How the ‘Funding Effect’ Leads to Bias in Science

Is funding bias stopping us from having access to more effective medication?

By Michela Reinink

Although the phrase “alternative facts” has been widely ridiculed, the reality is that scientific facts can be and are manipulated for monetary profit.

This is seen in “the funding effect,” a phenomenon which describes the correlation between the results of a scientific study and the desired outcome by the study’s funders. Previous reviews of biomedical journals have shown that industry sponsorship (as opposed to government-funded) was strongly associated with pro-industry conclusions. 

The funding effect can result from outright misconduct, predetermined conclusions, and publication bias. This is not to say that all science being performed is falsified, or that the scientists performing the studies were morally compromised. Rather, is it more likely that in many cases, executives and hired scientists are too closely linked to the products, and convincing themselves that any evidence of harm is trivial or misleading. 

​For instance, pharmaceutical companies often desirable results not by interfering with data, which would likely be found out during peer review, but by asking the “right” questions.

In drug development, there are many ways experimental questions can be posed in a way that can be used to a company’s advantage. For example; one may conduct a trial of their drug against too high a dose of a competitor’s drug, making your drug seem less toxic, or conduct trials with too small a sample size to be able to detect a difference from competing drugs. Or there may be several subsets of trials performed, but only favourable ones used for publication.

​In one instance of twisting data for monetary gain, the company Merck was defending its new pain reliever “Vioxx” against allegations that it caused an increased risk of heart attacks. The obvious choice drug to be used in experiments as a control to compare the effects of Vioxx would have been aspirin, but as aspirin is known to decrease risk of cardiovascular disease, naproxen (brand name Aleve) was selected instead. The results showed that Vioxx increased heart attack risk by 400%.

However, the results were initially presented to the public stating that naproxen was reducing risk of heart attack by 80%, not that Vioxx caused an increase risk. In the end, FDA-mandated double-blind studies with a placebo showed that Vioxx clearly causes increased risk of heart attack, but not before the drug was on the market for 5 years (1999-2004), causing an estimated 88,000-139,000 heart attacks, 30-40% of which were fatal. 

Documents were released showing Merck executives downplaying the risk of Vioxx, however given the individuals closeness to the situation it is difficult to believe that anyone employed at Merck genuinely thought that naproxen was reducing heart attack risk by 80%. And thus, one may assume that perhaps the funding effect resulted in impaired judgement so extreme that this dangerous drug was allowed to reach the market. 

This phenomenon merits further attention, as industry-funded research often goes unpublished if the results are not favourable to the funder, resulting in what is referred to as called publication bias. In these circumstances, the researchers who performed the research are often not permitted to publish the research without the consent of the sponsor, and so unfavourable research can be hidden.

To tackle this issue, it would be beneficial for journals to only publish works done under contracts wherein the investigators have the right to publish independently if the sponsor decides not to move forward with publication. If there is no such contract, that also should have to be disclosed. Additionally, federal agencies must ensure that data is independently verified.

In these ways, the funding effect on the pharmaceutical industry can be lessened, which will surely lead to safer and more effective drugs being developed.

Is funding bias stopping us from having access to more effective medication?