Iron deficiency continues to be one of the most common nutritional deficiencies in the world. An estimated 30% of the world's population is anemic, with just under half of these  approximately 600 million cases  due to impaired iron status (Carpenter and Mahoney 1992, Cook et al. 1994). Although anemia is its most clearly recognizable sign, iron deficiency can produce other adverse outcomes even before any noticeable drop in hemoglobin concentrations. Specifically, iron deficiency has been linked with decreased immune function and resistance to infection, diminished work capacity, and increased risk of delivery of preterm and low-birth-weight infants. In infants and children, it has also been associated with diminished cognitive development and learning capacity, with effects that may last into adulthood (Cook et al. 1994, National Research Council 1989).

Given the potential consequences of iron deficiency, determination of iron intakes is a useful strategy for assessing the adequacy of the diet to meet iron requirements. The amount of absorbed iron required to replace average daily basal losses in adult men is approximately 1 mg/d; for women, an additional 0.5 mg/d on average is required to replace menstrual loss (National Research Council 1989). Importantly, however, absorption of iron is highly variable, dependent not only on the iron status of the individual, but also on other factors in the diet that enhance or inhibit its absorption. As such, calculating iron availability is as important as determining total iron intakes when evaluating the adequacy of iron in the diet. Indeed, adjustment for bioavailability may provide a more realistic picture of whether or not iron requirements are met in a population than would a simple assessment of total iron intake.

The purpose of this article is threefold: 1 ) to describe the specific dietary information needs for conducting these analyses, 2 ) to present the analytical steps and computer algorithms needed to adjust for iron availability, and 3 ) to demonstrate the extent to which other dietary factors  specifically, meat, fish and poultry (MFP), vitamin C, tea, and phytates  affect calculated absorption of iron in a population at risk. For this last objective, we present results of adjustment for iron bioavailability in a sample of Russian women and children and compare results for urban and rural residents.

SUBJECTS AND METHODS

Data were coded by eating occasion for each person, with each observation representing one food item. Because of the different number of foods eaten per meal per person, the number of observations in the dataset varied for each individual in the study population. This data format was chosen to allow for meal-specific calculations and adjustments for enhancers and inhibitors. A representation of the data structure is provided in Appendix A-1.

Prior to nutrient conversion, the reported dietary intake data were examined for excessive total food intakes, food consumption with zero grams of intake recorded, and extreme intakes of individual foods. Excessive food intakes were either corrected in the file or replaced with median amounts depending on the source of the problem. Intakes of zero grams for any food reportedly consumed were interpreted as being indeterminate and were replaced with the median amount for the food. Finally, to address extreme intakes, the distribution of portion sizes of each food consumed was examined, and any portion size five times greater than the median portion size was replaced with the median amount. This was necessary for 0.5% of the total number of observations.

To adjust for bioavailability, the iron content of each food was first separated into its heme and nonheme components. The assumed proportion of total iron in each food that was nonheme was 60% for meats and 100% for nonmeats; any remaining iron was assumed to be heme. Mixed dishes were separated into meat and nonmeat parts based on official government recipes. Heme and nonheme iron values for each food were then incorporated into the food composition table.

Adjustment of iron availability for the effects of meat, fish and poultry (MFP) and vitamin C consumed in the same meal was performed using methods developed by Monsen et al. (1978) and Monsen and Balintfy (1982) for a reference individual with 500 mg of body iron stores. We felt this to be reasonable because per capita meat consumption was substantial prior to 1990 (Popkin et al. 1997). Heme iron availability was assumed to be 23%; nonheme iron availability was based on the presence of enhancing factors (EF ) in the meal, calculated as the sum of the grams of MFP and milligrams of vitamin C (see Figure 1 and Appendix B.) The assumed proportions on nonheme iron available in the meal after adjustment for enhancing factors ranged from 3 to 8% and are given by the following two formulas: Because the extent of iron stores in this population is not known, however, we also conducted a sensitivity analysis using Round 1 data in women aged 14 to 54 y, based on figures suggested by the FAO/WHO committee (1988) for individuals with less adequate iron stores. In this case, heme iron availability was assumed to be 25%. Nonheme iron availability was determined based on categories of meal-level MFP and vitamin C intakes shown in Table 1.

Table 1. Assumed percent absorption of nonheme iron based on meal-level intakes of meat, fish and poultry and vitamin C1 [View Table]

Table 2. Daily intakes of Russian women and children, 1992-19931 [View Table]

Further adjustments to available nonheme iron were made based on the amounts of tea and phytates consumed in the meal. Nonheme iron availability was reduced by 40% when the amount of tea consumed in the same meal exceeded 225 g (Singer et al. 1982). Phytate values were taken primarily from Harland and Oberleas (1987) and were supplemented with other values from Pennington and Church (1985). A variety of methods was used to determine phytate content of foods; these are described in greater detail by Harland and Oberleas (1987). Phytate values were then incorporated into a recipe file containing quantitative information on phytate-containing foods (see Appendix A-4). Approximate values and adjustments were used when Russian foods could not be closely matched with food items in the literature. For mixed dishes, phytate-containing ingredients were identified based on official government recipes in order to estimate the amount of phytate in the entire mixed dish. The amount of available nonheme iron was adjusted for phytates consumed in the meal using the following formula: (See Appendix B.) The formula was derived by fitting data from Hallberg et al. (1989) into a logarithmic regression model to estimate parameters. The phytate-phosphorous upon which the Hallberg et al. (1989) model was based was translated into phytate by assuming that phytate-phosphorous constitutes 28% of the hexaphosphate inositol molecule. Although the number of phosphate binding sites on the phytate molecule varies, hexaphosphate seems to be the dominant form (Hallberg et al. 1989). We therefore assume that both the phytate content tables and the absorption model from Hallberg et al. (1989) are based on the hexaphosphate inositol molecule.

RESULTS

Table 3. Proportion of total dietary iron available after adjustment for concurrently consumed enhancers and inhibitors [View Table]

Figure 3 shows the estimated iron availability in the study population relative to the physiologic range of daily iron needs without addition of the safety margin inherent in a recommended daily allowance. The entire range of intakes in most cases was lower than the lowest portion of the range of physiologic requirements.

When higher absorption rates were assumed for individuals with less adequate iron stores, a mean 12.0% was bioavailable after adjusting for MFP and vitamin C for women aged 14 to 54 y; mean amount of available iron was 1.55 mg/d, which was still at the bottom of the range of physiologic needs for women of reproductive age. Adjusting for phytates and tea consumed at the same meal, in addition to MFP and vitamin C, however, again resulted in a substantial reduction in available iron; mean amount of available iron following this adjustment was 0.69 mg/d, and mean level of bioavailability was only 5.4%.

Table 4. Total, heme, and bioavailable intakes in rural and urban Russian women and children, 1992-19931 [View Table]

DISCUSSION

Few nutritional epidemiology studies capture the information required for assessment of iron bioavailability  namely, data on dietary intake for each eating occasion and recipe files containing quantitative information for mixed dishes. The importance of this information is evidenced by our analyses showing that only a small proportion of total iron intake was bioavailable for this population of Russian women after adjustment for other factors consumed in the same meal. Estimates of mean percent bioavailability assuming adequate iron stores ranged from 8-11%, when adjusted for MFP and vitamin C, to as low as 3.3-4% following further adjustment for phytates and black tea consumption. Such bioavailability estimates have important implications if they are lower than the proportions of bioavailability assumed when setting recommendations; low bioavailability is especially of concern in developing countries, where diets may contain high levels of inhibitors of iron absorption.

Our research also indicates that iron intakes in this population were lower than both the recommended and the required levels of intake for Russian women of reproductive age, although the extent of inadequacy in the population is not readily estimable from these analyses alone. Total iron intake was 69% of the recommended amount of 18 mg/d for Russian women. Furthermore, after adjustment for both enhancers and inhibitors of iron absorption, the mean and 25th to 75th percentile range of estimated available iron tended to be less than even the lower end of the range of physiological iron requirements. This suggests that the level of iron inadequacy in this population may be considerably lower than what would be surmised based on total iron intakes alone.

Three previous studies adjusting for iron availability while assessing iron intakes have been based on Monsen's meal-based algorithms adjusting for MFP and vitamin C (Monsen and Balintfy 1982, Monsen et al. 1978). Raper et al. (1984), using data from USDA 1977-1978 Nationwide Food Consumption Survey, found that total iron intakes for women from 15 to 50 y old ranged from 55 to 59% of the Recommended Dietary Allowance. After adjustment for MFP and vitamin C, estimates were slightly lower; available iron ranged only from 45 to 48% of required amounts, and percent bioavailability averaged around 8%, less than the 10% assumed in setting the recommendations. Similarly, Viglietti and Skinner (1987), using data from food records of 224 adolescents, showed slightly less adequate levels of intake based on available iron than on total iron intakes; whereas iron intakes were 89 and 62% of the RDA for adolescent boys and girls, respectively, available iron after adjusting for MFP and vitamin C was 77% of the required amount for adolescent boys and only 51% that for adolescent girls. Again, percent bioavailability  between 8 and 9% for adolescent boys and girls  was less than the 10% assumed in setting the RDA. In another study, conducted in rural Mexico, Black et al. (1994) found total iron intakes that were two to three times the RDA for adult men and women. After adjustment for MFP and vitamin C, available iron was 120% of the RDA for men but 87% of the RDA for women; however, because of the high intakes of fiber and phytate in their population, the authors note that these are probably overestimates.

The studies by Raper et al. (1984) and Viglietti and Skinner (1987) found lower levels of total and available iron for women relative to recommended amounts than did our study. In these studies, total iron intakes were 55-62% of recommended amounts compared with 69% of the Russian recommendation in our study, whereas available iron after adjustment for MFP and vitamin C was 45-51% of required levels compared with about 76% in our study; percent bioavailability after adjustment for MFP and vitamin C was also slightly less favorable in these studies (8-9% compared with our finding of 9.5%). However, the results of the two studies, as well as the study by Black et al. (1994), are consistent with the finding that less favorable pictures of dietary iron adequacy are obtained after adjusting for bioavailability than from looking only at total iron intakes. In our study, iron intake expressed as a percentage of recommendation was actually lower than bioavailable iron adjusted for MFP and vitamin C expressed as a percentage of iron requirements; this is most likely attributable to the higher recommended intake of iron for women in Russia than in the United States.

Few previous studies have adjusted for factors other than MFP and vitamin C. In a study by Singer et al. (1982), iron intakes were categorized as having low, medium or high availability based on the presence of MFP and vitamin C in the same meal, according to an algorithm similar to that described by Monsen et al. (1978). Iron availability was further adjusted for tea by reducing iron intake by one category of bioavailability. Percent availability of heme and nonheme iron was then calculated based on the category in which it fell. In this study, further adjusting for tea did not seem to change mean levels of available iron substantially from mean levels obtained after adjusting for MFP and vitamin C alone.

In a study on toddlers in Egypt, Kenya and Mexico, Murphy et al. (1992) also adjusted for MFP and vitamin C using an algorithm based on Monsen's method (Monsen et al. 1978), with three major differences: 1 ) the MFP and vitamin C cutoff values used to categorize intakes as having low, medium or high availability were transformed into nutrient densities and applied to toddlers; 2 ) because meal-level information was not available, the algorithm was applied to daily intakes; and 3 ) the percent availability of nonheme iron for each category was changed to the levels suggested by FAO/WHO (1988), ranging from 5 to 15% as in our sensitivity analysis. To further adjust for tea, nonheme iron was then multiplied by a "tea factor" ranging from 1.00 to 0.40 depending on the amount of tea consumed. The authors found that estimated iron availability ranged from 5.5% in Mexico to 8.7% in Kenya, as compared with 7.7% in our study after adjustment for tea in addition to MFP and vitamin C. The reduction of nonheme iron availability due to tea ranged from 6 to 16%.

No previous studies have adjusted iron availability for the effects of phytates. One study by Cook et al. (1991) considered a large variety of enhancers and inhibitors  specifically, MFP, vitamin C, tea, coffee, whole wheat bread, bran muffin or bran cereal and eggs. However, these factors were not used to determine the amounts of iron available from total iron intake. Rather, they were used to create meal scores to rank nonheme iron availability, with larger scores representing meals with higher nonheme iron availability. Nevertheless, Cook et al. (1991) found that the meal scores correlated well with actual absorption, suggesting that these dietary factors are useful for ranking the quality of individual meals for iron bioavailability.

Several factors limited our ability to accurately estimate iron bioavailability in this study. Although the interactions of enhancers and inhibitors of iron absorption are complex (Hallberg et al. 1989, Siegenberg et al. 1991), our adjustments for the presence of MFP, vitamin C, tea and phytates were based on algorithms created from interpretations of the results of experimental work, primarily human feeding studies. Moreover, our algorithms assumed only additive effects and did not attempt to quantify interactions among dietary factors; currently, no model exists to estimate the effects of enhancers and inhibitors acting simultaneously on iron absorption. In addition, iron availability may be affected by still other factors not included in our analyses, such as coffee (Morck et al. 1983), calcium (Hallberg et al. 1991) or the proportion of iron derived from fortification (Hallberg and Rossander-Hulten 1991). Estimates of iron availability would be improved by further investigation into the effects and interactions of other dietary factors affecting iron absorption.

Our analyses were also limited by lack of information on iron status, another major determinant of iron absorption. The algorithms we used to adjust for iron availability were appropriate for a reference individual with well-supplied iron reserves; thus, the maximum level of nonheme iron bioavailability allowed by our algorithms was 8%. We believed this to be a reasonable assumption, because high intakes of red meat prior to economic reforms probably contributed significantly to long-term iron stores in this population. If iron status was less adequate than what we assumed, however, the actual level of iron bioavailability may be higher. Indeed, in a sensitivity analysis allowing a maximum level of nonheme iron bioavailability of 15%, mean available iron was considerably greater than estimates based on adequate iron stores. It is notable, nevertheless, that while mean available iron adjusting for MFP and vitamin C seemed to meet the estimated requirement for the population, it was still under half of the estimated requirement when further adjustment was made for phytates; thus, our conclusion that adjustment for enhancers and inhibitors of iron absorption should be considered in assessing bioavailability is not changed. However, future research in this area would benefit from the use of biochemical indicators both to help improve estimates of iron availability and to validate the extent of iron insufficiency in populations.

Despite these limitations, the ability to adjust for iron bioavailability is potentially useful in several respects. Because quality of the diet varies greatly across populations, an awareness of the proportion of total iron that is bioavailable in different settings would help inform decisions on recommended levels of intake and on patterns of food intake. Research in this area could identify the major as well as underutilized food sources of bioavailable iron in populations. These findings are important in nutrition education efforts  especially in areas with less varied diets where, because of the likely imbalance between enhancers and inhibitors of iron absorption, bioavailability more strongly determines iron status (Cook et al. 1991). In this case, altered patterns of tea consumption, enhanced intakes of sources of ascorbate at meals, and knowledge of phytate-rich foods could improve iron status significantly. Better understanding of adjustment for iron bioavailability also has important implications for policy recommendations. For example, under dietary conditions such as in Russia, where tea is consumed with every meal and phytate intakes are high, fortification of the food supply with iron will do less to improve iron nutriture than educational attempts at changing the meal pattern. In this case, again, educational measures may include encouraging the substitution of black tea with herbal teas or drinking tea at times other than during the nonheme-rich meals.

Our analyses suggest that only a small proportion of total iron intake is bioavailable after other factors in the meal, especially phytates, are considered. In addition, adjusting for other dietary factors may offer a less optimistic picture of the adequacy of dietary iron in a population compared with estimates based on total iron intake. In specific instances, such as comparisons of adequacy in rural and urban areas, the profile of concurrently consumed enhancers and especially inhibitors can profoundly affect conclusions.

Given the potential usefulness of adjusting for iron bioavailability, future research on the adequacy of dietary iron in populations should include some component to estimate available iron as well. Such research would necessarily involve collection of dietary data at the meal level. Because amounts of enhancers and inhibitors most likely vary across meals, information on daily intakes would not capture the effects of these dietary factors on iron availability; meal-level information may more accurately indicate the degree of adequacy or inadequacy of iron intakes in a population than would information on daily intakes. In addition, estimation of iron availability would also require fairly detailed information on food composition, including recipe files for mixed dishes, as well as information on dietary factors not normally included in nutrient databases, such as phytates. The health consequences of iron inadequacy are well known. More recently, iron excess has drawn greater attention as well. Because the estimation of iron availability could potentially offer a more accurate measure of the actual amount of iron received from the diet, adjustment for bioavailability is worth pursuing in future studies on iron intake.

ACKNOWLEDGMENTS