A Study of Conditional Median Survival in Advanced Cancer
* This Page is Dedicated to the Memory of Lynn Richards *
Despite the fact that Lynn Richards was fighting for her life against far advanced cancer, in late 2001 she took the time to bring my attention to this important study. It happened to be my birthday which is fitting since I knew as soon as I read her e-mail that it was a gift - she had found something very important, but it was five months before I finally incorporated it into CancerGuide. Unfortunately, Lynn did not live to see it, but if you find hope here, know well that it is a gift from Lynn Richards.
Reference on Conditional Median Survival in Advanced Cancer
Reference & Abstract
Conditional Median Survival Of Patients With Advanced Carcinoma: Surveillance, Epidemiology, And End Results Data
Kato I, Severson RK, Schwartz AG
Cancer 2001 ;92(8):2211-19
Department of Pathology, Karmanos Cancer Institute, Wayne State University, Detroit, Michigan 48201, USA. [email protected]
BACKGROUND: Conditional survival is clinically useful, particularly for patients with malignant disease who have a poor prognosis. However, there are no published data on the conditional median survival of patients with advanced carcinoma on a population basis.
METHODS: Data on 217,573 patients with breast, colorectal, lung, or prostate carcinoma who were newly diagnosed with distant disease between 1973 and 1995 and who were followed through the end of 1997 were extracted from the Surveillance, Epidemiology, and End Results (SEER) data base of the National Cancer Institute. The Kaplan-Meier method was employed to estimate conditional median survival and 95% confidence intervals at 0-5 years after the initial diagnosis.
RESULTS: The conditional median survival increased as time elapsed after the initial diagnosis. The increase was slowest and almost leveled off among patients with prostate carcinoma. The median survival of patients with breast carcinoma increased relatively linearly with time, i.e., 5-6 months per year. Conversely, there was a rapid increase in the conditional median survival according to the amount of time since diagnosis for patients with lung and colorectal carcinoma. The trend was most pronounced for patients with colorectal carcinoma. At 5 years after the initial diagnosis, the remaining median survival was longest for patients with colorectal carcinoma, almost 6 years (71.5 months), followed by patients with lung carcinoma (52.5 months), breast carcinoma (42.5 months), and prostate carcinoma (34.5 months). Although race was a correlate with initial survival, gender and age had more impact on late conditional survival.
CONCLUSIONS: The conditional median survival provides useful and encouraging information for patients who survive with advanced disease and for healthcare professionals who treat these patients. However, the information should be used carefully, taking the limitations of these data into account.
Commentary on Kato by Steve Dunn
This paper reviewed how the prognosis of people with common cancers who had metastatic disease at the time of their initial diagnosis changed as a result of their continued survival. The investigators discovered that the longer a patient survived, the longer her or she could expect to survive. For people who were fortunate enough to have long term survival, the increase in expected further survival was striking. This increase in expected survival was not bestowed in a single lump when patients reached the mythical five year mark, but rather manifested as a gradual increase in expected survival over time. In other words, from a statistical point of view, every day you live, the expected time you have left increases. For breast and prostate cancer there was a slow increase in the length of expected survival, and for lung and colorectal cancer the increase accelerated after about two years.
You should understand that median survival for metastatic lung and colorectal cancer is well less than a year. With that in mind, the increase in the median additional time that long term survivors of these diseases could expect is nothing less than spectacular, on the order of ten times the median these patients faced at the time of their diagnosis!
Although this paper didn’t examine the conditional survival for all cancers, the fact that increasing expected survival over time was found in all four cancers examined suggests that this is likely to be true for a variety of other cancers as well. It’s also worth noting that it can be hard to pick up this decreasing risk over time just by looking at the survival curve. As a result of difficulties in estimating the risk function from the survival curve by eye, this characteristic can apparently hold even when a casual glance at the survival curve appears almost unremittingly grim.
The existence of a small cohort of survivors far past the median in even cancers with a dire prognosis, is yet another reflection of the strong right skew in most survival distributions as explained by Stephen Jay in his wonderful essay, The Median Isn’t the Message. The improving prognosis of these survivors over time shows that the right tail that Gould speaks of is very long indeed!
Finally, on a realistic note, as a percentage long term survivors of these disease were a small minority. But this data still proves to me (and I hope to you) that hope is both real and realistic even when the prognosis appears to be bleak.