There is a scientific fraud epidemic – and we are ignoring the cure

THE dossier was so unsettling, one neurologist revealed, that he couldn’t sleep after reading it. It contained allegations that an experimental drug meant to curb damage from stroke – and eyed up for regulatory fast-tracking for fulfilling an unmet medical need – might instead have raised the risk of death among patients receiving it. 

The dossier, assembled by whistleblowers and obtained by an investigative journalist, was recently submitted to the US National Institutes of Health, which is finalising a US$30 million clinical trial into the medicine. The whistleblowers allege that the star neuroscientist driving the research, Berislav Zlokovic from the University of Southern California, pressured colleagues to alter laboratory notebooks and co-authored papers containing doctored data.

The university is investigating; Zlokovic is, according to his attorney, cooperating with the inquiry and disputes at least some of the claims.

The facts of this particular case, set out in the journal Science last week, are yet to be established but research is fast becoming a catalogue of mishaps, malfeasance and misconduct. Rooting out mistakes and manipulation should not have to depend on whistleblowers or dedicated amateurs who take personal legal risks for the greater good. Instead, science should apply some of its famed rigour to professionalising the business of fraud detection.

Zlokovic is not the only high-profile scientist to have hit the headlines for the wrong reasons. In June, Francesca Gino, a behavioural scientist at Harvard University, was accused of data irregularities by three US academics who run the Data Colada blog. Gino, on administrative leave, is now suing both Harvard and her accusers for defamation. The Data Colada trio have so far crowdsourced more than US$376,000 for a legal defence fund.

As the Oxford university psychologist Dorothy Bishop has written, we know only about the ones who get caught. In her view, our “relaxed attitude” to the scientific fraud epidemic is a “disaster-in-waiting”.

GET BT IN YOUR INBOX DAILY

Start and end each day with the latest news stories and analyses delivered straight to your inbox.

Sign UpVIEW ALL

The microbiologist Elisabeth Bik, a data sleuth who specialises in spotting suspect images, might argue the disaster is already here: her Patreon-funded work has resulted in over a thousand retractions and almost as many corrections.

That work has been mostly done in Bik’s spare time, amid hostility and threats of lawsuits. Instead of this ad hoc vigilantism, Bishop argues, there should be a proper police force, with an army of scientists specifically trained, perhaps through a masters degree, to protect research integrity.

It is a fine idea, if publishers and institutions can be persuaded to employ them (Spandidos, a biomedical publisher, has an in-house anti-fraud team). It could help to scupper the rise of the “paper mill”, an estimated US$1 billion industry in which unscrupulous researchers can buy authorship on fake papers destined for peer-reviewed journals. China plays an outsized role in this nefarious practice, set up to feed a globally competitive “publish or perish” culture that rates academics according to how often they are published and cited.

Peer reviewers, mostly unpaid, don’t always spot the scam. And as the sheer volume of science piles up – an estimated 3.7 million papers from China alone in 2021 – the chances of being rumbled dwindle. Some researchers have been caught on social media asking to opportunistically add their names to existing papers, presumably in return for cash.

Artificial intelligence (AI) is a godsend to this modern racket.

In 2021, a team of researchers tracked the rise of AI-generated “tortured phrases” in the literature, such as “counterfeit consciousness” in place of “artificial intelligence”. As language models improve, machine-generated text will no longer be obvious gibberish. When rotten results and dodgy data sets flood the literature, they become the flawed building blocks for further analyses – undermining the science yet to come.

Impropriety also ranges wider and deeper than blatant deception. It can be sloppy analysis or cherry-picked data. It could be the well-meaning conviction of an academic that, if only he can recruit the right patients and tailor his trials in the right way, he will crack the terrible problem of dementia. Factor in that medical regulators seem willing to revisit the cost-benefit trade-off in hard-to-treat conditions, and it becomes an environment in which the artful presentation of trial outcomes can potentially make or break a billion-dollar drug.

And that, really, is the problem: the lack of due diligence means the rewards for bending or breaking the scientific rules tend to outweigh the incentives to observe them. FINANCIAL TIMES