——> Follow us on TWITTER

The STROBE Statement, BMJ 2007

Certification of epidemiologists has been discussed from time to time and some have adopted such a policy and others have been reluctant for the reason that many good epidemiologists have achieved their skills in unorthodox ways. Success in science often depends upon many other things that the ability to pass exams and do what the professor tells you to do.

The STROBE (strengthening the reporting of observational studies in epidemiology) statements suggest adopting certain standards for reporting non experimental epidemiologic studies. Should we endorse these statements and try to harmonize our writing even more than it is at present?

At a talk at UCLA Somerset Maugham once said: ‘There are 3 rules for writing a novel. Unfortunately, nobody knows what they are.’ In reality he meant ‘fortunately’.

Here we get a whole set of rules; a full checklist. It will certainly make our papers more boring but will it make them more informative? Only 3 rules may be better for example of the type; 1) If you cannot write the introduction on 7 lines, reconsider if the paper is worth writing at all (the Philip Cole rule) 2) make sure the reader is informed about the most important sources of error 3) don’t overstate you conclusions.

A number of well known colleagues and editors have now written up much more detailed rules and call them ’a Statement’. Using the word ‘Statement’ is strong terminology that they themselves soften by adding the word ‘guidelines’ to the title of their paper which has been published in multiple journals 1.

Similar guidelines have on the other hand worked well for randomized trials and other experimental laboratory studies. The CONSORT Statement latest published in 2001 has increased the quality of reporting results from randomized clinical trials. Is a similar statement for none randomized epidemiologic studies what we have been waiting for, and perhaps something the IEA should have done long ago?

The first reason for being sceptical is that non-experimental studies cannot be put into a simple standard protocol. The best studies adapt to specific conditions that threaten validity of study results. Although non-experimental designs fall into certain categories (and there are many more than mentioned in the paper), variations within these categories are many.

Statements may replace critical thinking and may make peer reviewing of manuscripts more ‘mechanical’. It is easier to check if guidelines have been followed than to check if it all makes sense. A poor study may well appear better than it is because following the guidelines may hide the details that carry the most important information.

The authors come up with 22 items on their checklist. Most of them are quite general and can be found in any textbook on how to write a scientific paper. Most of these are sufficiently imprecise to do any harm. Some of these guidelines will probably help to improve the quality of reporting results from non experimental studies.

The Statement does not include a number of very important design options such as; the different case-only studies, case-time-crossover studies, Mendelian Randomization or other designs used in genetic studies. They do not even suggest reporting that allows a more refined classification of, for example, the case control study (case-non case, case-cohort or density sampling of cases and controls, or whether the study base is defined by the case series or defined a priori).

Their emphasis is on frequency type statistics where P values have done much harm. It is also unclear why generalizability should be emphasized outside studies that rest upon surveys.

In conclusion, let’s wait and see. The Statement may provide a useful checklist that will not be taken too literally and could therefore help in reporting findings from non-experimental epidemiologic studies. The editors of IJE (2007; 36:946-48) have suggested a positive approach and we may give the authors the benefit of the doubt.

Jorn Olsen, Shah Ebrahim, Chitr Sitthi-amorn

References

  1. See Lancet, 20 October 2007; Annals of Internal Medicine 16 October 2007; BMJ, 20 October 2007; Epidemiology November 2007; Preventive Medicine October 2007; PLoS Medicine 20 October 2007
Share Button

Comments are closed.