Sampling error in histopathology findings of nonalcoholic fatty liver disease: a post mortem liver histology study.

BACKGROUND Many clinical trials and natural history studies on nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH) rely heavily on liver histology to define their endpoints. There are many indications that the liver is not uniformly involved in NAFLD thus sampling error is a major concern. This study aims to evaluate the uniformity of various histologic features in livers affected with NAFLD. METHODS Samples from a forensic autopsy series were studied and subjects with NAFLD identified.  We took specimens from three different parts of each liver and recorded the degrees of steatosis, hepatocyte ballooning, lobular inflammation, portal inflammation, and fibrosis. A NASH activity index (NAI) which is the sum of scores of histologic features was also calculated. The agreement between the 3 samples from each liver was studied. RESULTS There were 945 autopsies performed; 896 were suitable for histologic evaluation and 283 had NAFLD. Of these, 146 livers were available to our study from which 438 samples were taken. Fibrosis[intra-class correlation (ICC) = 0.87], lobular inflammation (kappa = 0.83), and portal inflammation (kappa = 0.83)were fairly uniformly distributed in the damaged liver. Steatosis was less uniform (kappa = 0.64), and hepatocyte ballooning was least uniformly distributed (kappa = 0.57). The ICC for NAI was 0.86, which indicated good agreement. CONCLUSION The individual histologic features of NAFLD and NASH are not uniformly distributed in the liver. Hepatocyte ballooning is especially non-uniform. Such non-uniformity should be taken into account when interpreting results of studies that rely on paired biopsies. A summary score such as NAI is less affected by sampling error.