C-reactive protein in children: confounding by sex and linear regression modeling.