The ADNI Publication Policy: Commensurate recognition of critical contributors who are not authors

Abstract

An efficient approach to certain types of biomedical research requires a scale that precludes involvement of all critical contributors in all aspects of experimental design, execution, and as well as writing of most, if not all, derived works. Guarantors of both the integrity of the data and of its subsequent analyses are required. When separate groups are responsible for each of these activities, each should be readily identifiable both in the primary publication and in all subsequent citations. We describe the publication policy of the Alzheimer Disease Neuroimaging Initiative (ADNI), its origins and its acceptance by the editorial and scientific communities.

1. Introduction

Both recognition of and responsibility for scholarly efforts need to reflect accurately how such efforts were organized and accomplished, as discussed in the accompanying editorial by Rohlfing and Poline (2011), who correctly identify ADNI's requirement for “non-authorship credit on the author byline” as a condition for the use of our data. This requirement is based on our belief that the value of a scientific publication depends on the integrity of experimental design and execution, as well as of the analysis and discussion of the resultant data by its authors. When these functions are performed by different groups of individuals, one cannot serve as the guarantor of the other's contribution. The responsibility for each of these activities must be clearly identifiable not only in the original publication, but in all subsequent citations. We believe that this is best achieved by the identification of those who can vouch for the integrity of the data as well as of the authors who propose an interpretation, in perpetuity. Our policy has been accepted without exception by an unusually large number of authors as well as by the editorial boards of a wide range of respected scholarly journals.

We recognized the need for an appropriate publication policy that would adhere as much as possible to current editorial practice but would also respect the requirements imposed by our mandated departure from the traditional model of academic endeavor. This need is best appreciated by comparison of our model with those of historical precedents, both in research activities and in editorial policies.

2. Historical considerations

Over the years the organization of scholarship has evolved into many forms, but the means of acknowledgement, though evolving, remain rooted in the past. The original model was that of a single individual who was both investigator and author. The terms author and investigator were used interchangeably, without ambiguity or compromise of the usual meaning of either word. There was no need for codification or technical definition, as the meaning was self-evident. Such an author was a guarantor not only of his or her analysis but also of the underlying data.

Over the ensuing decades, an increasing proportion of scientific and clinical research has been undertaken by large teams of collaborators, each contributing special skills and effort to a common scientific goal. While all are clearly contributors, only by extension of the vernacular meaning could each be conferred the title of “author.” The need for such collaborative groups was originally driven by the use of complex and expensive tools for data acquisition and also by the utility of large datasets for meaningful analyses. The former has long been the case in disciplines such as high-energy physics, the latter for certain biological and clinical investigations, such as ADNI, large clinical trials and, perhaps most famously, the Human Genome Project. In such large consortia, there are many more critical contributors than can be authors in any meaningful sense of the word. The situation is even more complex when authors analyze data from studies designed and executed by others.

However, the convention that equates scholarly contribution with authorship persists, in large part to protect its value as an academic currency. This practice has had the unfortunate result of obscuring for the reader the responsibility for any given aspect of the work. As research groups have grown larger, the disparity between these terms has widened, sometimes to an awkward degree. Authorship, in its vernacular sense, is difficult to envision, much less accomplish, with more than a handful of individuals. We believe that a group cannot be an author, though certainly a small number of individuals in such a group can each be. The effort to preserve the original equation of contribution with authorship has prompted parsing into categories such as “corporate authors,” “authors writing on behalf of the group,” “group authors,” and others. Many of these have been given technical definitions.

Although the adoption of vernacular words into a technical lexicon has a long (if not always venerable) history, such a practice can only be endorsed if the technical meaning adds precision. Such is not the case with scientific authorship, even though the original meaning was clear and useful. The ability of an author to analyze data, to extract meaningful findings, to make appropriate generalizations, and to communicate these cogently to the greater community is unquestionably a skill necessary for the success of the scientific enterprise. But limiting prominent recognition only to authors, threatens either to distort the meaning of the word beyond recognition, to encourage inefficient balkanization of research efforts, or even worse, to result in suboptimal utilization of data acquired at great cost not only to investigators, but also to those who funded their efforts.

It is best to speak plainly. Not all critical contributors can be authors.

3. Publication policies for large scientific consortia

There is a suitable alternative to the convention that equates scholarly contribution with authorship. As early as 1997 the editors of the British Medical Journal (BMJ) commented on the growing disparity between the vernacular and technical definitions of authorship. They suggested that “the concept of authorship in science was so broken that it should be scrapped and replaced by something different. Instead of authors there should be contributors and guarantors.” (Smith, 1997) In doing so, they referred to an earlier proposal by Drummond Rennie, an editor of JAMA [Rennie et al. 1997], and its subsequent adoption by the Lancet. (Horton, 1997.)

More recently, bodies such as the Committee on Publication Ethics (COPE), the Council of Science Editors (CSE), and the International Committee of Medical Journal Editors (ICMJE) have issued similar though not identical, guidelines for what they variously described as “collaborative,” “corporate,” “collective,” or “group-author” articles. Some distinguish “authors writing on behalf of the group.” Perhaps the most definitive guidance is that issued by the ICMJE, which reserves the technical title of author to only those individuals who fulfill each of three criteria: (1) substantial contributions to conception and design, 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. This definition of an author, especially the second criterion, certainly helps to reconcile the growing disparity between the vernacular definition and its technical definition(s) in the scientific literature.

Over the ensuing decade, the BMJ position has evolved (BMJ Publication Policy.) The editors have abandoned their earlier proposal that the term author be scrapped entirely. Their current recommendation is the use of three separate terms (1) author, as defined by the ICMJE,(2) contributor, which requires explicit statement of “who has contributed what to the planning, conduct, and reporting of the work,” and (3) identification which contributor(s) are responsible for the overall content as guarantor(s). We agree with this view, but note that separate guarantors are required for the data and for its analysis if the responsible parties are distinct. However, only partially answered by the BMJ's current policy is the need for commensurate recognition of all three in citations in subsequent publications by others. Because we consider this to be important, we depart from the BMJ proposal that non-author contributors be named only in a separate acknowledgement section appended to the end of the text. Such acknowledgement will not be evident when the work is cited by another publication.

In summary, we agree with the ICMJE definition of author, and rather hope that the terms “corporate,” “collective” or “group-author” be abandoned. We are, of course, powerless to do so, except in our own publication policy. Furthermore, we agree with the BMJ's current proposal for the additional terms “contributor” and “guarantor”, with two key provisos in those instances when other individuals analyze ADNI data from which to author papers: (1) We in ADNI assume the responsibility of guarantor of the integrity of our raw and processed data, but cannot be guarantors of the authors' analyses. By the same token, non-ADNI authors cannot be guarantors of the integrity of our data. (2) We do not believe that acknowledgement of contributorship for the overall ADNI organization at the end of a paper is appropriate, as it will not be apparent in citations in subsequent works.

4. Considerations underlying the ADNI publication policy

As will be explained, the ADNI data use policy was designed to be very liberal, arguably uniquely so From the outset, its purpose is the promotion of unfettered access to high quality raw or processed data shortly after its acquisition, by as broad a range of qualified scholars as possible. Such access is meant to assure maximal utility of an unusually expensive resource. For an undertaking such as ADNI, expensive data acquisition precludes economically justifiable duplicate investments by multiple groups working independently. Our sponsors required that this database be made available to as wide a variety of investigators as could use this resource profitably. To the extent that this can be accomplished without having “too many cooks spoiling the broth,” wide availability helps ensure that the investments of our sponsors yield maximum scientific benefit. Ideally, investigators should have access to data without requiring explicit sponsorship, which could impose restraints, bias or even censorship.

ADNI intentionally departed from a traditional model of biological research in which the same small group of investigators not only designs the experiments, but also collects the data, analyzes the results, and authors the paper. In the traditional model, more often than not, some potentially useful data languishes in unpublished notebooks. As has been discussed, in the traditional model, credit can be appropriately apportioned according to the last piece, “authorship,” which has become convenient shorthand that implicitly subsumes the earlier three activities as well. What is to be done when academic work is undertaken by a very large group, such as ADNI, in which the responsibilities are divided? How do several hundred investigators “author” a paper, in any meaningful sense of the word? Furthermore, what is appropriate for an enterprise with a uniquely liberal policy on data sharing? In ADNI, data sharing is not an afterthought after primary publication, but central to our mission.

However, ADNI also has a responsibility to the research community, to our sponsors and to the public at large, to ensure that this data not be misrepresented, either unintentionally by those who lack the expertise to interpret it properly, or, intentionally, by those with other than serious scholarly intent. This responsibility was delegated to the Data and Publication Committee, the members of which are the authors of this article.

Our novel approach to research requires an equally novel approach to publication and acknowledgement. Nevertheless, we must comply with standards of publication and acknowledgement recognized within the greater scientific community. There has never been any intention of imposing a specific interpretation of the analyses performed by non-ADNI authors as would be required were ADNI to assume the mantle of author or guarantor of such analyses. By the same token, non-ADNI authors cannot serve as guarantors of the integrity of our experimental design or the resultant data. Both must be acknowledged separately in perpetuity.

5. The ADNI Publication Policy

The ADNI data use and publication policy rests on two principles: (1) the expectation of both the ADNI Executive Committee and the National Institute of Aging (NIA) of the NIH that de-identified data from ADNI be made available to the general scientific community within a very short timeframe, and (2) the need for tracking of the impact of the contribution by the ADNI organization and its appropriate acknowledgement. The global contribution of ADNI is the result of a coordinated effort of many groups; including (1) those that conceived the project, wrote the protocols (Snyder et al., 2002) and proposed it to a consortium of the federal government, academic institutions, private companies and patient advocacy groups, chiefly the Alzheimer's Association (Weiner et al., 2010) (2) those that shepherded it administratively (the NIH) and created the infrastructure for additional support (the Foundation for the NIH); (3) those who continue to lead and coordinate the individual cores (Trojanowski et al., 2010; Aisen et al., 2010) responsible for extensive processing, quality control, cross-referencing, curation and distribution of both raw and processed data, as well as (4) those who recruit subjects and acquire the primary data (Weiner et al., 2010.) The contribution of ADNI is that of the group.

A detailed description of the ADNI project has been published (Weiner et al., 2010) and is accessible online (ADNI Project description.) In brief, the goal of ADNI is to validate brain imaging and biomarkers for clinical trials in Alzheimer's disease. Initially, ADNI enrolled 821 subjects at 57 sites in the USA and Canada, and all subjects (controls, those with mild cognitive impairment, and those with dementia due to Alzheimer's disease) had longitudinal measurements of cognition, MRI, PET, as well as biomarkers in blood and cerebrospinal fluid. All data acquired by ADNI is uploaded electronically to a central data base, and once the quality of the data is verified (usually within several days) all data is available without embargo to “any qualified scientist” who requests a password to our website http://www.UCLA/LONI/ADNI. For example, if a subject has a neuropsychological test, or an MRI scan, this de-identified raw data is made available shortly after its acquisition to any scientist in the world to whom access has been granted. Such release of all data as it is being acquired may well be unprecedented in the Neuroimaging community, but follows the example of the Bermuda Principles (Bermuda Principles, 1996), which assert that “all human genomic sequence information, generated by centres funded for large-scale human sequencing, should be freely available and in the public domain. ” ADNI investigators are not afforded an advantage of any earlier access than that enjoyed by other scholars. However, it is important that their efforts be recognized properly and that they take responsibility for the integrity of their data.

We on the Data and Publications Committee (DPC) of ADNI have endeavored to develop a policy that ensures appropriate recognition of scholarly efforts and that defines the extent of ADNI's responsibility, as well as could be accomplished within the constraints posed by existing editorial and citation systems. As was explained, we elected an approach that is similar but not identical to, that currently advocated by the British Medical Journal. The DPC designed the data sharing policy for ADNI and is charged with responsibility both for the release of data to qualified investigators (ADNI Data Use Agreement) and for the publication policy relating to such use (ADNI Publication Policy.)

From the earliest planning stages of ADNI, the intent has been that ADNI be a critical contributor and guarantor of the quality of the data that it was to generate, but not to reserve the right of authorship to itself or its members. Built into the design of ADNI, was the expectation that the data be examined by the largest possible spectrum of qualified investigators, many of whom would go on to author papers based on our curated data. We have been absolutely transparent in this claim from the outset. Because of the constraints of current editorial practice, the explanation of our role as contributors is described with an asterisk that directs readers to this explicit statement:

Data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.ucla.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report.”

However, the terms ‘author-line” and “group-author’ had become so entrenched, that, indeed, they had unfortunately crept into some earlier ADNI publications (Weiner et al., 2010.) This was an unintended error. We do not consider ourselves co-authors on papers written by other investigators who downloaded ADNI data and authored manuscripts based on their analyses.

Nevertheless, we do require recognition of important contributions by and the responsibilities of those investigators who designed the study, collected and curated the data, but who are not authors of such works.

Clarity is better served with the terms “contributor” and “guarantor” as proposed by the BMJ (Smith 1997.) Although we will focus on our specific experience with ADNI, we believe that many of these same considerations are germane to other large biomedical consortia.

We believe that the contribution of the ADNI organization is critical to such publications and should be recognized as such when impact is assessed. To ensure recognition commensurate with our contribution and to formally accept responsibility attendant to that contribution, we require

1. Acknowledgement of the funding institutions for ADNI in the support acknowledgement section of the manuscript

2. Acknowledgement of data gathering by ADNI personnel in the methods section of the manuscript

3. Acknowledgement of the contribution of the ADNI organization in the author line.

A recommended wording for all three is given in our published data use policy Signed agreement is a requirement for release of ADNI data to the investigator. Compliance is overseen by the DPC, which is empowered to approve modifications of wording when requested by the investigator and deemed appropriate. The first two requirements have not been questioned, whereas the latter has drawn occasional comment, such as that by Rohlfing and Poline (2011) who express concern that acknowledgement on the “author-line” implies that we or our individual members are authors. It does not. There is simply no other convenient place to put the acknowledgement in a form that will survive citation in subsequent publications under the present system.

ADNI is a critical contributor, but not an author, group-author or otherwise. ADNI does not just provide a tool for others to use nor does it just provide raw data. The value of ADNI data derives from the formulation of the project, the design of the study, as well as the recruitment and vetting of subjects by expert clinicians. Scanning is performed using novel procedures that ensure a high degree of consistency between sites. Provision of meaningful data for the greater scientific community requires exquisite quality control as well as sophisticated data cleaning and processing.

In summary, we agree with the technical definition of authorship by the ICMJE .However, we also agree with the current thinking of the editorial board of the BMJ on the importance of both “guarantor” and “contributor” as well as their definition of authorship, which is concordant with the views explicitly expressed by the ICMJE, and, implicitly by COPE and the CSE. This delineation is, we believe, at the heart of the matter. In our view, a group cannot be an author, nor can hundreds of individual members be. The sheer size of the ADNI group as well as the limitations of the breadth of our expertise precludes our taking on either the responsibilities or resultant credit of authorship in cases where other investigators download our data. Indeed, we explicitly require the exact description of our contribution and its limitations

We believe strongly that maximal scientific benefit from ADNI's efforts will accrue from ready access to our data by a broad range of scholars as well as from uncensored publication without the restraints that might be posed by a requirement of ADNI sponsorship. We do not desire (nor, in many instances, are we even capable of) the role of author in independent studies using our downloaded data. We also recognize that such authors cannot be guarantors of ADNI data and therefore we take on that responsibility ourselves. Furthermore we do not wish to impose ourselves as guarantors of their analyses, with all that such an imposition would imply.

But then, why require acknowledgement of non-author contributors in the “author line?” We agree with the Council of Science Editors (CSE) that “Bibliographic databases should enable users to retrieve citations to original articles by searching for the group name.” However, we find the use by CSE of the “group-author” term suboptimal, and prefer the terms “contributor” and “guarantor”, as proposed by the BMJ. Furthermore, we also believe that it is important that the role of critical non-author contributors and guarantors, such as ADNI, be recognized when a work is cited by another text. Neither can be accomplished by acknowledgement only in the body of the text or in an appended acknowledgement section, nor by reliance on grant or Pubmed numbers, neither of which are entered into the reference section of other manuscripts. Tracking citations for funding bodies does not require author-line acknowledgment. Transparent communication with scientific colleagues does. We note that many editorial bodies are silent on the exact placement of acknowledgements of non-author contributors, while others, including the BMJ, recommend acknowledgement of non-author contributors in the body of the text

We disagree.

Under the present structure of citation, these goals can only be accomplished either by including ADNI in the title or in what continues to be referred to as the “author line.” Of the two, we believe that latter to be less unwieldy. Although we would prefer that this be renamed the ‘author, contributor and guarantor line,” it is not our place to do so. However, we believe that the inclusion of the word ADNI (rather than the names of individual members) will not cause confusion either to readers, promotions committees or ADNI members themselves when they are preparing their own CV's. There is no ambiguity about which named individuals are authors and which group is a contributor. While the distinction between an organization such as ADNI and individuals named in the author-line is obvious, it is further reinforced by the required asterisk directing the reader to an explicit explanation.

6. Acceptance of the ADNI Publication policy

We developed the data use and publication policy in order to ensure proper recognition of the extent of our organization's contribution and responsibilities while promoting orderly access, widespread usage and rapid publication of our data. We propose that other large consortia consider the transparent terms “contributor” and “guarantor' as a more straightforward alternative to the sometimes cumbersome parsing of authorship into subcategories such as “group authorship,” and the like.

Our policy has enjoyed wide acceptance by a variety of constituents. To date, 2198 investigators have consented to our data use and publication policy by submitting 1590 applications, of which 1463 were approved initially, as were an additional 113 after clarification. By an order of magnitude, the majority of these applicants were university-based researchers, both the absolute number and proportion of which has grown geometrically since the inception of our program in 2007. A smaller and rather flat proportion of requests for ADNI data continues to come from pharmaceutical companies, biotechnology firms, governmental agencies, scanner manufacturers and others. Data applicants agreeing to our publication policy have come from all six inhabited continents. As of April 2011, the total number of ADNI data downloads world-wide numbers 1,223,452.

The ADNI publication policy has been widely accepted not only by academic, commercial and governmental investigators, but also by the editorial bodies of the journals to which non-ADNI authored manuscripts have been submitted. As of April 2011, 305 manuscripts have utilized ADNI data. Of these, 175 have already been published in print and another 26 electronically. In the 52 journals that published these manuscripts (TABLE) the editors have agreed to our acknowledgement policy, albeit three only after a request for clarification. In addition, 34 published articles refer to ADNI, but do not analyze ADNI data. These latter publications do not require our committee's review, nor are they governed by our publication policy. It has been our experience that a very liberal data use policy promotes analysis by a wide variety of investigators, often with divergent approaches. There have been several illuminating examples of independent investigators bringing unique approaches to the same data, and coming to different conclusions. Furthermore, many investigations were explorations of issues not even envisioned by the originators of ADNI. This is as it should be.

Table. Journals in which ADNI data have been published by non-ADNI authors.

Acta Neuropathologica

Advances in Alzheimer's and Parkinson's Disease

Alzheimer's & Dementia : the Journal of the Alzheimer's Association

Alzheimer Disease and Associated Disorders

Alzheimer's Research and Therapy

American Journal of Medical Genetics

American Journal of Neuroradiology

Annals of Neurology

Archives of Neurology

Behavioural Neurology

BMC Neurology.

Brain

Brain Imaging and Behavior

Cellular Biochemical and Biophysics

Cerebral Cortex

Computer Science

Current Alzheimer Research

Dementia and Geriatric Cognitive Disorders

Drugs and Aging

European Radiology

Experimental Gerontology

Hippocampus

Human Brain Mapping

IEEE Transactions of Medical Imaging

Information Sciences

Information Processing and Medical Imaging

International Journal of Alzheimer Disease

Journal of Alzheimer's Disease

Journal of Neurology, Neurosurgery, and Psychiatry

Journal of Neuropsychology

Journal of Neuroscience Methods

Journal of Visualized Experiments (JoVE)

Magnetic Resonance Imaging

Medical Image Analysis

Medical Image Computing and Computer-Assisted Intervention

Medical Physics

Molecular Neuroscience

Molecular Psychiatry

Neurobiology of Aging

Neurocomputing

NeuroImage

Neurology

Neuropsychologia

NeuroReport

Neuroscience Letters

NeuroSignals

Proceedings of the Society of Photo-optical Instrumentation Engineers (SPIE)

Public Library of Science One (PLoS One)

Public Library of Science (Genetics) (PLoS Gen)

Proceedings of the National Academy of Science U.S.A.

Radiology

The Journal of Neuroscience

Finally, we are encouraged by other large consortia that have adopted very liberal data use and publication policies, similar to our own. Examples of these consortia include other ADNI-like initiatives in other countries, the Parkinson's Progression Markers Initiative (PPMI) and theAustralian Imaging, Biomarker & Lifestyle Flagship Study of Ageing (AIBL).

To promote optimal productivity by all investigators, ADNI has established a publication policy that not only permits commensurate acknowledgement of authors and other critical contributors, but also delineates their respective roles and encourages their independence. It is critical that scholars not abandon consortia such as ours for more modest efforts, solely to increase the likelihood of being granted the title of author. We believe the choice of endeavor should be driven by the likelihood of achieving substantive scholarly goals, not by the vagaries of recognition and acknowledgement. The wide acceptance of our policy by investigators and editors alike, suggests that it has some merit.

This is an important topic. The Principal Investigator of ADNI, Prof Michael W. Weiner, invites further discussion by eMail directly to his address: michael.weiner@ucsf.edu<mailto:michael.weiner@ucsf.edu>.