Quantitative CT Accurately Predicts Liver Iron Concentration in Transfusional Siderosis.

Hemoglobinopathies are among the most common genetic diseases in the work. Many hemoglobinopathy patients require lifeline transfusion, iron chelation, and careful monitoring of iron stores. Liver iron concentration (LIC) is an excellent metric of transfusional iron balance and total body iron stores(1). Noninvasive LIC estimation by MRI is gradually replacing liver biopsy but remains limited by cost and availability, particularly in regions where thalassemia is prevalent(2). Quantitative computed tomography (QCT) was proposed as a means to estimate LIC 30 years ago, but there has been surprisingly limited validation(3-5). QCT is cheaper and more available than MRI. Steady improvements in CT instrumentation and standardization warrant a re-evaluation of QCT for iron quantitation. In this study, we determined liver attenuation as a function of MRI-predicted liver iron concentration in 45 patients over a 6 year period.

This study represents a convenience sample of all iron-overloaded patients who had undergone both QCT for bone density and LIC measurement by MRI at Children's Hospital Los Angeles. 64 usable observations were obtained in 45 patients; 14 patients had multiple exams(range 2-6). MRI and QCT examinations were considered “paired” if the scans were less than 120 days apart (59 studies). MRI liver R2 and R2* examinations were performed and analyzed as previously described(2).

Quantitative CT was performed on a General Electric Hilite Advantage. A single axial 10 mm thick slice was collected at the L1 level using a KVp of 80 at 70 mA for 1 second. Three hydroxyappetite phantoms, calibrated to 0, 125, and 250 Hounsfield units, were placed in scanning platform (CT-T bone densitometry package; GE Medical Systems), approximately 7 cm from mid-vertebral body. Calibration curve was obtained from regions of interest drawn within the three phantoms, using linear regression calculated by custom MATLAB routines. Regions of interest in the liver were drawn in ∼ 9 cm² regions of the right and left lobe of the liver, as well as a region encompassing the entire cross-sectional area of the liver.

Most patients had thalassemia major and moderate to severe iron overload, with a LIC of 14.1 ± 14 mg/g dry weight and a cardiac R2* of 70.5 ± 95.0 Hz (median T2* of 30.9 ms). Patients who were receiving regular transfusions were well transfused, with a pre-transfusion hemoglobin of 9-9.5 g/dl. All chronically transfused patients were using deferoxamine until approximately 2005, with most switching to deferasirox in 1/2005.

Figure 1 demonstrates MRI-predicted LIC as a function of liver attenuation. There is a strong linear relationship having a slope of 0.591 mg/g dry weight of liver per HU. Normal liver attenuation ranges in non iron overload children and young adults is 57-76 HU. Upper limit of normal corresponds to a predicted LIC of 6 m/g, indicating an intrinsic lack of sensitivity for qCT at low iron concentrations. Time-courses of CT-iron relationship from 14 patients whom had serial evaluations paralleled the regression line and were well constrained by the 95% confidence intervals, suggesting the calibration is suitable for serial analysis (not shown). Whole liver attenuation values were unbiased with respect to values from the right and left lobe; coefficient of variation was 2.2-4.9%.

The present work represents the largest human validation of QCT for liver iron quantitation. QCT techniques have inadequate sensitivity to discriminate LIC values less than 6 mg/g but are not limited by high iron concentrations. High reproducibility makes them suitable for tracking serial LIC changes. QCT may be an acceptable surrogate for LIC in hospitals lacking the software, personnel, or financial resources to support MRI or SQUID LIC measurements.

Figure 1

View largeDownload slide

Plot of MRI-predicted LIC as a function of CT attenuation. 95% confidence intervals are ± 7 mg/g. Normal liver attenuation and LIC are indicated by the horizontal line segment.

Figure 1

View largeDownload slide

Plot of MRI-predicted LIC as a function of CT attenuation. 95% confidence intervals are ± 7 mg/g. Normal liver attenuation and LIC are indicated by the horizontal line segment.

Close modal

Acknowledgments: This work supported by NIH HL075592, CDC (U27/CCU922106) and GCRC (NIH #RR00043-43).

No relevant conflicts of interest to declare.