Many epidemiologists have not yet heard of STROBE. The acronym stands for “Strengthening the Reporting of Observational Studies in Epidemiology,” and represents an effort to provide guidelines on how epidemiologic results should be reported.
STROBE springs from the success of CONSORT, a set of guidelines for reporting randomized clinical trials (RTCs). There seems to be little question that the CONSORT guidelines have encouraged more complete and open reporting of clinical trials. Do such formulas apply equally well to observational research?
As the Editors of Epidemiology, we have reservations. In an earlier incarnation, the “S” in STROBE stood for “Standards”—suggesting judgmental intent. To their great credit, the authors of STROBE have retreated from the appearance of setting standards, and they provide instead a set of recommendations. We believe these recommendations deserve consideration. In this issue, we provide 2 documents on STROBE—a set of recommended guidelines1 (simultaneously published in BMJ, Bulletin of the World Health Organization, The Lancet, PLoS Medicine, and Preventive Medicine) and a longer explanatory document2 (appearing also in Annals of Internal Medicine and PLoS Medicine).
Epidemiology is not ready to endorse STROBE. We have concerns about any effort to formalize reporting in a field as heterogeneous as observational epidemiology. At the same time, we recognize STROBE as a serious effort to improve the quality of epidemiologic reports. Our purpose here is to inform epidemiologists about STROBE, and to encourage input on its further development. Toward this end, we provide not only the STROBE documents themselves, but also reflections on this effort from leading figures in our field.3–5 Jan Vandenbroucke, an author of STROBE (and former member of our Editorial Board), provides an insider's view on the dilemmas and tradeoffs that have shaped STROBE to date.6
RTC Origins and Remnants
The fact that STROBE has sprung from the success of CONSORT is itself a reason for concern. Clinical trials are narrowly constructed and thus well suited to explicit guidelines. By comparison, observational studies are idiosyncratic. Vandenbroucke6 describes his own concerns about STROBE's origins in the CONSORT experience. He expresses satisfaction that the transition to the world of observational studies has been successful. Nonetheless, there are odd remnants of clinical-trial thinking in STROBE. (For example, how many observational studies have actually been terminated because the results achieved statistical significance?2(section 10) We would welcome a more thorough purging of inapplicable RTC concepts, and a clear statement of the crucial distinctions between randomized trials and observational studies—and hence between CONSORT and STROBE.
On Reporting, Not Researching
We agree there is room for improvement of the presentation of epidemiologic data. As authors, we all experience the temptations to gloss over study shortcomings and to dwell on the importance of our results. Editors and reviewers have an essential function in counter-balancing such tendencies; Epidemiology has published specific guidance for this purpose.7 As Editors, we would be willing to consider a value-free set of recommendations that focuses narrowly on reporting—guidelines that do not include as their mission the improvement of the conduct of studies. It is unfortunate that STROBE ventures into matters of study conduct, with (for example) debatable opinions on the conditions under which matching should be done. Reporting guidelines should avoid attempts to define what constitutes “good” epidemiology.
From Recommendations to Judgment
Epidemiology is famously opportunistic. It was the accidental contamination of the Broad Street pump that provided John Snow with dramatic demonstration that cholera is water-bourne. This opportunism is a strength, but the corresponding lack of rigid structure in epidemiology makes our craft particularly vulnerable to criticism. One of the Editors' deepest concerns is that STROBE will evolve from a set of reporting guidelines into a tool for judging the studies themselves. STROBE is not an appropriate basis for the evaluation of studies. Although the authors of STROBE may agree, the documents lack a clear statement to this effect.
The Beginning, Not the End
Rothman and Poole3 suggest that any set of guidelines such as STROBE should be an evolving process. We concur. The authors of STROBE say that comments are welcome (although the specific process for developing future revisions is unclear). We present the complete STROBE documents to our readers in hopes of stimulating wide discussion—and ultimately improvement—of the STROBE recommendations.
Towards that end, we offer these suggestions:
1. STROBE should acknowledge the differences between randomized trials and observational studies, and the consequent differences between CONSORT and STROBE.
2. STROBE should explicitly restrict its function to the improvement of reports on observational studies. Efforts to improve the conduct of studies belong elsewhere.
3. STROBE should expressly discourage the use of its guidelines for the evaluation of studies or study results. STROBE's role is to provide aids for improving presentation, not for judging content.
4. Finally, and perhaps most important, STROBE should give clear emphasis to the importance of good judgment over any specific criteria. We should not allow the blindly applied rule to trump the creative exception.
1. von Elm E, Altman D, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. Epidemiology. 2007;18:800–804.
2. Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Epidemiology. 2007;18:805–835.
3. Rothman KJ, Poole C. Some guidelines on guidelines: they should come with expiration dates. Epidemiology. 2007;18:794–796.
4. MacMahon B, Weiss N. Is there a dark phase of this STROBE? Epidemiology. 2007;18:791.
5. Kuller LH, Goldstein BD. Suggestions for STROBE recommendations. Epidemiology. 2007;18:792–793.
6. Vandenbroucke JP. The making of STROBE. Epidemiology. 2007;18:797–799.
7. Sandler DP. On revealing what we'd rather hide: the problem of describing study participation. Epidemiology. 2002;13:117.
© 2007 Lippincott Williams & Wilkins, Inc.