MRI as an alternative to serum ferritin for diagnosis of iron overload in children in the context of immune response after stem cell transplantation

Multiple blood cell transfusions may cause iron overload or even liver fibrosis, requiring early diagnosis and intervention. SF is the standard for estimating iron levels in the body, but it also increases with inflammation. We hypothesized that T2* magnetic resonance (MR) relaxometry is a more accurate alternative for follow‐up in pediatric patients before and after allogenic SCT. Twenty‐three children (mean age 10.2 years, 10 female, 13 male) were evaluated prospectively before SCT as well as at least 1 year after SCT with T2* relaxometry on a 1.5 T MR‐scanner to estimate liver iron concentrations from the T2* values (“MR‐Fe”). The results were compared with SF, while also considering CRP, and correlated with the number of transfusions. Overall, 24.3 transfusions were administered in average, mainly within 100 days of SCT (mean 10.5 units). Both MR‐Fe and SF increased after SCT and decreased in the absence of new transfusions 1 year later without chelate therapy. This suggests regeneration of LP and iron loss, although the original states were not reached. Additionally, simultaneous peaks of CRP and SF were observed directly after SCT. MR‐Fe did neither reveal these peaks nor was it associated with CRP (P = .39). We postulate that these early CRP and SF peaks after SCT are probably related to inflammatory reactions and not to iron overload. Thus, SF is not reliable for iron overload diagnosis after SCT in every condition. Beside this interaction, SF and MR‐Fe revealed similar accuracy. MRI, however, has practical and economical disadvantages in routine estimation of iron.

1 year after SCT with T 2 * relaxometry on a 1.5 T MR-scanner to estimate liver iron concentrations from the T 2 * values ("MR-Fe"). The results were compared with SF, while also considering CRP, and correlated with the number of transfusions. Overall, 24.3 transfusions were administered in average, mainly within 100 days of SCT (mean 10.5 units). Both MR-Fe and SF increased after SCT and decreased in the absence of new transfusions 1 year later without chelate therapy. This suggests regeneration of LP and iron loss, although the original states were not reached. Additionally, simultaneous peaks of CRP and SF were observed directly after SCT. MR-Fe did neither reveal these peaks nor was it associated with CRP (P = .39). We postulate that these early CRP and SF peaks after SCT are probably related to inflammatory reactions and not to iron overload. Thus, SF is not reliable for iron overload diagnosis after SCT in every condition. Beside this interaction, SF and MR-Fe revealed similar accuracy.
MRI, however, has practical and economical disadvantages in routine estimation of iron.

K E Y W O R D S
iron overload, liver, MRI, SCT, T 2 * relaxometry

| INTRODUC TI ON
Children undergoing SCT often require many blood cell transfusions during bone marrow aplasia. This results in iron overload. The iv administration bypasses physiological regulation and leads to an accumulation of ferrous ions in tissues, when physiological transport and storage capacities are exhausted. 1 Heart, liver, and neuronal and endocrine organs may be affected with a preference for LP and heart muscle, 2-6 suggesting screening of these tissues. In terminal stages, non-reversible liver fibrosis and loss of liver function can be observed. 3 Liver biopsy has so far been the only reliable way to assess cellular liver iron concentration (LIC). This invasive method, however, is not always feasible and recommended, especially in young patients after SCT. Nevertheless, assessment is required to diagnose early iron overload and initiate possible interventions, such as chelate therapy. Accurate iron level determination is possible by applying superconducting quantum interference device biosusceptometry, but is only available at few locations for research purposes. 7 Estimation of accumulated iron is possible with T 2 * MR relaxometry as the T 2 *-relaxation time constants show an inverse relationship to LIC. 8,9 There is a very good correlation between LIC, calculated from biopsies, and T2*. 10,11 Alternatively, estimation of whole-body iron content is more easily performed with SF measurements if the patients are in good medical condition. 2,5,12 However, SF is far from being an ideal surrogate parameter of IL as it also increases in various immune responses, such as acute-phase reaction or macrophage activation. 2,13 Macrophages themselves can synthesize SF. Inflammation may be observed in the context of SCT due to recruitment of macrophages, especially during GvHD. Many other factors, like malnutrition and malignancy, or different liver and kidney diseases are known to affect SF levels as well 4,6,9,11,[14][15][16][17] : It would therefore not be very reliable to consider the upper normal range of SF as the limit for iron overload. Thus, an established, higher threshold for clinically significant iron overload has been defined when the SF level exceeds 1000 µg/L in two subsequent determinations. This criterion is used to indicate chelate therapy in patients receiving chronic transfusions, such as in cases of thalassemia. 2,5,12 Whereas the specificity of rising SF levels in SCT might be inadequate due to compromised immunology, this limitation has not been described for T 2 * MR relaxometry. Therefore, we aimed to evaluate its applicability to indicate clinically significant iron overload during follow-up after SCT and compared the results with SF levels in association with CRP levels, which served as surrogate marker of inflammation.

| MATERIAL AND ME THODS
We identified 58 patients, receiving SCT due to hematological/oncological indications at our institution between 2014 and 2018. The patients were included in this prospective study if they underwent allogenic SCT, if they were younger than 18 years at initial diagnosis, and if they received at least three MR relaxometry examinations.
There were 23 patients, meeting the inclusion criteria, who were included in the analyses. Depending on clinical indications, all patients received a different number of red blood cell transfusions prior to and after SCT. There was no patient receiving chelate therapy.
A myeloablative conditioning protocol was performed on every patient prior to the transplantation. They were scheduled for T2* MR relaxometry before SCT and at least at day 100 and day 365 afterward. SF and CRP levels were determined on scheduled visits that occurred more often than the MRI examinations, and these additional values were included to enable a more detailed longitudinal follow-up evaluation, especially early after SCT CRP values were recorded immediately to identify inflammation (normal value: CRP <7.5 mg/L).

| SF
Blood samples were analyzed by using a two-step immunoassay ("ARCHITECT," Abbott Laboratories). Usually, samples were taken twice a week after SCT, although this frequency also depended on the general health condition and related complications during the subsequent follow-up, making hospitalization necessary.
We selected a threshold of SF >1000 µg/L 2,12 as an indicator for relevant iron overload, which is above the normal range, a but was intended to increase the specificity in the context of inflammation.

| IL
The so-called "IL" estimates the maximum iron incorporation, which can result theoretically from blood transfusions. It is calculated from the number of packed red blood cell units, N pRCB , each carrying about 200 mg of ferrous ions, and the BW at the time of the measurement (see Equation 2). Small patients weighing <25 kg did not receive an entire unit of red blood cells at once, but between 10 and 20 mL per kg BW. The exact individual amount was taken into account for IL calculation.

| Statistical analysis
Statistical analyses were performed, and graphical artwork was created in IBM SPSS 25 and Microsoft Excel, respectively. MRI-based findings and SF levels were evaluated at the three different time points (before, 100 days, and 1 year after SCT). These descriptive analyses were complemented by Wilcoxon signed-rank tests to identify significant differences between grouped data ranks. The influence of CRP (>7.5 mg/L vs <7.5 mg/L) on MR-Fe or SF, adjusted for the time of measurement, was analyzed using GEEs to account for correlated data. A gamma distribution and a log link function were assumed. The back-transformed effect estimate is presented as the ratio of means with 95% CI and Wald test-based P-value. This analysis was performed with SAS version 9.4. ROC analysis was used to compare the accuracy of SF and MR-Fe in diagnosing relevant iron overload (we applied a threshold of SF >1000 µg/L, with reference to Ong et al 12 ). Spearman's ρ was used as a measure of their correlation.
The continuous follow-up of blood sample parameters (SF, CRP) and the number of transfusions were analyzed graphically in longitudinal plots for each patient. One typical example case is discussed below.

| RE SULTS
MRI was successfully performed in all 23 patients at the three time were mainly GvHD or relapses occurring during follow-up (see Table 1). We compared IL, MR-Fe, SF, and the ratio between SF and MR-Fe at the three different time points by using box plots (see Figure 1A-D and Table 2). As indicated by the mean values in Table 2, the highest transfusion requirement occurred between SCT (day 0) and day 100 to compensate for the loss of bone marrow function immediately after transplantation; the corresponding IL was strongly increased at day 100 compared with day 0, reflecting cumulative iron intake. Since only few transfusions were administered thereafter, IL remained rather constant (see Table 2 and Figure 1A). Both MR-Fe and SF values decreased at day 365 compared with day 100 (   ROC analysis including all patients revealed high accuracy between MR-Fe and SF (see Figure 2), as the AUC was nearly 1, when applying a limit of SF >1000 µg/L 12 for iron overload diagnosis.
We did not find a significant influence (P = .39) of increased CRP (>7.5 mg/L) on MR-Fe, adjusted for the significant differences between the points of measurement (P = .02) in GEE. The highest mean values of SF and MR-Fe were observed on day 100 after SCT and the lowest before SCT. The mean value of the SF/MR-Fe ratio was significantly (P = .001) higher at day 0 compared with 100 days and 1 year later (see Table 1). The observed reduced SF/MR-Fe ratio correlated significantly with lower mean CRP values 100 days after SCT compared with the initial values (P = .03; see Table 2 for mean values).
A high, non-parametric correlation was observed between SF and MR-Fe including all 23 patients (r s = .84; P < .001). This correlation was decreased after SCT (r = .67; P < .001) during the period of high immune responses (up to day 100) and was increased again at day 365 (r s = .78; P < .001).
All available SF and CRP values were compared longitudinally for each patient individually to evaluate immune responses to SCT in more detail than at just the three time points described above. One

| D ISCUSS I ON
Iron storage in LP appears to be partly reversible, as LIC by MR-Fe and SF mean values decreased 1 year after SCT ( Table 2). Both values were lower 1 year after SCT than at day 100, although chelate According to different studies, eLIC >7 mg Fe 2+ /g is considered as highly elevated. 5,18,23 Thus, 17 of 23 patients in our study revealed iron overload at least at one measurement during follow-up and the IL mean was consequently above this threshold at all 3 time points (see Table 3, center row). Estimated iron levels from MR relaxometry   Figure 1D) as were the mean CRP values. As MR-Fe is supposed to reflect LIC correctly, 10,11 we attribute these observations to inflammatory reactions. The CRP-related SF peaks in the longitudinal plots support this inference. Lower CRP hence implies more correct estimation of iron levels by SF. Its threshold for iron overload should therefore not be fixed on normal values for this patient population as they are probably less specific and frequently exceeded during inflammation. Defining a specific limit is difficult due to the great data variability. However, the clinically applied limit of SF >1000 µg/L for chronically transfused pa- In the light of these remarks, chelate therapy may still be initiated based on SF also in children after SCT. However, we would recommend a high limit (ie, >1000 µg/L) and additional determination of CRP levels; relying on SF determination within 100 days after SCT is not advisable. Ideally, the indication should be proved by MR relaxometry.
One relevant limitation of our study is that no liver biopsies were performed to confirm LIC histologically. MRI examinations were less frequent than SF determination and not scheduled within 1 month after SCT during preventive isolation of the patients after SCT.
Patients receiving autologous SCT were not included in this study, because the related immune reactions were not expected to be comparable. The study population was homogeneous regarding the myeloablative conditioning, but varies concerning the following immune responses. We assumed that differences between the immune responses and the intensity of complications might also be reflected by different SF levels after SCT. Nevertheless, these effects were independent from MR-Fe at all.

| CON CLUS ION
The iron intake from transfusions after SCT leads to high liver IL, suggesting chelate therapy to be reasonable even for these pediatric patients. Whereas SF is useful for iron estimation during stable disease, it shows reduced significance during periods of inflammation, as seen after SCT. MRI provides comparable results in iron overload diagnosis; moreover, it is not affected by inflammation.

ACK N OWLED G M ENTS
This study was partly subsidized by Novartis AG, Nuremberg, Germany.

CO N FLI C T O F I NTE R E S T
The authors report no conflicts of interests.