Not Your Median Patient:
How A Climate Scientist Faced Cancer
By John Unger Zussman
Stephen Schneider, the environmental scientist, died of a heart attack last month at the age of 65. He was a Stanford professor, a member of the Nobel Prize-winning Intergovernmental Panel on Climate Change, and a tireless and articulate advocate for action to counteract the threat of global warming. The blog Realclimate.org has posted a moving “scientific obituary” by Benjamin Santer of the Lawrence Livermore National Laboratory. There was even a tribute on the NPR program Science Friday.
What most of the obituaries mention only in passing (if at all) is that Schneider was a survivor of a rare and often deadly cancer called mantle cell lymphoma. He was diagnosed in 2001 and refused to accept both the medical establishment’s standard treatment and its dismal prognosis. Instead, he diligently applied to his cancer the same principles of decision analysis that he used as a climate scientist, and as a result persuaded his oncologist, Dr. Sandra Horning, to treat him more aggressively than the protocols dictated. In 2005, safely in remission, he published an account of that battle, The Patient from Hell.
I’ve written about cancer in these pages before, recounting what I learned when someone I love, “Bonnie,” was diagnosed with breast cancer, and discussing the role of environmental toxins in causing cancer. If you read that first piece, you might be surprised that Bonnie, who chose to “undertreat” her cancer, looked to Schneider, who chose to “overtreat” his, as a positive model. But she did, and in this essay I want to explain why.
Both Bonnie and Schneider found inspiration, in turn, from Stephen Jay Gould, the late, great evolutionary biologist. At age 41, Gould was diagnosed with abdominal mesothelioma, a cancer from which the median survival was only eight months after diagnosis. Two years later, he published a seminal essay entitled “The Median Isn’t the Message,” in which he interpreted that statistic and explained how he took hope from it. Beautifully written, it’s must reading—especially if you have been diagnosed with a serious illness.
The median, if you remember your college statistics, is a measure of the “average value” of a set of measurements that are distributed on a curve. It refers to the “middle” value if the measurements are sorted, high to low—half the scores are above, half below. It’s useful because it’s less sensitive to extremes than the mean, or mathematical average.
Gould realized that there was nothing magical about the median; it’s a measure of central tendency, but it doesn’t describe the distribution. His own survival might be any data point on that curve. Yes, 50% of abdominal mesothelioma patients survived eight months or less, but another 50% survived longer—some, given the characteristics of the distribution, significantly longer. Rather than despair, Gould set about figuring out how he could get himself on the long end of that curve.
And he did. He beat that cancer and had twenty more healthy and productive years—completing his magnum opus, The Structure of Evolutionary Theory—before he succumbed to an unrelated lung cancer in 2002.
Schneider’s insight was no less profound. He looked at the way the medical profession evaluates cancer treatments based on endpoints—survival or recurrence of the median patient at the end of a clinical trial of one treatment vs. another (or vs. a placebo). A good starting point, he said. But what if the disease is rare (like his) and there are no trials? Or what if a promising new treatment hasn’t yet been tested? Or what if you don’t resemble the median patient?
This situation reminded Schneider of climate science, in which you have a single patient (the Earth) whose symptoms are beginning to alarm you. Unfortunately, you lack good data on endpoints, your predictive models are imperfect, and the error ranges in your forecasts multiply upon each other. You have no other planets to run experiments with. “You can’t do controlled experiments on the future,” Schneider said in his last major interview. “What are we going to do, wait for it? Then apologize to posterity that we did nothing to slow it down?” Of course not. You collect as much data as you can about the climatic processes. You make your best estimates of the probabilities of each outcome, cognizant that they are only estimates, while continuing to monitor the data and update your models. And you make your best recommendations for policy that, you hope, will avoid the most catastrophic outcomes.
Applying these principles to medicine, in place of absent or inadequate clinical trials, or to supplement them, Schneider recommended process knowledge, Bayesian updating, and decision analysis. Process knowledge means that “your doctors should know how various treatments—both mainstream and not—work, how treatments for diseases similar to yours might work for you, what treatments are unlikely to be effective, and how your overall health could be affected.” Bayesian updating is a fancy statistical term for monitoring your response to treatment and adjusting it accordingly. Decision analysis means weighting potential outcomes, risks, and benefits by your doctors’ estimate of the likelihood of their occurrence as well as their confidence in that estimate.
Finally, Schneider, like Bonnie, insisted that his medical decisions were his to make. He treated his doctors as medical advisors, valuable for their knowledge, experience, and intuitions, but ultimately he had to make his own choices based on their advice.
If this sounds abstract, let me try to clarify with an example. One of the treatments Schneider’s doctors recommended, in addition to standard chemo, radiation, and a bone marrow transplant, was a new antibody drug called Rituxan. Rituxan targets a protein called CD20 that is expressed on the surface of B cells, the immune cells that cause mantle cell and other lymphomas. Essentially, Rituxan allows the body’s immune system to recognize B cells as foreign cells, which are then destroyed by the immune system’s NK (natural killer) cells.
The standard protocol was to administer a large dose of Rituxan (along with conventional chemo), then perform a bone marrow transplant, and then—to wait and see if the cancer came back, as detected by a CT scan. If it did—and 50% of patients on this protocol lost remission at least once within four years—they would try it again. But a second remission was harder to achieve than the first.
This didn’t sound like a good idea to Schneider. First, he questioned the means of monitoring his response to treatment. A CT scan won’t detect signs of cancer until there were already hundreds of millions of malignant cells and a detectable lump. Was there a way to monitor him more closely? It turned out there was, a highly sensitive molecular-based diagnostic test called PCR (polymeric chain reaction). PCR would provide a much more accurate measure of Schneider’s cancer cell count.
Schneider also realized that if the protocol didn’t kill the disease completely—and it was a coin toss whether it would—then the remaining cancer cells would keep reproducing, and in a few years he’d be right back where he started. Why not take a different tack, he reasoned, and presume the cancer is still present? It didn’t matter if his cancer was never cured, so long as the malignant cell count was kept below a dangerous level.
So Schneider requested what he called “maintenance therapy.” After the standard chemo, Rituxan, and bone marrow transplant, he wanted to receive low “maintenance” doses of Rituxan at periodic intervals. He also wanted periodic counts of his CD20 cells (to measure whether the Rituxan was wearing off) and cancer cells (via PCR), so that the dosage and interval could be adjusted if necessary. If his cancer cell count crept above negligible levels—signifying that his immune system wasn’t adequately controlling the cells—he would get another dose of Rituxan.
One of his doctors disapproved strongly. “We have no data whatsoever to suggest that low-dose interventions would have any benefit,” he said, “and operating without data would be foolhardy.” If there were no clinical trials showing that a particular treatment worked, it would be risky, idiotic, and even unethical to use it. Besides, side effects of the Rituxan might be damaging or even fatal.
So Schneider took his doctors through a decision analysis hypothetical. “Suppose you brought 100 people in off the street and gave them periodic maintenance doses of Rituxan,” he asked. “How many would die from the treatment?” The docs protested that they didn’t know because there were no trials. But Schneider persisted; he asked them to use their best clinical intuition and judgment. “Probably none, perhaps one,” Dr. Horning finally acknowledged.
“Now suppose you put 100 patients like me on the standard protocol,” Schneider asked. “How many would lose remission within five years?” “Probably eighty,” answered the docs. “And how many of those would not get back into remission with another chemo regime?” “Forty,” estimated the docs. “Who wouldn’t take that risk?” Schneider argued.
Schneider understood that he and his doctors were wandering into uncharted territory with Rituxan maintenance therapy. “There was no telling it would work,” he wrote in The Patient from Hell, “and no previous data that might help us develop a treatment plan.” But those are the risks you have to take when you are essentially running an experiment with incomplete data and a sample size of one. He viewed climate change the same way; with only one earth to experiment with, we’d better choose the treatment with the best chance of avoiding a catastrophic outcome.
This, I think, gives you a flavor of the way Schneider approached problems, both in climate science and in his battle with cancer. Ultimately, Dr. Horning agreed to put him on maintenance therapy—a low dose of Rituxan every three months—and it worked.
But Schneider was also an excellent writer and an engaging speaker, so I should get out of the way and let him explain it himself. Specifically, in 2008, Schneider gave a presentation at Stanford Medical School’s Café Scientifique entitled “Cancer and Climate Change: Parallels in Risk Management,” in which he outlined his approach to both issues. This talk is accessible and informative, and if you’re interested, I urge you to view it on YouTube or download it as a podcast from iTunes (available in the Medcast series from Stanford on iTunes U).
Near the end of the lecture (around 1:11), Schneider speculates, like Bonnie, that one reason doctors are so tied into the “standard treatment” has to do with liability issues. They are sometimes held accountable when, despite their best efforts, things go wrong. “I prescribed the standard treatment” is an almost universally successful defense against malpractice lawsuits. Schneider suggests that legislation is needed to exculpate doctors who, with their patients’ understanding and consent, depart from standard treatment to personalize their care.
One might wonder whether the aggressive treatment of Schneider’s cancer weakened his heart and eventually led to his heart attack. Of course, it’s impossible to know for certain. But it was a risk he took knowingly and voluntarily, believing that without that treatment he would likely be dead.
What Schneider and Bonnie share is their refusal to take the standard protocol on faith, their willingness to get their hands dirty with admittedly limited data, their insistence that their treatment be personalized, and their resolve to take responsibility for their own treatment decisions and outcomes. In the end, Schneider was able to spend his last nine years in good health, with his family, doing his life’s work—trying to make sure that we approach our climate the same way he approached his cancer. And that’s all any of us can ask for.
Copyright © 2010, John Unger Zussman. All rights reserved.
Not Your Median Patient:
How A Scientist Faced Cancer
By John Unger Zussman
Stephen Schneider, the environmental scientist, died of a heart attack last month at the age of 65. He was a Stanford professor, a member of the Nobel Prize-winning Intergovernmental Panel on Climate Change, and a tireless and articulate advocate for action to counteract the threat of global warming. The blog Realclimate.org has posted a moving “scientific obituary” by Benjamin Santer of the Lawrence Livermore National Laboratory. There was even a tribute on the NPR program Science Friday.
What most of the obituaries mention only in passing (if at all) is that Schneider was a survivor of a rare and often deadly cancer called mantle cell lymphoma. He was diagnosed in 2001 and refused to accept both the medical establishment’s standard treatment and its dismal prognosis. Instead, he diligently applied to his cancer the same principles of decision analysis that he used as a climate scientist, and as a result persuaded his oncologist, Dr. Sandra Horning, to treat him more aggressively than the protocols dictated. In 2005, safely in remission, he published an account of that battle, The Patient from Hell.
I’ve written about cancer in these pages before, recounting what I learned when someone I love, “Bonnie,” was diagnosed with breast cancer, and discussing the role of environmental toxins in causing cancer. If you read that first piece, you might be surprised that Bonnie, who chose to “undertreat” her cancer, looked to Schneider, who chose to “overtreat” his, as a positive model. But she did, and in this essay I want to explain why.
Both Bonnie and Schneider found inspiration, in turn, from Stephen Jay Gould, the late, great evolutionary biologist. At age 41, Gould was diagnosed with abdominal mesothelioma, a cancer from which the median survival was only eight months after diagnosis. Two years later, he published a seminal essay entitled “The Median Isn’t the Message,” in which he interpreted that statistic and explained how he took hope from it. Beautifully written, it’s must reading—especially if you have been diagnosed with a serious illness.
The median, if you remember your college statistics, is a measure of the “average value” (central tendency) of a set of measurements that are distributed along a scale. It refers to the “middle” value if the measurements are sorted, top to bottom—half the scores are above, half below. It’s useful because it’s less sensitive to extremes than the mean, or mathematical average.
Gould realized that there was nothing magical about the median; it’s a measure of central tendency, but it doesn’t describe the distribution. His own survival might be any data point on that curve (constrained by the shape of the distribution). Yes, 50% of abdominal mesothelioma patients survived eight months or less, but another 50% survived longer—some, given the characteristics of the distribution, significantly longer. Rather than despair, Gould set about figuring out how he could get himself on the long end of that curve.
And he did. He beat that cancer and had twenty more healthy and productive years—completing his magnum opus, The Structure of Evolutionary Theory—before he succumbed to an unrelated lung cancer in 2002.
Schneider’s insight was no less profound. He looked at the way the medical profession evaluates cancer treatments based on endpoints—survival or recurrence of the median patient at the end of a clinical trial of one treatment vs. another (or vs. a placebo). A good starting point, he said. But what if the disease is rare (like his) and there are no trials? Or what if a promising new treatment hasn’t yet been tested? Or what if you don’t resemble the median patient?
This situation reminded Schneider of climate science, in which you have a single patient (the Earth) whose symptoms are beginning to alarm you. Unfortunately, you lack good data on endpoints, your predictive models are imperfect, and the error ranges in your forecasts multiply upon each other. You have no other planets to run experiments with. “You can’t do controlled experiments on the future,” Schneider said in his last major interview. “What are we going to do, wait for it?” Of course not. You collect as much data as you can about the climatic processes. You make your best estimates of the probabilities of each outcome, cognizant that they are only estimates, while continuing to monitor the data and update your models. And you make your best recommendations for policy that, you hope, will avoid the most catastrophic outcomes.
Applying these principles to medicine, in place of absent or inadequate clinical trials, or to supplement them, Schneider recommends process knowledge, Bayesian updating, and decision analysis. Process knowledge means that “your doctors should know how various treatments—both mainstream and not—work, how treatments for diseases similar to yours might work for you, what treatments are unlikely to be effective, and how your overall health could be effective.” Bayesian updating is a fancy statistical term for monitoring your response to treatment and adjusting it accordingly. Decision analysis means weighting potential outcomes, risks, and benefits by your doctors’ estimate of the likelihood of their occurrence as well as their confidence in that estimate.
Finally, Schneider, like Bonnie, insisted that his medical decisions were his to make. He treated his doctors as medical advisors, valuable for their knowledge, experience, and intuitions, but ultimately he had to make his own choices based on their advice.
If this sounds abstract, let me try to clarify with an example. One of the treatments Schneider’s doctors recommended, in addition to standard chemo, radiation, and a bone marrow transplant, was a new antibody drug called Rituxan. Rituxan targets a protein called CD20 that is expressed on the surface of B cells, the immune cells that cause mantle cell and other lymphomas. Essentially, Rituxan allows the body’s immune system to recognize B cells as foreign cells, which are then destroyed by the immune system’s NK (natural killer) cells.
The standard protocol was to administer a large dose of Rituxan (along with conventional chemo), then perform a bone marrow transplant, and then—to wait and see if the cancer came back, as detected by a CT scan. If it did—and 50% of patients on this protocol lost remission at least once within four years—they would try it again. But a second remission was harder to achieve than the first.
This didn’t sound like a good idea to Schneider. First, he questioned the means of monitoring his response to treatment. A CT scan won’t detect signs of cancer until there were already hundreds of millions of malignant cells and a detectable lump. Was there a way to monitor him more closely? It turned out there was, a highly sensitive molecular-based diagnostic test called PCR (polymeric chain reaction). PCR would provide a much more accurate measure of Schneider’s cancer cell count.
Schneider also realized that if the protocol didn’t kill the disease completely—and it was a coin toss whether it would—then the remaining cancer cells would keep reproducing, and in a few years he’d be right back where he started. Why not take a different tack, he reasoned, and presume the cancer is still present? It didn’t matter if his cancer was never cured, so long as the malignant cell count was kept below a dangerous level.
So Schneider requested what he called “maintenance therapy.” After the standard chemo, Rituxan, and bone marrow transplant, he wanted to receive low “maintenance” doses of Rituxan at periodic intervals. He also wanted periodic counts of his CD20 cells (to measure whether the Rituxan was wearing off) and cancer cells (via PCR), so that the dosage and interval could be adjusted if necessary. If his cancer cell count crept above negligible levels—signifying that his immune system wasn’t adequately controlling the cells—he would get another dose of Rituxan.
One of his doctors disapproved strongly. “We have no data whatsoever to suggest that low-dose interventions would have any benefit,” he said, “and operating without data would be foolhardy.” If there were no clinical trials showing that a particular treatment worked, it would be risky, idiotic, and even unethical to use it. Besides, side effects of the Rituxan might be damaging or even fatal.
So Schneider took his doctors through a decision analysis hypothetical. “Suppose you brought 100 people in off the street and gave them periodic maintenance doses of Rituxan,” he asked. “How many would die from the treatment?” The docs protested that they didn’t know because there were no trials. But Schneider persisted; he asked them to use their best clinical intuition and judgment. “Probably none, perhaps one,” Dr. Horning finally acknowledged.
“Now suppose you put 100 patients like me on the standard protocol,” Schneider asked. “How many would lose remission within five years?” “Probably eighty,” answered the docs. “And how many of those would not get back into remission with another chemo regime?” “Forty,” estimated the docs. “Who wouldn’t take that risk?” Schneider argued.
Schneider understood that he and his doctors were wandering into uncharted territory with Rituxan maintenance therapy. “There was no telling it would work,” he wrote in The Patient from Hell, “and no previous data that might help us develop a treatment plan.” But those are the risks you have to take when you are essentially running an experiment with incomplete data and a sample size of one. He viewed climate change the same way; with only one earth to experiment with, we’d better choose the treatment with the best chance of avoiding a catastrophic outcome.
This, I think, gives you a flavor of the way Schneider approached problems, both in climate science and in his battle with cancer. Ultimately, Dr. Horning agreed to put him on maintenance therapy—a low dose of Rituxan every three months—and it worked.
But Schneider was also an excellent writer and an engaging speaker, so I should get out of the way and let him explain it himself. Specifically, in 2008, Schneider gave a presentation at Stanford Medical School’s Café Scientifique entitled “Cancer and Climate Change: Parallels in Risk Management,” in which he outlined his approach to both issues. This talk is accessible and informative, and if you’re interested, I urge you to view it on YouTube or download it as a podcast from iTunes (available in the Medcast series from Stanford on iTunes U).
Near the end of the lecture (around 1:11), Schneider speculates, like Bonnie, that one reason doctors are so tied into the “standard treatment” has to do with liability issues. They are sometimes held accountable when, despite their best efforts, things go wrong. “I prescribed the standard treatment” is an almost universally successful defense against malpractice lawsuits. Schneider suggests that legislation is needed to exculpate doctors who, with their patients’ understanding and consent, depart from standard treatment to personalize their care.
One might wonder whether Schneider’s aggressive treatment of his cancer weakened his heart and eventually led to his heart attack. Of course, it’s impossible to know for certain. But it was a risk he took knowingly and voluntarily, believing that without that treatment he would likely be dead.
What Schneider and Bonnie share is their refusal to take the standard protocol on faith, their willingness to get their hands dirty with admittedly limited data, their insistence that their treatment be personalized, and their resolve to take responsibility for their own treatment decisions and outcomes. In the end, Schneider was able to spend his last nine years in good health, with his family, doing his life’s work—trying to make sure that we approach our climate the same way he approached his cancer. And that’s all any of us can ask for.
Copyright © 2010, John Unger Zussman. All rights reserved.
As It Ought To Be 