Is selective reporting of clinical research unethical as well as unscientific?

Background: Studies that do not confirm their prior hypotheses, otherwise c alled "negative" studies, receive less interest from different parties incl uding authors, editors and sponsors, and so, not to publish such studies is a common phenomenon. Opinions differ on whether or not this phenomenon int roduces imprecision into the assessment of health research and care. Object ive: The current paper gives arguments against and in favor of publishing " negative" trials, and tries to give suggestions for a more balanced approac h to this problem. Results: Arguments against publishing "negative" trials include: we need not publish erroneously "negative" trials; we need not pub lish a "negative" study out of worry that the favored treatment is inferior ; full-length reports of "negative" trials devaluate the quality of literat ure, because the data are usually not so important, and generally receive l ittle interest from readers, and so, not to publish them is a more or less "natural" matter of course. Arguments in favor of publishing "negative" tri als include: no report reduces the flow of information because "negative" t rials provide at least some evidence and balance against the overwhelming p ower of positive data readily accepted for publication; no report violates the promise to patient participants; studies that do not confirm prior hypo theses are especially important; not-publishing leads to;unnecessary repeti tion of research. Initially, trials were frequently "negative" not only due to lack of power but also due to inappropriate hypotheses and poor designs . Currently, this is less so, and the issue of selective reporting, therefo re, needs to be reassessed. Suggestions for a more balanced approach to the problem of selective reporting might include: careful planning before the trial begins, reduces the chance of biased and erroneously "negative" trial s; any trial, "positive" or "negative", provides probabilities rather than truths; this notion does not explain away publication bias but does make it less of a problem; "negative" trials may not be appropriate for general jo urnals but very relevant to specialist journals as well as other organs of specialist groups; ethical committees and trial review boards should addres s the issue of publishing as part of their function. Conclusion: Data from properly executed trials should routinely be made available. However, we sh ould not forget that the empirical observations provided by clinical trials , are statistically tested, and that statistics are based merely on probabi lities. It means that we must consider a more philosophical attitude to cli nical trial evidence in terms of acceptance that scientific truths are rare ly absolute.