Industry is increasingly the source of funding for biomedical research world-wide, and thus is also increasingly responsible for the generation of biomedical information. In the U.S., industry's share of financial support and development reached 46% in 1990, while the National Institutes of Health share dropped to 32% [1]. Similar shifts in funding of biomedical research are also occurring in other countries.

It is now very common for researchers of a particular product or process to have financial or fiduciary ties with companies that have a financial interest that product or process. These ties can vary widely, including research funding, paid consulting, stockholding, advisory board membership, patent holding, or a combination of such relationships. These relationships create a conflict of interest. The researcher's duty to the integrity and dissemination of their research is in conflict with potential financial gain, either to the researcher personally, or to the lab. The researcher's relationship with the manufacturer may unduly influence the research or its publication, consciously or unconsciously. A big question for biomedical research policy is whether the increase in industry funding of research is affecting the quality and publication of research, and whether the effect is deleterious.

A study of research on calcium-channel blockers, recently published by Stelfox et al. in the New England Journal of Medicine [2] partially addresses this question. Calcium-channel blockers are widely prescribed to treat hypertension, and a number of studies suggesting they might actually increase risk of a heart attack [3] caused a stir in the medical community. Stelfox et al. found that authors whose studies supported the use of calcium-channel blockers were significantly more likely to have had a financial relationship with a manufacturer of this class of drug than authors whose studies were neutral or critical of calcium-channel blockers. Financial relationships were defined as support to attend a symposium, an honorarium to speak at a symposium, support to organize an educational program, support to perform research, and employment or consultation. Stelfox et al. also found that supportive authors were more likely than neutral or critical authors to have a financial relationship with a manufacturer of a competing product (such as beta-blockers, angiotensin-converting-enzyme inhibitors, diuretics, and nitrates), or with any drug manufacturer.

Furthermore, only 2 of 70 articles on calcium-channel blockers examined in this study disclosed the financial ties to manufacturers. The NEJM study concluded that guidelines for disclosing conflicts of interest are insufficiently developed and enforced. Clearly, journals can do better in collecting and disseminating authors' conflicts of interest. But focusing only on disclosure as a mechanism for managing conflicts of interest has ethical and scientific risks.

Several researchers have found that bias toward "positive" results exists even among articles where financial ties to industry were disclosed. Three studies showed pharmaceutical industry-supported drug studies are significantly more likely to report "positive" findings (i.e. the manufacturer-associated drug or the new, experimental drug was superior to the comparison drug or placebo) than non-industry funded studies [4] [5] [6]. Interestingly, Bero et al. found that tobacco industry-funded studies on the effects of secondhand smoke are more likely to be "negative" (i.e. showing that secondhand smoke has no health effects) than non-industry funded studies. [7] This suggests that industry funding is not associated with "positive" results per se, but with results favorable to the industry.

The tendency for "publication bias", where industry-favorable studies are more likely to get published than unfavorable ones, seems to be exacerbated by researchers' industry ties. The skewing of the biomedical literature as a whole has important implications for clinicians and researchers, especially those who conduct or depend on review studies. For example, systematic reviews of clinical treatments that are based on literature searches will tend to show that interventions are more often effective than they are in reality. Such systematic reviews are often the basis for clinical guidelines or research proposals. The Stelfox study indicates that it is not only ties with the manufacturer of the product under study that are associated with the publication bias effect, but ties with even a competitor company or any drug manufacturer. How does this happen?

There are several possible mechanisms to explain how industry ties may lead to publication bias. The most sinister is scientific fraud. That is, industry-funded researchers could be more likely to actually change their results to make them favorable to the industry. There is not evidence to support that fraud is more prevalent among industry-funded research than other research. However, there is evidence for other mechanisms that could exert influence on researchers. One is suppression of publication, either self-imposed or externally-imposed. Another is self-selection or industry selection of products or researchers who are more likely to get positive results.

Authors do not have much incentive to publish negative results, because they perceive journals to be uninterested in such studies. However, a study by Dickersin et al. found the major reason for not publishing negative results was not because articles were rejected, but because they were not submitted in the first place [8]. Thus, we have to examine why the articles are not being submitted, and why they are less likely to be submitted by authors with industry ties.

One explanation could be that authors with ties to manufacturers fear publishing a result unfavorable to a manufacturer could jeopardize future research funding, honoraria, or other financial benefit, either from that manufacturer or any manufacturer. Another explanation is that pharmaceutical companies actively try to suppress publication of negative results. An example of this was a widely-publicized incident over the thyroid drug Synthroid, in which the manufacturer of the drug temporarily succeeded in blocking publication of a peer-reviewed study even though it was accepted by the Journal of the American Medical Association. [9] The study, funded by the manufacturer but conducted by a university researcher, found that the name-brand drug Synthroid was bioequivalent to much less expensive generic brands. This finding was unexpected, given the researcher's earlier studies of Synthroid and generics, and threatened the primacy of Synthroid's market position.

Incidents like the Synthroid case are probably rare at academic institutions because of institutional academic freedom policies preventing sponsors from controlling content or dissemination of research information. However, the majority of drug research now takes place in settings such as contract research organizations, which, unlike traditional academic settings, do not necessarily have obligations to publish their findings. Thus, the proportion of unpublished negative results could be increasing.

Another element of the Synthroid case also illustrates how interpretations of data might differ, and these differences could be linked to differing allegiances or financial ties. While the manufacturer of Synthroid was blocking publication of the university researcher's paper in JAMA, employees of the manufacturer took the data and published it in another journal without the university researcher as co-author. This paper concluded that Synthroid was not bioequivalent to the generic drugs (Rennie, 1997). If we assume that no outright fraud occurred, this should tell us that we cannot take the naive position that "the data speak for themselves", and that data interpretation is potentially influenced by who researchers work for and how they are paid.

Researchers who tend to get positive results or products that actually show evidence of greater safety or effectiveness could be selected by manufacturers to receive research funding. In the Synthroid case, for example, the university researcher had gotten preliminary results that Synthroid might be superior to generics. Such results might lead a company to sponsor the researcher, or lead the researcher to seek sponsorship from the manufacturer.

Conversely, researchers who get negative results might be less likely to pursue manufacturer sponsorship, and these researchers or the products they studied might look less attractive to manufacturers. Furthermore, researchers who get results unfavorable to a manufacturer might actually be discouraged from pursuing these results. For example, pharmaceutical manufacturers allegedly tried to exert pressure on one of the researchers who initially found an association between calcium-channel blockers and myocardial infarction through pressure exerted on the researcher's dean, and allegedly tried to interfere with publication of the study (Deyo, et al. 1997). Similar pressure has also been exerted on researchers who found results unfavorable to the tobacco industry, via lawsuits, subpoenas, and threatening letters to university officials [10]. Thus, sometimes there also appear to be (unjustified) negative consequences to researchers who find results unfavorable to industry.

Publication bias is not the only potential effect of industry sponsorship on biomedical research. Other subtle, but damaging effects include secrecy among researchers and delays of publication. Blumenthal et al. found that university life science researchers engaged in commercialization of research (defined by having research that resulted in patent applications, patents, licenses, trade secrets, product under regulatory review or marketing, or a start-up company) were three times more likely to have delayed publication of research for six months or more, or two and a half times more likely to have refused to share information with other university scientists [11].

The NEJM study and others [12] showed that a significant number of conflicts of interest are not disclosed. Perhaps this information is provided to journal editors, but is not passed on to readers, or perhaps this information is incompletely collected in the first place. However, the studies by Davidson, Rochon, and Cho, which only examined articles that had disclosed funding, found a strong association between result and funding source. This suggests that disclosure alone will not eliminate publication bias. Do researchers have any options besides enforcing and publishing disclosure?

(1) Dilute the conflicting relationship - get funding from multiple sources (e.g. more than one manufacturer). However, the Stelfox study indicated that ties with any drug company made positive results more likely, suggesting this tack will not make the problem go away. Alternatively, researchers could get funding for any particular project from both industry and non-industry sources, rather than from industry sources alone. However, Bero et al. found that drug studies with such a mixture of sponsors were still significantly less likely to meet high standards of publication quality than studies with no drug company sponsors. [13]

(2) Manage the conflict - be sure researchers can and do publish studies regardless of results. Another step in this direction is to "register" studies in existing or new clinical trials registries or other databases upon inception of the studies, regardless of their results. This way, others could potentially determine study results even though the results may not be indexed in literature databases such as MEDLINE.

Other management strategies have included oversight of the research by a researcher without ties to the sponsor. However, this can create its own conflicts of interest if the overseer is in a vulnerable position with respect to the principal investigator (e.g. the PI is department chair and the overseer is a junior member of the department), or if the overseer is a competitor. A meaningful level of oversight probably takes more time than anyone is willing to give, and may foster an unhealthy "policing" relationship among scientists. On the other hand, openness and sharing data, design, and research methods may encourage healthy skepticism and scientific exchange. One major drawback of the oversight strategy as currently implemented is that it ignores the fact bias can occur well before the execution of a research project, in the choice of study questions and basic research design.

(3) Divesting from the conflicting relationship - don't do the research or get funding elsewhere. Not doing the research is probably an unacceptable option for most researchers, nor does this do the public much good. Obtaining research funding solely from government or foundation sources are probably not realistic options in most cases either. However, researchers can eschew additional financial ties that are not absolutely necessary to the pursuit of the research, such as accepting advisory board memberships, stock options, or consulting fees from companies sponsoring research. For example, although it may be necessary and desirable for a pharmaceutical company to have particular scientists as advisory board members because of their specific expertise, it is not necessary for the research, nor desirable in terms of research integrity, for those scientists to simultaneously receive research sponsorship from the company, or hold equity in the company.

In the end, disclosure alone is unsatisfactory because we're still left with the result that industry-funded studies are more likely to have positive results than non-industry-funded studies. Certainly other influences such as academic competition or the desire for fame drive publication bias, but the studies cited here suggest financial conflicts add to the bias.

Although it is difficult to think of solutions to this problem, there is a strong ethical argument based on professional duty for trying to think of creative ways to go beyond disclosure. That is, public trust in researchers could become undermined when financial conflicts of interest become the norm. Finally, there is also a scientific argument for minimizing or avoiding conflicts of interest, because introducing additional potential for bias through financial interests does not produce the best science [14].

IUPUI School Of Science: Policy On Conflicts Of Interest - while particular in some cases to Indiana University, this document provides information to researchers to help identify and avoid conflicts of interest.

Science Under Pressure: Dow-Funded Studies Say "No Problem" - from the Center for Media and Democracy, details how industry funding has affected the silicone breast implant controversy and other medical issues.

Thyroid Storm - Dr. Drummond Rennie's editorial on the Synthroid controversy. From JAMA, 16 April 1997.

The Ties That Bind - a Scientific American Explorations article that summarizes Dr. Sheldon Krimsky's study of the financial interests of scientific authors.

Science and Engineering Ethics - originally published Dr. Krimsky's article. Tables of contents and abstracts are available online.