Results

We considered results separately for each of the three ethnic groups because, as described in the report of the survey, the characteristics of the groups were quite different from each other.4

DIETARY AND LIFESTYLE QUESTIONNAIRE

Tables 1-3 present data obtained at the two year interview. Socioeconomic factors and data on feeding practices, diet, and growth have been brought forward from the previous stages of the infant feeding survey and used in the analysis described in this paper. The full results of the first four interviews are published in the report of that survey.4

Table 1 shows the milk drinking habits of the group at 2 years of age: 97% of Indian and Pakistani children and 92% of Bangladeshi children were given milk; 89% of Pakistani, 85% of Bangladeshi, and 79% of Indian children still drank milk from a bottle, even if only occasionally. Most children who were given milk in a bottle or a cup received whole milk, but 8% of Indian and Bangladeshi and 7% of Pakistani children were given semiskimmed milk (compared with 14% in the NDNS). One per cent or less in any group were given skimmed milk. Table 2 shows consumption of other drinks such as juices.

Table 3 shows how often children ate selected foods. Seventy two per cent of Indian children, 67% of Pakistani, and 62% of Bangladeshi children ate breakfast cereals daily. More than 90% of Indian and Pakistani children ate bread, naan, or chapati daily. Bangladeshi children ate them less frequently: 46% at least once a day, but 85% at least once a week. Twenty eight per cent of Indian children never ate meat. Bangladeshi children were the most likely (31%) to eat meat every day. Although Pakistani children did not eat meat as frequently as this, 96% ate meat at least once a week. Twenty two per cent of Bangladeshi children ate fish at least once a day. Only 1% of Pakistani and 2% of Indian children ate fish that frequently, but 75% of Pakistani and about 50% of the Indian children ate fish at least once a week. Thirty six per cent of Indian, 14% of Bangladeshi, and 12% of Pakistani children never ate fish.

At age 2 years, 46% of Indian, 43% of Bangladeshi, and 38% of Pakistani children were given vitamin or iron supplements. Three per cent of Pakistani and 2% of Indian and Bangladeshi children were given iron supplements. About a quarter of children in all groups were given Department of Health vitamin drops containing vitamins A, D, and C.

BLOOD RESULTS

The mean haemoglobin values for Bangladeshi and Indian children was 116.0 g/l and for Pakistani children 114.0 g/l, compared with a mean of 120.0 g/l for children aged 1.5 to 2.5 in the NDNS. Twenty nine per cent of Pakistani children, 25% of Bangladeshi, and 20% of Indian children had haemoglobin concentrations below 110.0 g/l compared with the NDNS results of 12% below 110.0 g/l (table 4).

The mean ferritin concentrations were 15.7 μg/l for Indian, 15.0 μg/l for Bangladeshi, and 13.9 μg/l for Pakistani children. The mean concentrations for children in the NDNS was 21.0 μg/l. If the lower limit of the normal range is taken as 10.0 μg/l,5 then 45% of Pakistani, 41% of Indian, and 39% of Bangladeshi children had values below this compared with 28% of children from the NDNS (table 5).

FACTORS AFFECTING HAEMOGLOBIN AND FERRITIN CONCENTRATIONS

Table 6 shows the characteristics of children with a haemoglobin concentration < 110.0 g/l and table 7 shows the characteristics of children with a ferritin concentration < 10 μg/l, which proved to be significant in bivariate analysis. Some of these factors may be interrelated. By using a modelling analysis (multiple regression) it is possible to identify the factors that have the greatest impact on, for example, the child’s haemoglobin concentration and those that have no significant impact when the main factors have been taken into account. The variables included in the analysis are those that might be expected from the bivariate analysis of data from this study or from previous studies (see discussion) to have a significant effect on the concentration of haemoglobin or ferritin. The distribution of ferritin was skewed so a log transformation was done.

Table 8 shows variables significant for at least one ethnic group in predicting haemoglobin and ferritin concentrations. Table 9 shows characteristics associated with haemoglobin concentration at age 2 years, and table 10 shows characteristics associated with ferritin concentration at this age. As only two thirds of the samples collected were sufficiently large to include a ferritin estimation, the model to analyse the characteristics associated with ferritin are based on a smaller sample than that described for haemoglobin.

The two most significant factors affecting haemoglobin concentrations of Bangladeshi and Pakistani children were a negative association between haemoglobin at age 2 and weight gain from birth to 15 months and whether they were mainly given cows’ milk at 9 months. The haemoglobin values of Pakistani children were also negatively related to the number of children in the family. Among Bangladeshi children, the amount of cows’ milk consumed in a bottle at 2 years of age was also a significant factor. Bivariate analysis showed that Bangladeshi children who were anaemic were significantly more likely to be bottle fed and consumed on average 600 ml or more of cows’ milk daily.

The significant factors affecting haemoglobin concentrations of Indian children differed from other groups. They were more likely to have a lower haemoglobin concentration if they were mainly given cows’ milk at 15 months or their mother was Muslim. Indian children were likely to have higher haemoglobin concentrations if they received a vitamin supplement or ate cereal at 2 years of age.

Among children in all groups, the two most significant factors that affected the concentration of ferritin at 2 years old were whether they were mainly given cows’ milk at 15 months and the amount of cows’ milk consumed in a bottle at 2 years of age. Pakistani children were also likely to have a lower concentration of ferritin if they were mainly given cows’ milk at 9 months.

Although the models indicate factors that affect haemoglobin and ferritin values, a large proportion of the variance is still unexplained. The proportion of the variance explained by the models to predict haemoglobin was 9–10% and those to predict ferritin levels was 11–17%. While these figures are not unusual for this type of modelling, the results should be used with care.

Discussion

IRON DEFICIENCY IN ETHNIC GROUPS

This is the first large, nationally representative survey of Asian children covering most areas with an Asian community in England. The prevalence of anaemia (20–29%) concords with the local smaller studies done in the past 10 years, which have described a high prevalence of anaemia (defined as haemoglobin < 110.0 g/l) in young Asian children compared with white children living in the same areas: Ehrhardt surveyed children aged between 6 months and 4 years admitted to hospital in Bradford and found a prevalence of 28% with a low haemoglobin concentration in 225 Asian children (mainly Pakistani) compared with 12% in 373 non-Asian children6; 68% of a sample of 55 Asian 2 year olds had a haemoglobin < 120.0 g/l in a study in Rochdale compared with 23% of a similar sample size of white children of the same age7; a study of 51 1 year old infants in London found a prevalence of 26% in Asian children (mainly Bangladeshi) and 12% in non-Asians8; a later study also in east London screened children aged 8 months to 2 years and found a prevalence of anaemia of 16% in Asian (mainly Pakistani) and 9% in white children (sample size 62 and 52 respectively)9; Marder et al found a prevalence of 39% anaemia in 44 Asian children from inner city Nottingham aged 15 months to 2 years and 16% in white children.10 A number of studies that did not include a white comparison group have described a high prevalence of haemoglobin < 110.0 g/l: Grindulis et al studied 135 Asian children aged 22 months in Birmingham and described a prevalence of 31% anaemia11; a prevalence of 24% was described in 138 Asian children in Sheffield aged 12 months to 2 years.12

FERRITIN

Serum ferritin is reduced in iron deficiency, and in adults correlates with the quantity of storage iron in the body.13 It would be expected to decrease before erythropoiesis would be affected, although this did not seem to be the case in these groups, as indicators of iron deficient erythropoiesis were detected in many more children than the 8–13% who had low ferritin values. Most studies where a high prevalence of haemoglobin < 110.0 g/l has been found also describe a high prevalence of ferritin values below 10 μg/l for Asian and white children; in the most recent study, 48% of children had low ferritin values and 24% also had low haemoglobin concentrations12; other studies in children have described a prevalence of low ferritin values of approximately double the prevalence of low haemoglobin values.6 8 In the NDNS for children aged 1.5 to 2.5 years a correlation between haemoglobin concentration and ferritin of 0.12 (p < 0.05) was found. Twenty eight per cent of children had low ferritin values compared with 12% with low haemoglobin values. A significant correlation of 0.37 (Bangladeshi), 0.32 (Pakistani), 0.21 (Indian) (p < 0.01) was found between haemoglobin and ferritin in this study, although we find it difficult to explain the unexpectedly high proportion who had normal ferritin concentrations. Infection is known to increase serum ferritin values, and it may be may be that the Asian children in this study had more infections than their white counterparts in the NDNS study, although children with a current or recent infection were not asked to give a blood sample. It has been noted in a previous study that Asian parents reported more infections in their children during their first year compared with white children.7

DIET

A number of dietary practices that may predispose to iron deficiency have been described in Asian populations in the UK. Compared with white mothers, Asian mothers tend to introduce solid foods later,7 8 14 15 and although we confirmed this finding,4 more than 90% of children from all groups were receiving some solids by 4 months of age. Asian children are reported to receive more sweet desserts and fewer savoury and meat based foods than white children during infancy,7 15 although Thomas and Avery showed that only Bangladeshi children ate desserts more frequently than white children.4 In the current study, meat consumption at 2 years of age was not found to be significantly associated with iron status. The Asian children consumed meat less frequently at 15 months than did white children; 53% of Bangladeshi, 30% of Pakistani, and 21% of Indian children ate meat daily compared with 69% of white children. Twenty two per cent of Indian households were vegetarian.

Apart from milk, the only food that was significantly associated with iron status was seen in Indian children, where those who ate breakfast cereal daily at 2 years of age were less likely to be anaemic. Data on iron content of cereals was not documented, but the majority of breakfast cereals in the UK are fortified with iron.

Although there appears to be no correlation between dietary iron intake and iron status,9 iron intake in Asian children has been described as being similar to,7 or lower than,9 12 13 similar samples of white children. Some attempts have been made to make a more general assessment of diet,16 and in a mainly white sample of children in Montreal, iron deficiency was associated with a “poor diet”.17 In our study, white children at 15 months ate a greater variety of foods in their diet than any of the Asian groups.4

A characteristic of Asian weaning diets is the emphasis and value put on cows’ milk as a staple food for young children. Asian mothers tend to introduce cows’ milk as a main drink earlier than white families do and encourage young children to drink larger quantities, with up to 60% of energy intake at 18 months derived from milk.12Several studies have described an increased risk of poor iron status in children who received cows’ milk as their main drink before 1 year of age. Most have found that the introduction of cows’ milk at 6 months of age results in a poorer iron status between 1 and 2 years of age compared with children who received an iron fortified formula until the age of 1,8 18 19 and one study showed that continuing with an iron fortified formula until 18 months had an advantage in terms of iron status at age two years.20 Two studies have not shown any association between age of introduction of cows’ milk and subsequent iron status.11 12 Mills described an effect of cows’ milk on iron status after introduction at 8 months of age.9 In our study, the introduction of cows’ milk at 9 months increased the risk of iron deficiency in Bangladeshi and Pakistani children. The iron status of Indian children was only affected if cows’ milk was given at age 15 months. This study endorses the recommendation by the Department of Health that cows’ milk should not be used as a main drink before the age of 1.21

The quantity of cows’ milk was shown to be negatively associated with iron status; Bangladeshi and Pakistani children who consumed more than 600 ml daily and Indian children who consumed more than 700 ml daily were found to be at greater risk of iron deficiency. These quantities are considerably smaller than the volumes associated with poor iron status in some earlier studies. Morton suggested that an intake of 900 ml was a risk factor for iron deficiency,8 although Kim found that as little as 500 ml at 6 months had an effect.19 Mills found no direct correlation between volume of milk consumed and haemoglobin concentrations but did report that iron deficiency was more common in Asian children consuming more than 1000 ml daily.9 The use of feeding bottles after the age of 1 was associated with higher milk consumption,4 and previous studies have described prolonged use of feeding bottles in Asian children.7 14 15 We found that at 15 months at least 90% of Asian children were still bottle fed, if only partially, compared with 75% of white children; 67% of white children drank milk from a cup at this age compared with 50% of Indian, 41% of Pakistani, and 26% of Bangladeshi children.4 The addition of cereal or sugar is more common among Asian mothers14; in the current survey approximately a third of Bangladeshi mothers added something to the milk of their 15 month old child: mainly rusk, baby rice, or other cereal; 10% of Pakistani mothers added sugar. It is likely that sweetening the milk and the use of a baby bottle compared with a cup is likely to encourage a high consumption of milk to the detriment of other foods. Milk is a valuable food for children and 600 ml is commonly recommended as being a suitable volume; further work needs to be done on the optimum quantity of milk for young children.

GROWTH AND IRON STATUS

Birth weight and growth velocity are considered to be major influences on iron status during 6–12 months of age. Faster growing children are theoretically at greater risk of deficiency; evidence is conflicting, although most studies have considered a younger age group than those in this study. No correlation was found between growth and iron status up to 1 year of age in one study,22 while Morton found that iron deficiency at 1 year of age was significantly correlated with a greater weight gain.8 Two studies of white children in Europe have described a negative effect of growth velocity on iron stores although not on haemoglobin values. In one study, infants were measured at 1 year,23 whereas the other study examined older children with a mean age of 7.1 years.24 In our study, we found no relation between weight gain up to 15 months and haemoglobin values, but ferritin values showed a significant relation with weight gain between birth and 15 months in the Bangladeshi group. We find this difficult to explain because Bangladeshi children were smaller than white children and children from the other Asian groups at 15 months of age,4 and although their increases in length and weight from birth to 15 months were larger than for the white group, the increase was still smaller than that seen in the two other Asian groups.

SOCIOECONOMIC AND ETHNIC FACTORS

We examined socioeconomic factors such as family size, income, education, and birthplace of mother. None of these factors showed an association with iron status for any of the Asian groups. The iron status of Pakistani children was negatively associated with the number of siblings in the family. Among Bangladeshi and Pakistani children, iron status was negatively associated with the mother being born outside of the UK; in addition, Bangladeshi children tended to have poorer iron status if their mother did not have a good understanding of English. Indian children were more likely to have poorer iron status if their mother’s religion was Muslim; this may reflect the socioeconomic characteristics of Indian Muslims who, in our study, tended to be less well educated than other Indian mothers.

We have difficulty in explaining the reason for different factors being associated with iron status in the three groups who all originated from the same sub-continent; the higher prevalence of iron deficiency in the Pakistani children is puzzling also because Pakistanis have usually been established in the UK for longer than Bangladeshis, who tend to be among the more recent immigrants. In the current study, 68% of Bangladeshi, 56% of Pakistani, and 50% of Indian mothers had arrived in the UK within the past 10 years.4 Differences in red cell indices between Afro-Caribbean and white ethnic groups has recently been reported.25 Wharton questions the possibility of a genetic influence on the expression of transferrin and transferrin receptors.26 The prevalence of iron deficiency is higher for children in Africa and southern Asia than in other areas of the developing world.27 It may be that there are racial differences, and our data require further analysis to determine whether it is appropriate to compare Asian children directly with white children.

CONCLUSIONS

There is considerable evidence of an association between iron deficiency and impaired mental and motor function in childhood in various ethnic groups,28 although a causative effect of iron deficiency has been questioned.29 We are concerned that our large national study has confirmed that up to one third of children from three different Asian groups in the UK have haemoglobin concentrations that may be associated with delayed development. The downward trend in the prevalence of iron deficiency seen in the United States during the 1970s seems not to have occurred here30; it is interesting to note that this trend appears to have reached a plateau with a prevalence of iron deficiency among children aged 1 to 2 years in the United States that is slightly lower (9%) than the NDNS study.31 No single factor or, indeed, no model using a number of factors explains more than a small proportion of the variance in the current study. The causes of low haemoglobin values in this group remain largely unexplained; further work needs to be done to determine appropriate normal values of indicators of iron status in populations of non-white children: whether there are differences in iron absorption in young Asian children and what are the specific effects of poverty and infection on iron status. A number of prevention strategies have been suggested,32 but before such strategies can be successful, more evidence is needed about the mechanisms underlying iron deficiency in young children.