The headline: “Cheap blood test detects pancreatic cancer before it spreads”.

The reality:

The nanosensor correctly identified healthy individuals 98% of the time, and identified people with pancreatic cancer with 73% accuracy. It always distinguished between individuals with cancer and those with other pancreatic diseases.

The 98% number means that two out of 100 healthy people who take the test would have a false positive result. It also misses more than 20 out of 100 people with cancer, giving them a false sense of security. If used in a mostly healthy population — a reasonable assumption to make for a screening test — a positive result would more likely than not be a false positive, and yet you would still miss plenty of actual cancers.

These are abysmal assay characteristics and the test should never see clinic, but you would never know it from the headline. (↬Tyler Cowen)