Comparing morphometric, biochemical, and visual measurements of macrovesicular steatosis of liver

Abstract

The degree of macrovesicular steatosis is typically evaluated in liver biopsies by visual estimation, which is subject to intraobserver and interobserver variations. Computer morphometry and biochemical measurement may provide more accurate results. Our aim was to develop a morphometry method and compare its results with visual and biochemical measurements. Twenty-six fresh frozen liver specimens were each divided into 4 aliquots. Three aliquots were processed biochemically to extract fat, and the fat content was defined as the weight percentage of fat. One aliquot was fixed in formalin, from which hematoxylin and eosin slides were made and reviewed by 3 pathologists to estimate fat content. Digital images of slides were analyzed by computer morphometry, which defined fat content as the percentage of area occupied by fat droplets. The results showed that individually, each method produced highly precise and reproducible measurements. Compared with each other, they showed very strong correlations (correlation coefficient r = 0.81-0.95). The range of fat content in all 26 specimens was 2.2% to 15% by biochemical, 0.8% to 82.5% by visual, and 0.3% to 19.6% by morphometry method. Visual estimation appeared to have a systematic bias, giving results nearly 4-fold higher than other methods. This may be because visual estimation denotes the fraction of hepatocytes containing fat droplets, instead of the true fraction of fat. Strong correlations between different methods suggest that all 3 are valid methods for measuring steatosis. Computer morphometry is easy to implement and not affected by the bias seen in visual estimation. It may serve as a potential supplemental or alternative method.

Figure 1

Linear regression plots between different methods