to provide ten by the authors themselves. In this prove the validity of these findings.
a more rigorous overview of past evi- study, I relied on these self-reports to A scientific research paper is an ex-
dence when considering their own find- find links between stated contribu- ercise in rhetoric7; that is, the paper is
ings has been pointed out before.6 tions and survey responses. No such as- designed to persuade or at least con-
©2002 American Medical Association. All rights reserved. (Reprinted) JAMA, June 5, 2002—Vol 287, No. 21 2777
2778 JAMA, June 5, 2002—Vol 287, No. 21 (Reprinted) ©2002 American Medical Association. All rights reserved.