To determine the repeatability of stiffness measurements in the liver using MR elastography (MRE) during the fasted and fed states. MRE has gained increased recognition as a noninvasive method to quantify fibrotic changes in the liver. It is well known that eating increases splanchnic blood flow, and fasting status of patients has been recognized as a factor that may affect hepatic stiffness measured with MRE.
Materials and Methods
Hepatic MRE stiffness and flow through the superior mesenteric vein (SMV) were measured in 12 healthy subjects in fasted and fed states, and measurements were repeated 5 weeks later. A linear mixed effects model was used to estimate the sources of variability in the data, which included day (exams on different days) and subject. Sources were combined to calculate the overall standard deviation of a single MRE measurement.
The total within-subject standard deviation of an MRE exam is 8.5% (standard error [SE] = 1.7%) or 9.0% (SE = 1.8%) for fasted and fed states, respectively. No significant differences between fasted/fed state stiffness and no correlation between hepatic stiffness and SMV flow were observed.
MAGNETIC RESONANCE ELASTOGRAPHY (MRE) has gained increased recognition as a noninvasive method to quantify hepatic stiffness as a surrogate biomarker of liver fibrosis (1–5). The elastic shear modulus of tissue is calculated by propagating acoustic strain waves through tissue using motion encoded phase difference images, and estimating the propagated wavelength through tissue using a wave inversion algorithm (1, 2). While this technique has been applied in many human tissues, such as the breast, kidney, brain, and muscle (3–6), hepatic stiffness measurements using MRE have shown great promise as a noninvasive surrogate of hepatic fibrosis (7–11). By quantifying liver fibrosis noninvasively, MRE has diagnostic potential by providing a noninvasive alternative to liver biopsy for quantifying liver fibrosis. In this way, this technique may be very relevant for a variety of disease processes, including viral hepatitis, nonalcoholic steatohepatitis, autoimmune hepatitis, primary biliary cirrhosis, and alcoholic cirrhosis (11).
The current gold standard for assessment of hepatic fibrosis is biopsy, which samples approximately 1/50,000th of the liver; however, given the heterogeneous nature of hepatic fibrosis, this technique is subject to sampling error (12–14). In addition, liver biopsy is under the subjective interpretation of a pathologist (15, 16) and also carries a small but significant risk of bleeding with an estimated 0.01–0.1% mortality (17). In contrast to liver biopsy, MRE has the ability to quantify stiffness over the entire liver, and thus may avoid sampling error.
Transient elastography is an ultrasound-based imaging technique that also noninvasively quantifies hepatic fibrosis. Transient elastography has been shown to have a diagnostic sensitivity of 87.5–100% and specificity of 81.5–93.5% for distinguishing varying stages of liver fibrosis in patients with chronic hepatitis C (18). Unlike MRE, however, transient elastography is also prone to sampling error (10, 18) because it does not evaluate the entire liver, and unlike MRE, it can be affected by the presence of steatosis, ascites, the ability to find an acoustic window, failing in up to 20% of patients, usually due to obesity (19).
Previous work has evaluated the sensitivity and specificity of MRE in quantifying the various degrees of liver fibrosis (10). However, ongoing studies have investigated the variability in the reproducibility of this diagnostic test that warranted further investigation (20). Specifically, the largest source of variability in the repeatability of MRE occurs for individuals on two different days and may reflect the effects of diurnal variation. In this past work, the effects of meals were not controlled and may explain this source of variability.
Splanchnic blood flow is well known to increase in the postprandial state, and the increase in portal blood flow is associated with a decrease in intrahepatic vascular resistance in healthy individuals (21, 22). Preliminary work by Yin et al has shown that hepatic stiffness may increase in cirrhotic patients after a meal (23), and this phenomenon has also been shown using transient elastography by Mederacke et al (24). As such, changes in blood flow due to fasting status could potentially affect the accuracy and reproducibility of MRE hepatic stiffness measurements.
A primary purpose of MRE may be to distinguish normal versus abnormal stiffness in the liver, as defined by Yin et al as having a cutoff of 2.93 kPa to distinguish normal from abnormal livers (10). Therefore, the repeatability and sources of variability for hepatic MRE must be explicitly defined in normal subjects to further determine the normal liver response to MRE assessment, which may help further distinguish normal from abnormal hepatic stiffness measurements. Healthy subjects provide a controlled manner to investigate the effects of blood flow and fasting status because it is easily assumed that individuals without liver disease should not exhibit any changes in shear stiffness over short periods of time (days), and assist in the definition of the normal physiological response of meals to MRE measurements.
Like any biomarker, the clinical utility of MRE will depend greatly on its repeatability (precision), which may be affected by physiological variability such as hepatic blood flow. If the repeatability of this technique is altered by blood flow and/or fasting status, these effects must also be quantified to gain insight into the repeatability of MRE. Thus, the purpose of this work is to quantify and determine the impact of blood flow and fasting status on the repeatability of MRE stiffness measurements in the liver in normal subjects.
MATERIALS AND METHODS
Twelve healthy subjects (nine men, three women) with no known liver disease were included in this study. Mean age was 30.3 years (range, 23–39 years), and mean body mass index was 23.8, (range, 20.3–26.1). All imaging was performed after obtaining institutional review board approval and informed consent. Subjects were without food or liquid (fasted) for 4 h before the imaging exam, per clinical routine. The meal before fasting was not standardized, although subjects were asked to eat meals consistent with normal routine. The imaging exam consisted of a localizer, MRE sequence, balanced steady-state free precession (bSSFP), and 2D CINE phase contrast (CINE-PC) as described below.
Imaging was performed on a clinical 1.5 Tesla (T) scanner (HDx TwinSpeed, GE Healthcare, Waukesha, WI) using an eight-channel phased array cardiac coil, which is the most commonly used coil for liver imaging at our institution. The MRE sequence used a passive pneumatic driver 19 cm in diameter positioned on the rib cage and attached to an acoustic waveform generator; the driver was on the subject for the entirety of the exam. A 60 Hz waveform was applied to the driver. An investigational two-dimensional (2D) gradient echo MRE sequence acquired anatomical magnitude and unwrapped phase difference wave images using the following image parameters: echo time/repetition time (TE/TR) = 24.2/100.0 ms, flip = 30°, BW = ± 31.25 kHz, slice = 10 mm, 256 × 128 matrix, 4 slices, and an asymmetric 75% field of view (FOV) adjusted to fit each subject (range, 34–40 cm in the readout dimension).
A stack of axial 2D bSSFP scout images were acquired to locate the superior mesenteric vein (SMV) using the following parameters: TE/TR = 1.7/5.5 ms, flip = 45°, BW = ± 125 kHz, slice = 10 mm, 256 × 192 matrix, 16–18 slices, FOV = 35 × 35 cm, and 1 signal average for a total imaging time of 20 s acquired in one breath-hold. Next, an axial CINE-PC acquisition was prescribed from the bSSFP images to measure blood flow through the SMV in the through-slice direction, using the following parameters: TE/TR = 6.7/33 ms, flip = 30°, BW = ± 15.63 kHz, slice = 5 mm, 256 × 192 matrix, 1 slice, FOV = 28 × 28 cm, 1 signal average, and velocity encoding (“VENC”) of 40 cm/s (22, 25). The sequence used cardiac gating and respiratory compensation and 16 interpolated phases were reconstructed throughout the cardiac cycle. Total imaging time was approximately 4 min and 30 s, based on respiratory rate.
After the exam, subjects were removed from the table and ate a standardized meal consisting of 20 ounces of non-diet soda and one piece of pepperoni pizza. In total, subjects ingested an average of 931 calories, 22.2 g fat, 42.4 mg cholesterol, 1817.5 mg sodium, 164.5 g carbohydrates (77.5 g sugar), and 30 g protein. This meal was chosen because it is more representative of a normal meal to the population and provided a greater gastrointestinal impact compared with a liquid meal challenge. Approximately 60 min after the initiation of the meal (30 min to eat, followed by 30 min of rest), the entire imaging exam was repeated to document an increase in mesenteric blood flow (21, 24, 26, 27). A VENC of 80 cm/s was used for the flow sequence to avoid phase aliasing from anticipated increases in mesenteric blood flow. Five weeks after the initial scans (range, 2–8 weeks), the entire process was repeated again such that four exams over 2 days were acquired. The same operator performed all exams.
MRE phase difference images were phase unwrapped to generate “wave images,” imaging masking was applied, and a mathematical inversion of the wave images produced individual shear stiffness maps for each slice (28). Stiffness maps express the calculated shear stiffness in kPa on a pixel-by-pixel image and ROI measurements are made using these images. All postprocessing was performed using an on-line reconstruction software package, and all images were transferred to an Advantage Workstation (AW 4.2, GE Healthcare, Waukesha, WI) for region of interest (ROI) analysis.
After imaging, the operator made measurements of the stiffness images from all MRE exams of all subjects. Because a previous study showed that the number of readers has minimal impact on MRE variability, only one reader was used (20). Using the wave images, ROI were selected in regions of waves relatively free of reflections and interference patterns, as described previously (20). The largest ROI that could be drawn in such a region was attained, and copied and placed onto the exact location in the stiffness map. The corresponding stiffness measurement was recorded, and one ROI per slice was measured. Average stiffness measurements and standard deviations were weighted by relative ROI area and calculated for each exam of all subjects.
The operator also made flow measurements from the magnitude and velocity CINE-PC images. Flow was determined through segmentation of the SMV using CV Flow (Medis, Raleigh, NC) where ROIs were placed around the SMV on each of the 16 magnitude images, and were simultaneously placed on the velocity images. ROIs were then manually adjusted to best fit the SMV. The area integral of the velocity within each time frame determined the flow waveform, which was then integrated over time to estimate the total flow through the SMV in one cardiac cycle (mL/cardiac cycle).
Linear regression was performed between hepatic stiffness and SMV flow for all subjects to display any patterns between measured stiffness and flow after averaging within subjects. A linear mixed effects (LME) model was used to estimate the sources of variability in the data, which included day (exams on different days) and subject. The sources of variability in the data have been termed “component variabilities,” because they are combined to calculate the standard deviation of a single MRE measurement (“total variability”). Standard errors for percent variability were calculated using a parametric bootstrap. All statistical calculations were carried out in the R statistical programming language (29).
All exams had both sufficient image quality and sufficient wave amplitude for analysis. Figure 1 displays representative MRE stiffness maps and wave and magnitude images from a healthy 39-year-old male subject before and after eating on both days of the study. No differences between measured stiffness were seen after each meal and between days, even though flow through the SMV more than doubled after eating in this individual. Day 1 fasted (fed) measured stiffness and flow were 2.32 ± 0.31 kPa (2.31 ± 0.40 kPa) and 7.67 mL/cycle (16.04 mL/cycle), respectively, and day 2 fasted (fed) measured stiffness and flow were 2.34 ± 0.34 kPa (2.42 ± 0.40 kPa) and 4.81 mL/cycle (20.5 mL/cycle), respectively.
For all subjects, average flow before eating was 3.5 ± 2.0 mL/cycle, and after eating was 12.0 ± 6.4 mL/cycle. According to Burkart et al, flow should increase by at least 100% in healthy individuals (22). Although the flow through the SMV, averaged across subjects, more than tripled after eating, no differences in stiffness were seen before and after a meal. Average fasting and fed stiffness from all subjects were 2.41 kPa (standard error [SE] = 0.06 kPa) and 2.38 kPa (SE = 0.06 kPa), respectively, and no statistically significant differences were observed between stiffness measured before and after eating (P = 0.65).
Figure 2 displays the change observed in measured stiffness after a meal versus the change observed in measured flow after a meal for all subjects on day 1 and day 2, where average fasting and fed state stiffness values were 2.41 kPa and 2.38 kPa, respectively, and average flow before and after eating was 3.5 mL/cycle and 12.0 mL/cycle, respectively. Lines connect individual subjects between day 1 and day 2. For both days, no agreement is seen between changes in shear stiffness and SMV flow for either day, as the line of best fit for day 1 (day 2) is as follows: slope = 0.0038 ± 0.01 (0.012 ± 0.01), intercept = 0.014 ± 0.1 (0.085 ± 0.1), r2 = 0.015 (0.088). When fasted and fed stiffness and flow measurements from both days are pooled for all subjects, obtained slope, intercept and r2 are −0.0011 ± 0.0054, 2.41 ± 0.05, and 0.0009, respectively. Measured stiffness remains constant for subjects despite considerable increases in the measured flow increases as a result of eating. The stiffness and flow measurements are also summarized in Table 1.
Table 1. MRE Hepatic Stiffness and SMV Flow Measurements in Healthy Subjects*
Shear stiffness = average ± standard deviation, with units in kPa. SMV flow is given in units of mL/cardiac cycle. Subject 12 (italic font) has been excluded from analysis due to abnormally high stiffness values.
2.82 ± 0.38
2.43 ± 0.30
2.53 ± 0.41
2.45 ± 0.41
2.32 ± 0.30
2.31 ± 0.40
2.34 ± 0.34
2.42 ± 0.40
2.26 ± 0.38
2.31 ± 0.26
2.52 ± 0.59
2.98 ± 0.62
2.82 ± 0.48
2.78 ± 0.38
2.69 ± 0.41
2.66 ± 0.49
2.31 ± 0.43
2.32 ± 0.31
2.47 ± 0.36
2.21 ± 0.38
1.98 ± 0.35
2.40 ± 0.44
2.11 ± 0.36
2.46 ± 0.31
2.61 ± 0.31
2.53 ± 0.30
2.51 ± 0.46
2.14 ± 0.26
2.31 ± 0.34
2.29 ± 0.43
2.20 ± 0.25
2.36 ± 0.33
2.35 ± 0.47
2.37 ± 0.46
2.24 ± 0.35
1.94 ± 0.33
1.99 ± 0.37
2.33 ± 0.42
2.15 ± 0.34
2.53 ± 0.36
2.56 ± 0.47
2.56 ± 0.33
2.53 ± 0.48
3.74 ± 0.47
3.86 ± 0.54
3,71 ± 0.53
3.91 ± 0.61
Results from one subject were excluded from analysis due to abnormally elevated hepatic stiffness according to Yin et al (10) that could not be explained by the subject's medical history or from possible technical failure. This subject is included in Table 1 with italic font, but has been excluded from all analysis. Of note, however, measured stiffness for this subject followed the same pattern of indifference toward SMV flow. Fasting and fed stiffness (flow) measurements for this subject were 3.74 ± 0.47 kPa (7.96 mL/cycle) and 3.86 ± 0.54 kPa (13.36 mL/cycle), respectively, on day 1, and 3.71 ± 0.53 kPa (2.86 mL/cycle) and 3.91 ± 0.61 kPa (12.94 mL/cycle), respectively, on day 2.
From the LME model, standard deviations (i.e., coefficient of variation, expressed as a percent of the mean measured stiffness) were calculated and are summarized in Table 2. The component standard deviation due to subject (i.e., within an individual) for the fasting state was 8.5% and fed state was 6.9%. The component subject variability in the fasted and states are not significantly different (P = 0.97). Similarly, the component standard deviation due to day of imaging for the fasted state was 4.5% and for the fed state was 5.1%, and the variability due to day of exam in fed and fasted states are also not significantly different (P = 0.28). The residual standard deviation, which includes all sources of variability not explicitly accounted for in the LME model, was 3.8% (SE = 0.5%).
Table 2. Sources of Variability, Expressed as Percent of Measured Stiffness
Tests whether fasted and fed state standard deviations are significantly different. A P value greater than 0.05 indicates no statistically significant difference between fasted and fed state standard deviations. n/a = not applicable.
Due to day, component
Total within subject standard deviation
The total within subject standard deviation of a MRE measurement was calculated by summing the residual variability and the component variability due to day, and has been described in detail previously (20). The total within subject standard deviation of a single MRE measurement was calculated to be 8.5% (SE = 1.7%) in the fasted state, and 9.0% (SE = 1.8%) in the fasted state. The within subject standard deviation is not statistically significantly different in the fasted or fed states (P = 0.66).
In this work, we have demonstrated that the total within subject standard deviation was not significantly different in fed (8.5%, SE = 1.7%) or fasted (9.0%, SE = 1.8%) states. The total within subject standard deviation is the standard deviation or variability of a single MRE measurement and because no clinically meaningful difference is seen between fasted or fed states, these results show that the repeatability of MRE stiffness measurements in the liver does not change with eating, or increased blood flow through eating in subjects. These findings are supported by the fact that no significant differences in the MRE stiffness measurements themselves were observed before and after eating in healthy subjects, despite large increases in observed blood flow to the liver. Average fasting and fed stiffness measurements were 2.41 kPa and 2.38 kPa, respectively. Lastly, no significant relationship between blood flow through the superior mesenteric vein and measured stiffness was seen.
One limitation of this study is that the results are applicable in healthy subjects only and before contrast agent administration. However, these findings have great clinical relevance because one of the main utilities of MR elastography is to differentiate normal from abnormal stiffness. Therefore, it is of critical importance to understand the repeatability of hepatic stiffness measurements in healthy subjects, in both fed and fasting states. Our observations suggest that the normal auto-regulatory mechanisms of the healthy liver, in response to a meal, maintain constant hepatic stiffness. Due to potential physiologic derangements in cirrhotic or fibrotic patients, however, these conclusions may be different. In another study, it was shown that statistically significant differences exist between fasted and fed states with increasing levels of fibrosis (23), although blood flow was not measured to document increases in mesenteric blood flow in these subjects. Differences in hepatic stiffness between fasted and fed states could be due to impaired auto-regulation of blood flow in patients with liver fibrosis. In the subset of normal subjects in this study, no differences were seen between fasted and fed states. However, unlike Yin et al, the aim of this study was to investigate the effects of postprandial state on the repeatability of MRE in the liver—we have demonstrated that a controlled fasting or fed status has no effect on the repeatability (precision) of this technique in healthy subjects.
The variability due to day in this study (fasted 4.5 ± 1.5%, fed 5.1 ± 1.5%) was lower than that reported previously by Hines et al using a larger cohort (8.5%) (20). In the latter, the effects of meals were not controlled and it was theorized that this may explain the largest source of variability, which was the variability due to day. The results of this study suggest that this variability is reduced by scanning subjects in a known fasted or fed state, although this study cannot be directly compared with Hines et al due to study design.
Additional work in patients with known liver disease, correlating increases in splanchnic blood flow after meal with changes in hepatic stiffness remains to be performed. Such work in patients would aid in defining expected within subject standard deviations in fasted and fed states for various fibrotic and cirrhotic changes, and stages of each. However, the repeatability of MRE in meal challenge studies in stages of fibrotic and cirrhotic patients is beyond the scope of this work, as we sought to investigate and define and quantify normal changes in MRE hepatic stiffness in response to a meal, which is essential for distinguishing normal versus abnormal hepatic stiffness.
In conclusion, this work has demonstrated no clinically significant differences in hepatic stiffness between fed and fasting states, with no relationship significance from portal blood flow, in healthy subjects. Moreover, we determined the repeatability of MRE and estimated that the total standard deviation for an individual undergoing an MRE exam is 8.5% (SE = 1.7%) in the fasted state and 9.0% (SE = 1.8%) in the fed state. The difference of the fasted and fed standard deviations is not significantly different. Scanning in either a fasting or fed status provides reliable measurement error (standard deviation) with repeatable stiffness estimates. However, scanning in the fasted state may still be preferred to reduce gastric and bowel motion and potential discomfort when using the driver on a full stomach, or for safety reasons if contrast agents are administered.