A Network Science Approach to Conflicts of Interest: Metrics, Policies, and Communication Design Project Description 1. Overview Research on the effects of conflicts of interest (COI) in biomedical science have long- established, significant causes for concern. For example, industry-funded trials are significantly more likely to return results that support patient use of tested drugs. As early as 2003, a JAMA study of 370 randomized controlled trials found that industry-sponsored studies were 5.3 times more likely to return results favorable to industry.[1] Subsequent research has corroborated these sponsorship effects.[2] Furthermore, funding relationships come in the form of advisory board positions, speaking fees, consulting fees, free lunches, prescription pads, and ink pens, and these smaller, financial relationships also have significant, undesirable effects on biomedical research, including an 8.4 factor increase in the likelihood of results favorable to industry.[3] The primary purpose of this project is to develop new metrics and mechanisms for the evaluation and communication of COI risks in the biomedical research enterprise. The research will use a network science approach to understand the circulation and accumulation of COI in biomedical decision systems. Network science offers an ideal framework for assessing influence within the biomedical research enterprise. The approach also moves beyond understandings of COI as about individual researchers or funders to encompass systemic risk and network-level effects. To advance the science of COI mitigation, this research will improve and enhance a machine-learning-based system that identifies and classifies COI across published disclosure statements. We will apply this system to COI disclosure statements representing the current state of biomedical research for 240 of the most commonly used prescription drug products and distill COI networks for each drug. The team will test candidate COI and COI-network metrics against drug safety data drawn from the Food and Drug Administration's Adverse Events Reporting System. The results will inform the development of new recommendations for COI policies in biomedical research and recommendations for evidence-based disclosure practices. 2. Intellectual Merit 2.1 Significance Given the growing recognition of the problems of COI, many federal agencies, universities, professional medical associations, and biomedical journals have adopted policies to mitigate the potentially harmful effects of these relationships. However, despite the consensus that policy interventions are a necessary and appropriate response, policies are inconsistent.[4] In some cases, institutions have outright bans on any financial relationships with industry. In other cases, institutions have complex requirements for vetting each individual COI. The FDA, for example, assesses each potential advisory committee member COI for ?directness? and ?predictability.? So, if a potential advisory committee member owns stock in a company that is seeking a new drug approval, regulators can identify a direct and predictable COI (stock values may rise based on the regulatory decision). The most common COI intervention across institutions is some form 12
