"Unreliable
By Csaba Szabo
Columbia, 328 pages, $30
The largest repositories of biomedical research in the U.S. and Europe, PubMed and Europe PMC, contain 84 million articles between them, and add a million more each year.
According to recent estimates, up to 90% of those papers -- 75 million total -- contain information that's either misleading, wrong or completely fabricated.
Over the past 20 years, certain branches of science have endured a so-called reproducibility crisis, in which countless papers have been exposed as shoddy if not bogus. Sometimes these revelations are merely embarrassing, but in biomedical research, incorrect publications can cost lives as doctors and drugmakers rely on them to treat patients.
In "Unreliable: Bias, Fraud, and the Reproducibility Crisis in Biomedical Research," Csaba Szabo -- a physician with doctorates in physiology and pharmacology -- dissects the ways he's seen research go wrong in his 30 years in academia and industry: data manipulation, poor experimental design, statistical errors and more. It's a sobering, often depressing, litany.
The most titillating accounts concern fraud, where scientists either invent data or purchase bylines in ready-made articles for as little as $500. Dr. Szabo slams these efforts as "brazen insolence and serious criminal activity" which make "an utter mockery" of the scientific search for truth.
Nevertheless, such activities are thriving: One scientific sleuth estimated that so-called paper mills churned out a third of neuroscience papers in 2020 and a quarter of medical papers.
The biggest problem, however, lies with scientists who strive to do good work but feel pressured to cut corners.
Scientists cannot work without grant money, but of the 70,000 applications the National Institutes of Health receive each year, only 20% get funded.
Leading journals reject up to 99% of papers submitted, and only one in 200 doctoral graduates ever becomes a full professor.
Even with tenure, professors can suffer salary cuts or have their labs handed to higher-performing colleagues if they don't keep pulling in cash.
Some sadistic research professors even pit their graduate students against each other in "dogfights" -- they run the same experiment, but only the first to get results publishes.
No wonder researchers massage data or fudge images: Forget "publish or perish." It's "fib or forgo your career."
And that doesn't address mistakes: mislabeled solutions, impure lab water, results that turn on what time of day you feed your mice.
The human mind is also a bottomless source of unwitting error. We read deep meaning into random data, shoehorn anomalous results into theories we favor, and overhype the impact of the results. These biases aren't "scientific misconduct," Dr. Szabo notes, yet there's "a baseline level of subjectivity and error that contributes to the unreliability of the literature."
Given this tsunami of mistakes, the author points out that cynical types have suggested we treat all biomedical research as fraudulent unless proved otherwise.
The cost is staggering: The U.S. wastes tens of billions of dollars annually on useless research, shortening or even costing patient lives. Most scientists can't even reproduce their own data half the time, and the number of papers retracted rose to 10,000 in 2023 from 500 in 2010. As Dr. Szabo asks plaintively, "how can anyone conduct meaningful research" under such circumstances?
He has some ideas.
The first involves abandoning the notion that science can police itself.
He's scathing about for-profit journals and universities, where editors and administrators have strong incentives to look away from bad practices and even outright fraud, both to protect the reputation of their "brands" and pad their own salaries -- for every $160 brought in as research funds, we are told, $60 supports university facilities and administration.
In cases of outright fraud, Dr. Szabo suggests prison for theft of public tax dollars. Alarmingly, he also proposes installing cameras in every lab and keystroke monitors on all computers -- an invasion I cannot imagine any scientist tolerating.
Artificial intelligence could potentially flag common problems, but it's also likely to introduce new ones: In a darkly humorous appendix, Dr. Szabo reprints the results of asking ChatGPT to write an article on the medical effects of a made-up compound, "nanocapsulated zumbroid." According to ChatGPT, it's a miracle cancer cure.
Most importantly, Dr. Szabo calls for systematic changes in how science gets done.
Nowadays, scientists spend about 30% of their time writing grant proposals.
If the government handed block grants to universities instead, he suggests, that time could be freed up for replication studies -- which aren't fun but would sure beat paperwork. Scientists who refuse to participate could be made ineligible for grants and publication themselves.
Dr. Szabo also suggests producing fewer scientists, culling the field to those who show real passion for lab work.
Others could focus on teaching or become professional peer reviewers to ease the burden on lab researchers.
Given the garbage science that's already being produced, he suspects that this wouldn't slow the production of authentic knowledge.
"Unreliable" is somewhat uneven. Dr. Szabo goes on for pages about his own research and indulges in grouchy complaints about practices that, while unseemly, don't affect research reliability -- inflating citation counts, schmoozing with grant reviewers. Still, budding scientists should contemplate whether they're willing to put up with all this. (The book should be kept away from the unscrupulous, though; at times it reads like a playbook for gaming the system.)
Ultimately, Dr. Szabo advocates for slower but surer science -- a tough sell in a biomedical industry with big dollars and lives at stake. But he finds hope in the fact that people no longer ignore the problem of reproducibility. Above all, he despises the broken status quo, where "everybody acts politely . . . keeps their mouths shut, and acts like the whole process is functioning perfectly well."
---
Mr. Kean is the author, most recently, of "The Icepick Surgeon." His next book, "Dinner With King Tut," will be published in July.” [1]
1. Reaching For Results.Kean, Sam. Wall Street Journal, Eastern edition; New York, N.Y.. 25 Mar 2025: A15.
Komentarų nėra:
Rašyti komentarą