Why we are proud to publish well-performed negative clinical studies?

1.1 Ethically, it is equally important to publish well-performed negative studies

The excellent paper by Jarkko Kalliomäki et al. [1] is at least as important as studies that report the new drug to be effective. The patients have been exposed to the drug, and therefore it is unethical not toreport about a negative study. Previously, pharmaceutical companies sometimes – possibly much more often than we knew – decided to delay or cancel publication of negative results. A consortium of major pharmaceutical companies have recently agreed to submit all phase III clinical trial results to a peer-reviewed journal no later than 18 months after the study is completed. The ball is now in the hands of the journals and their editors. Attitudes to clinical trials with negative results have recently changed to the point, where several journals dedicatedto publishing only negative results have emerged.

1.2 The pressure to publish positive findings

The pressure to publish positive findings used to be so strong that changing the primary outcome of a study, which seemed to be negative, to something else that gave p-values indicating efficacy, has been one of the most common forms of scientific misconduct. Compulsory registration of clinical trial plans to databases before recruitment of the patients can be helpful here. Fiddling with the primary outcome can be found out by comparing the goals given in the registered trial protocol to those of the manu scriptor published paper. In a sample of 147 papers in cardiology, rheumatology, and gastroenterology, 31% had some discrepancy between the registered and the published outcomes [2].

1.3 What is wrong about changing the primary outcome?

It takes away the basis for statistical analysis. Think about a sunny beach. Good. Then think about throwing a handful of pebbles randomly in the air. The likelihood of all landing within a small circle drawn in the sand is small. But, if you first throw the stones and then draw circles around them,the like lihood is very high. Changing the primary outcome of a clinical trial after the statistical analysis is similar. Post hoc analyses or analysing unplanned subgroups can givegreat ideas for fur ther research, but cannot prove anything. The reader of ascientific report mustbeable to know if the analysis was really planned in advance.

1.4 Is it better to just keep the negative results in the file drawer?

Publication bias in psychiatry was discussed recently; it was shown that adding unpublished data to meta-analyses significantly reduced the estimated efficacy of antidepressants [3]. A study on publication bias in anaesthesiology journals with data from 2008 to 2009 demonstrated that it is more likely to get a study published, especially in the journals with the highest impact factor, if the results are positive [4]. In fact, the problem is not limited to getting negative results published. Negative studies are less likely to be published rapidly (time lag bias), less likely to be published in English language, and less likely to be cited (http://www.cochrane-net.org/openlearning/html/mod15-2.htm).

Additionally, negative studies are less likely to be published more than once (multiple publication bias), which is less of a problem, as this most often is a form of scientific fraud. Naturally, it may be that some of the negative studies have not been adequately powered, or there may have been some other methodological problems.

In the field of cannabinoid research, the pressure to show positive results has been especially strong. Patient associations and lobbying groups in North America and elsewhere have created strong pressure to do research on the medical use of cannabinoids and, more importantly, to produce positive results. Arguments for more liberal medicinal use of cannabinoids in the press and in the internet are often basedonopinion rather than evidence. In clinical studies in acute postoperative pain, the analgesic effect of cannabinoids has been modest [5,6,7]. Furthermore, in some of the studies cannabinoids have produced significant central adverse effects, such as sedation, dizziness, dry mouth, and cognitive impairment [8].