Statistical analysis exposes dubious data in 2% of anaesthesia trials

Statistical analysis of over 5,000 published trials has highlighted 90 that contain dubious data.


This is equivalent to around 2% of the trials examined using statistical tools to identify anomalies hidden within the data.


The authors of the study into anaesthesia trials point out that there could be many reasons for these anomalies – such as poor trial methodology, but clearly fraud is a possibility too.


All of these trials had been published in prestigious journals, including: Anaesthesia; Anaesthesia and Analgesia; European Journal of Anaesthesiology; Journal of the American Medical Association; and the New England Journal of Medicine.


John Carlisle, a consultant anaesthetist at Torbay Hospital, carried out the analysis. He had previously used similar statistical tools to discover a high profile case of scientific fraud involving a Japanese anaesthesiologist who was found to have repeatedly fabricated data.


The tool works by comparing the baseline data, such as the height, sex, weight and blood pressure of trial participants, to known distributions of these variables in a random sample of the populations.


Carlisle reviewed data from 5,087 clinical trials published during the past 15 years. He concluded that 90 published trials had underlying statistical patterns that were unlikely to appear by chance in a credible dataset.


The data in 43 of the trials had only about a one in a quadrillion chance of having occurred by chance.


I wonder how long it will be before a statistical tool is developed to massage falsified data so it is harder to detect?


Not that I’m a pessimist of course…


Read more [here]


                                    +21,000 STUDENTS        +9,400 COMPANIES          +100 COUNTRIES


Novartis logo                        NHS logo                        Takeda logo                        Roche logo                        DHL logo                        Baxter logo                        King's College logo                        US AID logo                        Novo Nordisk logo                           Grunenthal logo                           Wellcome logo                           Ipsen logo                           BTG logo