Editorial decisions are often considered to be a source of publication bias, simply because editors seem to favour manuscripts with positive or significant results. Critics of this approach argue (justifiably, in our view) that negative or non-significant findings in clinical pharmacology studies can be equally important to funding bodies, researchers and study participants, provided the study is sufficiently powered to exclude a biologically meaningful difference between active and comparator. (It is for this reason that we favour presenting the difference together with its 95% confidence interval for important negative findings.) A frustrated scientist once complained that he would spend months fruitlessly testing a particular molecule, only to discover, in a conversation with another group, that previous studies (unpublished) had been consistently negative. Why waste resources on further developing and experimenting with interventions where the available evidence indicates little prospect of clinical utility?Looking back at 2008, BJCP has not shied away from publishing 'non-significant' but important studies. Take, for example, a randomized controlled trial of dexibuprofen in 255 children with upper respiratory tract infection [1]. Yoon and colleagues found no significant advantage in antipyretic efficacy or tolerability for dexibuprofen as compared to plain old ibuprofen, despite theoretical pharmacodynamic and pharmacokinetic reasons why the dexibuprofen enantiomer might be preferable. The findings of this study are consistent with trials of dexibuprofen in osteoarthritis which failed to demonstrate superior efficacy or safety over established analgesics. Is further research into dexibuprofen worth pursuing, or is it time to decide that 'as good as (but no better than) other analgesics' is simply not good enough for a crowded therapeutic arena?Likewise, a systematic review reports little success in the hunt for genetic factors that might predict aspirin resistance [2]. Goodman and colleagues reviewed 31 studies covering 50 polymorphisms in 11 genes in 2834 subjects, and found only a heterogeneous and weak link between the PlA1/A2 variant and aspirin resistance in healthy subjects. The potential association was not robust, as evidenced by considerable variation in the odds ratio (and statistical significance) arising from different laboratory assay methods, presence of cardiovascular disease, and choice of statistical model for the meta-analysis. Before going on to pursue polymorphism number 51, it would be better to standardize laboratory assays, exclude noncompliance with study protocols, and reach a consensus on the definition of aspirin resistance.Non-significant results may be helpful in reassuring patients and medical practitioners in the field of drug safety. Grosso and colleagues applied an innovative technique, the self-controlled case-series, to investigate serious adverse reactions with strontium ranelate [3]. This novel approach uses patients as their own controls, comparing the relative incidenc...

No abstract