Monitoring Iron Overload: Relationship between R2* Relaxometry of the Liver and Serum Ferritin under Different Therapies

INTRODUCTION

Different methods are available for the assessment of a patient's iron status,.[1] Serum ferritin is one of the most important clinical parameters[2] and its determination can be seen as a routine laboratory assay.[3] Despite the previously described positive correlation between serum ferritin and total iron stores,[4] there are known confounding factors so that serum[56] can also be elevated in other contexts.[47] Consequently, serum ferritin values only provide a rough indication of total body iron burden[8] and may not always correlate with its changes.[9]

The most accurate method to define iron overload is the estimation of the liver iron concentration (LIC) with its linear relation to the total body iron.[10] LIC is increasingly assessed by magnetic resonance imaging (MRI) providing a fast, noninvasive, and previously well-investigated and validated diagnostic method.[11]

To prevent possible severe clinical consequences in patients with iron overload, an adequate therapeutic management is required, thereby all the available therapies need accurate monitoring, in which serum ferritin and LIC are the actual parameters of interest.[12] Some studies have already pointed out that serum ferritin is a suboptimal parameter to measure the total iron at any time;[13] nevertheless, its trends are persistently used to estimate changes in the LIC.[4]

We evaluated the relationship between MRI-based R2* relaxometry of the liver and serum ferritin in therapy monitoring of patients with iron overload. Further, we investigated a possible influence of different therapies (phlebotomy or chelation) and other factors, such as hepatic fat or even the underlying disease.

MATERIALS AND METHODS

RESULTS

A total of 42 patients (28 males and 14 females; age range: 10–80 years, mean age: 48.10 years) were enrolled. About 19/42 had HFE-associated hemochromatosis, 3/42 aceruloplasminemia, and 20/42 were affected from secondary iron overload due to various reasons including dysmetabolic iron overload syndrome, repeated transfusion therapies due to various underlying diseases, as well as repeated iron infusions in the context of hemodialysis.

Twenty-seven patients had two and 15 patients had more than two follow-up examinations including MRI and blood analysis, respectively. Of the 15 patients with more than two consecutive examinations, five patients had three, four patients had four, two patients had five, one patient had nine and three patients had 10 follow-up examinations. This resulted in a total of 92 paired examinations, whereby 27 paired examinations were from 27 patients, 10 from five patients, 12 from four patients, 8 from two patients, 8 from 1 patient, and 27 paired exams from three patients. The maximum time interval between hepatic R2* determination and laboratory analysis was 60 days (average time interval: 7.3 days), whereby 27 patients had imaging and the laboratory examination on the same day.

The mean liver R2* value was 284.65 1/s (median: 219.03, range: 74–948). All patients had significant iron overload (R2* >70 1/s) at the time of their first examination. About 13/42 patients had concomitant hepatic steatosis. Detailed information for patients with hepatic iron and concomitant fat is provided in Table 1. Altogether, 21 patients were treated with phlebotomy, 12 patients received a chelation therapy, and nine patients were handled with no therapy (“watchful waiting”). Among 18/42 patients, R2* values normalized (R2* <=70 1/s) during the course of therapy.

Table 1 R2* and fat fraction values for patients with a pathological R2* (>70 1/s) and an increased hepatic fat fraction (>5.6%)

The relation between liver R2* (1/s) and serum ferritin values (ng/ml) for all patients and examinations is shown in Figure 1. A moderate positive, statistically significant correlation of r = 0.509 (P < 0.001) and rho = 0.624 (P < 0.001) was found.

Close

Figure 1

Correlation between liver R2* values and serum ferritin values of all patients and examinations. The solid line represents the best linear fit to the data; the dotted lines correspond to the 95% confidence interval of the linear regression

Export to PPT

The relation between absolute changes as well as percentage changes of liver R2* values and ferritin for each pair of consecutive examinations (n = 92) is shown in Figures 2 and 3, respectively. Thereby, an almost moderate correlation between absolute R2* changes and absolute ferritin changes (r = 0.497, P < 0.001; rho = 0.626, P < 0.001) and a strong correlation (r = 0.712, P < 0.001; rho = 0.759, P < 0.001) for percentage changes were found.

Close

Figure 2

Correlation between absolute changes of pathological liver R2* values and ferritin changes. Only pairs of consecutive examinations and only patients with R2* >70 1/s at their first examination were considered. The solid line represents the best linear fit to the data; the dotted lines correspond to the 95% confidence interval of the linear regression. Negative values represent an increase of the respective value between examinations.

Export to PPT

Close

Figure 3

Correlation between relative changes of hepatic R2* measurements and serum ferritin levels. Only pairs of consecutive examinations and only patients with R2* >70 1/s at their first examination were taken into account. The solid line represents the best linear fit to the data. The dotted lines correspond to the 95% confidence interval of the linear regression and the dashed lines represent the variability of ferritin (±14%) which was used to accept a real change of this parameter. Negative values represent an increase of the respective value between examinations.

Export to PPT

As described in materials and methods section, a threshold of 14% was used to classify a change of ferritin values as a real (effective) change and a threshold of 20 1/s to classify a change of R2* as a real change. With this, in 41/92 consecutive examinations, a concurrent decrease of R2* and serum ferritin was observed; hepatic steatosis was found for 12 of these 41 examinations. About 29/92 consecutive examinations showed stable or increased R2* and serum ferritin values. In none of these patients, hepatic steatosis was found. For 12 consecutive examinations (associated with 8/42 patients), R2* decreased and ferritin either stayed stable or increased and, in three of these cases, a pathological fat content was found. Finally, in ten consecutive examinations (associated with 9/42 patients), R2* was stable or increased and ferritin decreased, whereby three patients had hepatic steatosis. In total, thus for 22 paired examinations (23.9%), a discordance between ferritin and R2* was found.

By defining cases with both a decrease in R2* and serum ferritin as true positive, cases with a stable or increasing R2* and serum ferritin as true negative, cases with stable or increasing ferritin, but decreasing R2* as false positive, and cases with decreasing ferritin but stable or increasing R2* as false negative, we obtain a specificity of 0.707, a sensitivity of 0.804, a positive predictive value of 0.774, and an accuracy of 76%.

Table 2 summarizes all individual true- and false-positive results as well as true- and false-negative results (number of examinations) regarding the three therapeutic possibilities based on the above definitions (e.g., a stable or increasing R2* and serum ferritin were defined as true negative).

Table 2 Number of true-positive and false-positive classifications as well as true- and false-negative classifications regarding ferritin and R2* changes for the three therapeutic possibilities

Figure 4a and b shows examples of liver R2* and ferritin time courses for two different patients during therapy.

Close

Figure 4

(a) Ferritin and R2* values over time in one patient with HFE-associated hemochromatosis treated with phlebotomy – no relevant differences in the time course between ferritin and R2* values are found.(b) Ferritin and R2* values over time in another patient with iron overload due to aceruloplasminemia treated with chelation. Discrepancies with opposite gradients between ferritin and R2* are seen in this patient.

Export to PPT

Comparison of R2* and transferrin values resulted in a weak negative correlation (r = −0.342, P < 0,001). Analysis of the relative changes of transferrin values with changes of pathological R2* showed a moderate negative correlation (−0.513, P < 0.001). A weak correlation was found between liver R2* and transferrin saturation (r = 0.163, P = 0.0457). Considering the relative changes of these two parameters, no improvement was observed (r = 0.164, P = 0.119).

Figure 5 depicts the particular therapy which was applied between the respective consecutive examinations for each data point of Figure 3.

Close

Figure 5

Correlation between relative changes of hepatic R2* measurements and serum ferritin levels. In this figure, the same data as in Figure 3 are shown; however, taking into account the different applied therapies, the shown lines represent the best linear fit to the respective data: dashed line – phlebotomy data; dash-dotted line – chelation; and dotted line – no therapy.

Export to PPT

In 26/92 data points, the therapy consisted of phlebotomy, in 44/92 of chelation, and, in 22/92 cases, no therapy (watchful waiting) was applied. Analyzing the relation between relative changes of pathological liver R2* values and relative ferritin changes taking into account the different applied therapies, we found a strong correlation for phlebotomy (r = 0.855, P < 0.001), a strong correlation for chelation (r = 0.727, P < 0.001), and a moderate correlation for no applied therapy (watchful waiting) (r = 0.536, P = 0.0102). No significant difference with respect to slope and intercept was found between the linear regression models of phlebotomy and chelation; however, there was a significant difference for the intercept between phlebotomy and no therapy (P = 0.0043) as well as chelation and no therapy (P = 0.0016).

Calculating the linear regression between relative ferritin change and relative R2* change in patients with and without a hepatic steatosis did not result in a significant difference (no fat: r = 0.718 [P < 0.001], fat: r = 0.503 [P = 0.0796]; significance of difference in the slope: P =0.698 and significance of difference in the intercept: P =0.078). Between patients with and without HFE-associated hemochromatosis, the linear regression between relative ferritin change and relative R2* change also did not show a significant difference (HFE negative r = 0.631 (P < 0.001), HFE positive r = 0.795 (P < 0.001), significance of difference in the slope: P =0.139, and significance of difference in the intercept: P =0.35).

DISCUSSION

Quantifying hepatic iron by means of MRI is proposed as the method of choice for treatment monitoring of iron overload due to the known linear correlation of LIC with total body iron.[10] Even though hepatic iron quantification with MRI has already been proven to be a very reliable method, serum ferritin is still frequently used for monitoring therapy in a clinical setting.[21] The purpose of our study was to compare the relationship of R2* relaxometry with clinical biochemical parameters as used in our clinical setting in treated patients under monitoring (including watchful waiting).

We found a strong correlation between relative R2* and ferritin changes for patients receiving phlebotomy and chelation and a moderate correlation when no therapy (watchful waiting) was applied. However, in 22 of 92 paired examinations, a discordance of R2* and ferritin was found. in 12 examinations, R2* decreased and ferritin either stayed stable or increased and, in 10 examinations, R2* was stable or increased and ferritin decreased. Only relying on ferritin changes alone, this could have led to an unnecessary increase in treatment (overtreatment) in 21% of the associated patients or to a reduction in treatment (undertreatment) in 19% of the patients during some point of therapy. This has a crucial impact on clinical management and should be considered as an even more important result than the direct correlation between R2* and ferritin changes. Especially in patients with unexpected ferritin changes under therapy, that might be interpreted as a failure, and hence MRI with R2* should be considered for confirmation. Puliyel et al.,[22] used a very similar approach comparing R2* values with laboratory parameters. Their study concluded that neither ferritin nor the change in ferritin showed accurate concordance with LIC or changes of LIC and should therefore be interpreted with caution when used to draw clinical decisions such as adjusting chelation management in chronically transfused patients. Porter et al.,[23] concluded in a recent study that instead of using serum ferritin trends alone, the use of liver MRI is preferable to differentiate true from apparent nonresponders to deferasirox.

Rombout-Sestrienkova et al.,[24] postulated that depending on the treatment procedure, blood iron parameters can change differently. To the best of our knowledge, no study evaluated different therapies (phlebotomy vs. chelation) with respect to monitoring by ferritin or R2* relaxometry. We could show a strong correlation of relative serum ferritin and hepatic R2* changes for phlebotomy and chelation, but only a moderate correlation when no therapy was conducted (watchful waiting). No significant difference between phlebotomy and chelation but a significant difference between phlebotomy/chelation and no therapy (watchful waiting) was found. Interestingly, we found an increase or stable ferritin with a decrease of R2* in eight examinations under chelation but in only two examinations under phlebotomy. This implies that especially when dealing with chelation, R2* of the liver should be performed in cases with unexpected increase of ferritin. Beutler et al.,[25] pointed out that when effective chelation therapy is started, the serum ferritin falls more rapidly than body iron. This may be in some extent due to the improvement in liver function and on the other hand because ferritin may predominantly reflect reticular endothelial iron rather than parenchymal iron in the liver and other organs.[26]

Moreover, the underlying disease should also be considered. The study by Puliyel et al.,[22] did only include patients with secondary iron overload (chronically transfused patients, predominantly with thalassemia and sickle syndromes), while our study also considered patients with primary iron overload (hereditary hemochromatosis). Both patient groups may differ from each other in severity, course of disease, and even therapeutic possibilities as well as therapeutic response. Nevertheless, we found no significant difference between patients with primary and secondary iron overload, which implies that the underlying disease does not determine the choice of the therapy monitoring tool. Transferrin saturation is highly valuable for the diagnosis of hereditary iron overload.[27] Wood et al.,[11] described transferrin and transferrin saturation as valuable in tracking the therapeutic response to iron removal therapies. However, due to the many shortcomings, their use as a sole marker for chelator efficacy is not recommended. Our correlation analysis between hepatic R2* values and transferrin showed a weak negative correlation based on the absolute changes, improving to a moderate negative correlation regarding the relative changes. Transferrin saturation revealed a weak positive correlation for absolute as well as relative changes. This may be because transferrin can also be influenced by confounding factors and many chronically transfused patients have fully saturated transferrin, making it unsuitable for adequate measurement in the follow-up.

Steatosis is a common co-factor in liver injury in patients with iron overload.[2829] Powell et al.,[30] stated that these patients go along with higher serum ferritin levels and experience an acceleration of liver injury. Recently, Idilman et al.,[31] indicated that in conventional MRI, the detection and quantification of iron and fat becomes difficult because, for example, IP and out-of-phase images can cause diagnostic confusion. We systematically analyzed the simultaneous incidence of steatosis in patients with manifest iron overload and found 30% (13/42) concomitant steatohepatitis (SH). We found no significant differences regarding relative ferritin change versus relative R2* change in patients with and without concomitant steatosis. Therefore, hepatic fat does not seem to influence ferritin measurements for monitoring iron overload. This study did not focus on a possible inflammatory component as in nonalcoholic SH (NASH). We are aware of this limitation, but the diagnosis of NASH is based on criteria that were not available for our study.

Another limitation that should be addressed is the fact that the MRI measurement of LIC and the laboratory test was not on the same day in all patients. Although 27/50 patients had MRI and the laboratory analysis on the same day, the maximum tolerated interval was 60 days. Furthermore, no liver biopsies have been performed due to their invasive characteristic with known limitations such as hemorrhage risk and sampling errors.[32] Therefore, no histological verification/correlation of LIC measurements was carried out. Last but not the least, as described in detail, we included 42 patients and evaluated the change from one examination to the subsequent examination so that altogether 92 paired examinations could be analyzed. This means that as presented in the results' part some of the paired examinations were derived from follow-up examinations of the same patients. Thereby, for example, the three patients with aceruloplasminemia treated with chelation who underwent ten follow-up examinations accounted for 27 of all the 92 paired examinations. However, as treatment decisions are based on changes of ferritin or R2* between consecutive examinations, this fact should not influence our results.

CONCLUSION

A good correlation was observed between serum ferritin and MRI R2* relaxometry in monitoring iron overload. Nevertheless, in 22/92 (23.9%) examinations, taking only serum ferritin into account would have let to a misinterpretation of the used therapy. Further, neither hepatic fat nor the underlying disease had an influence on the correlation between ferritin changes under therapy and R2*. In general, ferritin is an acceptable and far cheaper tool for monitoring iron overload, but should be used with caution. MRI should be performed for confirmation, especially in the case of unexpected ferritin changes under therapy, particularly in patients treated with chelation.