In 2017, professor Andrea Riebler published the paper “Projecting the future burden of cancer: Bayesian age–period–cohort analysis with integrated nested Laplace approximations” together with professor Leonhard Held at the University of Zurich.
Riebler had developed software for calculating age-specific forecasts of disease incidence or mortality based on historical data. The source code was freely available, and she had documented how to use it.
In the years that followed, the paper was cited frequently enough to demonstrate that it had made an impact among other researchers. But then something happened. The number of citations suddenly soared—from 70 in 2024 to more than 250 the following year.
Researchers are usually delighted when their work attracts large numbers of citations, but something didn't seem right. The new papers followed a pattern. Almost all came from Chinese researchers, and most were based on data from the international Global Burden of Disease (GBD) project.
The subjects ranged from global, national and regional stroke risk among adults over 20 to testicular cancer in young adults, falls among older people in China, and colorectal cancer associated with very low wholegrain consumption, among others.
“Many of the papers had the same structure. Their titles were similar, with variations relating to different diseases," says Riebler. “Some of the graphics were based on the same model we used, and their design was very similar to the one I used in my software package. That's unlikely to be a coincidence.”
Riebler and colleagues have now published a paper titled “Paper mill or paper mine? A tentative answer to the sharp increase in research papers based on the Global Burden of Disease database” in the Journal of Clinical Epidemiology.
“In one of the more than 700 GBD papers we examined, we found a reference to a commercial software package from China. The package includes instructions for doing exactly what has been done in a great many of the other papers,” said Riebler.
In the Guangzhou region of China, a company is selling software, AI-powered writing assistance, and instructions that appear to be based on Riebler's research. The company advertises that its tools make it possible to write research funding applications in just a few minutes, or produce scientific research suitable for publication in under two hours. The authors also identified two other software packages serving similar purposes, sold by Chinese companies.
“This opens up a new way of mass-producing ‘research’,” said Riebler. “We're not saying that the results are wrong. The question is whether each individual paper is scientifically relevant and useful, particularly given the enormous volume of publications we're now seeing.”
For 99 percent of the papers Riebler and her colleagues examined, the source code was not provided. This means the research cannot be reproduced, violating a fundamental scientific principle intended to ensure accountability, transparency and quality.
The researchers are calling for better monitoring of online platforms that offer ready-made research tools.
“Sharing code should be mandatory,” said Riebler.
Data from Norwegian University of Science and Technology