Commentary: over-correction for regression dilution bias? Not for blood pressure vs coronary heart disease.