Thursday, December 3, 2015

Use and Misuse of Chemotherapy at the End of Life

by: Danny Cox (@dannyMD)

The goal of palliative chemotherapy for patients with terminal cancer is to prolong survival and improve quality of life. A major ongoing concern among many physicians is that chemotherapy given to those near the end of life might actually be doing more harm than good. The 2012 American Society of Clinical Oncology guidelines recommend against the use of chemotherapy for those with an ECOG score of ≥3 (ECOG is a graded scale of functional status; see table). The implicit assumption is that those with better functional status will be better able to tolerate chemotherapy and derive benefit.
An eye-opening new study entitled “Chemotherapy Use, Performance Status, and Quality of Life,” recently published by Prigerson et al. in the September 2015 edition of JAMA Oncology casts doubt on the perceived benefits of chemotherapy in terminally ill cancer patients even with good functional status and highlights the need for a more nuanced approach.

In this multi-center, longitudinal cohort study, 312 patients with progressive metastatic cancer with a life expectancy of ≤6 months with failure of at least 1 prior chemotherapy regimen were enrolled and followed prospectively until death. Significantly, half of these terminally ill cancer patients (50.6%) were receiving chemotherapy when enrolled in this study at a median of 4 months before death. And here’s the kicker: patients with good performance status at baseline (EGOG = 1), i.e. those most likely to “benefit” from chemotherapy, had worse QOL near death than those who did not receive chemotherapy. Patients with poorer performance status (ECOG ≥2) did not derive a QOL benefit from chemotherapy. Moreover, there was no survival benefit to chemotherapy in this study population.

So why is this happening? Why are our sickest cancer patients receiving toxic chemotherapeutics near the end of life when common sense, and now this new evidence, would dictate that there is limited benefit and even harm? One reason is that a physician, when faced with a patient in a terrible situation who is desperate to live, feels compelled to offer hope, to do something. It may be easier to write the next chemo order that the patient is requesting than to have a very difficult conversation acknowledging that a patient’s prognosis is limited and the treatments that have been tried have not worked.

Given this new evidence, physicians must have more candid conversations with their patients about the limited benefits and real risks of chemotherapy in terminal cancer. This way, patients and their families can make informed decisions about what is likely to happen rather than what they hope will happen. Medical education and professional development need to fill the gaps in communication skills to train our current and future oncologists to have these compassionate, courageous, and informative conversations. And together with their palliative care colleagues, oncologists must be armed with more tools in their proverbial tool belts like symptom relief and psychosocial support so that they can meaningfully offer something at that critical juncture when their terminally ill patient is crying for help and chemotherapy will not be effective.

A final take-away question introduced by this study: should tumor shrinkage or patient experience be the primary endpoint by which we judge the efficacy of any chemotherapeutic? Currently, most chemotherapy drug trials look at end-points like tumor size and time to progression of disease. Almost none examine QOL. The problem with the current end-point analysis paradigm is that you can make a tumor smaller, but the patient might not feel any better.

In fact, they sometimes feel worse.


Judith Henry said...

Until more people accept that death is a consequence of living, and a good life and a good death are not mutually exclusive, we'll continue to see physicians authorizing useless and often harmful treatments.

Dan Matlock said...

We reviewed this study in our palliative care journal club a month ago. I think we need to be very cautious about blaming the chemo for the poor QOL seen here.

This was observational - they compared those who chose to get chemo to those who chose not to get chemo (this wasn't randomized). It is just as likely that the people who choose chemo (and thus make a statement to go down "fighting") have a lower QOL because of the type of people they are, not because of the chemo. This could be all treatment selection bias.

My hypothesis - people who chose to stop chemo months before their death approach the whole concept of death differently and generally have a higher QOL.

Shirley Otis-Green said...

Thanks for this thoughtful post and call for more (and earlier) conversations regarding patient's end-of-life care. As suggested, a nuanced conversation in such a situation is challenging, in part, because our words carry often unintended meaning. I noted that the author described the study with this line: "...312 patients with progressive metastatic cancer with a life expectancy of ≤6 months who had failed at least 1 prior chemotherapy regimen..." and wanted to highlight the unintended language that implies "the patient failed chemotherapy" (vs the "chemotherapy failed the patient").
As clinicians, we need to be mindful of how our words have power and may convey unintended meanings...

Shirley Otis-Green, LCSW (

Holly Prigerson said...

The hypothesis that people who stop chemo have higher QOL was disproved by the data in the JAMA Onc report. In fact, those who got chemo not only had significantly higher performance status, mood, optimism but also QOL. This bias was used in the propensity weighted analysis that neutralized the differences between the groups and the observed association with worse QOL was still found.

DannyMD said...

Shirley -- thank you for your mindfulness. I corrected my unintended language, but your comment is a useful reminder of how often we use language that negatively affects patients.

Dan -- please see Holly Prigerson's (lead author on paper) response to your post, as this was taken into account.

I look forward to more studies that examine QOL in this setting, specifically those that are able to serially follow QOL over the entire arc of treatment rather than just in the last week of life.

Dan Matlock said...

Thanks for this string. Don't get me wrong, it is an excellent paper. The problem with propensity adjustment is you can only adjust for what you can measure. I'm not sure mood, performance status, or QOL really capture the deeper constructs around acceptance (which I agree we don't quite know how to measure and Holly's PEACE measure is really good). Adjusting for optimism also gets closer to it.

As far as observational work goes, this is the gold standard. I also intuitively agree with the results. I still can't get past the thought there are unmeasured confounders between those who choose chemo and those who do not that may have influenced these results.

This is why the research must continue!

A great paper highlighting the limitations of propensity adjustment. Analysis of Observational Studies in the Presence of Treatment Selection Bias. Thérèse A. Stukel, AMA.J2007;297(3):278-285.

Holly Prigerson said...

Thanks for your thoughtful comments, Dan.

We agree that there is a need to measure things like illness understanding better as well as understanding of treatment options of patients and caregivers. We have done some of this better in Coping with Cancer II and in work under development. We are now proposing to examine a wide variety of end-stage/palliative interventions including the newer treatments, but I am wondering if there is something missing that should be included so that we can measure those unmeasured factors better than we currently do. Given the ethical and logistical challenges of an RCT, observational studies of this may be the best we can do. We have measures of cognitive and emotional acceptance of relevant constructs, but is there something else you believe is important to add to the "equation" so that these observational studies can better account for an important influence on these outcomes?

For the record, the data in our BMJ 2014 report did factor into analyses of palliative chemotherapy outcomes and after adjusting for them (which I hear you do not have much confidence in) they do not explain away our observed effects on quality of care (or life) in the patient's final week:

Sociodemographic characteristics—Patients reported sex, age, marital status, race/ethnicity, religion, health insurance coverage, and last grade completed in school.

Health status measures—We measured patients’ functional status with the Karnofsky score and Eastern Cooperative Oncology Group (ECOG) performance status, comorbid medical conditions with the Charlson Comorbidity Index, and quality of life with the McGill Quality of Life Index.

Treatment preferences, terminal illness acknowledgment, coping, communication, and advance care planning—We used validated measures to elicit patients’ preferences for chemotherapy, mechanical ventilation, admission to an intensive care unit near death, and life extending care versus comfort care. We also assessed patients’ acknowledgment of terminal illness, as this has been associated with higher rates of advance care planning and hospice use.23 The Brief COPE and RCOPE surveys assessed active, emotion focused, maladaptive, and positive religious coping with cancer related stress. We also asked patients whether they recalled having an end of life discussion with their physician, had a strong therapeutic alliance with their providers, and had completed a living will, durable power of attorney, or do not resuscitate order

Dan Matlock said...

I'm excited to hear about this next phase of work. I think exploring the acceptance measures and coping measures will be really interesting.

In thinking about other measures:

We've also been starting to work with Terror Management Theory. In short, the theory basically argues that much of human behavior is related to fear of death. Much of this literature is based on college kids and shows that when college kids are exposed to death (a mortality salience exercise), they are more likely to do things to avoid death while the death thoughts are in the conscious. However, after a period of time, those thoughts move to the subconscious and then they are more likely to do YOLO type behaviors. Very little of this work has been done among people who are actually facing illness. We've been exploring how we might explore these hypotheses.

Also, we're considering a longitudinal qualitative study of people with advanced illness (perhaps a diary method if it is feasible) to see if the qualitative findings can help with the interpretation of the quantitative findings.

We're also working on a conceptual model around acceptance of death but that is also proving to be really difficult.