Yoshitaka Fujii, a Japanese researcher in anesthesiology and ophthalmology, published more than 200 papers between 1993 and 2011, mostly studying the effect of drugs intended to prevent nausea after surgery.
A whopping 183 of these papers have now been retracted — 7 percent of all retracted papers between 1980 and 2011. It's been shown that Fujii simply made up the data for at least 171 of them, making him the most prolific fraudster in modern science.
Writing at Nautilus, Adam Marcus and Ivan Oransky — co-founders of the site Retraction Watch — tell the fascinating story of how Fujii got caught. It's definitely worth reading in full and explains a lot about the blind spots in modern science that allow this type of fraud to happen.
As the Nautlius piece describes, one of Fujii's papers first raised eyebrows back in 2000, when other researchers noticed his data looked too "neat": it didn't show the random noise and variation present in real-world data sets. Still, the journal Anesthesia & Analgesia published the paper and went on to publish several more of his over the years.
But Fujii switched fields, publishing mostly in ophthalmology, seemingly to avoid suspicion — and many scientists continued to quietly doubt the truth of his data. In 2010, the journal Anaesthesia published an editorial on the broader problem of identifying fraud — citing Fujii's 2010 paper as an example — prompting British anesthetist John Carlisle to write in, lamenting that the paper was still part of the scientific literature. Anaesthesia's editor challenged Carlisle to look at the rest of Fujii's work, and when he did, he found the vast majority of it was equally suspect.
Carlisle's analysis involved looking at how unrelated variables —such as height and blood pressure — differed between placebo and experimental groups before the start of Fujii's clinical trial. In 171 papers, it turned out, the two groups showed huge differences in these variables that you simply wouldn't expect to find among real people. Carlisle found them because the data didn't come from real people: Fujii had made it all up.
How science allows fraud to happen
One of the bigger problems here is that as part of the peer-review process, reviewers seldom look at raw data or perform these sorts of analyses to see if it holds up. Just yesterday, news came out that a highly publicized paper on same-sex marriage, published in the prestigious journal Science, was based on entirely fabricated data.
Fraudsters can also get fake data published in other ways. The Taiwanese researcher Peter Chen, for instance, managed to get 60 fraudulent papers on acoustics published between 2010 and 2014, with an elaborate scheme using fake email accounts and identities so he could review his own submitted papers.
Carlisle isn't the only person using statistics to identify this sort of behavior. Uri Simonsohn, a University of Pennsylvania professor, has done similar analyses to show that a handful of different high-profile social sciences and psychology papers were fraudulent.
On the whole, it's very hard to say how big of a problem fraud is in science. Vox's Susannah Locke notes that about 0.04 percent of the 1.4 million scientific papers published annually get retracted, with two-thirds of these retractions due to outright misconduct.
But the amount of intentional fraud is dwarfed by subtler, more systematic problems in science. Julia Belluz and Steven Hoffman recently detailed this huge range of issues, such as bias, poor study designs, and unfounded analyses. If anything, these sorts of problems can be even harder for peer reviews to catch — so while fraudsters like Fujii make headlines, these deeper issues do more to contaminate the scientific literature.