Measurement of Liver Stiffness and Liver Fat Using Vibration-Controlled Transient Elastography Has Excellent Reproducibility and Intra-Operator Reliability in an Unselected Community-Based Cohort

Ian Downs, MD

Authors: Ian A. Downs, Robert Wilechansky, and Michelle T. Long

Background: Vibration-controlled transient elastography (VCTE) is a non-invasive method for the assessment of hepatic fibrosis via liver stiffness measurement (LSM) and hepatic steatosis via the controlled attenuation parameter (CAP). We aimed to determine the success and reliability of VCTE measurements in the Framingham Heart Study (FHS), a single-center, community-based, longitudinal cohort study with multiple operators.

Methods: A subset of participants of the FHS who underwent VCTE to characterize liver stiffness and fat between April 2016 and April 2019 underwent VCTE twice in one day. Using the Bland-Altman plot analysis, significant disagreement between readings was defined as the absolute value of the difference > 95% limits of agreement between the 2 readings. We also used two-way mixed effects models to determine the intra-operator correlation coefficients (ICC) for LSM and CAP.

Results: Overall, 97.8% (3334/3410) of the VCTE examinations performed were considered valid (defined by interquartile range/median LSM ≤ 30%). A total of 31 FHS participants were assessed using VCTE twice in one day with 18 (58%) participants receiving duplicate VCTE by the same operator and 13 (42%) participants receiving duplicate VCTE by two different operators. Each operator performed > 500 VCTE examinations. A total of 21 participants underwent duplicate VCTE examinations using the same probe size (M and M or XL and XL). The Bland-Altman plot for the LSM showed that 95% of the difference between the 2 readers occurred within -1.8kPa to 2.1kPa. Significant disagreement between LSM readings was noted for 2/21 (9.5%) of examinations. Using linear regression, the difference between the 2 LSM readings was not associated with the mean of the 2 LSM readings (p=0.82). The Bland-Altman plot for the CAP showed that 95% of the difference between the 2 readings occurred within -59.5 dB/m to 74.9 dB/m. Significant disagreement between CAP readings was noted for 1/21 (4.7%) of examinations. Using linear regression, the difference between the 2 CAP readings was not associated with the mean of the 2 CAP readings (p=0.78). For participants undergoing repeat VCTE with the same probe size, intra-operator reliability was excellent for LSM (ICC=0.91) and good for CAP (ICC=0.76).

Conclusion: In our unselected, community-based cohort, the overall failure rate for obtaining valid LSM and CAP measurements was very low. Significant disagreement between multiple readings for LSM and CAP were rare and in general, the intra-operator reliability was excellent. Additional studies are needed to determine the factors associated with discrepancies between or within operators.