Statistical model for predicting non-heme iron bioavailability from vegetarian meals.

Availability of non-heme iron has been extensively discussed when meals comprise heme as well as non-heme iron, but seldom so for exclusively vegetarian meals. The present study aimed to develop a statistical model for predicting non-heme iron availability from a composite vegetarian meal. Radioisotopic measurements of in vitro iron dialyzability of 208 out of 274 meals representing vegetarian diets from Asia, Africa, Europe and Latin America and the meal contents of iron, zinc, copper, ascorbic acid, beta-carotene, riboflavin, thiamin, folic acid, tannic acid, fiber and degraded phytate forms (IP6-IP1) were used for development of the model. A multiple regression model weighted for calorie contents was developed for the percentage iron dialyzability with the possible predictors as meal contents along with plausible interaction terms. The model was validated with in vitro iron dialyzability of 66 meals and in vivo iron absorption in five ileostomized adults. Application of the model was demonstrated using data on the daily dietary intake of 215 young adults whose hemoglobin levels were estimated twice in 3 weeks. Weighted multiple regression model was: ln(% Fe dialyzability)=1.340-0.259xln(IP2 [mg])+0.188xln(IP3 [mg])-0.278xln(IP5 [mg])+0.0912xln(ascorbic acid [mg])+0.06693xln(tannins [mg])+0.09552xln(beta-carotene [microg])+0.137xln(hemicellulose [g]) (P<0.01, R2=0.51). Good agreement was seen between observed and predicted dialyzability (r=0.90) and human absorption (r=0.89). The model would be useful to estimate bioavailable iron intakes of vegetarian populations and to identify at-risk individuals.