When median doesn’t mean what it seems

When interpreting brain scans, the eyes can deceive. A numbers dispute over a potential Alzheimer's disease test in Eli Lilly & Co.'s pharmaceutical pipeline broke out in the pages of a prestigious medical journal last week. The test, which holds the potential to diagnose the presence of Alzheimer's-linked protein accumulation in the brain in living patients, has to be interpreted by doctors. Three physicians were enlisted by researchers conducting a Lilly-funded study to put a number on what they saw in scans, and they didn't always see eye to eye. The debate, and the fate of this test, matters greatly to people with dementia and their loved ones. It also sheds light on the use and potential abuse of median, a common statistical concept. While median can be very useful in analyzing and summarizing a large data set, it might obscure the results for individual cases.