Iron bioavailability from maize and beans: a comparison of human measurements with Caco-2 cell and algorithm predictions

Background: An in vitro digestion and Caco-2 cell model may predict iron bioavailability to humans; however, direct comparisons are lacking. Objective: The objective was to test the differences in iron bioavailability between 2 maize varieties and 2 bean varieties (white beans and colored beans) by comparing human, Caco-2, and algorithm results. Design: Two randomized, 2 x 2 factorial experiments compared women's iron absorption from 2 maize varieties (ACR and TZB; n = 26) and 2 bean varieties (great northern and pinto; n = 13), each fed with and without ascorbic acid (AA) from orange juice. Nonheme iron bioavailability was determined from 2-wk retention of extrinsic radioiron tracers and was compared with Caco-2 cell and algorithm results from identical meals. Results: Without AA supplementation, women absorbed only about 2% of the iron from the maize or bean meals. The results were unaffected by the variety of either maize or beans. Adding AA (15-20 molar ratios of AA:iron) roughly tripled the iron absorption (P < 0.0001) from all test meals. Although the Caco-2 model predicted a slightly improved bioavailability of iron from ACR maize than from TZB maize (P < 0.05), it accurately predicted relative iron absorption from the maize meals. However, the Caco-2 model inaccurately predicted both a considerable difference between bean varieties (P < 0.0001) and a strong interaction between bean varieties and enhancement by AA (P < 0.0001). The algorithm method was more qualitatively than quantitatively useful and requires further development to accurately predict the influence of polyphenols on iron absorption. Conclusions: Caco-2 predictions confirmed human iron absorption results for maize meals but not for bean meals, and algorithm predictions were only qualitatively predictive.