Accuracy and optimal proton density fat fraction threshold of magnitude- and complex-based magnetic resonance imaging for diagnosis of hepatic steatosis in obese patients using histology as reference

Tydus Thai¹, William Haufe¹, Yesenia Covarrubias¹, Alexandria Schlein¹, Curtis N. Wiens², Alan McMillan², Nathan S. Artz^2,3, Rashmi Agni⁴, Michael Peterson⁵, Luke Funk⁶, Guilherme M. Campos⁷, Jacob Greenberg⁶, Santiago Horgan⁸, Garth Jacobson⁸, Tanya Wolfson¹, Jeffrey Schwimmer⁹, Scott Reeder^{2,10,11,12,13}, and Claude Sirlin¹

¹Liver Imaging Group, Radiology, University of California-San Diego, San Diego, CA, United States, ²Radiology, University of Wisconsin-Madison, Madison, WI, United States, ³Radiological Sciences, St. Jude Children's Research Hospital, Memphis, TN, United States, ⁴Pathology, University of Wisconsin-Madison, Madison, WI, United States, ⁵Western Washington Pathology and Multicare Health System, Tacoma, WA, United States, ⁶Surgery, University of Wisconsin-Madison, Madison, WI, United States, ⁷Surgery, Virginia Commonwealth University, Richmond, VA, United States, ⁸Surgery, University of California-San Diego, San Diego, CA, United States, ⁹Pediatrics, University of California-San Diego, San Diego, CA, United States, ¹⁰Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ¹¹Biomedical Engineering, Madison, WI, United States, ¹²Medicine, Madison, WI, United States, ¹³Emergency Medicine, Madison, WI, United States

Synopsis

The purpose of this prospective cross-sectional study was to determine the accuracy of optimal MRI-M- and MRI-C-determined PDFF thresholds for diagnosis of hepatic steatosis using contemporaneous histology as reference in obese adults without previously known NAFLD. The excellent performance parameters of the Youden-index PDFF thresholds for MRI-M and MRI-C (5.3% and 7.7%, respectively) further support the use of these techniques for the quantitative and non-invasive diagnosis of HS. If validated by additional prospective studies, these PDFF thresholds could be used for diagnosing HS in obese adults non-invasively.

Purpose

Magnitude- and complex-based magnetic resonance imaging (MRI) techniques (MRI-M and MRI-C, respectively) were developed to estimate the proton density fat fraction (PDFF), a quantitative biomarker for liver fat content. Some studies have shown that PDFF closely correlates with histologically-determined hepatic steatosis (HS) grades and have suggested that it can be used to diagnose HS.^1-5 A recently published MRI-M-based PDFF threshold of 6.4% had 86% sensitivity and 83% specificity to diagnose HS.² While these results suggest that PDFF can diagnose HS accurately, that study had limitations. The cohort included only patients with known or suspected nonalcoholic fatty liver disease (NAFLD). At biopsy, HS was absent in only 7% of patients, which caused the confidence intervals (CIs) for the specificity estimate to be wide (95% CI: 36, 100). The enrichment with patients with known NAFLD may have also altered the optimal diagnostic threshold and reduced study generalizability. Finally, the study did not include MRI-C.

Therefore, the primary aim of this study was to determine the accuracy of optimal MRI-M- and MRI-C-determined PDFF thresholds for diagnosis of HS using contemporaneous histology as reference in obese adults without previously known NAFLD. This cohort was selected because it represents an at-risk population in which PDFF may be a beneficial biomarker for non-invasive diagnosis. A secondary aim was to compare the performance characteristics of MRI-M and MRI-C.

Methods

This was an IRB-approved, HIPAA-compliant, cross sectional analysis of patients enrolled in a prospective, longitudinal, dual-center study of obese adults undergoing clinical care weight loss surgery. The analysis focused on MRI exams performed immediately before surgery, since intraoperative biopsy served as reference. MRI-M and MRI-C were performed at 3T (GE Healthcare) and used to calculate liver PDFF. As described previously,^6-9 these advanced MRI techniques estimate PDFF by minimizing T1 bias, permitting fat-water separation while correcting for T2* decay, and using a multipeak spectral model to account for the multifrequency interference effects of triglyceride protons. Additionally,^10-11 MRI-C corrects for eddy currents and noise-related bias.

For analysis, regions of interest were placed at Couinaud liver segments 2-9 on source images and later propagated onto PDFF maps. PDFF values from the segments were averaged for each patient and technique.

Biopsy specimens were scored for HS based on the Nonalcoholic Steatohepatitis Clinical Research Network scoring system.¹²

Using histology-determined presence of HS as reference, receiver operator characteristic (ROC) analyses and bootstrapped-based pairwise tests were performed to calculate and compare the area under the ROC curve (AUC) for the two MRI techniques as well as the Youden-index diagnostic PDFF thresholds and their associated performance parameters.

Results

This study included 65 patients (55 female, 10 male, mean age 49.2, mean BMI at MRI 40.9). 44 patients (67%) had histology-confirmed HS.

Figure 1 illustrates MRI-M and MRI-C PDFF maps in representative patients with and without HS. As shown in Figure 2, both techniques had high AUCs. As summarized in Table 1, the Youden-index MRI-M PDFF threshold of 5.3% had a sensitivity of 0.909 (95% CI: 0.783, 0.975) and specificity of 0.905 (95% CI: 0.696, 0.988). This threshold misclassified 6 out of 65 patients (4 false negatives and 2 false positives). The Youden-index MRI-C PDFF threshold of 7.7% had a sensitivity of 0.773 (95% CI: 0.622, 0.885) and specificity of 0.952 (95% CI: 0.762, 0.999). This threshold misclassified eleven patients (10 false negatives and 1 false positive). Table 2 summarizes results from the bootstrapped-based comparisons of four performance characteristics (AUC, sensitivity, specificity, and total accuracy). Differences among the parameters were statistically insignificant except for AUC in which MRI-M had a borderline significantly higher value (p-value = 0.05).

Discussion

In a population of obese adults without previously known NAFLD, both MRI-M and MRI-C demonstrated excellent accuracy (AUC ≥ 0.93) in differentiating those with versus without HS. The optimal PDFF thresholds for the two techniques provided moderate-to-high sensitivity and high specificity for detecting HS. The optimal threshold for MRI-M was 2% lower than that for MRI-C, in keeping with prior reports that MRI-M tends to slightly underestimate PDFF compared to MRI-C.¹³Some MRI-histology misclassifications conceivably could be attributable to sampling variability and interpretation subjectivity of the histology reference standard.^14-15

Conclusion

This study further supports the use of MRI-M and MRI-C instead of biopsy and histological analysis for diagnosing HS. Since the study cohort was not enriched with patients with known NAFLD, the PDFF thresholds in this study may be more applicable for diagnosing HS in the general at-risk population. The optimal thresholds for MRI-M and MRI-C are similar but not identical.

Acknowledgements

Supported by R01 DK088925

References

1. Tang A, Tan J, Sun M, et al. Nonalcoholic fatty liver disease: MR imaging of liver proton density fat fraction to assess hepatic steatosis. Radiology. 2013; 267(2): 422–431.

2. Tang A, Desai A, Hamilton G, et al. Accuracy of MR Imaging–estimated Proton Density Fat Fraction for Classification of Dichotomized Histologic Steatosis Grades in Nonalcoholic Fatty Liver Disease. Radiology. 2015; Feb; 274(2): 416-25. doi: 10.1148/radiol.14140754

3. Permutt Z, Le TA, Peterson MR, et al. Correlation between liver histology and novel magnetic resonance imaging in adult patients with non-alcoholic fatty liver disease - MRI accurately quantifies hepatic steatosis in NAFLD. Aliment Pharmacol Ther. 2012; 36(1): 22–29.

4. Schwimmer JB, Middleton MS, Behling C, et al. Magnetic Resonance Imaging and Liver Histology as Biomarkers of Hepatic Steatosis in Children With Nonalcoholic Fatty Liver Disease. Hepatology. 2015 Jun; 61(6): 1887-95. doi: 10.1002/hep.27666.

5. Idilman IS, Aniktar H, Idilman R, et al. Hepatic steatosis: quantification by proton density fat fraction with MR imaging versus liver biopsy. Radiology. 2013; 267(3): 767–775.

6. Yokoo T, Shiehmorteza M, Hamilton G, et al. Estimation of Hepatic Proton-Density Fat Fraction by Using MR Imaging at 3.0 T 1. Radiology. 2011; 258(3):749-59. doi: 10.1148/radiol.10100659

7. Yu H, Shimakawa A, McKenzie CA, Brodsky E, Brittain JH, Reeder SB. Multiecho water-fat separation and simultaneous R2* estimation with multifrequency fat spectrum modeling. Magn Reson Med. 2008; 60: 1122–1134.

8. Hines CD, Yu H, Shimakawa A, McKenzie CA, Brittain JH, Reeder SB. T1 independent, T2* corrected MRI with accurate spectral modeling for quantification of fat: validation in a fat-water-SPIO phantom. J Magn Reson Imaging. 2009; 30: 1215–1222.

9. Bydder M, Yokoo T, Hamilton G, et al. Relaxation effects in the quantification of fat using gradient echo imaging. Magn Reson Imaging. 2008; 26 (3): 347–359.

10. Yu H, Shimakawa A, Hines CD, et al. Combination of complex-based and magnitude-based multiecho water-fat separation for accurate quantification of fat-fraction. Magn Reson Med. 2011; 66: 199–206.

11. Liu CY, McKenzie CA, Yu H, Brittain JH, Reeder SB. Fat quantification with IDEAL gradient echo imaging: correction of bias from T(1) and noise. Magn Reson Med. 2007; 58: 354–364.

12. Kleiner DE, Brunt EM, Van Natta M, et al. Design and validation of a histological scoring system for nonalcoholic fatty liver disease. Hepatology. 2005; 41: 1313–21.

13. Artz NS, Haufe WM, Hooker BS, et al. Reproducibility of MR-based liver fat quantification across field strength: Same-day comparison between 1.5T and 3T in obese subjects. J Magn Reson Imaging. 2015; 42: 811–817. doi: 10.1002/jmri.24842 DOI: 10.1002/jmri.24842.

14. El-Badry AM, Breitenstein S, Jochum W, et al. Assessment of hepatic steatosis by expert pathologists: the end of a gold standard. Ann Surg. 2009; 250(5): 691–697.

15. Ratziu V, Charlotte F, Heurtier A, et al. Sampling variability of liver biopsy in nonalcoholic fatty liver disease. Gastroenterology. 2005;128(7):1898–1906.

Figures

Figure 1: MRI-M and MRI-C PDFF maps of two representative patients, one with steatosis and one without steatosis, overlaid with the final PDFF estimates (averaged from PDFFs of Couinaud liver segments 2-9). The PDFF dynamic ranges for MRI-M and MRI-C were 0-50% and 0-100%, respectively (see scale bars at right).

Figure 2: Combined ROC curves for MRI-M and MRI-C with AUCs and p-values overlaid. Both techniques demonstrated excellent accuracy (AUC ≥ 0.93) for differentiating patients with versus without HS.

Table 1: Summary of performance parameters for the MRI-M- and MRI-C-determined Youden-index PDFF thresholds. Note—number of patients in parentheses were used to calculate percentages. Numbers in brackets are 95% CI. PPV = positive predictive value; NPV = negative predictive value.

Table 2: Results from bootstrapped-based comparisons of four performance parameters for MRI-M and MRI-C. Estimates were obtained via the difference between MRI-M and MRI-C values. Differences were statistically insignificant except for AUC, in which MRI-M had a borderline significantly higher value (p-value = 0.05).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

0358