The Median Is Not the (Only) Message

When Stephen Jay Gould developed a particularly aggressive form of mesothelioma in 1982, his review of the medical literature led him to conclude—prematurely, as it turned out—that he had only 8 months to live. In a popular article published 3 years later, Gould famously reminded his fellow patients that “the median is not the message” (1), and he argued that median survival is ambiguous at best and is both misleading and discouraging at worst. As increasingly sophisticated predictive models have advanced the science of prognostication, Gould's warning has become even more relevant. In addition, it is a particularly useful background to 2 important articles in this issue (2, 3).

The article in this issue by Gross and colleagues (2) on screening for colon cancer offers an important contribution to previous research (4, 5). The authors show that life expectancy after a diagnosis of colorectal cancer depends on the patient's age and comorbid diseases at the time of diagnosis. Gross and colleagues' study, like Walter and Covinsky's and Welch and colleagues' earlier works (4, 5), provides a way to make individualized predictions of survival, which is a critical component of patient-specific screening decisions.

However, if it is clear that physicians should consider comorbid conditions in decisions about screening, how they should integrate these data into discussions with patients is less clear. This is the broader question that underlies Gross and colleagues' study and future research in this area. The point is not that the median survival isn't the message, …