Multiple authorship : growth or infl ationary bubble ?

OBJECTIVE: To analyze the increase in number of authors per article in Brazilian scientifi c journals on public health. METHODS: Articles published between 1999 and 2010 in six journals on public health and one medical journal (for comparison) from SciELO with Qualis (Capes) classifi cation equal or superior to B-1, were searched on the LILACS database. The evolution of the median number of authors/article and the proportion of articles with more than four authors were evaluated. The association between the the triennium of publication and the presence of four or more authors per paper was estimated through the Mantel-Haenzel odds ratio, adjusted for the type of journal. RESULTS: An increase of the median number of authors and the proportion of articles with more than four authors was observed in all journals, especially in the last triennium. The odds ratio for articles with four or more authors, adjusted for the type of journal, were: second triennium 1.3 (95%CI 1.1;1.4); third triennium 1.5 (95%CI 1.3;1.8), fourth triennium 2.39 (95%CI 2.1;2.8). CONCLUSIONS: Scientifi c journals on public health have shown an increase in the number of authors per article over the years, regardless of editorial orientation. DESCRIPTORS: Authorship. Journal Article. Public Health. Authorship and Co-Authorship in Scientifi c Publications. Scientifi c and Technical Publications. Scientifi c Publication Ethics.


INTRODUCTION
The number of authors per article has increased in scientifi c journals, a peculiar worldwide phenomenon in recent decades.Studies using different techniques, involving diverse journal groups and varied reference periods have unequivocally verifi ed this fact. 1,8,15,18,22 ne of these studies 24 quantifi ed the authorship of millions of articles in fi ve decades and showed that such phenomenon occurs in all areas of knowledge, including the social sciences, albeit with less intensity.
Most of these studies merely identify and describe the phenomenon, but there are those that seek to understand its determinants. 2,5,12,17,23Analysis performed on a sample of authors of 896 articles published in leading medical journals (Annals of Internal Medicine, JAMA, Lancet, Nature Medicine, New England Journal of Medicine and PLoS Medicine) showed a reasonable proportion (17.6%) of cases in which individuals who were included as authors had not contributed suffi ciently to merit this designation, characterizing the so-called "honorary authorship".Considering only research articles, this proportion reached 25%.
One of the mentioned studies 5 showed that the main contribution to the growing number of authors per article in the British Medical Journal (BMJ), over 20 years, was the relative increase in the number of senior authors.This fi nding becomes worrying in the view of the "white bull effect" (a reference to the myth of seduction of Europa by Zeus, disguised as a white bull), described by Kwok 14 as the self-imposition of a given individual as "honorary author" to a researcher in a weaker position.For instance, the negotiation of senior researchers with students under their guidance.How much of the senior authors' contribution in the study on the BMJ articles would be attributable to the "white bull" is something that must be questioned.A frequent explanation for the increase in the average number of authors is that the greater complexity of studies would demand more and more cooperative work. 1,7,8,17While this is possible and even probable, Papatheodorou et al 17 disagree that this fact explains all the situations in which such an increase is observed.Their research showed an increase in the number of authors over the years, in both randomized and non-randomized studies, with the effect of year of publication remaining signifi cant after adjustment for other factors, such as the topic of study, multinational study and population size.
One reason for the growing number of authors would be the answer to the pressures on researchers to increasingly publish articles (the so-called "publish or perish").However, in some situations a large number of authors is justifi ed, including multicenter and/or multidisciplinary studies, and research involving complex and laborious designs.Therefore, the analyses presented are justifi ed to protect and adequately characterize these situations.Such considerations led to the question: Would this growth also be observed in Brazilian journals on public health?Hence, the objective of this study is to analyze the increasing number of authors per article in Brazilian scientifi c journals on public health.

METHODS
A study was carried out in 2010 with Brazilian journals on public health included in the ScieLO collection, with classifi cation Qualis/CAPES equal or superior than B1: Revista de Saúde Pública (RSP), Cadernos de Saúde Pública (CSP), Ciência & Saúde Coletiva (C&SC), Revista Brasileira de Epidemiologia (RBE), Physis e Interface.The Brazilian Journal of Medical and Biological Research (BJMBR), also holding a Qualis B1 classifi cation in public health a (2010) and found in SciELO, was included for comparison.For each journal, all references in the LILACS database (1999 to 2010) were selected.The initial year of the period was chosen for the journals that were already being published for at least one year, and 2010 represented the last year with all issues published and present in the database at the time of the review.
a For purposes of evaluation of postgraduate programs, the Brazilian Ministry of Education created a ranking of journals based on different criteria and specifi c to each area of knowledge.Further information at http://www.capes.gov.br/avaliacao/qualis The references were exported in RIS format and the number of authors per paper was calculated by an application developed for this purpose.Analyses were performed according to the four triennia of the studied period to ensure a greater stability in the estimates.The median number of authors per article (interquartile range) and the proportion of the number of articles with four or more authors were calculated for each triennium.The association between the triennium of publication and the presence of four or more authors per paper was estimated through the Mantel-Haenzel odds ratio, adjusted for the type of journal.The journals were grouped into four categories 3 : 1) clinical journal (BJMBR); 2) journals with a higher proportion of epidemiological articles (RBE, CSC, RSP); 3) journal without a clear predominance of a specifi c study fi eld (C&SC); 4) journals with a higher proportion of articles on the humanities (Physis and Interface).The analyses were performed with the Stata program (version 9.0).

RESULTS
A higher median of authors for the BJMBR, followed by group 2 journals (RBE, CSC, RSP), was observed in all periods (Table 1).Group 4 (Physis and Interface) and group 3 (C&SC) journals had the lowest medians.An increase in the median number of authors in the last triennium was observed for all journals, when compared to the fi rst triennium.The proportion of articles with four or more authors had a roughly similar distribution to the median of authors related to journal category, although C&SC has shown higher proportions than the group 4 journals for this indicator (Figure 1).The growth of this proportion was observed for all journals analyzed.Taking the fi rst triennium as a reference, the odds ratios are presented for articles with four or more authors, adjusted according to journal group (Table 2).This analysis also reveals growth.The chance of having four or more authors was 2.39 higher for articles published in the last three years compared to those published in the fi rst three years (Table 2).

DISCUSSION
An increase in the number of authors per paper was observed, particularly in the last triennium, based on two indicators: median number of articles and proportion of articles with four or more authors.
As this phenomenon is confi rmed in the analyzed journals, a question remains: Is it the result of an increased cooperation between Brazilian authors or "honorary authorship"?
19,21 An editorial published in Nature 9 reports the diffi culty of fi nding mechanisms to control the situation.Another 10 connects this problem with the use of quantitative indicators of scientifi c production, as the productivity incentive would lead to the proliferation of authors as a way to cheat the system, at least partially.
A particularly intense exchange began after a joint editorial by the then editors of Lancet and BMJ 11 (Richard Horton and Richard Smith, respectively), who analyzed their concerns regarding this problem, and announced a seminar to discuss proposals to address the situation.An editorial in the BMJ 19 points to a radical strategy, based on the discussions from the seminar previously mentioned: articles would have contributors instead of authors, who would be identifi ed according to their participation, just as in the credits of a movie.To ensure ethical responsibility on the printed content, the articles would have a "guarantor".These ideas prompted the then editor of the American Journal of Public Health (AJPH), Mervyn Susser, to publish an editorial 21 that endorsed such proposal, asking the readers for their opinions.Responses were published in the May 1998 issue, occupying practically the entire letters section of the AJPH.The positions were varied, and no consensus was reached.
Publications 4,6,7,16,20 make revisions of proposed solutions, trying to offer control models to limit the possibility of undue authorship.Such solutions are divided into two major groups, both based on exhaustive lists of possible actions (contributions) in the making of an article.The fi rst treats the list as a checklist, requiring minimum number of contributions (usually three) for the authorship to be considered.The other works with complex score systems assigned to each type of contribution, requiring a minimum total value, which varies according to the scheme adopted.One of the consulted papers 13 sought to evaluate the operation of such schemes.The authors of 181 articles published in the Croatian Medical Journal, from January to July 2005, were consulted, by randomly using instruments based on the checklist or scoring system.It was concluded that the latter was more sensitive for determining authorship.
These proposals are attempts to implement the minimum requirements for authorship by the International Committee of Medical Journal Editors (ICMJE), b which states that "authorship credit should be based on 1) substantial contributions towards the conception and design, data acquisition, or analysis and interpretation of data; 2) article writing or critical revision with important intellectual content; and 3) fi nal approval of the version to be published.Authors should meet conditions 1, 2 and 3." This guidance resulted from the discovery of a fraud case (known as "Darsee affair") in the 1980s, which exposed several cases of "honorary authorship". 19e review article analyzed 16 points out to usual and mistaken roles for attributing authorship, such as being administratively in charge of a research group or department, and raising funds for a project without being involved with it otherwise.In addition, there are other roles that could be acknowledged, but no authorship would be given to those revising or editing a manuscript, performing manual data collection (exceptional circumstances could change this), cleaning data, as well as providing resources (e.g.reagents or basic processes involved in the research that have not been specifi cally developed for it), basic maintenance and management of equipment/ instruments (equipment/tools developed specifi cally for the considered project could, nonetheless, qualify for authorship).The studied journals adopt the ICMJE criteria with small systematic variations, demanding the authors a statement of authorship responsibility, without presenting any kind of checklist or score.The CNPq itself recently reported problems in the dissemination of scientifi c research under its funding.
Considering that overcoming such problems would require the formulation of specific internal rules, nonexistent at that time, a committee responsible for their elaboration was created.Their publication on the organization's website followed.c Several rules relate to the question of authorship (referring to the ICMJE criteria), indicating that, at least, there is some concern over this issue in our circles.
We emphasize that there are multiple authors per article, and even though the average number of authors is growing, it does not translates as an irregularity.defi ning who can appear as an author in a given publication, and this is not adequately discussed by Brazilian authors and editors.
It is not possible to determine, due to the limitations of this study, whether or not the honorary authorship was occurring in the publications analyzed.The increasing number of authors per article, regardless of the journal's editorial orientation, indicates the need to further explore this issue through more extensive studies that include other variables, allow to qualify more precisely the type of study that originated the article and also assess authorship qualitatively.
Particularly in the absence of effective controls, the attribution of "honorary authorship" may be a considerable temptation in an environment which stimulates the production of increasing numbers of articles.It is a task for editors, authors and readers to ensure compliance with the ethical principles governing authorship, thus avoiding the situation when the basic currency of academic credibility will suffer from infl ationary devaluation, as expressed by Papatheodorou et al. 17 These authors made a clear allusion to what repeatedly occurred in recent decades with certain products under accelerated growth: the collapse of their markets, showing that their apparent value was unreal, an infl ationary bubble.
Figure.Proportion of articles with four or more authors according to the analyzed journals and triennia of publication, 1999 to 2010.

Table 1 .
Distribution of the number of authors published according to triennium, 1999 to 2010

Table 2 .
Association between triennium of publication and occurrence of four or more authors, 1999 to 2010.