Tuesday, October 7, 2008
JAMA (huzzah) has published what will likely be remembered as the core publication out of the Coping with Cancer Study (CWCS)(see our posts here, here, and here for more). This analysis looks at the association between end-of-life (EOL) discussions and several outcomes including patient mental health and aspects of EOL care. Briefly: the CWCS was a prospective, multi-institutional (several cancer centers in Northeast US and Texas) cohort study of advanced cancer patients (metastatic; 'failed' first line chemotherapy) and their family caregivers. It involved an intensive psychosocial assessment including disease understanding at baseline; patients were followed over time and bereavement outcomes were measured in family members of those who died (most ~6 months post-loss). This paper is based on the ~320 who died by the time the analysis was done (total enrollment was a little over 600): median time from enrollment to death was 4.4 months.
This study compared outcomes in those who reported having a discussion about EOL issues at baseline vs those who didn't. The presence of an EOL discussion was defined by patients answering positively to the question: Have you and your doctor discussed any particular wishes you have about the care you would want to receive if you were dying? They used 'propensity score weighted' adjustments to balance differences between groups: I can't say that I understand exactly what this means - it seems to be different from the use of propensity scores I discussed previously in the blog. Their use here seems to be, and I trust any statisticians out there will forgive me for saying this, a propensity-score enhanced regression model to adjust for baseline differences between groups. Short story: they used linear and logistic regression models to try to control for underlying differences in the +EOL conversation group and the -EOL conversation group.
37% of these patients (who had a median 4.4 month prognosis!) answered they had EOL discussions at baseline. The differences at baseline, for which they adjusted, were: performance status, symptom burden, survival time, and recruitment site. About recruitment site: rates of EOL conversations varied from 16% to 62% depending on which cancer center you were at. Even if you believe the rest of the results here are bogus that wide variation in patient care (because this is a patient care issue) is astounding and cause for alarm.
Anyway - reporting an EOL conversation at baseline was associated with major differences (improvements) in EOL care and bereavement and no significant differences in adverse mental health outcomes. Rates of depressive and anxiety disorders, PTSD, subscale scores looking at depressive and anxiety symptoms, etc. were the same between groups, even after adjustment. And the results keep on getting better: in another analysis in which they adjusted further for baseline treatment preferences, terminal illness acceptance, and self-reported desire for prognostic information, rates of ICU admission/ventilator use/resuscitation prior to death were lower and rates of hospice length of stays over 1 week were higher in the baseline +EOL conversation group. These were pretty substantial differences in the relative rates (e.g. 11 vs 1.6% for ventilator use; 66% vs 45% for longer hospice stays, etc.). Ok, I guess that last one is a pretty good difference in absolute rate as well - 20%.
These differences in care at the end of life were associated with substantial changes in bereaved caregivers ratings about patients' quality of life (better in those who had less 'aggressive' interventions and who had longer hospice lengths of stay). Bereaved caregivers of patients with more aggressive interventions reported more depression, regret, and other unwelcome outcomes.
Because I like the outcomes I am trying to be cautious in interpreting them. The two big questions that CWCS is trying to answer is 1) does having these conversations hurt our patients by causing emotional harm, and 2) does having these conversations help our patients and their loved ones by improving the care they receive as they die? And these findings overwhelmingly show that patients who report having these conversations are no worse off than those who don't report such conversations and the care received by these patients as they die is measurably different/better and these differences are associated with improvements in outcomes in loved ones. One could also argue that it's not just having these conversations (because one assumes that most patients had an EOL talk prior to their death if not at baseline) but that it's having these conversations relatively early that made the difference.
I think the major caveat here, which is one that can't be overcome without a randomized trial (which will never happen, and this sort of study design will likely be the best level of evidence we'll get on this topic), is of course confounding from unmeasured patient characteristics. Two possible sources immediately come to mind 1) patients who willingly report talking about dying with their docs to a relative stranger (a member of the research team) vs. those who don't (e.g. what is actually being measured is not rates of EOL conversations per se but patients' willingness to talk about those conversations, or something along those lines), and 2) patients who reported EOL conversations were more likely to want to have them in the first place and, again, the 'real' difference between the groups is something along the lines of innate patient characteristics regarding openness to EOL discussions, talking about the future, etc. You could imagine that these 'open' patients were more amenable to less aggressive care as they died, may have had more open/sharing relationships with their caregivers which resulted in better bereavement outcomes, etc. The other issue is that they may be measuring a broader pattern of physician behavior, and this reporting of an EOL discussion (or not) really reflects substantial differences in overall physician care for their patients, being more or less likely to offer/recommend aggressive care, etc. and that "-EOL conversation" is a proxy here for, bluntly, bad end of life care.
Reflecting on how concerned I am about confounding, I'm asking myself how would I have responded to this if they found that early EOL conversations had no effect on EOL care (would have believed it, and wondered if the measurable differences were too small to be appreciated with an N of 320) or emotionally 'harmed' patients such as by increasing levels of anxiousness (I think I would have believed that too, although it wouldn't have led me to conclude we shouldn't do it but look more into how it's done, etc.). Given that, the real benefits which seem to be associated with such early conversations, and the disappointing rates of these conversations and disparity across treatment centers, I think that the CWCS has given real backbone to the idea that 'early' EOL discussions should be a standard of care for advanced cancer patients, and that this is not some nice fantasy of self-righteous palliative care people who think that talking about dying is 'just the right thing to do' but something backed by good data.
It's not a 'moral' question, but a clinical one of ensuring the best possible outcome for our sickest patients.
Alexi A. Wright, Baohui Zhang, Alaka Ray, et al (2008). Associations Between End-of-Life Discussions, Patient Mental Health, Medical Care Near Death, and Caregiver Bereavement Adjustment. JAMA, 300 (14), 1665-1673