Interobserver and intraobserver variability in the sonographic assessment of fatty liver

The purpose of this study was to evaluate interobserver and intraobserver variability in the sonographic assessment of the presence and severity of fatty liver.

We retrospectively evaluated the static images of 168 adult patients who had undergone abdominal sonography. Three experienced radiologists independently graded the hepatic images as normal, mild steatosis, moderate steatosis, or severe steatosis. Assessment of liver steatosis was repeated on the same set of images 1 month later under the same conditions and blinded to the initial reading. Weighted kappa statistics were used to analyze interobserver and intraobserver agreement, and the agreement percentages were calculated.

The mean interobserver and intraobserver agreement rates for the presence of fatty liver were 72% (kappa = 0.43) and 76% (kappa = 0.54). For severity of fatty liver, the initial reading for pairs of observers had 47-59% (kappa = 0.40-0.51) interobserver agreement. The interobserver agreement for the second reading was 59-64% (kappa = 0.43-0.54). The mean agreement rates for pairs of observers were 53% (kappa = 0.47) and 62% (kappa = 0.50) on the first and second readings. Intraobserver agreement for severity of fatty liver ranged from 55% to 68% (kappa = 0.51-0.63).

Subjective visual assessment of fatty liver on sonography has substantial observer variability. There is a need for a more objective quantitative method of grading fatty liver on sonography that would be easily available and applicable in routine clinical practice.