Body-mass index and mortality.

摘要

The Prospective Studies Collaboration' states that it did not correct for regression dilution bias "since one BMI [body-mass index] measurement is highly correlated with the long-term usual BMI". However, I and colleagues have found in two separate Cox regression analyses that use of usual BMI rather than one BMI measurement caused substantial increases in the hazard ratio for obesity.These increases were larger than achieved by dividing the regression coefficient for the one BMI measurement by the regression dilution ratio, unless we excluded individuals with serious illness and smokers. The Collaboration did not exclude smokers nor individuals with a history of cancer or emphysema in some of their analyses, suggesting that these analyses might have been affected by regression dilution.