How Do Cancer Doctors Frame Information?
Can we learn from Kahneman & Tversky?
It is my pleasure to feature this article by Dr. Raj Chakraborty. In it he discusses a patient of his who decided to try one more cancer drug option, and sadly never made it home. In the essay he contemplates different ways he might have presented the data to yield different choices.
Like all oncologists, I know Raj’s pain; We spend countless hours thinking about our patients, and puzzling over the eternal question: what if I had done it differently? Spoke differently, acted differently, recommended differently?
Ultimately, I believe that patient’s decide, but doctors advise, and a trusted advisor yields a mighty power and responsibility. I hope you enjoy his essay as much as I did
Vinay Prasad MD MPH
How Do Cancer Doctors Frame Information?
Can we learn from Kahneman & Tversky?
On a busy Monday morning in myeloma clinic, I got a message from the Intensive Care Unit resident. My patient Bernice (name changed for anonymity), whom I was expecting to see, had been admitted overnight for a bleed inside her brain due to low platelet counts secondary to complication from her chemotherapy.
Bernice was diagnosed with multiple myeloma, a cancer arising from antibody producing cells in the bone marrow, about a decade ago at the dawn of arrival of several effective chemotherapy agents. Although she responded well to treatments and enjoyed a good quality of life, recently, her cancer relapsed several times over the past few years, with remissions becoming progressively shorter in duration.
The last time this happened, it was clear we were running out of options. Furthermore, she has become progressively weaker physically and her kidneys had suffered significant damage, which made her ineligible for any clinical trial or intensive treatment.
At that point, I had several discussions with her regarding potential options including supportive and palliative care alone versus trying one of the chemotherapy agents with marginal benefit. She chose the latter after learning about the potential risks and benefits. Unfortunately, she did not benefit from the drug and experienced several adverse events leading to recurrent hospitalizations that culminated in the aforementioned ICU admission for intracranial bleed that ultimately took her life.
As oncologists, our job is not only to help patients navigate the diagnosis and treatment as smoothly as possible, but also to enable a good death. The last couple of months of Bernice’s journey made me feel as if I had failed to foresee the unfortunate turn of events and failed to help her make the decision that may have been in her best interest. Despite the futility of aggressive interventions toward the end of life, why do patients and physicians struggle to make rational decisions?
Some potential solutions can be found in the writings of Daniel Kahneman and Amos Tversky, and described in their seminal article titled “Choices, Values, and Frames” that was presented at the American Psychological Association meeting in 1983. Two critical tenets of prospect theory (backed by a wealth of empirical data) can enlighten us about decision making for risky interventions in medicine.
First, most humans are risk-averse in the face of gain and risk-seeking in the face of loss. The latter scenario is extremely pertinent to patients with advanced cancer who are facing imminent death unless we deploy an effective intervention, i.e. an intervention that has robust evidence on prolonging life or improving the quality of remaining life. However, the situation gets complicated when there is significant uncertainty regarding the evidence of clinical benefit from the intervention, especially if the intervention has serious and life-threatening toxicities. For example, let’s recreate in our mind one of the conversations I may have had with Bernice when her cancer was relapsing. I could have presented her the options in two different ways:
Option 1: If we do not pursue any further chemotherapy and focus on supportive and palliative care, our best guess is that most patients in this situation live another 6 months on average. If we try this chemotherapy, there is 70% probability that it would not work and may decrease the quality of your remaining life due to side-effects, including 10% chance of dying from side-effects that could be sooner than 6 months. There is 30% probability of your cancer regressing, which may or may not translate into you living longer than 6 months.
Option 2: If we do not pursue any further chemotherapy and focus on supportive and palliative care alone, our best guess is that most patients in your situation will die in around 6 months. If we try this chemotherapy agent, there is 30% probability of your cancer regressing, which may translate into you living longer than 6 months. However, there is 70% probability that it would not work and may decrease the quality of your remaining life due to side-effects, including 10% chance of dying from side-effects that could be sooner than 6 months.
In the first option, information is provided in gain-frame (live or survive for 6 months), and in the second option, in loss-frame (die in 6 months). Extrapolation of data from Kahneman and Tversky’s work would suggest that a patient is more likely to be risk-averse in option 1 and risk-seeking (i.e. accept chemotherapy) in option 2.
Hence, if we believe that the chances of benefit from an intervention is slim and risk of harm from side-effects is large based on our clinical judgement, should we present information to our patients in gain-frame? Is this a strategy we should deploy to protect our patients from taking dangerous risks that may shorten their life and decrease the quality of their remaining life? Of course, the situation would be much different if the evidence on life-prolonging benefit of an intervention is robust such that we would want our patients to take the risk and accept the intervention.
The second tenet of prospect theory that is the category-boundary effect, which implies that a change from impossibility to possibility (e.g. increase in response rate of a chemotherapeutic agent from 0% to 10%) and possibility to certainty (e.g. increase in response rate of a chemotherapeutic agent from 90% to 100%) has an outsized impact on decision-making compared to changes in the middle of the scale.
Although the latter situation is rare, approximately one in five cancer drugs that are approved based on the endpoint of response rate, which is an imperfect surrogate of living longer or living better, have an overall response rate of less than 20%. Hence, it’s plausible that patients may unrealistically overweigh their odds of benefit and seek risk with improbable gains in these situations, especially when the alternative is no treatment (i.e. response rate of 0%).
Let’s imagine that a drug has real side-effects which can impair the quality of life, add financial toxicity due to high out-of-pocket costs, and/or time-toxicity (i.e. more time away from things they may love) due to frequent contact with healthcare. If this drug has a response rate of around 15% with a large uncertainty on whether that translates into liver longer or better, would it be in the best interest of a patient to have this an option?
Of course, there is no one right answer to these complex questions. It depends on many things, patients’ values and preferences being a critically important factor. However, as a physician with unique content expertise on the risk and benefit of an intervention in a particular context and (hopefully) deep insight into the values, preferences, and expectations of their patient, they do shoulder the responsibility of framing the outcomes in a way such that patients do not reflexively take dangerous risks in the face of improbable gain or be extremely risk-averse in the face of likely loss. As oncologists, we often have long-term relationship with our patients and their families and become emotionally attached to their outcomes. Hence, knowledge of our inherent heuristics and biases in decision making under uncertainty may be even beneficial to us in having a greater insight into the fallacies of our recommendations.
As I continue to struggle with framing options to my patients in difficult circumstances when there is no one best answer, I recall my visits with Bernice and the decision we made together. What if I had presented the information to Bernice in a way that might have nudged her away from taking on the risk of trying chemotherapy as a desperate effort to prolong her life?
If we had just focused on her comfort and pain control, would she have been able to spend the Holidays with her family one last time?
How will I handle similar situations with my future patients?
Rajshekhar Chakraborty, MD is Assistant Professor at Columbia Medical Center.