Duke University is committed to advancing translational research and innovation across our research enterprise. Projects and initiatives span a broad range of scholarly activities from biomedical and health focused programs – such as MEDx, Precision Genomics, the Center for Health Informatics, and the Regeneration Center - to innovative solutions to advance society and public policy like Duke AI Health the Nicholas Institute for Environmental Policy Solutions, Center for Politics (POLIS), Wilson Center for Science and Justice, and Duke Arts, Law and Markets Initiative (DALMI) etc.
Translational research and innovation come with risks as well as ethical and regulatory challenges that must be managed with care and accountability.
How can we reduce risk and maximize benefit while implementing cutting edge interventions? How do we minimize potential harm in translational research? Can we ethically adopt innovative solutions, technologies, and policies when their impact or benefit have not yet been demonstrated? Duke researchers and experts joined us on September 28, 2022, to discuss these important questions in the Duke Office of Scientific Integrity’s monthly research town hall series.
The event, attended by almost 100 people, was moderated by Dr. Geeta Swamy, Associate VP for Research, Vice Dean for Scientific Integrity, and Distinguished Professor of Women’s Health. In her welcoming remarks, Dr. Swamy welcomed the audience and acknowledged the outstanding speakers, representing many disciplines and departments across Duke.
Kathy Nightingale: How to improve diagnostic imaging without increasing safety risks
The field of Engineering and AI was represented at the town hall by Kathy Nightingale, “Theo Pilkington” Distinguished Professor of Biomedical Engineering, and Boyuan Chen, PhD, Assistant Professor in the Thomas Lord Department of Mechanical Engineering and Materials Science.
Professor Nightingale has developed an innovative elasticity imaging technology. “We inform clinicians of the stiffness of the tissue because how hard or stiff tissue is tends to be correlated with regions of pathology. As we focus sound to push on the tissue, that creates a shear wave that propagates in the tissue; the speed of the wave is related to the underlying stiffness of the tissue. So, now, before getting a liver biopsy to assess hepatic fibrosis, we can perform an imaging test and, if the wave speed is very low, there is no fibrosis and thus no need for biopsy,” Prof. Nightingale explained. Her laboratory is using the same technology to provide imaging guidance for targeting prostate cancer and also hopes to develop elasticity-based biomarkers to assess the health of muscle and to characterize the progression of muscle disease.
Prof. Nightingale described the ethical challenges of her work in the context of the innovation pathway, from idea to the market. The idea comes first, then the research team works on sequencing and algorithms. Next, the technology is assessed in phantoms and/or preclinical studies. If all goes well, the demonstrations of the technology are performed in the hospital, on patients. Finally, if the technology is proven to be effective, it moves into the commercialization phase. Although this technology is non-invasive and does not involve significant risk, there is potential for harm which needs to be managed. For example, her team realized that using elevated acoustic output beyond the FDA approved guidelines for diagnostic testing would be useful for better assessing the shear wave speed. “We worked with clinicians and scientists to determine if it would be reasonably safe to increase our output in some of our studies. We found that by increasing the Mechanical Index beyond what the suggested FDA guidelines allow – which was based on historical measures rather than reported bioeffects – would lead to considerable diagnostic benefit without creating any damage. We established our own threshold (below reported bioeffects) and then we performed a risk-benefit analysis to see whether we would be comfortable asking patients if we could make these measurements on them.” she continued. Similarly, in some prostate clinical studies, patients are asked whether they agree to have additional biopsy cores taken, which could increase the risk for infection, while also potentially providing diagnostic benefits. “My first bar when I'm designing a study, is whether I would participate myself in such a study. The next question is would I be willing to ask a family member to participate in the study? If my answer is no, I wouldn't go ahead with a study.” Professor Nightingale added.
Boyuan Chen: Ethical challenges in robotics have to do with securities and mutations of the technology
Boyuan Chen’s area of research is robotics. He is studying robots and how to build their adaptability to the environment. “We want to pursue a path of world modeling which can be summarized into this very simple formula. If you have a given context and a plan for the robot, can the robot predict the consequences of an action? If so, then the robot can have a basic understanding of the world.” - Prof. Chen described. He is also studying how quickly robots learn to adapt to a change in the environment – for example if a motor on the robot’s body breaks – and, furthermore, how robots can be designed to compensate – as humans do - when other robots don’t work well. He is also exploring how humans and machines can team up together. There are many ethical challenges that need to be overcome in robotics, many of them related to “the securities and the mutations of the technology”, Dr. Chen stated. How the data is collected, saved, preserved and analyzed are other key considerations for designing robots that are safe and easy to work with. While humans can anticipate other people's intentions or perspectives, robots don't really have this capability when they perform actions. “For example, if you use a robot cleaner on your floor, oftentimes we will have to save the robot when it is stuck in the corner, or we have to get out of the robot's way. We need to be careful how we design objectives for robots to be successful. Sometimes we may design objectives that will conflict with each other – and we need to avoid that”, Chen added. While today, AI scientists have to navigate how to advance the field with maximum caution, Dr. Chen is certain that there will soon come a day when robots will be a commodity, “just like internet or electricity that you don’t even feel it and that is everywhere”.
David Harpole: It takes an entire team to manage the risk for misconduct in cancer research
David Harpole, MD, “George Barth Geller” Distinguished Professor for Research in Cancer, Professor of Surgery and Professor in Pathology, opened the Medicine and Global Health section of the town hall. He emphasized the importance of building a trustworthy research team when conducting translational research successfully. This is particularly true in cancer research – he said – because it is multidisciplinary: computational experts, proteomic experts, surgeons and data analysts utilize each other’s expertise to advance the science. “Cancer research is a team sport”, he said. Many ethical challenges are related to biomarkers, Dr. Harpole continued. “Even biomarkers that are commercially available were not correctly validated,” Dr. Harpole said. Because omics platforms use terabytes of data and generate thousands of data points, there is also potential for tens or hundreds of errors. Many times errors happen because of rapid technology advancement. “They have developed so rapidly that there's no validated or standardized analytical method on how you process and how you handle the data.” he explained. And, given that computer expertise is “outside of bounds”, the risk for misconduct can be high. “It’s like not being able to understand a black box about how the analysis is done”. A great resource that has been created to navigate this risk is the Early Detection Research Network, a program of the National Cancer Institute that validates collaborative discovery and validates biomarkers for cancer risk and early detection.
To ensure the integrity of data and minimize the potential for errors, Dr. Harpole emphasized the need to share the data on an open platform with the entire laboratory, and to use an independent statistician for data analysis. A third step is to create a multi-institutional data set where other centers have followed your SOPs, generated the bio-specimens, and analyzed them using your methodology. Then they come up with a prediction and use an independent statistical team for analysis. “It is very rigorous data isolation to eliminate potential biases and potential misconduct out of the realm of predictors,” he concluded.
Jacob Schroeder: the huge men-women discrepancy with respect to heart assist devices
Jacob Schroeder, MD, Assistant Professor of Surgery, Duke Heart Transplant Center, discussed the ethics of innovation in cardio-thoracic surgery. According to the CDC, there are 6.2 million Americans who have heart failure, and would qualify for heart failure therapy, such as a left ventricular assist device (LVAD) or heart transplantation. Unfortunately, heart transplants are limited by the number of donors, given that the only FDA approved artificial heart, a 1980 technology, has a high rate of adverse events. Dr. Schroeder was part of the Duke team of heart surgeons, led by Dr. Carmelo Milano, who, in 2021, performed the first artificial heart implant ever in the US, using the Carmat total artificial heart, a new product developed and tested in Europe with promising results. “One of the most gratifying parts of my career was sitting at the press conference that we had when we put this device in and listening to the patient's wife talking about this life changing experience for her husband and the entire family,” Schroeder shared. One of the biggest ethical dilemmas about artificial heart implantation is related to resource allocation, given how high the costs are – up to $ 100,000 per patient, life expectancy and durability. A detailed analysis of these challenges, by Dr. Carmelo Milano, can be read in an Integrity Interview on DOSI’s website. Dr. Schroeder laid out a new ethical issue: the size of the devices, which are designed based on an average patient, mostly based on an average male patient, but they can be too big for an average female person. “There is a huge discrepancy between men and women which goes across pretty much all of adult cardiac surgery. This is a problem on which we need to focus significantly more energy,” Schroeder continued.
Garry Bennett: Digital health innovation has great impact on reaching marginalized communities
In the context of global health research, innovation also comes with challenges, especially in the new and rapidly developing realm of digital health. Gary Bennett, PhD, “Bishop-MacDermott Family” Professor of Psychology & Neuroscience, Professor of Global Health and Medicine, Director of the Duke Global Digital Health Science Center and Vice Provost for Undergraduate Education shared his work on obesity in underserved populations. He emphasized the complexity of the “do not harm” ethical principle in the digital health space where a lot of discussion has focused on data breaches and privacy concerns given that data collected about someone’s health can be used not only to make predictions about medication and adherence across a wide range of conditions, but also about mental health outcomes.
Bennett also mentioned that there are communication challenges in the context of informed consent: It is hard to provide information about how data is being used in language that is accessible for patients. “Another big challenge is that the vast majority of accumulated evidence in digital health is related to obesity, physical activity, smoking, cessation, alcohol, mental health or metabolic conditions for which self-management is indicated. Almost all of those approaches work better when there's a human involved in the process. This leads to the fear that someone would grab the app and modify their health outcomes.” Bennett said. But managing these challenges is important, because of the high impact of digital health particularly in reaching and engaging vulnerable communities. “Most people in many of the vulnerable communities have a mobile. Using digital health care can be very effective particularly in reaching the disadvantaged and marginalized groups,” Bennett ended.
Brandon Garrett: From policy engagement to cutting edge research to inform criminal justice reforms
Brandon L. Garrett, JD, the inaugural L. Neil Williams, Jr. Professor of Law and Director of the Wilson Center for Science and Justice is conducting cutting edge law & policy research. Much of his work focused on the criminal justice system. For example, he examined wrongful convictions – information which is publicly available – to analyze what are the causes of the errors and how can they be reduced.
“We ask questions such as why do witnesses sometimes make mistakes. Why do false confessions happen? How do jurors appreciate scientific evidence like forensic evidence, fingerprint evidence? We would like to ask jurors what they found powerful about the evidence. So there is a large body of experimental work that we perform at the Wilson Center in collaboration with many experts, including psychologists and psychiatrists,” Garrett said.
Another wide area of his research is focused on studying the inequities in the criminal system, including mass incarceration and harsh sentencing. He is also exploring how behavioral health can be better integrated with the criminal justice system, so that individuals with mental illness can be directed to adequate treatment rather than to jail.
The Wilson Center is also filing amicus briefs on behalf of experts to educate the courts on cases that involve specialized issues. “For example, we just filed a brief for educating an Illinois court about the research around the validity of fingerprint evidence and what the court needs to understand in terms of the limitations of that evidence. There actually is research on how accurate those kinds of pattern recognition calls are, and it's misleading if the jurors are just told “I am an expert and this is what I think”,” Garrett explained. While filing Amicus Briefs is policy engagement and not scholarly research, the impact of this work is notable and it is very helpful to judges who often don't have the opportunity to hear from the scientific community.
ABCDS Oversight: a Duke program focused on the governance, evaluation and monitoring of algorithms deployed at Duke Health
Nicoleta Economou, PhD, Program Director for the Algorithm-Based Clinical Decision Support (ABCDS) Oversight Committee, presented the importance of this organization for patient safety and high quality care. Her talk introduced Duke Health’s ABCDS Oversight program, which ensures that algorithmic healthcare applications deployed at Duke Health, including those that use machine learning or artificial intelligence, are safe, accurate, fair and equitable.
Recent growth in the use of algorithmic tools in healthcare, where sophisticated models are applied to enormous volumes of patient data available in electronic health records and other sources of patient data. Economou discussed that although algorithmic tools have enormous potential for improving patient outcomes, care delivery, and operational efficiency, there have already been examples of such tools failing to perform properly in “real-world” environments.
The ABCDS Oversight Committee provides a framework for continuous oversight and evaluation of algorithmic healthcare applications at Duke Health, with the goal of ensuring that such tools are fit for purpose throughout their entire lifecycles and do not introduce error or bias into patient care.
The ABCDS Oversight Committee, which includes expertise drawn from technical and clinical specialties, looks at clinical outcome and performance metrics, how the method is evaluated, who the owner is, and who covers the costs. The Committee also determines how technologies are integrated with clinical workflows and whether it is part of the standard of care. All algorithms that could impact patient care at Duke Health are overseen by the ABCDS Oversight Committee and must undergo registration, Dr. Economou added.