Purpose

This work was performed to study effects of patient characteristics, i.e. disease, age, and gender, on the relation between liver-to-muscle signal intensity ratio (SIR) values calculated using gradient echo (GRE) data, and reference liver iron concentration (LIC) values obtained by spin-echo (SE).

Methods

168 patients (71 f, 97 m, age range 2.3 – 79.9 years, mean 33.6 ± 20.8 y) suspected for iron overload due to various diseases were investigated for liver iron concentration by MRI. Transversal slices of the liver were acquired with two different protocols. For reference, the Ferriscan® protocol was performed (Resonance Health, Claremont, WA, Australia), consisting of five SE series with TE of 6, 9, 12, 15 and 18 ms. Results of centralized analysis were used as reference LIC. Additionally, GRE data were acquired in a single breathhold at TE/TR/FA 4.7/120ms/90°. GRE images were analyzed by manually placing three ROIs in liver parenchyma free of vessels and artefacts, and two in paraspinal muscles, avoiding potential pulsation artefacts. Liver-to-muscle SIR was calculated from median ROI values, and its natural logarithm (ln) correlated to reference LIC. To investigate influence of disease, patients were divided into groups without Thalassemia, Thal. major and other forms of Thalassemia. Since age differed between groups (cf. tab. 1), correlation was performed in four steps: a) linear regression between ln (SIR) and reference LIC was performed in all patients together and separated by gender, b) divided by disease and gender regardless of age, c) in a subgroup including only patients of age between 10 and 40 years, separated by disease and gender, and d) in patients with no Thalassemia above age of 40 years. Coefficient of determination R² of all correlations was determined as a measure of correlation quality. Parameters of linear regression (slope, intercept) and their uncertainties were checked for differences related to patient characteristics.

Results

Correlation between ln (SIR) and reference LIC was good with R² of 0.86 for all patients, 0.92 for Thal. maj. and 0.85 for patients with no Thalassemia. Slopes of regression lines did not differ significantly. Considering correlation separated by gender, R² was 0.80 for female and 0.92 for male patients. Corresponding scatter plots and regression lines are depicted in fig. 1. Differences in slope, although more pronounced than for different diseases, were not significant. Separating the Thal. maj. group by gender, we got R² of 0.97 for female and 0.95 for male Thal. maj. patients. Slope differences of regression lines were increased, cf. fig. 2, compared to gender differences considering all patients (fig. 1). Restricting analysis to age range of 10 to 40 years, which means a comparable age distribution in all subgroups, (for patient numbers in subcategories cf. Tab. 2), yielded further increased R² in Thal. maj. patients (0.98 for female and 0.96 for male patients) and significant slope differences, cf. fig. 3. However, in this age range, R² was reduced in patients with no Thalassemia. On the other hand, for patients without Thalassemia above 40 years of age (52 pts, 20 f, 32 m; diseases in this age group were predominantly different forms of anemia, leukemia, and MDS), we found R² of 0.92 (0.86 for female, 0.95 for male), i.e. larger coefficients of determination than considering all ages. Fig. 4 gives a scatter plot of these data. Correlation parameters showed no gender dependence in patients above 40 years without Thalassemia.

Discussion

Improved correlation in smaller subgroups indicate significant difference due to subgroup characteristic. Therefore, we conclude that disease, gender and age have to be taken into account when attempting to determine LIC from GRE data using the SIR method. Since Thalassemia was our largest patient subgroup, we studied this disease compared to others in a first step. It remains unclear due to small patient numbers whether there are also differing correlations of ln (SIR) to reference LIC between other diseases. Also, it is worthwhile to figure out whether different correlation between ln (SIR), derived from GRE data, and reference LIC, obtained from SE acquisitions, reflects solely effects of different acquisition strategies (GRE vs. SE) or has to do with underlying iron storage mechanisms, namely aggregated vs. dispersed iron (1).

Conclusion

Patient characteristics like disease, gender and age have to be considered in SIR analysis of GRE MRI when attempting to determine LIC. Further studies are needed to check whether this also applies to R2* analysis of GRE data.

Acknowledgements

We acknowledge Prof. G. Grön for help with statistics.

References

1. Jensen JH, Tang H, Tosti CL, Swaminathan SV, Nunez A, Hultman K, et al. Separate MRI quantification of dispersed (ferritin-like) and aggregated (hemosiderin-like) storage iron. Mag Reson Med 2010; 63(5):1201-9. PubMed PMID: 20432291.