Writing in the Journal of Clinical Epidemiology, a Dr Yvo Smulders of the Netherlands makes a proposal: A two-step manuscript submission process can reduce publication bias Smulder's point is that scientific manuscripts should be submitted for peer review with the results and discussion omitted. The reviewers would judge the submission on the strength of the methods and the introduction alone. If they recommended publication, the authors would then send them the full paper. The reviewers would then have a chance to change their mind and reject it, or ask for further experiments to be carried out, but the 'bar' for this to happen would be high. Hence the scope for reviewer-based publication bias, the tendency to favour 'positive' results, would be reduced. Reviewers would have to make a decision on the basis of the experiment itself, regardless of whether the results were positive or not. Smulders says that it would also ease the burden on reviewers in terms of the volume of material they'd need to digest. It's a clever notion (and, as Smulders points out, not a new one; it dates to the 1970s, but has never taken off.) The proposal is reminiscent of the preregistration with peer pre-review modelwhich I've advocated. The difference is that in the latter case, the authors submit the introduction and methods before the study has been conducted while in 'two-step' submission, the results are already there, just not revealed until later in the process.
The difference is that unlike preregistration, two-step review would not prevent publication bias (or other questionable practices) on the author's side. Two-step would, however, reduce the incentive for such bias - why fish so hard for a positive result if you know your study would make it into a good journal on the strengths of its methods? But the proposal would certainly be a step in the right direction and, in fact, could form a natural stepping-stone to a preregistration system.
Smulders YM (2013). A two-step manuscript submission process can reduce publication bias. Journal of Clinical Epidemiology PMID: 23845183
