Editor's Choice | This Week in BMJ | Press releases



BMJ No 7122 Volume 315

Education and debate Saturday 13 December 1997


Your letter failed to win a place...

Eyal Shahar

See Commentary, p 1609

Letters that comment on published work are treated differently from the original article itself. They are rarely subject to peer review, and scientific explanations are not usually given when they are rejected. Professor Shahar argues that this is unjustified and counterproductive to scientific inquiry, and that criticism of published work should be subject to peer review.

The quality of published scientific work is evaluated at least twice - by a handful of reviewers and editors during peer review and by an unknown number of readers after publication. Editorial peer review sometimes helps authors to improve their manuscripts, but more often it helps editors to decide between acceptance and rejection. Just as important - or perhaps even more important - are the unsolicited opinions of readers. Many of the readers are as qualified as the reviewers whose opinions contributed to the editorial decision.(1,2)

The voice of the reader is heard through letters written by the relatively few who formalise their critique in writing. Letters are sometimes better thought out than the original article. They may identify inaccuracies that were missed by formal peer review or uncover flaws in design, analysis, or interpretation. Peer review does not preclude error.(1) Although letter columns are considered important by editorial boards,(3) the fate of correspondence on published work is rarely determined by peer review. The editor(s) usually make the decision whether to publish, and rejection notes to authors are often standardised and contain little, if any, scientific explanation for the decision. Vague statements such as "in the face of fierce competition, your letter failed to win a place" or "many worthwhile contributions must be declined simply for lack of space" are typical.

Anecdotal experience

The example I provide below illustrates the shortcomings of editorial practices. It is a "letter to the editor" (coauthored by Paul G McGovern) that was rejected with no specific explanation. Since the letter challenges unequivocally the main conclusion of an article, either the challenge or the original conclusion must have been faulty. The letter can, therefore, only be rejected on the assumption that its contents are faulty. I am asking the reader to judge this assertion (ignoring the question of whether the letter's content is true or false). I also provide the letter as a test case of my proposal for peer review of correspondence, inviting concrete evaluation of its content that will either support or oppose the editorial verdict in this case.

Letter to the editor

Sir
Andreotti et al conclude that "among patients with acute myocardial infarction, those with prodromal unstable angina ... have remarkably faster responses to treatment with tissue plasminogen activator than those without such symptoms."(4) We take issue with their inference.

The study by Andreotti et al was an observational cohort study of 23 patients who had suffered a myocardial infarction. Of these patients, fourteen had experienced preinfarction angina and 9 had not ("exposed" and "unexposed" groups, respectively, in epidemiological terminology.) The only acceptable inference from this small study is that exposure to unstable angina before a myocardial infarction was associated with more rapid reperfusion and smaller infarcts than were observed in the absence of prodromal unstable angina. That the two groups of patients happened to be treated with tissue plasminogen activator (and by other means) further defines the cohort characteristics, but is irrelevant to the question of whether thrombolytic therapy is more effective in the presence of preinfarction angina than in its absence. As the authors acknowledge, reperfusion occurs spontaneously during the course of myocardial infarction and therefore its rate is far from being entirely determined by treatment with tissue plasminogen activator. For example, Andreotti et al would have observed exactly the same results if treatment with tissue plasminogen activator had identical effects in both groups of patients yet infarctions that follow unstable angina are associated with faster rates of spontaneous reperfusion (and tend to be smaller) than infarctions without prodromal unstable angina.

To show a differential effect of treatment with tissue plasminogen activator in the presence (versus absence) of preinfarction angina, one should demonstrate a statistically significant interaction between two effects: the effect of treatment with tissue plasminogen activator in myocardial infarction patients who had preinfarction angina and the effect of treatment with tissue plasminogen activator in those who did not. Each of these effects can only be estimated by comparing the reperfusion rate and infarction size in patients treated with tissue plasminogen activator to these measures in patients who were not, preferably by a randomised design. Unfortunately, such a design is no longer feasible since it is ethically unacceptable to withhold thrombolytic therapy from patients who should receive it.

Discussion

Only two legitimate reasons exist for the journal to reject this letter - most of its content was judged faulty or the editor(s) preferred to publish another letter with a similar message. Lack of space should not justify rejection because space should be made available for corrections, even at the expense of delaying the publication of new original articles. I saw no letter with a similar message in follow up correspondence.(5,6)

When a manuscript is rejected by a journal, the authors may get it published elsewhere. When a letter criticising a published article is rejected no such remedy is usually available. In this sense, an erroneous editorial decision to reject a letter may be more damaging to scientific progress than an erroneous decision to reject a manuscript.

The differential treatment of scientific correspondence and manuscripts is not unique to scientific journals. Scientists rarely cite criticism of original research,(2) and academic institutions give little or no credit for published letters.

The most truthful message in any particular case cannot be deduced from some general rule of importance, even if there were an empirical way of substantiating such a rule. What is important is not the origin of the message (for example, authors of a manuscript or authors of a letter) but the message itself. Is it scientific or perhaps pseudoscientific? Does it survive logical criticism or not? Differential treatment of scientific communications introduces a potential prejudice into the search for objective knowledge.

Peer review of manuscripts is based on criteria such as clarity, validity, originality, and relevance. Peer review of correspondence could follow the same path. Letters to the editor (including the reply of the authors of the original article, which usually escapes rejection) should be evaluated for their scientific merit, and their fate should be determined on specific grounds. A letter may be rejected, for example, because its argument is judged to be rhetorical, its content faulty, or the thoughts of the author poorly articulated. Unexplained decisions leave too much room for speculation and, sometimes, suspicion.

Opponents of my suggestion for peer review of correspondence may argue that the process is lengthy and that it is essential to publish follow up correspondence quickly, while the original article is still fresh in the reader's memory. Scientific progress, however, is not a race against an arbitrary deadline.

It might be argued that formulating specific criteria for evaluating letters might be difficult. But a scientific communication - that is, one that claims to advance objective knowledge - should lend itself to critical appraisal, above and beyond just "feeling" for its merit.

Some might claim that peer review for letters could lead to an infinite, regressive process of publishing letters that comment on letters, and that such a process would have to be stopped arbitrarily anyway. Most debate, however, fades naturally away after one or two rounds, and if it does not, peer review should be responsible for identifying reiterative stages of a correspondence and for stopping it. Remember too that arbitrary termination also happens with peer review of manuscripts since the reviewers' critiques are not subject to peer review.

In a recent article, Bhopal and Tonks asked, "If published critical comment is considered integral to research should it not be peer reviewed?"(2) My answer is "Yes, it should." Editors would do justice to science if they solicited peer review of correspondence, including peer review of the reply from the authors of the original article. Editors who object should provide authors of rejected letters to the editor with their own scientific review to support their decision.

"Errors may lurk even in our best tested theories. It is the responsibility of the professional to search for these errors."

Neil McIntyre, Karl Popper

I thank Lori Vitelli, Maureen Smith, Jacqueline Dekker, the anonymous reviewer, and the editorial board of the BMJ for helpful comments on earlier versions of the article.

(Accepted 4 March 1997)

Division of Epidemiology,
University of Minnesota,
Minneapolis,
MN 55454-1015,
USA
Eyal Shahar, associate professor

email: shahar@epivax.epi.umn.edu

References

1 Spodick D H. The peer review system and the editor's correspondence. Arch Intern Med 1981;141:1121.

2 Bhopal R S, Tonks A. The role of letters in reviewing research: always look for letters that follow original papers. BMJ 1994;308:1582-3.

3 International Committee of Medical Journal Editors. Statements from the Vancouver Group. BMJ 1989;299:1394-5.

4 Andreotti F, Pasceri V, Hackett D R, Davies G J, Haider A W, Maseri A. Preinfarction angina as a predictor of more rapid coronary thrombolysis in patients with acute myocardial infarction. N Engl J Med 1996;334:7-12.

5 Migrino R Q, Moliterno D J, Topol E J. Preinfarction angina. N Engl J Med 1996;335:59.

6 Kloner R A, Gibson M, Cannon C, Braunwald E. Preinfarction angina. N Engl J Med 1996;335:59-60.


Home | Current issue | Past issues | Classified ads | Career Focus | Feedback
Collections | About this site | About the BMJ | BMA | Medline