Identification Sheet

Title: The effects of micronutrient interactions on iron status using the NDNS survey of children (AN0848)

Project Leader: Dr Barrie Margetts

Institute of Human Nutrition

University of Southampton

Biomedical Sciences Building

Bassett Crescent East

Southampton

SO16 7PX

Project duration:1 April 1998 to 31 March 1999

Total Project costs:£51 453

Staff Input:Dr Rachel Thompson RA2A.151.0year

CLE40.2

Scientific Objectives:

To investigate the effects of micronutrient interactions on iron status using data from National Diet and Nutrition Survey of children1 ½ to 4 ½ years of age.

To assess the effect of under-reporting on the interactions investigated above.

Specifically:

1)Is dietary iron related to iron status only in children with low iron status?

2)Do the food sources of iron, and riboflavin, vitamin A and zinc, together with the meal patterns, affect iron status.

3)Does marginal vitamin A, zinc and riboflavin status influence iron status?

4)Do children with a higher infectious load have poorer iron status?

5)Assess the impact under-reporting has on objectives 1-4

6)Publish results in peer reviewed journals

Primary milestones:

01April-JuneClean, organise and clarify extent of under-reporting

02July-OctoberAnalysis and development of model

03NovemberAssess the impact of under-reporting on model

04MarchFinal analysis, preparation of papers, submission of papers for consideration for publication
Executive Summary

Iron deficiency and iron deficiency anaemia are widely prevalent and thought to underlie much morbidity and impaired development. Chemically iron is highly reactive and its metabolism is closely regulated. Functionally, a limitation in the availability of iron at the point of its metabolic use might be due to limited intake, metabolic sequestration, or the limited availability of another nutrient required for its effective utilisation. As the body content of iron is regulated by absorption, an apparent iron deficiency might be accounted for by a dietary deficiency, interactions with other components in the diet, limitation in the availability of other nutrients, or other pathological processes associated with infection or an inflammatory response. The present analysis sought to clarify the effects of the interactions of markers of iron status with lifestyle factors, diet, biochemical markers for other nutrients and markers of inflammation. Three markers were used to indicate iron status: haemoglobin as a functional marker of the ability to use iron in the longer term; ferritin as a marker for iron in storage; and zinc protoporphyrin (ZPP) as a marker of the availability of iron at the site of haemoglobin formation.

Specifically we sought to address the following questions:

  1. What are the main factors affecting iron status in healthy children?
  2. Is dietary iron related to iron status only in children with low iron status?
  3. Do the food sources of iron, and riboflavin, vitamin A and zinc, together with the meal patterns, affect iron status?
  4. Does marginal vitamin A, zinc and riboflavin status influence iron status?
  5. Do children with evidence of current infection have poorer iron status?
  6. Might under-reporting of dietary intakes impact on the interpretation of the above?

From our analyses we found that:

  • When considered on their own, food patterns and nutrient intakes were associated with measures of iron status, but when they were included in multiple regression models with included biochemical, social and anthropometric measures, statistically they became less important in well children.
  • For haemoglobin, vitamin C (positively) and n-6 polyunsaturated fats (negatively) were the only dietary measures which were included in an explanatory model, with biochemical measures of retinol, zinc, vitamin D, and body weight being more strongly associated than dietary measures.
  • For ferritin, children with a diet of poorer quality (more cakes, and sugary drinks) were more likely to have lower levels of plasma ferritin, and there was a stronger statistical association with plasma folate.
  • For ZPP, there was a negative association with the consumption of diet soft drinks and with plasma zinc.
  • In the children with low haemoglobin levels (8%), dietary iron intake was associated with haemoglobin.
  • A higher body weight was associated with better measures of iron status, but within the present data it is not possible to determine the extent to which this potentially important association is causal in one direction or the other.
  • Evidence of current infection was not consistently associated with each of the measures of iron status, suggesting a more complex interaction than exposed in previous analyses.
  • Dietary under-reporting, although present, had little effect on associations reported.

These data show important interactions amongst indices of micronutrient status and markers of iron status. There has been a tendency to draw a direct relationship between dietary iron and iron status, which has been translated into dietary fortification or supplementation programmes. The implications of these data are that during childhood more complex interactions amongst nutrients might be of equal or greater importance, as differences in blood concentrations of retinol, zinc, and folate appeared to be more important influences on haemoglobin than simply dietary iron intake. When different markers for iron status, haemoglobin, ferritin, and ZPP were used, the pattern of associations were different and the implications of these observations need to be explored more fully. They suggest that the metabolic handling of iron is affected by other nutrients, and that factors that determine bioavailability within the gastrointestinal tract differ from those that influence either the storage of iron or its use in specific metabolic pathways.

The associations found between anthropometric measures, markers of inflammation and measures of iron storage appear of particular importance and the nature of the causal relationships need to be determined.

Scientific Objectives:

The objective of this study was to use the data obtained in the National Dietary and Nutritional Survey of Children between 1½ and 4½ years of age (Gregory et al, 1995) to investigate the effects of micronutrient interactions on iron status.

Specifically, the following questions were addressed:

1)What are the main factors affecting iron status in healthy children?

2)Is dietary iron related to iron status only in children with low iron status?

3)Do the food sources of iron, and riboflavin, vitamin A and zinc, together with the meal patterns, affect iron status.

4)Does marginal vitamin A, zinc and riboflavin status influence iron status?

5)Do children with evidence of current infection have poorer iron status?

5)Might under-reporting impact on the interpretation of the above?

Methods

Between 1992 and 1993 a national dietary and nutritional survey was carried out in a nationally representative sample of 2102 children between 1½ and 4½ years of age (Gregory et al, 1995). The survey included an interview with the parent on the food habits and lifestyle consideration of the child, a weighed dietary record of all food and drink consumed over a period of four days, anthropometry, and a blood sample. For the present analysis, three markers of iron status were identified: haemoglobin concentration, a functional measure of iron utilisation; ferritin, a measure of stored iron; and zinc protoporphyrin, a measure of the metabolic availability of iron at the site of haemoglobin formation. Inflammation is known to influence iron absorption and metabolism. A number of biochemical indices might be used as measures of an inflammatory response, -1 antichymotrypsin (ACT) is the only measure which is not directly, or indirectly, related to other aspects of iron metabolism or nutrient availability (eg iron for ferritin and copper for caeruloplasmin). Amongst the range of complex metabolic interactions, the interactions of vitamin A and riboflavin with the availability and utilisation of iron within the body have been well described. Under-reporting of the food consumed is always a concern in dietary studies, and in the present analysis the potential effect this might have on the interpretations was assessed. It was assumed that if under-reporting werepresent for any individual, the reported energy consumption would be less than 1.2 times the estimated basal metabolic rate.

General

The data were cleaned and checked before analysis began to ensure that the variables and data we used were the same as those reported. In the published report it stated that 74 children were excluded for quality control reasons, but we could only account for 72 such children. Although only a small discrepancy, it took us a considerable time to track these cases down and clarify the anomalies. The normality of the distribution of all continuous variables was assessed. Distributions that were skewed were transformed to approximate more closely to normal. Distributions where the skewness statistic was greater than or equal to 1 were defined, for this purpose, as being skewed, and requiring transformation. The most common transformation was base10 logarithmic and square root transformations.

  1. Is dietary iron intake related to iron stores only in children with low iron status?

Spearman rank correlation coefficients were used to assess the relationship between dietary iron intake (from food sources and all sources) and iron status (haemoglobin, ferritin and zinc protoporphyrin (ZPP)). The main outcome measure for this analysis was haemoglobin. Ferritin was used as a proxy measure for iron stores. ZPP was used as an indictor of inadequate iron supply at the site of haem formation. The relationship between dietary intake and iron status measures was assessed in children with haemoglobin above and below 11g/dl (the level used by the World Health Organisation for defining anaemia). A cut-off 10µg/l was used for ferritin. To assess the relationship between different levels of iron status and dietary iron intake without using a pre-defined cut-off the distributions for haemoglobin, ferritin and ZPP were divided into thirds and the relationship between iron status and iron intake assessed for each third separately.

  1. Do the food sources of iron and the meal patterns affect iron status?

In order to compute one variable describing education, income and occupation of the head of the household a cluster analysis was performed on highest qualification achieved, income (in thirds) and occupation (manual or non-manual or never worked). To identify food patterns, a principal component analysis was carried out. The analysis was set so that only components with an Eigen factor of greater than one were extracted. A varimax with Kaiser normalisation rotation method was used. The original food groups were combined to give 25 food groups (appendix A). Food groups were chosen à priori to cover the whole diet but also to include food groups that may affect iron absorption such as tea and coffee and fruit juices. Individual food groups and nutrients were correlated with haemoglobin and ferritin and those with a statistically significant Spearman correlation coefficient were selected for further analysis (appendix B). As correlations were higher for nutrient intakes from food sources rather than all sources, values excluding supplements were used. A stepwise multiple regression analysis was carried out with individual foods, food patterns, nutrients and a number of other important variables as independent variables and an iron status measure (haemoglobin, ferritin or ZPP) as the dependent variable. The other important variables were the social cluster variable, age, gender, body weight, birth weight, height and whether the child was given vitamin supplements.

  1. Does marginal vitamin A, zinc and riboflavin status influence iron status?

Lower haemoglobin concentrations may be a result of poor intakes of other nutrients such as vitamin A, riboflavin and zinc. Vitamin A and riboflavin both affect iron absorption and utilisation and zinc is important in the synthesis of haemoglobin and affects the bioavailability of iron. The main dietary source of vitamin A, riboflavin and zinc is milk. To determine whether children with low milk consumption also had low intakes of vitamin A, zinc and riboflavin, nutrient intakes for each fifth of milk consumption were computed. The association between dietary and blood measures of vitamin A, riboflavin and zinc was assessed using Spearman rank correlation coefficients. To assess whether blood or dietary variables were more strongly related to iron status, Spearman rank correlation coefficients were computed for dietary and blood measures (vitamin A, zinc and riboflavin) with measures of iron status (haemoglobin, ferritin and ZPP). To assess whether children with lower intakes and blood levels of vitamin A, zinc and riboflavin also had a poorer iron status independently of their iron intake an analysis of variance was carried out. Dietary and blood measures of zinc, vitamin A and riboflavin were divided into fifths and then grouped into the bottom fifth and remaining fifths. The model also included age, gender, social cluster, birth weight, current body weight and iron intake.

  1. Do children currently experiencing an inflammatory response load have poorer iron status?

Infections elicit metabolic responses, including the inflammatory response, with widespread, co-ordinated alterations in metabolism, which includes a change in the pattern of proteins secreted by the liver. There is an increase in the synthesis and secretion of “acute phase reactants” such as ACT, and the magnitude of the increase in circulating concentration might be used as a crude measure of the infective load. Infection may be associated with a lower haemoglobin. In addition, data are available for ACT, albumin, caeruloplasmin, and ferritin. As ACT is the only acute phase reactant not directly or indirectly related to iron metabolism, this has been used as the independent indicator of the presence of infection. The association between the iron status measures and measures of inflammation was assessed using Spearman rank correlation coefficients for ACT, albumin, caeruloplasmin (both as continuous variables and in thirds), and iron status measures (haemoglobin, ferritin and ZPP). To determine which factors accounted for the variation in iron status a stepwise multiple regression analysis was carried out with iron status as the dependent variable. The independent variables included foods, food components, nutrients, social cluster variable, age, gender, body weight, birth weight, height and whether child took vitamin supplements, blood measures of retinol, zinc, vitamin B12, folate, vitamin C,  tocopherol, vitamin D and inflammation markers, ACT, albumin, caeruloplasmin and ferritin.

  1. Assess the impact of under-reporting

The Schofield equations for the relevant age groups (Schofield et al 1985) were used to calculate basal metabolic rate. In order to assess the effect of under-reporting a cut-off of 1.2 for energy intake to basal metabolic rate ratio (EI/BMR) was used. It is recognised that this is an arbitrary cut-off and that under-reporting is possible across the whole range of intakes therefore the distribution of EI/BMR was also divided into thirds.

To determine whether there was differential under-reporting in the toddlers the social and demographic characteristics of the toddlers above and below the 1.2 cut-off for EI/BMR and across thirds of the distribution were compared. The effect of under reporting on nutrient and food intakes was assessed by comparing mean intakes in children above and below the cut-off. To assess the effect of under-reporting and being unwell on the final results the multiple regression analysis as used in 4) was re-run on sub-groups of the population. Firstly those children with biochemical evidence of infection (ACT > 0.65) were excluded. For the remainder of children beta coefficients were computed separately for those who were well during the diary, those above the 1.2 cut-off for EI/BMR and those who were both well and above the 1.2 cut-off. The final group was considered to be better reporters of dietary habits and were well at the time of the diary and thus were considered to be the group of the most representative of normal health. Complete data on all variables were not available for all children. To assess the generalisability of the results the social and demographic variables of those toddlers with complete dietary and blood data were compared to children without complete data.

Results

  1. Is dietary iron related to iron status only in children with low iron status?

Table 1 shows the correlations for haemoglobin and ferritin with dietary iron. For all subjects small but statistically significant correlations were found. In general correlations were stronger for food sources of iron rather than all sources which included dietary supplements. When children were divided into thirds of the distribution for ferritin only those children below the cut-off and in the lowest third of the distribution showed a statistically significant association with iron intake. The same was true for haemoglobin and ZPP. When the 11g/dl cut-off for haemoglobin was used statistically significant correlations were seen for children above the cut-off although the correlation between iron intake and haemoglobin was stronger for those below the cut-off. This may be because few children had haemoglobin levels below 11g/dl. Table 2 shows similar results for ZPP and shows statistically significant negative correlations with iron from food sources for all subjects and those in the highest third. As single measures of iron status may not be good indicators of iron deficiency those children with haemoglobin below 11g/dl and ferritin less than 10g/l were selected. Only 31(3%) children had low values for both haemoglobin and ferritin. Correlations in this group were 0.47 for haemoglobin with iron from food sources and 0.37 with iron from all sources (both were statistically significant). Correlations with ferritin were lower and did not reach statistical significance (0.23 for food sources and 0.25 for all sources). Respective correlations with ZPP were –0.21 and –0.03.