Noninvasive assessment of liver fibrosis: The clinical context and question are important


  • Potential conflict of interest: Nothing to report.

To the Editor:

We read the recent article on noninvasive assessment of liver fibrosis, and we would like to make some comments.1

First, when using a noninvasive marker in a specific patient, both the clinical context and question are important. The sensitivity, specificity, and area under the receiver operating characteristic curve of a noninvasive test are not enough. In our recent meta-analysis of transient elastography (TE), we incorporated pretest probabilities as well as summary sensitivities and specificities to calculate post-test probabilities and, therefore, clinically translate the performance of such markers.2 If we assume that TE is used as a screening tool to detect cirrhosis (pretest probability = 0.5), a measurement above the cirrhosis cut-off value would mean a 90% probability of cirrhosis, whereas a measurement below this value would suggest a cirrhosis risk of 16%. Obviously, by manipulating the cut-off, either sensitivity or specificity can be increased at the expense of the other, depending on the clinical question, but this has not been adequately addressed, to date. For instance, if noninvasive tests are to be used as population-screening tests for fibrosis, then the emphasis should be on enhancing sensitivity. In an individual patient, if the index of suspicion for fibrosis/cirrhosis is low, one would need enhanced specificity, whereas with a high suspicion index, an enhanced sensitivity is more useful.

Second, noninvasive markers do not provide unexpected additional or alternative diagnoses. In this specific case scenario, there would be no information on steatosis, which is associated with fibrosis severity and treatment outcome.

Third, because the evaluation of noninvasive markers is shifting to clinical outcomes, it is important to remember that histological staging is descriptive and categorical and not designed as a prognostic tool. Therefore, regarding clinical outcomes, noninvasive fibrosis markers should be compared to established prognostic models. If histology is to be used a comparator for prognostic studies, then a quantitative histological assessment, such as collagen proportionate area or other prognostically validated morphometric approach, should be used.3

Last, sensitivities and specificities of noninvasive fibrosis markers are probably overestimated, because failed measurements are not calculated in the estimation of these parameters. This needs to be addressed in future studies.

In conclusion, noninvasive markers should be used and interpreted in the specific context of individual patients, because, clearly, one size does not fit all. Regarding the patient in the case scenario, we would advocate the performance of liver biopsy, because noninvasive tests can only potentially diagnose, but not exclude, cirrhosis and cannot give any other reliable information apart from this.

Emmanuel A. Tsochatzis*, Giacomo Germani*, Amar P. Dhillon†, Andrew K. Burroughs F.MED.SCI.*, * The Royal Free Sheila Sherlock Liver Center and University Department of Surgery Royal Free Hospital and UCL, London, UK, † Department of Cellular Pathology, UCL Medical School, London, UK.