点击显示 收起
1 From the Institute of Internal Medicine, the Department of Clinical Nutrition, the University of Göteborg, Annedalsklinikerna, Sahlgrenska University Hospital, Göteborg, Sweden.
2 Supported by the Swedish Medical Research Council (project B94-19X-04721-19A), the Swedish Council for Forestry and Agriculture Research (50.0120/95, 997/881, and 113:3), and the Swedish Dairy Association. 3 Reprints not available. Address correspondence to L Hallberg, Department of Clinical Nutrition, University of Göteborg, Annedalsklinikerna, Sahlgrenska University Hospital, S-41345 Göteborg, Sweden.
ABSTRACT
Background: Dietary iron absorption from a meal is determined by iron status, heme- and nonheme-iron contents, and amounts of various dietary factors that influence iron absorption. Limited information is available about the net effect of these factors.
Objective: The objective was to develop an algorithm for predicting the effects of factors known to influence heme- and nonheme-iron absorption from meals and diets.
Design: The basis for the algorithm was the absorption of iron from a wheat roll (22.1 ± 0.18%) containing no known inhibitors or enhancers of iron absorption and adjusted to a reference dose absorption of 40%. This basal absorption was multiplied by the expected effect of different amounts of dietary factors known to influence iron absorption: phytate, polyphenols, ascorbic acid, meat, fish and seafood, calcium, egg, soy protein, and alcohol. For each factor, an equation describing the dose-effect relation was developed. Special considerations were made for interactions between individual factors.
Results: Good agreement was seen when measurements of iron absorption from 24 complete meals were compared with results from use of the algorithm (r2 = 0.987) and when mean iron absorption in 31 subjects served a varied whole diet labeled with heme- and nonheme-iron tracers over a period of 5 d was compared with the mean total iron absorption calculated by using the algorithm (P = 0.958).
Conclusions: This algorithm has several applications. It can be used to predict iron absorption from various diets, to estimate the effects expected by dietary modification, and to translate physiologic into dietary iron requirements from different types of diets.
Key Words: Humans • iron absorption • heme iron • nonheme iron • algorithm • diet • meals • bioavailability • iron status • iron requirements • phytate • polyphenols • ascorbic acid • meat • soy protein • alcohol • eggs • calcium
INTRODUCTION
Knowledge about the absorption of iron from the diet and about factors influencing absorption has increased considerably since the extrinsic tag was introduced to label dietary iron in meals (1, 2). The amount of iron absorbed from a meal is determined by iron status, the content of heme and nonheme iron, and the bioavailability of the 2 kinds of iron, which in turn is determined by the balance between dietary factors enhancing and inhibiting the absorption of iron, especially nonheme iron (3). It is well known that the variation in dietary iron absorption from meals is due more to differences in the bioavailability of the iron, which can lead to a >10-fold variation in iron absorption, than to a variation in iron content.
Therefore, several attempts have been made to devise algorithms to estimate the bioavailability of the dietary iron content of meals. The aim of the first attempt was to illustrate the fact that the composition of meals greatly influences the absorption of dietary nonheme iron (4). Later, attempts were made to improve the algorithm (5, 6). A simpler method using a score system to estimate the expected bioavailability of dietary nonheme iron was also suggested (7). In this model, factors inhibiting iron absorption were also considered.
Several dietary factors (eg, ascorbic acid, meat, fish, and poultry) enhance iron absorption, whereas other factors [eg, inositol phosphates (phytate), calcium, and certain structures in polyphenols] inhibit iron absorption. In the present study, we analyzed the dose-response relation between amounts of these factors and their effects on nonheme-iron absorption. All of these factors must be considered in an algorithm to predict the amount of iron absorbed from a meal. For almost all of the factors, it has been possible to develop continuous functions related to the amounts of each in the meal. Moreover, interactions between different factors have been examined and considered.
The hypothesis tested in the present algorithm was that the bioavailability of iron in a meal is a product of all factors present in the meal that inhibit or enhance iron absorption. A starting point for the present work was to find a food or meal that contained no known inhibiting or enhancing components and then use this food as a basis for evaluating the effects of different factors added in different amounts. For many years we used, as a control, wheat rolls made of low-extraction wheat flour and fermented to such an extent that no inositol phosphates could be detected. Various factors to be tested were added in different amounts to such rolls and iron absorption was measured from the rolls, when served with or without a specific factor in known and various amounts, after the rolls were labeled with 2 different radioiron isotopes. Iron status in each fasted subject was measured by using the absorption from a standard reference dose of ferrous iron to describe the iron status of the individuals studied. The reference dose was introduced by Layrisse et al (8) and the entire procedure was described in detail (9). Iron absorption can also be related to log serum ferritin as suggested by Cook et al (10).
Numerous studies on factors influencing the bioavailability of dietary iron have been published by several research groups (discussed below), in addition to the studies by our group. It has only been possible, however, to use some of the data from their studies. This is true also for some of the older data from our laboratory. The reason is simply that there is a lack of information about the content of phytate and sometimes that of polyphenols in the meals studied.
METHODS
The method used to predict dietary iron absorption is based on an algorithm containing the value for iron absorption (relative to 40% of the absorption of the reference dose of iron) from a single basal meal ([low-extraction (40%) wheat flour] that contained no components known to inhibit or enhance iron absorption. This basal value was then multiplied by factors expressing the effect of different dietary components present in the meal known to influence iron absorption: phytate, polyphenols, soy protein, calcium, eggs, ascorbic acid, meat (including fish and seafood), and alcohol. For each factor, an equation was derived that also considered interactions between components in the meal.
Iron absorption from a basal meal
The basal meal was composed of wheat rolls served with margarine and water on 2 mornings while subjects were in a fasting state. The rolls were made of a special low-extraction (40%) wheat flour and the dough was fermented for 2 periods (30 + 10 min) to ensure that no inositol phosphates could be detected with a sensitive method (11). The iron content of the rolls was adjusted to 4.1 mg by adding ferrous sulfate to the dough. The rolls were labeled with an extrinsic radioiron tracer. Iron absorption was measured as described previously (9, 12).
The rolls were included in different studies of factors influencing iron absorption. Rolls were served with and without a factor to be studied in specific amounts and were labeled with 2 different radioiron isotopes (13–15). Iron absorption from these rolls was measured in 310 subjects (194 female and 116 male volunteers). In each subject, iron absorption from a reference dose containing 3 mg Fe as ferrous sulfate, given while subjects were in a fasting state on 2 consecutive mornings, was also measured. All absorption values were adjusted to correspond to an absorption of 40% from the reference dose. Thus, absorption measurements from the same meal could be pooled from different groups of subjects with different iron statuses. The mean (±SEM) absorption of iron from the rolls in all studies, adjusted to a 40% reference dose absorption, was 22.1 ± 0.18%.
Effect of phytate and other inositol phosphates
The effect of different amounts of phytate on iron absorption was examined when wheat rolls were served with and without different amounts of added sodium phytate. Seven groups of subjects (n = 63) were studied and the added phosphorus as phytate (phytate-P) varied from 2 to 250 mg (14). A similar study was performed in another laboratory in which the basal wheat rolls contained 10 mg phytate-P (n = 57). Four different amounts of phytate-P (14–58 mg) were added (16). Because the effect of 10 mg phytate-P was examined in the previous study, it was possible to recalculate the effect of the added phytate-P. The effect of phytate was similar in the 2 studies. When the data from the 2 studies were pooled, the following relation was found:
DISCUSSION
It has been nearly 20 y since the first simple algorithm for estimating iron absorption was published (4). Since then, much new knowledge has accumulated about dietary iron absorption, as emphasized in a recent review (52). It is thus probable that new information will lead to modifications of the present algorithm. Instead of waiting for the "final version," we developed an algorithm based on as much present knowledge as possible and we think the present algorithm has many practical applications.
Note that the method of measuring iron absorption from the whole diet with tracers has been validated. In each subject, a comparison was made between the absorption measured and iron requirements. In men, requirements were calculated from body weight and in women from body weight and measured menstrual losses of iron (53). The comparison in study 1 clearly showed that iron absorption estimated with the algorithm agreed well with measured iron absorption.
Nonheme-iron absorption was estimated for the 24 meals in study 1 by using the 2 previously published algorithms, in which effects of both enhancers and inhibitors were included. In the earliest study (7), there was a significant relation between observed and estimated absorption (r2 = 0.192, P = 0.032). There was also a significant relation between observed and estimated nonheme-iron absorption (r2 = 0.256, P = 0.0116) when a more recent algorithm was used (6). These correlation coefficients are thus considerably lower than that obtained with the present algorithm (r2 = 0.987) for estimated and observed nonheme-iron absorption. Probable reasons are that, in contrast with the 2 previous algorithms mentioned, the present algorithm 1) is based on continuous variables for content of enhancers and inhibitors, 2) takes into consideration interactions between factors, and 3) includes more factors. In study 2, the same mean heme- and nonheme-iron absorption values were seen despite the expected markedly varying bioavailability of iron in the 20 different meals included (Table 3).
An important difference between the 2 validation studies was that each absorption value in study 1 was the mean of 10 subjects (observed and calculated by using the algorithm; Table 1), whereas each absorption value in study 2 was the mean of 31 subjects measured over 5 d (Table 4). In study 1 the slope of the regression line did not differ from the identity line and there were no statistically significant differences between observed absorption and absorption estimated by using the algorithm at the same iron status. In study 2, the total amounts of observed and calculated (algorithm) iron absorbed from the whole diet were not significantly different after adjustment to the same iron status (Table 4).
Effect of meal size and iron content of meals
It may seem obvious that the size of a meal should be taken into account in an algorithm for estimating iron absorption. A certain amount of ascorbic acid, for example, should be expected to have a greater effect in a small meal than in a large meal because the concentration would be higher in the small meal. Meal size, however, is an ambiguous concept because it can be interpreted in terms of volume, weight, or content of energy or iron. The concentration of a nutrient may also be influenced by the amount of beverage consumed with the meal. Another factor that can influence absorption is the rate of gastric emptying and, in turn, the volume of the meal and its fat content. Meal size as well as body size can influence the absorption of iron from a specific meal; however, we did not observe either in our adult volunteers.
There was almost a 4-fold variation in the content of both energy and iron and a 3-fold variation in nutrient density (nonheme iron/energy) in the meals in study 1 (Table 1). Despite these variations, the relation between calculated and observed absorption was more similar than we had expected. Thus, the balance of evidence indicates that meal size per se had no major systematic effect on the validity of the algorithm. The algorithm may need modification when used in infants and small children. In a recent study, however, direct comparison of iron absorption from a formula given to adults and infants showed no difference in absorption (54). Moreover, a 3-fold increase in meal size (and iron content) in adults did not change fractional iron absorption (54). It is thus reasonable to assume that the algorithm will also be useful in infants.
A linear relation was observed between log amounts of iron administered and log amounts of iron absorbed (see references 55 and 56 for a review). Most of these studies used therapeutic doses of iron or pure iron solutions; iron given with food seems to behave differently. In one of our early studies, we found that the percentage iron absorption from a meal was the same despite an almost 5-fold difference in iron content (57). This result is thus compatible with the results mentioned above (54), probably because the concentration of iron in the gastrointestinal lumen is many times lower when a certain amount of iron is present in a meal than when the iron is provided as a salt without food.
Iron absorption from single meals compared with that from the whole diet
To estimate iron absorption from the whole diet, absorption measurements from all the single meals consumed over a certain time period are summed. Almost all studies of factors influencing iron absorption are based on single meals served in a fasting state, with and without a factor to be studied given in different amounts. Note that direct measurements have shown that a preceding meal has no effect on the absorption of iron from a subsequent meal. In studies of 4 diets, it was shown that iron absorption from a meal served in the morning after an overnight fast was the same as that from a meal eaten during the day at lunch or supper (58). Similarly, we found that iron absorption was the same from a hamburger meal served in the morning or after breakfast (with or without added calcium) 2 or 4 h earlier (59).
It has been suggested that the variation in iron absorption from single meals under laboratory conditions would exaggerate the variation in iron absorption from the whole diet (7). The variation in iron absorption between single meals of different compositions may be much greater than the variation in iron absorption from whole diets composed of several single meals. The iron content and bioavailability of single meals varies markedly, whereas iron absorption from whole diets is the mean absorption of several single meals. The expected lower variation in iron absorption from the whole diet than from single meals was documented previously (7) and in the 3 studies of iron absorption from whole diets in our laboratory (46, 50, 51).
Some investigators seem to have misinterpreted these results and assumed that the absorption of iron from single meals per se, for some unknown reason, would be falsely high or low. The present result that the sum of the calculated iron absorption from 4 different meals served for 5 d (ie, 20 meals in 31 men for a total of 620 meals) did not differ significantly from that obtained from meals in which heme and nonheme iron were homogenously labeled with 2 different tracers, clearly indicates the validity of basing total dietary iron absorption on the sum of iron absorption from single meals. This issue was also discussed in our previous review (53).
Some applications of the algorithm
The algorithm can be used to evaluate the nutritional value of meals with respect to iron, for example, in school-lunch programs, in catering programs for the elderly, and for military services. The algorithm may be used to translate data from dietary surveys into amounts of iron expected to be absorbed. The main requirement for such calculations is that detailed information is available about the meal composition and its variation over a representative and sufficiently long period of time. A 7-d record, for example, may not represent the iron absorption from the habitual diet.
The algorithm can be used to estimate the expected effects of different dietary modifications that can be considered realistic in both developed and developing countries. In developed countries, the main concerns are low energy expenditure and, thus, low energy intakes. To adequately provide for high iron needs, especially in infants, adolescents, and menstruating women, a high nutrient density and a high bioavailability is required. The algorithm can also be used to examine the overall effects of a higher extraction of flour (increasing the intake of both intrinsic phytate and iron) on bioavailability and iron content. It can be used to estimate the expected effects of iron fortification or increased intakes of fruit, vegetables, and meat in the diet. In developing countries, the problems are similar but the knowledge about the chemical composition of foods and its variation is even more limited; for example, knowledge is limited about the contents of phytate and iron binding polyphenols in common foods, including spices and condiments. Evaluation of the expected effects on iron absorption and iron balance resulting from modification of food-preparation methods may also be required.
An important use of the algorithm would be to translate physiologic iron requirements into dietary requirements under different dietary conditions known to prevail in a certain population. In the Food and Agriculture Organization/World Health Organization recommendations, 3 levels of bioavailability (5%, 10%, and 15%) were used arbitrarily for this translation (60). The validity of choices of representative bioavailability values can be tested by using the algorithm. It is obvious from the present results that there is marked variation in the bioavailability of different types of diets in developed countries. The recommended dietary allowances (61) for different groups of subjects with different physiologic iron requirements should, therefore, not be given as single values, but rather as 3–4 values adjusted for different types of diets (eg, vegan or vegetarian, low-meat, and high meat). The algorithm can then be used to make rough estimates of the bioavailability of diets in some groups in the population with different dietary habits. The algorithm may be useful in the future search for realistic recommendations to be used in food-based strategies to improve iron nutrition in developing countries. However, more knowledge about the composition and properties of diets in developing countries is needed.
In the screening for unknown dietary factors influencing iron absorption, new starting points can be obtained by comparing actual absorption values from a certain meal with absorption values estimated from the content of presently known factors. A significant discrepancy would indicate that some unknown nutritionally important factor is present.
Importance of correct values for the factors included in the algorithm
One problem with the application of the algorithm is limited knowledge about the content of factors such as phytate and iron binding polyphenols in different foods. An extensive report on the phytate content in foods was published previously (62). Note that even low phytate contents play an important role in the bioavailability of iron, but are often not detectable with the current method used by the Association of Official Analytical Chemists, which was used in that report. A simple modification of the current method of the Association of Official Analytical Chemists was made to determine low phytate contents in foods and was calibrated against HPLC (11).
Another practical problem in applying the algorithm is the difficulty in estimating the ascorbic acid content in a meal at the time of consumption because cooking times and food-preparation methods markedly influence the final phytate content. In Appendix A, we provide data for some common foods. Appendix A also contains data from our laboratory about the content of total iron and heme iron in different kinds of meat. More detailed food-composition tables are needed. The lack of knowledge of the presence of different factors in different foods is even more obvious when the algorithm is applied to diets in developing countries.
APPENDIX A
Observed iron absorption at a certain iron status adjusted to the absorption expected at another iron status
The relation between log serum ferritin and the reference dose absorption was examined in 1066 subjects in whom an adequate serum ferritin standard had been used in the calibration. In these studies we found that a reference dose of 40% corresponded to a serum ferritin concentration of 23 µg/L. Data from a recent study of iron absorption from whole diets in men (n = 31) showed that there is a linear relation between log iron absorption and log serum ferritin (SF) (r2 = 0.720):
REFERENCES