Interobserver discrepancy in liver fibrosis using transient elastography

Abstract

Transient elastography is a useful method to assess liver fibrosis, but uncertainties still exist regarding reliability and reproducibility of the technique. We aimed to improve knowledge on interobserver variability with the procedure and tried to find factors associated with such variability. This was a cross-sectional study to compare the results of transient elastography performed by two different operators, one test made just after the other. We assessed both results with correlation tests and with repeated parametric or nonparametric tests, as needed. We also carried out a multivariate analysis to find factors associated with discrepancy in the results obtained by the two operators. We included a total of 333 pairs of transient elastography tests, belonging to 274 different patients. A total of 325 pairs of tests (97.6%) were valid. Results of the first and the second tests were, respectively, median (and interquartile range) of direct measurement 6.2 (4.6–10.6) and 6.0 (4.4–10.1) kPa (P = 0.012), and mean ± standard deviation of log10 of direct measurement 0.892 ± 0.316 and 0.871 ± 0.324 (P = 0.001). In 87 pairs of tests (26.7%), a discrepancy of at least 2 kPa between both results was found, and in 15 pairs of tests (4.6%), a discrepancy of at least 10 kPa was found. Discordance of at least one stage between both measurements was noted in 74 pairs of tests (22.8%). An association was found between higher stiffness and discrepancy between both operators (P < 0.001). Although transient elastography is a very convenient test to assess liver fibrosis in clinical practice, interobserver discrepancy in results is common and represents a significant problem with the technique. Discrepant results are more common in patients with higher values of stiffness.