New tool flags randomised controlled trials that may contain false dataBMJ 2017; 357 doi: https://doi.org/10.1136/bmj.j2743 (Published 07 June 2017) Cite this as: BMJ 2017;357:j2743
- Nigel Hawkes
Dozens of randomised controlled trials may contain false or fraudulent data, a statistical analysis of more than 5000 papers has shown.1
John Carlisle, a consultant anaesthetist at Torbay Hospital in Devon, has developed a statistical method for assessing whether the baseline data in a reported trial are consistent with the distribution people would expect. The method can be used to identify papers in which demographic data—such as age or body weight—exhibit either too wide or too narrow a distribution to be true.
In earlier papers Carlisle examined the work of Yoshitaka Fujii, a Japanese anaesthetist with more than 190 retractions to his name, and found that his baseline data often failed this test. Carlisle has now searched 5087 randomised controlled trials published from 2000 to 2015 in six anaesthesia journals, as well as in JAMA and the New …