Accuracy of imaging methods for steatohepatitis diagnosis in non-alcoholic fatty liver disease patients: A systematic review.

BACKGROUND & AIMS: Non-invasive tests to diagnose non-alcoholic steatohepatitis (NASH) are urgently needed. This systematic review aims to evaluate imaging accuracy in diagnosing NASH among non-alcoholic fatty liver disease (NAFLD) patients, using liver biopsy as reference.

METHODS

Eligible studies were systematic reviews and cross-sectional/cohort studies of NAFLD patients comparing imaging with histology, considering accuracy and/or associations. MEDLINE, Scopus, EMBASE and Cochrane Library databases were searched up to April 2018. Studies were screened on title/abstract, then assessed for eligibility on full-text. Data were extracted using a predesigned form. Risk of bias was assessed using Quality Assessment of Diagnostic Accuracy Studies-2 tool.

RESULTS

Of the 641 studies screened, 61 were included in scoping review, 30 of which (with accuracy results) in data synthesis. Imaging techniques included: elastography (transient elastography-TE, acoustic radiation force impulse-ARFI, magnetic resonance elastography-MRE), ultrasound (US), magnetic resonance (MR), computed tomography and scintigraphy. Histological NASH definition was heterogeneous. In 28/30 studies, no prespecified threshold was used (high risk of bias). AUROCs were up to 0.82 for TE, 0.90 for ARFI, 0.93 for MRE and 0.82 for US scores. MR techniques with higher accuracy were spectroscopy (AUROC = 1 for alanine), susceptibility-weighted imaging (AUROC = 0.91), multiparametric MR (AUROC = 0.80), optical analysis (AUROC = 0.83), gadoxetic acid-enhanced MR (AUROCs = 0.85) and superparamagnetic iron oxide-enhanced MR (AUROC = 0.87). Results derived mostly from single studies without independent prospective validation.

CONCLUSIONS

There is currently insufficient evidence to support the use of imaging to diagnose NASH. More studies are needed on US and MR elastography and non-elastographic techniques, to date the most promising methods.