- Top of page
Objectives Aboriginal children in tropical Australia have a high prevalence of both iron deficiency and acute infections, making it difficult to differentiate their relative contributions to anaemia. The aims of this study were to compare soluble transferrin receptor with ferritin in iron deficiency anaemia (IDA), and to examine how best to distinguish the effect of iron deficiency from infection on anaemia.
Methods We conducted a prospective study of 228 admissions to Royal Darwin Hospital in children from 6 to 60 months of age. Transferrin receptor concentrations were measured by a particle-enhanced immunoturbidimetric assay and ferritin by a microparticle enzyme immunoassay.
Results On multiple regression, the best explanatory variables for haemoglobin differences (r2 = 33.7%, P < 0.001) were mean corpuscular volume (MCV), red cell distribution width (RDW) and C-reactive protein (CRP); whereas transferrin receptor and ferritin were not significant (P > 0.4). Using ≥2 abnormal indices (MCV, RDW, blood film) + haemoglobin <110 g/l as the reference standard for IDA, transferrin receptor produced a higher area under the curve on receiver operating characteristic curve analysis than ferritin (0.79 vs. 0.64, P < 0.001) or the transferrin receptor–ferritin index (0.77). On logistic regression, the effect of acute infection (CRP) on haemoglobin was significant (P < 0.001) at cut-offs of 105 and 110 g/l, but not at 100 g/l when only iron deficiency indicators (MCV, RDW, blood film) were significant.
Conclusions Transferrin receptor does not significantly improve the diagnosis of anaemia (iron deficiency vs. infection) over full blood count and CRP, but in settings with a high burden of infectious diseases and iron deficiency, it is a more reliable adjunctive measure of iron status than ferritin.
- Top of page
Anaemia, iron deficiency and infections are common problems for Aboriginal children living in remote areas of the Northern Territory (Kruske et al. 1999). Iron deficiency is also an important contributor to enteropathy (Molla et al. 1973; Berant et al. 1992), which is another major health concern in Aboriginal children in tropical Australia (Kukuruzovic & Brewster 2002; Kukuruzovic et al. 2002, 2003). The diagnosis of iron deficiency anaemia (IDA) in Aboriginal children can be difficult as the high burden of acute infectious disease may lower the haemoglobin and confound the laboratory markers used to assess iron stores as inflammation and infection elevate ferritin levels, depress serum iron and reduce transferrin saturation levels (Reeves et al. 1984; Olivares et al. 2000).
The transferrin receptor is a membrane glycoprotein which facilitates the uptake of iron into cells (Ponka & Lok 1999). The plasma contains small amounts of soluble transferrin receptor, which is regulated by cellular iron demands and erythropoietic activity (Feelders et al. 1999). Unlike ferritin, which is a marker of body iron stores, and serum transferrin which reflects iron transport in blood, transferrin receptor closely reflects the functional iron needs of the tissues (Ahluwalia 1998). The concentration of transferrin receptor has now been validated as a quantitative measure of functional iron deficiency and has proven to be a reliable diagnostic tool to assess iron status in children (Dimitriou et al. 2000). Transferrin receptor has distinct advantages over ferritin in our setting as it is unaffected by an acute phase response, and hereditary haematological disorders such as sickle cell anaemia, thalassaemia and other chronic haemolytic anaemias that also increase transferrin receptor concentrations are rare in Aboriginal people.
The aim of this study was to compare the diagnostic accuracy of transferrin receptor and ferritin using a reference standard based on haematological measurements of iron deficiency. A secondary aim was to see whether transferrin receptor could distinguish IDA from anaemia of acute infection (inflammation) better than the existing protocol of reliance upon mean corpuscular volume (MCV). This is an important clinical issue in our setting because IDA cases are treated with iron injections (Kruske et al. 1999), which are not without risk, so we were seeking evidence to restrict iron injections to true cases of iron-deficient erythropoiesis causing anaemia rather than iron-deficient children with anaemia associated with acute infection.
- Top of page
Children admitted to the Royal Darwin Hospital between the ages of 6 and 60 months who had blood collected for full blood count (FBC) and C-reactive protein (CRP) on presentation were eligible for the study. Exclusion criteria were chronic haemolytic anaemia, chronic infection or those who had received parenteral or oral iron therapy within 3 months of admission (n = 13). Nearly all Aboriginal children admitted to Royal Darwin Hospital are from remote communities in the Top End of the Northern Territory, whereas non-Aboriginal children are mainly from the urban area. Although a convenience sample, our hospital subjects had typical disease and demographic characteristics of all acute paediatric admissions from our catchment population, both Aboriginal and non-Aboriginal.
On admission a total of 1.5 mL of venous blood was collected from each subject, of which 1 mL was placed into a paediatric lithium heparin tube for CRP, serum ferritin and transferrin receptor analysis, and 0.5 mL into a paediatric ethylenediaminetetraacetic acid (EDTA) tube for FBC. Hospital records of all patients in the study were reviewed during the admission and basic demographic data including age, sex, ethnicity and place of residence recorded. Primary diagnosis was documented after discharge from the case record summary. Community records were examined by nursing staff from remote clinics to help identify children who had received oral or parenteral iron therapy. The project was approved by the Human Research Ethics Committee of NT Department of Health and Community Services and Menzies School of Health Research, including an Aboriginal subcommittee.
Samples for FBC and CRP were analysed immediately after collection, while plasma ferritin specimens were batched and measured within 72 h. Transferrin receptor samples were frozen for assay in batches. Automated red cell, white cell, platelet counts, haemoglobin concentration and MCV were performed on a Coulter MaxM (Beckman Coulter Inc., Fullerton, CA, USA) using EDTA-anticoagulated plasma.
Soluble transferrin receptor concentrations were measured using the Orion IDeA transferrin receptor IT kit (Orion Corporation Orion Diagnostica, Espoo, Finland) adapted to the Immage Immunochemistry System (Beckman Coulter Inc.). The manufacturer's normal range for the test is <3.3 mg/l, with iron-deficient children showing elevated values. In a pilot study, we compared transferrin receptor concentrations measured in serum, plasma heparin and plasma EDTA in 31 normal subjects. Serum had the consistently lowest values with a mean of 1.73 (1.31–2.16) followed by plasma heparin at 1.95 (1.50–2.39) and EDTA at 2.18 (1.60–2.73), which were not significantly different (P = 0.4). This report is based on results for plasma heparin, which may differ from serum and EDTA results in other reports.
Ferritin concentrations were assayed in heparinized plasma using an Abbott AxSYM Chemistry System and Abbott Ferritin Reagent Pack (Abbott Laboratories, IL, USA). CRP concentrations in heparin plasma were determined by dry-chemistry immunoassay, using Vitros CRP slide kits (CAT8723496) and Vitros Chemistry System (Ortho-Clinical Diagnostics, Johnson & Johnson Company, Rochester, NY, USA). Laboratory staff performing transferrin receptor and ferritin concentrations were blinded to clinical details and other results.
Data were entered and analysed using Stata 7.0 (http://www.stata.com, College Station, TX, USA). Normally distributed data are expressed as arithmetic means with 95% confidence intervals (CI) and skewed data as geometric means with 95% CI after logarithmic transformation (for transferrin receptor, ferritin, CRP and age). Continuous variables with homogeneity of variance were compared using the t-test or anova, and categorical variables using χ2. Multiple linear regression analyses were carried out on haemoglobin as a continuous variable, to establish which variables were associated with differences in haemoglobin. Stepwise multiple logistic regression analysis was applied to different haemoglobin cut-off values as dependent variables to evaluate the contribution (expressed as odds ratios) of normally distributed continuous independent variables such as transferrin receptor, ferritin, MCV and CRP. Figure 1 was constructed using Prism 3.0 (GraphPad Software Inc., San Diego, CA, USA).
Figure 1. Receiver operating characteristic (ROC) curve for various cut-off levels of transferrin receptor and ferritin in differentiating iron deficiency anaemia (haemoglobin <110 g/l, with at least two of: MCV < 70 fL, RDW > 14.5 and hypochromic-microcytic red cells, n = 49) from no iron deficiency anaemia (n = 179). Transferrin receptor produced a significantly higher AUC (0.79; 0.71–0.86) than ferritin (0.64; 0.54–0.73, P = 0.001).
Download figure to PowerPoint
- Top of page
This study was conducted over a 12-month period from January 2002 until January 2003 and included a total of 228 children. The baseline characteristics of the 173 (75.8%) Aboriginal and 55 (24.2%) non-Aboriginal children are shown in Table 1. Aboriginal admissions were younger with significantly lower MCV, higher RDW and more hypochromic-microcytic blood films reflecting the higher prevalence of iron deficiency in Aboriginal children, yet surprisingly did not have significantly higher transferrin receptor or lower ferritin values than non-Aboriginal children. Transferrin receptor was not significantly related to age (P = 0.16), sex (P = 0.23) or CRP (P = 0.7) differences in the group of 119 subjects with normal indices, but ferritin levels varied directly with CRP (P < 0.001) in subjects with both normal and abnormal indices. The mean (range) haemoglobin in our subjects was 113 g/l (62–177 g/l) and was normally distributed.
Table 1. Admission characteristics in Aboriginal and non-Aboriginal children
|Characteristic||Aboriginal (n = 173)||Non-Aboriginal (n = 55)||P-value†|
|Mean age (months)|| 15.0 (13.7–16.4)|| 20.0 (17.1–23.5)||0.001|
|Sex (male)|| 94 (54.3%)|| 35 (63.6%)||0.22|
| Acute gastroenteritis|| 52 (30.0%)|| 10 (18.1%)||0.08|
| Lower respiratory infection|| 47 (27.1%)|| 11 (20%)||0.28|
| Soft tissue infection|| 8 (4.6%)|| 1 (1.8%)||0.35|
| Viral infection|| 11 (6.3%)|| 7 (12.7%)||0.12|
| Seizure|| 2 (1.1%)|| 3 (5.4%)||0.05|
| Malnutrition|| 9 (5.2%)|| 0||0.08|
|Clinical infection||143 (82.6%)|| 30 (76.3%)||0.29|
|WBC toxic changes|| 79/132 (59.8%)|| 19/41 (46.3%)||0.12|
|Haemoglobin (g/l)||112.4 (110.0–114.8)||115.7 (111.0–119.1)||0.16|
|Haemoglobin <110 g/l|| 68 (39.3%)|| 17 (30.9%)||0.26|
|MCV (fL)|| 71.4 (70.3–72.4)|| 76.2 (74.4–78.0)||<0.001|
|MCV <70 fL|| 61 (35.2%)|| 7 (12.7%)||0.001|
|Blood film*|| 75/132 (56.8%)|| 11/41 (26.8%)||0.001|
|RDW (%)|| 16.2 (15.6–16.8)|| 14.3 (13.8–14.9)||<0.001|
|RDW >14.5%||104 (60.1%)|| 21 (38.2%)||0.004|
|Transferrin receptor (mg/l)|| 2.8 (2.6–3.0)|| 2.4 (2.2–2.7)||0.07|
|Transferrin receptor >3.3 mg/l|| 57 (33.0%)|| 14 (25.5%)||0.30|
|Ferritin (μg/l)|| 56.3 (47.9–66.7)|| 47.0 (35.2–62.2)||0.26|
|Ferritin ≤20 μg/l|| 26 (10.4%)|| 10 (12.7%)||0.57|
|CRP (mg/l)|| 20.3 (16.9–24.3)|| 19.3 (14.4–25.7)||0.76|
|CRP >8 mg/l||121 (69.9%)|| 35 (63.6%)||0.38|
In order to explain the relative contributions of iron deficiency and infection on haematopoiesis in our setting, we examined the correlations of haemoglobin with MCV, RDW, transferrin receptor, ferritin and CRP. On stepwise multiple linear regression of haemoglobin level (Table 2), the best model which explained 33.7% of the variance (r2) included MCV (P < 0.001), RDW (P < 0.001) and CRP (P = 0.002), adjusted for age and aboriginality, with no significant interactions between variables. Adding transferrin receptor and/or ferritin did not improve the model, but there were significant interactions (P = 0.001) between transferrin receptor and MCV. Yet, substituting transferrin receptor in the model for MCV (excluding RDW) explained significantly less of the variability in haemoglobin (r2 reduced from 26.4% with MCV to 13.2% with transferrin receptor). On logistic regression of haemoglobin (as a dichotomous variable), the relative contributions of MCV, RDW, transferrin receptor and ferritin (for iron deficiency) compared with CRP (for infection) were assessed at various cut-off levels of haemoglobin (e.g. <100 vs.≥100 g/l) in all 228 subjects (Table 3). The results confirm that MCV, RDW and CRP are the best explanatory variables for haemoglobin differences, with the highest area under the curve (AUC) (0.82) for a haemoglobin cut-off of 100 g/l. Infection (CRP), on the other hand, is only significantly associated with anaemia at higher haemoglobin cut-off levels (105–115 g/l) but with no significant effect for more severe anaemia (<100 g/l). Thus, the FBC model of MCV and RDW combined with CRP correlated better with changes in haemoglobin from iron deficiency than the transferrin receptor and CRP model (Table 3). This is confirmed on logistic regression of transferrin receptor (Table 4) which shows that high levels correlate with MCV, RDW and ferritin, but not with haemoglobin.
Table 2. Linear regression of haemoglobin (g/l) for single unadjusted variables and adjusted for multivariate estimation (n = 228, d.f. = 5, r2 = 33.7%)
|β-Coefficient (±SEM)||r2 (%)||P-value||β-Coefficient (±SEM)||t-value||P-value|
|MCV (fL)||1.03 ± 0.12||23.8||<0.001||0.57 ± 0.16||3.63||<0.001|
|RDW (%)||−2.23 ± 0.25||25.8||<0.001||−1.56 ± 0.32||−4.95||<0.001|
|CRP* (mg/l)||−2.44 ± 0.87||3.4||0.006||−2.35 ± 0.74||−3.20||0.002|
|Age* (months)||3.38 ± 1.68||1.8||0.05||−0.30 ± 1.48||−0.20||0.840|
|Aboriginality (0, 1)||−3.35 ± 2.39||0.86||0.16||2.37 ± 2.08||1.14||0.256|
|Transferrin receptor*||−10.05 ± 2.06||9.5||<0.001|| || || |
|Ferritin* (μg/l)||0.95 ± 0.95||0.44||0.32|| || || |
|Blood film (0, 1)†||−9.76 ± 2.30||9.5||<0.001|| || || |
Table 3. Odds ratios from logistic regressions of four different haemoglobin (Hb) cut-off values for the relative contributions of iron deficiency (FBC, transferrin receptor, ferritin) and infection (CRP) on haematopoiesis (n = 228)
| ||Cut-off categories|
| ||Hb 100 g/l||Hb 105 g/l||Hb 110 g/l||Hb 115 g/l|
|Numbers†||30 vs. 198||56 vs. 172||85 vs. 143||119 vs. 109|
|FBC model (gold standard)|
| MCV||0.90 (0.83–0.97)**||0.89 (0.84–0.95)***||0.92 (0.87–0.97)**||0.93 (0.88–0.98)*|
| RDW||1.29 (1.12–1.50)***||1.22 (1.07–1.39)**||1.21 (1.07–1.38)**||1.15 (1.01–1.30)*|
| CRP||1.41 (0.94–2.10)||1.74 (1.26–2.39)***||1.77 (1.33–1.37)***||1.63 (1.24–2.13)***|
|Transferrin receptor model|
| Transferrin receptor||4.74 (2.00–11.27)***||3.05 (1.51–6.14)**||3.21 (1.63–6.33)**||2.15 (1.15–4.01)*|
| CRP||1.24 (0.88–1.75)||1.52 (1.16–2.00)**||1.64 (1.27–2.13)***||1.59 (1.23–2.05)***|
| Ferritin (μg/l)||0.71 (0.46–1.09)||0.65 (0.46–0.93)*||0.63 (0.46–0.86)**||0.72 (0.54–0.95)*|
| CRP||1.38 (0.95–2.01)||1.78 (1.30–2.44)***||1.96 (1.46–2.64)***||1.82 (1.37–2.41)***|
Table 4. Logistic regression of soluble transferrin receptor (≥3.3 vs. <3.3 mg/l) for single unadjusted variables and adjusted for multivariate estimation (n = 228, AUC = 0.78)
|Odds ratio||CI||AUC||P-value||Odds ratio||CI||P-value|
|MCV (<70 fL)||8.50||4.4–16.3||0.72||<0.001||4.27||2.0–8.8||0.001|
|Ferritin (<20 μg/l)||5.31||2.5–11.3||0.62||<0.001||3.61||1.5–8.3||0.004|
|Blood film (0, 1)*||5.21||2.6–10.6||0.69||<0.001|| || ||0.63|
|Haemoglobin (<110 g/l)||2.47||1.4–4.4||0.61||0.002|| || ||0.52|
|CRP (>20 mg/l)||1.11||0.6–2.0||0.51||0.71|| || ||0.22|
|Age (months)||0.60||0.4–0.96||0.59||0.04|| || ||0.99|
|Aboriginality (0, 1)||1.44||0.7–2.9||0.53||0.30|| || ||0.42|
As a result of these regressions, we chose the combination of haemoglobin (<110 g/l) and at least two of the indices MCV <70, RDW >14.5 and hypochromic-microcytic red cells on film as our gold standard for IDA. The capacity of ferritin and transferrin receptor to accurately predict IDA in our setting was assessed by receiver operating characteristic (ROC) curve analysis from the 228 observations, comparing the AUC. Among anaemic children, transferrin receptor produced a significantly higher AUC than ferritin (0.85 vs. 0.70, P = 0.004). For all 228 subjects, the AUC for transferrin receptor vs. ferritin was 0.79 vs. 0.64 (P < 0.001, Figure 1). The combination of ferritin and transferrin receptor (TfR-F index) did not improve the AUC over transferrin receptor alone (0.77 vs. 0.79, P = 0.43). On logistic regression analysis of IDA, high transferrin receptor (≥3.3 mg/l) was a better predictor (odds ratio 6.1; 2.8–13.2, P < 0.001) than ferritin (0.65; 0.43–0.99, P = 0.04), adjusted for CRP, aboriginality and age. Further evidence against ferritin was that only 13 of 42 (31.7%) Aboriginal IDA subjects had a ferritin ≤20 μg/l. The mean ferritin in IDA cases was 37.0 μg/l (25.1–54.6) compared with 60.0 (51.9–69.3) for non-IDA children (P < 0.001). Increasing the ferritin reference range to ≤50 μg/l did not improve significantly the low sensitivity (6–9%).
Transferrin receptor levels were significantly higher (P < 0.001) in IDA cases with a geometric mean of 4.0 (3.3–4.8) compared with 2.4 (2.3–2.5) for non-IDA cases. Using the ROC curve for IDA at various cut-off levels of transferrin receptor and ferritin (Figure 1), the best cut-off for soluble transferrin receptor to diagnose IDA was 3.3 mg/l (sensitivity 63%, specificity 79%, positive likelihood ratio 3.0), which corresponds to the manufacturer's recommendation for this test. However, for iron deficiency (with or without anaemia) the best cut-off was lower at 2.7 mg/l with sensitivity 67%, specificity 64%. Consequently, we examined whether an intermediate range of transferrin receptor (2.7–3.2) was detecting iron deficiency at an earlier stage than the FBC gold standard. Table 5 shows that the 42 children with intermediate values have similar FBC values to the low transferrin receptor group, and ferritin showed only a non-significant trend (P = 0.29) to intermediate values (if stratified for CRP). A third of these 42 children were already identified as iron deficient by FBC criteria, but the remaining 30 children had a mean ferritin of 73.0 μg/l (56.8–93.1, range 28–475) and mean CRP of 24.3 (15.9–37.7).
Table 5. Low, intermediate and high values for soluble transferrin receptor
| ||Low transferrin receptor (≤2.6 mg/l)||Intermediate transferrin receptor (2.7–3.2)||High transferrin receptor (≥3.3)|
|Aboriginal, n (%)|| 83 (72)||33 (78)||57 (80)|
|Iron deficient*, n (%)|| 29 (25)||12 (38)||49 (69)|
|MCV (fL)|| 75.4 (74.4–76.5)||74.2 (72.4–76.0)||67.0 (65.2–68.8)|
|Haemoglobin (g/l)||114.3 (112.0–116.7)||118.0 (114.3–121.7)||108.6 (103.9–113.4)|
|Ferritin (μg/l)||(n = 64 vs. 51)||(22 vs. 20)||(37 vs. 34)|
| If CRP < 20||52.4 (40.6–67.6)||40.3 (26.4–61.7)||24.5 (16.2–37.0)|
| If CRP ≥ 20||103.2 (81.0–131.4)||81.9 (59.5–112.7)||49.2 (35.1–69.1)|
As our results clearly showed that transferrin receptor performed better than ferritin, we compared four different clinical classifications based upon tests of iron status (MCV, RDW, transferrin receptor and blood film) and anaemia (Table 6). The level of agreement (κ) between the FBC gold standard and transferrin receptor classifications was 0.59 compared with 0.43 for ferritin. The best agreement with transferrin receptor was a combination of abnormal MCV + RDW (0.63). All of the classifications confirm that younger Aboriginal children are likely to be iron deficient, and that CRP is high in non-iron-deficient anaemia, which is related to acute infection.
Table 6. Four diagnostic groups based upon haemoglobin, red cell mean corpuscular volume (MCV), red cell distribution width (RDW), blood film appearance and soluble transferrin receptor (STR)
|Classification||Normal||Iron deficiency||IDA||Non-IDA anaemia|
|Gold standard||Haemoglobin ≥110 g/l||Haemoglobin <110 g/l|
| || ||≥2 indicators of iron deficiency|| |
|Aboriginal, n (%)*||64 (65)||41 (91)||42 (85)||26 (77)|
|Age (months)**||19.1 (16.8–21.7)||13.2 (11.5–15.2)||12.7 (11.1–13.3)||17.5 (14.3–21.4)|
|Transferrin receptor**||2.4 (2.2–2.6)||2.8 (2.5–3.1)||4.0 (3.3–4.6)||2.2 (1.9–2.5)|
|Ferritin**||61.3 (57.5–73.1)||48.8 (34.1–69.8)||37.0 (25.1–54.6)||73.2 (53.3–100.5)|
|CRP**||16.8 (13.5–21.0)||14.4 (10.4–19.3)||27.8 (17.6–33.9)||58.1 (24.4–59.6)|
|Transferrin receptor||Haemoglobin ≥110 g/l||Haemoglobin <110 g/l|
| || ||Transferrin receptor ≥3.3 mg/l|| |
|Aboriginal, n (%)||79 (72.5)||26 (76.5)||31 (83.8)||37 (77.1)|
|Age (months)||17.4 (15.5–19.6)||15.9 (13.0–19.6)||12.8 (10.5–15.6)||16.6 (14.0–19.5)|
|MCV**||75.9 (74.8–76.9)||71.2 (69.3–73.1)||63.1 (60.7–65.2)||73.3 (71.9–74.9)|
|Ferritin**||60.6 (50.7–72.5)||47.2 (32.0–69.5)||25.5 (17.3–37.3)||82.5 (60.8–111.9)|
|RDW**||14.3 (13.9–14.7)||16.2 (15.3–17.0)||19.4 (17.7–21.0)||16.0 (15.0–16.9)|
|CRP**||15.1 (12.4–18.5)||19.6 (13.6–28.3)||22.3 (15.4–32.3)||36.7 (25.3–53.3)|
|Abnormal blood film, n (%)*||24 (31.2)||12 (54.6)||31 (88.6)||19 (48.7)|
Among Aboriginal children, iron deficiency was present in 61.0% and inflammation (CRP ≥ 20 mg/l) in 55.8% among anaemic children, compared with 38.1% and 38.1%, respectively, without anaemia. Out of the 85 anaemic children in the study, both iron deficiency and inflammation were present in 25 (30%) and neither in 12 (14%), so it was in these two groups that soluble transferrin receptor might add to existing tests. Two of 11 (16%) anaemic children with neither IDA nor inflammation had a high transferrin receptor (≥3.3 mg/l) consistent with mild IDA, whereas 10 of 25 (40%) with both IDA and inflammation had normal transferrin receptor levels (<3.3 mg/l) consistent with acute infection rather than IDA as the cause of anaemia. Thus, transferrin receptor enabled possible reclassification of 12 of 85 (14.1%) anaemic children, but it has low sensitivity (65%) for IDA and even lower for iron deficiency (52%) compared with the FBC gold standard.
Figure 2 is a flow chart which shows the diagnostic implications of the study findings. We believe that (a) a high transferrin receptor may help classify non-IDA cases with normal CRP to IDA (two of our cases) and (b) a normal transferrin receptor may help classify IDA with high CRP to anaemia of infection (10 of our cases). Our flow chart left 10 cases unclassified, which upon examining all the evidence, we diagnosed two with mild IDA, seven as probably normal (haemoglobin 105–109 g/l) and one remained unclassified.
Figure 2. A diagnostic protocol for iron deficiency anaemia from the results of the study. Numbers in square brackets are numbers of children in the study as classified.
Download figure to PowerPoint
- Top of page
Anaemia in our paediatric admissions tends to be related to either iron deficiency or acute infections, with mild and transient infections often reducing the haemoglobin concentration below 110 g/l even in a community study in our setting (Kruske et al. 1999). This community study documented haemoglobin falls below 110 g/l within 12 weeks of iron repletion intramuscularly in association with minor infections not requiring hospitalization in 62% of Aboriginal children. The American National Health and Nutritional Examination Survey has documented the importance of iron deficiency and inflammatory conditions in both anaemic and non-anaemic children in a low prevalence setting (Yip & Dallman 1988). Inherited haemoglobinopathies and chronic haemolytic anaemias have not been documented in Aboriginal children and folate or vitamin B12 deficiency is uncommon with no case of macrocytic anaemia documented in our study. Paediatric admissions in Aboriginal children are nearly all acute admissions from remote communities, so resemble the pattern of admissions in the developing world, except for the absence of malaria.
The presence of microcytosis in anaemic children increases the likelihood of iron deficiency. In this study, microcytosis (MCV < 70 fL) was found in 52.6%vs. 16.4% of Aboriginal vs. non-Aboriginal children, whereas hypochromic-microcytic red cells on blood film were reported in 56.8%vs. 26.8%, respectively (Table 1). Due to the length of RBC survival, MCV is slow to respond to changes in iron status, so one aim of the study was to examine whether transferrin receptor identified IDA cases better than MCV. We were not able to document much benefit of transferrin receptor over MCV, but found that raised RDW and CRP also correlated highly with anaemia.
Measurement of soluble transferrin receptor concentration has been found a reliable adjunctive measure of iron status both in adults (Skikne et al. 1990; Cook et al. 1993; Punnonen et al. 1997; Feelders et al. 1999; Ponka & Lok 1999) and children (Dimitriou et al. 2000; Olivares et al. 2000). It is most useful in differentiating IDA from anaemia of chronic disease in adults (Ferguson et al. 1992; Cook 1999; Ho 2001), and from anaemia of inflammation in children (Malope et al. 2001). Clinical studies have reported that soluble transferrin receptor can detect the development of IDA in both adults and children with chronic infection or inflammation (Pettersson et al. 1994; Suominen et al. 1998; Revel-Vilk et al. 2000). Not all studies, however, have been in agreement, prompting questions about both the discriminating power and clinical usefulness of soluble transferrin receptor above standard laboratory indices of iron deficiency (Junca et al. 1998; Junca & Oriol 2001). Despite claims that soluble transferrin receptor measurement has its greatest potential in the evaluation of iron deficiency where overt or subclinical infection is present (Ahluwalia 1998) only a few large published clinical trials among adults have been conducted in the developing world setting (Kuvibidila et al. 1994, 1996; Zhu & Haas 1998). Even fewer studies are available in children, where this problem is more likely to be encountered. In one study of rural South African children from 1 to 6 years of age, the combination of soluble transferrin receptor and ferritin was shown to clearly differentiate between anaemia of inflammation and IDA but was unable to exclude IDA in the presence of inflammation (Malope et al. 2001). The role of soluble transferrin receptor as an index of iron deficiency in children with malaria also needs to be determined, although it appears from a small study involving 40 Zairian children to be a more sensitive index than serum ferritin (Kuvibidila et al. 1995).
Measurement of ferritin is often used as a confirmatory marker of IDA in hospitalized children. The main disadvantage of this approach is that in the presence of infection or inflammation, ferritin is disproportionately elevated and, by acting as an acute phase reactant, loses sensitivity as an index of iron status (Lipschitz et al. 1974). A simple alternative to abandoning ferritin altogether in our setting would be to increase the cut-off in the presence of infection (Guyatt et al. 1992). A ferritin level of 50 μg/l approximates the upper 95% CI for all children with IDA in our sample, but it still has very low sensitivity. In contrast, we have confirmed that transferrin receptor is not influenced by the presence of infection, represented by the level of CRP.
As ferritin is a good indicator of tissue stores and transferrin receptor of tissue needs, the combination of measurements as the ratio of transferrin receptor/log ferritin (the TfR-F index) has been shown to be a more sensitive index for the diagnosis of IDA than transferrin receptor alone (Malope et al. 2001). By using ROC curve analysis, we were able to compare the overall diagnostic accuracy of the two tests against our reference standard. We found that transferrin receptor alone offered better discriminatory power for IDA than either ferritin or the TfR-F index because it is unaffected by infection. Transferrin receptor reflects tissue iron needs in parallel with changes in MCV, and appears most sensitive when iron-restricted erythropoiesis is present (Ahluwalia 1998).
Another aim of the study was to see how reliable transferrin receptor was in distinguishing IDA from anaemia of acute infection. Consequently, we have used regression techniques to look at the determinants of haemoglobin in this setting of a high burden of iron deficiency, acute infections and no chronic haemolytic anaemias. Our results suggest that transferrin receptor levels have a lower correlation with anaemia than MCV, RDW and blood film, and that CRP as a proxy for infection only contributes to mild anaemia (Hb 105–110 g/l) whereas iron deficiency indices (e.g. low MCV, high transferrin receptor) correlate with more severe anaemia (Table 3).
Measurement of transferrin receptor requires minimal blood volume (200 μL) and is easily performed using commercially available kits. The transferrin receptor assay is neither sensitive nor specific enough to stand alone as a laboratory test of iron deficiency, however it does provide an additional tool to assist in the diagnosis. The subset of patients most likely to benefit from measurement of transferrin receptor is those with suspected iron deficiency and concurrent infection, where ferritin is likely to be normal or raised (Figure 2). In our assessment, transferrin receptor does not provide sufficient benefit over FBC and film to justify the cost of its routine use in the developing world.
The limitations of this study relate to sample selection, the gold standard for IDA and the use of CRP as a proxy for acute infection. This is not a population-based study, but our hospital cases are representative of acute paediatric admissions in Darwin, and are similar to children in the developing world with a high prevalence of iron deficiency and acute infections, except for the absence of malaria. In spite of the hospital bias, our Aboriginal subjects are similar to those in our previous community study (Kruske et al. 1999). We have used as our gold standard for IDA, a haemoglobin <110 g/l with at least two of the following FBC indices: MCV <70 fL, RDW >14.5 and hypochromic-microcytic blood film. Confirmation that iron rather than infection is limiting erythropoieses would require the invasive examination of bone marrow for iron, and a therapeutic trial of iron makes it difficult to differentiate a response to iron from resolving infection (Kruske et al. 1999). We used CRP as a proxy for infection to examine the correlations between anaemia and infection vs. indices of iron deficiency. Although CRP may not be a perfect measure of the impact of infection on haemoglobin, it correlated better with anaemia than clinical infection (P = 0.01 vs.P = 0.53, respectively). However, the association between haemoglobin level and CRP does not prove causality, and there may be unrecorded confounding variables.
In conclusion, our study confirms that the soluble transferrin receptor is a more reliable measure of iron status than ferritin in settings with a high prevalence of acute infections. Does transferrin receptor help us differentiate the effect of iron deficiency from acute infection on haemoglobin level so we can limit the use of iron injections? Our transferrin receptor results suggest that in anaemic children, one in six children (two of 12) with neither IDA (by FBC gold standard) nor high CRP does indeed have IDA, whereas 40% (10 of 25) of IDA cases with high CRP may instead have anaemia due to acute infection (Figure 2). But this assumes that a normal transferrin receptor level with high CRP excludes IDA, which may not be warranted as a third (eight of 24) of IDA cases with low CRP had normal transferrin receptor levels (false negatives). We are not convinced that our results warrant the introduction of soluble transferrin receptor as an additional (or replacement) test in assessing IDA in conjunction with FBC (RDW, MCV, blood film) and CRP. The newly described hormone, hepcidin, is an acute phase reactant made by hepatocytes which appears to regulate iron homeostasis, including enterocyte iron absorption and iron sequestration in macrophages (Ganz 2003; Roy et al. 2003). Inappropriate expression of hepcidin with the high burden of infection seen in Aboriginal children may mediate the anaemia of inflammation and high prevalence of iron deficiency, but further studies are needed to clarify its role.