In Search of Equipoise

In 1974, Summerlin proudly displayed white mice with black patches of fur, claiming that he had successfully transplanted tissue from unrelated animals without rejection.¹ He confessed to deception when laboratory assistants discovered that the transplanted sites were blemished by alcohol, leading to use of the phrase “painting the mice” to indicate research fraud. In 2005, Korean scientist Hwang was lauded when he proclaimed that he had created the world's first stem cells from a human embryo. The praise was short-lived when his own university concluded that his work was also fraudulent.

The efforts of Wakefield and Walker-Smith to falsify data linking vaccines to childhood autism are the latest prominent misrepresentations of science as fact.² The fraud of Wakefield and Walker-Smith epitomizes how fabricated research can lead to a domino effect of tragic consequences, characterized by patients fearing lifesaving interventions, clinicians altering practice, and scientists and governments wasting precious resources to evaluate researchers' claims.

Falsified research is perhaps so profoundly corrosive because it strikes at a fundamental tenet of ethical research—equipoise. Introduced as a term by Freedman in 1987, equipoise is “a state of genuine uncertainty on the part of the clinical investigator regarding the comparative therapeutic merits of each arm in a trial.” On the moral compass of scientific research, equipoise points due north, providing a framework of ethical principles and research integrity beyond reproach.³

While elegant in theory, equipoise is elusive in reality. The very definition is ambiguous. Is equipoise a state of consciousness or conscientiousness that a researcher seeks? Is it an implementation framework? Or, perhaps, an underlying philosophy of research, a code of conduct? Notwithstanding these uncertainties, transparency and objectivity are prerequisites for equipoise. As graphically illustrated by Wakefield and others, the emotions and convictions of researchers are frequently at odds with this requirement.

How then to rationalize fraudulent (or, yet more common, biased) research proffered as medical science? In a multinational, multibillion-dollar health care industry encompassing numerous entities and vested financial interests, the stakes and complexity of modern medical research have never been greater or more confounding. Results of influential studies are regarded as the milestone of prestige and success and affect professional discourse more than ever. In this context, one hypothesis is that research equipoise is corroded by bias from 3 main sources: researchers, presentation, and publication.

To maintain equipoise, investigators must theoretically remain indifferent to the results of a trial, ensuring that the conduct—not the outcome—of a study is most valued. In practice, this is difficult to achieve. The mere act of posing a research question may violate this principle; for instance, researchers typically select questions based on the relevance or perceived importance of a problem, skepticism regarding current knowledge, or the prospect of personal or political gain. Although they ask questions because the answer seems inherently appealing, researchers often do not resolve whether having any answer vs only a specific answer is acceptable. The former is equipoise; the latter, bias.

Similarly, research methods may be manipulated to deviate from equipoise. For example, selecting specific samples (sampling or selection bias) or measuring only desirable outcomes (outcome bias) are but a few means by which researchers can influence study results. Even at a conceptual stage, a specific null hypothesis (the outcome that represents the lack of an association between the predictor and outcome variable) may be proposed to ensure that statistical testing rejects it, facilitating acceptance of preconceived beliefs (the alternative hypothesis). This caveat usually remains obscured by statistical tests of significance, leading to questions about whether a quantitative method based on the science proving itself wrong, not right, is inherently compatible with equipoise.⁴ Bayesian theory, in contrast, advances equipoise by maintaining that study interpretation is a qualitative argument embodying the investigators' logic, rationale for the experiment, biological plausibility of the findings, and context of the findings in relation to other studies. This prismatic viewpoint is used infrequently in designing as well as decoding clinical studies.

In an era of increasing competition for funding and publication, researchers face mounting pressure to report the results they wish to see.⁵ Against this backdrop, misrepresentation of study results is a veritable threat to research equipoise. In a study comparing proposed vs published outcomes of randomized clinical trials, selective reporting of results was found in up to one-third of included studies in general medical and subspecialty journals.⁶ Another study demonstrated that many authors distort results of their findings—perhaps because of simple unawareness of the facts, subconscious biases, or purposeful intent to deceive.⁷ Researchers also may tailor a study to cater to a specific audience or agenda, interpreting data to arrive at conclusions not directly supported by the evidence.⁸

These nuances in interpretation have led to the phenomenon described as “spin” in reporting research studies. Although universal conflict-of-interest reporting is one technique by which to prevent this problem, a more radical solution is to mandate the public release of all available data in connection with a study (as opposed to specific data associated with outcomes). This requirement (likely to face opposition from researchers owing to the presumed value of research data) would enable reviewers to derive their own conclusions using the totality of study evidence, freeing them from reliance on author presentation.

While engagement in presentation bias can influence reporting of studies by researchers, peer-reviewed journals paradoxically further contribute to this phenomenon. Publication bias, wherein medical journals accept and publish only those studies with statistically significant results, can be an obstacle to attaining equipoise. By selectively filtering for studies with statistically significant results, consideration and dissemination of research pertaining to findings on both sides of a research question are jeopardized. A Cochrane review confirmed that a significant delay exists in the publication of studies with statistically nonsignificant results; such clinically important “negative” studies are often not published in journals with the widest audiences.⁹ Furthermore, it is conceivable that this phenomenon also influences investigators before research begins. As researchers contemplate study designs to challenge or extend accepted scientific knowledge, fear that diffusion of their work may be compromised may lead them away from lines of inquiry in which the risk of a negative study appears high, despite the fact that balanced evidence is often most critical in these very areas.

With so many sources, it appears that bias inevitably goes hand in hand with clinical research. Can research equipoise be attained? New approaches on numerous fronts are necessary. For instance, public and private sponsors of research must shift the focus of trials from outcome to process, affirming and championing the notion that study veracity is more valuable than result. This movement is gaining momentum, evidenced by calls to tie research integrity to the funding of investigations.¹⁰ Furthermore, reviewers, editors, and publishers of medical journals must move toward a model in which robust methodology, scientific discipline, and clinical relevance are valued in addition to statistical significance. Readers play a critical role in this dynamic and must demand that all sides of a research question are equally represented.

Ultimately, however, the quest for equipoise rests squarely on the shoulders of the researchers whose work carries science forward. To identify and celebrate equipoise, researchers must be trained and unwaveringly encouraged to recognize, appreciate, and root out biases in their work. They must learn to disseminate their data freely and fiercely to support their conclusions and their tenacity. Without such efforts, the domino effect of fraud will continue. Society and science can ill afford the costs.