Iron absorption prediction equations lack agreement and underestimate iron absorption

A number of algorithms have been developed to predict the bioavailability of iron from mixed meals and diets, but their direct validity in predicting change in iron status remains questionable. Throughout the course of conducting a large feeding trial in 10 convents in Manila, we collected weighed food intake data (n = 317) and directly compared the performance of these prediction equations to each other and to the change in serum ferritin (SF). Dietary weighed food intakes were measured on d 3 every 2 wk for each woman and iron status determined at baseline, 4.5 mo, and 9 mo. The Monsen and Balintfy equation predicted higher median absorption efficiency (7.3%) than did the equations of Hallberg and Hulthen (6.1%) and Reddy et al. (5.8%). In contrast, the predictions that used the equations of Bhargava et al. (3.8%), Tseng et al. (2.9%), and Du et al. (2.6%) were significantly lower. The iron absorption efficiencies calculated using the Monsen and Balintfy equation correlated with those using the Hallberg and Hulthen equation (r = 0.91, P < 0.001). This slope did not differ from unity, whereas all other equations underestimated iron absorption efficiency relative to Monsen and Balintfy's equation. The median efficiency of absorption, based on change in SF in 114 subjects, was 17.2%, suggesting that these equations underestimate iron absorption. The inhibitory and enhancing factors in the published prediction equations were quantitatively either too large or perhaps too small to correctly predict apparent iron bioavailability over a 9-mo period. The causes of the lack of agreement between change in iron status estimated by SF change and absorption predicted by algorithms are open to discussion and will need to be resolved.