P1016 : Factors associated with the development of fatty liver or glucose intolerance

Abstract

measuring different quantities and so bias cannot be assessed directly. The recently cleared (FDA, CE Mark, and TGA) HepaFatScan MRI method reports a volumetric fraction of liver tissue that is fat, a measure that can be directly compared with measurements from biopsy histological sections. This study measures the degree of bias of the HepaFat-Scan method against unbiased stereological measurements from biopsies. Stereological analysis (SA) is a method of obtaining unbiased estimates of volume fractions of a phase of material in a 3D matrix from a cross section through the matrix. Methods: Volumetric fractions of fat in liver tissue of 59 patients (autoimmune hepatitis 3, alcoholic liver disease 2, viral hepatitis 16, NAFLD 10, NASH 17, normal 3, primary sclerosing cholangitis 4, other 4) recruited from the hepatology outpatient clinics at Fremantle and Sir Charles Gairdner Hospitals (Western Australia) were measured by SA of digital images of histological sections of liver biopsies and with non-invasive HepaFat-Scan measurements. Bland–Altman (BA) statistics were used to assess the bias between the two methods of measurement. A steatosis grade for each biopsy was also assessed by an experienced hepatopathologist. ROC curve analysis was used to determine sensitivity and specificity of HepaFat-Scan for predicting steatosis grade. Results: Figure 1 shows a plot of the volumetric fraction of fat measured by HepaFat-Scan plotted against that measured by SA of biopsy histological sections. The straight line is the line of equivalence (not a fitted regression line). BA analysis indicates a small systematic bias with HepaFat-Scan reporting volumetric liver fat fractions 1.4% higher than SA of biopsy. Sensitivities, specificities and areas under ROC curve for HepaFat-Scan predicting steatosis grade >0 were 97%, 96%, and 0.963; >1 were 100%, 94%, and 0.996; and >2 were 100%, 88%, and 0.971. Conclusions: The 1.4% bias between HepaFat-Scan and SA of biopsy is not clinically significant since it represents less than 5% of the overall range of fat fractions measured. The observed high sensitivities and specificities indicate that overall accuracy of HepaFat-Scan is sufficient for clinical patient management.