Article Text

Download PDFPDF

The Meaning of Author Order in Medical Research
  1. Mark Otto Baerlocher,
  2. Marshall Newton,
  3. Tina Gautam,
  4. George Tomlinson,
  5. Allan S. Detsky
  1. From the University of Toronto Radiology Residency Training Program (M.O.B.); Departments of Health Policy Management and Evaluation and Medicine (M.N., T.G., A.S.D.) and Medicine and Public Health Services (G.T.), University of Toronto; Department of Medicine (M.N., T.G., A.S.D., G.T.), Mount Sinai Hospital; and Department of Medicine (M.N., T.G., A.S.D., G.T.), University Health Network, Toronto, ON.
  1. Address correspondence to: Dr. Allan S. Detsky, Mount Sinai Hospital, 427-600 University Avenue, Toronto, ON M5G 1X5; e-mail: allan.detsky{at}uhn.on.ca.

Abstract

Background Manuscript authorship and author placement have important implications for accountability and allocation of credit. The objective of this study was to assess the relationship between an author's place in the author list and the type of contribution reported by that author. This pattern was then used to develop a method by which author responsibility and accountability can be clarified.

Methods The published contributions of each author of original research articles with a minimum of four authors published in the Journal of the American MedicalAssociation, the British Medical Journal, TheLancet, and the Canadian Medical Association Journal in a 3-year period after author contribution forms were required were coded into 1 of eleven contribution categories. The contributions were grouped according to first, second, middle, and last author and compared by position.

Results For most categories of contribution, the levels of participation were highest for first authors, followed by last and then second authors. Middle authors had lower levels particularly in conception, drafts of the manuscript, supervision, and being a guarantor.

Conclusions Current patterns of author order and contribution suggest a consistent theme. Based on the results, a proposal is put forth by which author accountability is clarified. In this proposal, authors are classified as either “primary,” “contributing,” or “senior or supervisory,” each with specified contributions. More than one author may be classified into each author category.

Key words
  • authorship
  • biomedical research
  • publication

Statistics from Altmetric.com

Key words

Authorship and author order can have important implications.1-3Consider the recent extreme example of Dr. Eric Poehlman, who admitted in court documents to falsifying numerous federal grant applications and the data on which many of his manuscripts were based and is now facing up to 5 years in prison and $250,000 in fines.4The question of what consequences the coauthors should face arises. Furthermore, authorship and authorship order (eg, first vs second) have significant implications in academia and are used in determining academic achievement for the purposes of promotion, allocated research time, and funding.

According to the International Committee of Medical Journal Editors (ICMJE), the designation of “author” should be based only on “1) substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; 2) drafting the article or revising it critically for important intellectual content; and 3) final approval of the version to be published.”5All three conditions must be met. Furthermore, “acquisition of funding, the collection of data, or general supervision of the research group, by themselves, do not justify authorship.”5

Rennie and colleagues published a commentary in the Journal of the American Medical Association (JAMA) that noted the increasing decline of meaning and utility of the concept of “author” owing to increasing complexity in relationships between coauthors and more specialized contributions and evidenced by an increase in the number of coauthors per manuscript.6The authors concluded that a concept of “contributors” rather than coauthors had become necessary for readers to more accurately assess credit and assess accountability, both crucial components of medical literature. Furthermore, Rennie and colleagues argued for the necessity of a guarantor for the entirety of the manuscript and for the data on which the manuscript is based. The notion of public and published specific contributions per author had thus become center stage, whereas previously relative contribution seemed to have had more importance.

In an effort to ensure true authorship, TheLancet (Lancet) began to require signatory statements of manuscript and research contribution,7followed by the British Medical Journal (BMJ),8 Radiology,9and the American Journal of Public Health.10The initial experience was very positive.11A large number of medical journals now require authorship contribution forms prior to publication. Each author must declare and satisfy a minimum number of contributions and conditions to merit coauthorship.

This brings up the issue of author order. Perhaps it is possible to differentiate the “true” or “deserving” authors based on author order. With the exception of a minority of journals that list authors alphabetically, most list authors according to author-rated level of contribution. It is our impression that in most medical cultures, the first author contributed more than the second author, and so forth, whereas the author in the last position tends to be the “senior” author or “principal investigator” and therefore made a substantial contribution. However, the decision about author order may actually have much to do with author seniority rather than contribution.

The objective of this study was to assess the relationship between an author's place in the author list and the type of contribution reported by that author and to develop a proposal to modify the practice of author order. We included articles from JAMA, BMJ, the Canadian Medical Association Journal (CMAJ), and Lancet, each of which has a slightly different method by which authorship contribution information is detailed.

Methods

We examined the published contributions of each of the listed coauthors of each original research article with a minimum of four authors; this required minimum meant that for each publication, an author could be classified as first, second, last, or middle author. The corresponding author was also noted, and this author's contributions were analyzed separately. We selected articles published in the first issue each month of JAMA, BMJ, CMAJ, or Lancet in a 3-year period after signed statements of contribution became mandatory. These journals were chosen because they are the four general medical journals with the highest journal impact factor rating that require explicit author contribution disclosure. Studies that included group authors and research letters were excluded. Using the published contribution byline, we recorded each author's contribution within the following categories: (1) conception and design of the study; (2) data management; (3) interpretation of data and statistical analyses; (4) data acquisition; (5) drafting of the manuscript; (6) manuscript revision; (7) obtaining funding; (8) administration, technical, or material support; (9) supervision; (10) principal investigator or guarantor; and (11) other (see the Appendixfor a list of our interpretations of the published contribution terms). Multiple contributions were allowed per author.

This method of classification of authorship roles was based partly on JAMA's authorship contribution form, as well discussion from the ICMJE on what constitutes authorship.5The categories were modified after pilot testing and then reviewed by an independent researcher with experience in a major medical journal office for further modification.

Contributions for any one article were coded by one of two authors (T.G., M.N.). To assess interobserver variability, 10% of all manuscripts were independently reviewed by both raters, and the kappa values were calculated for each contribution category. A third author (M.O.B.) re-reviewed 5% of all studies. In cases in which there was difficulty in assigning contributions or disagreement between abstractors, the senior author (A.S.D.) made the final decision.

To test whether each author position had a different likelihood of reporting a given contribution category, a bayesian random effects multiple logistic regression model was fitted to the data, where for each author, the outcome was whether she or he contributed to a given category. Odds ratios were estimated to compare contribution frequency between author positions.

Contribution information was used to determine the percentage of authors, for each author position and journal, who met both criteria one and two of the ICMJE guidelines for authorship: (1) substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data and (2) drafted the article or revised it critically for important intellectual content. The third criterion, final approval of the version to be published, could not be assessed as this information was commonly not published; it was therefore assumed that all authors gave their final approval for publication of the manuscript (as did Bates and colleagues in their analysis of author contributions12). Authors not fulfilling ICMJE criteria one and two were assumed to be “honorary,” that is, not meriting authorship status.

Results

A total of 109, 62, 106, and 94 articles were analyzed from JAMA (2001-2003), CMAJ (2001-2003), BMJ (1998-2000), and Lancet (1998-2000), respectively. Contributions for 2,586 authors were recorded (371 for each of first, second, and last authors and 1,473 for middle authors). The kappa values for agreement on studies that were read by two raters were high for all contribution categories, ranging from 0.92 to 1.00.

Figure 1 shows the percentage of authors in each position who participated in the various categories of contribution. Visual examination reveals several patterns. The first authors had the greatest participation in most categories and were generally responsible for planning the study, performing the study, including data acquisition and analysis, and drafting the manuscript. Authors in the last position were generally responsible for planning the study, critically revising the manuscript, and supervising. The second authors, and in particular the corresponding authors, played dominant roles in most aspects of the research, comparable to that of the first author (Figure 2).

Figure 1

Author contribution of original research manuscripts published in JAMA (2001-2003), Lancet (1998-2000), CMAJ (2001-2003), and BMJ (1998-2000).

Figure 2

Contributions of corresponding authors (all journals combined; n = 370).

Combining data from all journals, few authors specified a contribution to “data management” or “obtaining funding” or reported that they were the “principal investigator or guarantor” (see Figure 1).

Statistical analysis showed that some of the differences in the levels of contribution by author order shown in Figure 1 were statistically significant. For three of the categories (interpretation of data or statistical analyses, drafting the manuscript, and supervision), there is a high probability (> 95%) that all author positions have different contribution frequencies. Each of the other categories has at least one pair of author positions with similar levels of contribution frequency (ie, a 95% confidence interval for an odds ratio that crosses 1). This includes the comparisons of second with last author for conception and design all author positions for data management; first with second author and second with middle author for data acquisition; second with middle author and middle with last author for manuscript revision; first with last author for obtaining funding; all author positions for administrative, technical, or material support; and second with last author for principal investigator or guarantor.

Authors publishing in JAMA were most likely to fulfill the ICMJE criteria for authorship, whereas authors publishing in Lancet were least likely (Table 1). Middle authors were least likely to fulfill authorship criteria (see Table 1). Combining data from the four journals (JAMA, Lancet, CMAJ, and BMJ), 93.5% of corresponding authors (n = 370) fulfilled ICMJE criteria 1 and 2 for authorship.

Table 1

Percentage of Listed Authors Meriting Authorship (Fulfilling Criteria 1 and 2 of the International Committee of Medical Journal Editors' Guidelines for Authorship)*

Figure 2 shows the pattern of contribution of corresponding authors. Corresponding authors were most commonly in the first author position (76.8%, n = 284), followed by the last position (13.5%, n = 50), second position (7.6%, n = 28), and middle position (2.2%, n = 8). One article had two corresponding authors, whereas two articles had none.

Discussion

In this study, we were able to demonstrate two findings. First, author order correlates with a pattern of contribution indicating that the first, second, and last authors had the highest levels of participation in most areas. Second, according to data published by the authors themselves, a significant minority of authors do not merit the designation “authorship” according to the ICMJE criteria.

JAMA authors demonstrated the highest proportion of authors following the ICMJE guidelines. This is a result of the method of eliciting participation using a specific checklist. The JAMA checklist outlines how many categories have to be designated for someone to be considered an author, thereby ensuring higher compliance. This may or may not reflect a true higher rate of participation. This also represents a limitation of using the data collected to make interjournal comparisons as some journals had more extensive contribution bylines than others.

It was also clear that the majority of articles published did not indicate any author responsible for obtaining funding (see Figure 1). This would imply that the majority of published clinical research did not require any funding. This seemed odd to us. The corresponding author, who deals with the journal editors and readers, was most commonly the first or last author.

It was not too long ago that the idea of “contributorship” gained momentum, wherein all authors would descriptively detail the parts of the study for which they were responsible.6-8,13A great deal of hope was placed on this new proposal, and it was incorporated to varying degrees by many medical journals thereafter. It was clear, however, that its effect would have to be monitored and its utility demonstrated,2given that it created yet one more step in what was already often a laborious process of manuscript submission.

Based on our results, we would like to put forth a new proposal that may clarify the accountability of authors, as well as alleviate the issue of honorary authorship. We propose that journals eliminate the traditional method by which authors are listed, in which the first and the last positions are generally (but not always) viewed as the most significant, followed by the second position, third position, fourth position, and so forth. Instead, this method should be replaced with a system in which there are only three designations: primary author(s), senior or supervisory author(s), and contributing author(s). For each designation, more than one author should be allowed, if deemed appropriate by the study investigators.

Primary authors are those who (1) plan the study, (2) execute the study, (3) write the manuscript (ie, are primarily responsible for the drafting and redrafting of the manuscript), and (4) guarantee the accuracy of the data or manuscript. Senior or supervisory authors are those who (1) plan the study, (2) substantially revise the manuscript, (3) supervise, and (4) guarantee the accuracy of the data or manuscript. Finally, contributing authors are those who (1) follow the ICMJE authorship criteria for authorship but (2) do not fulfill the criteria for either the primary author or senior or supervisory author (Table 2). Authors should fulfill all criteria of their respective author category to receive that designation.

Table 2

Proposal for Authorship Appropriation and Categorization

Primary and senior or supervisory authors should be responsible for guaranteeing the accuracy of the data on which the manuscript is based; to the best of their knowledge ensure that there was no fraud or error; and be able to explain the results of the entire manuscript. Contributing authors should not be responsible for having the ability to explain the results of the entire manuscript or for guaranteeing the accuracy of all of the data. We would therefore still recommend the requirement of signed declarations of authorship for all authors and of guarantor only for the primary and senior or supervisory authors. By doing so, primary and senior authors would understand that they would all be held accountable for any error or fraud contained in the paper.

The definitions of these three authorship positions still should follow the ICMJE authorship guidelines. Manuscripts may list primary author(s) and contributing author(s), without any senior or supervisory author(s), if this was indicated.

There are several reasons why we believe that our proposed system is an improvement over the current method of listing authors. First, it would standardize the patterns of contribution that seem to be currently followed by authors as observed in this study and in that way would help new or inexperienced authors in providing guidance about what author order means. For example, it would standardize the position of the senior author, whereas now the senior author might sometimes be the second and at other times the last author. Second, it would allow a clearer gradation of authors' relative contribution. Third, by allowing more than one primary author and more than one senior or supervisory author, there is much greater flexibility in assigning relative rank. We are certain that there are many cases in which the decision about which single author to list in the first and last positions leads to strife and potentially arbitrary choices. Fourth, this system would help clarify situations in which there are only two authors as to whether the second author is more of a primary author or more of a senior author. Fifth, this system would clarify which authors are accountable for error and fraud and which are not. The current method of author listing is vague in this regard. Finally, this system will make it easier for those who evaluate relative authors' contributions during processes such as promotion without having to retrieve each paper to examine the author's listed contributions. Many investigators currently annotate their curriculum vitae outlining their role in the paper. A lack of standardization and unusual designations (eg, problems such as declaring yourself to be a senior author when you are in the middle of the pack) would be greatly aided by our proposed system.

There has been somewhat of an analogous reformation occurring in the world of research grant applications. For example, the National Institutes of Health recently sought “input and advice from the scientific community on various concepts associated with permitting more than one Principal Investigator… to be associated with an NIH funded grant, contract, or cooperative agreement” and is now allowing more than one principal investigator from some studies.14This was due to recognition that research has become increasingly complex, often involving collaboration within a multidisciplinary team of researchers.

Of course, if there are advantages to our proposed change, there must also be disadvantages. Some of these include deciding how to order the lists of first and senior authors if there is more than one, the potential to exacerbate the problem of honorary authors in the contributory category, and the potential to dilute the credit given to those who are now first and last authors. As well, there would need to be discussion regarding how journal indexers would handle the new requirements, which this proposal does not address.

Our proposed authorship system would be a great improvement over the current approach by clearly indicating the relative role and contribution of each listed author, without the ambiguity that exists with the current system of authorship. We recognize that this proposal would ask the scientific community to change a standard practice, similar to the introduction of the standardized abstract about 10 years ago, but believe that the benefits in terms of clarity and flexibility outweigh the inconvenience. Furthermore, the presented proposal is not intended to be definitive and may be refined by other researchers, journal editors, and authors. The proposal is intended to provoke and kick-start a discussion on the topic.

Appendix: Classification of Author Contribution Byline

Advised/consulted on study methods = technical support

All aspects of the study = everything (except other)

Analytic design = statistics and data analysis

Assessments (eg, angiography, polymerase chain reactions); assisted or helped? = acquisition or technical

Assisted in research = other (a vague example)

Clinical aspects of the study = data acquisition

Clinical input = technical support

Conduct of the study or research = other

Consulting = support

Coordinated data collection = data acquisition

Coordinated data analysis = analysis

Coordinated of the project = supervision

Coordinated a particular aspect of the study (ie, reading a test, radiograph) (unless it is the key part of the study, then it is data acquisition) = technical support

Data abstraction = data acquisition

Data coding = data management

Data extraction = acquisition

Designed a certain aspect of the study (survey, questionnaires) = support (unless it falls into a specific category)

Development of the database or application for data collection = data management

Directed the study = supervision and guarantor

Did the randomization = technical support

Epidemiologically related = technical support

Ethics and ethical work = technical

Expertise = support

Execution of the study or research = other

Formulated the research question = conception and design

Implemented the study or protocol = data acquisition

Initiated the study = design and concept

Key role in presenting the data = other

Managed data collection or data collection = data acquisition

Managed, organized, or prepared the data = data management

Monitored adverse events = technical

Organized the project = study design and supervisor

Overall or study management = supervision

Patient recruitment or accrual or follow‐up = data acquisition

Planning = study design

Principal analyst = data interpretation

Programmed the database = technical support

Protocol design, development, writing = design

Statistical expertise = statistical analyses

Supervised data analysis = data analysis

Supervised data collection = data acquisition

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
View Abstract

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.