Bias in scientific publishing rife

The reluctance for negative or neutral results to be published in scientific or medical journals continues to bias research and encourage the proliferation of false medical truths, found an international team of researchers.

The negative results conundrum has plagued scientific publication for many years as journals tend to be more inclined to publish significant or positive results rather than the results of a study that has found no significant effect.

Acknowledgement of this phenomenon resulted in the creation of a series of journals of 'negative results', although they are still significantly outnumbered by journals of 'positive results'.

In this latest study, researchers over the span on six years surveyed 525 unique publications looking at 1,359 animal experiments focusing on 16 stroke interventions.

They found a substantial self selection bias of the papers published, with 214 experiments that had been conducted but which were not published due to negative or neutral results.

“We estimate that around one sixth of experiments remain unpublished,” said Professor David Howells of the Florey Neuroscience Institutes.

“The consequences of drawing erroneous conclusions, based on incomplete scientific literature, are particularly troubling if unnecessary and potentially risky clinical trials are conducted as a result.”

The effect of the self selection bias was an overstatement of the efficacy of the treatments being studied. In fact, with the unreported experiments factored in, the efficacy is reduced by a significant seven per cent, from 31.2 per cent to 23.8 per cent.

“If data about experiments in animal models of human diseases is being collected but not published, we are concerned that we are creating a false medical truths about treatments that will affect us," said Howells.

"These false truths eventually are filtered down to being taught to our undergraduate science and medical students, and become reading materials for existing practitioners.

Howells also believes that this phenomenon isn't restricted to studies of stroke treatments, but is rife amongst other aspects of scientific and medical research.

This study was published in the journal Public Library of Science (PLoS) Biology.