Exploring the Use of Iron and Hematologic Indicators for Surveillance of Iron Deficiency in Pregnant and Nonpregnant Women in the United States.

Abstract

Background: Clinical data, such as electronic health records, may be useful for iron deficiency (ID) surveillance.

Objective: Our objective was to compare iron and hematologic indicators commonly measured in clinical settings to the World Health Organization (WHO)-recommended iron indicator, serum ferritin (SF), to assess ID among a population of adult women aged 20-44 y.

Methods: We evaluated sensitivity, specificity, and area under the receiver operating characteristics (ROC) curve of iron and hematologic indicators commonly measured in clinical settings for ID: hemoglobin (Hb) <120 g/L (nonpregnant), Hb <110 g/L or <105 g/L (pregnant, depending on trimester); mean corpuscular volume (MCV) <80 fL; serum iron <40 μg/dL; total iron binding capacity (TIBC) >400 μg/dL; transferrin saturation (TSAT) <15%, compared with a reference (SF <30 μg/L for pregnant women and inflammation-adjusted SF <15 μg/L for nonpregnant women) using the United States NHANES 2003-2006 and 2017-2018.

Results: Among pregnant women (n = 730), sensitivity ranged from 8.1% (MCV) to 87.2% (TIBC), and specificity ranged from 63.0% (TIBC) to 97.5% (MCV), and area under the ROC curve ranged from 0.553 (MCV) to 0.816 (TIBC). Among nonpregnant women (n = 3345), sensitivity ranged from 32.1% (MCV) to 71.9% (TSAT), specificity ranged from 85.7% (TIBC) to 96.1% (Hb), and area under the ROC curve ranged from 0.785 (MCV) to 0.853 (TSAT). ID prevalence using clinical indicators varied widely compared with prevalence using SF. Among pregnant women, ID prevalence ranged from 6.7% (95% confidence interval [CI]: 3.6, 9.7) (MCV) to 51.8% (95% CI: 46.0, 57.5) (TIBC), compared with 47.0% (95% CI: 39.5, 54.5) using SF. Among nonpregnant women, prevalence of ID using clinical indicators ranged from 6.4% (95% CI: 5.4, 7.3) (MCV) to 24.7% (95% CI: 22.4, 26.9) (TSAT), compared with 15.1% (95% CI: 13.4, 16.6) using SF.

Conclusions: Iron and hematologic indicators routinely used in clinical settings can under- or overestimate ID, suggesting that they are not ideal for ID surveillance.