Shabbir M.H. Alibhai, MD, MSc, FRCPC
Geriatrics & Aging
Recently I saw a 72-year-old woman with metastatic uterine cancer in consultation. She was receiving palliative chemotherapy and our team felt she would benefit from short-term inpatient rehabilitation. Within a week of this consult, I saw an 81-year-old woman with an acute myocardial infarction. My team elected to start her on lipid-lowering therapy.
What do these two patients have to do with one another, and what does the latter patient have to do with this issue's theme of cancer? It is, simply put, a matter of life and death.
Let me explain. Many readers know that I do research in prostate cancer. In my editorial in last year's cancer issue,1 I discussed the importance of assessing the impact of treatment on both length and quality of remaining life. Whenever clinicians make treatment decisions, we explicitly or (more commonly) implicitly consider life expectancy. For both patients that I alluded to, estimating life expectancy was fundamental to the management decision.
Several randomized trials have demonstrated improvement in survival for certain malignancies, including some gynaecologic tumours, from "palliative" (i.e. non-curative) chemotherapy. The question that clinicians and patients must struggle with is whether the expected improvement in survival (for example, two to six months) is clinically important and worth the risks of treatment. Lipid-lowering therapy has been shown in numerous large, well-designed, randomized trials to prolong survival in patients with normal or elevated lipid levels after a myocardial infarction. However, most studies suggest a survival benefit is observable after 12-18 months of therapy. Conversely, lipid-lowering therapy is generally not useful if patients have a life expectancy of less than 12 months. We estimated that our 81-year-old patient would likely survive at least a year after her infarct and thus started her on a statin.
While most of this sounds straightforward, there is a potentially flawed underlying premise--namely, that physicians can accurately estimate life expectancy. How does one do this in theory? Life expectancy can be calculated by taking age-specific annual mortality risks and modifying them by factoring in risks of dying from specific illnesses from which the patient is suffering. For example, my 81-year-old woman post infarct has an average one year risk of 9.4% of dying because of her age. This number is obtained from actuarial life tables.2 Depending on her cardiac status post infarct (left ventricular function, presence of heart failure, deficits on cardiac perfusion study), she might have a 20% risk of dying from her coronary artery disease. Thus her overall risk of dying would be some combination of these two risks, assuming she is otherwise well.
So how good are we at estimating life expectancy? It is interesting to note that there are very few published studies on this subject. Yet in prostate cancer, for example, clinicians are encouraged to explicitly consider life expectancy when making treatment decisions--the so-called ten-year rule.3 This rule suggests that if patients have ten or more years of life left, excluding their prostate cancer, then curative therapy is indicated, and vice versa. There may be similar rules in other areas of cancer. Beyond these guidelines, there is little published data to guide physicians when estimating remaining life.
Two studies from the field of palliative care are provocative. The more recent one, published last year,4 asked a group of palliative care doctors in Chicago to estimate remaining life expectancy for 504 patients newly admitted to a hospice program. Patients were followed until they died. Patients were relatively old (mean age 69) and had advanced disease (median survival 24 days). If accuracy was defined as being within 0.5 and 2.0 times the actual survival, physicians were accurate 34% of the time, optimistic 55% of the time, and pessimistic 11% of the time. Even when we consider fairly terminal patients, physicians do not perform as well as expected.
This does not mean we should forget about estimating life expectancy. Rather, it behooves us to perform more research in this area to identify better models for predicting survival. Furthermore, we must realize that disease severity and comorbid illnesses are far more powerful predictors of life expectancy than is chronological age. Until we are far better at estimating survival, guidelines like the ten-year rule in prostate cancer do little service to ourselves and our patients.
Regular readers will realize that this is the third year we are featuring the theme of cancer in Geriatrics & Aging. Much new information continues to be published about cancer screening, diagnosis, treatment, and prognosis in the elderly. In this issue, we have chosen to focus on acute myelogenous leukemia, bladder cancer, multiple myeloma, and gastro-esophageal malignancies. As always, we hope you enjoy it.
- Alibhai SMH. Oncology & Aging--Bitter Truths and Misguided Paternalism. Geriatrics & Aging 2000; 3(3):3.
- Statistics Canada. Life tables, Canada and provinces. Health Reports 1990; 2 (4(Suppl.13)):17.
- Walsh PA, Partin AW, Epstein JI. Cancer control and quality of life following anatomical radical retropubic prostatectomy: results at 10 years. J Urol 1994; 152:1831-1836.
- Christakis NA, Lamont EB. Extent and determinants of error in doctors' prognoses in terminally ill patients: prospective cohort study. BMJ 2000; 320:469-473.