Reporting Research, Retraction of Results, and Responsibility

	ARTICLE

TOP ARTICLE REFERENCES

When a manuscript is submitted to a scientific journal, proof for its verity and worth rests completely with the author. If the journal accepts the manuscript, it is conferring a nihil obstat, if not an imprimatur, on the work. And when published, the paper carries the journal's aegis and authority (1). A portion of responsibility for the presented work now shifts from the author to the journal, and the editor accepts responsibility that the paper meets the journal's standards for originality, validity, and importance. Authors view acceptance of a manuscript as the completion of a piece of workbut for the research content of the paper, it is only the beginning. To contribute to the research enterprise, other investigators need to take the findings and build on them. But the construction of science requires solid bricks, not cardboard.

When scientists discover knowledge, they have a responsibility to publish it. If the findings are later found to contain fraudulent data, journals have a responsibility to make it known to readers. Some editors claim that scientific misconduct falls outside their mandate (2). But editors share responsibility for the integrity of work published in their journal, and, when necessary, for correcting the scientific record (3).

In this issue of AJRCCM, a group of authors describe every researcher's nightmare: they found a member of their team fabricating data. As a result, they checked data collected by that person in previously published papers. They found reasons to doubt the verity of about half the data of a previous publication in AJRCCM and a lesser proportion in a second article in the Journal. Yes, it is disappointing that scientific misconduct occurred in this laboratory, but we recognize the authors for honesty, accepting responsibility, and faithfully informing the Journal and its readers.

With most cases of scientific misconduct, the entire dataset is deemed invalid (4). What is unusual in this case is that the authors considered half or more of the data to be solid, and when reanalyzed, they claim the conclusions do not change from the original reports. What is also unusual is that the authors left it for the editor to decide whether to publish an erratum, retraction or reanalysis in AJRCCM, or direct readers to reanalyses posted on their institution's website. This posed a dilemma for me, because accepted guidelines for handling this particular problem do not exist (7). Having sought advice from biostatisticians, ethicists, jurists, sociologists and other editors, I decided that neither paper, considered as an entity in itself and as originally published in AJRCCM, was still valid. To correct the scientific record, I asked the authors to write retractions for both papers. For the data they considered solid, I asked the authors to submit two new manuscripts based on reanalyses. I did not consider it sufficient for AJRCCM's readers to have access to the reanalyses only through the website of the authors' institution: the posted material would not be peer-reviewed and the Journal would not control the permanency of the record. The new manuscripts underwent peer review but did not meet the standards for publication in AJRCCMpartly because the new data sets have less statistical power.

This is the first time that a paper previously published in AJRCCM has been retracted. When a journal retracts a paper it is saying to readers to treat the paper as if it had never existed (1). Editors handle retractions in a very uneven manner (6). Some refuse to publish any notice and others bury them in unnumbered pages (5). As a consequence, researchers may unwittingly build on specious work (8). The International Committee of Medical Journal Editors has developed guidelines for handling retractions (9). A retraction should be labeled as such, published in a prominent place in the journal, listed in the table of contents, the original article referenced, and the reason for retracting the paper stated. The term retraction has a specific procedural advantage. The National Libray of Medicine applies a tag to the specified paper (6), and in all future electronic searches, readers will be alerted to the problem with the previously published paper (10). The tag does not distinguish between honest errors and fabricated data (3), such a distinction is not the purpose of the tag, but, instead, to protect researchers against building on hollow publications. This process for correcting the scientific record underscores how science is a communal enterprise, and one based on the sharing of information.

Scientific misconduct holds enormous interest for the public and the media (11). In part, this is because science is an endeavor devoted to the discovery of truth. Also, the public wants to feel confident about the professionals in whom it invests. Before the 1980s, researchers handled scientific misconduct internally. But attempts to conceal and minimize the problem (12) caused the public to become disillusioned. The age of paternalism in science has passed, and we live in a time of transparency. To maintain the public trust, scientific institutions, including journals, need to accept openness and accountability. Anything less suggests complacency and a disregard for public investment. And attempts to limit the press's interest only enhance their suspicions (11, 15).

All participants in the research enterprise must accept responsibility for maintaining the integrity of published literature, or trust in science will erode. When fraud comes to light after a paper has been published, authors have a responsibility for informing the journal (15)as the present authors have done. When an allegation of scientific misconduct is raised, institutions have a responsibility for investigating it (16). Funding agencies are responsible for seeing that the investigation is thorough and that the respondent receives due process (7). Editors have a responsibility for contacting involved parties when an allegation is made, asking to be informed about the results of the investigation, and setting the publication record straight (3, 9). But editors are not responsible for the conduct of investigations or identification of miscreants (7), and they must not allow themselves to be viewed as science police. The quality of a scientific journal depends on authors being able to look at the editor as a colleague, without which the fabric of trust in scientific publishing is impaired.

How can we do better? The peer-review system is rarely successful in detecting fabricated data (17). While referees are extremely skeptical about an author's scientific method and the worth of the presented data, they trust that the manuscript is based on honest observations. A number of well publicized cases of alleged misconduct in the early 1990s led to shrill calls for closer audits. Despite good intentions, further regulatory constraints would do more harm than good and they have a chilling effect on honest researchers (18). Developing rules and procedures with the intent of catching every falsifier would be damaging to sciencea communal enterprise based on trust. But the scientific community needs to recognize that scientific misconduct will not disappear (21)serious students of the subject say the number of miscreants is probably irreducible (16). And the community needs to do a better job at setting and internalizing standards for ethical research practiceshigher than the merely legaland to proactively address issues of accountability and the handling of misconduct (2, 12, 16). Certainly, moral indignation and self-righteousness have no place in the handling of these episodesinstead, the events emphasize everyone's vulnerability and the need to stay humble.

	References

TOP ARTICLE REFERENCES

1. Hammerschmidt, D. E., and A. G. Gross. 1994. Withdrawal of aegis? So what's that? J. Lab. Clin. Med. 123: 792-794 [Medline].

2. Woolf, P. K.. 1987. Ensuring integrity in biomedical publication. J.A.M.A. 258: 3424-3427 [Abstract/Free Full Text].

3. Caelleigh, A. S.. 1993. Role of the journal editor in sustaining integrity in research. Acad. Med. 68: S23-S29 [Medline].

4. Budd, J. M., M. E. Sievert, and T. R. Schultz. 1998. Phenomena of retraction: reasons for retraction and citations to the publications. J.A.M.A. 280: 296-297 [Abstract/Free Full Text].

5. Friedman, P. J.. 1990. Correcting the literature following fraudulent publication. J.A.M.A. 263: 1416-1419 [Abstract/Free Full Text].

6. Parrish, D. M.. 1999. Scientific misconduct and correcting the scientific literature. Acad. Med. 74: 221-230 [Medline].

7. Office of Research Integrity. 2000. Managing allegations of scientific misconduct: a guidance document for editors. Office of Public Health and Science, U.S. Department of Health and Human Services.

8. Whitely, M. P., D. Rennie, and A. W. Hafner. 1994. The scientific community's response to evidence of fraudulent publication: the Robert Slutsky case. J.A.M.A. 272: 170-173 [Abstract/Free Full Text].

9. International Committee of Medical Journal Editors. 1997. Uniform requirements for manuscripts submitted to biomedical journals. N. Engl. J. Med. 336:309-315 and http://www.acponline.org/journals/resource/unifreqr.htm

10. Lindberg, D.. 1987. Retraction of research findings. Science 235: 1308 [Medline].

11. Center for Health Communication Harvard School of Public Health. 1996. Report of the conference on "science, technology, and the media." President and Fellows of Harvard College.

12. Woolf, P.. 1981. Fraud in science: how much, how serious? Hastings Cent. Rep. 212: 9-14 .

13. Wilmshurst, P.. 1997. The code of silence. Lancet 349: 567-569 [Medline].

14. LaFollette, M. C. 1996. Paychecks on a Saturday night: a brief history and analysis of the politics of integrity in the United States. In S. Lock and F. Wells, editors. Fraud and Misconduct in Medical Research, 2nd ed. BMJ Publishing Group. 1-13.

15. Angell, M., and J. P. Kassirer. 1994. Setting the record straight in the breast-cancer trials. N. Engl. J. Med. 330: 1448-1450 [Free Full Text].

16. Gunsalus, C. K.. 1993. Institutional structure to ensure research integrity. Acad. Med. 68: S33-S38 [Medline].

17. Relman, A. S. 1990. The value of peer review. In J. C. Bailer III, M. Angell, S. Boots, E. S. Myers, N. Palmer, N. Shipley, and P. Woolf, editors. Ethics and Policy in Scientific Publication. Council of Biology Editors, Inc., Bethesda, MD. 272-277.

18. Kelves, D. J. 1998. The Baltimore Case: A Trial of Politics, Science, and Character. Norton, New York. 135-151, 289-308.

19. Healy, B.. 1993. The Dingell hearings on scientific misconduct: blunt instruments indeed. N. Engl. J. Med. 329: 725-727 [Free Full Text].

20. Kennedy, D.. 1996. The Baltimore affair: let's not forget what went wrong. Nature Med. 2: 843-844 [Medline].

21. Lock, S.. 1997. Fraud in medical research. J. Royal Coll. Physicians Lond. 31: 90-94 .