The Death Predictor: A Helpful New Tool or an Ethical Morass?
Whenever Eric Karl Oermann has to tell a patient about a terrible prognosis, their first question is always: "how long do I have?" Oermann would like to offer a precise answer, to provide some certainty and help guide treatment. But although he's one of the country's foremost experts in medical artificial intelligence, Oermann is still dependent on a computer algorithm that's often wrong.
Doctors are notoriously terrible at guessing how long their patients will live.
Artificial intelligence, now often called deep learning or neural networks, has radically transformed language and image processing. It's allowed computers to play chess better than the world's grand masters and outwit the best Jeopardy players. But it still can't precisely tell a doctor how long a patient has left – or how to help that person live longer.
Someday, researchers predict, computers will be able to watch a video of a patient to determine their health status. Doctors will no longer have to spend hours inputting data into medical records. And computers will do a better job than specialists at identifying tiny tumors, impending crises, and, yes, figuring out how long the patient has to live. Oermann, a neurosurgeon at Mount Sinai, says all that technology will allow doctors to spend more time doing what they do best: talking with their patients. "I want to see more deep learning and computers in a clinical setting," he says, "so there can be more human interaction." But those days are still at least three to five years off, Oermann and other researchers say.
Doctors are notoriously terrible at guessing how long their patients will live, says Nigam Shah, an associate professor at Stanford University and assistant director of the school's Center for Biomedical Informatics Research. Doctors don't want to believe that their patient – whom they've come to like – will die. "Doctors over-estimate survival many-fold," Shah says. "How do you go into work, in say, oncology, and not be delusionally optimistic? You have to be."
But patients near the end of life will get better treatment – and even live longer – if they are overseen by hospice or palliative care, research shows. So, instead of relying on human bias to select those whose lives are nearing their end, Shah and his colleagues showed that they could use a deep learning algorithm based on medical records to flag incoming patients with a life expectancy of three months to a year. They use that data to indicate who might need palliative care. Then, the palliative care team can reach out to treating physicians proactively, instead of relying on their referrals or taking the time to read extensive medical charts.
But, although the system works well, Shah isn't yet sure if such indicators actually get the appropriate patients into palliative care. He's recently partnered with a palliative care doctor to run a gold-standard clinical trial to test whether patients who are flagged by this algorithm are indeed a better match for palliative care.
"What is effective from a health system perspective might not be effective from a treating physician's perspective and might not be effective from the patient's perspective," Shah notes. "I don't have a good way to guess everybody's reaction without actually studying it." Whether palliative care is appropriate, for instance, depends on more than just the patient's health status. "If the patient's not ready, the family's not ready and the doctor's not ready, then you're just banging your head against the wall," Shah says. "Given limited capacity, it's a waste of resources" to put that person in palliative care.
The algorithm isn't perfect, but "on balance, it leads to better decisions more often."
Alexander Smith and Sei Lee, both palliative care doctors, work together at the University of California, San Francisco, to develop predictions for patients who come to the hospital with a complicated prognosis or a history of decline. Their algorithm, they say, helps decide if this patient's problems – which might include diabetes, heart disease, a slow-growing cancer, and memory issues – make them eligible for hospice. The algorithm isn't perfect, they both agree, but "on balance, it leads to better decisions more often," Smith says.
Bethany Percha, an assistant professor at Mount Sinai, says that an algorithm may tell doctors that their patient is trending downward, but it doesn't do anything to change that trajectory. "Even if you can predict something, what can you do about it?" Algorithms may be able to offer treatment suggestions – but not what specific actions will alter a patient's future, says Percha, also the chief technology officer of Precise Health Enterprise, a product development group within Mount Sinai. And the algorithms remain challenging to develop. Electronic medical records may be great at her hospital, but if the patient dies at a different one, her system won't know. If she wants to be certain a patient has died, she has to merge social security records of death with her system's medical records – a time-consuming and cumbersome process.
An algorithm that learns from biased data will be biased, Shah says. Patients who are poor or African American historically have had worse health outcomes. If researchers train an algorithm on data that includes those biases, they get baked into the algorithms, which can then lead to a self-fulfilling prophesy. Smith and Lee say they've taken race out of their algorithms to avoid this bias.
Age is even trickier. There's no question that someone's risk of illness and death goes up with age. But an 85-year-old who breaks a hip running a marathon should probably be treated very differently than an 85-year-old who breaks a hip trying to get out of a chair in a dementia care unit. That's why the doctor can never be taken out of the equation, Shah says. Human judgment will always be required in medical care and an algorithm should never be followed blindly, he says.
Experts say that the flaws in artificial intelligence algorithms shouldn't prevent people from using them – carefully.
Researchers are also concerned that their algorithms will be used to ration care, or that insurance companies will use their data to justify a rate increase. If an algorithm predicts a patient is going to end up back in the hospital soon, "who's benefitting from knowing a patient is going to be readmitted? Probably the insurance company," Percha says.
Still, Percha and others say, the flaws in artificial intelligence algorithms shouldn't prevent people from using them – carefully. "These are new and exciting tools that have a lot of potential uses. We need to be conscious about how to use them going forward, but it doesn't mean we shouldn't go down this road," she says. "I think the potential benefits outweigh the risks, especially because we've barely scratched the surface of what big data can do right now."
Jamie Rettinger was still in his thirties when he first noticed a tiny streak of brown running through the thumbnail of his right hand. It slowly grew wider and the skin underneath began to deteriorate before he went to a local dermatologist in 2013. The doctor thought it was a wart and tried scooping it out, treating the affected area for three years before finally removing the nail bed and sending it off to a pathology lab for analysis.
"I have some bad news for you; what we removed was a five-millimeter melanoma, a cancerous tumor that often spreads," Jamie recalls being told on his return visit. "I'd never heard of cancer coming through a thumbnail," he says. None of his doctors had ever mentioned it either. "I just thought I was being treated for a wart." But nothing was healing and it continued to bleed.
A few months later a surgeon amputated the top half of his thumb. Lymph node biopsy tested negative for spread of the cancer and when the bandages finally came off, Jamie thought his medical issues were resolved.
Melanoma is the deadliest form of skin cancer. About 85,000 people are diagnosed with it each year in the U.S. and more than 8,000 die of the cancer when it spreads to other parts of the body, according to the Centers for Disease Control and Prevention (CDC).
There are two peaks in diagnosis of melanoma; one is in younger women ages 30-40 and often is tied to past use of tanning beds; the second is older men 60+ and is related to outdoor activity from farming to sports. Light-skinned people have a twenty-times greater risk of melanoma than do people with dark skin.
"When I graduated from medical school, in 2005, melanoma was a death sentence" --Diwakar Davar.
Jamie had a follow up PET scan about six months after his surgery. A suspicious spot on his lung led to a biopsy that came back positive for melanoma. The cancer had spread. Treatment with a monoclonal antibody (nivolumab/Opdivo®) didn't prove effective and he was referred to the UPMC Hillman Cancer Center in Pittsburgh, a four-hour drive from his home in western Ohio.
An alternative monoclonal antibody treatment brought on such bad side effects, diarrhea as often as 15 times a day, that it took more than a week of hospitalization to stabilize his condition. The only options left were experimental approaches in clinical trials.
Early research
"When I graduated from medical school, in 2005, melanoma was a death sentence" with a cure rate in the single digits, says Diwakar Davar, 39, an oncologist at UPMC Hillman Cancer Center who specializes in skin cancer. That began to change in 2010 with introduction of the first immunotherapies, monoclonal antibodies, to treat cancer. The antibodies attach to PD-1, a receptor on the surface of T cells of the immune system and on cancer cells. Antibody treatment boosted the melanoma cure rate to about 30 percent. The search was on to understand why some people responded to these drugs and others did not.
At the same time, there was a growing understanding of the role that bacteria in the gut, the gut microbiome, plays in helping to train and maintain the function of the body's various immune cells. Perhaps the bacteria also plays a role in shaping the immune response to cancer therapy.
One clue came from genetically identical mice. Animals ordered from different suppliers sometimes responded differently to the experiments being performed. That difference was traced to different compositions of their gut microbiome; transferring the microbiome from one animal to another in a process known as fecal transplant (FMT) could change their responses to disease or treatment.
When researchers looked at humans, they found that the patients who responded well to immunotherapies had a gut microbiome that looked like healthy normal folks, but patients who didn't respond had missing or reduced strains of bacteria.
Davar and his team knew that FMT had a very successful cure rate in treating the gut dysbiosis of Clostridioides difficile, a persistant intestinal infection, and they wondered if a fecal transplant from a patient who had responded well to cancer immunotherapy treatment might improve the cure rate of patients who did not originally respond to immunotherapies for melanoma.
The ABCDE of melanoma detection
Adobe Stock
Clinical trial
"It was pretty weird, I was totally blasted away. Who had thought of this?" Jamie first thought when the hypothesis was explained to him. But Davar's explanation that the procedure might restore some of the beneficial bacterial his gut was lacking, convinced him to try. He quickly signed on in October 2018 to be the first person in the clinical trial.
Fecal donations go through the same safety procedures of screening for and inactivating diseases that are used in processing blood donations to make them safe for transfusion. The procedure itself uses a standard hollow colonoscope designed to screen for colon cancer and remove polyps. The transplant is inserted through the center of the flexible tube.
Most patients are sedated for procedures that use a colonoscope but Jamie doesn't respond to those drugs: "You can't knock me out. I was watching them on the TV going up my own butt. It was kind of unreal at that point," he says. "There were about twelve people in there watching because no one had seen this done before."
A test two weeks after the procedure showed that the FMT had engrafted and the once-missing bacteria were thriving in his gut. More importantly, his body was responding to another monoclonal antibody (pembrolizumab/Keytruda®) and signs of melanoma began to shrink. Every three months he made the four-hour drive from home to Pittsburgh for six rounds of treatment with the antibody drug.
"We were very, very lucky that the first patient had a great response," says Davar. "It allowed us to believe that even though we failed with the next six, we were on the right track. We just needed to tweak the [fecal] cocktail a little better" and enroll patients in the study who had less aggressive tumor growth and were likely to live long enough to complete the extensive rounds of therapy. Six of 15 patients responded positively in the pilot clinical trial that was published in the journal Science.
Davar believes they are beginning to understand the biological mechanisms of why some patients initially do not respond to immunotherapy but later can with a FMT. It is tied to the background level of inflammation produced by the interaction between the microbiome and the immune system. That paper is not yet published.
Surviving cancer
It has been almost a year since the last in his series of cancer treatments and Jamie has no measurable disease. He is cautiously optimistic that his cancer is not simply in remission but is gone for good. "I'm still scared every time I get my scans, because you don't know whether it is going to come back or not. And to realize that it is something that is totally out of my control."
"It was hard for me to regain trust" after being misdiagnosed and mistreated by several doctors he says. But his experience at Hillman helped to restore that trust "because they were interested in me, not just fixing the problem."
He is grateful for the support provided by family and friends over the last eight years. After a pause and a sigh, the ruggedly built 47-year-old says, "If everyone else was dead in my family, I probably wouldn't have been able to do it."
"I never hesitated to ask a question and I never hesitated to get a second opinion." But Jamie acknowledges the experience has made him more aware of the need for regular preventive medical care and a primary care physician. That person might have caught his melanoma at an earlier stage when it was easier to treat.
Davar continues to work on clinical studies to optimize this treatment approach. Perhaps down the road, screening the microbiome will be standard for melanoma and other cancers prior to using immunotherapies, and the FMT will be as simple as swallowing a handful of freeze-dried capsules off the shelf rather than through a colonoscopy. Earlier this year, the Food and Drug Administration approved the first oral fecal microbiota product for C. difficile, hopefully paving the way for more.
An older version of this hit article was first published on May 18, 2021
All organisms have the capacity to repair or regenerate tissue damage. None can do it better than salamanders or newts, which can regenerate an entire severed limb.
That feat has amazed and delighted man from the dawn of time and led to endless attempts to understand how it happens – and whether we can control it for our own purposes. An exciting new clue toward that understanding has come from a surprising source: research on the decline of cells, called cellular senescence.
Senescence is the last stage in the life of a cell. Whereas some cells simply break up or wither and die off, others transition into a zombie-like state where they can no longer divide. In this liminal phase, the cell still pumps out many different molecules that can affect its neighbors and cause low grade inflammation. Senescence is associated with many of the declining biological functions that characterize aging, such as inflammation and genomic instability.
Oddly enough, newts are one of the few species that do not accumulate senescent cells as they age, according to research over several years by Maximina Yun. A research group leader at the Center for Regenerative Therapies Dresden and the Max Planck Institute of Molecular and Cell Biology and Genetics, in Dresden, Germany, Yun discovered that senescent cells were induced at some stages of regeneration of the salamander limb, “and then, as the regeneration progresses, they disappeared, they were eliminated by the immune system,” she says. “They were present at particular times and then they disappeared.”
Senescent cells added to the edges of the wound helped the healthy muscle cells to “dedifferentiate,” essentially turning back the developmental clock of those cells into more primitive states.
Previous research on senescence in aging had suggested, logically enough, that applying those cells to the stump of a newly severed salamander limb would slow or even stop its regeneration. But Yun stood that idea on its head. She theorized that senescent cells might also play a role in newt limb regeneration, and she tested it by both adding and removing senescent cells from her animals. It turned out she was right, as the newt limbs grew back faster than normal when more senescent cells were included.
Senescent cells added to the edges of the wound helped the healthy muscle cells to “dedifferentiate,” essentially turning back the developmental clock of those cells into more primitive states, which could then be turned into progenitors, a cell type in between stem cells and specialized cells, needed to regrow the muscle tissue of the missing limb. “We think that this ability to dedifferentiate is intrinsically a big part of why salamanders can regenerate all these very complex structures, which other organisms cannot,” she explains.
Yun sees regeneration as a two part problem. First, the cells must be able to sense that their neighbors from the lost limb are not there anymore. Second, they need to be able to produce the intermediary progenitors for regeneration, , to form what is missing. “Molecularly, that must be encoded like a 3D map,” she says, otherwise the new tissue might grow back as a blob, or liver, or fin instead of a limb.
Wound healing
Another recent study, this time at the Mayo Clinic, provides evidence supporting the role of senescent cells in regeneration. Looking closely at molecules that send information between cells in the wound of a mouse, the researchers found that senescent cells appeared near the start of the healing process and then disappeared as healing progressed. In contrast, persistent senescent cells were the hallmark of a chronic wound that did not heal properly. The function and significance of senescence cells depended on both the timing and the context of their environment.
The paper suggests that senescent cells are not all the same. That has become clearer as researchers have been able to identify protein markers on the surface of some senescent cells. The patterns of these proteins differ for some senescent cells compared to others. In biology, such physical differences suggest functional differences, so it is becoming increasingly likely there are subsets of senescent cells with differing functions that have not yet been identified.
There are disagreements within the research community as to whether newts have acquired their regenerative capacity through a unique evolutionary change, or if other animals, including humans, retain this capacity buried somewhere in their genes.
Scientists initially thought that senescent cells couldn’t play a role in regeneration because they could no longer reproduce, says Anthony Atala, a practicing surgeon and bioengineer who leads the Wake Forest Institute for Regenerative Medicine in North Carolina. But Yun’s study points in the other direction. “What this paper shows clearly is that these cells have the potential to be involved in tissue regeneration [in newts]. The question becomes, will these cells be able to do the same in humans.”
As our knowledge of senescent cells increases, Atala thinks we need to embrace a new analogy to help understand them: humans in retirement. They “have acquired a lot of wisdom throughout their whole life and they can help younger people and mentor them to grow to their full potential. We're seeing the same thing with these cells,” he says. They are no longer putting energy into their own reproduction, but the signaling molecules they secrete “can help other cells around them to regenerate.”
There are disagreements within the research community as to whether newts have acquired their regenerative capacity through a unique evolutionary change, or if other animals, including humans, retain this capacity buried somewhere in their genes. If so, it seems that our genes are unable to express this ability, perhaps as part of a tradeoff in acquiring other traits. It is a fertile area of research.
Dedifferentiation is likely to become an important process in the field of regenerative medicine. One extreme example: a lab has been able to turn back the clock and reprogram adult male skin cells into female eggs, a potential milestone in reproductive health. It will be more difficult to control just how far back one wishes to go in the cell's dedifferentiation – part way or all the way back into a stem cell – and then direct it down a different developmental pathway. Yun is optimistic we can learn these tricks from newts.
Senolytics
A growing field of research is using drugs called senolytics to remove senescent cells and slow or even reverse disease of aging.
“Senolytics are great, but senolytics target different types of senescence,” Yun says. “If senescent cells have positive effects in the context of regeneration, of wound healing, then maybe at the beginning of the regeneration process, you may not want to take them out for a little while.”
“If you look at pretty much all biological systems, too little or too much of something can be bad, you have to be in that central zone” and at the proper time, says Atala. “That's true for proteins, sugars, and the drugs that you take. I think the same thing is true for these cells. Why would they be different?”
Our growing understanding that senescence is not a single thing but a variety of things likely means that effective senolytic drugs will not resemble a single sledge hammer but more a carefully manipulated scalpel where some types of senescent cells are removed while others are added. Combinations and timing could be crucial, meaning the difference between regenerating healthy tissue, a scar, or worse.