Camilo A Campo1, Diego Hernando1, Tilman B Schubert1, Utaroh Motosugi1,2, Samir D Sharma1, Shane A Wells1, and Scott B Reeder1,3,4,5,6
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, 2Radiology, University of Yamanashi, Yamanashi, Japan, 3Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, 4Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States, 5Medicine, University of Wisconsin-Madison, Madison, WI, United States, 6Emergency Medicine, University of Wisconsin-Madison, Madison, WI, United States
Synopsis
This study evaluated the accuracy of liver
proton-density fat-fraction (PDFF) quantification in the presence of
ferumoxytol. Seven healthy subjects underwent MRI scans immediately before
injection of ferumoxytol and one day after injection. Our results indicate that
confounder-corrected chemical shift-encoded MRI PDFF estimates exhibit a small
but significant bias in the presence of ferumoxytol. This bias could be due to
differential shortening in the T2* of water and fat signals. Therefore,
investigators attempting to create human models of iron and fat within the
liver by administering ferumoxytol to patients with fatty liver should be aware
of this potential source of bias.Introduction
Chemical shift-encoded MRI (CSE-MRI) measurements
of proton-density fat-fraction (PDFF) in the liver have emerged as accurate
imaging biomarkers for the non-invasive detection and quantification of hepatic
steatosis in patients with non-alcoholic fatty liver disease.1
Because hepatic steatosis commonly occurs concurrently with liver iron
accumulation,2 studies have utilized intravenous injections of
superparamagnetic iron oxide (SPIO) contrast agents to evaluate the accuracy of
liver PDFF measurements in the presence of iron-caused R2* signal decay.3
Ferumoxytol is an SPIO-based agent used in iron
replacement therapy for patients with anemia, and it is growing in popularity for
off-label use as an MR contrast agent.4 Despite the research and
clinical interest in ferumoxytol, no studies have been conducted to investigate
the performance of PDFF measurements in the presence of ferumoxytol to the
best of our knowledge. Therefore, the purpose of this study was to evaluate the
accuracy of liver PDFF quantification in subjects receiving ferumoxytol injections.
Methods
Seven healthy subjects (mean[range] age: 47.1[32–60] years; 3M/4F) were recruited for this study after IRB
approval and informed written consent. Three subjects were given a dose of 2mg/kg
of ferumoxytol. Four subjects were given a dose of 4mg/kg of ferumoxytol. Imaging
was performed at 1.5T and 3T (HDxt and MR750, respectively, GE Healthcare, Waukesha,
WI) immediately before ferumoxytol injection, and
one day following injection.
Imaging was performed at each field strength using
an investigational version of a CSE-MRI technique based on a multi-echo 3D
spoiled gradient echo acquisition, using a 32-channel phased-array torso coil.
Imaging parameters for 1.5T were: TR=15.5ms, TE1=1.24ms,
ΔTE=2.06ms, echoes=6, FOV=400x360mm, slice
thickness=8mm, flip angle=5°, slices=28, and receiver bandwidth=±125kHz. Imaging parameters for 3T were:
TR=8.02ms, TE1=1.24ms, ΔTE=1.01ms,
echoes=6, FOV=400x360mm, slice thickness=8mm, flip angle=4°, slices=28, receiver bandwidth=±125kHz. Additionally, a multi-TE Stimulated Echo
Acquisition Mode (STEAM) single voxel MR Spectroscopy (MRS) was performed to
provide a T2-corrected reference for PDFF.5 Voxel size was
20×20×20-30×30×30mm and it was located on the right liver lobe (segment 6–7).
CSE imaging data were processed using a
confounder-corrected R2* and PDFF mapping algorithm.6,7 R2* and PDFF
measurements were obtained by drawing a single region-of-interest (ROI) on the
right liver lobe (segment 6–7).
This ROI was co-localized on both R2* and PDFF maps for every subject and
magnetic field strength.
Jarque-Bera statistical tests were done to test
the normality of the distribution of R2* and PDFF differences pre- and
post-injection. Student’s paired t-tests were done to test the statistical
significance of the R2* and PDFF differences pre- and post-injection.
Bland-Altman analysis was done to assess the
accuracy of PDFF measurements performed with MRI (MRI-PDFF) by analyzing their
agreement with PDFF measurements obtained with MRS (MRS-PDFF) pre- and
post-injection.
Results
Liver R2* measurements (s-1) pre- and post-injection had a mean ± SD[range] of 68.6 ± 55.8[27.8–232.8] and 246.1 ± 84.4[123.6–371.4], respectively. Liver
MRI-PDFF measurements (%) pre-
and post-injection had a mean ± SD[range] of 7.2 ± 5.9[1.1–18.4] and 9.0 ± 6.4[1.6–19.7], respectively. Liver
MRS-PDFF measurements (%) pre-
and post-injection had a mean ± SD[range] of 7.0 ± 6.3[0.42–17.7] and 7.1 ± 7.9[0.61–23.0], respectively.
R2*
and PDFF differences were normally distributed pre- and post-injection at both
field strengths. Figure 1 shows the
R2* and PDFF maps of Subject 1 pre- and post-injection at 3T. Figure 2 shows liver R2*, MRI-PDFF, and
MRS-PDFF measurements from all subjects pre- and post-injection. MRI-PDFF was
significantly higher post-injection at both field strengths (1.5T, p=0.02; 3T,
p=0.014). MRS-PDFF was not significantly different post-injection at either
field strength (1.5T, p=0.8; 3T, p=0.44). The Bland-Altman plots in Figure 3 show the agreement between
MRI-PDFF and MRS-PDFF pre- and post-injection at 1.5T and 3T. MRI-PDFF and
MRS-PDFF measurements pre- and post-injection were not significantly different at
either field strength (Figure 4).
Discussion and Conclusions
Our results indicate that confounder-corrected
CSE-MRI estimates of PDFF exhibit a small but significant bias in the presence of
ferumoxytol. Possible reasons for this bias could be related to differential
shortening in the T2* of water and fat signals. Although estimates of PDFF are
R2*-corrected, the signal model assumes a common R2* value for water and fat,
which is a good assumption in vivo,
8 even in the presence of
endogenous iron overload.
9 Further, the T1 shortening effects of
this agent could also potentially introduce additional unforeseen bias.
Overall, however, the bias introduced into PDFF estimates is small, but
investigators attempting to create human models of iron and fat within the
liver by administering ferumoxytol to patients with fatty liver should be aware
of this potential source of bias. Independent R2*-correction may be needed for
accurate estimation of PDFF in this case.
Acknowledgements
We acknowledge the support of NIH (R01 DK083380,
R01 DK088925, R01 DK100651) and GE Healthcare.References
1) Reeder Hepatology 2013 2) Sirlin Magn Reson Imaging Clin N Am 2010 3) Liau Magn Reson Imaging 2013 4) Bashir J Magn
Reson Imaging 2015 5) Hamilton NMR Biomed 2011 6) Yu
Magn Reson Med 2008 7) Meisamy Radiology 2011 8) Horng J Magn Reson Imaging
2013 9) Horng ISMRM 2014