IN THIS issue of the ARCHIVES, Klassen et al1 contribute an interesting article on the problem of publication bias. They examined abstracts presented at Society for Pediatric Research meetings from 1992 through 1995 and found that only 59% of randomized controlled trials presented at the meetings were subsequently published. Abstracts were more likely to be published as a full study if, using some criteria for statistical significance, they either reported that the outcomes of the treated group were superior to those of the control group or reported that a newer therapy was at least equivalent to an older treatment. To rephrase this in the other direction, studies were more often sent to oblivion if they failed to report good news about newer therapies.
Rivara FP, Cummings P. Publication BiasThe Problem and Some Suggestions. Arch Pediatr Adolesc Med. 2002;156(5):424-425. doi:10.1001/archpedi.156.5.424