MRE vs. CPRs for Noninvasive Diagnosis of Fibrosis in NAFLD
Utilising a prospective cohort design, our study demonstrated that 2D-MRE is highly accurate for the diagnosis of advanced fibrosis in adults with NAFLD. This is consistent with previous studies showing 2D-MRE to be highly accurate for advanced fibrosis prediction. We also demonstrated 2D-MRE to be superior to commonly available CPRs in predicting advanced fibrosis.
The non-invasive diagnosis of advanced fibrosis remains a major unmet need in the care of NAFLD patients. Although CPRs have been developed to predict advanced fibrosis in NAFLD patients, their AUROCs are significantly below that of 2D-MRE. Also, FIB-4, the NAFLD fibrosis score, and the Lok Index have indeterminate ranges. We propose that 2D-MRE is more accurate than CPRs and may be used in addition to CPRs for the non-invasive diagnosis advanced fibrosis in NAFLD patients, especially in patients whose CPRs score may be indeterminate.
The results of this study are consistent with previously published studies that demonstrated 2D-MRE to be effective in diagnosing advanced fibrosis. Kim et al. retrospectively showed 2D-MRE to be superior to five of the eight CPRs we examined (AST:ALT ratio, APRI, FIB-4, NAFLD fibrosis score and BARD) for diagnosing advanced fibrosis in a retrospective cohort of NAFLD patients, with 2D-MRE having an AUROC of 0.954 and with FIB-4 and the NAFLD fibrosis score having the highest AUROCs of the CPRs at 0.829 and 0.799 respectively. Our analysis showed a comparable AUROC at 0.957 for 2D-MRE and also showed FIB-4 to have the highest AUROC of the eight CPRs we examined, at 0.861, with the NAFLD fibrosis score ranking lower at 0.818. Two additional prediction rules not included in Kim et al., the Bonacini CDS and the Lok Index, also performed well for predicting advanced fibrosis with AUROCs of 0.826 and 0.838 respectively. The Bonacini CDS and the Lok Index were both originally developed for predicting hepatitis C cirrhosis and are thus far poorly validated in NAFLD patients. However, the performances of all eight CPRs we examined were still significantly below that of 2D-MRE.
Two-dimensional magnetic resonance elastography had a significantly better AUROC than FIB-4, the best-performing CPR in our analysis. FIB-4 has two cut-offs, an upper cut-off at 2.67 with good specificity and PPV and a lower cut-off at 1.30 with good sensitivity and NPV. Although FIB-4 was originally developed with an upper cut-off of 3.25 for use in patients with HIV and hepatitis C co-infection, for our analysis we used an alternative upper cut-off of 2.67 that was validated in a separate cohort of NAFLD patients. Although using a combination of the upper FIB-4 cut-off to rule in advanced fibrosis and the lower cut-off to rule out advanced fibrosis would have resulted in only five out of 102 misclassifications in our cohort (three false negatives and two false positives), it also results in 33 patients (32.4% of total) falling into the FIB-4 indeterminate range. 2D-MRE accurately classified 25 out of 33 patients in the FIB-4 intermediate range who would otherwise not have been classified using FIB-4 alone. Therefore, while FIB-4 may have diagnostic utility in the prediction of NAFLD-associated advanced fibrosis, it is less accurate than 2D-MRE. In addition, 2D-MRE can achieve high diagnostic accuracy even in patients with indeterminate FIB-4 scores.
The NAFLD fibrosis score, originally developed for NAFLD patients, is the most widely validated of all the CPRs we examined, and has been recommended as a screening tool to identify NAFLD patients with higher likelihoods of advanced fibrosis and/or cirrhosis in the NAFLD practice guideline from the American Association for the Study of Liver Diseases, American College of Gastroenterology, and the American Gastroenterological Association. Our study shows that 2D-MRE is also significantly better than the NAFLD fibrosis score for detecting advanced fibrosis in NAFLD patients. Better diagnostic accuracy may be achieved if 2D-MRE is used instead of the NAFLD fibrosis score as a screening tool for advanced fibrosis in NAFLD patients. However, the cost-effectiveness of 2D-MRE vs. the NAFLD fibrosis score in the context of improved diagnostic outcomes remains unknown.
The main strength of this study lies in its use of a prospective cohort of patients with biopsy-proven NAFLD and clinical indications for liver biopsy who met strict inclusion and exclusion criteria. The NASH CRN histologic scoring system, well-validated in NAFLD patients, was used as the gold standard. Biopsy, 2D-MRE and clinical data in each subject were done at close time intervals to one another. Data collection and 2D-MRE imaging were performed by experienced personnel at a specialised centre with a high degree of expertise in clinical and radiologic NAFLD research.
However, this study has several limitations. It was performed by highly trained personnel at a single centre specialising in NAFLD research, and the generalisability of its results in other clinical settings remains unknown. This study is cross-sectional and does not provide evidence on the longitudinal benefit of 2D-MRE vs. CPRs in detecting changes in liver fibrosis. Although 2D-MRE is cheaper than biopsy at our centre, it remains more expensive than clinical exams. Further studies are needed to determine if the additional diagnostic utility of 2D-MRE offsets its increased costs compared to the use of CPRs, although additional refinements in magnetic resonance technology may lower their costs in the future.
This prospective study shows 2D-MRE is superior to CPRs for diagnosing advanced fibrosis in patients with biopsy-proven NAFLD. 2D-MRE may be helpful in identifying patients with higher risks of advanced fibrosis, especially if the patients have indeterminate CPR scores. This study is also consistent with previous studies showing 2D-MRE to be highly accurate for detecting NAFLD-associated advanced fibrosis. Future studies should focus on the utility of 2D-MRE and CPRs to detect and monitor fibrosis progression in multicentre, progressive, and longitudinal cohorts, as well as the cost-effectiveness of 2D-MRE compared to CPRs, especially in patients with indeterminate CPR scores. Finally, the utility of even more novel and advanced imaging modalities, such as 3D-MRE, should be evaluated against that of 2D-MRE and CPRs for the non-invasive diagnosis of fibrosis in NAFLD patients.
Discussion
Main Findings
Utilising a prospective cohort design, our study demonstrated that 2D-MRE is highly accurate for the diagnosis of advanced fibrosis in adults with NAFLD. This is consistent with previous studies showing 2D-MRE to be highly accurate for advanced fibrosis prediction. We also demonstrated 2D-MRE to be superior to commonly available CPRs in predicting advanced fibrosis.
The non-invasive diagnosis of advanced fibrosis remains a major unmet need in the care of NAFLD patients. Although CPRs have been developed to predict advanced fibrosis in NAFLD patients, their AUROCs are significantly below that of 2D-MRE. Also, FIB-4, the NAFLD fibrosis score, and the Lok Index have indeterminate ranges. We propose that 2D-MRE is more accurate than CPRs and may be used in addition to CPRs for the non-invasive diagnosis advanced fibrosis in NAFLD patients, especially in patients whose CPRs score may be indeterminate.
In Context of Published Literature
The results of this study are consistent with previously published studies that demonstrated 2D-MRE to be effective in diagnosing advanced fibrosis. Kim et al. retrospectively showed 2D-MRE to be superior to five of the eight CPRs we examined (AST:ALT ratio, APRI, FIB-4, NAFLD fibrosis score and BARD) for diagnosing advanced fibrosis in a retrospective cohort of NAFLD patients, with 2D-MRE having an AUROC of 0.954 and with FIB-4 and the NAFLD fibrosis score having the highest AUROCs of the CPRs at 0.829 and 0.799 respectively. Our analysis showed a comparable AUROC at 0.957 for 2D-MRE and also showed FIB-4 to have the highest AUROC of the eight CPRs we examined, at 0.861, with the NAFLD fibrosis score ranking lower at 0.818. Two additional prediction rules not included in Kim et al., the Bonacini CDS and the Lok Index, also performed well for predicting advanced fibrosis with AUROCs of 0.826 and 0.838 respectively. The Bonacini CDS and the Lok Index were both originally developed for predicting hepatitis C cirrhosis and are thus far poorly validated in NAFLD patients. However, the performances of all eight CPRs we examined were still significantly below that of 2D-MRE.
Two-dimensional magnetic resonance elastography had a significantly better AUROC than FIB-4, the best-performing CPR in our analysis. FIB-4 has two cut-offs, an upper cut-off at 2.67 with good specificity and PPV and a lower cut-off at 1.30 with good sensitivity and NPV. Although FIB-4 was originally developed with an upper cut-off of 3.25 for use in patients with HIV and hepatitis C co-infection, for our analysis we used an alternative upper cut-off of 2.67 that was validated in a separate cohort of NAFLD patients. Although using a combination of the upper FIB-4 cut-off to rule in advanced fibrosis and the lower cut-off to rule out advanced fibrosis would have resulted in only five out of 102 misclassifications in our cohort (three false negatives and two false positives), it also results in 33 patients (32.4% of total) falling into the FIB-4 indeterminate range. 2D-MRE accurately classified 25 out of 33 patients in the FIB-4 intermediate range who would otherwise not have been classified using FIB-4 alone. Therefore, while FIB-4 may have diagnostic utility in the prediction of NAFLD-associated advanced fibrosis, it is less accurate than 2D-MRE. In addition, 2D-MRE can achieve high diagnostic accuracy even in patients with indeterminate FIB-4 scores.
The NAFLD fibrosis score, originally developed for NAFLD patients, is the most widely validated of all the CPRs we examined, and has been recommended as a screening tool to identify NAFLD patients with higher likelihoods of advanced fibrosis and/or cirrhosis in the NAFLD practice guideline from the American Association for the Study of Liver Diseases, American College of Gastroenterology, and the American Gastroenterological Association. Our study shows that 2D-MRE is also significantly better than the NAFLD fibrosis score for detecting advanced fibrosis in NAFLD patients. Better diagnostic accuracy may be achieved if 2D-MRE is used instead of the NAFLD fibrosis score as a screening tool for advanced fibrosis in NAFLD patients. However, the cost-effectiveness of 2D-MRE vs. the NAFLD fibrosis score in the context of improved diagnostic outcomes remains unknown.
Strength and Limitations
The main strength of this study lies in its use of a prospective cohort of patients with biopsy-proven NAFLD and clinical indications for liver biopsy who met strict inclusion and exclusion criteria. The NASH CRN histologic scoring system, well-validated in NAFLD patients, was used as the gold standard. Biopsy, 2D-MRE and clinical data in each subject were done at close time intervals to one another. Data collection and 2D-MRE imaging were performed by experienced personnel at a specialised centre with a high degree of expertise in clinical and radiologic NAFLD research.
However, this study has several limitations. It was performed by highly trained personnel at a single centre specialising in NAFLD research, and the generalisability of its results in other clinical settings remains unknown. This study is cross-sectional and does not provide evidence on the longitudinal benefit of 2D-MRE vs. CPRs in detecting changes in liver fibrosis. Although 2D-MRE is cheaper than biopsy at our centre, it remains more expensive than clinical exams. Further studies are needed to determine if the additional diagnostic utility of 2D-MRE offsets its increased costs compared to the use of CPRs, although additional refinements in magnetic resonance technology may lower their costs in the future.
Implication for Future Research
This prospective study shows 2D-MRE is superior to CPRs for diagnosing advanced fibrosis in patients with biopsy-proven NAFLD. 2D-MRE may be helpful in identifying patients with higher risks of advanced fibrosis, especially if the patients have indeterminate CPR scores. This study is also consistent with previous studies showing 2D-MRE to be highly accurate for detecting NAFLD-associated advanced fibrosis. Future studies should focus on the utility of 2D-MRE and CPRs to detect and monitor fibrosis progression in multicentre, progressive, and longitudinal cohorts, as well as the cost-effectiveness of 2D-MRE compared to CPRs, especially in patients with indeterminate CPR scores. Finally, the utility of even more novel and advanced imaging modalities, such as 3D-MRE, should be evaluated against that of 2D-MRE and CPRs for the non-invasive diagnosis of fibrosis in NAFLD patients.