The Art of Statistics: Learning from Data

by David Spiegelhalter

But much of the improvement seen in people who do not receive any active treatment may be regression-to-the-mean, since patients are enrolled in trials when they are showing symptoms, and many of these would have resolved anyway.