Denominator problem needs to be addressed

EDITOR—During my personal exchange of material with Kjaergard and Als-Nielsen (I provided them with the equipoise scale, our data extraction forms, etc), I commented that their association between authors' competing interests and conclusions could have been explained by at least two additional types of bias. The first is pervasive authors' self selection bias (the authors tend to send their best pieces to a high impact journal such as the BMJ) and the second is the bias of the BMJ's editors.1

The major problem with this type of research is that its results can never be completely reliable until the problem of denominators is addressed. The journal editors' decision accepting or rejecting a given paper for publication can potentially seriously skew the distribution of studies with negative and positive results. I believe that the editors of the BMJ have a unique opportunity to inform …