Assessment of iron deficiency in US preschool children and nonpregnant females of childbearing age: National Health and Nutrition Examination Survey 2003-2006

Background: A new index to determine body iron promises a simpler approach to monitoring iron deficiency (ID) prevalence. Objective: Our objective was to compare ID defined as body iron, <0 mg/kg and calculated from the log ratio of transferrin receptor to ferritin (the body iron model) to ID defined as >= 2 of 3 abnormal concentrations in ferritin, transferrin saturation, or erythrocyte protoporphyrin (the ferritin model). Design: We used measures of iron status and inflammation from 486 children aged 1-2 y, 848 children aged 3-5 y, and 3742 nonpregnant females aged 12-49 y from the National Health and Nutrition Examination Survey 2003-2006. Results: ID prevalences (+/- SE) based on the body iron model in children (1-2 and 3-5 y) and in females (12-19 and 20-49 y) were 14.4 +/- 1.9%, 3.7 +/- 0.8%, 9.3 +/- 1.0%, and 9.2 +/- 1.6%, respectively. ID prevalences based on the ferritin model in children (3-5 y) and females (12-19 and 20-49 y) were 4.5 +/- 0.9%, 15.6 +/- 1.2%, and 15.7 +/- 0.8%, respectively. The kappa statistics for agreement between the 2 models were 0.5-0.7. Among females (12-49 y) the positive predictive values of ID based on the body iron model and the ferritin model for identifying anemia were 43 +/- 3% and 30 +/- 2%, respectively, whereas negative predictive values did not differ. C-reactive protein was elevated in 28.8 +/- 3.1% of females with ID by the ferritin model but not by the body iron model and in 0% of persons with ID by the body iron model but not by the ferritin model. Conclusions: The agreement between the 2 indexes was fair to good. Among females, the body iron model produced lower estimates of ID prevalence, better predicted anemia, and appeared to be less affected by inflammation than the ferritin model. Am J Clin Nutr 2009;89:1334-42.