Home>Uncategorized>What do we know on diagnostic accuracy of elastography and magnetic resonance imaging in NAFLD?

Comment to: Selvaraj EA, Mózes FE, Ajmer Jayaswal AN, Zafarmand MH, Vali Y, Lee JA, Levick CK, Joseph Young LA, Palaniyappan N, Liu CH, Aithal GP, Romero-Gómez M, Brosnan MJ, Tuthill TA, Anstee QM, Neubauer S, Harrison SA, Bossuyt PM, Pavlides M; LITMUS Investigators.
Diagnostic accuracy of elastography and magnetic resonance imaging in patients with NAFLD: a systematic review and meta-analysis.
J Hepatol. 2021 May 12:S0168-8278(21)00309-3. doi: 10.1016/j.jhep.2021.04.044. Epub ahead of print. PMID: 33991635.

The LITMUS (Liver Investigation: Testing Marker Utility in Steatohepatitis) project investigators recently evaluated the diagnostic accuracy of the following techniques for evaluation of liver fibrosis and non-alcoholic steatohepatitis (NASH): vibration-controlled transient elastography (VCTE), point shear wave elastography (pSWE), two-dimensional shear wave elastography (2DSWE), magnetic resonance elastography (MRE), and magnetic resonance imaging (MRI). The authors realized a systematic review and meta-analysis searching for studies examining the diagnostic accuracy of these index tests against histology as the reference standard in adult patients with non-alcoholic fatty liver disease (NAFLD). Summary estimates of sensitivity, specificity and area under the curve (sAUC) were calculated for fibrosis stages and NASH, using a random-effects bivariate logit-normal model.

Eighty-two studies for a total of 14,609 patients were analyzed. Meta-analysis for diagnosing fibrosis stages was possible in 53 VCTE, 11 MRE, 12 pSWE and four 2DSWE studies, and for diagnosing NASH in four MRE studies.

The authors assert that most of the primary studies reported the experience of the index test operators, however, data on intra- or inter-observer agreements were lacking, even when variability in interpretation is well known. In particular, they suggest that, if liver histology is the appropriate reference standard, it would be preferable that at least two histopathologists review the liver biopsy specimen, preferably with consensus.

The investigators identified several areas where data in the literature are lacking. They scored most studies as high risk of bias in at least one domain since very few studies conducted validation of pre-defined cut-offs or intention-to-diagnose analyses. The true applicability is also difficult to evaluate as many studies did not report success rates of index tests.

Furthermore, the available data address the use of these index tests exclusively for diagnosis of the target conditions in secondary care. Data for their performance in other contexts (e.g. screening for a target condition in primary care, or using an index test to indicate prognosis or predict treatment response) are therefore needed. In addition, for the two other ultrasound elastography techniques (pSWE and 2DSWE) there is also a lack of data on validated reliability criteria.

Another area where data are lacking is the diagnostic performance of these tests for NASH. While traditionally the modalities examined in this review were developed for fibrosis assessment, it is becoming increasingly important to identify patients with significant fibrosis and NASH, as these patients are considered more likely to benefit from pharmacological treatments.

When elastography index tests are gained successfully, they have acceptable diagnostic accuracy for advanced fibrosis and cirrhosis. The potential clinical impact of these index tests cannot be assessed fully as intention-to-diagnose analyses and validation of pre-specified thresholds are lacking.

The authors conclude that, in patients with NAFLD where liver stiffness can be measured successfully, VCTE, MRE, pSWE and 2DSWE have good diagnostic accuracy for the assessment of fibrosis, but only MRE and pSWE meet the minimum acceptable criteria of at least 80% sensitivity and specificity for the diagnosis of advanced fibrosis. Selvaraj et al. therefore wonder whether these promising results were overestimating the true diagnostic accuracy as intention-to-diagnose analyses and validation of pre-specified cut-offs are lacking from the literature.

Related papers