Article Text
Abstract
Objective To evaluate reticulocyte haemoglobin content (CHr), compared with ferritin, transferrin saturation (TS) and mean corpuscular volume (MCV), as a marker of iron deficiency (ID).
Design Retrospective analysis of clinically indicated blood samples taken between February 2010 and October 2012.
Setting Single-centre neonatal care unit.
Patients 210 very preterm (gestational age <32 weeks) or very low birthweight infants (birth weight <1500 g) at 3–4 months corrected age.
Main outcome measures Complete blood count, CHr, ferritin and TS determined as part of a standard follow-up examination. To detect the optimal CHr cut-off, ID was defined by the presence of more than two of the following three criteria: MCV <75 fL, TS <10%, ferritin <30 µg/L.
Results 210 preterm infants were included at a corrected age of (median (IQR)) 3.5 (3.0–4.0) months and with a CHr of 29.7 (28.6–30.7) pg. There were correlations between CHr and MCV (r=0.54, p <0.0001) and between CHr and TS (r=0.44, p <0.0001). There were 27 (13.4%) iron-deficient infants, and two infants (1%) fulfilled criteria of ID-anaemia. CHr was lower in infants with ID (26.4 (23.8–28.7) pg) than in those without (29.9 (29.0–30.8) pg, p <0.0001). The optimal CHr cut-off for detecting ID was 29 pg (sensitivity 85%, specificity 73%). Areas under the receiver operating characteristic curve for detection of ID tended to be higher for CHr compared with ferritin (0.92 vs 0.75), TS (0.90 vs 0.82) and MCV (0.81 vs 0.72).
Conclusions CHr seems to be a suitable marker for latent ID in preterm infants at 3–4 months corrected age and may be superior to ferritin, TS and MCV.
- Biochemistry
- Haematology
- Monitoring
- Neonatology
- Neurodevelopment
Statistics from Altmetric.com
What is already known on this topic?
-
Reticulocyte haemoglobin content (CHr) is provided by many analysers along with a whole blood count without additional blood loss or costs.
-
CHr is a strong and, compared with haemoglobin, early predictor of iron deficiency (ID) in older infants and adults.
-
Very preterm and very low birthweight infants are at high risk of ID.
What this study adds?
-
The CHr cut-off of 29 pg seems appropriate to detect ID and, potentially, also to guide iron supplementation in former preterm infants.
-
As a marker of ID, CHr may be superior to ferritin, transferrin saturation and mean corpuscular volume in this setting.
-
At 3–4 months corrected age, the CHr cut-off does not have to be adjusted for gestational age or birth weight if iron supplementation is given.
Background
Iron deficiency (ID) is still common, and infants are particularly at risk.1 Preterm infants are even more vulnerable because of low iron stores at birth,2 their exceptionally high postnatal growth velocity3 and uncompensated iatrogenic blood and, hence, loss of iron.4 Untreated ID in infancy and childhood may lead to neurodevelopmental impairment that cannot be corrected by later iron supplementation (reviewed5). Therefore, all low birthweight infants should receive an early6 ,7 enteral iron supplementation (at least 2–4 mg/kg),8–10 which should be continued for at least 12 months.9 Despite iron supplementation, some preterm infants develop ID,7 hence, early detection and treatment of ID is required to avoid future neurological impairments.
Bone marrow aspiration is the gold standard to diagnose ID, but was replaced by less invasive iron status parameters, such as soluble transferrin receptor, zinc protoporphyrin/haem ratio, transferrin saturation (TS) and ferritin, all with parameter-specific limitations (reviewed11) and confounded by developmental changes.12 The haemoglobin concentration (Hb) is widely used as a screening test for ID. As Hb is derived from the entire red blood cell population with a lifespan of approximately 120 days, the detection of ID will be delayed. Since reticulocytes persist in the circulation for only 24–48 h, measurements of the haemoglobin content of the reticulocytes (CHr) better reflect the current iron supply to the bone marrow (reviewed13 ,14). Moreover, CHr is reported as part of the reticulocyte count by certain haematological analysers without additional blood loss or costs, and results from different analysers seem to correlate well.15
Since very preterm as well as very low birthweight (VLBW) infants are at the highest risk of developing ID, we aimed to analyse their iron status at 3–4 months corrected age, establish a CHr cut-off value for the detection of latent ID and for guidance of iron supplementation.
Methods
Study population
This retrospective analysis was carried out at University Children's Hospital Tübingen (Germany) evaluating laboratory data gathered as part of the clinical routine follow-up examination from February 2010 to October 2012 in consecutively admitted very preterm (gestational age (GA) <32 weeks) and VLBW infants (birth weight <1500 g) at 3–4 months of age corrected for prematurity. Infants with haematological diseases were excluded from this analysis.
In Germany, institutional review board approval and informed consent are not required for this type of retrospective analysis of anonymised clinical data.
As part of the follow-up examination, venous blood samples were drawn for analysis of red blood cell count and iron status parameters.
According to institutional guidelines in place, all preterm infants received enteral iron supplementation (ferrous glycin sulfate complex) at a dose depending on their birth weight (<1500g → 6 mg/kg/d, 1500–1999g → 4 mg/kg/d, 2000–2500g → 2 mg/kg/d) starting as soon as full enteral feeds (≥150 mL/kg/d) were tolerated. At discharge, parents were advised to continue the iron supplementation at the same absolute dose until the follow-up examination at 3–4 months corrected age, and then to continue at a dose of 3 mg/kg/d until 12 months corrected age. All infants received fortified breast milk (1.3 mg iron/100 mL) or preterm infant formula (1.8 mg iron/100 mL) until hospital discharge, and parents were advised to continue breast milk fortification or preterm formula until the follow-up examination.
Generally, delayed cord clamping (45 s) was practiced in all preterm infants (GA <32 weeks).
Laboratory methods
Complete blood count and reticulocyte parameters were measured with the ADVIA 120 blood analyser (Siemens, Eschborn, Germany). Ferritin was measured with ADVIA Centaur, transferrin with BN Prospec and iron with ADVIA 1800 (all from Siemens, Eschborn, Germany). TS was calculated from iron and transferrin levels (TS=iron (µg/dL) / transferrin (mg/dL) × 70.9).
Computation of SD scores for weight
Based on anthropometric data, SD scores for weight (SDSW) were computed using the Microsoft Excel add-in LMS growth (V.2.77; http://www.healthforallchildren.co.uk/; 2012). The reference population is the British 1990 growth reference fitted by maximum penalised likelihood.16 ,17 Small for GA (SGA) was defined by an SDSW at birth <−2.
Definition of ID-anaemia and ID
For primary assessment of CHr at the age of 3–4 months, ID was defined by two or more of the following three indicators of iron status being below the reference range: mean corpuscular volume (MCV) <75 fL,18 TS <10%,19 ,20 ferritin <30 µg/L.21 ,22 Iron status was defined as ‘unknown’ if two out of three parameters were not available for analysis or if one parameter was not available and the remaining two were contrary to each other (ie, one parameter indicating ‘ID’ and the other ‘within the reference range’). For the diagnosis of ID-anaemia, infants had to fulfil the criteria for ID as well as that for anaemia (Hb<10 mg/L19 ,20).
When comparing the diagnostic accuracy between CHr and each of MCV, TS and ferritin, respectively, the diagnosis of ID was made if both the remaining two iron parameters were below the reference range (eg, for comparison of CHr with ferritin, both, MCV and TS had to be low to classify an infant as ‘iron-deficient’), iron status was defined as ‘unknown’ if more than one of the remaining two iron status parameters were not available, or the two parameters revealed contradictory results (ie, one parameter indicating ‘ID’ and the other parameter being ‘within the reference range’). Patients with iron status classified as ‘unknown’ were excluded from these comparisons.
Statistical analysis
Normality was examined by the Shapiro–Wilk test. Normally distributed data are shown as mean and SD. Because logarithmic (LOG10(y), LN(y)) as well as other transformations (1/(y), root(y)) failed to convert non-normally distributed data into normal distribution, non-parametric tests were applied and data reported as median (IQR). Univariate between-group comparisons were made using the Wilcoxon test, difference between medians and 95% CIs was computed by the Hodges–Lehmann estimate,23 and the Spearman correlation coefficient is reported to describe correlations. Multivariate analysis of covariance was performed to determine a possible influence of GA, SDSw at birth, gender and weight gain since birth, on CHr. Sensitivity, specificity and negative and positive predictive values were calculated along with 95% CI. p Values <0.05 were considered statistically significant.
Receiver operating characteristic (ROC) analysis was used to graphically assess the relationship of sensitivity and specificity for detection of ID at different cut-off values. The ‘optimal’ CHr cut-off was chosen where (sensitivity–(1–specificity)) reached the maximum value.24
The diagnostic accuracy of CHr was compared with that of ferritin, TS and MCV by comparison of areas under the ROC curve (AUC) using MedCalc V.12.7.7 (MedCalc Software, Ostend, Belgium). The multivariate analysis was performed using SAS V.9.2 (SAS Institute, Cary, North Carolina, USA). All other analyses were performed using JMP V.10.0.0 (SAS Institute).
Results
A total of 210 infants were enrolled; none needed to be excluded due to a haematological disease. GA at birth ranged from 23+1/7 to 36+0/7 weeks. The distribution of GA and SDSW at birth is shown in figure 1.
Characteristics of the infants as well as indicators for iron status and red blood cell count are summarised in table 1. Missing measurements were due to insufficient blood sample volumes.
CHr as an iron status parameter
CHr was positively correlated with MCV (r=0.54, n=209, p <0.0001, 95% CI 0.44 to 0.63, figure 2), mean corpuscular haemoglobin (r=0.56, n=209, p<0.0001, 95% CI 0.46 to 0.65) and TS (r=0.44, n=194, p<0.0001, 95% CI 0.31 to 0.54, figure 2), whereas only weak correlations were found with Hb (r=0.26, n=209, p<0.0001, 95% CI 0.13 to 0.38), haematocrit (r=0.17, n=209, p<0.05, 95% CI 0.03 to 0.30), reticulocyte count (r=0.28, n=208, p<0.0001, 95% CI 0.15 to 0.40), and no correlation was found with ferritin (r=0.12, n=204, p=0.08, 95% CI −0.01 to 0.26, figure 2).
ID-anaemic infants (n=2) had lower CHr levels (20.6 pg) compared with non-ID-anaemic infants (n=207, 29.7 (28.7–30.7) pg, median difference (95% CI) −9.1 (−5.9 to −11.7), p <0.05) and infants with ID (n=27) also had lower levels of CHr (26.4 (23.8–28.7) pg) compared with iron-replete infants (n=174, 29.9 (29.0–30.8) pg, median difference (95 % CI) 3.3 (2.2 to 4.4), p<0.0001).
There was no difference between infants with and without ID in weight, length and head circumference.
The 2.5th percentile for CHr in our population was 23.2 pg. The CHr cut-off value of 29.0 pg was established as ‘ideal’ to identify ID based on the optimal combination of 85 % sensitivity (95 % CI 68 % to 94 %) and 73 % specificity (95 % CI 66 % to 79 %), resulting in a negative predictive value of 97 % (95 % CI 92 % to 99 %) and a positive predictive value of 33 % (95 % CI 23 % to 45 %). The AUC of CHr for the diagnosis of ID was 0.86 (figure 3).
AUCs for CHr tended to be greater than those for ferritin (0.922 vs 0.750, n=156, p=0.25), TS (0.904 vs 0.817, n=79, p=0.15) and MCV (0.809 vs 0.721, n=80, p=0.25) (figure 4). The number of infants evaluated in each comparison varied because of different numbers of infants with ‘unknown’ iron status (based on contradictory results of iron status parameters, as detailed in the Methods section).
No apparent association between CHr at 3–4 months corrected age and GA at birth, SDSw at birth, weight gain since birth and gender
There were no correlations between CHr at 3–4 months of age and GA at birth (r=−0.11, n=209, p=0.12, 95% CI −0.24 to 0.03), between CHr and birth weight (r=−0.07, n=209, p=0.34, 95% CI −0.20 to 0.07), and between CHr and weight gain since birth (r=0.06, n=209, p=0.40, 95% CI −0.08 to 0.19). CHr levels were similar in infants with SDSw at birth <−2 versus ≥−2 (n=48, 29.7 (28.6–31.0) pg), versus n=161, 29.7 (28.6–30.6) pg, median difference (95% CI) −0.1 (−0.6 to 0.5), p=0.73), and there was no gender difference in CHr (female: n=110, 29.8 (28.7–30.6) pg, vs male: n=99, 29.5 (28.3–30.7) pg, median difference (95% CI) −0.2 (−0.6 to 0.3), p=0.45).
Multivariate analysis of covariance showed that GA, SDSw at birth, gender and weight gain since birth were not associated with CHr (all p values >0.05). This model explained only about 2% of the variability of CHr (data not shown).
Discussion
This is the first study providing cut-off values for CHr to detect ID in former very preterm and VLBW infants at 3–4 months corrected age.
Previously reported CHr cut-off values (at 2.5th percentile) of preterm infants at birth (mean GA: 30 weeks) were at 31.5 pg.25 CHr cut-off values (at 2.5th percentile) of term infants in cord blood were 33.3 pg26 and at age 6 weeks were 27.7 pg, at 4 months 23.1 pg,27 at 6 months 21.6 pg,27 at 10 months 23.6 pg,28 at 3 years 23.7 pg29 and at 5 years 23.2 pg.30
Our study population had similar CHr values (at 2.5th percentile: 23.2 pg) compared with iron-supplemented (9 mg/d) marginally low birthweight infants (24.5 pg)27 and not iron-supplemented term infants (23.1 pg) at the age of 4 months,27 indicating that adequately iron-supplemented preterm infants have similar iron status compared with healthy term infants. Similarly, the incidence of ID at the age of 6 months was greater among non-iron-supplemented term infants compared with iron-supplemented marginally low birthweight infants.27
Moreover, in this study population, there was no association between CHr and GA at birth and SGA status at birth. Thus, provided that adequate iron supplementation is administered, it seems that the CHr cut-off value at 3–4 months corrected age does not have to be adjusted for GA or birth weight. By contrast with non-iron-supplemented infants in a previous study,27 higher weight gain was not associated with lower CHr levels in iron-supplemented very preterm and VLBW infants in this study.
For the classification of ID versus non-ID infants, developmental changes in red blood cell count and iron status parameters over the first 12 postnatal months were considered, and appropriate cut-off values for the definition of ID chosen.18 We determined a CHr cut-off value of 29.0 pg as ‘ideal’ to identify ID using a statistical method based on an optimal overall combination of sensitivity and specificity and verifying whether the resulting sensitivity and negative predictive value would be appropriate in our context. This is a more liberal CHr cut-off compared with the aforementioned 2.5th percentile and compared with 26.9 pg (determined at 4 months as a sensitive predictor of anaemia at 6 months) as described by Torsvik et al.27 Since anaemia is a late sign of ID, and even latent untreated ID in infancy may contribute to neurodevelopmental impairment that cannot be corrected by later iron supplementation (reviewed5), we aimed to determine a CHr cut-off to detect latent ID rather than iron-deficient anaemia. Moreover, the cut-off for CHr was selected to achieve a high sensitivity (to identify most infants with ID) and a high negative predictive value, accepting that by applying this cut-off, some infants may be treated without having ID. This seems justifiable because iron supplementation does not seem to have adverse effects in preterm infants in developed countries.6 ,10 ,13 ,20
CHr estimates the amount of Hb in reticulocytes and is an indirect measure of the functional iron availability for current red blood cell production in bone marrow (reviewed31) and, therefore, an early indicator of ID. By contrast, a decrease in Hb is a late finding of ID. CHr can also identify a response to enteral iron supplementation much earlier than Hb.32 ,33
Similar to other studies,14 ,29 CHr seemed to have higher diagnostic accuracy in detecting ID than ferritin, TS and MCV and, therefore, may be suitable as a screening tool to identify ID and to optimise iron supplementation. When comparing the AUC of the different iron status parameters, strict (and for each comparison, identical) definitions of ID were chosen, resulting in smaller sample sizes and, hence, loss of power, which probably can explain why quite large differences in AUC did not reach statistical significance.
Since <0.5 mL blood is needed to determine complete red blood cell count and reticulocyte parameters within a few minutes, determination of CHr requires less blood than other iron status parameters at competitive costs. In fact, the cost-effectiveness of anaemia prevention with CHr screening has been evaluated and was considered to be an affordable strategy.34
Determination of CHr requires a suitable haematological analyser. Moreover, CHr has diagnostic limitations: it can be falsely low in case of haemoglobinopathies causing microcytic anaemia-like β thalassemia35 or falsely high in patients with megaloblastic anaemia.36 However, the prevalence of haemoglobinopathies in Germany is low (1/200–1/500),37 and our patients routinely receive supplements of folic acid and Vitamin B12 according to recent recommendations. It is, therefore, unlikely that the results presented here are confounded by either haemoglobinopathy or megaloblastosis.
Hence, CHr appears to be useful in monitoring iron status in former preterm infants, in whom iron supplementation should be continued for at least 1 year after hospital discharge,9 but further studies are required to confirm that CHr-guided individualisation of iron supplementation in former very preterm infants will improve haematological (and eventually long-term) outcomes.
Conclusion
CHr with a cut-off level of 29.0 pg was a sensitive and specific marker to identify ID in former very preterm and VLBW infants at 3–4 months corrected age. Further studies are needed to evaluate if the use of CHr to detect ID followed by an adjustment of the iron supplementation dosage decreases the risk of ID-anaemia and long-term neurocognitive deficits in former preterm infants.
Acknowledgments
The authors thank Renate Fundel for her valuable assistance and her dedicated fieldwork, and gratefully acknowledge the support by Dr Corinna Engel (Center for Pediatric Clinical Studies) in performing multivariate statistical analyses.
References
Footnotes
-
Contributors LL designed the study, analysed the data and wrote the first draft of the manuscript; KB, AW-G collected data and blood samples and contributed to the writing of the manuscript; JA and AP collected and analysed the data, and contributed to the manuscript; CFP contributed to the study design and the manuscript; ARF designed the study, contributed to data and blood sample collection and data analysis and wrote the manuscript.
-
Funding AKF-grant number 283-0-0 from the Faculty of Medicine, University of Tübingen.
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.