Transaminase abnormalities and adaptations of the liver lobule manifest at specific cut-offs of steatosis

There is little documented evidence suggesting that liver fat is responsible for liver injury in the absence of other disease processes. We investigated the relationships between liver fat, aminotransferases and hepatic architecture in liver biopsies with simple steatosis. We identified 136 biopsies with simple steatosis from the Royal Free Hospital Archives with both clinical data and sufficient material. Digital image analysis was employed to measure fat proportionate area (mFPA). Hepatocyte area (HA) and lobule radius (LR) were also measured. There were significant increases in ALT (p < 0.001) and AST (p = 0.013) with increased fat content and evidence to suggest both 5% and 20% mFPA as a cut-off for raised ALT. In liver with increased fat content there were significant increases in HA (p < 0.001). LR also increased as mFPA increased to 10% (p < 0.001), at which point the lobule ceased to expand further and was counterbalanced with a decrease in the number of hepatocytes per lobule (p = 0.029). Consequently there are mechanisms of adaption in the liver architecture to accommodate the accumulation of fat and these are accompanied by significant increases in transaminases. These results support the generally accepted cut-off of 5% fat for steatosis and indicate 20% as a threshold of more severe liver injury.

Fax: 0207 4353289 e-mail: andrewhall1@nhs.net Method for the determination of an adequate sample size for hepatocyte size measurements: We used H&E sections from 5 patients with a range of median hepatocyte areas,  µm 2 ). Non-overlapping images were taken at x20 objective magnification (OM) using a Zeiss Axioskop 40 with a Zeiss IcC5 camera in which the whole tissue area was sampled.
The hepatocytes were measured in 2 dimensions. The first measurement was along the maximum diameter of the hepatocyte and the second the maximum diameter perpendicular to the first. All the hepatocytes in the biopsy were measured if nucleus and the cell membranes were identifiable. Hepatocyte area was calculated as an ellipse from the 2 diameter measurements.
At least 2000 hepatocytes were measured per biopsy. A custom written MATLAB script was written that randomly selected a number of hepatocyte measurements and compared their average area to a reference value emulating and improving on previous work we have done on sample size calculation 1 . The reference value for the biopsy was the median hepatocyte area for the entire biopsy (median hepatocyte area for the 5 biopsies were 199, 216, 236, 299 and 457µm 2 ). In brief, the script would, for example, start by randomly (without replacement) selecting 2 hepatocytes from the 2000 and comparing the mean of their areas to the reference value and record the amount by which it deviated from the reference. The script repeated this random selection process 1000 times for each sample size from 1 to 2000 hepatocytes. In total, for all 5 biopsies, the script calculated 2 x10 6 random samples for each biopsy, in total 10x10 6 samples. The plots below show that a sample size of approximately 50 HA measurements taken randomly from throughout the biopsy is representative of average hepatocyte size for a biopsy. Figure S1. A scatter plot showing the proportion of samples that were 95% accurate (i.e. deviated by less than 5% of the reference value) for each of the sample sizes from 1-2000 cells measured.

Supplementary
97-99.6% of samples were 95% accurate (i.e. deviated by less than 5% of the reference value) when a sample size of 50 hepatocytes was randomly selected from the each biopsy. Therefore to establish a protocol for average hepatocyte size we randomly selected 50 hepatocytes from each biopsy and calculated their median hepatocyte size.

Method for the determination of an adequate sample size for mFPA
From H&E stained sections of 5 patients with a range of mFPA, (3-43%), 50 non-overlapping images were taken at x20 OM (Zeiss Axioskop 40 with a Zeiss IcC5 camera). Using the same script described for the determination of sample adequacy of hepatocyte size the mFPA results from the 50 images per biopsy were analysed for sample sizes from 1-50. The plots show that a sample size of approximately 21 non-overlapping images taken at x20 OM is representative of the biopsy mFPA.
Therefore, to assess mFPA of a biopsy, 21 randomly selected non-overlapping areas were imaged at x20 objective magnification (OM). Altogether 2100 at x20 OM were taken.