Perhaps every industry-funded study should include a Surgeon General-style warning: This research may contain funding bias hazardous to your decision-making, health and welfare.
When companies pay for research on their products and services, there is a strong incentive to rig the results in their favor, say critics. Variously called “sponsorship bias,” “funding effect,” or “funding bias,” the favoritism can be baked deep into the design or methodology.
“It’s not something you can test very easily,” Sheldon Krimsky, a Tufts University professor who authored a seminal paper on funding bias, told Tarbell. In his paper, he notes that a comprehensive, independent review of literature on tobacco indicates a funding effect. That’s because non-industry research was more likely over time to pinpoint harmful consequences.
The only way to know for sure if bias is at play is to thoroughly interview the authors to gauge whether there was prejudice in the methodology or data interpretation, Krimsky said. And even then, “it’s not easy to pick out actual bias unless it’s so obvious you cannot deny it.”
Since extensive debriefs of researchers are usually not feasible, he recommends looking for clues that might indicate bias. The structure of an experiment, exclusion of certain data, and interpretation of results are key areas to evaluate. “You can [introduce] bias by doing the testing in a way that is least likely going to get you an adverse effect,” he explained.
[embed_content url=”//“]
But it’s not just discernable bias that worries Krimsky. “I’ve argued that you’re either going to get the appearance of bias or you’re going to get actual bias,” he said. “And both of those conditions are not good.”
Genna Reed, lead science and policy analyst at the Union of Concerned Scientists, cautions that just because a study is industry-funded doesn’t automatically mean the conclusions reflect the sponsor’s interests. But she does see “at least the potential” for bias in any industry-led research. “There is this inherent bias that does exist because there is that direct financial link.”
Another way to measure for bias is to compare the results of studies conducted by for-profit companies with those prepared by non-profits, such as government institutions. Krimsky used the example of research on risks associated with a new pharmaceutical.
If the for-profit studies are more likely to interpret the risks as minimal or insignificant while the non-profit studies find greater risk, this could hint at funding prejudice, he explained.
For example, independent analysis of the food sweetener sucralose, sold as Splenda, has raised cancer alarms not identified by industry-backed studies.
How can such bias be reduced or eliminated? Reed thinks more transparency would help. She’d like to see every academic journal and database require disclosure about researchers’ financial ties to funders and whether authors provided expert testimony on behalf of sponsors, she said. The federal database PubMed, which publishes scholarly papers and abstracts, already requires this information in conflict-of-interest statements.
Reed urges journals to go further by requiring companies to reveal any involvement in editing the manuscript.
Dr. Jerome Kassirer, former editor of the New England Journal of Medicine, is not convinced. “Disclosure is a not a solution to the conflict,” he said. “It only gives you a hint that there could be a bias.”
[callout type=”newsletter-occasional” title=”Don’t miss other Tarbell investigations”]
Another idea that has traction among academics involves the creation of an impenetrable firewall between the funder and scientists testing a product. “That would be the greatest assurance that there would not be” actual bias, or even the appearance of it, Krimsky said.
For example, drug companies could rely on a non-profit, national institute of drug testing to select researchers who would assess whether a medication should win FDA approval. Companies could still conduct research, but they would not be the exclusive evaluators of their products.
At the moment, no such institute exists.
In an email exchange with Tarbell, Marion Nestle, the nation’s foremost specialist on food industry studies, also suggested a firewall as the best remedy. But she was realistic about its prospects, stating, “Companies don’t want to fund research on those terms.” Nestle is publisher of the Food Politics blog and author of Food Politics: How the Food Industry Influences Nutrition and Health.
Short of a firewall, Krimsky recommends rock-solid contracts that bar corporate sponsors from influencing methodology, data interpretation, publication schedule, or any other aspect of research.
“In other words, if they just funded the study, and left the investigators completely on their own to do the work,” he said. “You still might have an appearance, but you may not have any actual biasing results.”
