The Center for Bioethics at Harvard Medical School recently held an online conference covering the current state of the field. All the standard ethical issues we talk about in medicine appear in the new genetic and biological research, often with greater urgency: safety and effectiveness, equitable access, privacy and informed consent, data sharing, and individual dignity. In addition, immense philosophical questions beyond the scope of medicine are raised. I’ll look at each of these issues in the article.
Safety and Effectiveness
Time-tested processes for assuring safety and effectiveness don’t always extend to new biological techniques, because the techniques’ potential effects are so hard to predict. As one example, Jun Wu cited research showing that the foreign cells introduced into a person for treatment might be broken down by cell-cell competition, a normal metabolic process leading to the death of cells (apoptosis). Treatments that don’t account for this process may seem to work in the lab and then have no effect in real life.
And when the techniques are not amenable to clinical trials, even more risk gets injected into the protoplasm. Dr.Timothy Yu and Jacob S. Sherkow said many modern experiments involve diseases so rare that it would be either unfeasible or unethical to set some patients aside as controls. Follow-up studies can provide a lot of useful data, but comparisons between patients in retrospect are less conclusive than a standard clinical trial. Jonathan Kimmelman said that at best, N-of-1 studies are “on the path to” Phase 1 studies.
One survey of data scientists and machine learning developers found a shortfall in health care expertise at their firms. Although the developers recognized the excess of “hype” about artificial intelligence in health care, they also failed to understand the extent to which their products could produce or exacerbate harm.
Predictably, the articifial intelligence firms also reject responsibility for harm, placing that responsibility on the end-users. This is a fixture of both the medical and software industries. If an electronic health record presents an incorrect choice in its user interface, the doctor who makes that choice is the one who is considered legally liable. And if you read the terms of service on almost any software product or service, you’ll see that they back off from any liability as well.
Equitable Access
The past few years have given us a wealth of unsettling but much needed research into disparate access to health care on the basis of race, gender, and geography. Whether it’s pain relief or treatment for heart disease, white men fare much better in the health care system than others.
Revelations that pulse oximetry devices are less effective for dark-skinned people have shamed the medical device industry, an indictment that became even more stark as these devices proved critical for tracking COVID-19 risk. Tina Eliassi-Rad pointed out that the racial disparity was identified in the 1950s, but was just ignored for over 60 years.
Bias in machine learning has been widely reported, so I don’t have to summarize the issues here. As Eliassi-Rad said, possible problems arise at every point: scoping out the problem to be solved, data collection, choice of parameters, model design, testing, application, and updates. Even the definition of “fairness” is impossible to resolve, as I have reported elsewhere. Dr. Erich S. Huang says that to fix bias in health care AI, we need to fix bias in the health care system.
One positive advance can be cited: Eliassi-Rad says that machine learning algorithms now can provide both accuracy and transparency so that raising one does not lower the other.
New technologies can exacerbate access in many ways, just like any technology that is initially expensive. It’s not just the problem of paying for treatments that are experimental and uncovered by insurers. Who is well enough integrated into the health care system to get a diagnosis for a rare disease in the first place? Who has doctors eager to stand up for the patient and guide the family through uncharted clinical terrains?
The most familiar complaint about inequities in genetic progress is the “designer babies” colloquialism. Although we have no idea whether we’ll ever be able to pinpoint a genetic source of intelligence, physical strength, or other qualities, an eery fear has spread of the rich developing a super-race such as anticipated by H.G. Wells’s novel The Time Machine.
Privacy and Informed Consent
The unpredictable paths taken by technology also call into question our procedures for getting consent, which were always beset by controversy. Unless researchers throw away data and tissues after a single experiment, they are asking patients to give them the right to do things that they and the patient might regret later.
Furthermore, as J. Benjamin Hurlbut pointed out, assumptions about the possibilities of technology change over time. A procedure may have been ruled out because it was unavailable or dangerous, and then later become feasible. Already, many people can be identified from their genome, thanks to widespread databases that may contain their relatives. The genome is therefore considered to not to be anonymizable.
Religion plays an important role both in individual decisions and social policy. Many countries have banned the use of human embryos in research, because their legislators count these embryos are legitimate human beings. These legal measures generate a lot of resentment in the field. Insoo Hyun said that the definition of an embryo changes from law to law and from country to country. M. William Lensch reported a poor understanding in laws about the biological differences between chimeras (where genetic material from species is injected into another) and hybrids (where an organism such as a mule is created by breeding two species).
Data Sharing
Scientific progress has always depended on data sharing, and in recent decades expectations have gone far beyond publications and patents. Researchers are expected to share raw data and the procedures they developed for their research, as I have written about elsewhere. But to researchers, this is like asking Marvel Comics to share the Avengers movie on which they spent 356 million dollars. A lot of research time and money goes into collecting the data and developing the procedures.
Except for some hints of a shared data repository, I heard nothing about this problem at the bioethics conference.
Individual Dignity
Rarely do we force treatment on a person who doesn’t want it. Psychiatrists can confine people at risk of harming themselves and others, but even the most florridly psychotic patients can refuse treatment. The dilemma of modern medicine is to create treatments that seem miraculous, but also give patients the heebie-jeebies.
The “designer baby” issue is also a challenge to human dignity. In this speculative scenario, deafness and LGBTQ status are cultures commonly cited as in danger of being wiped out.
Matthew Sample cited a more immediate and subtle example of our dilemma. Imagine a person with a neural problem that renders him unable to walk, so he requires accommodations at work. There may exist today a brain-computer-interface (BCI) that can improve his condition. However, he may fear having something he doesn’t understand entering his brain and changing his behavior. So he may choose to stay disabled.
Sample asks how the patient’s friends and coworkers will react to this decision. Will they refuse to help him, punishing him for refusing the treatment? Sample reframed the dilemma. Instead of case of disability versus health, he suggested that people around the disabled person are trying to force him to be more like them, dismissing his right to a life of his choosing.
More generally, many ethicists want to ban genetic interventions that change the genes for future generations, because it removes self-determination from them.
Big steps into the philosophical mists
Several sessions and poster sessions at the conference ventured into the nearly metaphysical. What is life, and what are the rights of living things? Will artifical intelligence progress to the point where we have to grant independent agency to applications? How about the organoids generated in the labs from brain cells?
We should not be surprised that new biological discoveries force these difficult questions on us. Technological advances routinely challenge our assumptions. As an example that’s easy to grasp, HIPAA was passed in 1996 and covered all the organizations that were expected at that time to handle patient data: clinical settings, payers, and their business associates. The framers of the law did not anticipate fitness devices that would beam our vital signs into the cloud, or social media sites where people would share intimate details.
And over the past couple decades, we’ve learned more about the amazing world of animals and plants, where intelligence, planning, communication, feeling, and empathy are much more widespread than earlier generations of our own species were willing to accept. So our struggle to set rules for modern biological research calls on us to re-examine some of our most cherished values.
