As a statistician, I hear this phrase a lot, and I believe it is mis-understood and over-used. There is a small theoretical possibility that an improvement in an outcome in a clinical trial, observed in a placebo group, could be due to regression to the mean. This would be the case when subjects were selected based on a criteria based on severity of disease. Of course, measurement of severity of disease has measurement error, and it would be the overestimation of the severity due to the measurement error that would cause the regression to the mean effect in this case. In my experience, this effect is tiny compared with psychological or "meaning" effects.