SUBJECTS AND METHODS

Subjects were part of a prospective study designed to examine the effects of diet on growth and development in preterm, low birthweight infants after hospital discharge.^17,18 Healthy, low birthweight (< 1750 g), preterm (< 34 weeks gestation) infants were recruited immediately before hospital discharge. Informed consent was obtained from the subject's parents, and the study received local ethical approval.

Infants were seen at discharge, at 38 weeks of postconceptional age, at their expected date of delivery, monthly to 6 months of corrected age, and at 9 months of corrected age. At each visit, blood was taken and preserved with sodium EDTA. Complete blood counts were obtained using a Coulter Counter (model STKS), which was calibrated daily with “4C” calibration solution. Whole blood ZPP/H was measured on unwashed red cells using an Aviv ZPHematofluorometer (Lakewood, New Jersey, USA). Plasma ferritin was measured by immunoradiometric assay (Ferritin Mab; ICN Pharmaceuticals, Orangeburg, New York, USA). All assays were carried out in the clinical haematology laboratories of the Royal Victoria Infirmary, Newcastle upon Tyne.

Statistical analysis was carried using the StatView statistical package for Macintosh (version 5.01; SAS Institute, Cary, North Carolina, USA). Simple and multiple regression, analysis of variance, and analysis of covariance were used as appropriate. Results were considered significant at p < 0.05. ZPP/H ratios are expressed as μg/g haemoglobin (Hb) and plasma ferritin as μg/l. ZPP/H and plasma ferritin were log transformed (to the base 10) before analysis to normalise the distribution, as both were positively skewed. Values are quoted as mean (SD), unless stated otherwise.

To study the between subject correlation between ZPP/H and plasma ferritin values, log (ZPP/H) and log (ferritin) were regressed against postnatal age and the square of postnatal age (both centred at postnatal age 180 days). Residuals around these regressions were calculated and averaged for each subject. Finally, all the individual average log (ZPP/H) residuals were correlated with the corresponding average log (ferritin) residuals. A significant negative correlation would suggest that subjects with higher than average serum ferritin would have lower than average ZPP/H ratio (and vice versa).

To study the within subject correlation between ZPP/H and plasma ferritin, log (ZPP/H) and log (ferritin) were regressed against postnatal age, the square of postnatal age (both centred at postnatal age 180 days). A subject's log (ZPP/H) residuals and log (ferritin) residuals were correlated, and an average correlation for all subjects was calculated. A significant negative correlation would support the hypothesis that changes in ZPP/H in a particular subject would be mirrored by reciprocal changes in plasma ferritin (and vice versa).

The relation between ZPP/H and plasma ferritin at 6–9 months of postnatal age and 6–9 months of corrected age were assessed by simple regression, as was the relation between log (ZPP/H) or log (ferritin) and other measures of iron status (haemoglobin concentration, mean cell volume, mean cell haemoglobin, and red cell distribution width). If there was more than one data point for a given subject during either of these time periods, only the first one was used in this part of the analysis.

RESULTS

A total of 109 infants were recruited, with a mean birth weight of 1345 (271) g, and a gestation of 214 (15) days. Almost half (48%) were boys, and most (60%) received no blood transfusions before discharge. The remainder received between one and five blood transfusions (median one). No infant had a blood transfusion after hospital discharge.

ZPP/H ratios

Altogether, 637 measurements of ZPP/H were made in the first year of life. Up to 11 measurements were made in each subject (range 1–11; median 6). ZPP/H was initially high, fell rapidly, then stabilised (fig 1). Multiple regression analysis shows that log (ZPP/H) was related to postnatal age (t = −3.74, p = 0.0002), but not postconceptional age (t = 1.68, p = 0.094). Most of the decline in ZPP/H occurred in the first 6 months of postnatal life (fig 1). Analysis of covariance with subject as a covariant showed that log (ZPP/H) varied significantly between subjects (F = 4.6, p < 0.001) and as a function of postnatal age (F = 138, p < 0.0001).

Plasma ferritin

In the later part of the study, plasma ferritin was measured at the same time as ZPP/H. A total of 357 samples were taken in 86 subjects. Between 0 and 10 measurements were made in each infant (median four). Plasma ferritin declined with increasing postnatal age (fig 2). It was less than 20 μg/l in 96 instances (27%), less than 15 μg/l in 58 instances (16%), and less than 10 μg/l in 25 instances (7%).

Relation between ZPP/H and plasma ferritin during the first year of life

There was a weak, but significant, correlation between ZPP/H and plasma ferritin (r = +0.26, p < 0.0001; fig 3), but it is important to note that the slope of the regression line is positive, not negative (as expected from theoretical considerations).

Residuals of log (ZPP/H) and log (ferritin) were calculated from a multiple regression model that used postnatal age and the square of postnatal age (both centred to a value of postnatal age 180 days). There was a weak (r = −0.22) but significant (p = 0.038) negative correlation between the average log (ZPP/H) residual for a subject, and the corresponding log (ferritin) residual. This is consistent with the hypothesis that subjects with higher ferritin concentrations have lower ZPP/H ratios (and vice versa).

A similar model that included “subject” was carried out. When a subject's ZPP/H residuals and ferritin residuals (around their own individual line of best fit) were correlated, the average correlation coefficient was −2.91, but did not differ significantly from zero (95% confidence interval −7.14 to + 1.33; p = 0.18). In other words, in any subject, changes in ZPP/H were poorly mirrored by changes in plasma ferritin.

ZPP/H and plasma ferritin at 6–9 months of age

ZPP/H and plasma ferritin was measured at 6–9 months of postnatal age (175–270 days of postnatal age, 362–477 days of postconceptional age) in 54 subjects. There was a non-significant negative correlation between log (ZPP/H) and log (ferritin) (r = −0.21, p = 0.12). Log (ZPP/H) tended to correlate negatively with haemoglobin concentration (y = 13.0 − 1.76x, r = −0.24, p = 0.08), and correlated significantly with mean cell volume (y = 88.1 − 17.2x, r = −0.49, p = 0.002), mean cell haemoglobin (y = 30.1 − 5.45x, r = −0.37, p = 0.006), and red cell distribution width (y = 10.4 + 5.5x, r = +0.43, p = 0.001). There was, however, no correlation between log (ferritin) and haemoglobin concentration, mean cell volume, mean cell haemoglobin, or red cell distribution width (p > 0.50 for all comparisons).

ZPP/H and ferritin was measured at 6–9 months of corrected age (459–561 days of postconceptional age in 51 subjects). Once again there was no significant correlation between log (ZPP/H) and log (ferritin) (r = +0.19, p = 0.18). However, there was a significant negative correlation of log (ZPP/H) with haemoglobin concentration (r = −0.48, p = 0.004), mean cell volume (r = −0.39, p = 0.005), and mean cell haemoglobin (r = −0.45, p = 0.0008) and a significant positive correlation with red cell distribution width (r = +0.39, p = 0.005). As before, there was no correlation between log (ferritin) and haemoglobin concentration, mean cell volume, mean cell haemoglobin, or red cell distribution width (p > 0.50 for all comparisons).

DISCUSSION

In adults, ZPP/H is inversely correlated with plasma ferritin,^12,14 and inversely related to the amount of stainable iron in the marrow.¹³ We observed a fall in plasma ferritin in the first year of life, which agrees with the known changes in body iron stores in neonates.^11,19 However, rather than seeing the expected increase in ZPP/H during this period, a decrease was seen. Indeed ZPP/H and ferritin were directly, not inversely, related. When the entire dataset was examined using multiple regression, with age included in the model, there was no relation between log (ZPP/H) and log (ferritin). However, there was an inverse relation between average ZPP/H residuals for each subject, from a multiregression model, and corresponding log (ferritin) residuals. This suggests that subjects with a higher than average plasma ferritin tend to have lower ZPP/H (the between subject correlation). However, the relation was relatively weak (r = −0.22). When a similar model was used that incorporated “subject”, there was no significant relation between an individuals's ZPP/H residuals (around their own line of best fit) and their ferritin residuals, when the correlation coefficients for all subjects were averaged. In other words, changes in a subject's ZPP/H correlated poorly with changes in their plasma ferritin (the within subject correlation). The positive correlation between log (ZPP/H) and log (ferritin) on simple regression analysis appears to be an artefact of the developmental changes in ZPP/H, which was high in the early part of the study.

In our study, ZPP/H was measured using a photofluorimetric assay, which can be falsely increased by other fluorophores such as bilirubin.²⁰ However, none of our subjects were visibly jaundiced. The effect of other fluorophores can be overcome by using washed red cells,^21,22 but a small study in term infants, using washed red cells, has shown similar ZPP/H ratios to our own.²¹

Plasma ferritin is not a perfect measure of iron stores, and may be falsely raised in infective conditions, and there is only indirect evidence of its value as a measure of iron stores in preterm infants. However, low values are felt to be highly suggestive of iron deficiency, and it remains the most widely used measure of iron status.^10,11 None of the subjects in the study were considered to have any infectious condition at the time of blood sampling.

The American Academy of Pediatrics recommends screening for iron deficiency between 6 and 9 months of age.⁷ We could not show a significant inverse relation between log (ZPP/H) and log (ferritin) at either 6–9 months of postnatal age or 6–9 months of corrected age. However, during both time periods, log (ZPP/H) correlated significantly with other measures of iron status (haemoglobin concentration, mean cell volume, mean cell haemoglobin, and red cell distribution width), whereas no such relation could be shown for log (ferritin). Higher ZPP/H ratios (suggestive of poorer iron status) correlated with lower haemoglobin concentration, lower mean cell volume, lower mean cell haemoglobin, and an increased red cell distribution width. The changes in haemoglobin, mean cell volume, mean cell haemoglobin, and red cell distribution width are those expected in iron deficiency. These findings also agree with the results of one small study that has shown a positive correlation between ZPP/H and red cell distribution width in preterm infants at the time of hospital discharge.²³ Although making correlations between different measures of iron status overlooks the different stages of iron deficiency,⁹ these results suggest that ZPP/H may be a more useful measure of iron status than plasma ferritin concentration in this age range.

It has been suggested that ZPP/H may only begin to increase once body iron stores are exhausted and iron delivery to erythroid precursors becomes the rate limiting step in haemoglobin synthesis,^9,22,24 although this is disputed.¹⁵ The former view does not explain the good correlation between ZPP/H and plasma ferritin in adults over a wide range of ferritin concentrations.^12,14

We have shown that subjects with higher ZPP/H levels tend to have lower plasma ferritins, although changes in ZPP/H, for a given subject, did not correlate significantly with changes in plasma ferritin. In the later part of the first year of life, ZPP/H correlated significantly with other measures of iron status, whereas plasma ferritin did not. It is possible therefore that the ZPP/H ratio may be a better measure of iron status in this later age group than plasma ferritin concentration. These results underline the difficulty in assessing iron status in preterm infants.