Materials and methods

STUDY SAMPLE

All births in the former Avon Health Authority area with an expected date of delivery between 1 April 1991 and 31 December 1992 were eligible for the ALSPAC study. Over 80% of the known births from the geographically defined catchment area were included, resulting in a total cohort of 14 138 surviving live births. From the population cohort, we invited a 10% randomly selected sample of parents whose babies were born within the last six months of the survey to bring their 12 and 18 month old children to a research clinic (children in focus), where a number of clinical, physiological, and psychological assessments were carried out. Of those invited, 1213 (88%) parents attended with their 12 month old children and 1151 (87%) parents attended when their children were 18 months old. The mean ages of the children attending the 12 and 18 month clinics were 54 weeks and 80 weeks, respectively. Permission was given by 940 and 827 of these parents for blood to be taken from their 12 and 18 month old children, respectively. Twins were excluded from this analysis.

BLOOD ASSAYS

A heel prick sample of capillary blood was taken from the children and collected into an EDTA capillary tube. The haemoglobin concentration was assayed using the HEMOCUE B-Hb photometer. The blood was then centrifuged and the plasma removed and frozen. Plasma ferritin was assayed using the DELFIA time resolved fluoroimmunoassay system. If the samples clotted they were discarded. Sufficient blood was obtained to measure the haemoglobin concentration for 900 and 796 children at 12 and 18 months of age, respectively. Sufficient plasma was obtained to assay ferritin for 734 and 718 children at 12 and 18 months of age, respectively.

The quality assurance scheme operating within the laboratory and the measures taken to test the stability of the samples were reported previously by Emond et al,7when the 8 month blood samples were analysed.

Ethical approval for the study was obtained from the ALSPAC ethics committee, and the local ethics committees of United Bristol, Southmead, and Frenchay Health Care Trusts. Following their advice, the results of those children whose haemoglobin was found to be less than 90 g/l were given to the mothers, with a letter to hand to their general practitioner suggesting further investigation.

STATISTICAL ANALYSES

All data were analysed within the SAS statistical software package (SAS Institute, Cary, North Carolina, USA). Visual inspection of the histograms of haemoglobin concentrations within the sample of 12 and 18 month old children participating in this study suggested that the data were normally distributed; however, ferritin concentrations were skewed and so were appropriately transformed (natural logarithm). Normality was confirmed using the Kolmogorov-Smirnov statistic, with a Lilliefors significance level. Pearson’s correlation coefficients were calculated to determine the extent of association between continuous variables. Generalised linear models using type III sums of squares were constructed to investigate factors associated with haemoglobin and ferritin concentrations in 12 and 18 month old children.

Growth velocity was expressed as the difference in weight of the child between two time points over the time between measurements. The change in ferritin concentrations was calculated as the difference in ferritin concentrations between two time points over the time between measurements.

Results

Data gathered from infants at 8 months of age are included in this paper for comparison with the data collected from children at 12 and 18 months of age. Analyses of the 8 month data have been reported in an earlier paper.7

HAEMOGLOBIN CONCENTRATIONS

Table 1 gives the centiles of haemoglobin ranked in ascending order for this cohort of children at 8, 12, and 18 months of age. It can be seen that median values remained fairly constant with age and between sexes, but the fifth and 10th centile cut off points rose with age and the 90th and 95th centile cut off points diminished, implying a narrowing of the distribution. Haemoglobin concentrations were most highly correlated between 12 and 18 month old children (r = 0.371; p < 0.0001; n = 696). The correlation was lower between 8 and 12 month old children (r = 0.258; p < 0.0001; n = 785) and between 8 and 18 month old children (r = 0.217; p < 0.0001; n = 700). Of those who attended all three clinics, 30 children had haemoglobin concentrations below the fifth centile at any one time. However, only two children had haemoglobin concentrations consistently below the fifth centile at all three clinic visits.

FERRITIN CONCENTRATIONS

Ferritin concentrations were measured when sufficient blood plasma remained after haemoglobin had been assayed. A total of 734 and 718 such samples were obtained from singletons at 12 and 18 months of age, respectively. The distribution of ferritin in both age groups is approximately log normal and the centiles are given in table 2. Both the median and the other centiles dropped with age, implying that iron stores were falling with age.

Log_e ferritin concentrations in the blood at 8, 12, and 18 months of age were correlated positively. Pearson’s correlation coefficient for the association between log_e ferritin concentrations at 8 and 12 months old wasr = 0.463 (p < 0.0001; n = 479); between 8 and 18 months old, r = 0.241 (p < 0.0001; n = 470); and between 12 and 18 months old,r = 0.334 (p < 0.0001; n = 518). Ferritin measures were available on 351 children at all three time points. Seventeen children had ferritin concentrations below the fifth centile at any one time point. However, only one child had ferritin concentrations consistently below the fifth centile on all three occasions.

FACTORS ASSOCIATED WITH HAEMOGLOBIN AND FERRITIN CONCENTRATIONS

Factors found to be associated with haemoglobin and log_e ferritin concentrations in the infants in this cohort at 8 months of age were investigated using multiple regression analyses. These comprised sex of the study child (boy or girl), birth weight (kg), weight on attendance at clinic (kg), and study mother’s highest educational qualification (CSE, vocational, O level, A level or degree). In addition, as potential confounders, we included ethnicity of the child (white or non-white), and whether the child had a recent infection before, or during, their visit to the clinic (yes or no) because these factors are known to affect ferritin concentrations. Table 3 shows unadjusted (geometric) mean haemoglobin and ferritin concentrations for the above factors and table 4 gives adjusted regression coefficients from the multiple regression analyses.

HAEMOGLOBIN CONCENTRATIONS

At 12 months of age, none of the variables in table 3 contributed significantly to the variation in haemoglobin concentrations within the cohort. By 18 months of age, however, sex of the study child, and the study mother’s highest educational qualification were significantly associated with haemoglobin concentrations. Boys had significantly lower haemoglobin concentrations than girls (p = 0.04; adjusted mean difference (95% confidence interval (CI)), 0.141 (0.004 to 0.319)). The effect of maternal education on haemoglobin concentrations of their children at 18 months of age appeared to follow a U shaped curve with the highest concentrations in children whose mothers attained a degree and the lowest concentrations in those children whose mothers gained a vocational qualification (p = 0.028).

FERRITIN CONCENTRATIONS

Log_e ferritin concentrations in infants at 12 months of age were associated with the sex of the study child, birth weight, and weight on attendance at the clinic. Girls had 8% higher ferritin concentrations than boys (p = 0.018). Birth weight was associated positively (p < 0.001), and weight on attendance at the clinic (p < 0.001) was associated inversely with log_e ferritin concentrations. Neither recent infection, ethnicity, or mother’s highest educational qualifications were associated with log_e ferritin concentrations at 12 months of age.

At 18 months of age, the birth weight of the study child and the weight on attendance at the clinic were significantly associated with log_e ferritin concentrations. Birth weight was associated positively, and weight on attendance at the 18 month clinic was associated inversely, with log_e ferritin concentrations. Children with reported recent infection at 18 months had 9% higher ferritin concentrations than those who had not reported a recent infection, but this difference was not significant (p = 0.066). There was no difference between boys and girls at this age, nor was there any effect of maternal education or ethnicity on ferritin concentrations.

Within the statistical model, when all else remains constant, an increase of 1 kg in birth weight (but no increase in clinic weight) would be associated with an increase in ferritin concentrations of 22% and 14% in 12 and 18 month old children, respectively. An increase of 1 kg in 12 and 18 month weight in children with the same birth weight would be associated with a decrease of 7% and 4% in ferritin concentrations at 12 and 18 months of age, respectively.

GROWTH VELOCITY AND DEPLETION OF FERRITIN CONCENTRATIONS

Using weights obtained at 8, 12, and 18 months, the growth velocities (g/week) were calculated for all children. The rates of ferritin depletion (μg/l/week) were calculated from 8, 12, and 18 month blood samples. In the period between 8 and 12 months of age, growth velocity correlated with the rate of change in ferritin concentrations (r = −0.154; p = 0.004). The more rapid the growth between 8 and 12 months of age, the more depleted were the iron stores. However, between 12 and 18 months of age, no significant relation was found, although a similar trend was noted (r = −0.074; p = 0.168).

Discussion

These data were derived from the “children in focus” study, a randomly selected 10% subsample of the geographically defined total population cohort enrolled in ALSPAC.10 Using the 1991 national census data, the children in focus sample has been compared with UK national figures, and found to be very similar. The main difference affecting this particular research is that children from ethnic minority backgrounds are underrepresented in the children in focus sample compared with the UK. This is an important difference because other studies11 12 have suggested that iron deficiency anaemia is more common in ethnic minority groups than in the white population of the UK. However, the small number of non-white children in our sample were not significantly different from the white children.

Because concentrations of haemoglobin within the samples of 12 and 18 month old children were approximately normally distributed, we have defined a one sided reference range from the fifth to 100th centile comprising 95% of the distribution from this healthy population. We defined the fifth centile, at ∼ 100 g/l, as the cut off point for anaemia. However, ferritin was log normally distributed and, for iron deficiency, the fifth centiles—16 μg/l at 12 months and 12 μg/l at 18 months—were used as logical cut offs.

We have shown that between 12 and 18 months of age haemoglobin concentrations in children do not change considerably, although there was a marked change in ferritin concentrations with age. Therefore, if ferritin is going to be used for defining iron deficiency anaemia, it is important that age specific norms are used.

The most widely known guidelines for the detection of iron deficiency anaemia have been produced by the WHO.6 In the USA, the Institute of Medicine suggests cut off points of 110 g/l for haemoglobin and 10 μg/l for ferritin, but these values are independent of age and do not take into account the differences in haemoglobin and ferritin concentrations from infancy to old age.11 Several authors have suggested alternative cut off points for infants and young children—for example, Millman9 and Freeman.8 Using our data, a logical cut off would be the fifth centile values for haemoglobin (100 g/l) and ferritin (16 μg/l at 12 months and 12 μg/l at 18 months). Table 5 compares the percentage of children in our study who would be identified as having iron deficiency anaemia using our proposed criteria based on the fifth centile and four other recommended criteria. The comparison shown in table 5 suggests that a cut off of 100 g/l for haemoglobin might be more appropriate at this age than the threshold of 110 g/l recommended by WHO and others.6 For ferritin, the comparison in table 5 suggests that the cut offs recommended by other authors might be too strict for a healthy population at this age. It is even more difficult to define appropriate cut offs using a combined haemoglobin and ferritin measure.

The lack of association between haemoglobin concentrations and birth weight or current weight was not surprising and was similar to what had been found at 8 months of age in the same cohort.7 The sex difference in haemoglobin concentrations identified at 18 months of age, although significant, might be of little importance clinically owing to the small size of the effect. At 12 months of age, we could not show an association between haemoglobin concentrations and educational level of the mother, but there was a weak relation at 18 months when haemoglobin concentrations were found to be higher in those children whose mothers were educated to university level compared with those children whose mother’s highest educational achievement was a vocational qualification. There was not a linear effect with other levels of maternal education, however, and this statistical difference might have arisen by chance.

The relation between other factors and ferritin concentrations were more complicated. The significant sex difference in mean log ferritin concentrations seen at 8 months of age was also found at 12 months but not at 18 months of age. In addition, similar to the findings at 8 months of age, the log ferritin concentrations in the children at 12 and 18 months of age were associated positively with the birth weight of the child, yet associated negatively with the weight of the child at 12 and 18 months of age, respectively. The most likely explanation for this finding is that ferritin stores are depleted by rapid growth in infancy, and that those infants most at risk of iron deficiency are infants with a low birth weight and low iron stores who show rapid catch up growth in the 1st year of life.

There is evidence from this and our previous paper that rapid growth in infancy leads to a depletion of iron stores, and it has been shown elsewhere that treatment with iron increases weight gain.13 To explore this finding further, we examined the relation between growth rate in early infancy and ferritin depletion rates. Using unadjusted growth velocity from 8 to 12 months and from 12 to 18 months, we found a significant relation between rate of growth and rate of ferritin depletion only between 8 and 12 months. Hence, relatively large infants by 1 year of age, particularly those who have crossed up the centiles, are more likely to have depleted iron stores and be at risk of developing iron deficiency anaemia.

CONCLUSION

Our study showed that haemoglobin concentrations in the blood do not vary greatly between the ages of 8 and 18 months, and neither are they strongly dependent on sex, birth weight, or weight of the child when blood was taken. Caution is needed when using ferritin as an indicator of overall iron status in infants because ferritin concentrations change rapidly in the first 18 months of life. We have demonstrated that age, sex, and the growth process in early infancy are important determinants of ferritin concentrations in the blood, resulting in different children being iron deficient at different ages. Follow up of these children is needed to identify important outcomes, such as cognitive function, to determine whether it is appropriate to screen for iron deficiency, and at which age(s).