Soluble transferrin receptor-ferritin index (TfR-F index) and the Thomas plot have high sensitivity and specificity for ID in anaemic patients with concomitant inflammation [
15,
16], and bone marrow aspiration with subsequent iron staining can be used as a supplementary diagnostic feature for ID during inflammation [
3‐
5]. Currently, the best way to identify iron deficiency during concurrent inflammation is probably to use the TfR-F index, that is, sTfR/log ferritin. In a study by Punnonen and co-workers [
15] AUC
ROC for this index was 1.0 when identifying anaemic patients with depleted iron stores. This index is also superior to analysis of ferritin alone in the diagnosis of iron depletion in patients with inflammatory bowel disease [
17] and in the prediction of responsiveness to intravenous iron supplementation in patients with chronic kidney disease [
18]. In addition, in a prospective multicenter evaluation of the TfR-F index it was shown that this ratio was superior to both analysis of ferritin and of soluble transferrin receptor in the diagnosis of iron depletion in patients with concurrent AI [
19]. Since we did not analyse soluble transferrin receptor in this study, we used bone marrow aspiration as a supplementary diagnostic feature. The importance of using this criterion for ID in our patient population is evident when considering that the mean ferritin in the ID-AI population was 37 μg/L. If the conventional ferritin cut-offs for ID of 15 and 25 μg/L for women and men, respectively, had been used, 80% (8/10) of the iron-deficient patients in this study would have been misdiagnosed as iron replete. A ferritin cut-off for ID of 30 μg/L is commonly used in different clinical situations and in research [
20‐
22] but even if this ferritin cut-off had been used 60% of the patients with ID in the present study would have been classified as iron replete. The optimal ferritin cut-off for ID in this study was 87 μg/L. This cut-off value is approximately two times higher than we observed in a previous study with a similar design [
10,
11] and that others have observed [
15,
23]. The reason for this difference is not known. Mass spectrometry analysis of hepcidin is specific for biologically active hepcidin-25, and differentiates between patients with IDA and AI [
24,
25]. In contrast to the ELISA developed by Butterfield and co-workers. [
26], the competitive ELISA assay used in this study is not specific for hepcidin-25 [
27], but it cross-reacts with hepcidin-20 and -22. However, hepcidin levels determined by this assay correlated significantly with hepcidin-25 determined by mass spectrometry in the study by Dahlfors and co-workers [
27]. Due to cross-reactivity with hepcidin-20 and -22, the level of hepcidin was approximately 30% higher when hepcidin was analysed using the ELISA assay compared to the level using hepcidin-25 specific mass spectrometry [
27]. Although not specific for hepcidin-25, the mean hepcidin level was significantly lower in subjects with ID compared to healthy controls using this ELISA assay [
27]. In the study reported here, we show a significant difference in hepcidin levels when comparing patients with ID-AI to those with AI. The mean hepcidin levels were 8.5 and 44 μg/L for the ID-AI and AI populations, respectively. The mean hepcidin was higher in the ID-AI population than that in the iron-deplete subjects in the Dahlfors study [
27], probably because the iron-deplete patients in our study suffered from concomitant inflammation. The problem of identifying an underlying ID in the context of AI has been addressed in several populations, using ELISA assays. For example, the first commercially available hepcidin ELISA assay did not differentiate between IDA and AI in geriatric patients [
28]. To our knowledge, only three other such published studies addressing this problem have used the absence of stainable bone marrow iron as the diagnostic criterion for ID [
29‐
31]. In two of these studies hepcidin levels were considerably higher in all patient groups than we observe in this report. In addition, Shu et al. (30) reported an optimal hepcidin cut-off for ID of 83 μg/L, approximately four times higher than the optimal cut-off of 21 μg/L reported in this study. In a recent study using mass spectrometry-based hepcidin analysis we showed an optimal hepcidin cut-off for ID of 31 μg/L in a similar patient population with ID-AI or AI (11). One possible explanation for this difference is that another ELISA kit (Sandwich ELISA Uscn Life Science; hepcidin antibody E1979HU; mean hepcidin concentration 79 μg/L in controls) was used [
29,
30]. In the third study, Barsan and co-workers used the same hepcidin ELISA kit as we did. They did not find that hepcidin was superior to ferritin analysis in its ability to differentiate between iron deficiency anaemia and anaemia of inflammation in patients with chronic kidney disease [
31]. This is in agreement with our results reported here. However, in that report [
31] the optimal hepcidin cut-off for ID was 82 μg/L. This discrepancy can at least to some extent be explained by the fact that hepcidin is excreted in the urine, and that patients with renal failure have higher levels of hepcidin in serum than healthy controls [
9]. Applying a hepcidin cut-off of approximately 80 μg/L on our patient population would yield a much lower specificity for ID and a lower Yuoden Index. Sensitivity and specificity data for ID for different hepcidin cut-offs are presented in Table
5. We also confirm the positive correlation between expression of hepcidin and ferritin that we and others have described previously [
9,
11,
29]. IL-6 is the main positive regulator of hepcidin expression [
32], but we did not find a significant difference in mean IL-6 levels between patients with ID-AI and AI and only a weak positive correlation between hepcidin and IL-6. This indicates that other pro-inflammatory cytokines are involved in the regulation of hepcidin, since we did detect a significant difference in CRP between the two groups. It has been shown previously that increased erythropoiesis rather than serum iron negatively regulates hepcidin expression [
33,
34]. Since we could not detect any differences in iron or reticulocyte count between the ID-AI and AI groups, we suggest that hepcidin is mainly regulated positively by iron stores and inflammation in the patient population reported here. We have previously shown that hepcidin analysis by mass spectrometry does not appear to be superior to ferritin in the differential diagnosis of ID with concurrent inflammation and AI in elderly patients [
11]. In conclusion, in the present report we show that hepcidin analysed by a competitive ELISA is not superior to ferritin, using a modified optimal ferritin cut-off, in detecting iron deficiency in this patient category.
Table 5
Sensitivity, specificity, predictive values and Yuoden Indices for different hepcidin cut-offs for iron deficiency
18.5 | 82 | 78 | 79 | 81 | 0.60 |
19.5 | 81 | 67 | 71 | 79 | 0.48 |
21.0 | 100 | 67 | 75 | 100 | 0.67 |
24.5 | 100 | 56 | 69 | 100 | 0.56 |
35.0 | 100 | 50 | 67 | 100 | 0.50 |
76.5 | 100 | 17 | 55 | 100 | 0.17 |
89.0 | 100 | 11 | 53 | 100 | 0.11 |